text
stringlengths 254
1.16M
|
|---|
---
title: 'Prevalence of overweight, obesity, and associated factors among healthcare
workers in the Gaza Strip, Palestine: A cross-sectional study'
authors:
- Joma Younis
- Hong Jiang
- Yahui Fan
- Lina Wang
- Zhaofang Li
- Majed Jebril
- Mei Ma
- Le Ma
- Mao Ma
- Zhaozhao Hui
journal: Frontiers in Public Health
year: 2023
pmcid: PMC9998069
doi: 10.3389/fpubh.2023.1129797
license: CC BY 4.0
---
# Prevalence of overweight, obesity, and associated factors among healthcare workers in the Gaza Strip, Palestine: A cross-sectional study
## Abstract
### Background
Overweight and obesity are multifactorial conditions that are prevalent in developing and developed countries. They are emerging as a significant public health concern among healthcare workers (HCWs). We aimed to estimate the prevalence of overweight and obesity and their associated factors among HCWs in the Gaza Strip.
### Methods
A cross-sectional study was conducted to recruit 1,850 HCWs aged 22 years and older. Interviews were carried out to collect sociodemographic information, nutritional information, and physical activity. Anthropometric measurements [height, weight, and waist circumference] were conducted with the HCWs. The body mass index was computed to determine the prevalence of overweight and obesity. Chi-square, t-test, and one-way ANOVA were used to compare the variables, and logistic regression was used to examine the associated factors of overweight and obesity.
### Results
The combined prevalence of overweight and obesity among HCWs was $65\%$. The result of logistic regression showed the risk of being overweight and obesity increased within the age group of 40–49 years (OR = 3.20; $95\%$ CI: 2.37–4.32; $P \leq 0.001$). Male participants had more risk of obesity than female participants (OR = 1.77; $95\%$ CI: 1.45–2.15). Married participants had a significantly higher risk of being overweight and obese (OR = 2.52; $95\%$ CI: 2.05–3.28; $$P \leq 0.001$$). Increased monthly income was significantly associated with the risk of being overweight and obese (OR = 2.16; $95\%$ CI: 1.22–3.83; $$P \leq 0.008$$). In addition, hypertension (OR = 2.49; $95\%$ CI: 1.65–3.78; $P \leq 0.001$) and type 2 diabetes (OR = 2.42; $95\%$ CI: 1.21–4.85; $$P \leq 0.012$$) were associated with overweight and obesity. Finally, a family history of NCDs was associated with overweight and obesity (OR = 1.69; $95\%$ CI: 1.38–2.07; $P \leq 0.001$).
### Conclusion
This study showed a high prevalence of overweight and obesity among HCWs. Age, monthly income, marital status, known hypertension, type 2 diabetes, and eating habits were associated with the prevalence of overweight and obesity compared to other variables that were not associated with overweight and obesity such as profession, vegetables, fruit consumption, and physical activity. Urgent action is needed to tackle overweight and obesity among HCWs.
## Introduction
Around the world, 650 million adults, 340 million adolescents, and 39 million children are obese. This estimation is still rising. According to the WHO, 167 million adults and children will experience deteriorating health by 2025 as a result of being overweight or obese [1]. Obesity is a significant contributor to poor health, including type 2 diabetes, cardiovascular disease, cancer, decreased life expectancy, and mortality [2], which could impact all classes of communities.
Most body systems, including the endocrine, gastrointestinal, neurological, and cardiovascular systems, are severely impacted by overweight and obesity, increasing a person's risk of contracting infectious diseases [3]. Obesity-related complications increase the number of morbidities that need to be managed by the declining number of healthcare professionals [4]. Healthcare workers (HCWs) who have direct contact with patients and often influence their behaviors ought to have a healthy physique to show a good role model in front of patients, and this revealed the importance of a healthy physique toward disease prevention [5]. Overweight and obese HCWs might have difficulty counseling the patients even if the patients clinically state with weight increase [6], which could be one of the barriers that can affect the patient consultation due to the same health issue.
Obesity and overweight rates are rapidly rising in developed and developing countries, including Palestine [7]. Overweight [body mass index (BMI) of 25.0–29.9 kg/m2] and obesity (BMI ≥ 30 kg/m2) increases the risk of non-communicable diseases (NCDs) [8]. A meta-analysis study in Middle East countries found that the prevalence of obesity and overweight was 21.17 and $33.14\%$, respectively [9]. A recent survey conducted in Palestine concluded that the prevalence of overweight and obesity is 23.6 and $19.5\%$ in the Gaza Strip and $26.1\%$ in the West Bank [10].
Due to the Israeli siege, people in Gaza live in unfavorable and dangerous conditions. The political events there, where Palestinian refugees, in particular, and Palestinian lives, in general, are concerned, have contributed to widespread unemployment, malnutrition, food insecurity, a lack of income, and a continuous decline in the quality of care for all patient categories [11, 12]. These factors undoubtedly could have an impact on people with NCDs. Healthcare workers are exposed to a wide range of difficulties that prevent the provision of health services as required, including lack of human resources, lack of medical supplies, workload, and financial burden, and these factors may negatively affect the health of health sector workers and lead to the development of work-related illness [13].
Understanding the viewpoints of healthcare professionals is crucial before creating a successful healthcare intervention [14]. A healthy society depends on the efforts of an important set of professionals, the HCWs [15]. Unfortunately, little health research has been done on managing NCDs. Previous studies revealed that sedentary jobs, long periods of sitting, and shift work all greatly raise the risk of obesity [16]. Healthcare providers include certified medical personnel (e.g., doctors, nurses, medical scientists, pharmacists, and technicians) and non-clinical support staff (e.g., the administrative class) [17]. Because of their specialized training, healthcare professionals are supposed to have a high level of knowledge and awareness of their health condition and the effects of lifestyle changes on their health [18, 19]. In addition, HCWs are responsible for promoting appropriate lifestyle changes that affect disease prevention and serve as role models for the general population by leading healthy lifestyles [20].
Therefore, few studies have been conducted to evaluate overweight, obesity, and the risk factors linked to these conditions among Palestinian HCWs working in the Gaza Strip. Because they are overlooked in research studies while being identified as high-risk populations, this study was done to evaluate the prevalence of overweight and obesity and its associated factors among HCWs.
## Study design and setting
A cross-sectional study was conducted from Feb to May 2020 in the Gaza Strip, Palestine, among a representative sample of Palestinian HCWs (physicians, nurses, paramedics, and non-medical) who work in hospitals and the primary healthcare (PHCs) of the ministry of health. Gaza *Strip is* divided into five smaller governorates, which include North Gaza, Gaza City, Mid Zone, Khan Younis, and Rafah. The whole number of hospitals was 10 and PHCs was 51 [21]. The study sample was distributed according to the number of hospitals and PHCs in each governorate.
## Sampling methods and participants
The consistency and completeness of all 1,900 responses were rigorously checked. The final analysis had 1,850 responses after 50 were excluded since they were considered incomplete or inconsistent. The participants were selected by multistage stratified random sampling. First, we dived the population into five governorates, and second, we selected two hospitals and three PHCs for each governorate. Considering the population distribution in each area, we have increased the sample size to be representative of the overall HCWs in the Gaza Strip. The sample size was determined using the formula n = Z2 P (1–P)/d2 at $95\%$ CI [22].
## Study criteria
Our inclusion criteria were: All HCWs working at the ministry of health in the Gaza Strip, with at least 1 year of experience, aged 22 years and older. Pregnant and lactating women, in addition to those workers under an unemployment program, were excluded.
## Data collection
Using a self-constructed face-to-face interview questionnaire, data were collected with multistage stratified random sampling. The data included detailed demographic and socioeconomic information (sex, marital status, educational levels, workplace, experience, monthly income, etc.), lifestyle involved (sleep duration, workload, work routine, and physical activity [measured by the International Physical Activity Questionnaire, IPAQ-short version]) [23]; the health profile involved [dietary patterns adapted from questions used in the food frequency questionnaire [24, 25]], menstrual cycle, family history of the disease, medical records, etc. After analyzing the literature on the subject, the questionnaire was well-prepared. The questionnaire's validity was tested by sending the completed questionnaire and a cover letter explaining the study's goal to 10 experts in various health professions (associated professors, hospital directors, managers of health departments, and academic teachers) who were asked to comment on the questionnaire. From the original English edition, all questions were translated into Arabic (forward and English and Arabic mother-tongue speakers performed backward translations). The study's research team consisted of five workers, physicians, and nurses, who collected the questionnaires by filling out the printed sheets.
Expert nurses measured each participant's anthropometric data by using standard protocols [26, 27], and height and weight were measured using stadiometers and weighing scales, respectively [26]. A measuring tape was positioned 1 cm below the umbilicus and at the iliac crest to measure the circumferences of the waist and hips, respectively [27]. The body mass index (BMI) formula (kg/m2) = Weight (kg)/Height squared (m2) [28] was used to calculate the BMI. We defined obesity according to WHO criteria; where underweight people had a BMI of <18.5 kg/m2, normal weight had a BMI of 18.5–24.9 kg/m2, overweight had a BMI of 25.0–29.9 kg/m2, and obese were over 30.0 kg/m2 [29].
The blood pressure (BP) values were taken with a sphygmomanometer. After participants had rested in a sitting position for at least 10 min, experienced nurses took two measurements on the right arm at a properly sized cuffed 1-min interval, with the arm supported at heart level and feet flat on the floor [30].
## Ethical considerations
All subjects signed consent forms before participating in the study. The ethical committee of Xi'an Jiao tong University Health Science Center approved the study, which was carried out following the Declaration of Helsinki. The Palestinian Health Research Committee at the Directorate General of Human Resources Development, Ministry of Health, Gaza (PHRC/HC/$\frac{663}{19}$) approved the protocol. The data were analyzed in an anonymous and non-linked manner, with no participant names being used. Moreover, there are no physical risks as there is no intervention such as blood sampling during the study.
## Statistical analysis
SPSS V.26 (Statistical package of social science) was used to carry out all data analyses. Continuous variables were represented by mean values and standard deviations (SD), while categorical variables were described by frequency and percentage. For categorical and continuous variables, chi-square and t-tests were used. If three or more groups were studied, one-way variance analysis (ANOVA) was used to compare demographic characteristics between groups.
The odds ratio (OR) and $95\%$ confidence interval (CI) of overweight and obesity were calculated by using univariate logistic regression, with the predictors associated (gender, age, marital status, work experience, monthly income, known hypertension, type 2 diabetes, and family history of NCDs). The statistical significance was set as a two-sided level, $P \leq 0.05.$
## Sociodemographic characteristics of the HCWs
A total of 1,850 HCWs were included in this study, 1,146 ($61.9\%$) were male participants, 704 ($38.1\%$) were female participants, and most of the participants were in the age group (30–39) with $49.3\%$. About $78.1\%$ of participants were married, while $68.6\%$ had a first degree (Bachelor's). The HCWs included 226 physicians ($12.2\%$), 1,208 nurses ($65.3\%$), 334 paramedics ($18.1\%$), and 82 non-medical ($4.4\%$). Most of the participants ($33.8\%$) had work experience of 10–15 years. The average monthly income of most participants was less than (2,000 NIS) per month at $61.8\%$.
The means anthropometric measurements for the participants were as follow: height 170.39 ± (8.86) cm, weight 78.63 ± (14.92) Kg, BMI 27.09 ± (4.77) Kg/m2, waist circumference 107.14 ± (11.63) cm, hip circumference 99.83 ± (7.02) cm, systolic blood pressure (SBP) 118.16 ± (10.13) mmHg, and diastolic blood pressure (DBP) 73.04 ± (10.01) mmHg. Moreover, $65\%$ of the participants were overweight or obese. BMI had significant associations with gender, age, marital status, profession, experience, monthly income, and anthropometric measurements (weight, height, waist circumference, hip circumference, SBP, and DBP) ($P \leq 0.05$; Table 1).
**Table 1**
| Variable | All | Underweight | Normal | Overweight | Obese | P-value |
| --- | --- | --- | --- | --- | --- | --- |
| | n =1,850 | n = 18 (1.0%) | n = 630 (34.0%) | n = 727 (39.3%) | n = 475 (25.7%) | |
| Gender | Gender | Gender | Gender | Gender | Gender | Gender |
| Male | 1146 | 8 (0.7) | 336 (29.3) | 490 (42.8) | 312 (27.2) | <0.001a |
| Female | 704 | 10 (1.4) | 294 (41.7) | 237 (33.7) | 163 (23.2) | |
| Age group (years) | Age group (years) | Age group (years) | Age group (years) | Age group (years) | Age group (years) | Age group (years) |
| 22–29 | 334 | 2 (0.6) | 184 (55.1) | 98 (29.3) | 50 (15.0) | <0.001a |
| 30–39 | 912 | 10 (1.1) | 304 (33.3) | 390 (42.8) | 208 (22.8) | |
| 40–49 | 440 | 6 (1.4) | 118 (26.8) | 184 (41.8) | 132 (30.0) | |
| 50–61 | 164 | - | 24 (14.6) | 55 (33.6) | 85 (51.8) | |
| Marital status | Marital status | Marital status | Marital status | Marital status | Marital status | Marital status |
| Unmarried | 406 | 4 (1.0) | 201 (49.5) | 135 (33.2) | 66 (16.3) | <0.001a |
| Married | 1444 | 14 (1.0) | 429 (29.7) | 592 (41.0) | 409 (28.3) | |
| Educational level | Educational level | Educational level | Educational level | Educational level | Educational level | Educational level |
| Diploma | 354 | 5 (1.4) | 129 (36.5) | 142 (40.1) | 78 (22.0) | 0.258a |
| Bachelor | 1269 | 12 (0.9) | 433 (34.1) | 498 (39.3) | 326 (25.7) | |
| Post-graduated | 227 | 1 (0.4) | 68 (30.0) | 87 (38.3) | 71 (31.3) | |
| Profession | Profession | Profession | Profession | Profession | Profession | Profession |
| Physician | 226 | 4 (1.8) | 68 (30.1) | 94 (41.6) | 60 (26.5) | 0.033a |
| Nurse | 1208 | 6 (0.5) | 414 (34.3) | 483 (40.0) | 305 (25.2) | |
| Paramedics | 334 | 8 (2.4) | 112 (33.5) | 122 (36.5) | 92 (27.5) | |
| Non-medical | 82 | – | 36 (43.9) | 28 (34.1) | 18 (22.0) | |
| Experience (years) | Experience (years) | Experience (years) | Experience (years) | Experience (years) | Experience (years) | Experience (years) |
| <5 | 381 | 4 (1.0) | 192 (50.4) | 132 (34.7) | 53 (13.9) | <0.001a |
| 5–10 | 575 | 6 (1.0) | 192 (33.4) | 238 (41.4) | 139 (24.2) | |
| 10–15 | 626 | 6 (1.0) | 190 (30.3) | 258 (41.2) | 172 (27.5) | |
| >15 | 268 | 2 (0.8) | 56 (20.9) | 99 (36.9) | 111 (41.4) | |
| Monthly income (NIS) | Monthly income (NIS) | Monthly income (NIS) | Monthly income (NIS) | Monthly income (NIS) | Monthly income (NIS) | Monthly income (NIS) |
| <2,000 | 1144 | 14 (1.2) | 432 (37.8) | 453 (39.6) | 245 (21.4) | <0.001a |
| >2,000 | 706 | 4 (0.6) | 198 (28.0) | 274 (38.8) | 230 (32.6) | |
| | Mean (SD) | Mean (SD) | Mean (SD) | Mean (SD) | Mean (SD) | |
| Height | 170.39 (8.86) | 170.44 (10.39) | 170.05 (8.97) | 171.46 (8.95) | 169.19 (8.33) | <0.001b |
| Weight | 78.63 (14.92) | 52.67 (6.41) | 65.36 (8.45) | 80.28 (8.61) | 94.69 (11.68) | <0.001b |
| BMI | 27.09 (4.77) | 18.07 (0.37) | 22.53 (1.64) | 27.24 (1.24) | 33.25 (3.78) | <0.001b |
| WC | 107.14 (11.63) | 82.22 (2.41) | 97.25 (5.58) | 110.51 (9.96) | 116.03 (9.20) | <0.001b |
| HC | 99.83 (7.02) | 82.72 (4.53) | 95.61 (5.08) | 100.32 (5.29) | 105.30 (7.02) | <0.001b |
| SBP | 118.16 (10.13) | 113.33 (8.40) | 115.72 (8.76) | 118.51 (10.02) | 121.04 (11.18) | <0.001b |
| DBP | 73.04 (10.01) | 67.78 (9.42) | 70.40 (9.44) | 73.55 (9.48) | 75.94 (1.61) | <0.001b |
## Prevalence of overweight and obesity among HCWs
Based on gender, the prevalence of overweight and obesity among HCWs was $43.4\%$ for male participants and $21.6\%$ for female participants. Moreover, the total prevalence in both genders was $65\%$ for the whole study participants and $35\%$ had normal weight. There is a statistical association between gender and BMI ($P \leq 0.05$; Table 1).
The prevalence of overweight and obesity among HCWs stratified by age group also offers the most age group at risk of overweight and obesity between 30 and 39 years with $32.32\%$, followed by 40 and 49 years with $17.09\%$, and the lowest prevalence for the age group of 50–61 years with $7.56\%$. There is a statistically significant association between age and BMI ($P \leq 0.05$; Table 1).
## Lifestyle factors of the healthcare workers and BMI
Approximately $18.9\%$ of the study participants were smokers, and $65.5\%$ had low physical activity. The consumption of fruits <3 days/week was $92.9\%$, and the consumption among overweight and obese, as well as for vegetables, was $96.6\%$, <3 days/week. However, there is no relationship between fruit/vegetable consumption and the prevalence of overweight, and obesity. While $91.6\%$ of the participants ate three meals daily, $8.9\%$ ate more than three times per week out of home, and $65.2\%$ ate with more than five people. The prevalence of hypertension was $8.4\%$, type 2 diabetes was $2.9\%$, and family history of NCDs among participants was $67.6\%$. There are statistically significant associations between BMI and the number of meals, the number of meals eaten outdoors, the total number of members who ate together, known hypertension, type 2 diabetes, and family history of NCDs ($P \leq 0.05$; Table 2).
**Table 2**
| Variable | All | Underweight | Normal | Overweight | Obese | P-value |
| --- | --- | --- | --- | --- | --- | --- |
| | n = 1,850 | N = 18 (1.0%) | n = 630 (34.0%) | n = 727 (39.3%) | n = 475 (25.7%) | |
| Smoking | Smoking | Smoking | Smoking | Smoking | Smoking | Smoking |
| Yes | 350 | 2 (0.6) | 100 (28.6) | 152 (43.4) | 96 (27.4) | |
| No | 1500 | 16 (1.1) | 530 (35.3) | 575 (38.3) | 379 (25.3) | |
| Physical activity | Physical activity | Physical activity | Physical activity | Physical activity | Physical activity | Physical activity |
| Low | 1211 | 12 (1.0) | 421 (34.8) | 456 (37.6) | 322 (26.6) | 0.353 |
| Moderate | 540 | 5 (0.9) | 174 (32.2) | 236 (43.7) | 125 (23.2) | |
| Hight | 99 | 1 (1.0) | 35 (35.4) | 35 (35.4) | 28 (28.2) | |
| Fruits consumption in a week | Fruits consumption in a week | Fruits consumption in a week | Fruits consumption in a week | Fruits consumption in a week | Fruits consumption in a week | Fruits consumption in a week |
| ≤3 days | 1718 | 18 (1.0) | 584 (34.0) | 677 (39.4) | 439 (25.6) | 0.654 |
| >3 days | 132 | – | 46 (34.8) | 50 (37.9) | 36 (27.3) | |
| Vegetable consumption in a week | Vegetable consumption in a week | Vegetable consumption in a week | Vegetable consumption in a week | Vegetable consumption in a week | Vegetable consumption in a week | Vegetable consumption in a week |
| ≤3 days | 1788 | 18 (1.0) | 604 (33.8) | 707 (39.5) | 459 (25.7) | 0.457 |
| >3 days | 62 | – | 26 (41.9) | 20 (32.3) | 16 (25.8) | |
| Number of meals (day) | Number of meals (day) | Number of meals (day) | Number of meals (day) | Number of meals (day) | Number of meals (day) | Number of meals (day) |
| ≤3 | 1694 | 16 (0.9) | 570 (33.7) | 685 (40.4) | 423 (25.0) | 0.009 |
| >3 | 156 | 2 (1.3) | 60 (38.5) | 42 (26.9) | 52 (33.3) | |
| How many days do you usually eat at home every week? | How many days do you usually eat at home every week? | How many days do you usually eat at home every week? | How many days do you usually eat at home every week? | How many days do you usually eat at home every week? | How many days do you usually eat at home every week? | How many days do you usually eat at home every week? |
| ≤3 | 164 | 4 (2.5) | 44 (26.8) | 62 (37.8) | 54 (32.9) | 0.015 |
| >3 | 1686 | 14 (0.8) | 586 (34.8) | 665 (39.4) | 421 (25.0) | |
| How many people do you usually eat together? | How many people do you usually eat together? | How many people do you usually eat together? | How many people do you usually eat together? | How many people do you usually eat together? | How many people do you usually eat together? | How many people do you usually eat together? |
| ≤5 | 1206 | 12 (1.0) | 428 (35.5) | 498 (41.3) | 268 (22.2) | <0.001 |
| >5 | 644 | 6 (0.9) | 202 (31.4) | 229 (35.6) | 207 (32.1) | |
| Known hypertension | Known hypertension | Known hypertension | Known hypertension | Known hypertension | Known hypertension | Known hypertension |
| Yes | 155 | – | 29 (18.7) | 55 (35.5) | 71 (45.8) | |
| No | 1695 | 18 (1.1) | 601 (35.5) | 672 (39.6) | 404 (23.8) | |
| Known type 2 diabetes | Known type 2 diabetes | Known type 2 diabetes | Known type 2 diabetes | Known type 2 diabetes | Known type 2 diabetes | Known type 2 diabetes |
| No | 1796 | 15 (0.8) | 623 (34.7) | 705 (39.3) | 453 (25.2) | <0.001 |
| Yes | 54 | 3 (5.6) | 7 (13.0) | 22 (40.7) | 22 (40.7) | |
| Family history of NCDs | Family history of NCDs | Family history of NCDs | Family history of NCDs | Family history of NCDs | Family history of NCDs | Family history of NCDs |
| Yes | 1250 | 8 (0.6) | 380 (30.4) | 505 (40.4) | 357 (28.6) | |
| No | 600 | 10 (1.7) | 250 (41.7) | 222 (37.0) | 118 (19.6) | |
The prevalence of overweight and obesity among HCWs was $65\%$. Approximately $21.6\%$ of them were female participants and $43.4\%$ were male participants. The most age group of overweight and obesity was 30–39 years with $32.3\%$. Moreover, nurses were the most HCWs who had overweight and obese, with $42.6\%$. In addition, the prevalence of type 2 diabetes and hypertension was higher in a group of overweight and obese, thus, there is a statistically significant association between BMI and type 2 diabetes and hypertension ($P \leq 0.05$; Table 2).
## Association between overweight, obesity, and associated risk factors
Univariate analysis using logistic analysis showed the significant predictors associated (gender, age, marital status, work experience, monthly income, known hypertension, type 2 diabetes, and family history of NCDs) with overweight and obesity among HCWs.
The odds ratio of overweight and obesity increased 1.77 times for male participants than female participants ($95\%$ CI: 1.45–2.15). This relationship was statistically significant ($P \leq 0.001$). Moreover, the age group (40–49) was associated with overweight and obesity by almost three times (OR = 3.20; $95\%$ CI: 2.37–4.32; $P \leq 0.001$). The OR of overweight and obesity among married participants was two times greater than that of unmarried participants (OR = 2.30; $95\%$ CI: 1.84–2.88; $P \leq 0.001$).
Doctors were associated with an increase in overweight and obesity than non-medical professions by 1.67 times (OR = 1.67; $95\%$ CI: 0.99–2.81). This relationship was not statistically significant ($$P \leq 0.051$$). The work experience increased overweight and obesity by 3.8 times for more than 15 years (OR = 3.83; $95\%$ CI: 2.69–5.46; $P \leq 0.001$). Furthermore, the increase in monthly income was associated with overweight and obesity by 1.59 times (OR = 1.59; $95\%$ CI: 1.30–1.95; $P \leq 0.001$).
Hypertension was strongly associated with overweight and obesity (OR = 2.49; $95\%$ CI: 1.65–3.78; $P \leq 0.001$) as well as overweight and obesity was associated with type 2 diabetes (OR = 2.42; $95\%$ CI: 1.21–4.85; $$P \leq 0.012$$). Finally, a family history of NCDs was associated with overweight and obesity (OR = 1.69; $95\%$ CI: 1.38–2.07; $P \leq 0.001$; Table 3).
**Table 3**
| Variables | Regression coefficient B | OR (95% CI) | Wald statistics (df) | P-value |
| --- | --- | --- | --- | --- |
| Gender | Gender | Gender | Gender | Gender |
| Male | 0.572 | 1.77 (1.45–2.15) | 32.90 | <0.001 |
| Female | | 1 | | |
| Age group | Age group | Age group | Age group | Age group |
| 22–29 | | 1 | | |
| 30–39 | 0.873 | 2.393 (1.85–3.09) | 44.83 | <0.001 |
| 40–49 | 1.164 | 3.203 (2.37–4.32) | 57.99 | <0.001 |
| 50–61 | 1.992 | 7.331 (4.51–11.89) | 65.12 | <0.001 |
| Marital status | Marital status | Marital status | Marital status | Marital status |
| Unmarried | | 1 | | |
| Married | 0.835 | 2.305 (1.84–2.88) | 53.17 | <0.001 |
| Profession | Profession | Profession | Profession | Profession |
| Physician | 0.515 | 1.674 (0.99–2.81) | 3.80 | 0.051 |
| Nurse | 0.384 | 1.468 (0.93–2.30) | 2.77 | 0.096 |
| Paramedics | 0.333 | 1.396 (0.85–2.27) | 1.77 | 0.182 |
| Non-medical | | 1 | | |
| Experience (years) | Experience (years) | Experience (years) | Experience (years) | Experience (years) |
| <5 | | 1 | | |
| 5–10 | 0.702 | 2.017 (1.54–2.62) | 27.04 | <0.001 |
| 10–15 | 0.843 | 2.324 (1.78–3.02) | 39.66 | <0.001 |
| >15 | 1.340 | 3.836 (2.69–5.46) | 55.59 | <0.001 |
| Monthly income (NIS) | Monthly income (NIS) | Monthly income (NIS) | Monthly income (NIS) | Monthly income (NIS) |
| <2,000 | | 1 | | |
| >2,000 | 0.466 | 1.594 (1.303–1.951) | 20.50 | <0.001 |
| Number of meals (day) | Number of meals (day) | Number of meals (day) | Number of meals (day) | Number of meals (day) |
| ≤3 | | 1 | | |
| >3 | −0.044 | 0.957 (0.83–1.09) | 0.42 | 0.515 |
| How many people do you usually eat together? | How many people do you usually eat together? | How many people do you usually eat together? | How many people do you usually eat together? | How many people do you usually eat together? |
| ≤5 | | 1 | | |
| >5 | 0.186 | 1.204 (0.98–1.47) | 3.23 | 0.072 |
| Known hypertension | Known hypertension | Known hypertension | Known hypertension | Known hypertension |
| No | | 1 | | |
| Yes | 0.916 | 2.499 (1.65–3.78) | 18.66 | <0.001 |
| Type 2 diabetes | Type 2 diabetes | Type 2 diabetes | Type 2 diabetes | Type 2 diabetes |
| No | | 1 | | |
| Yes | 0.885 | 2.424 (1.21–4.85) | 6.26 | 0.012 |
| Family history of NCDs | Family history of NCDs | Family history of NCDs | Family history of NCDs | Family history of NCDs |
| NO | | 1 | | |
| YES | 0.530 | 1.699 (1.38–2.07) | 26.68 | <0.001 |
## Discussion
Obesity among HCWs is an important issue as it impacts the morbidity of HCWs. In the present study, the overall prevalence of overweight and obesity among HCWs was $65\%$ ($39.3\%$ overweight and $25.7\%$ obesity). These findings are comparable with the meta-analysis in the Middle Eastern countries, which reported the prevalence of overweight and obesity were 33.14 and $21.17\%$, respectively [9]. Moreover, a previous study conducted in the Gaza Strip said the prevalence of obesity was $19.5\%$ [10]. The current research has found a higher prevalence of overweight and obesity among HCWs than the prevalence of overweight and obesity reported in the previous study. When compared to the overall adult population, healthcare workers have a higher prevalence of obesity. This may be because they are more susceptible to the disease due to irregular and extended work hours, poor food, and workplace stress.
There is a significant association between the elevation of the anthropometric measure (Height, weight, WC, HC, SBP, and DBP) and being overweight or obese. It was also noted that the participant's elevation in anthropometric measurements had a significant positive association with being overweight and obese. The higher prevalence of overweight and obesity among HCWs might be influenced by sedentary behaviors, which might be contributed to the work environments and the socioeconomic status that may encourage the adaption of less physical activity and eating habits [31, 32]. One of the contributing factors may be eating a poor diet while watching TV, especially sugary snacks. Other researchers have discovered a link between screen time and sugary, high-energy snacks [33]. Due to the work environment, the HCW's low frequency of intake of meals, low intake of fruits and vegetables, and low physical activity increased overweight and obesity. Further research is needed to investigate dietary habits and socioeconomic status and their association with overweight and obesity among HCWs.
Furthermore, the prevalence of overweight and obesity increased with age. Similar results were reported by Low et al. [ 34], Addo et al. [ 35], Kishawi et al. [ 36], and Firouzbakht et al. [ 37], that the prevalence of overweight and obesity increased as age increased. This places a future load of illness on the medical workforce. In addition, the previous study found that the estimated peak increase in the prevalence of overweight and obesity as age increased was 40–50 years in developing countries while it was 50–60 years in developed countries [34], which is in line with the current study. The peak increase in the prevalence of obesity with age was 30–39 years. In the present study, the peak age is lower than in the previous study. Most of the studies confirmed the association between age increasing and susceptibility to non-communicable diseases and overweight and obesity as one of the risk factors of illness. Overweight and obesity are associated with increased age due to decreased physical activity, routine daily activities, comorbidities, and dietary habits. Therefore, the high ratio of youth and the predominance of the HCWs category found in this study represent the actual age and job category distribution among HCWs in the Gaza Strip.
In addition, the findings from this study show that married participants had a higher prevalence of overweight and obesity than single participants. The results support the findings reported by Dagen et al. [ 38] and Tzotzas et al. [ 39] that married adults had a higher prevalence of being overweight and obese than single adults and hypothesized that the increase in BMI among married couples is due to the increased social support, along with regularly eating dense food that increases the risk of being overweight or obese [39].
Our results revealed that the prevalence of overweight and obese increased among patients who had hypertension and type 2 diabetes. In addition, participants with a positive family history of NCDs had a higher prevalence of overweight and obesity than those without a family history of NCDs, in which similar findings have been found in previous studies (40–43).
Similar to the findings of our investigation, numerous studies have shown that nurses had a higher risk of obesity than workers in other occupations. In their academic careers in Scotland and England [44, 45], it is covered that the risk of obesity was lower for different HCW job categories than for nurses. Unfortunately, Hegde et al. in their studies in Tamil Nadu, India, found results that contradicted our study, suggesting that the burden of obesity was greater among physicians than nurses [46]. All the studies mentioned earlier agree that different job categories were associated with obesity among HCWs. However, the aforementioned studies did not share comparable sociodemographic populations to that of the Gaza Strip. Most studies among HCWs did not specifically analyze different types of HCW occupations and obesity. Hazmi et al. [ 47] only mentioned the overall prevalence of obesity without further analysis according to job category, while Mustafa et al. [ 48] and Ramli [49] only divided the occupations of HCWs into professional vs. ancillary jobs. In contrast, our study is the first in the Gaza Strip to work with HCWs and determine the prevalence of overweight and obesity and classify HCW occupations into four main groups, doctors, nurses, paramedical, and non-medical categories, with the prevalence of overweight and obesity. Thus, we know nurses are at a greater risk of becoming obese than doctors and other job categories in the Gaza Strip.
Our findings reported that eating behavior was significantly associated with increased overweight and obesity, thus, the participants <3 meals per day had increased the prevalence of overweight and obesity in other groups (Table 2). Incomparable to previous studies, eating behavior was reported as one of the leading factors in the development of overweight and obesity with carried gender and age differences (50–53). In addition, no local studies reported an association between eating behavior and the prevalence of obesity.
In addition, physical inactivity is identified as a risk factor for overweight and obesity [54, 55]. However, the current study revealed that all healthcare workers with low or moderate activity had no significant association between physical activity and overweight and obesity. A similar result was reported by Firouzbakht et al. [ 37] and El Kishawi et al. [ 36]. These previous studies recognized the low physical activity due to the working shift of HCWs, limited availability of exercise facilities, and workload.
However, the current study showed significant predictors associated with an increase in the prevalence of overweight and obesity among HCWs. These associated factors included gender, age, marital status, monthly income, known hypertension, type 2 diabetes, family history of NCDs, and elevation of anthropometric measurements. On the other hand, the current study did not find any significant associations between educational level, smoking status, physical activity, and fruit and vegetable consumption.
## Strengths and limitations of the study
This study has several strengths. The first study estimated the prevalence of overweight and obesity among HCWs in the Gaza Strip. This study defined the associated factors with an increase in the prevalence of overweight and obesity; therefore, this study will be a baseline for subsequent studies among HCWs. There are some limitations in the study that need to be considered. First, cross-sectional data do not explore the causal pathways that underlie the reported association. Second, no availability of previous studies about the prevalence of overweight and obesity to interpret our results. Third, recall bias is also possible by using food frequency. Fourth, the majority of our participants were nurses compared with other professions.
## Conclusion
This study showed a high prevalence of overweight and obesity among HCWs. Age, monthly income, marital status, known hypertension, type 2 diabetes, and eating habits were associated with the prevalence of overweight and obesity. These findings appear to show an emerging problem in HCWs. A wellness program should be developed by decision-makers throughout mass-level educational awareness campaigns for HCWs to prevent and manage the modifiable risk factors that increase overweight or obesity.
## Data availability statement
The original contributions presented in the study are included in the article/supplementary material, further inquiries can be directed to the corresponding author.
## Ethics statement
The studies involving human participants were reviewed and approved by Palestinian Health Research Committee at the Directorate General of Human Resources Development, Ministry of Health, Gaza (PHRC/HC/$\frac{663}{19}$). The patients/participants provided their written informed consent to participate in this study.
## Author contributions
LM, JY, and MaM generated the idea for the study and formulated a research plan. JY wrote the original draft preparation and reviewed and edited. MJ, HJ, YF, LW, ZL, ZH, and MeM revised the manuscript and interpreted the data and editing. LM and MaM supervised the study. All authors acquired, analyzed, and interpreted the data. All authors contributed to the article and approved the submitted version.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher's note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## References
1. Utkirzhonovna N. **Prevalence of the main risk factors for overweight and obesity in young people**. *Res J Trauma Disabil Stud.* (2022) **1** 14-25
2. Li Y, Schoufour J, Wang DD, Dhana K, Pan A, Liu X. **Healthy lifestyle and life expectancy free of cancer, cardiovascular disease, and type 2 diabetes: prospective cohort study**. *BMJ.* (2020) **368** 16669. DOI: 10.1136/bmj.l6669
3. Caterson ID. **Medical management of obesity and its complications**. *Ann Acad Med Singapore.* (2009) **38** 22-7. DOI: 10.47102/annals-acadmedsg.V38N1p22
4. Kana M. **From brain drain to brain circulation**. *Jos J Med.* (2009) **4** 8-10. DOI: 10.4314/jjm.v4i1.55092
5. Seibert DJ, Speroni KG, Oh KM, DeVoe MC, Jacobsen KH. **Knowledge, perceptions, and practices of methicillin-resistant**. *Am J Infect Control.* (2014) **42** 254-9. DOI: 10.1016/j.ajic.2013.09.005
6. Lymer UB, Richt B, Isaksson B. **Blood exposure: factors promoting health care workers' compliance with guidelines in connection with risk**. *J Clin Nurs.* (2004) **13** 547-54. DOI: 10.1111/j.1365-2702.2004.00897.x
7. Elessi K, Albaraqouni L. **Prevalence of obesity and overweight in Palestine: a systematic review**. *Lancet.* (2019) **393** S20. DOI: 10.1016/S0140-6736(19)30606-3
8. Riquelme R, Rezende LF, Guzmán-Habinger J, Chávez JL, Celis-Morales C, Ferreccio C. **Non-communicable diseases deaths attributable to high body mass index in Chile**. *Sci Rep.* (2021) **11** 15500. DOI: 10.1038/s41598-021-94974-z
9. Okati-Aliabad H, Ansari-Moghaddam A, Kargar S, Jabbari N. **Prevalence of obesity and overweight among adults in the middle east countries from 2000 to 2020: a systematic review and meta-analysis**. *J Obes.* (2022) **2022** 8074837. DOI: 10.1155/2022/8074837
10. Salah RO, Ghandour R, Husseini A. **Prevalence of overweight, obesity, and associated factors among adolescents in the occupied Palestinian territory: a cross-sectional study**. *Lancet.* (2021) **398** S46. DOI: 10.1016/S0140-6736(21)01532-4
11. Cavazzoni F, Fiorini A, Shoman H, Diab M, Veronese G. **The role of gender and living context in shaping Palestinian children's agency and well-being**. *Gender, Place & Culture.* (2022) **29** 222-47. DOI: 10.1080/0966369X.2021.1882954
12. AlKhaldi M, Kaloti R, Shella D, Al Basuoni A, Meghari H. **Health system's response to the COVID-19 pandemic in conflict settings: policy reflections from Palestine**. *Glob Public Health.* (2020) **15** 1244-56. DOI: 10.1080/17441692.2020.1781914
13. Lasalvia A, Amaddeo F, Porru S, Carta A, Tardivo S, Bovo C. **Levels of burn-out among healthcare workers during the COVID-19 pandemic and their associated factors: a cross-sectional study in a tertiary hospital of a highly burdened area of north-east Italy**. *BMJ Open.* (2021) **11** e045127. DOI: 10.1136/bmjopen-2020-045127
14. Sentell T, Vamos S, Okan O. **Interdisciplinary perspectives on health literacy research around the world: more important than ever in a time of COVID-19**. *Public Health.* (2020) **17** 3010. DOI: 10.3390/ijerph17093010
15. Haldane V, De Foo C, Abdalla SM, Jung A-S, Tan M, Wu S. **Health systems resilience in managing the COVID-19 pandemic: lessons from 28 countries**. *Nat Med.* (2021) **27** 964-80. DOI: 10.1038/s41591-021-01381-y
16. Ruettger K, Varela-Mato V, Chen Y-L, Edwardson CL, Guest A, Gilson ND. **Physical activity, sedentary time, and cardiometabolic health in heavy goods vehicle drivers**. *J Occup Environ Med.* (2022) **64** e217-23. DOI: 10.1097/JOM.0000000000002484
17. Philpot LM, Ebbert JO, Hurt RT. **A survey of the attitudes, beliefs and knowledge about medical cannabis among primary care providers**. *BMC Fam Pract.* (2019) **20** 17. DOI: 10.1186/s12875-019-0906-y
18. Roy D, Tripathy S, Kar SK, Sharma N, Verma SK, Kaushal V. **Study of knowledge, attitude, anxiety & perceived mental healthcare need in Indian population during COVID-19 pandemic**. *Asian J Psychiatry.* (2020) **51** 102083. DOI: 10.1016/j.ajp.2020.102083
19. Zhao W, Jones C, Wu ML, Moyle W. **Healthcare professionals' dementia knowledge and attitudes towards dementia care and family carers' perceptions of dementia care in China: an integrative review**. *J Clin Nurs.* (2022) **31** 1753-75. DOI: 10.1111/jocn.15451
20. Perry L, Xu X, Gallagher R, Nicholls R, Sibbritt D, Duffield C. **Lifestyle health behaviors of nurses and midwives: the ‘fit for the future'study**. *Int J Environ Res Public Health.* (2018) **15** 945. DOI: 10.3390/ijerph15050945
21. Chobanian AV, Bakris GL, Black HR, Cushman WC, Green LA, Izzo JL. **The seventh report of the joint national committee on prevention, detection, evaluation, and treatment of high blood pressure: the JNC 7 report**. *JAMA.* (2003) **289** 2560-71. DOI: 10.1001/jama.289.19.2560
22. Kadam P, Bhalerao S. **Sample size calculation**. *Int J Ayurveda Res.* (2010) **1** 55. DOI: 10.4103/0974-7788.59946
23. Rodriguez-Munoz S, Corella C, Abarca-Sos A, Zaragoza J. **Validation of three short physical activity questionnaires with accelerometers among university students in Spain**. *J Sports Med Phys Fitness.* (2017) **57** 1660-8. DOI: 10.23736/S0022-4707.17.06665-8
24. Imaeda N, Goto C, Sasakabe T, Mikami H, Oze I, Hosono A. **Reproducibility and validity of food group intake in a short food frequency questionnaire for the middle-aged Japanese population**. *Environ Health Prev Med.* (2021) **26** 28. DOI: 10.1186/s12199-021-00951-3
25. Hamdan M, Monteagudo C, Lorenzo-Tovar M, Tur J, Olea-Serrano F, & Mariscal-Arcas. **Development and validation of a nutritional questionnaire for the Palestine population**. *Public Health Nutr.* (2014) **17** 2512-8. DOI: 10.1017/S1368980013002711
26. Consultation WE. **Appropriate body-mass index for Asian populations and its implications for policy and intervention strategies**. *Lancet.* (2004) **363** 157-63. DOI: 10.1016/S0140-6736(03)15268-3
27. 27.World Health Organization. Waist Circumference and Waist-Hip Ratio: Report of a WHO Expert Consultation. Geneva, 8-11 December 2008.. WHO (2011).. *Waist Circumference and Waist-Hip Ratio: Report of a WHO Expert Consultation. Geneva, 8-11 December 2008* (2011)
28. Rocha S, Mota J, Furtado G, Santos C, Dantas E, Coutinho A. **Association between body mass index and chronic non-communicable diseases among the elderly**. *Moj Gerontol Geriatr.* (2017) **1** 120-6. DOI: 10.15406/mojgg.2017.01.00026
29. Anuurad E, Shiwaku K, Nogi A, Kitajima K, Enkhmaa B, Shimono K. **The new BMI criteria for asians by the regional office for the western pacific region of WHO are suitable for screening of overweight to prevent metabolic syndrome in elder Japanese workers**. *J Occup Health.* (2003) **45** 335-43. DOI: 10.1539/joh.45.335
30. Lemogoum D, Seedat YK, Mabadeje AF, Mendis S, Bovet P, Onwubere B. **Recommendations for prevention, diagnosis and management of hypertension and cardiovascular risk factors in sub-Saharan Africa**. *J Hypertens.* (2003) **21** 1993-2000. DOI: 10.1097/00004872-200311000-00003
31. Choi B, Schnall PL, Yang H, Dobson M, Landsbergis P, Israel L. **Sedentary work, low physical job demand, and obesity in US workers**. *Am J Ind Med.* (2010) **53** 1088-101. DOI: 10.1002/ajim.20886
32. Ismail IM, Kulkarni AG, Kamble SV, Borker SA, Rekha R, Amruth M. **Prevalence of hypertension and its risk factors among bank employees of Sullia Taluk, Karnataka**. *Sahel Med J.* (2013) **16** 139. DOI: 10.4103/1118-8561.125553
33. Kerkadi A, Sadig AH, Bawadi H, Al Thani AAM, Al Chetachi W, Akram H. **The relationship between lifestyle factors and obesity indices among adolescents in Qatar**. *Int J Environ Res Public Health.* (2019) **16** 4428. DOI: 10.3390/ijerph16224428
34. Low S, Chin MC, Deurenberg-Yap M. **Review on epidemic of obesity**. *Ann Acad Med Singap.* (2009) **38** 57. DOI: 10.47102/annals-acadmedsg.V38N1p57
35. Addo PN, Nyarko KM, Sackey SO, Akweongo P, Sarfo B. **Prevalence of obesity and overweight and associated factors among financial institution workers in Accra Metropolis, Ghana: a cross sectional study**. *BMC Res Notes.* (2015) **8** 599. DOI: 10.1186/s13104-015-1590-1
36. Kishawi E, Rafiq R, Soo KL, Abed YA, Muda WAMW. **Prevalence and predictors of overweight and obesity among women in the Gaza strip-Palestine: a cross-sectional study**. *BMC Public Health.* (2020) **20** 864. DOI: 10.1186/s12889-020-08966-1
37. Firouzbakht M, Riahi ME, Hajian-Tilaki K, Ebadi A, Tirgar A, Nikpour M. **Relationship of social capital with overweight and obesity among female health care workers**. *Caspian J Intern Med.* (2019) **10** 281. PMID: 31558989
38. Dagne S, Gelaw YA, Abebe Z, Wassie MM. **Factors associated with overweight and obesity among adults in northeast Ethiopia: a cross-sectional study**. *Diabetes, Metab Syndr Obes.* (2019) **12** 391. DOI: 10.2147/DMSO.S179699
39. Tzotzas T, Vlahavas G, Papadopoulou SK, Kapantais E, Kaklamanou D, Hassapidou M. **Marital status and educational level associated to obesity in Greek adults: data from the national epidemiological survey**. *BMC Public Health.* (2010) **10** 732. DOI: 10.1186/1471-2458-10-732
40. Aung TNN, Shirayama Y, Moolphate S, Lorga T, Jamnongprasatporn W, Yuasa M. **Prevalence and risk factors for hypertension among Myanmar migrant workers in Thailand**. *Int J Environ Res Public Health.* (2022) **19** 3511. DOI: 10.3390/ijerph19063511
41. Paquissi FC, Manuel V, Manuel A, Mateus GL, David B, Béu G. **Prevalence of cardiovascular risk factors among workers at a private tertiary center in Angola**. *Vasc Health Risk Manag.* (2016) **12** 497. DOI: 10.2147/VHRM.S120735
42. Rahmouni K, Correia ML, Haynes WG, Mark AL. **Obesity-associated hypertension: new insights into mechanisms**. *Hypertension.* (2005) **45** 9-14. DOI: 10.1161/01.HYP.0000151325.83008.b4
43. Capingana DP, Magalhães P, Silva AB, Gonçalves MA, Baldo MP, Rodrigues SL. **Prevalence of cardiovascular risk factors and socioeconomic level among public-sector workers in Angola**. *BMC Public Health.* (2013) **13** 732. DOI: 10.1186/1471-2458-13-732
44. Kyle RG, Wills J, Mahoney C, Hoyle L, Kelly M, Atherton IM. **Obesity prevalence among healthcare professionals in England: a cross-sectional study using the health survey for England**. *BMJ Open.* (2017) **7** e018498. DOI: 10.1136/bmjopen-2017-018498
45. Kyle RG, Neall RA, Atherton IM. **Prevalence of overweight and obesity among nurses in Scotland: a cross-sectional study using the Scottish health survey**. *Int J Nurs Stud.* (2016) **53** 126-33. DOI: 10.1016/j.ijnurstu.2015.10.015
46. Hegde SKB, Sathiyanarayanan S, Venkateshwaran S, Sasankh A, Ganeshkumar P, Balaji R. **Prevalence of diabetes, hypertension and obesity among doctors and nurses in a medical college hospital in Tamil Nadu, India**. *World.* (2015) **1** 504-11
47. Hazmi H, Ishak WRW, Jalil RA, Hua GS, Hamid NF, Haron R. **Traditional cardiovascular risk-factors among healthcare workers in Kelantan, Malaysia**. *Southeast Asian J Trop Med Public Health.* (2015) **46** 504-11. DOI: 10.22088/cjim.10.3.281
48. Mustafa J, Salleh NM, Isa ZM, Ghazi HF. **Overweight problem among primary health care workers in Suburban District of Hulu Langat, Selangor, Malaysia**. *Pakistan J Nutr.* (2013) **12** 291. DOI: 10.3923/pjn.2013.291.296
49. Ramli A. **Obesity and habitual physical activity level among staffs working in a military hospital in Malacca, Malaysia**. *IIUM Med J Malaysia.* (2013) **12** 53-7. DOI: 10.31436/imjm.v12i1.2003
50. Kiro LS, Zak MY, Chernyshov OV, Nikolenko AE, Iakovenko NO. **Eating behaviour and obesity: gender-age features**. *Wiadomosci Lekarskie.* (2021) **74** 1114-9. DOI: 10.36740/WLek202105112
51. Gross AC, Fox CK, Rudser KD, Foy AM, Kelly AS. **Eating behaviours are different in youth with obesity and severe obesity**. *Clin Obes.* (2016) **6** 68-72. DOI: 10.1111/cob.12127
52. Boriak KR, Vesnina LE. **Relationship of the eating behavior features with anthropometric indicators and energy value of the diet in young people with normal weight and overweight**. *Wiad Lek.* (2020) **73** 2586-90. DOI: 10.36740/WLek202012106
53. Shevchenko Y, Vesnina L, Kaydashev I. **Features of eating behavior in persons with normal and increased body weight**. *Fiziolohichnyi Zhurnal.* (2015) **61** 51-8. DOI: 10.15407/fz61.03.051
54. Shayo GA, Mugusi FM. **Prevalence of obesity and associated risk factors among adults in Kinondoni municipal district, Dar es Salaam Tanzania**. *BMC Public Health.* (2011) **11** 365. DOI: 10.1186/1471-2458-11-365
55. Alghamdi AS, Yahya MA, Alshammari GM, Osman MA. **Prevalence of overweight and obesity among police officers in Riyadh City and risk factors for cardiovascular disease**. *Lipids Health Dis.* (2017) **16** 1-5. DOI: 10.1186/s12944-017-0467-9
|
---
title: On the origin of the functional versatility of macrophages
authors:
- Adam Bajgar
- Gabriela Krejčová
journal: Frontiers in Physiology
year: 2023
pmcid: PMC9998073
doi: 10.3389/fphys.2023.1128984
license: CC BY 4.0
---
# On the origin of the functional versatility of macrophages
## Abstract
Macrophages represent the most functionally versatile cells in the animal body. In addition to recognizing and destroying pathogens, macrophages remove senescent and exhausted cells, promote wound healing, and govern tissue and metabolic homeostasis. In addition, many specialized populations of tissue-resident macrophages exhibit highly specialized functions essential for the function of specific organs. Sometimes, however, macrophages cease to perform their protective function and their seemingly incomprehensible response to certain stimuli leads to pathology. In this study, we address the question of the origin of the functional versatility of macrophages. To this end, we have searched for the evolutionary origin of macrophages themselves and for the emergence of their characteristic properties. We hypothesize that many of the characteristic features of proinflammatory macrophages evolved in the unicellular ancestors of animals, and that the functional repertoire of macrophage-like amoebocytes further expanded with the evolution of multicellularity and the increasing complexity of tissues and organ systems. We suggest that the entire repertoire of macrophage functions evolved by repurposing and diversification of basic functions that evolved early in the evolution of metazoans under conditions barely comparable to that in tissues of multicellular organisms. We believe that by applying this perspective, we may find an explanation for the otherwise counterintuitive behavior of macrophages in many human pathologies.
## Introduction
The human body is made up of more than two hundred types of cells (Castillo-Armengol et al., 2019). Unlike most cell types, macrophages display a striking level of functional versatility and an extraordinary degree of autonomy (Locati et al., 2020).
Macrophages represent the front line of the immune system, responsible for the recognition, phagocytosis, and elimination of pathogens and to control the inflammatory response by instructing other branches of the immune system via cytokine signaling (Cole et al., 2014). However, macrophage function is not limited to protection against foreign organisms. Macrophages are also involved in many homeostatic processes in the body (Biswas and Mantovani, 2012; Theret et al., 2019). Every day, millions of cells die in the human body and the constant substitution of cells and reconstitution of the extracellular matrix (ECM) governed by macrophages is fundamental for the health of any tissue in the body (Kwon et al., 2019; Batista-Gonzalez et al., 2020; Sender and Milo, 2021; Witherel et al., 2021).
Macrophages exhibit many specific characteristics predisposing them to be highly effective in the above functions. Macrophages are highly motile and crawl through the organism toward the site where they are needed (Xuan et al., 2015). Once in place, macrophages are sensitive to external signals and respond according to external conditions (Lavin and Merad, 2013). Their functional repertoire includes engulfing pathogens and removing damaged, senescent, or apoptotic cells. Internalized cellular material is processed and metabolically degraded in the phagolysosome. To this end, macrophages exhibit many specific metabolic pathways for processing and interconversion of phagocytosed organic material. In addition to sensing external signals, macrophages also excel in the production of a broad spectrum of signaling factors. Macrophages are central producers of cytokines in the body and are actively involved in interorgan signaling and regulation of homeostasis in healthy and pathological conditions (Arango Duque and Descoteaux, 2014).
The ability of macrophages to perform such a wide repertoire of functions is largely due to their metabolic plasticity. Sentinel macrophages typically reside in a quiescent state, referred to as M0, which serves as a baseline metabolic profile. From this state, macrophages can undergo metabolic polarization into various forms in response to different stimuli. Thus, various external factors trigger a specific macrophage expression program that leads to the modulation of major metabolic pathways to generate sufficient energy and specific metabolites required for an adequate functional response (Galvan-Pena and O'Neill, 2014). Therefore, metabolic polarization allows macrophages to adopt a specific functional polarization phenotype and perform unique functions efficiently (Liu et al., 2021). It was originally described that macrophages adopt two polarization phenotypes, defined as bactericidal (also known as pro-inflammatory; classically activated or M1) or healing (also known as anti-inflammatory; alternatively activated or M2) (Viola et al., 2019). However, more recent research has revealed many divergences from the polarized M1 and M2 types, such as metabolically activated macrophages (MMe) or macrophages activated by oxidized phospholipid (Mox) (Coats et al., 2017). Currently, the prevailing view is that M1 and M2 macrophages represent the two extremes of the entire continuum of all possible polarization phenotypes.
In addition to the general pro-inflammatory and homeostatic functions common to all macrophages, the population of tissue-resident macrophages found in virtually all tissues of the human body often perform highly specialized tasks (Nobs and Kopf, 2021). Among many others, some of the most well-studied tissue-resident macrophages include Kupffer cells in the liver, and microglia in the central nervous system, alveolar macrophages in the lungs, Langerhans cells in the skin, or peritoneal and adipose tissue macrophages (Wu and Hirschi, 2021). The progenitors of these tissue-resident macrophages migrate to destination tissue during embryonic development and their populations are sustained throughout the life of the individual by self-replication (Davies et al., 2013; Munro and Hughes, 2017). Tissue resident macrophages are functionally shaped by signaling factors characteristic for their particular tissue environment and exhibit distinct functional and morphological phenotypes. The role of tissue-resident macrophages ranges from fundamental functions, such as antibacterial responses and removal of dead and senescent cells, to advanced functions, such as promoting stem cell proliferation, regulating local and systemic metabolism, promoting lipid metabolism and thermogenesis, controlling sinoatrial node action potential, governing hematopoiesis, regulating synaptic pruning, inducing vascularization, and removing amyloid plaques and other potentially harmful substances from the extracellular space (Gordon and Plüddemann, 2017).
From the preceding paragraphs, it is clear that the mononuclear phagocyte system represents a central system for maintaining homeostasis that controls many physiological processes. However, the role of macrophages in the organism is not beneficial in all circumstances, and macrophages also play a significant role in the induction of several pathological conditions (Sica et al., 2015).
Macrophages may become inadequately activated in response to external stimuli, resulting in behavior that may appear counterintuitive in certain situations (Parisi et al., 2018). Excessive production of pro-inflammatory factors or excessive deposition of ECM components often leads to tissue and organ dysfunction and progressive development of pathology.
Excessive pro-inflammatory macrophage polarization is typically observed in obesity, non-alcoholic fatty liver disease, atherosclerosis, and neurodegenerative diseases (Lauterbach and Wunderlich, 2017; Mammana et al., 2018; Barreby et al., 2022). Likewise, chronic adoption of M2 macrophage polarization is associated with liver fibrosis, chronic obstructive pulmonary disease, Alzheimer’s disease, or cancer (Wang L. et al., 2019; Zhou et al., 2020). Pathologies in which macrophage activation plays a critical role are not limited to those listed here. In fact, lack of macrophage polarization plasticity in any tissue inevitably progresses to pathology. Nevertheless, the rationale behind the switch from the primarily protective role of macrophages to induction of pathology remains largely undetermined.
Fascinated by the functional versatility of macrophages, we seek to understand why macrophages have such an unusual degree of autonomy and responsibility. Understanding the evolutionary origins of macrophages may provide insight into how they have acquired critical properties necessary for their protective and homeostatic roles.
To reveal the origin of macrophages and their functional versatility, we decided to trace the characteristic features of mammalian macrophages back in the evolution of the animals. While investigating the origin of macrophage-like cells in the animal phyla, we realized that macrophage-like amoebocytes are present in virtually all multicellular animals.
We surmise that macrophage functional versatility reflects the ancient origin of these cells in free-living unicellular animals and that macrophage functional repertoire has further expanded with the emergence of multicellularity and the increasing complexity of the body plan of multicellular animals.
Information regarding unicellular animals and the emergence of multicellular animals is fragmented and can be inferred only from indirect evidence. Therefore, we decided to investigate the functional analogy between mammalian macrophages and free-living predatory amoeba (Acanthamoeba; Protists) (Tice et al., 2016). We then combined this with knowledge from the clades represented by unicellular animals (Choanoflagellatea, Filasterea, Ichtyosporea; Holozoa) (Hehenberger et al., 2017) to formulate an idea of what functions may have already been present in the unicellular free-living ancestor of animals.
We next set out to compare the characteristic features of mammalian macrophages with those observed in a social facultative multicellular amoeba (Dictyostelium discoideum; Amoebozoa) (Romeralo et al., 2011) to explore the possibility that the emergence of multicellularity has gone along with the expansion of the functional repertoire of macrophage-like amoebocytes.
Following this idea, we compare the functions known in mammalian macrophages with those observed in macrophage-like amoebocytes in sponges (Porifera; Holozoa), which represent multicellular animals without yet fully differentiated tissues and organs (Nielsen, 2019) and can thus provide some indication of what functions might be present in macrophage ancestors at the emergence of multicellular organisms.
Subsequently, we analyzed the characteristics of primitive macrophage-like plasmatocytes in the fruit fly (Drosophila melanogaster; Metazoa, Insecta) as a representative of a simple animal with fully developed tissues and organs at a level of complexity comparable to that of mammals (Cheng et al., 2018). For a historical perspective on the early discoveries of macrophage functional variability, see Box 1. The phylogenetic relationship of the compared clades and lineages is shown in the Figure 1.BOX 1 Metchnikoff’s predictions and the discovery of macrophagesMore than a century has passed since Metchnikoff formulated his theory of phagocytosis as the central mechanism of the immune response, for which he was awarded the Nobel Prize (Kaufmann, 2008). The attention this hypothesis attracted in the scientific community has unfortunately overshadowed many of the other postulates Metchnikoff made regarding the function of macrophages in the body. These speculations become particularly interesting in light of current knowledge about the function of macrophages, which goes far beyond their bactericidal function in the organism (Tauber, 2003).Metchnikoff discovered the immune role of macrophages when studying the function of mesodermal amoeboid cells moving freely in the body of primitive multicellular organisms. In doing so, he paid close attention to the role these cells play in nutrient acquisition in organisms that do not have a digestive cavity and identified how these cells shape multicellular organisms during evolution and ontogeny (Merien, 2016). Metchnikoff proposed that complex multicellular organisms are inherently disharmonious and that macrophages induce physiological inflammation to achieve a harmonious whole (Tauber, 2017).Metchnikoff’s exceptional observational skills and work ethic led him to recognize the importance of macrophages in maintaining nutritional, metabolic, and tissue homeostasis more than a century before the confirmation of this phenomenon by current molecular biological research.
**FIGURE 1:** *The typical macrophage behavior in tissues can be split into several consecutive phases. Macrophages first perceive activation signals indicating changes in tissue homeostasis. Subsequently, macrophages migrate against the concentration gradient toward the source of activating signals. Macrophages infiltrating disharmonious tissue are exposed to local signals in the form of pathogen-associated molecular patterns (PAMPs), danger-associated molecular patterns (DAMPs), and a cocktail of pro-inflammatory and anti-inflammatory cytokines. Based on extrinsic cues and intrinsic predetermination, macrophages adjust their metabolic setup and functionally polarize to M1 and M2 polarization phenotypes (in a simplistic view of the problematics). Macrophages, in an effort to resolve a stressful situation, engulf and eliminate pathogenic bacteria, or remove senescent and dysfunctional cells to restore tissue homeostasis.*
## Macrophage versatility is based on a few fundamental macrophage features
Although macrophages perform many functions in the body, their behavior can be divided into several basic features that make them distinctly different from all other cells in the body. Generally, macrophages reside in the tissue in a quiescent state and calmly perceive signals from the environment (Holt and Grainger, 2012). Macrophages are equipped with a number of receptors for the recognition of chemoattractants and signaling substances that originate from indisposed cells and tissues, other immune cells, or produced by bacterial pathogens as their secondary metabolites. Most of the receptors recognizing the chemoattractant signals belong to the class of G-protein coupled receptors (GPCRs), such as formyl peptide receptor, folate receptor, adenosine receptor, purinergic receptors, and various chemokine receptors (Kim, 2018; O’Callaghan et al., 2021).
Upon chemokine recognition, the GPCR activates intracellular signaling that constitutes G-protein and arrestin as second messengers and leads to the activation of common stress response-related signaling cascades, such as PKC, PI3K-Akt, MAPK-ERK, AP, JAK-STAT, etc. The induced transcriptional program leads to increased cytoskeleton reorganization, cell shape changes, directed motility, secretion of lysosomal enzymes, phagocytosis, and activation of the respiratory burst (Wang X. et al., 2019).
Macrophages are chemotactically guided through the environment against the concentration gradient of extracellular chemical stimuli, such as chemokines, polyunsaturated fatty acid metabolites (leukotrienes and eicosanoids), components of the complement cascade (C3a, C5a), or formyl peptides (Sokol and Luster, 2015). Unlike most cell types in the mammalian organism, macrophages exhibit active migration, facilitated by rapid remodeling of the actin cytoskeleton. Macrophages primarily use two distinct types of migration, namely amoeboid and mesenchymal. Amoeboid migration is a rapid movement driven by an actin-rich pseudopod at the leading edge, hydrostatically generated blebs, and a highly contractile uropod at the trailing edge. This movement is characterized by weak or absent adhesion to the substrate and low-level proteolysis of the ECM. In contrast, mesenchymal movement is characterized by cell adhesion to the substrate via integrins, cadherins, or fibronectins and requires enzymatic disruption of binding to the ECM (Pizzagalli et al., 2022).
To effectively distinguish various pathogens from the body’s own cells, macrophages must sense and recognize specific pathogen-associated antigens on the surface of the foreign cells. These molecules are recognized by immune-cell-specific receptors called pattern recognition receptors (PRRs) (Amarante-Mendes et al., 2018). Mammalian macrophages exhibit a wide spectrum of PPRs, categorized into several classes according to their structure. Many of these receptors, such as toll-like receptor family, scavenger receptors, c-type lectins, or NOD-like receptors, are evolutionarily ancient, and their ability to recognize pathogen-associated molecular patterns (PAMPs) has been shaped over the billions of years of coevolution between pathogen and host (Li and Wu, 2021). Antigen binding to PRR activates macrophage immune-related cascades, such as NFĸB, ERK, JNK, and p38, which initiate complex signaling cascades that allow remodeling of the macrophage cytoskeleton and formation of membrane invaginations to engulf the particle and form a phagosome. Subsequently, the primary phagosome fuses with acidic lysosomes, which contain a mixture of enzymes that cleave the phagocytosed material. During the respiratory burst, the NADPH oxidase NOX2 pumps massive amounts of reactive oxygen species (ROS) into the phagolysosome to destroy its contents. Elimination of pathogenic bacteria is enhanced by the activity of natural resistance-associated macrophage proteins (NRAMP) transporters, which pump divalent ions onto the phagolysosome lumen. Additionally, macrophages polarize toward a pro-inflammatory state, releasing a mixture of pro-inflammatory cytokines and opsonizing factors (Mogensen, 2009).
The underlying mechanism enabling these changes is the modification of cellular metabolism. Strikingly, pro-inflammatory macrophages adopt aerobic glycolysis as the predominant method of ATP production, driven by the stabilization of the transcription factor HIF1α (Wang et al., 2017). While oxidative phosphorylation in mitochondria generates significantly more ATP per glucose molecule, M1 macrophages favor aerobic glycolysis, likely due to the rate of ATP production. In addition, aerobic glycolysis allows increased NADPH production in the pentose phosphate pathways, which is used as a building block for many biomolecules. Since pyruvate is converted to lactate by lactate dehydrogenase and excreted from the cell, the TCA cycle is supplemented with glutamine causing it to be “interrupted” or “rewired”. As a result, TCA cycle intermediates accumulate and contribute to further stabilization of HIF1α. At the same time, mitochondria, which are liberated from generating ATP in oxidative phosphorylation, instead generate ROS by the reversed electron flux at the respiratory chain complex1 (Viola et al., 2019). M1 polarization is also characterized by the specific utilization of arginine, which is converted by L-arginase to citrulline, and growth-inhibiting NO, which is transported to the phagolysosome (Palmieri et al., 2020). M1 polarization is associated with the production of pro-inflammatory cytokines, such as IL-1, IL-6, or INFγ, which further inform other cells of danger (Nonnenmacher and Hiller, 2018). Once the pathogen is eliminated, the immune response is not yet complete, M2 macrophages need to be recruited to promote the resolution of inflammation and restore homeostasis.
In addition to pathogenic activation, macrophages are activated by signals produced by damaged, metabolically stressed cells and tissues, known as DAMPs (danger-associated molecular patterns), leading to M2 macrophage polarization (Ferrante and Leibovich, 2012). While the functions of M1 macrophages are relatively simple, the functions of M2 macrophages are more diverse. The main goals of M2 macrophages are to resolve inflammation, protect against viral and fungal infections, promote angiogenesis, facilitate ECM remodeling, support tissue healing, and regeneration, and remove senescent and damaged cells by efferocytosis (Wang L. et al., 2021). One of the important properties of M2 macrophages is the maintenance of immunological tolerance, i.e., the prevention of an immune reaction against host antigens. Thus, their function is particularly crucial in organs that must tolerate foreign antigens, such as those of the developing fetus or developing spermatids in the testis (Porta et al., 2009). This tolerogenic property also allows the presence of symbiotic bacteria. However, excessive adoption of M2 macrophage polarization may become detrimental as it induces tissue fibrosis, leading to chronic infections and promotion of tumor cell growth (Lin et al., 2019).
M2 macrophages differ significantly from their M1 counterparts in cellular metabolism, which determines their different function. While the amino acid arginine serves as a substrate for iNOS in M1 macrophages, as it is essential for the production of ROS (Rodriguez et al., 2017), M2 macrophages primarily use arginine as a substrate for arginase, promoting its conversion to ornithine and urea. Ornithine is subsequently used as a substrate for forming ECM components, making M2 macrophages essential contributors to tissue regeneration and wound healing (Szondi et al., 2021). Hence, after the elimination of pathogenic invaders, pro-inflammatory macrophages are gradually replaced by M2 macrophages, which trigger the regeneration of the wounded tissue and promote vascularization, ECM synthesis, and inflammation resolution. In addition, M2 macrophages participate in ECM remodeling by producing matrix metalloproteases, cathepsins, and other enzymes that reorganize collagen fibers and by modulating fibroblast function (Witherel et al., 2021).
M2 macrophages are also responsible for maintaining tissue homeostasis under physiological conditions by detecting and removing apoptotic and damaged cells through efferocytosis. The term “efferocytosis” was introduced by deCathelineau and Henson, [2003] to describe the phagocytosis of apoptotic cells. Unlike phagocytosis of foreign objects, which triggers inflammation and antigen presentation, efferocytosis of apoptotic cells upregulates anti-inflammatory cytokines and compounds promoting tissue healing. During efferocytosis, macrophages are guided chemotactically to apoptotic and senescent cells through the detection of “find me” signals, such as nucleotides (ATP, ADP, or UDP), lysophosphatidylcholine, or sphingosine-1-phosphate (Ravichandran, 2010). The receptors responsible for recognizing apoptotic cells differ from those involved in phagocytosis. Subsequently, macrophages respond to “eat me” signal molecules, such as phosphatidylserine, oxidized phospholipids, DNA, or annexin A1, exposed on the surface of the cells destined for efferocytosis. While the engulfment process resembles macropinocytosis, the machinery fusing the efferosome with the lysosome is analogous to phagolysosome formation. Therefore, M2 macrophages exhibit a wide spectrum of enzymes that can metabolize phospholipids and DNA fragments and neutralize otherwise dangerous modified lipids and proteins (Martin et al., 2014).
M2 macrophages can be divided into different polarization subtypes, such as M2a, M2b, M2c, and M2d, based the on the specific cocktail of chemokines, cytokines, and growth factors they polarize with and subsequently produce (Ross et al., 2021). In addition, plethoras of polarization phenotypes have also been described in the context of hypertrophic adipocytes or atherosclerotic plaques. For example, ingestion of heme by macrophages leads to the adoption of the Mhem polarization phenotype, internalization of hemoglobin to M (Hb), and the exposure of oxidized lipids to Mox (Lin et al., 2021). Since these macrophage subsets are often characterized only by mammalian-specific surface markers and do not exhibit a characteristic functional profile, tracking them during evolution is impossible (Natoli and Monticelli, 2014). For this reason, in this paper, we focus only on the functionally well-characterized macrophages M1 and M2 as representatives of the two phenotypic extremes.
Overall, macrophages are truly unique cells of the animal body that can play many roles in different tissues and body contexts by combining several specific properties. In particular, macrophages are exceptional at sensing chemotactic signals, exhibiting controlled active motility, recognizing molecular patterns associated with pathogen or tissue damage, and adopting metabolic and functional polarization accordingly. These properties predispose them to deal with stressful situations in the body (Figure 2).
**FIGURE 2:** *Simplified phylogenic tree of Holozoa and their relatives. We have analyzed the occurrence of characteristic features of mammalian macrophages in unicellular free-living amoeba (Acanthamoeba; 1), social facultative-multicellular amoeba (Dictyostelium; 2), macrophage-like amoebocytes in primitive multicellular animal lacking true tissues and organs (Porifera; 3), and macrophage-like plasmatocytes of simple multicellular organisms with fully developed organs and tissues (Drosophila; 4). We propose that many macrophage characteristics are inherited from unicellular ancestors of animals. The functional repertoire of macrophages then diversified with the emergence of multicellularity and increasing complexity of body plan and development of organ systems.*
## Free-living predatory amoebas share many similarities with M1 macrophages
By comparing the characteristics of mammalian macrophages with the prey-hunting strategies of free-living amoebae, we can find surprising similarities. Acanthamoeba and macrophages share the principal mechanisms used for chemotaxis towards bacteria, motility, interaction with bacteria, phagocytosis, the killing of bacteria in the phagolysosome, and production of antimicrobial peptides (Siddiqui and Khan, 2012a).
The underlying molecular mechanisms show a remarkable degree of similarity, documented by the fact that human intracellular pathogens use the same strategies to escape the bactericidal mechanism in the macrophage and Acanthamoeba (Molmeret et al., 2005). Therefore, *Acanthamoeba is* often viewed as a training ground for microbial organisms to become successful human and animal pathogens and a melting pot for horizontal gene transfer between different bacterial strains (Salah et al., 2009).
Acanthamoeba is a free-living heterotrophic Protist that specializes in hunting microbes for its nutritional needs. Acanthamoeba has two life stages; an active trophozoid or a dormant double-walled cyst, which can withstand adverse environmental conditions for long periods of time (Siddiqui and Khan, 2012b). In terms of life strategy, Acanthamoeba as professional phagocytic bactericidal omnivores do not differ significantly from the basal groups of Holozoa and are not expected to substantially differ from unicellular ancestors of animals (Lang et al., 2002).
Immediately, we can discern similarities between Acanthamoeba and mammalian macrophages with respect to size, behavior, cellular ultrastructure, and chemical composition (Rayamajhee et al., 2022). Like mammalian macrophages, Acanthamoeba can sense chemical signals from the environment and approach the signal source by chemotaxis through motility based on actin and myosin remodeling (Swart et al., 2018). In-depth studies of chemotactic factors have identified various bacterial metabolic products, such as formyl-methionyl-leucyl-phenylalanine, lipopolysaccharide, lipoteichoic acid, cAMP, lipid A, or N-acetylglucosamine. In analogy to mammalian macrophages, the perception of chemotactic signals in *Acanthamoeba is* mediated via GPCRs (Schuster and Levandowski, 1996). Most of these signals are products of bacterial metabolism or fragments of surface bacterial macromolecules and also serve as potent chemoattractants for mammalian macrophages and neutrophils (Nadesalingam et al., 2005).
A detailed study of crawling in free-living amoebae revealed that the migratory mechanisms used by macrophages and amoebae are identical (Campolo et al., 2021), indicating their ancient origin in the common ancestor of Amoebozoa and Opisthokonta.
Once macrophages approach the site of origin of chemotactic signals, they must recognize which cells are to be engulfed and eliminated in the phagolysosome. Many of the receptors used by macrophages to recognize pathogenic bacteria can also be found in some form in Acanthamoeba. For instance, the C-type lectin mannose receptor, which is abundantly expressed by mammalian macrophages, is used by Acanthamoeba to identify prey and engulf it (Allen and Dawidowicz, 1990).
Acanthamoeba recognizes and binds the bacteria, and the subsequent processes of phagocytosis and destruction of the pathogen show a high degree of similarity to mammalian macrophages. Pathogen recognition leads to massive reorganization of F-actin filaments in both macrophages and Acanthamoeba, resulting in dynamic probing, disruption of the cortical F-actin layer, nucleation and polymerization of F-actin filaments, phagosome closure, and particle internalization (Bowers, 1977; Alsam et al., 2005). Internalized bacteria are inactivated and enzymatically processed in the phagolysosome. Ultrastructural analysis of Acanthamoeba revealed that they contain many lysosomes containing a cocktail of degradative enzymes (Alsam et al., 2005; Salah et al., 2009). After fusing the phagosome with the lysosome, V-ATPases embedded in the phagolysosomal membrane pump hydrogen ions inside the phagolysosome to acidify the phagolysosomal lumen (Akya et al., 2009). The bacteria are then exposed to superoxide ions and hydrogen peroxide, in a process called oxidative burst. The active form of oxygen is produced in the lumen of the phagolysosome by NADPH oxidase activity, supported by altered mitochondrial metabolism (Rayamajhee et al., 2022). To further inhibit the ability of bacteria to avoid the phagolysosome, additional transporters are housed in the phagolysosomal membrane. NRAMPs transport sequestered divalent ions (Mn2+, Fe2+, Zn2+, and Cu2+) outside the phagolysosomes, thereby limiting the ability of engulfed bacterial to use metalloenzymes required to escape the phagolysosome (Siddiqui et al., 2019).
Overall, mammalian macrophages and Acanthamoeba display striking similarities in the molecular mechanisms involved in directional motility, recognition, binding, engulfment, and phagolysosome processing of bacteria.
To get a better idea of the characteristics of the last unicellular common ancestor of animals, we can compare the genomic information of primitive multicellular animals with their unicellular relatives forming the basal clades in a phylogenetic tree of Holozoa, such as Filasterea, Ichthyosporea, and Choanoflagellata (Ros-Rocher et al., 2021). We can assume that the genes shared by these groups were already present in the unicellular ancestors of modern multicellular animals. Therefore, we can expect that the last unicellular ancestor of animals already possessed a wide repertoire of genes required for multicellularity, such as molecules for intercellular adhesion, communication, and interaction with the ECM (King, 2004). We speculate that many of these genes are analogous to those characteristically used by macrophages to carry out similar functions.
Given that most of the characteristic features of macrophages observed in Acanthamoeba are associated with bacterial recognition, endocytosis, and elimination, we hypothesize that these abilities later evolved into a protective bactericidal function as part of the host immune response in multicellular animals (Hartenstein and Martinez, 2019). This suggests that the evolutionary origin of the bactericidal function of mammalian macrophages arose prior to the branching of Amoebozoa and Opisthokonta, most likely in the environment of a free-living unicellular amoeboid cell.
Moreover, this implies that the features underlying the function of M1 bactericidal macrophages represent an ancestral macrophage phenotype and that M2-like macrophage features arose later in evolution, potentially coinciding with the emergence of multicellularity, as discussed in the following paragraphs.
## Macrophage homeostatic features arose along with multicellularity
Based on comparisons of macrophages with free-living predatory Acanthamoeba and basal unicellular relatives of metazoans, we hypothesize that many specific features of macrophages associated with their bactericidal function derive from unicellular animal ancestors. However, Acanthamoeba does not possess analogous homeostatic, regulatory, and metabolic functions as mammalian macrophages. We, therefore, explore the possibility that the functional repertoire of macrophages has expanded substantially with the evolution of multicellularity.
We explore the analogy between the features observed in mammalian macrophages and D. discoideum, a close relative of Acanthamoeba, used as a model organism to investigate facultative multicellularity (Bozzaro, 2013). Dictyostelium possesses a complex life cycle. Typically, Dictyostelium resides in the vegetative state of free-living haploid amoebae that divide periodically by mitosis and prey on microbes for nutrition. When food becomes scarce, starving vegetative amoebae enter a social life form, or a sexual cycle. During the social cycle, the amoebae aggregate to form a multicellular pseudoplasmodium (also known as a slug). The slug conforms to all the parameters of a multicellular organism. The originally amoeboid vegetative cells differentiate into four distinct cell types that coordinate their behavior and give rise to a fruiting body that produces resistant spores (Flowers et al., 2010).
In terms of their biology, the vegetative cells of Dictyostelium resemble the trophozoids of Acanthamoeba. Therefore, it is not surprising that, like Acanthamoeba, vegetative cells of Dictyostelium also share many features with mammalian pro-inflammatory macrophages (Bozzaro and Eichinger, 2011). Observations from Dictyostelium vegetative cells conveniently complement our previous statements, discussed in the following paragraph. Interestingly, despite the vegetative cells of Dictyostelium being freely motile, we can observe a certain degree of sociality. The behavior of these vegetative amoebae is coordinated by mutual communication of soluble signaling factors, which could provide the basis for the later emergence of cytokine signaling in macrophages. In fact, similar communication has been observed in Acanthamoeba (Golé et al., 2011).
Vegetative cells of Dictyostelium perceive signals from their environment and localize bacteria as a nutrient source through a gradient of their secondary metabolites, such as folic acid, retinoic acid, lipopolysaccharides, and lysophosphatidic acid (Iglesias, 2012). The perception of these chemotactic signals is mediated by GPCRs (e.g., folic acid receptor, retinoic acid receptor) that trigger strong chemotaxis and foraging behavior (Iglesias, 2012). Recently, it was shown that vegetative Dictyostelium cells are also attracted to signaling factors of a non-biological nature. Exposure of vegetative cells to a gradient of Mg2+, Zn2+, or hydrogen peroxide induces high chemotactic motility (Consalvo et al., 2019). Most factors that activate vegetative amoebae of Dictyostelium also have a strong activating and chemotactic effect on mammalian macrophages (Cammer and Cox, 2014). This is consistent with the observation that the vast majority of receptors carried by vegetative cells of Dictyostelium are retained in mammalian macrophages. Indeed, exposure of macrophages to the chemoattractants mentioned above leads to increased macrophage motility (Xu et al., 2021).
Dictyostelium is equipped with a wide spectrum of receptors that recognize pathogens and other cells to be engulfed, which are classified as (PRRs). These surface receptors show substantial homology to many mammalian PRRs, such as scavenger receptors (LIMP-2), toll-like receptors (tirA, tirB), leucine-rich repeats receptors (LrrA), and C-type lectin receptors. Activation of these receptors triggers intracellular signaling cascades initiating phagocytosis, phagosome maturation and bacterial killing, and stress-related cascades and detoxification response (Dunn et al., 2018).
The process of F-actin remodeling and phagolysosome formation starts with the activation of one of the GPCRs. For example, activation of the folate receptor or the homolog of the toll-like receptor tirA leads to activation of conserved RAS-PI3K and ERK-MAPK signaling, resulting in induction of actin polymerization, increased motility and phagocytosis (Chen et al., 2007). Actin nucleation and branching are mediated by actin remodeling complexes consisting of WASp Arp$\frac{2}{3}$ and SCAR/WAVE proteins (Vogel et al., 1980). The mechanism described above in Dictyostelium resembles that observed in mammalian macrophages, in which activation of surface toll-like receptors (TLRs) or Fc receptors analogously initiates increased motility, phagocytosis, and production of pro-inflammatory factors (Schmitz et al., 2004). The detailed mechanism of phagolysosome maturation in *Dictyostelium is* now well described (Cosson and Lima, 2014). Interestingly, this mechanism is principally homologous to that of mammalian macrophages. Internalized bacteria are eliminated in the phagolysosome by the sequestration of divalent ions by the activity of NRAMT transporters and by ROS production by the mitochondrial NADH-dependent oxidase NOX2 (Lardy et al., 2005). Maturation of the phagolysosome containing indigestible bacterial remnants leads to their exocytosis and neutralization of the phagolysosome. Alternatively, ingested bacterial remnants are processed by autophagy, which is particularly important during starvation and infection by intracellular pathogens (Mesquita et al., 2017). The remarkable analogy of these processes between Dictyostelium and the mammalian macrophage rule out the possibility of convergent evolution and further supports the adoption of features characteristic of bactericidal macrophages already in our unicellular ancestors. *In* general, many features of vegetative amoebas of Dictyostelium resemble those observed in M1 mammalian macrophages.
In certain situations, vegetative amoeboid cells can switch from unicellular to multicellular life. Amoeboid vegetative cells constantly coordinate cell growth and division through signals that inform each other about their density and nutrient availability (Loomis, 2014). Nutritionally supplied cells continuously produce prestarvation factor (PSF), which inhibits cell behavior leading to aggregation. When PSF production decreases due to nutrient deficiency, cells begin to produce conditioned medium factor (CMF), which triggers the release of a pulse of cAMP. The cAMP signal is further amplified by surrounding cells, creating a concentration gradient that allows aggregation (Clarke and Gomer, 1995).
The cellular cascade that transduces the extracellular cAMP signal is of particular interest. Extracellular cAMP binds to the G-protein-coupled chemoattractant receptor cAR1, which serves as a docking receptor for β-Arrestin. This interaction triggers signaling through second messengers well known from mammalian cells, such as GSK3, ERK, Ras/GTP, and PI3K, and leads to activation of the effectors PKB, PKA, STAT, and TORC2, which drive an expression program controlled by the GATA family transcription factors (Loomis, 2014; Singer et al., 2019).
The transition from the unicellular to the multicellular life stage is associated with significant transcriptomic changes. These changes are achieved primarily through the propagation of repressive epigenetic modifications that functionally shape amoeboid cells to become more cooperative. ATAC-seq. analysis of vegetative cells undergoing transition revealed that the most significantly enhanced genes are classified as factors regulating ECM organization, cell adhesion, differentiation, and morphogenesis (Wang S. Y. et al., 2021). Recently, it has been shown that alternation of mitochondrial metabolism is a prerequisite for adopting tolerogenic cell behavior and multicellularity (Glöckner et al., 2016; Singer et al., 2019; Kelly et al., 2021). This process highly resembles cAMP tolerogenic behavior of mammalian myeloid cells required for macrophages to perform tissue homeostatic tasks. ( Sciaraffia et al., 2014).
When transitioning to the social phase of the life cycle, Dictyostelium cells inevitably encounter many problems common to multicellular animals, indicating an increased need for self-recognition and regulation. Previously, it has been described that the social life stage of *Dictyostelium is* associated with various types of cellular relationships, such as cheating and allocheating, but also altruism and self-sacrifice (Strassmann and Queller, 2011).
The multicellular body of the pseudoplasmodium consists of thousands of cells. Most of the cells in the body are destined to form future morphological structures of the sorocarp, such as stem cells, cup cells, and spores (Jang and Gomer, 2011). However, when tracing the evolution of macrophage-like features, a fourth subpopulation of sentinel cells deserves particular attention. Sentinel cells have protective, homeostatic, and regulatory functions and, therefore, resemble the primitive immune system of multicellular organisms. Sentinel cells are free-moving cells that phagocytose bacteria and toxins until they are eventually eliminated. Compared to other slug cells, sentinel cells show increased expression of the gene coding for Toll-interleukin receptor domain-containing protein (tirA), which is analogous to the mammalian toll-like receptors (Brock et al., 2016a).
Sentinel cells protect the snail from potentially pathogenic bacteria by releasing extracellular DNA traps and producing ROS to the external space (Zhang and Soldati, 2016). In case of infection by intracellular bacteria, sentinel cells cleanse the slug of infected cells, keeping the rest of the organism healthy and giving rise to uninfected spores (Farinholt et al., 2019). In addition to their protective role, sentinel cells exhibit a high degree of tolerogenic behavior and can discriminate between genetically related and unrelated cells in aggregation (Hirose et al., 2011). Thus, in the multicellular stage of life, only closely related cells are nourished by sentinel cells. Indeed, their tolerogenic internal predetermination is represented by the rather unexpected observation that the multicellular stage of Dictyostelium can maintain commensal bacteria, in a specific form of farming for nutritional symbiosis (Brock et al., 2013; Brock et al., 2016b). By these features, the sentinel cells of the slug resemble the functions of mammalian M2 macrophages.
Collectively, the features observed in Dictyostelium cells during the transition from the unicellular to the multicellular life stage may provide critical insight into how macrophage-like features emerged with multicellularity in animals.
To explore this idea, we took inspiration from a study that compared the genomes of multicellular animals and their unicellular relatives to identify the genes present in the last common multicellular ancestor of animals which expanded upon the emergence of multicellularity (Ros-Rocher et al., 2021). *Such* genes are mostly related to intercellular signaling, signal transduction, adhesion molecules, and regulators of the cytoskeleton. Furthermore, multicellular animals also show an increase in the repertoire of transcription factors and genes mediating epigenetic modifications, suggesting the need for temporal functional plasticity and restriction of specific traits to certain subpopulations of cells in the multicellular body (Hinman and Cary, 2017; Herron et al., 2018).
As mentioned, we may assume that the transition to multicellularity is conditioned by several adaptations on various levels of regulation, including epigenetic remodeling, transcriptional programming, metabolism, and cell behavior. The most significant changes are related to enhanced expression of adhesive molecules, signaling factors, enzymes involved in remodeling of ECM, and adoption of tolerogenic predetermination.
Comparison of macrophage-like properties in unicellular vegetative amoebae and sentinel cells in multicellular slugs reveals a functional shift of macrophage-like properties, from clearly pro-inflammatory and bactericidal, to protective but also tolerogenic and regulatory. Furthermore, we speculate that many of the functions that arose in unicellular amoebae to hunt microbes were functionally repurposed and served as a solid basis for the evolution of multicellularity. For an overview of the evolving hypotheses concerning the primary cell type in animals, see Box 2.BOX 2 macrophages in perspective of emerging multicellularityThe emergence of multicellular animals is a fascinating event in the evolution of metazoans. Formulation of the theory of common descent in the nineteenth century led many famous evolutionary and developmental biologists to seek a thorough explanation of what the hypothetical last common ancestor of all animals (the mysterious “Urmetazoan”) may have looked like (King, 2004). Among the most famous is Earnest Haeckel, whose theories suggested that the most ancestral animal cell was the amoeboid cell, which, under certain conditions, could have progressed to the colonial stage of life (Brunet and King, 2022). However, this theory was challenged by Elie Metchnikoff, who was convinced that the most ancestral animal cell was equipped with a flagellum, as is observed in basal groups of Holozoa, such as Porifera and Choanoflagellata (Brunet and King, 2022). However, Metchnikoff’s theory had major discrepancies, as it failed to explain the striking similarity between the amoeboid cells observed in animals and the unicellular Protista. Recently, this obstacle has been resolved by the discovery that Choanoflagellata are able to switch to amoeboid cells under certain circumstances (Brunet et al., 2021). In addition, it has been found that amoeboid cells can give rise to all other cell types in Porifera (Müller, 2006). This suggests that amoeboid cells represent the most ancestral cell type in metazoans, and that the amoeboid cell type has been retained and is present throughout the metazoan phylogenetic tree, rather than being evolutionarily discontinued (Brunet and King, 2017). According to the current generally accepted theory, the ancestor of animals was a facultative multicellular organism that alternated cell types between free-moving social amoebae and a multicellular stage in which amoebocytes differentiate into collar containing flagellated cells (Brunet and King, 2017). As the complexity of multicellular organisms increased, as did the need for molecules responsible for cell colony cohesion, signaling, cell differentiation, and maintenance of homeostasis (Brooke and Holland, 2003; Grau-Bové et al., 2017).
## Macrophage-like amoebocytes perform both M1 and M2 features within Porifera
In the previous section, we described that in the multicellular stage of the social amoeba D. discoideum, subpopulations of sentinel cells retain features of professional phagocytes, and play a protective, regulatory, and homeostatic role in the pseudoplasmodium. This raises the question of whether the presence of amoeboid cells fulfilling these tasks is essential for the functioning of multicellular organisms. Virtually every known multicellular animal has a highly motile professional phagocyte that performs protective, healing, regenerative, regulatory, and homeostatic functions in the organism (Brunet et al., 2021).
We can gain a comprehensive understanding of the range of functions that professional phagocytes can perform in a primitive multicellular organism by studying sponges (Porifera), which represent a phylum of basal multicellular organisms with incomplete tissues and organ systems (Feuda et al., 2017). Members of Porifera phylogenetically represent the most ancestral metazoans. They are primitive multicellular heterotrophic organisms and represent the sister group of multicellular animals. These aquatic creatures depend on filtering water from which they obtain nutrients. Although they lack distinct tissues and organs, such as nervous, digestive, or circulatory systems, they possess several cell types with specialized functions (Thacker et al., 2014).
The structure of the sponge body is relatively simple. The body is formed by a gel-like, amorphous matrix called the mesohyl, sandwiched between two layers of cells, the outer pinacoderm and the inner choanoderm. The mesohyl is composed of ECM components commonly found in other animals, such as collagen, dermatopontin, galectin, and fibronectin-like glycoproteins (Dahihande and Thakur, 2021). Most sponges live a sedentary lifestyle and filter nutrients from the water using specialized cells called choanocytes. Choanocytes are equipped with flagella, whose movement creates water flow, and cilia, which form a filtering collar to trap food particles. The food particles are internalized by the choanocytes by nutritive phagocytosis and processed in food vacuoles (Laundon et al., 2019). Nutrients must then be distributed throughout the body, from choanocytes to other cell types. This function is performed by archaeocytes, which receive nutrients from choanocytes and transport them, by virtue of their high motility, throughout the mesophyll to the nutritionally demanding cells (Hartenstein and Martinez, 2019).
As already mentioned, the protective role of macrophages originates from the wild unicellular ancestors of animals, in which it evolved as a nutritional phagocytosis of bacteria. Choanocytes and archaeocytes are professional phagocytic cells in Porifera. The identity of these cells is not completely fixed and both cells can undergo a change to the opposite cell type under certain conditions. As such, it is difficult to distinguish which of these 2 cell types represents the ancestor of macrophages in bilaterians (Nakanishi et al., 2014). Since archaeocytes are freely motile and play a protective role in sponges, they show functional similarities to macrophages of bilaterians, therefore, it is feasible that archaeocytes represent the ancestors of these cells. The mechanism of nutrient uptake by choanocyte-like cells and nutrient distribution by freely motile amoebocytes is highly conserved in animals, with the exception of vertebrates and insects (Hartenstein and Martinez, 2019).
Archaeocytes, also called amoebocytes, are macrophage-like cells dispersed in the mesophyll of the sponge. Archaeocytes are unique from other sponge cells because they retain a significant degree of totipotency and can give rise to any other cell type. An isolated suspension of archeocytes can regenerate the entire body of sponges, suggesting that they represent their ancestral cell type (Ereskovsky et al., 2021).
Archaeocytes were originally described by Ellie Metchnikoff in 1892 and denoted as macrophages of the sponge by Van de Vyver more than a century later (Muller, 2003). Sponges are exposed to a many potential pathogens and foreign particles from filtering the water and need an effective system for their elimination (Dzik, 2010). Archaeocytes play a central role in the protection of sponges from pathogens. Sequencing of the Porifera genome revealed that sponges exhibit a broad spectrum of pathogen recognition receptors that are homologous to the main PRR groups found in mammals, such as GPCRS, NOD-like receptors, cysteine-rich receptors, scavenger receptors, and receptors from the immunoglobulin superfamily (Wiens et al., 2005; Srivastava et al., 2010). A recent study also documented the presence of the TLR-mediated signaling cascade (Germer et al., 2017).
Upon recognition of PAMPs, archaeocytes activate the signal transduction pathway in which MyD88 acts as a second messenger and activates effector transcription factors known in mammalian immune response, such as IRAK, TRAFs, and NFĸB (Müller et al., 2009). Activation of these immune-related pathways induces the production of galectins, perforins, and ROS as molecules participating in the opsonization of the pathogen and its elimination (Wiens et al., 2005). Until now, 39 different lectins have been identified in the genomes of the Porifera phylum, including C-type lectins, tachylectin-like, F-type lectins, and galectins (Gardères et al., 2015). Thus, archaeocytes, after their activation by pathogens, exhibit features, and behavior with a high degree of homology to mammalian proinflammatory macrophages.
However, many situations require an advanced level of coordination and tolerant behavior of archaeocytes. In the following paragraphs, we will discuss the indispensable role of archaeocytes in immune tolerance, healing, regeneration, self-identification, and reproduction.
Archaeocytes display surprising tolerogenic potential, as commensal bacteria do not invoke bactericidal behavior. However, the tolerogenic mechanism has not yet been satisfactorily elucidated (Maldonado, 2016; Carrier et al., 2022). Archaeocytes are also indispensable for healing and tissue regeneration (Boury-Esnault, 1977). During healing, the wound is infiltrated by archaeocytes and damaged cells are cleared from the local environment. Archaeocytes then secrete components of ECM and differentiate into other cell types, giving rise to the regular structure of the body. At his point, the archaeocytes may also phagocytose the grey cells, which contain large amounts of glycogen and osmiophilic inclusions and thus serve as a nutrient reservoir (Fernàndez-Busquets et al., 2002).
Sponges possess the ability of whole body regeneration, either from a body fragment or by aggregation of dissociated cells. After the cells of the sponge body are dissociated to a cell suspension, the cells dedifferentiate to amoebocyte morphotypes, and the archaeocytes represent the most abundant cell type in the suspension. Subsequently, the cells aggregate, presumably due to pseudopodial activity, and differentiate into the appropriate cell types to form the body of the sponge (Buscema et al., 1980). Strikingly, if the bodies of two distinct sponges are dissociated into single-cell suspension, the cells sort in a species-specific manner and the two individuals are eventually reconstituted. Moreover, archaeocytes are sufficient to reconstitute functional sponges without any other cell type (Lavrov and Kosevich, 2014). Hence, they represent the totipotent stem cells of the organism. These observations undeniably demonstrate Porifera’s ability to recognize its own genetically related cells from others.
Transplantation studies have further contributed to the understanding of this phenomenon. Whether a graft is accepted or rejected depends on the phylogenetic distance between the recipient and the donor. It has been shown that a graft comprised of cell from the same species and strain fuses with the recipient and is eventually accepted. Transplantation of an allograft causes the formation of a barrier between the transplanted tissues or a cytotoxic reaction at the graft interface, leading to the separation of the allograft cells (Smith and Hindeman, 1986). A small subset of cell types are involved in allograft rejection. Archaeocytes and lophocytes, which are recruited from the mesoglea and migrate along the border of both tissues, either phagocytose healthy donor cells to separate the tissues or exhibit cytotoxic activity to destroy cells in contact (Gaino et al., 1999; Fernàndez-Busquets et al., 2002).
Overall, the presented information indicates that archaeocytes perform characteristic functions of M1 bactericidal and M2 tolerogenic macrophages within signal organisms according to the situational context. Particularly, archaeocyte display an exceptional level of totipotency and autonomy (Zhang et al., 2003; Müller, 2006). We may speculate that archaeocytes execute important regulatory tasks in the sponge body and thus functionally precedes the role of neural and endocrine system.
We observe that as the complexity of multicellular organisms increases, the repertoire of functions performed by macrophage-like amoebocytes increases. This can be attributed to the need for a higher degree of regulation and maintenance of homeostasis or to the fact that specialized cells (in this case choanocytes) have taken over the original nutritional function of amoebocytes, thus providing macrophages with the opportunity to acquire additional diverse functions.
## Macrophage-like cells in animals with specialized tissues display rich repertoire of functions
Considering macrophage functions have diversified in animal evolution with the increasing complexity of the body, it is important to pay close attention to the macrophage-like plasmatocytes in D. melanogaster, a simple animal with clearly defined tissues and organs.
Drosophila is a simple, genetically tractable model organism, often used to model human diseases. Over a century of genetic and molecular biological research has led to many fundamental discoveries and a knowledge base that is unparalleled by any other invertebrate model used for biological research (Jennings, 2011). Research on the innate immune system of Drosophila has provided one of the major breakthroughs in immunology, the discovery of the Toll receptor and downstream immunity-related signaling cascade (Lemaitre et al., 1996). Since then, Drosophila has become a widely used model organism for research on host-microbe interactions, immune signaling pathways, wound healing, phagocytosis, clearance of apoptotic and damaged cells, tissue repair, immuno-metabolism, etc. ( Razzell et al., 2011).
Most innate immune pathways known in mammals are highly conserved in Drosophila, including PRRs, second messengers, transcription factors, and effector molecules (Govind, 2008). Given that the conservation of immune pathways has been extensively described in many previous works, we mention them only briefly with emphasis on their evolutionary development and instead focus on immune-unrelated properties of plasmatocytes, such as their role in morphogenesis, regulation of metabolism, their tissue-specific roles, and their ability to phenotypically polarize.
Compared to basal clades of animals, the Drosophila immune system shows several significant advances. Firstly, the cellular branch of the immune system is represented by three distinct cell types with characteristic immunity-related functions. Crystal cells and lamellocytes are essential for the melanization reaction and the encapsulation of foreign objects that cannot be simply phagocytized, such as parasitoid eggs. Plasmatocytes are professional phagocytes that resemble macrophages in many of their properties (Gold and Brückner, 2015) with a high degree of molecular conservation of the underlying mechanisms (Melcarne et al., 2019).
The evolutionary novelty of the tunable immune response can be further documented by the variation of immune cascade activation following the recognition of different pathogens by PRRs on plasmatocytes. While fungi and Gram-positive bacteria elicit an immune response by activating the Toll receptor, Gram-negative bacteria predominantly activate the peptidoglycan receptor PGRP-LC and the downstream immune cascade IMD (De Gregorio, 2002). The components of the Toll and IMD immune cascades are highly conserved and show homology with downstream signaling from Toll-like receptors, NOD, GPCRs, and TNFR (Martinelli and Reichhart, 2005). Importantly, the diversity of immune-related signaling pathways enables the production of a cocktail of destructive effector molecules specifically tailored to the given pathogen, leading to an effective immune response while limiting immune-mediated damage to the host.
Moreover, the immune-related signaling pathways are accompanied by the production of various signaling factors that further influence other branches of the immune system and modify the function of other organs and tissues. Many of these factors can be denoted as true cytokines because their mammalian homologues are important regulators of the immune response, such as unpaired3 (IL-6) or eiger (TNFα) (Vanha-aho et al., 2016).
In addition to their protective functions, plasmatocytes also have many macrophage-like properties essential for tissue homeostasis. They are responsible for clearing apoptotic, senescent, and damaged cells through efferocytosis and express various genes required for the remodeling of the ECM (Preethi et al., 2020). These abilities predispose them to play important roles in fundamental processes of multicellular organisms, such as embryonic morphogenesis, tissue healing, and regeneration. Indeed, plasmatocytes are essential for the patterning and developmental morphogenesis of the ventral nerve cord, intestine, heart, and skeletal muscle (Yarnitzky and Volk, 1995; Olofsson and Page, 2005).
However, the participation of plasmatocytes in embryonic morphogenesis can be disrupted by the production of danger signals. In an experimental model of laser-induced injury in the Drosophila embryo, plasmatocytes are attracted to the site of the wound by oxygen peroxide produced by the injured cells. Plasmatocytes infiltrating the wounded tissue clear the damaged cells and provide ECM components and growth factors necessary for tissue regeneration (Wood et al., 2002). Interestingly, the underlying mechanism of wound healing that includes transcription factors, actin organization, cell infiltration, and morphogenesis appears to be conserved between Drosophila and mammals at the molecular level (Belacortu and Paricio, 2011).
Although it has been well documented that Drosophila plasmatocytes can perform both bactericidal and healing functions, the question whether plasmatocytes adopt functional and metabolic polarization has not yet been satisfactorily answered. Upon bacterial infection in adult flies, plasmatocytes enter a state that closely resembles the pro-inflammatory polarization of mammalian macrophages. Plasmatocytes stimulated by streptococcal infection exhibit increased transcriptional activity of hypoxia-inducible factor 1α (HIF1α), a master regulator of metabolic reprogramming in mammalian M1 macrophages. The transcriptional program directed by HIF1α is required for the infection-induced increase in glycolytic flux, glucose consumption, and accelerated conversion of pyruvate to lactate in Drosophila plasmatocytes. This metabolic reprogramming, which closely resembles aerobic glycolysis in mammalian M1 macrophages, is essential for increased bactericidal activity of plasmatocytes and resistance of flies to bacterial infection (Krejčová et al., 2019). Transcriptomic data obtained in an independent experimental system indicate that metabolic rearrangement of plasmatocytes may be a general prerequisite for the bactericidal function of these cells (Ramond et al., 2020).
Plasmatocyte polarization that resembles mammalian M2 macrophages has been observed in an experimental model of retinal tissue injury. In this scenario, plasmatocytes infiltrate the wound and promote tissue healing, presumably by increasing the expression of arginase, an enzyme promoting the conversion of arginine to ornithine necessary for tissue regeneration (Neves et al., 2016), which is the hallmark of M2 macrophages polarization in mammalian macrophages (Rath et al., 2014). Whether plasmatocytes adopt an M2-like phenotype during other situations, such as during efferocytosis or wound healing, remains to be investigated.
The idea of functional diversification of plasmatocytes in Drosophila also finds support in single-cell transcriptomic analysis of Drosophila immune cells. The available data suggest that despite the morphological uniformity, plasmatocytes consist of more than ten distinct subpopulations that differ markedly in their expression pattern and expression of characteristic markers (Cho et al., 2020; Tattikota et al., 2020). Albeit, functional confirmation of these observations is currently lacking. In depth analysis of single-cell data revealed that a particular population of plasmatocytes express a substantial number of genes related to lipid metabolism, lipid catabolism, and sphingolipid processing, indicating certain adipocyte features are also present in plasmatocytes (Tattikota et al., 2020).
Of the many situations in the fly life cycle, the most important metabolic role of plasmatocytes is, arguably, during metamorphosis. During the transition of the larva into the adult, the lymph gland is broken down, and plasmatocytes are released into the circulation (Kharrat et al., 2022). The vast majority of larval tissues undergo extensive histolysis, and adult tissues form de novo from imaginal discs. However, the energy accumulated during the larval life stage must be transferred to the adult (Merkey et al., 2011). Thus, cells that are no longer needed are removed during metamorphosis by plasmatocytes infiltrating the histolysis-undergoing tissues. Within a short period of time, thousands of cells must be efferocytosed and the building material recycled into a suitable, reusable form (Storelli et al., 2019). In this situation, cells predisposed to efficient processing of lipids and sphingolipids may be highly desirable.
Plasmatocytes not only serve as metabolically active cells per se, but also significantly regulate the metabolism of other tissues. During bacterial infection, plasmatocytes produce the factor called *Imaginal morphogenesis* protein-Late2 (ImpL2), which reduces insulin signaling in the fat body. In turn, the fat body produces lipoproteins and carbohydrates that replenish activated immune cells (Krejcova et al., 2019). Thus, plasmatocytes orchestrate metabolic homeostasis and nutrient redistribution during the stress response. Their role in regulating systemic metabolism has been further documented in flies fed a high-fat diet. During excessive energy intake, plasmatocytes exposed to excessive lipids secrete IMPL2, leading to increased circulating glucose levels. Therefore, suppression of ImpL2 in plasmatocytes improves metabolism in obese flies (Morgantini et al., 2019). The pro-inflammatory effect of lipids on plasmatocytes was confirmed in an independent study. Plasmatocytes exposed to excessive amounts of lipids engulf the lipids through the activity of the scavenger receptor croquemort, which is homologous to mammalian CD36. Lipid accumulation in the plasmatocyte cytosol leads to increased production of the cytokine unpaired3 (upd3) and systemic attenuation of insulin signaling via JAK/STAT signaling (Woodcock et al., 2015). Interestingly, upd3 production by plasmatocytes may have an adaptive significance in addition to its pathological role analogically to ImpL2. It has been shown experimentally that sustained production of UPD3 by plasmatocytes is required for the regular distribution of lipids between tissues in the body and that a missing UPD3 signal leads to lipid accumulation in muscles (Kierdorf et al., 2020).
It is unclear whether Drosophila has tissue macrophages as we know them in mammals. Functionally, the cells that most closely resemble the concept of tissue macrophages in Drosophila are cells that can be functionally considered as microglia. Microglia are the resident macrophages of the mammalian central nervous system (CNS) and are responsible for the immune protection of neurons and elimination of toxic and harmful substances, and for the maintenance, neuronal pruning, and proper functioning of synapses in CNS (Lee et al., 2021).
Glial cells in Drosophila, similar to their mammalian counterpart, form the brain-blood barrier and maintain homeostasis of the CNS of flies. Although no plasmatocytes reside in the Drosophila brain under physiological conditions, glial cells display molecular parallels regarding their phagocytic receptors six microns under (simu) and draper (drpr) (Kim et al., 2020). Glial cells expressing Simu and drpr are required for clearance of the impaired neurons and neuronal debris, and the lack of expression these receptors leads to neurodegeneration (Elliott and Ravichandran, 2008). Moreover, a microglia-like glial subtype called MANF (Mesencephalic Astrocyte Derived Neurotrophic Factor) immuno-reactive cells has been described in the Drosophila brain during metamorphosis under certain conditions. These cells are extremely rich in lysosomes and express drpr (Stratoulias and Heino, 2015). In addition, cortex glia and ensheathing cells are non-professional phagocytes engulfing apoptotic cells during the development of the nervous system and degenerating axons, respectively (Doherty et al., 2009; Kurant, 2011).
Since insects lack the adaptive immune system that evolved 500 million years ago in jawed fish, they must rely solely on innate immunity. Non-etheless, it has been described that the innate immune system can also be “trained”, and display certain memory traits. The phenomenon of “innate immune memory” was proposed by Netea et al, [2020], who conducted this research on mammalian models. This concept has also been addressed in Drosophila. It has been documented that fruit flies display enhanced survival of streptococcal infection if re-encountered by an otherwise lethal dose of the same bacteria and that this protective mechanism lies in the action of phagocytes and the Toll signaling pathway (Pham et al., 2007). However, such protection could not be invoked against all the bacteria examined.
As evidenced by advances made in recent years, plasmatocytes in Drosophila perform a strikingly wide range of roles that encompass the functional repertoire of M1 and M2 macrophages. In addition, experimental data demonstrate that plasmatocytes are capable of entering different polarization phenotypes over time. Moreover, several lines of evidence suggest that plasmatocytes are not a uniform population and consist of many distinct subpopulations of plasmatocyte phenotypes. Particularly, their ability to regulate the metabolism of other tissues via signaling factors might be of interest. In terms of the pathological role of mammalian macrophages, it is an interesting observation that exposure of plasmatocytes to excessive amounts of lipids may lead to macrophage polarization, reminiscent of Mox polarization in mammalian macrophages. Whether Drosophila possesses a functional analogy to mammalian tissue-resident macrophages remains to be determined.
## Macrophage functional versatility as a legacy of animal origin
In the previous paragraphs, we have discussed how the rising complexity of body plans corresponds with the adoption of crucial macrophage features (Figure 3). Nevertheless, the question of why macrophages are predisposed to exceptional functional versatility remains to be addressed.
**FIGURE 3:** *Mammalian macrophages share many features with Acanthamoeba, Dictyostelium, archaeocytes in Porifera, and plasmatocytes in Drosophila. While the bacteria hunting unicellular Acanthamoeba resembles the M1 polarization phenotype of mammalian macrophages, Dictyostelium exhibits M1 or M2 features of mammalian macrophages depending on its life stage. The archaeocytes of aquatic sponges have both M1 and M2 macrophages within a single organism, phenotypically responding to situation context rather than life stage. Plasmatocytes of Drosophila exhibit a wide range of highly specialized roles in the organism in addition to M1 and M2 polarization. Mϕs, macrophages.*
Recently, the theoretical concept of the origin of multicellular animals has been revisited. It is generally accepted that amoebocytes represent the most ancestral cell type of all Holozoa. Primitive facultative multicellular animals consisted of a cluster of a few cell types, temporally forming multicellular colonies (Ros-Rocher et al., 2021).
The observations from Dictyostelium and Porifera indicate that amoebocytes, the last common ancestor of multicellular animals, represent the ancestral super-ordinated cells that must have been widely distributed and capable of functionalities performed in more complex animals by specialized tissues and organs. They were most likely able to differentiate into all other cell types, control their number, and govern protective, nutritional, regulatory, and homeostatic functions. Therefore, we expect these cells to be already highly functionally versatile with a certain level of plasticity and autonomy.
Given that macrophage-like amoebocytes represent an archetypal cellular type in animals (Cavalier-Smith, 2017), the second cell type commonly diversified in early multicellular organisms are cells specialized for acquiring nutrients from the environment (Sogabe et al., 2019). We can hypothesize that as amoebocytes no longer needed to obtain nutrients for themselves, they evolved to perform other functions in the multicellular body. However, the differentiation of individual specialized cells imposed the requirement to evenly distribute resources, coordinate the function of individual cells, and maintain homeostasis in response to changing external biotic and abiotic factors (Bich et al., 2019). These requirements demand a certain level of regulation, and before the development of the circulatory, endocrine, and neuronal systems, the amoebocytes were predisposed to perform such functions (Dyakonova, 2022).
We believe that macrophage functional versatility may be a heritage of their origin in unicellular and early multicellular animals, where the universality of macrophage-like amoebocytes was essential for resistance to different types of environmental and biological stress. Over millions of years of evolution from amoebocytes to macrophages, macrophage-like cells have taken advantage of their initial versatility and gradually achieved their full functional repertoire along with the increasing complexity of the animal body. Although we might assume that the functional variability of macrophages would decrease with the emergence of organ systems, the opposite is true. Every tissue in the mammalian body contains a population of tissue-resident macrophages that often perform highly specialized functions (Mass et al., 2016). Collectively, maintaining functionally versatile amoeboid cells that can easily change their functional repertoire to suit emerging needs seems to be an adaptive strategy.
## The origin of macrophage functions may explain their pathological effect in mammals
*In* general, it can be assumed that the acquisition of new functions of macrophages in evolution can be achieved by changing their original archetypal role and adapting it to the current context (Brosius, 2019). This can be documented, for instance, by a functional shift from mechanisms evolved to hunt bacteria to an antibacterial protective role of macrophages. The mechanism required by unicellular amoeboid cells to identify, approach, phagocytose, and digest bacteria in the phagolysosome for nutritional reasons was later shown to be advantageous for macrophages in multicellular organisms for protection against pathogenic bacteria (Hartenstein and Martinez, 2019). Another such example is the rich repertoire of genes originally used in the unicellular ancestor of animals for amoeboid crawling and attachment to surface structures, which evolved into a broad repertoire of surface receptors and adhesion molecules used by macrophages in multicellular organisms for sensing surrounding tissues and motility (Hynes and Zhao, 2000).
Thus, many signaling pathways in macrophages and other myeloid cells may carry remnants of their evolutionary origin without retaining their initial adaptive function in a complex multicellular organism. Such vestigial molecular relationships may underlie the pathological behavior of macrophages. For example, the folate receptor, formyl-peptide receptor, or cAMP signaling represent the shift of adaptive functions originally developed in macrophage-like amoebocytes to their pathological effect in macrophages. We believe that many analogous comparisons can be found when applying this perspective to human pathologies.
Folate, a secondary metabolite of bacteria, is a potent chemoattractant for amoeboid vegetative cells of Dictyostelium (Driel, 1981). The folate gradient is perceived via a G-protein coupled folate receptor at nanomolar concentrations and leads to the activation of chemotaxis and machinery required for phagocytosis and bacterial processing in the phagolysosome (Pan et al., 2016). Interestingly, increased expression of folate receptors are a hallmark of pro-inflammatory mammalian macrophages (Steinz et al., 2022). In particular, folate receptor β (FR-β) has been identified as a specific surface receptor for highly pro-inflammatory macrophages, such as those found in the synovial tissue of arthritic patients, in atherosclerotic plaques, or in pulmonary fibrosis (Chandrupatla et al., 2019). Activation of macrophage FR-β leads to their pro-inflammatory polarization and production of cytokines that further perpetuate the chronic inflammatory state. Inhibition of the folate receptor has thus been recognized as a possible avenue for treating arthritis and atherosclerosis, making FR-β agonist Methotrexate the first-choice treatment for these diseases (Xia et al., 2009).
Similar functional dualism can be observed for other GPCRs carried by mammalian myeloid cells, such as the formyl peptide receptor (FPR) abundantly expressed by macrophages and neutrophils (Chen et al., 2017). Activation of FPR serves as a potent signal leading to enhanced directional motility, the production of ROS, the release of pro-inflammatory cytokines, and acceleration of phagocytic and bactericidal machinery (Liang et al., 2020). Since formyl peptides are released by bacteria as their secondary metabolite, the response mediated via the FPR receptor is important for resistance to bacterial pathogens (Dorward et al., 2015). However, under stress conditions, formyl peptides are released from the mitochondria of stressed and damaged tissues, leading to infiltration of the affected tissue by macrophages and neutrophils, which induce inflammation even under sterile conditions (Wenceslau et al., 2013). Therefore, excessive activation of FPR on macrophages and neutrophils underpins the progressive development of many human inflammatory diseases, such as neurodegeneration, cardiovascular diseases, and pulmonary fibrosis (Trojan et al., 2020; Caso et al., 2021).
In both cases, inadequate activation of receptors, initially designed to detect bacterial secondary metabolites and tracking bacteria in the environment, causes pathology in a complex multicellular organism, where their activation can occur even under sterile conditions (Lu et al., 2021).
Non-etheless, the repurposing of ancestral signaling is not limited to bacterial detection and localization mechanisms. As described previously, metabolically stressed vegetative amoebae of Dictyostelium produce cAMP as a potent aggregation chemoattractant (Singer et al., 2019). Sensing of extracellular cAMP leads to the activation of stress-related cellular pathways, remodeling of cellular metabolism, and epigenetic remodeling, resulting in a transition to multicellularity, increased production of ECM components, and tolerogencity (Wang S. Y. et al., 2021). Interestingly, many lines of evidence suggest that extracellular cAMP (ex-cAMP) strongly effects the recruitment and reprograming of monocytes and macrophages and induces efferocytosis of damaged or exhausted surrounding cells (Negreiros-Lima et al., 2020). Exposure of monocytes to ex-cAMP enhances the production of cytokines with known anti-inflammatory effects, such as IL-6 and IL-10, and ameliorates response to pro-inflammatory stimuli (Sciaraffia et al., 2014).
cAMP in the extracellular space is cleaved by ectonucleotidases to extracellular adenosine and sensed by the adenosine receptor abundantly expressed by macrophages and other myeloid cells (Haskó and Pacher, 2012). Adenosine and cAMP are released from damaged, hypoxic, and metabolically stressed tissues. Activation of adenosine receptors causes potent anti-inflammatory effects and plays an essential role in tissue regeneration and maintenance of tissue homeostasis (Pasquini et al., 2021). Interestingly, cAMP is an important secondary messenger in mammalian immune cells that activates identical downstream cascades in Dictyostelium amoebocytes, leading to the inhibition of NFĸB and activation of anti-inflammatory tolerogenic polarization (Tavares et al., 2020).
Thus, we can assume that cAMP signaling, which appeared in evolution at the origin of multicellular animals, may play an adaptive role in the immune system up to the present day. However, adopting a tolerogenic program through the activation of adenosine and cAMP signaling also has a role in pathology. Increased adenosine and cAMP production by metabolically demanding and often hypoxic neoplastic tumors leads to the induction of tolerogenic polarization in surrounding immune cells (Strakhova et al., 2020). Hence, tumor-associated macrophages often promote tumor growth, instead of elimination, by providing nutrients and growth factors and promoting vascularization (Moeini and Niedźwiedzka-Rystwej, 2021).
Thus, the repurposing of the features of the macrophage ancestors may be adaptive, as evidenced by protection against pathogenic bacteria, but may contribute to the development of many pathologies.
## Discussion
The hypotheses we present here are speculative, convincing evidence that documents events that took place in the distant past in evolution is limited. However, this may change significantly with the growing list of organisms with fully sequenced genomes and well-annotated transcriptomes. Many of these newly sequenced species provide information allowing speculation regarding the nature of the last unicellular and first multicellular ancestors of animals. These data provide evidence of genes that were prerequisites for the emergence of multicellularity and the development of advanced multicellular body structures. An interesting example of such an approach can be found in the work of Ros-Rocher and colleagues, and it is feasible that analogous analyses can yield valuable information in the future (Ros-Rocher et al., 2021). Regarding the origin of macrophage functional versatility, the effort requires tracing macrophage characteristic features in evolution. Recent work conducted by Nagahata and colleagues, which, on the genetic level, supports the hypotheses that many functions typical of bactericidal macrophages evolved from a common ancestor of animals, and that many characteristic macrophage features are adaptations of free-living unicellular bacterivorous amoebae (Nagahata et al., 2022; Rayamajhee et al., 2022).
The majority of macrophage-like features that are observed in amoebae resemble those of bactericidal (M1) macrophages. This indicates that the bactericidal macrophage polarization represents an ancestral polarization type and the protective function of macrophages evolved from hunting microbes for nutritional reasons (Desjardins et al., 2005; Hartenstein and Martinez, 2019).
Inspired by the currently revised theory of the origin of animal multicellularity (Brunet and King, 2017), we believe that amoebocytes, as the ancestral type of animal cells, play a central role in the origin of multicellularity. Amoebocytes display several features that may be considered prerequisites for the emergence of multicellularity, such as the ability to deposit and remodel ECM, remove senescent and damaged cells, respond to various signals, regulate the function of other cells by signaling factors, and recognize genetically related cells in the colony (Misevic, 1999). These features are required by multicellular organisms and also resemble the characteristics of healing (M2) macrophages. Therefore, we suggest that along with the emergence of multicellular animals, the macrophage-like amoeboid cells acquired macrophage-like properties characteristic of M2 macrophages.
Given that the last common ancestor of animals likely switched between free-living and colonial life stages during its life cycle in response to extrinsic cues (Brunet et al., 2021), it is possible that the macrophage-like amoebocytes had the capacity for phenotypic polarization in a context-dependent manner, before the emergence of multicellularity. While the wild-type social amoeba shows features observed predominantly in M1 macrophages, amoebocytes participating in colonial life stages have changed their biology and acquired features characteristic predominantly of M2 macrophages.
We can hypothesize that the divergence of macrophage functions in emerging multicellular organisms was driven by the differentiation of specialized cell types for obtaining nutrition, the growing need for harmonizing force, and the recognition of cellular identity. The diversification of the functional repertoire of macrophages in primitive multicellular organisms (Porifera) suggests that, together with the increasing complexity of body plans, there is a need for polarization of macrophage-like archaeocytes into both phenotypes within a single organism, with the polarization phenotype depending on context rather than life stage (Degnan et al., 2015). We may hypothesize that the regulatory role of macrophage-like amoebocytes precedes the function of the endocrine and nervous systems in primitive multicellular organisms, indicating a superior regulatory role of amoebocytes over other cells in the body.
Upon the evolution of complex multicellular organisms, macrophages expanded their functions, reaching their full potential, participating in development, organogenesis, immune protection, self-recognition, tissue and metabolic homeostasis maintenance, and many tissue- and context-specific tasks (Mase et al., 2021). It is likely that these functions will be revealed in future research in virtually all complex multicellular animals.
We present a perspective of evolutionary biology, combined with knowledge from modern biomedical research. Both approaches can be mutually inspiring in future research on the biology of macrophage-like cells. One critical feature which contributes to the functional versatility of macrophages is their ability to adopt distinct metabolic polarization phenotypes, which are determined by epigenetic modifications and activation of specific signaling cascades. Therefore, investigating polarization phenotypes of unicellular and facultative multicellular relatives of true animals would be of interest. A seminal study addressing this was performed on Dictyostelium, carried out in the laboratory of Erika Pearce, one of the leading scientists working on macrophage immuno-metabolism (Kelly et al., 2021). This work demonstrated a link between cell metabolism and the transition to the multicellular stage. In addition, the traits of early relatives of animals can be explored for their significance to mammalian macrophage biology. One such example is the potentially conserved nutritional role of macrophage-like amoebocytes in Dictyostelium and sponges in the macrophages of higher animals and humans.
In addition, we believe that understanding the role of ancestral macrophage-like cells may help to understand the biology of mammalian macrophages and possibly discover new functions. Given the ancestral origin of macrophage functions, some difficult-to-understand pathological behaviors of macrophages can be explained by the activation of ancient vestigial functions that may appear counterintuitive in a specific context in a complex multicellular body. We believe that this perspective may shed new light on the function and pathogenesis of macrophages in animals and humans.
## Data availability statement
The original contributions presented in the study are included in the article/supplementary material, further inquiries can be directed to the corresponding authors.
## Author contributions
AB and GK participated in the conceptualization of the manuscript, preparation of the figures, and writing of the text and its graphic design.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## References
1. Akya A., Pointon A., Thomas C.. **Mechanism involved in phagocytosis and killing of Listeria monocytogenes by Acanthamoeba polyphaga**. *Parasitol. Res.* (2009) **105** 1375-1383. DOI: 10.1007/s00436-009-1565-z
2. Allen P. G., Dawidowicz E. A.. **Phagocytosis inAcanthamoeba: I. A mannose receptor is responsible for the binding and phagocytosis of yeast**. *J. Cell. Physiol.* (1990) **145** 508-513. DOI: 10.1002/jcp.1041450317
3. Alsam S., Sissons J., Dudley R., Khan N. A.. **Mechanisms associated with Acanthamoeba castellanii (T4) phagocytosis**. *Parasitol. Res.* (2005) **96** 402-409. DOI: 10.1007/s00436-005-1401-z
4. Amarante-Mendes G. P., Adjemian S., Branco L. M., Zanetti L. C., Weinlich R., Bortoluci K. R.. **Pattern recognition receptors and the host cell death molecular machinery**. *Front. Immunol.* (2018) **9** 2379. DOI: 10.3389/fimmu.2018.02379
5. Arango Duque G., Descoteaux A.. **Macrophage cytokines: Involvement in immunity and infectious diseases**. *Front. Immunol.* (2014) **5** 491. DOI: 10.3389/fimmu.2014.00491
6. Barreby E., Chen P., Aouadi M.. **Macrophage functional diversity in NAFLD — More than inflammation**. *Nat. Rev. Endocrinol.* (2022) **18** 461-472. DOI: 10.1038/s41574-022-00675-6
7. Batista-Gonzalez A., Vidal R., Criollo A., Carreño L. J.. **New insights on the role of lipid metabolism in the metabolic reprogramming of macrophages**. *Front. Immunol.* (2020) **10**. DOI: 10.3389/fimmu.2019.02993
8. Belacortu Y., Paricio N.. **Drosophila as a model of wound healing and tissue regeneration in vertebrates**. *Dev. Dyn.* (2011) **240** 2379-2404. DOI: 10.1002/dvdy.22753
9. Bich L., Pradeu T., Moreau J.-F.. **Understanding multicellularity: The functional organization of the intercellular space**. *Front. Physiol.* (2019) **10** 1170. DOI: 10.3389/fphys.2019.01170
10. Biswas S. K., Mantovani A.. **Orchestration of metabolism by macrophages**. *Cell Metab.* (2012) **15** 432-437. DOI: 10.1016/j.cmet.2011.11.013
11. Boury-Esnault N.. **A cell type in sponges involved in the metabolism of glycogen. The gray cells**. *Cell Tissue Res.* (1977) **175** 523-539. DOI: 10.1007/BF00222416
12. Bowers B.. **Comparison of pinocytosis and phagocytosis in Acanthamoeba castellanii**. *Exp. Cell Res.* (1977) **110** 409-417. DOI: 10.1016/0014-4827(77)90307-x
13. Bozzaro S., Eichinger L.. **The professional phagocyte**. *Curr. Drug Targets* (2011) **12** 942-954. DOI: 10.2174/138945011795677782
14. Bozzaro S.. **The model organism**. *Methods Mol. Biol.* (2013) **983** 17-37. DOI: 10.1007/978-1-62703-302-2_2
15. Brock D. A., Callison W. É., Strassmann J. E., Queller D. C.. **Sentinel cells, symbiotic bacteria and toxin resistance in the social amoeba**. *Proc. R. Soc. B Biol. Sci.* (2016a) **283** 20152727. DOI: 10.1098/rspb.2015.2727
16. Brock D. A., Jones K., Queller D. C., Strassmann J. E.. **Which phenotypic traits of**. *Symbiosis* (2016b) **68** 39-48. DOI: 10.1007/s13199-015-0352-0
17. Brock D. A., Read S., Bozhchenko A., Queller D. C., Strassmann J. E.. **Social amoeba farmers carry defensive symbionts to protect and privatize their crops**. *Nat. Commun.* (2013) **4** 2385. DOI: 10.1038/ncomms3385
18. Brooke N. M., Holland P. W.. **The evolution of multicellularity and early animal genomes**. *Curr. Opin. Genet. Dev.* (2003) **13** 599-603. DOI: 10.1016/j.gde.2003.09.002
19. Brosius J.. **Exaptation at the molecular genetic level**. *Sci. China Life Sci.* (2019) **62** 437-452. DOI: 10.1007/s11427-018-9447-8
20. Brunet T., Albert M., Roman W., Coyle M. C., Spitzer D. C., King N.. **A flagellate-to-amoeboid switch in the closest living relatives of animals**. *Elife* (2021) **10** e61037. DOI: 10.7554/eLife.61037
21. Brunet T., King N.. **The origin of animal multicellularity and cell differentiation**. *Dev. Cell* (2017) **43** 124-140. DOI: 10.1016/j.devcel.2017.09.016
22. Brunet T., King N.. **The single-celled ancestors of animals**. *The evolution of multicellularity* (2022) 251-278
23. Buscema M., Sutter D., Vyver G.. **Ultrastructural study of differentiation processes during aggregation of purified sponge archaeocytes**. *Wilhelm Roux’s Arch. Dev. Biol.* (1980) **188** 45-53. DOI: 10.1007/BF00848609
24. Cammer M., Cox D.. **Chemotactic responses by macrophages to a directional source of a cytokine delivered by a micropipette**. *Methods Mol. Biol.* (2014) **1172** 125-135. DOI: 10.1007/978-1-4939-0928-5_11
25. Campolo A., Harris V., Walters R., Miller E., Patterson B., Crary M.. **Continuous real-time motility analysis of Acanthamoeba reveals sustained movement in absence of nutrients**. *Pathogens* (2021) **10** 995. DOI: 10.3390/pathogens10080995
26. Carrier T. J., Maldonado M., Schmittmann L., Pita L., Bosch T. C. G., Hentschel U.. **Symbiont transmission in marine sponges: Reproduction, development, and metamorphosis**. *BMC Biol.* (2022) **20** 100. DOI: 10.1186/s12915-022-01291-6
27. Caso V. M., Manzo V., Pecchillo Cimmino T., Conti V., Caso P., Esposito G.. **Regulation of inflammation and oxidative stress by formyl peptide receptors in cardiovascular disease progression**. *Life* (2021) **11** 243. DOI: 10.3390/life11030243
28. Castillo‐Armengol J., Fajas L., Lopez‐Mejia I. C.. **Inter‐organ communication: A gatekeeper for metabolic health**. *EMBO Rep.* (2019) **20** e47903. DOI: 10.15252/embr.201947903
29. Cavalier-Smith T.. **Origin of animal multicellularity: Precursors, causes, consequences—the choanoflagellate/sponge transition, neurogenesis and the cambrian explosion**. *Philos. Trans. R. Soc. B Biol. Sci.* (2017) **372** 20150476. DOI: 10.1098/rstb.2015.0476
30. Chandrupatla D. M. S. H., Molthoff C. F. M., Lammertsma A. A., van der Laken C. J., Jansen G.. **The folate receptor β as a macrophage-mediated imaging and therapeutic target in rheumatoid arthritis**. *Drug Deliv. Transl. Res.* (2019) **9** 366-378. DOI: 10.1007/s13346-018-0589-2
31. Chen G., Zhuchenko O., Kuspa A.. **Immune-like phagocyte activity in the social amoeba**. *Science* (2007) **317** 678-681. DOI: 10.1126/science.1143991
32. Chen K., Bao Z., Gong W., Tang P., Yoshimura T., Wang J. M.. **Regulation of inflammation by members of the formyl-peptide receptor family**. *J. Autoimmun.* (2017) **85** 64-77. DOI: 10.1016/j.jaut.2017.06.012
33. Cheng L., Baonza A., Grifoni D.. **Drosophila models of human disease**. *Biomed. Res. Int.* (2018) **1–2** 7214974. DOI: 10.1155/2018/7214974
34. Cho B., Yoon S.-H., Lee D., Koranteng F., Tattikota S. G., Cha N.. **Single-cell transcriptome maps of myeloid blood cell lineages in Drosophila**. *Nat. Commun.* (2020) **11** 4483. DOI: 10.1038/s41467-020-18135-y
35. Clarke M., Gomer R. H.. **PSF and CMF, autocrine factors that regulate gene expression during growth and early development ofDictyostelium**. *Experientia* (1995) **51** 1124-1134. DOI: 10.1007/BF01944730
36. Coats B. R., Schoenfelt K. Q., Barbosa-Lorenzi V. C., Peris E., Cui C., Hoffman A.. **Metabolically activated adipose tissue macrophages perform detrimental and beneficial functions during diet-induced obesity**. *Cell Rep.* (2017) **20** 3149-3161. DOI: 10.1016/j.celrep.2017.08.096
37. Cole J., Aberdein J., Jubrail J., Dockrell D. H.. **The role of macrophages in the innate immune response to Streptococcus pneumoniae and**. *Adv. Microb. Physiol.* (2014) **65** 125-202. DOI: 10.1016/bs.ampbs.2014.08.004
38. Consalvo K. M., Rijal R., Tang Y., Kirolos S. A., Smith M. R., Gomer R. H.. **Extracellular signaling in Dictyostelium**. *Int. J. Dev. Biol.* (2019) **63** 395-405. DOI: 10.1387/ijdb.190259rg
39. Cosson P., Lima W. C.. **Intracellular killing of bacteria: Is D**. *Cell. Microbiol.* (2014) **16** 816-823. DOI: 10.1111/cmi.12291
40. Dahihande A. S., Thakur N. L.. **Differences in the structural components influence the pumping capacity of marine sponges**. *Front. Mar. Sci.* (2021) **8**. DOI: 10.3389/fmars.2021.671362
41. Davies L. C., Jenkins S. J., Allen J. E., Taylor P. R.. **Tissue-resident macrophages**. *Nat. Immunol.* (2013) **14** 986-995. DOI: 10.1038/ni.2705
42. De Gregorio E., Spellman P. T., Tzou P., Rubin G. M., Lemaitre B.. **The Toll and Imd pathways are the major regulators of the immune response in Drosophila**. *EMBO J.* (2002) **21** 2568-2579. DOI: 10.1093/emboj/21.11.2568
43. deCathelineau A. M., Henson P. M.. **The final step in programmed cell death: Phagocytes carry apoptotic cells to the grave**. *Essays Biochem.* (2003) **39** 105-117. DOI: 10.1042/bse0390105
44. Degnan B. M., Adamska M., Richards G. S., Larroux C., Leininger S., Bergum B.. **Porifera**. *Evolutionary developmental biology of invertebrates* (2015)
45. Desjardins M., Houde M., Gagnon E.. **Phagocytosis: The convoluted way from nutrition to adaptive immunity**. *Immunol. Rev.* (2005) **207** 158-165. DOI: 10.1111/j.0105-2896.2005.00319.x
46. Doherty J., Logan M. A., Tasdemir O. E., Freeman M. R.. **Ensheathing glia function as phagocytes in the adult Drosophila brain**. *J. Neurosci.* (2009) **29** 4768-4781. DOI: 10.1523/JNEUROSCI.5951-08.2009
47. Dorward D. A., Lucas C. D., Chapman G. B., Haslett C., Dhaliwal K., Rossi A. G.. **The role of formylated peptides and formyl peptide receptor 1 in governing neutrophil function during acute inflammation**. *Am. J. Pathol.* (2015) **185** 1172-1184. DOI: 10.1016/j.ajpath.2015.01.020
48. Driel R.. **Binding of the chemoattractant folic acid by**. *Eur. J. Biochem.* (1981) **115** 391-395. DOI: 10.1111/j.1432-1033.1981.tb05250.x
49. Dunn J. D., Bosmani C., Barisch C., Raykov L., Lefrançois L. H., Cardenal-Muñoz E.. **Eat prey, live:**. *Front. Immunol.* (2018) **8** 1906. DOI: 10.3389/fimmu.2017.01906
50. Dyakonova V. E.. **Origin and evolution of the nervous system: New data from comparative whole genome studies of multicellular animals**. *Russ. J. Dev. Biol.* (2022) **53** 55-64. DOI: 10.1134/s1062360422010088
51. Dzik J. M.. **The ancestry and cumulative evolution of immune reactions**. *Acta Biochim. Pol.* (2010) **57** 443-466. DOI: 10.18388/abp.2010_2431
52. Elliott M. R., Ravichandran K. S.. **Death in the CNS: Six-Microns-Under**. *Cell* (2008) **133** 393-395. DOI: 10.1016/j.cell.2008.04.014
53. Ereskovsky A., Borisenko I. E., Bolshakov F. V., Lavrov A. I.. **Whole-body regeneration in sponges: Diversity, fine mechanisms, and future prospects**. *Genes (Basel).* (2021) **12** 506. DOI: 10.3390/genes12040506
54. Farinholt T., Dinh C., Kuspa A.. **Microbiome management in the social amoeba**. *Int. J. Dev. Biol.* (2019) **63** 447-450. DOI: 10.1387/ijdb.190240ak
55. Fernàndez-Busquets X., Kuhns W. J., Simpson T. L., Ho M., Gerosa D., Grob M.. **Cell adhesion-related proteins as specific markers of sponge cell types involved in allogeneic recognition**. *Dev. Comp. Immunol.* (2002) **26** 313-323. DOI: 10.1016/s0145-305x(01)00079-9
56. Ferrante C. J., Leibovich S. J.. **Regulation of macrophage polarization and wound healing**. *Adv. Wound Care* (2012) **1** 10-16. DOI: 10.1089/wound.2011.0307
57. Feuda R., Dohrmann M., Pett W., Philippe H., Rota-Stabelli O., Lartillot N.. **Improved modeling of compositional heterogeneity supports sponges as sister to all other animals**. *Curr. Biol.* (2017) **27** 3864-3870.e4. DOI: 10.1016/j.cub.2017.11.008
58. Flowers J. M., Li S. I., Stathos A., Saxer G., Ostrowski E. A., Queller D. C.. **Variation, sex, and social cooperation: Molecular population genetics of the social amoeba**. *PLoS Genet.* (2010) **6** e1001013. DOI: 10.1371/journal.pgen.1001013
59. Gaino E., Bavestrello G., Magnino G.. **Self/non‐self recognition in sponges**. *Ital. J. Zool.* (1999) **66** 299-315. DOI: 10.1080/11250009909356270
60. Galván-Peña S., O'Neill L. A. J.. **Metabolic reprograming in macrophage polarization**. *Front. Immunol.* (2014) **5**. DOI: 10.3389/fimmu.2014.00420
61. Gardères J., Bourguet-Kondracki M.-L., Hamer B., Batel R., Schröder H., Müller W.. **Porifera lectins: Diversity, physiological roles and biotechnological potential**. *Mar. Drugs* (2015) **13** 5059-5101. DOI: 10.3390/md13085059
62. Germer J., Cerveau N., Jackson D. J.. **The holo-transcriptome of a calcified early branching metazoan**. *Front. Mar. Sci.* (2017) **4** 81. DOI: 10.3389/fmars.2017.00081
63. Glöckner G., Lawal H. M., Felder M., Singh R., Singer G., Weijer C. J.. **The multicellularity genes of dictyostelid social amoebas**. *Nat. Commun.* (2016) **7** 12085. DOI: 10.1038/ncomms12085
64. Gold K. S., Brückner K.. **Macrophages and cellular immunity in**. *Semin. Immunol.* (2015) **27** 357-368. DOI: 10.1016/j.smim.2016.03.010
65. Golé L., Rivière C., Hayakawa Y., Rieu J.-P.. **A quorum-sensing factor in vegetative Dictyostelium discoideum cells revealed by quantitative migration analysis**. *PLoS One* (2011) **6** e26901. DOI: 10.1371/journal.pone.0026901
66. Gordon S., Plüddemann A.. **Tissue macrophages: Heterogeneity and functions**. *BMC Biol.* (2017) **15** 53. DOI: 10.1186/s12915-017-0392-4
67. Govind S.. **Innate immunity in Drosophila: Pathogens and pathways**. *Insect Sci.* (2008) **15** 29-43. DOI: 10.1111/j.1744-7917.2008.00185.x
68. Grau-Bové X., Torruella G., Donachie S., Suga H., Leonard G., Richards T. A.. **Dynamics of genomic innovation in the unicellular ancestry of animals**. *Elife* (2017) **6** e26036. DOI: 10.7554/eLife.26036
69. Hartenstein V., Martinez P.. **Phagocytosis in cellular defense and nutrition: A food-centered approach to the evolution of macrophages**. *Cell Tissue Res.* (2019) **377** 527-547. DOI: 10.1007/s00441-019-03096-6
70. Haskó G., Pacher P.. **Regulation of macrophage function by adenosine**. *Arterioscler. Thromb. Vasc. Biol.* (2012) **32** 865-869. DOI: 10.1161/ATVBAHA.111.226852
71. Hehenberger E., Tikhonenkov D. V., Kolisko M., del Campo J., Esaulov A. S., Mylnikov A. P.. **Novel predators reshape holozoan phylogeny and reveal the presence of a two-component signaling system in the ancestor of animals**. *Curr. Biol.* (2017) **27** 2043-2050.e6. DOI: 10.1016/j.cub.2017.06.006
72. Herron M. D., Ratcliff W. C., Boswell J., Rosenzweig F.. **Genetics of a de novo origin of undifferentiated multicellularity**. *R. Soc. Open Sci.* (2018) **5** 180912. DOI: 10.1098/rsos.180912
73. Hinman V., Cary G.. **The evolution of gene regulation**. *Elife* (2017) **6** e27291. DOI: 10.7554/eLife.27291
74. Hirose S., Benabentos R., Ho H.-I., Kuspa A., Shaulsky G.. **Self-recognition in social amoebae is mediated by allelic pairs of tiger genes**. *Science* (2011) **333** 467-470. DOI: 10.1126/science.1203903
75. Holt D. J., Grainger D. W.. **Senescence and quiescence induced compromised function in cultured macrophages**. *Biomaterials* (2012) **33** 7497-7507. DOI: 10.1016/j.biomaterials.2012.06.099
76. Hynes R. O., Zhao Q.. **The evolution of cell adhesion**. *J. Cell Biol.* (2000) **150** F89-F96. DOI: 10.1083/jcb.150.2.f89
77. Iglesias P. A.. **Chemoattractant signaling in Dictyostelium: Adaptation and amplification**. *Sci. Signal.* (2012) **5** pe8. DOI: 10.1126/scisignal.2002897
78. Jang W., Gomer R. H.. **Initial cell type choice in Dictyostelium**. *Eukaryot. Cell* (2011) **10** 150-155. DOI: 10.1128/EC.00219-10
79. Jennings B. H.. **Drosophila – A versatile model in biology & medicine**. *Mat. Today* (2011) **14** 190-195. DOI: 10.1016/s1369-7021(11)70113-4
80. Kaufmann S. H. E.. **Immunology’s foundation: The 100-year anniversary of the Nobel prize to Paul ehrlich and Elie Metchnikoff**. *Nat. Immunol.* (2008) **9** 705-712. DOI: 10.1038/ni0708-705
81. Kelly B., Carrizo G. E., Edwards-Hicks J., Sanin D. E., Stanczak M. A., Priesnitz C.. **Sulfur sequestration promotes multicellularity during nutrient limitation**. *Nature* (2021) **591** 471-476. DOI: 10.1038/s41586-021-03270-3
82. Kharrat B., Csordás G., Honti V.. **Peeling back the layers of lymph gland structure and regulation**. *Int. J. Mol. Sci.* (2022) **23** 7767. DOI: 10.3390/ijms23147767
83. Kierdorf K., Hersperger F., Sharrock J., Vincent C. M., Ustaoglu P., Dou J.. **Muscle function and homeostasis require cytokine inhibition of AKT activity in Drosophila**. *Elife* (2020) **9** e51595. DOI: 10.7554/eLife.51595
84. Kim C. H.. **Immune regulation by microbiome metabolites**. *Immunology* (2018) **154** 220-229. DOI: 10.1111/imm.12930
85. Kim T., Song B., Lee I.-S.. **Drosophila glia: Models for human neurodevelopmental and neurodegenerative disorders**. *Int. J. Mol. Sci.* (2020) **21** 4859. DOI: 10.3390/ijms21144859
86. King N.. **The unicellular ancestry of animal development**. *Dev. Cell* (2004) **7** 313-325. DOI: 10.1016/j.devcel.2004.08.010
87. Krejčová G., Danielová A., Nedbalová P., Kazek M., Strych L., Chawla G.. **Drosophila macrophages switch to aerobic glycolysis to mount effective antibacterial defense**. *Elife* (2019) **8** e50414. DOI: 10.7554/eLife.50414
88. Kurant E.. **Keeping the CNS clear: Glial phagocytic functions in Drosophila**. *Glia* (2011) **59** 1304-1311. DOI: 10.1002/glia.21098
89. Kwon D. H., Lee H., Park C., Hong S.-H., Hong S. H., Kim G.-Y.. **Glutathione induced immune-stimulatory activity by promoting M1-like macrophages polarization via potential ROS scavenging capacity**. *Antioxidants* (2019) **8** 413. DOI: 10.3390/antiox8090413
90. Lang B. F., O’Kelly C., Nerad T., Gray M. W., Burger G.. **The closest unicellular relatives of animals**. *Curr. Biol.* (2002) **12** 1773-1778. DOI: 10.1016/s0960-9822(02)01187-9
91. Lardy B., Bof M., Aubry L., Paclet M. H., Morel F., Satre M.. **NADPH oxidase homologs are required for normal cell differentiation and morphogenesis in**. *Biochim. Biophys. Acta - Mol. Cell Res.* (2005) **1744** 199-212. DOI: 10.1016/j.bbamcr.2005.02.004
92. Laundon D., Larson B. T., McDonald K., King N., Burkhardt P.. **The architecture of cell differentiation in choanoflagellates and sponge choanocytes**. *PLOS Biol.* (2019) **17** e3000226. DOI: 10.1371/journal.pbio.3000226
93. Lauterbach M. A. R., Wunderlich F. T.. **Macrophage function in obesity-induced inflammation and insulin resistance**. *Pflügers Arch. - Eur. J. Physiol.* (2017) **469** 385-396. DOI: 10.1007/s00424-017-1955-5
94. Lavin Y., Merad M.. **Macrophages: Gatekeepers of tissue integrity**. *Cancer Immunol. Res.* (2013) **1** 201-209. DOI: 10.1158/2326-6066.CIR-13-0117
95. Lavrov A. I., Kosevich I. A.. **Sponge cell reaggregation: Mechanisms and dynamics of the process**. *Russ. J. Dev. Biol.* (2014) **45** 205-223. DOI: 10.1134/s1062360414040067
96. Lee E., Eo J.-C., Lee C., Yu J.-W.. **Distinct features of brain-resident macrophages: Microglia and non-parenchymal brain macrophages**. *Mol. Cells* (2021) **44** 281-291. DOI: 10.14348/molcells.2021.0060
97. Lemaitre B., Nicolas E., Michaut L., Reichhart J.-M., Hoffmann J. A.. **The dorsoventral regulatory gene cassette spätzle/Toll/cactus controls the potent antifungal response in Drosophila adults**. *Cell* (1996) **86** 973-983. DOI: 10.1016/s0092-8674(00)80172-5
98. Li D., Wu M.. **Pattern recognition receptors in health and diseases**. *Signal Transduct. Target. Ther.* (2021) **6** 291. DOI: 10.1038/s41392-021-00687-0
99. Liang W., Chen K., Gong W., Yoshimura T., Le Y., Wang Y.. **The contribution of chemoattractant GPCRs, formylpeptide receptors, to inflammation and cancer**. *Front. Endocrinol. (Lausanne)* (2020) **11** 17. DOI: 10.3389/fendo.2020.00017
100. Lin P., Ji H.-H., Li Y.-J., Guo S.-D.. **Macrophage plasticity and atherosclerosis therapy**. *Front. Mol. Biosci.* (2021) **8** 679797. DOI: 10.3389/fmolb.2021.679797
101. Lin Y., Xu J., Lan H.. **Tumor-associated macrophages in tumor metastasis: Biological roles and clinical therapeutic applications**. *J. Hematol. Oncol.* (2019) **12** 76. DOI: 10.1186/s13045-019-0760-3
102. Liu Y., Xu R., Gu H., Zhang E., Qu J., Cao W.. **Metabolic reprogramming in macrophage responses**. *Biomark. Res.* (2021) **9** 1. DOI: 10.1186/s40364-020-00251-y
103. Locati M., Curtale G., Mantovani A.. **Diversity, mechanisms, and significance of macrophage plasticity**. *Annu. Rev. Pathol. Mech. Dis.* (2020) **15** 123-147. DOI: 10.1146/annurev-pathmechdis-012418-012718
104. Loomis W. F.. **Cell signaling during development of Dictyostelium**. *Dev. Biol.* (2014) **391** 1-16. DOI: 10.1016/j.ydbio.2014.04.001
105. Lu Y. J., Wheeler L. W., Chu H., Kleindl P. J., Pugh M., You F.. **Targeting folate receptor beta on monocytes/macrophages renders rapid inflammation resolution independent of root causes**. *Cell Rep. Med.* (2021) **2** 100422. DOI: 10.1016/j.xcrm.2021.100422
106. Maldonado M.. **Sponge waste that fuels marine oligotrophic food webs: A re‐assessment of its origin and nature**. *Mar. Ecol.* (2016) **37** 477-491. DOI: 10.1111/maec.12256
107. Mammana S., Fagone P., Cavalli E., Basile M., Petralia M., Nicoletti F.. **The role of macrophages in neuroinflammatory and neurodegenerative pathways of alzheimer’s disease, amyotrophic lateral sclerosis, and multiple sclerosis: Pathogenetic cellular effectors and potential therapeutic targets**. *Int. J. Mol. Sci.* (2018) **19** 831. DOI: 10.3390/ijms19030831
108. Martin C. J., Peters K. N., Behar S. M.. **Macrophages clean up: Efferocytosis and microbial control**. *Curr. Opin. Microbiol.* (2014) **17** 17-23. DOI: 10.1016/j.mib.2013.10.007
109. Martinelli C., Reichhart J.-M.. **Evolution and integration of innate immune systems from fruit flies to man: Lessons and questions**. *J. Endotoxin Res.* (2005) **11** 243-248. DOI: 10.1179/096805105X37411
110. Mase A., Augsburger J., Brückner K.. **Macrophages and their organ locations shape each other in development and homeostasis – a Drosophila perspective**. *Front. Cell Dev. Biol.* (2021) **9** 630272. DOI: 10.3389/fcell.2021.630272
111. Mass E., Ballesteros I., Farlik M., Halbritter F., Günther P., Crozet L.. **Specification of tissue-resident macrophages during organogenesis**. *Science* (2016) **80** aaf4238. DOI: 10.1126/science.aaf4238
112. Melcarne C., Lemaitre B., Kurant E.. **Phagocytosis in Drosophila: From molecules and cellular machinery to physiology**. *Insect biochem. Mol. Biol.* (2019) **109** 1-12. DOI: 10.1016/j.ibmb.2019.04.002
113. Merien F.. **A journey with Elie Metchnikoff: From innate cell mechanisms in infectious diseases to quantum biology**. *Front. Public Heal.* (2016) **4** 125. DOI: 10.3389/fpubh.2016.00125
114. Merkey A. B., Wong C. K., Hoshizaki D. K., Gibbs A. G.. **Energetics of metamorphosis in**. *J. Insect Physiol.* (2011) **57** 1437-1445. DOI: 10.1016/j.jinsphys.2011.07.013
115. Mesquita A., Cardenal-Muñoz E., Dominguez E., Muñoz-Braceras S., Nuñez-Corcuera B., Phillips B. A.. **Autophagy in Dictyostelium: Mechanisms, regulation and disease in a simple biomedical model**. *Autophagy* (2017) **13** 24-40. DOI: 10.1080/15548627.2016.1226737
116. Misevic G. N.. **Molecular self-recognition and adhesion via proteoglycan to proteoglycan interactions as a pathway to multicellularity: Atomic force microscopy and color coded bead measurements in sponges**. *Microsc. Res. Tech.* (1999) **44** 304-309. DOI: 10.1002/(SICI)1097-0029(19990215)44:4<304::AID-JEMT9>3.0.CO;2-X
117. Moeini P., Niedźwiedzka-Rystwej P.. **Tumor-associated macrophages: Combination of therapies, the approach to improve cancer treatment**. *Int. J. Mol. Sci.* (2021) **22** 7239. DOI: 10.3390/ijms22137239
118. Mogensen T. H.. **Pathogen recognition and inflammatory signaling in innate immune defenses**. *Clin. Microbiol. Rev.* (2009) **22** 240-273. DOI: 10.1128/CMR.00046-08
119. Molmeret M., Horn M., Wagner M., Santic M., Abu Kwaik Y.. **Amoebae as training grounds for intracellular bacterial pathogens**. *Appl. Environ. Microbiol.* (2005) **71** 20-28. DOI: 10.1128/AEM.71.1.20-28.2005
120. Morgantini C., Jager J., Li X., Levi L., Azzimato V., Sulen A.. **Liver macrophages regulate systemic metabolism through non-inflammatory factors**. *Nat. Metab.* (2019) **1** 445-459. DOI: 10.1038/s42255-019-0044-9
121. Müller W. E. G., Custódio M. R., Wiens M., Zilberberg C., Châtel A., Müller I. M.. **Effect of bacterial infection on stem cell pattern in Porifera**. *Stem cells in marine organisms* (2009)
122. Muller W. E. G., Muller I. M.. **Origin of the metazoan immune system: Identification of the molecules and their functions in sponges**. *Integr. Comp. Biol.* (2003) **43** 281-292. DOI: 10.1093/icb/43.2.281
123. Müller W. E. G.. **The stem cell concept in sponges (Porifera): Metazoan traits**. *Semin. Cell Dev. Biol.* (2006) **17** 481-491. DOI: 10.1016/j.semcdb.2006.05.006
124. Munro D. A. D., Hughes J.. **The origins and functions of tissue-resident macrophages in kidney development**. *Front. Physiol.* (2017) **8** 837. DOI: 10.3389/fphys.2017.00837
125. Nadesalingam J., Dodds A. W., Reid K. B. M., Palaniyar N.. **Mannose-binding lectin recognizes peptidoglycan via the N -acetyl glucosamine moiety, and inhibits ligand-induced proinflammatory effect and promotes chemokine production by macrophages**. *J. Immunol.* (2005) **175** 1785-1794. DOI: 10.4049/jimmunol.175.3.1785
126. Nagahata Y., Masuda K., Nishimura Y., Ikawa T., Kawaoka S., Kitawaki T.. **Tracing the evolutionary history of blood cells to the unicellular ancestor of animals**. *Blood* (2022) **140** 2611-2625. DOI: 10.1182/blood.2022016286
127. Nakanishi N., Sogabe S., Degnan B. M.. **Evolutionary origin of gastrulation: Insights from sponge development**. *BMC Biol.* (2014) **12** 26. DOI: 10.1186/1741-7007-12-26
128. Natoli G., Monticelli S.. **Macrophage activation: Glancing into diversity**. *Immunity* (2014) **40** 175-177. DOI: 10.1016/j.immuni.2014.01.004
129. Negreiros-Lima G. L., Lima K. M., Moreira I. Z., Jardim B. L. O., Vago J. P., Galvão I.. **Cyclic AMP regulates key features of macrophages via PKA: Recruitment, reprogramming and efferocytosis**. *Cells* (2020) **9** 128. DOI: 10.3390/cells9010128
130. Netea M. G., Domínguez-Andrés J., Barreiro L. B., Chavakis T., Divangahi M., Fuchs E.. **Defining trained immunity and its role in health and disease**. *Nat. Rev. Immunol.* (2020) **20** 375-388. DOI: 10.1038/s41577-020-0285-6
131. Neves J., Zhu J., Sousa-Victor P., Konjikusic M., Riley R., Chew S.. **Immune modulation by MANF promotes tissue repair and regenerative success in the retina**. *Science* (2016) **80** aaf3646. DOI: 10.1126/science.aaf3646
132. Nielsen C.. **Early animal evolution: A morphologist’s view**. *R. Soc. Open Sci.* (2019) **6** 190638. DOI: 10.1098/rsos.190638
133. Nobs S. P., Kopf M.. **Tissue-resident macrophages: Guardians of organ homeostasis**. *Trends Immunol.* (2021) **42** 495-507. DOI: 10.1016/j.it.2021.04.007
134. Nonnenmacher Y., Hiller K.. **Biochemistry of proinflammatory macrophage activation**. *Cell. Mol. Life Sci.* (2018) **75** 2093-2109. DOI: 10.1007/s00018-018-2784-1
135. O’Callaghan A. A., Dempsey E., Iyer N., Stiegeler S., Mercurio K., Corr S. C.. **Intestinal metabolites influence macrophage phagocytosis and clearance of bacterial infection**. *Front. Cell. Infect. Microbiol.* (2021) **11** 622491. DOI: 10.3389/fcimb.2021.622491
136. Olofsson B., Page D. T.. **Condensation of the central nervous system in embryonic Drosophila is inhibited by blocking hemocyte migration or neural activity**. *Dev. Biol.* (2005) **279** 233-243. DOI: 10.1016/j.ydbio.2004.12.020
137. Palmieri E. M., Gonzalez-Cotto M., Baseler W. A., Davies L. C., Ghesquière B., Maio N.. **Nitric oxide orchestrates metabolic rewiring in M1 macrophages by targeting aconitase 2 and pyruvate dehydrogenase**. *Nat. Commun.* (2020) **11** 698. DOI: 10.1038/s41467-020-14433-7
138. Pan M., Xu X., Chen Y., Jin T.. **Identification of a chemoattractant G-protein-coupled receptor for folic acid that controls both chemotaxis and phagocytosis**. *Dev. Cell* (2016) **36** 428-439. DOI: 10.1016/j.devcel.2016.01.012
139. Parisi L., Gini E., Baci D., Tremolati M., Fanuli M., Bassani B.. **Macrophage polarization in chronic inflammatory diseases: Killers or builders?**. *J. Immunol. Res.* (2018) **2018** 1-25. DOI: 10.1155/2018/8917804
140. Pasquini S., Contri C., Borea P. A., Vincenzi F., Varani K.. **Adenosine and inflammation: Here, there and everywhere**. *Int. J. Mol. Sci.* (2021) **22** 7685. DOI: 10.3390/ijms22147685
141. Pham L. N., Dionne M. S., Shirasu-Hiza M., Schneider D. S.. **A specific primed immune response in Drosophila is dependent on phagocytes**. *PLoS Pathog.* (2007) **3** e26. DOI: 10.1371/journal.ppat.0030026
142. Pizzagalli D. U., Pulfer A., Thelen M., Krause R., Gonzalez S. F.. *Front. Immunol.* (2022) **12** 804159. DOI: 10.3389/fimmu.2021.804159
143. Porta C., Rimoldi M., Raes G., Brys L., Ghezzi P., Di Liberto D.. **Tolerance and M2 (alternative) macrophage polarization are related processes orchestrated by p50 nuclear factor κB**. *Proc. Natl. Acad. Sci.* (2009) **106** 14978-14983. DOI: 10.1073/pnas.0809784106
144. Preethi P., Tomar A., Madhwal S., Mukherjee T.. **Immune control of animal growth in homeostasis and nutritional stress in Drosophila**. *Front. Immunol.* (2020) **11** 1528. DOI: 10.3389/fimmu.2020.01528
145. Ramond E., Dudzic J. P., Lemaitre B.. **Comparative RNA-Seq analyses of Drosophila plasmatocytes reveal gene specific signatures in response to clean injury and septic injury**. *PLoS One* (2020) **15** e0235294. DOI: 10.1371/journal.pone.0235294
146. Rath M., Mã¼ller I., Kropf P., Closs E. I., Munder M.. **Metabolism via arginase or nitric oxide synthase: Two competing arginine pathways in macrophages**. *Front. Immunol.* (2014) **5** 532. DOI: 10.3389/fimmu.2014.00532
147. Ravichandran K. S.. **Find-me and eat-me signals in apoptotic cell clearance: Progress and conundrums**. *J. Exp. Med.* (2010) **207** 1807-1817. DOI: 10.1084/jem.20101157
148. Rayamajhee B., Willcox M. D. P., Henriquez F. L., Petsoglou C., Subedi D., Carnt N.. **Acanthamoeba, an environmental phagocyte enhancing survival and transmission of human pathogens**. *Trends Parasitol.* (2022) **38** 975-990. DOI: 10.1016/j.pt.2022.08.007
149. Razzell W., Wood W., Martin P.. **Swatting flies: Modelling wound healing and inflammation in Drosophila**. *Dis. Model. Mech.* (2011) **4** 569-574. DOI: 10.1242/dmm.006825
150. Rodriguez P. C., Ochoa A. C., Al-Khami A. A.. **Arginine metabolism in myeloid cells shapes innate and adaptive immunity**. *Front. Immunol.* (2017) **8** 93. DOI: 10.3389/fimmu.2017.00093
151. Romeralo M., Cavender J. C., Landolt J. C., Stephenson S. L., Baldauf S. L.. **An expanded phylogeny of social amoebas (Dictyostelia) shows increasing diversity and new morphological patterns**. *BMC Evol. Biol.* (2011) **11** 84. DOI: 10.1186/1471-2148-11-84
152. Ros-Rocher N., Pérez-Posada A., Leger M. M., Ruiz-Trillo I.. **The origin of animals: An ancestral reconstruction of the unicellular-to-multicellular transition**. *Open Biol.* (2021) **11** 200359. DOI: 10.1098/rsob.200359
153. Ross E. A., Devitt A., Johnson J. R.. **Macrophages: The good, the bad, and the gluttony**. *Front. Immunol.* (2021) **12** 708186. DOI: 10.3389/fimmu.2021.708186
154. Salah I. B., Ghigo E., Drancourt M.. **Free-living amoebae, a training field for macrophage resistance of mycobacteria**. *Clin. Microbiol. Infect.* (2009) **15** 894-905. DOI: 10.1111/j.1469-0691.2009.03011.x
155. Schmitz F., Mages J., Heit A., Lang R., Wagner H.. **Transcriptional activation induced in macrophages by toll-like receptor (TLR) ligands: From expression profiling to a model of TLR signaling**. *Eur. J. Immunol.* (2004) **34** 2863-2873. DOI: 10.1002/eji.200425228
156. Schuster F. L., Levandowsky M.. **Chemosensory responses of Acanthamoeba castellanii: Visual analysis of random movement and responses to chemical signals**. *J. Eukaryot. Microbiol.* (1996) **43** 150-158. DOI: 10.1111/j.1550-7408.1996.tb04496.x
157. Sciaraffia E., Riccomi A., Lindstedt R., Gesa V., Cirelli E., Patrizio M.. **Human monocytes respond to extracellular cAMP through A2A and A2B adenosine receptors**. *J. Leukoc. Biol.* (2014) **96** 113-122. DOI: 10.1189/jlb.3A0513-302RR
158. Sender R., Milo R.. **The distribution of cellular turnover in the human body**. *Nat. Med.* (2021) **27** 45-48. DOI: 10.1038/s41591-020-01182-9
159. Sica A., Erreni M., Allavena P., Porta C.. **Macrophage polarization in pathology**. *Cell. Mol. Life Sci.* (2015) **72** 4111-4126. DOI: 10.1007/s00018-015-1995-y
160. Siddiqui R., Khan N. A.. **Acanthamoeba is an evolutionary ancestor of macrophages: A myth or reality?**. *Exp. Parasitol.* (2012a) **130** 95-97. DOI: 10.1016/j.exppara.2011.11.005
161. Siddiqui R., Khan N. A.. **Biology and pathogenesis of Acanthamoeba**. *Parasit. Vectors* (2012b) **5** 6. DOI: 10.1186/1756-3305-5-6
162. Siddiqui R., Roberts S. K., Ong T. Y. Y., Mungroo M. R., Anwar A., Khan N. A.. **Novel insights into the potential role of ion transport in sensory perception in Acanthamoeba**. *Parasit. Vectors* (2019) **12** 538. DOI: 10.1186/s13071-019-3785-0
163. Singer G., Araki T., Weijer C. J.. **Oscillatory cAMP cell-cell signalling persists during multicellular Dictyostelium development**. *Commun. Biol.* (2019) **2** 139. DOI: 10.1038/s42003-019-0371-0
164. Smith L. C., Hildemann W. H.. **Allograft rejection, autograft fusion and inflammatory responses to injury in Callyspongia diffusa (Porifera; Demospongia)**. *Proc. R. Soc. Lond. Ser. B. Biol. Sci.* (1986) **226** 445-464. DOI: 10.1098/rspb.1986.0003
165. Sogabe S., Hatleberg W. L., Kocot K. M., Say T. E., Stoupin D., Roper K. E.. **Pluripotency and the origin of animal multicellularity**. *Nature* (2019) **570** 519-522. DOI: 10.1038/s41586-019-1290-4
166. Sokol C. L., Luster A. D.. **The chemokine system in innate immunity**. *Cold Spring Harb. Perspect. Biol.* (2015) **7** a016303. DOI: 10.1101/cshperspect.a016303
167. Srivastava M., Simakov O., Chapman J., Fahey B., Gauthier M. E. A., Mitros T.. **The Amphimedon queenslandica genome and the evolution of animal complexity**. *Nature* (2010) **466** 720-726. DOI: 10.1038/nature09201
168. Steinz M. M., Ezdoglian A., Khodadust F., Molthoff C. F. M., Srinivasarao M., Low P. S.. **Folate receptor beta for macrophage imaging in rheumatoid arthritis**. *Front. Immunol.* (2022) **13** 819163. DOI: 10.3389/fimmu.2022.819163
169. Storelli G., Nam H.-J., Simcox J., Villanueva C. J., Thummel C. S.. **Drosophila HNF4 directs a switch in lipid metabolism that supports the transition to adulthood**. *Dev. Cell* (2019) **48** 200-214.e6. DOI: 10.1016/j.devcel.2018.11.030
170. Strakhova R., Cadassou O., Cros-Perrial E., Jordheim L. P.. **Regulation of tumor infiltrated innate immune cells by adenosine**. *Purinergic Signal* (2020) **16** 289-295. DOI: 10.1007/s11302-020-09701-6
171. Strassmann J. E., Queller D. C.. **Evolution of cooperation and control of cheating in a social microbe**. *Proc. Natl. Acad. Sci.* (2011) **108** 10855-10862. DOI: 10.1073/pnas.1102451108
172. Stratoulias V., Heino T. I.. **MANF silencing, immunity induction or autophagy trigger an unusual cell type in metamorphosing Drosophila brain**. *Cell. Mol. Life Sci.* (2015) **72** 1989-2004. DOI: 10.1007/s00018-014-1789-7
173. Swart A. L., Harrison C. F., Eichinger L., Steinert M., Hilbi H.. **Acanthamoeba and Dictyostelium as cellular models for Legionella infection**. *Front. Cell. Infect. Microbiol.* (2018) **8** 61. DOI: 10.3389/fcimb.2018.00061
174. Szondi D. C., Wong J. K., Vardy L. A., Cruickshank S. M.. **Arginase signalling as a key player in chronic wound pathophysiology and healing**. *Front. Mol. Biosci.* (2021) **8** 773866. DOI: 10.3389/fmolb.2021.773866
175. Tattikota S. G., Cho B., Liu Y., Hu Y., Barrera V., Steinbaugh M. J.. **A single-cell survey of Drosophila blood**. *Elife* (2020) **9** e54818. DOI: 10.7554/eLife.54818
176. Tauber A. I.. **A history of the immune self**. *Immunity* (2017) 23-56
177. Tauber A. I.. **Metchnikoff and the phagocytosis theory**. *Nat. Rev. Mol. Cell Biol.* (2003) **4** 897-901. DOI: 10.1038/nrm1244
178. Tavares L. P., Negreiros-Lima G. L., Lima K. M., E Silva P. M. R., Pinho V., Teixeira M. M.. **Blame the signaling: Role of cAMP for the resolution of inflammation**. *Pharmacol. Res.* (2020) **159** 105030. DOI: 10.1016/j.phrs.2020.105030
179. Thacker R. W., Díaz M. C., Kerner A., Vignes-Lebbe R., Segerdell E., Haendel M. A.. **The Porifera ontology (PORO): Enhancing sponge systematics with an anatomy ontology**. *J. Biomed. Semant.* (2014) **5** 39. DOI: 10.1186/2041-1480-5-39
180. Theret M., Mounier R., Rossi F.. **The origins and non-canonical functions of macrophages in development and regeneration**. *Development* (2019) **146** dev156000. DOI: 10.1242/dev.156000
181. Tice A. K., Shadwick L. L., Fiore-Donno A. M., Geisen S., Kang S., Schuler G. A.. **Expansion of the molecular and morphological diversity of Acanthamoebidae (Centramoebida, Amoebozoa) and identification of a novel life cycle type within the group**. *Biol. Direct* (2016) **11** 69. DOI: 10.1186/s13062-016-0171-0
182. Trojan E., Bryniarska N., Leśkiewicz M., Regulska M., Chamera K., Szuster-Głuszczak M.. **The contribution of formyl peptide receptor dysfunction to the course of neuroinflammation: A potential role in the brain pathology**. *Curr. Neuropharmacol.* (2020) **18** 229-249. DOI: 10.2174/1570159X17666191019170244
183. Vanha-aho L.-M., Valanne S., Rämet M.. **Cytokines in Drosophila immunity**. *Immunol. Lett.* (2016) **170** 42-51. DOI: 10.1016/j.imlet.2015.12.005
184. Viola A., Munari F., Sánchez-Rodríguez R., Scolaro T., Castegna A.. **The metabolic signature of macrophage responses**. *Front. Immunol.* (2019) **10** 1462. DOI: 10.3389/fimmu.2019.01462
185. Vogel G., Thilo L., Schwarz H., Steinhart R.. **Mechanism of phagocytosis in dictyostelium discoideum: Phagocytosis is mediated by different recognition sites as disclosed by mutants with altered phagocytotic properties**. *J. Cell Biol.* (1980) **86** 456-465. DOI: 10.1083/jcb.86.2.456
186. Wang L., Li H., Tang Y., Yao P.. **Potential mechanisms and effects of efferocytosis in atherosclerosis**. *Front. Endocrinol. (Lausanne)* (2021a) **11** 11. DOI: 10.3389/fendo.2020.585285
187. Wang L., Zhang S., Wu H., Rong X., Guo J.. **M2b macrophage polarization and its roles in diseases**. *J. Leukoc. Biol.* (2019a) **106** 345-358. DOI: 10.1002/JLB.3RU1018-378RR
188. Wang S. Y., Pollina E. A., Wang I.-H., Pino L. K., Bushnell H. L., Takashima K.. **Role of epigenetics in unicellular to multicellular transition in Dictyostelium**. *Genome Biol.* (2021b) **22** 134. DOI: 10.1186/s13059-021-02360-9
189. Wang T., Liu H., Lian G., Zhang S.-Y., Wang X., Jiang C.. **HIF1 α -induced glycolysis metabolism is essential to the activation of inflammatory macrophages**. *Mediat. Inflamm.* (2017) **2017** 1-10. DOI: 10.1155/2017/9029327
190. Wang X., Iyer A., Lyons A. B., Körner H., Wei W.. **Emerging roles for G-protein coupled receptors in development and activation of macrophages**. *Front. Immunol.* (2019b) **10**. DOI: 10.3389/fimmu.2019.02031
191. Wenceslau C. F., McCarthy C. G., Goulopoulou S., Szasz T., NeSmith E. G., Webb R. C.. **Mitochondrial-derived N-formyl peptides: Novel links between trauma, vascular collapse and sepsis**. *Med. Hypotheses* (2013) **81** 532-535. DOI: 10.1016/j.mehy.2013.06.026
192. Wiens M., Korzhev M., Krasko A., Thakur N. L., Perović-Ottstadt S., Breter H. J.. **Innate immune defense of the sponge**. *J. Biol. Chem.* (2005) **280** 27949-27959. DOI: 10.1074/jbc.M504049200
193. Witherel C. E., Sao K., Brisson B. K., Han B., Volk S. W., Petrie R. J.. **Regulation of extracellular matrix assembly and structure by hybrid M1/M2 macrophages**. *Biomaterials* (2021) **269** 120667. DOI: 10.1016/j.biomaterials.2021.120667
194. Wood W., Jacinto A., Grose R., Woolner S., Gale J., Wilson C.. **Wound healing recapitulates morphogenesis in Drosophila embryos**. *Nat. Cell Biol.* (2002) **4** 907-912. DOI: 10.1038/ncb875
195. Woodcock K. J., Kierdorf K., Pouchelon C. A., Vivancos V., Dionne M. S., Geissmann F.. **Macrophage-Derived upd3 cytokine causes impaired glucose homeostasis and reduced lifespan in Drosophila fed a lipid-rich diet**. *Immunity* (2015) **42** 133-144. DOI: 10.1016/j.immuni.2014.12.023
196. Wu Y., Hirschi K. K.. **Tissue-resident macrophage development and function**. *Front. Cell Dev. Biol.* (2021) **8**. DOI: 10.3389/fcell.2020.617879
197. Xia W., Hilgenbrink A. R., Matteson E. L., Lockwood M. B., Cheng J.-X., Low P. S.. **A functional folate receptor is induced during macrophage activation and can be used to target drugs to activated macrophages**. *Blood* (2009) **113** 438-446. DOI: 10.1182/blood-2008-04-150789
198. Xu X., Pan M., Jin T.. **How phagocytes acquired the capability of hunting and removing pathogens from a human body: Lessons learned from chemotaxis and phagocytosis of**. *Front. Cell Dev. Biol.* (2021) **9** 724940. DOI: 10.3389/fcell.2021.724940
199. Xuan W., Qu Q., Zheng B., Xiong S., Fan G.-H.. **The chemotaxis of M1 and M2 macrophages is regulated by different chemokines**. *J. Leukoc. Biol.* (2015) **97** 61-69. DOI: 10.1189/jlb.1A0314-170R
200. Yarnitzky T., Volk T.. **Laminin is required for heart, somatic muscles, and gut development in the Drosophila embryo**. *Dev. Biol.* (1995) **169** 609-618. DOI: 10.1006/dbio.1995.1173
201. Zhang X., Cao X., Zhang W., Yu X., Jin M.. **Primmorphs from archaeocytes-dominant cell population of the spongehymeniacidon perleve: Improved cell proliferation and spiculogenesis**. *Biotechnol. Bioeng.* (2003) **84** 583-590. DOI: 10.1002/bit.10811
202. Zhang X., Soldati T.. **Of amoebae and men: Extracellular DNA traps as an ancient cell-intrinsic defense mechanism**. *Front. Immunol.* (2016) **7** 269. DOI: 10.3389/fimmu.2016.00269
203. Zhou J., Tang Z., Gao S., Li C., Feng Y., Zhou X.. **Tumor-associated macrophages: Recent insights and therapies**. *Front. Oncol.* (2020) **10** 188. DOI: 10.3389/fonc.2020.00188
|
---
title: Exploiting the mediating role of the metabolome to unravel transcript-to-phenotype
associations
authors:
- Chiara Auwerx
- Marie C Sadler
- Tristan Woh
- Alexandre Reymond
- Zoltán Kutalik
- Eleonora Porcu
journal: eLife
year: 2023
pmcid: PMC9998083
doi: 10.7554/eLife.81097
license: CC BY 4.0
---
# Exploiting the mediating role of the metabolome to unravel transcript-to-phenotype associations
## Abstract
Despite the success of genome-wide association studies (GWASs) in identifying genetic variants associated with complex traits, understanding the mechanisms behind these statistical associations remains challenging. Several methods that integrate methylation, gene expression, and protein quantitative trait loci (QTLs) with GWAS data to determine their causal role in the path from genotype to phenotype have been proposed. Here, we developed and applied a multi-omics Mendelian randomization (MR) framework to study how metabolites mediate the effect of gene expression on complex traits. We identified 216 transcript-metabolite-trait causal triplets involving 26 medically relevant phenotypes. Among these associations, $58\%$ were missed by classical transcriptome-wide MR, which only uses gene expression and GWAS data. This allowed the identification of biologically relevant pathways, such as between ANKH and calcium levels mediated by citrate levels and SLC6A12 and serum creatinine through modulation of the levels of the renal osmolyte betaine. We show that the signals missed by transcriptome-wide MR are found, thanks to the increase in power conferred by integrating multiple omics layer. Simulation analyses show that with larger molecular QTL studies and in case of mediated effects, our multi-omics MR framework outperforms classical MR approaches designed to detect causal relationships between single molecular traits and complex phenotypes.
## Introduction
Genome-wide association studies (GWASs) have identified thousands of single nucleotide polymorphisms (SNPs) associated with a wide range of complex traits (MacArthur et al., 2017; Visscher et al., 2017). However, the path from GWAS to biology is not straightforward as most SNPs implicated by GWASs reside in non-coding regions of the genome (MacArthur et al., 2017) and do not directly inform on the functional mechanism through which variants exert their effect on phenotypes.
GWASs have been performed on gene expression (Võsa et al., 2021), DNA methylation (Min et al., 2021), protein (Sun et al., 2018), and metabolite (Shin et al., 2014; Lotta et al., 2021) levels, identifying genetic variants influencing molecular traits, commonly referred to as molecular quantitative trait loci (molQTLs). The large overlap between complex and molecular trait-associated variants suggests that integrating these data can help interpreting GWAS loci (Vandiedonck, 2018; Taylor et al., 2019; Ongen et al., 2017). Advances in the field of transcriptomics make gene expression the best studied molecular phenotype, thanks to the presence of large expression QTL (eQTL) studies (e.g., eQTLGen Consortium [Võsa et al., 2021] N>30,000). Availability of these datasets fostered the development of summary statistic-based statistical approaches aiming at identifying associations between transcripts and complex traits (Zhu et al., 2016; Porcu et al., 2019; Hormozdiari et al., 2014; Gusev et al., 2016), prioritizing genes from known GWAS loci for functional follow-up, and inferring the directionality of these relations (Porcu et al., 2019; Porcu et al., 2021b). However, the cascade of events that mediates the effect of genetic variants on complex traits involves more than one molecular trait. Although approaches used for gene expression can be extended to other molecular data, investigating whether these molecular traits reside along the same causal pathway remains under-explored and only recently have studies applied colocalization and Mendelian randomization (MR) to methylation, gene expression, and protein levels data (Giambartolomei et al., 2018; Wu et al., 2018; Gleason et al., 2020; Sadler et al., 2022) and to a lesser extent to metabolic QTLs (mQTL) (Yin et al., 2022).
Metabolites are often the final products of cellular regulatory processes and the most proximal omic layer to complex phenotypes. Their levels could thus represent the ultimate response of biological systems to genetic and environmental changes. For instance, the metabolic status of organisms reflects disease progression, as metabolic disturbances can often be observed several years prior to the symptomatic phase (Shah et al., 2010; Wang et al., 2011; Sabatine et al., 2005). Therefore, using metabolomics to identify early-stage biomarkers of complex phenotypes, such as prediabetes and COVID-19 susceptibility, has gained increased interest (Wang-Sattler et al., 2012; Julkunen et al., 2021). While two-sample MR approaches using metabolites as single exposure have revealed biomarkers for several diseases (Qian et al., 2021; Lord et al., 2021; Porcu et al., 2021a), these analyses focused on the prediction of disease risk rather than on deciphering the mechanisms of discovered associations.
Integrating transcriptomics with metabolomics data can provide insights into how metabolites are regulated, elucidating targetable functional mechanisms. Here, we develop a framework based on established MR methodology that hypothesizes a mediating role of the metabolome in the transcript-to-phenotype axis, with the primary exposure being defined as an upstream omic layer, namely gene expression. Specifically, our integrative MR analysis combines summary-level multi-omics (i.e., GWAS, eQTL, and mQTL) data to compute the indirect effect of gene expression on complex traits mediated by metabolites in three steps (Figure 1). First, we map the transcriptome to the metabolome by identifying causal associations between transcripts and metabolites. Next, we screen metabolites for downstream causal effects on 28 complex phenotypes, resulting in the identification of gene expression → metabolite → phenotype cascades (Figure 1A). In parallel, we prioritize trait-associated genes by testing the association of transcripts with phenotypes (Figure 1B). Third, for transcripts found to causally influence either a metabolite (A) or a complex phenotype (B), we test whether the identified target genes exert their effect on the phenotype through the metabolite using multivariable MR (MVMR; Figure 1C). Finally, we carried out extensive power analyses to determine under which conditions the mediation analysis (Figure 1C) outperforms the conventional exposure-outcome MR framework (Figure 1B).
**Figure 1.:** *Workflow overview.(A) Estimation of the causal transcript-to-metabolite and metabolite-to-phenotype effects through univariable Mendelian randomization (MR). (B) Estimation of the causal transcript-to-phenotype effects through univariable transcriptome-wide MR (TWMR). (C) Estimation of the direct (i.e., not mediated by the metabolites) and mediated effect of transcripts on phenotypes through multivariable MR (MVMR) by accounting for mediation through the metabolome.*
## Mapping the transcriptome onto the metabolome
We applied univariable MR to identify metabolites whose levels are causally influenced by transcript levels in whole blood (Figure 1A). Summary statistics for cis-eQTLs stem from the eQTLGen Consortium meta-analysis of 19,942 transcripts in 31,684 individuals (Võsa et al., 2021), while summary statistics for mQTLs originate from a meta-analysis of 453 metabolites in 7824 individuals from two independent European cohorts: TwinsUK ($$n = 6056$$) and KORA ($$n = 1768$$) (Shin et al., 2014). After selecting SNPs included in both the eQTL and mQTL studies, our analysis was restricted to 7884 transcripts with ≥3 instrumental variables (IVs) (see Methods, Figure 1—figure supplement 1A) and 242 metabolites with an identifier in the Human Metabolome Database (HMDB) (Wishart et al., 2022) (see Methods, Supplementary file 1a). By testing each gene for association with the 242 metabolites, we detected 96 genes whose transcript levels causally impacted 75 metabolites, resulting in 133 unique transcript-metabolite associations (FDR $5\%$ considering all 1,907,690 instrumentable gene-metabolite pairs; Supplementary file 1b). Most involved genes ($86\%$; $\frac{83}{96}$) were causally influencing the level of a single metabolite, with some notable exceptions acting as mQTL hubs, such as TMEM258 and FADS2 both affecting the same 11 metabolites, followed by FADS1 affecting a subset of six metabolites. While only 5 ($3.8\%$) of the 133 associations were reported in HMDB, an automated literature review (see Methods) identified a match for 22 ($16.5\%$) of the identified transcript-metabolite pairs (Supplementary file 1b).
## Mapping the metabolome onto complex phenotypes
Univariable metabolome-wide MR (MWMR; Figure 1A) was used to identify causal relationships between 48 metabolites with ≥3 IVs (Figure 1—figure supplement 1B) and 28 complex phenotypes. The latter include a wide range of anthropometric traits, cardiovascular assessments, and blood biomarkers, whose summary statistics originate from the UK Biobank (UKB) (Bycroft et al., 2018;Supplementary file 1c). Overall, 34 metabolites were associated with at least one phenotype (FDR $5\%$ considering all 1344 metabolite-phenotype pairs), resulting in 132 unique metabolite-phenotype associations (Supplementary file 1d).
## Mapping the transcriptome onto complex phenotypes
We applied univariable transcriptome-wide MR (TWMR [Porcu et al., 2019] Figure 1B) to identify associations between expression levels of 10,435 transcripts from the eQTLGen Consortium with ≥3 IVs (Figure 1—figure supplement 1C) measured in both exposure and outcome datasets and the same 28 UKB phenotypes described in the previous section (Supplementary file 1c). In total, 5140 transcripts associated with at least one phenotype (FDR $5\%$ considering all 292,170 gene-phenotype pairs) resulting in 13,141 unique transcript-phenotype associations (Supplementary file 1e).
## Mapping metabolome-mediated effects of the transcriptome onto complex phenotypes
The mapping of putative causal effects performed in the previous steps provides the opportunity to infer the mediating role of the metabolome in biological processes leading to transcript-phenotype associations. We combined the 133 transcript-metabolite (FDR ≤$5\%$) and 132 metabolite-trait (FDR ≤$5\%$) associations to pinpoint 216 transcript-metabolite-phenotype causal triplets (FDR = 1–0.952 = $9.75\%$) (Supplementary file 1f). Among the 37 triplets for which the transcript and metabolite had previously been linked through automated literature review, none remained after incorporating a third term for the phenotype in the search and manually removing abstracts for which the search terms were used in an erroneous context. Relaxing the search criteria by omitting the metabolite term, $\frac{13}{37}$ ($35\%$) triplets returned at least one match for the gene-trait association.
For each of these 216 putative mechanisms, an MVMR approach to compute the direct effect of gene expression on the phenotype was applied (see Methods; Figure 1C; Supplementary file 1f). Regressing direct effects (αd) on total effects (αTP) and accounting for regression dilution bias (see Methods; Figure 2A), it was estimated that $77\%$ [$95\%$ CI: 70–$85\%$] of the transcript effect on the phenotype was direct and thus not mediated by the metabolites (Figure 2B).
**Figure 2.:** *Direct and mediated effects.(A) Graphical representation of the multivariable Mendelian randomization (MVMR) framework for mediation analysis: DNA represents genetic instrumental variables (IVs) chosen to be directly associated with either the exposure (transcript; βeQTL) or the mediator (metabolite; βmQTL) through summary statistics. The effect of these IVs on the outcome (phenotype; βGWAS) originates from genome-wide association studies (GWASs) summary statistics. Total effects αTP of transcripts on phenotypes are estimated by transcriptome-wide Mendelian randomization (TWMR), while direct effects αd are estimated by MVMR. Total effects αTP are assumed to equal the sum of the direct αd and indirect αi (i.e., mediated) effects, the two former being depicted in B. (B) Direct (αd ; y-axis) and total (αTP ; x-axis) effects for the 216 transcript-metabolite-trait causal triplets. The dashed line represents the identity, while the purple line represents the regression line with a shaded 95% confidence interval. Data related to Figure 2 panel B are available in Figure 2—source data 1.
Figure 2—source data 1.Direct and mediated effects.Total (αTP ; transcript-to-phenotype effect) and direct (αd ; direct_effect) effects for the 216 transcript-metabolite-trait causal triplets involving the listed transcript (Gene_ID), metabolite (Shin_ID), and complex phenotype. This file relates to Figure 2B.*
## Molecular mechanisms of genotype-to-phenotype associations
Dissecting causal triplets allows gaining mechanistc insights into biological pathways linking genes to phenotypes. For instance, expression of TMEM258 [MIM: 617615], FADS1 [MIM: 606148], and FADS2 [MIM: 606149], all mapping to a region on chromosome 11 (Figure 3A), were found to influence a total of 17 complex phenotypes through modulation of 1-arachidonoylglycerophosphocholine (LPC(20:4); HMDB0010395; αTMEM258→LPC(20:4)=−1.02; $$P \leq 8.0$$×10–81; αFADS1→LPC(20:4)=−0.39; $$P \leq 4.6$$×10-15; αFADS2→LPC(20:4)=−0.63; $$P \leq 5.1$$×10–62), 1-arachidonoylglycerophosphoethanolamine (LPE(20:4);HMDB0011517; αTMEM258→LPE(20:4)=−0.68; $$P \leq 1.1$$×10–37; αFADS1→LPE(20:4)=−0.30; $$P \leq 1.4$$×10–07; αFADS2→LPE(20:4)=−0.37; $$P \leq 1.2$$×10–18), and 1-arachidonoylglycerophosphoinositol (LPI(20:4); HMDB0061690; αTMEM258→LPI(20:4)=−0.51; $$P \leq 8.2$$×10–18; αFADS2→LPI(20:4)=−0.28; $$P \leq 6.3$$×10–16) levels (Figure 3B–C). These results align with the known pleiotropy of the region (i.e., >6000 associations reported in the GWAS Catalog as of May 2022). Interestingly, involved metabolites are complex lipids synthesized from arachidonic acid, a product of the rate-limiting enzymes encoded by FADS1 and FADS2 (Figure 3B). Recently, polymorphisms affecting the expression of these genes were shown to associate with the levels of over 50 complex lipids, including the ones identified by our study (Reynolds et al., 2020). Overall, this example illustrates how our method can capture meaningful biological associations and shed light on underlying molecular pathways of pleiotropy.
**Figure 3.:** *Molecular pleiotropy at the FADS locus.(A) Genome browser (GRCh37/hg19) view of the genomic region on chromosome 11 encompassing TMEM258, FADS1, and FADS2 (red). (B) Diagram of the mediation signals detected for TMEM258, FADS1, and FADS2. Two of the implicated genes encode enzymes involved in arachidonic synthesis (purple). Involved genes impact 17 phenotypes (pink) through alteration of the levels of three metabolites, 1-arachidonoylglycerophosphocholine (LPC(20:4)), 1-arachidonoylglycerophosphoethanolamine (LPE(20:4)), and 1-arachidonoylglycerophosphoinositol (LPI(20:4)) whose structure is depicted (orange). (C) Network of the 65 transcript-metabolite-trait causal triplets involving TMEM258, FADS1, and FADS2. Nodes represent genes (purple), metabolites (orange), or phenotypes (pink). Edges indicate the direction of the effects estimated through univariable Mendelian randomization. Width of edges is proportional to effect size and color indicates if the effect is positive (red) or negative (blue).*
## Power analysis
Importantly, only $42\%$ ($\frac{90}{216}$) of the causal triplets showed a significant total transcript-to-phenotype effect (i.e., estimated by TWMR), suggesting that the method lacks power under current settings. To characterize the parameter regime where the power to detect indirect effects is larger than it is for total effects, we performed simulations using different settings for the mediated effect. In each scenario we evaluated 500 transcripts and 80 metabolites and varied two parameters characterizing the mediation: Transcripts were simulated with $6\%$ heritability (i.e., median h2 in the eQTLGen data) and a causal effect of 0.035 (i.e., ~$65\%$ of power in TWMR at α=0.05) on a phenotype. Each scenario was simulated 10 times and results were averaged to assess the mean difference in power (see Methods).
Simulations show that with current sample sizes (i.e., NGWAS=300,000, NeQTL=32,000, and NmQTL=8000), when αMP>αTM (i.e., σ<1), TWMR has increased power to detect significant transcript-to-phenotype associations, especially when ρ>0 (i.e., direct and total effect have the same direction (Figure 4A)). However, for all 216 causal triplets, we observed σ>1 (Figure 4—figure supplement 1). Under this condition, and assuming that the total effect of the transcript on the phenotype is dominated by the effect mediated by the metabolite (i.e., ρ<0.5 and ρ>1.5), TWMR had less power than the approach identifying mediators (Figure 4A), confirming that significant associations were missed by TWMR due to power issues related to the proportion of mediated effect.
**Figure 4.:** *Power comparison between transcriptome-wide Mendelian randomization (TWMR) and multivariable Mendelian randomization (MVMR).Heatmap showing the difference in statistical power between TWMR and mediation analysis through MVMR at current (A; N=8000) and realistic future (B; N=90,000) metabolic quantitative trait loci (mQTL) dataset sample sizes. The x-axis shows the proportion (ρ) of direct (αd) to total (αTP) effect (i.e., effect not mediated by the metabolite) ranging from –2 to 2, arrows indicating increasing proportion of direct effect. The y-axis shows the ratio (σ) between the transcript-to-metabolite (αTM) and the metabolite-to-phenotype (αMP) effects, ranging from 0.1 to 10. Red vs. gray indicates higher power for TWMR vs. mediation analysis, respectively, while white represents equal power between the two approaches. Data related to Figure 4 panels A and B are available in Figure 4—source data 1 and Figure 4—source data 2, respectively.
Figure 4—source data 1.Difference in statistical power between transcriptome-wide Mendelian randomization (TWMR) and mediation analysis at N = 8000 metabolic quantitative trait locus (mQTL) dataset sample size.Each cell represents the mean difference in power between TWMR and mediation analysis for a given scenario across 10 simulations. Rows reflect decreasing ratio between transcript-to-metabolite and metabolite-to-phenotype effects from 10 to 0.1 (sigma). Columns reflect increasing proportion of direct to total effect from –2 to 2 (rho). This file relates to Figure 4A.
Figure 4—source data 2.Difference in statistical power between transcriptome-wide Mendelian randomization (TWMR) and mediation analysis at N = 90,000 metabolic quantitative trait locus (mQTL) dataset sample size.Each cell represents the mean difference in power between TWMR and mediation analysis for a given scenario across 10 simulations. Rows reflect decreasing ratio between transcript-to-metabolite and metabolite-to-phenotype effects from 10 to 0.1 (sigma). Columns reflect increasing proportion of direct to total effect from –2 to 2 (rho). This file relates to Figure 4A.*
Repeating the simulations with an mQTL sample size of 90,000, nearing state-of-the-art sample sizes (Lotta et al., 2021), leads to a strong shift in the above-described trends (Figure 4B). Specifically, when the effect of the transcript on the phenotype is dominated by the effect mediated by the metabolite (ρ<0.3 and ρ>1.7), mediation analysis has more power than TWMR when σ>0.2. For larger proportions of direct effect, TWMR has increased power the more σ differs from 1. In line with the increased power of mediation analysis with larger mQTL datasets, the gain in power of mediation analysis over TWMR decreases with decreasing mQTL dataset sample sizes (ranging between $$n = 1000$$ and $$n = 4000$$; Figure 4—figure supplement 2), indicating that our approach is dependent on large sample sizes to reach its full potential.
## Identifying new genotype-to-phenotype associations
The 126 triplets that were not identified through TWMR due to power issues represent putative new causal relations. This is well illustrated by a proof-of concept example involving ANKH [MIM: 605145] and calcium levels, for which 48 publications were identified through automated literature review (Supplementary file 1f). While the TWMR effect of ANKH expression on calcium levels was not significant (αANKH→calcium=-0.02; $$P \leq 0.03$$), ANKH expression decreased citrate levels (αANKH→citrate=-0.30; $$P \leq 2.2$$×10–06), which itself increased serum calcium levels (αcitrate→calcium=0.07; $$P \leq 6.5$$×10–0). Mutations in ANKH have been associated with several rare mineralization disorders [MIM: 123000, 118600] (Williams, 2016) due to the gene encoding a transmembrane protein that channels inorganic pyrophosphate to the extracellular matrix, where at low concentrations it inhibits mineralization (Ho et al., 2000). Recently, a study proposed that ANKH instead exports ATP to the extracellular space (where it is then rapidly converted to inorganic pyrophosphate), along with citrate (Szeri et al., 2020). Citrate has a high binding affinity for calcium and influences its bioavailability by complexing calcium-phosphate during extracellular matrix mineralization and releasing calcium during bone resorption (Granchi et al., 2019). Together, our data support the role of ANKH in calcium homeostasis through regulation of citrate levels, connecting previously established independent links into a causal triad.
In another example, SLC6A12 [MIM: 603080], which encodes the betaine/GABA transporter-1, involved in betaine and GABA uptake (Borden et al., 1995), was identified as a negative regulator of betaine (αSLC6A12→betaine=-0.37; $$P \leq 8.2$$×10–08). While blood betaine levels negatively impacted serum creatinine levels (αbetaine→creatinine=-0.06; $$P \leq 1.7$$×10–07), the effect of SLC6A12 expression on creatinine was not significant (αSLC6A12→creatinine=0.02; $$P \leq 1.5$$×10–03). This observation is particularly interesting given that betaine acts as a protective renal osmolyte whose plasma and kidney tissue concentration were found to be downregulated in renal ischemia/reperfusion injury (Jouret et al., 2016; Wei et al., 2014) and whose urine levels have been proposed as a biomarker for chronic kidney disease progression (Gil et al., 2018). As both renal conditions are commonly monitored through serum creatinine levels, our data support the critical role of osmolyte homeostasis in renal health.
## Discussion
In this study, we combined MR approaches integrating eQTL, mQTL, and GWAS summary statistics to explore the role of the metabolome in mediating the effect of the transcriptome on complex phenotypes. Applied to 28 medically relevant traits, our approach revealed 216 causal transcript-metabolite-phenotype triplets. Our automated literature review indicates that while some detected associations were previously reported, a large fraction, especially among the triplets, appears to be novel. It should be noted that the number of previously reported associations is likely underestimated as our approach does not account for all synonyms of a given feature and requires the terms to appear in the title or abstract of the publication. This makes it more likely for hypothesis-driven studies, inherently biased toward well-studied genes and metabolites, to be identified. Conversely, high-throughput, hypothesis-free studies that report the given association in a supplemental table are likely to be missed. Furthermore, due to its automated nature, our search is context-blind, so that some of the identified studies might report negative results, associations only under specific conditions (e.g., different organisms, experimental settings), or usage of the search term with a different meaning. To attenuate the latter, we also performed manual review of the retained abstracts for transcript-to-phenotype searches. While flawed, this rough estimate of the amount of existing evidence supporting our findings can be interpreted in combination with other lines of evidence. For instance, among the 90 signals that were also identified through TWMR, $93\%$ showed a directionally concordant effect between the transcript-to-phenotype, transcript-to-metabolite, and metabolite-to-phenotype estimates (i.e., sign of product of the transcript-to-metabolite and metabolite-to-phenotype effects agrees with the sign of the transcript-to-phenotype effect). In these situations, dissection of causal effects provides clues as to the molecular mechanism through which involved genes modify complex phenotypes. This information is particularly valuable to identify key molecular mediators of highly pleiotropic genetic regions, such as the TMEM258/FADS1/FADS2 locus (Figure 3). While transcript levels of these genes affected eleven metabolites, three complex lipids were highlighted as strong molecular mediators of the transcript-to-phenotype effects.
Strikingly, $58\%$ of the 216 causal transcript-metabolite-phenotype triplets were missed by TWMR – an approach that only considers gene expression and GWAS data. We highlight two novel but biologically plausible mechanisms linking ANKH to calcium levels through modulation of citrate and SLC6A12 to serum creatinine levels through regulation of the renal osmolyte betaine. Simulation analyses showed that these signals were likely missed by TWMR due to lack of power, as mediation analysis is better suited to detect associations with a low direct to total effect proportion and stronger transcript-to-metabolite than metabolite-to-phenotype effect. Promisingly, our simulations showed that mediation analysis becomes increasingly powerful over a wider range of parameter settings as the sample size of the mediator QTL study increases, highlighting the importance of generating large and publicly available molQTL datasets that can help to unravel functional gene-to-phenotype mechanisms.
As illustrated through the selected examples, a large fraction of detected mediations involves genes encoding metabolic enzymes or transporters/channels, with an enrichment for ‘secondary active transmembrane transporter activity’, for example (GO:0015291; FDR=0.021; background: 7884 genes with ≥3 IVs assessed through TWMR; STRING database). Matching the finding that the most likely effector genes of mQTLs are enriched for pathway-relevant enzymes and transporters (Smith et al., 2022), these results are not surprising given that the proteins encoded by these genes directly interact with metabolites, making it more likely that the effect of changes in their expression is mediated by metabolites. While our method is well suited to detect such effects, interpretation of discovered mediations is limited by the lack of spatial resolution of the mQTL data. Access to metabolite concentrations in different cellular compartments (e.g., extracellular space, cytosol, mitochondrial matrix, etc.) would generate more fine-tuned mechanistic hypotheses that consider the directionality of metabolite fluxes.
The observation that $77\%$ of the transcript’s effect on the phenotype is not mediated by metabolites suggests that either true direct effects are frequent or that other unassessed metabolites or molecular layers (e.g., proteins, post-translational modifications, etc.) play a crucial role in mediation. It is to note that in the presence of unmeasured mediators or measured mediators without genetic instruments, our mediation estimates are lower bounds of the total existing mediation. In addition, unmeasured mediators sharing genetic instruments with the measured ones can modify result interpretation as some of the observed mediators may simply be correlates of the true underlying mediators. While this is a limitation of all MR methods, metabolic networks may harbor particularly large number of genetically correlated metabolite species. Similarly, owing to linkage disequilibrium and regulatory variants affecting multiple genes, transcripts from adjacent genes might appear to be involved in the same signals, as exemplified with the TMEM258/FADS1/FADS2 locus (Figure 3). While literature supports the role of the FADS genes, one cannot exclude a role for TMEM258, nor disentangle the specific function of FADS1 and FADS2. Thanks to the flexibility of the proposed framework, we expect that in the future and upon availability of ever larger and more diverse datasets, our method could be applied to estimate the relative contribution of currently unassessed mediators in translating genotypic cascades.
Another consideration is that complex phenotypes can have a stronger impact on gene expression than the opposite (Porcu et al., 2021b). Due to the lack of genome-wide trans-eQTL association summary statistics, our method does not investigate reverse causality on metabolites and gene expression, nor the role of metabolites as regulators of gene expression. Metabolites might also integrate the effect of several transcripts (i.e., multiple transcripts causally impact the levels of the same metabolite) before affecting complex phenotypes (Supplementary file 1g) or multiple metabolites may jointly mediate the impact of a single transcript. Modelling the latter phenomenon, which is beyond the scope of our current work, requires the development of structural equation models accounting for such effects and will eventually lead to a more comprehensive modelling of causal relations in complex biological networks, nuancing the interpretation of the molecular mechanisms shaping complex traits.
In conclusion, we developed a modular MR framework that has increased power over classical MR approaches to detect causal transcript-to-phenotype relationships when these are mediated by alteration of metabolite levels and is likely to become increasingly powerful upon release of larger molQTL datasets.
## Univariable MR analyses
TWMR and MWMR (Porcu et al., 2019) were used to estimate the causal effects of transcript and metabolite levels (exposure) on various outcomes. For each transcript/metabolite, using inverse-variance weighted (IVW) method for summary statistics (Burgess et al., 2013), the causal effect of the molecular traits on the outcome was defined as[1]α^=(β′C−1β)−1(β′C−1γ) Here, β is a vector of length n containing the standardized effect size of n independent SNPs on the gene/metabolite, derived from eQTL/mQTL studies, with β` being the transpose of β. γ is a vector of length n containing the standardized effect size of each SNP on the outcome. C is the pairwise LD matrix between the n SNPs. The standardized effect sizes for molecular and outcome GWASs were obtained from Z-score of summary statistics standardized by the square root of the sample size to be on the same standard deviation scale.
IVs were selected as autosomal, non-strand ambiguous, independent (r2 <0.01), and significant (PeQTL<1.8×10−05 / PmQTL<1.0×10−07) eQTL/mQTLs available in the UK10K reference panel (Huang et al., 2015) using PLINK (v1.9) (Chang et al., 2015). As retained SNPs are independent, we used the identity matrix to approximate C. SNPs with larger effects on the outcome than on the exposure were removed, as these potentially indicate violation of the MR assumptions (i.e., likely reverse causality and/or confounding).
The variance of α can be calculated approximately by the Delta method[2]var(α^)= (∂α^∂β)2∗var(β^)+(∂α^∂γ)2∗var(γ^)+(∂α^∂β)∗(∂α^∂γ)∗cov(β^,γ^) where cov(β,γ) is 0 if β and γ are estimated from independent samples. The causal effect Z-statistic for transcript/metabolite i was defined as α^iSE(α^i), where SE(α^i)=var(α^)i,i.
The IVW method provides an unbiased estimate under the assumption that all genetic variants are valid IVs, that is, all three MR assumption hold. However, the third assumption (no pleiotropy) is easily violated, leading to inaccurate estimates when horizontal pleiotropy occurs (Verbanck et al., 2018). To test for the presence of pleiotropy, we used Cochran’s Q test (Bowden et al., 2015; Burgess et al., 2017) to assess whether there were significant differences between the MR effects of an instrument (i.e., αβi) and the estimated effect of that instrument on phenotype/metabolite levels (γi). We defined[3]di=γi− αβi and its variance as[4]var(di)=var(γi)+(βi)2∗var(α)+ var(γi)∗(α)2+ var(βi)∗var(α) Next, the deviation of each SNP was tested using the test statistic[5]Ti=di2var(di) ∼ χ12 When $p \leq 0.05$, the SNP with largest | di | was removed and the test was repeated.
## Mediation analysis through MVMR analyses
An MVMR approach was used to dissect the total causal effect of transcript levels on phenotypes (αTP) into a direct (αd) and indirect (αi) effect measured through a metabolite. Through inclusion of a metabolite and its associated genetic variants (r2 <0.01, pmQTL<1 × 10–07), the direct effect of gene expression on a phenotype can be estimated using a multivariable regression model (Burgess et al., 2013) as the first element of[6]α^=(B′C−1B)−1(B′C−1γ) where B is a matrix with two columns containing the standardized effect sizes of n IVs on transcript levels in the first column and on the metabolite levels in the second column, γ is a vector of length n containing the standardized effect size of each SNP on the phenotype, and C is the pairwise LD matrix between the n SNPs.
The proportion of direct effect (ρ) is calculated by regressing direct effects (αd) on total effects (αTP) and then correcting for regression dilution bias:[7]ρcorrected= ρ1−∑(SE(αTP))2∑αTP2
## Omics and traits summary statistics
Expression QTL data originated from the eQTLGen Consortium (Võsa et al., 2021) ($$n = 31$$,684), which includes cis-eQTLs (<1 Mb from gene center, two-cohort filter) for 19,250 transcripts (16,934 with at least one significant cis-eQTL at FDR <0.05 corresponding to $p \leq 1.8$ × 10–05). mQTL data originate from Shin et al., 2014, which used ultra-high performance liquid chromatography-tandem mass spectrometry to measure 486 whole blood metabolites in 7824 European individuals. Association analyses were carried out on ~2.1 million SNPs and are available for 453 metabolites at the Metabolomics GWAS Server (http://metabolomics.helmholtz-muenchen.de/gwas/). Among these metabolites, 242 were manually annotated with the HMDB identifiers (Supplementary file 1a) and used in this study. GWAS summary statistics for 28 outcome traits measured in the UKB originate from the Neale Lab (http://www.nealelab.is/uk-biobank/). Protein interactions with metabolites were downloaded from HMDB v5.0 (https://hmdb.ca/downloads/) and were used to annotate transcript-metabolites associations.
## Automated literature review
An automated literature review of all transcript-metabolite associations (Supplementary file 1b) was conducted in PubMed (September 20, 2022) following the scheme:[8](Gene[Title/Abstract]) AND ((MetHMDB[Title/Abstract]) OR (MetShin[Title/Abstract])) With Gene being the name of the gene whose transcript is involved in the association, MetHMDB the involved metabolite’s common name on HMDB, and MetShin the involved metabolite’s name as reported in Shin et al., 2014. Returned PubMed identifiers were retrieved (Supplementary file 1b).
For transcript-metabolite associations involved in a causal triplet and for which the transcript-metabolites returned at least one publication (Supplementary file 1f), the search was extend by (i) adding an additional search term for the trait (i.e., AND (trait[Title/Abstract])) and (ii) substituting the metabolite term for the trait term. Returned PubMed identifiers were retrieved and corresponding abstracts were manually curated to exclude abstracts in which the search terms were used in a meaning other than the intended one (Supplementary file 1f).
## Simulation analyses
Simulation analyses were conducted to assess the gain in power upon inclusion of metabolomics data in the MR framework. In the simulated scenario, a transcript has an effect on a phenotype mediated by a metabolite. Two parameters were allowed to vary: the proportion (ρ) of direct effect (i.e., effect not mediated by the metabolite) and the ratio (σ) between the effect of the transcript on the metabolite (αTM) and of the metabolite on the phenotype (αMP). Other parameters were fixed, including the heritability of the transcript at hT2=0.06 (corresponding to the median h2 in the eQTLGen data), the number of IVs NIVs at 6 (corresponding to the median number of IVs used in TWMR analyses). Effect sizes βeQTL are from a normal distribution βeQTL∼N(0,hT2NIVs). The causal effect of the transcript on the phenotype (αTP) was fixed to 0.035, which results in ∼$65\%$ power to detect a significant effect with TWMR. These quantities allowed to define βGWAS as βGWAS= αTP* βeQTL+ εP, where εP∼N(0,1NGWAS) with NGWAS=300,000 to reflect the sample size of UKB GWASs. The same vector of βeQTL was used to define βmQTL and estimate the causal effect of the transcript on the metabolite. βmQTL was defined as βmQTL=αTM∗βeQTL+εM, where εM∼N(0,1NmQTL) and NmQTL=8000 to reflect the sample size of the mQTL study used in this work. Simulations were also performed at NmQTL=90,000, to reflect sample size of potential future studies and NmQTL=1000, NmQTL=2000 and NmQTL=4000, to compare the two approaches’ power were the developed framework to be applied on existing smaller mQTL datasets. The total effect αTP can be expressed as αTP=αTM*αMP+ αdirect, where αdirect represents the direct effect of the transcript on the phenotype and αTM*αMP is the indirect effect mediated by the metabolite. Equivalently, αTM*αMP=αTP*(1-ρ) where ρ=αdirectαTP. To assess the ratio between the effect of the transcript on the metabolite and the effect of the metabolite on the phenotype (i.e., σ=αTM/αMP), αTM can be expressed as αTM=αTP∗(1−ρ)∗σ. Similarly, to estimate the effect of the metabolite on the phenotype, a metabolite with heritability hM2=0.04 (corresponding to the median of h2 in the KORA +TwinsUK mQTL data) and NIVs=5 (corresponding to the median number of IVs used in MWMR analyses) is considered. Effect size βmQTL are from a normal distribution βmQTL∼N(0,hM2NIVs). These quantities allowed to define βGWAS as βGWAS= αTP*1-ρ/σ* βmQTL+ εP, where εP∼N(0,1NGWAS). Ranging ρ and σ from –2 to 2 and from 0.1 and 10, respectively, we run each simulation for 500 transcripts measuring 80 metabolites at each run and performed TWMR and MWMR starting from above-described βeQTL, βmQTL, and βGWAS. For each MR analysis the power to detect a significant association as well as the difference in power between TWMR and the mediation analyses (i.e., powerTP- powerTM*powerMP) was calculated. Each specific scenario was repeated 10 times and the average difference in power across simulation was plotted as a heatmap.
## Data and code availability
All data used in this study are publicly available. GWAS summary statistics for outcome traits measured in the UKB originate from the Neale Lab (http://www.nealelab.is/uk-biobank/). eQTL data originated from the eQTLGen Consortium (https://www.eqtlgen.org) and was published in Võsa et al., 2021. mQTL data originate from Shin et al., 2014, and are available at the Metabolomics GWAS Server (http://metabolomics.helmholtz-muenchen.de/gwas/). The HMDB was used to annotate metabolites and the v5.0 release from November 9, 2021, of the ‘All proteins’ file was downloaded to extract transcript-metabolite interactions (https://hmdb.ca/downloads). PubMed was used for the automated literature review (https://pubmed.ncbi.nlm.nih.gov). The UCSC Genome Browser (https://genome.ucsc.edu/) was used to visualize the FADS locus, while the GWAS Catalog was used to assess the number of reported GWAS signals in the region (https://www.ebi.ac.uk/gwas/). The STRING database was used for the enrichment analysis (https://string-db.org/). Produced data is available as Supplementary file 1 and Source Data. Code used to perform analyses is freely available at https://github.com/eleporcu/Gene_Metab_Pheno; (Porcu, 2022 copy archived at swh:1:rev:c6bff8d094e369ff0d399751fc85fcd5ea250134).
## Funding Information
This paper was supported by the following grants:
## Data availability
All data used in this study are publicly available. GWAS summary statistics for outcome traits measured in the UK Biobank originate from the Neale Lab (http://www.nealelab.is/uk-biobank/). eQTL data originated from the eQTLGen Consortium (https://www.eqtlgen.org) and was published in Vosa et al., 2021 [3]. mQTL data originate from Shin et al. 2014 [6], and are available at the Metabolomics GWAS Server (http://metabolomics.helmholtz-muenchen.de/gwas/). The Human Metabolome Database (HMDB) was used to annotate metabolites and the v5.0 release from 2021-11-09 of the "All proteins" file was downloaded to extract transcript-metabolite interactions (https://hmdb.ca/downloads). PubMed was used for the automated literature review (https://pubmed.ncbi.nlm.nih.gov). The UCSC Genome Browser (https://genome.ucsc.edu/) was used to visualize the FADS locus, while the GWAS Catalog was used to assess the number of reported GWAS signals in the region (https://www.ebi.ac.uk/gwas/). The STRING database was used for the enrichment analysis (https://string-db.org/). Produced data is available as Supplementary File 1 and Source Data. Code used to perform analyses is freely available at https://github.com/eleporcu/Gene_Metab_Pheno; (copy archived at swh:1:rev:c6bff8d094e369ff0d399751fc85fcd5ea250134).
## References
1. Borden LA, Smith KE, Gustafson EL, Branchek TA, Weinshank RL. **Cloning and expression of a betaine/GABA transporter from human brain**. *Journal of Neurochemistry* (1995) **64** 977-984. DOI: 10.1046/j.1471-4159.1995.64030977.x
2. Bowden J, Davey Smith G, Burgess S. **Mendelian randomization with invalid instruments: effect estimation and bias detection through egger regression**. *International Journal of Epidemiology* (2015) **44** 512-525. DOI: 10.1093/ije/dyv080
3. Burgess S, Butterworth A, Thompson SG. **Mendelian randomization analysis with multiple genetic variants using summarized data**. *Genetic Epidemiology* (2013) **37** 658-665. DOI: 10.1002/gepi.21758
4. Burgess S, Bowden J, Fall T, Ingelsson E, Thompson SG. **Sensitivity analyses for robust causal inference from mendelian randomization analyses with multiple genetic variants**. *Epidemiology* (2017) **28** 30-42. DOI: 10.1097/EDE.0000000000000559
5. Bycroft C, Freeman C, Petkova D, Band G, Elliott LT, Sharp K, Motyer A, Vukcevic D, Delaneau O, O’Connell J, Cortes A, Welsh S, Young A, Effingham M, McVean G, Leslie S, Allen N, Donnelly P, Marchini J. **The UK Biobank resource with deep phenotyping and genomic data**. *Nature* (2018) **562** 203-209. DOI: 10.1038/s41586-018-0579-z
6. Chang CC, Chow CC, Tellier LC, Vattikuti S, Purcell SM, Lee JJ. **Second-Generation PLINK: rising to the challenge of larger and richer datasets**. *GigaScience* (2015) **4**. DOI: 10.1186/s13742-015-0047-8
7. Giambartolomei C, Zhenli Liu J, Zhang W, Hauberg M, Shi H, Boocock J, Pickrell J, Jaffe AE, Pasaniuc B, Roussos P. **A bayesian framework for multiple trait colocalization from summary association statistics**. *Bioinformatics* (2018) **34** 2538-2545. DOI: 10.1093/bioinformatics/bty147
8. Gil RB, Ortiz A, Sanchez-Niño MD, Markoska K, Schepers E, Vanholder R, Glorieux G, Schmitt-Kopplin P, Heinzmann SS. **Increased urinary osmolyte excretion indicates chronic kidney disease severity and progression rate**. *Nephrology, Dialysis, Transplantation* (2018) **33** 2156-2164. DOI: 10.1093/ndt/gfy020
9. Gleason KJ, Yang F, Pierce BL, He X, Chen LS. **Primo: integration of multiple GWAS and omics QTL summary statistics for elucidation of molecular mechanisms of trait-associated snps and detection of pleiotropy in complex traits**. *Genome Biology* (2020) **21**. DOI: 10.1186/s13059-020-02125-w
10. Granchi D, Baldini N, Ulivieri FM, Caudarella R. **Role of citrate in pathophysiology and medical management of bone diseases**. *Nutrients* (2019) **11**. DOI: 10.3390/nu11112576
11. Gusev A, Ko A, Shi H, Bhatia G, Chung W, Penninx BWJH, Jansen R, de Geus EJC, Boomsma DI, Wright FA, Sullivan PF, Nikkola E, Alvarez M, Civelek M, Lusis AJ, Lehtimäki T, Raitoharju E, Kähönen M, Seppälä I, Raitakari OT, Kuusisto J, Laakso M, Price AL, Pajukanta P, Pasaniuc B. **Integrative approaches for large-scale transcriptome-wide association studies**. *Nature Genetics* (2016) **48** 245-252. DOI: 10.1038/ng.3506
12. Ho AM, Johnson MD, Kingsley DM. **Role of the mouse ANK gene in control of tissue calcification and arthritis**. *Science* (2000) **289** 265-270. DOI: 10.1126/science.289.5477.265
13. Hormozdiari F, Kostem E, Kang EY, Pasaniuc B, Eskin E. **Identifying causal variants at loci with multiple signals of association**. *Genetics* (2014) **198** 497-508. DOI: 10.1534/genetics.114.167908
14. Huang J, Howie B, McCarthy S, Memari Y, Walter K, Min JL, Danecek P, Malerba G, Trabetti E, Zheng HF, Gambaro G, Richards JB, Durbin R, Timpson NJ, Marchini J, Soranzo N. **Improved imputation of low-frequency and rare variants using the UK10K haplotype reference panel**. *Nature Communications* (2015) **6**. DOI: 10.1038/ncomms9111
15. Jouret F, Leenders J, Poma L, Defraigne J-O, Krzesinski J-M, de Tullio P. **Nuclear magnetic resonance metabolomic profiling of mouse kidney, urine and serum following renal ischemia/reperfusion injury**. *PLOS ONE* (2016) **11**. DOI: 10.1371/journal.pone.0163021
16. Julkunen H, Cichońska A, Slagboom PE, Würtz P. **Metabolic biomarker profiling for identification of susceptibility to severe pneumonia and COVID-19 in the general population**. *eLife* (2021) **10**. DOI: 10.7554/eLife.63033
17. Lord J, Jermy B, Green R, Wong A, Xu J, Legido-Quigley C, Dobson R, Richards M, Proitsi P. **Mendelian randomization identifies blood metabolites previously linked to midlife cognition as causal candidates in alzheimer’s disease**. *PNAS* (2021) **118**. DOI: 10.1073/pnas.2009808118
18. Lotta LA, Pietzner M, Stewart ID, Wittemans LBL, Li C, Bonelli R, Raffler J, Biggs EK, Oliver-Williams C, Auyeung VPW, Luan J, Wheeler E, Paige E, Surendran P, Michelotti GA, Scott RA, Burgess S, Zuber V, Sanderson E, Koulman A, Imamura F, Forouhi NG, Khaw K-T, Griffin JL, Wood AM, Kastenmüller G, Danesh J, Butterworth AS, Gribble FM, Reimann F, Bahlo M, Fauman E, Wareham NJ, Langenberg C. **A cross-platform approach identifies genetic regulators of human metabolism and health**. *Nature Genetics* (2021) **53** 54-64. DOI: 10.1038/s41588-020-00751-5
19. MacArthur J, Bowler E, Cerezo M, Gil L, Hall P, Hastings E, Junkins H, McMahon A, Milano A, Morales J, Pendlington ZM, Welter D, Burdett T, Hindorff L, Flicek P, Cunningham F, Parkinson H. **The new NHGRI-EBI catalog of published genome-wide association studies (GWAS catalog)**. *Nucleic Acids Research* (2017) **45** D896-D901. DOI: 10.1093/nar/gkw1133
20. Min JL, Hemani G, Hannon E, Dekkers KF, Castillo-Fernandez J, Luijk R, Carnero-Montoro E, Lawson DJ, Burrows K, Suderman M, Bretherick AD, Richardson TG, Klughammer J, Iotchkova V, Sharp G, Al Khleifat A, Shatunov A, Iacoangeli A, McArdle WL, Ho KM, Kumar A, Söderhäll C, Soriano-Tárraga C, Giralt-Steinhauer E, Kazmi N, Mason D, McRae AF, Corcoran DL, Sugden K, Kasela S, Cardona A, Day FR, Cugliari G, Viberti C, Guarrera S, Lerro M, Gupta R, Bollepalli S, Mandaviya P, Zeng Y, Clarke T-K, Walker RM, Schmoll V, Czamara D, Ruiz-Arenas C, Rezwan FI, Marioni RE, Lin T, Awaloff Y, Germain M, Aïssi D, Zwamborn R, van Eijk K, Dekker A, van Dongen J, Hottenga J-J, Willemsen G, Xu C-J, Barturen G, Català-Moll F, Kerick M, Wang C, Melton P, Elliott HR, Shin J, Bernard M, Yet I, Smart M, Gorrie-Stone T, Shaw C, Al Chalabi A, Ring SM, Pershagen G, Melén E, Jiménez-Conde J, Roquer J, Lawlor DA, Wright J, Martin NG, Montgomery GW, Moffitt TE, Poulton R, Esko T, Milani L, Metspalu A, Perry JRB, Ong KK, Wareham NJ, Matullo G, Sacerdote C, Panico S, Caspi A, Arseneault L, Gagnon F, Ollikainen M, Kaprio J, Felix JF, Rivadeneira F, Tiemeier H, van IJzendoorn MH, Uitterlinden AG, Jaddoe VWV, Haley C, McIntosh AM, Evans KL, Murray A, Räikkönen K, Lahti J, Nohr EA, Sørensen TIA, Hansen T, Morgen CS, Binder EB, Lucae S, Gonzalez JR, Bustamante M, Sunyer J, Holloway JW, Karmaus W, Zhang H, Deary IJ, Wray NR, Starr JM, Beekman M, van Heemst D, Slagboom PE, Morange P-E, Trégouët D-A, Veldink JH, Davies GE, de Geus EJC, Boomsma DI, Vonk JM, Brunekreef B, Koppelman GH, Alarcón-Riquelme ME, Huang R-C, Pennell CE, van Meurs J, Ikram MA, Hughes AD, Tillin T, Chaturvedi N, Pausova Z, Paus T, Spector TD, Kumari M, Schalkwyk LC, Visscher PM, Davey Smith G, Bock C, Gaunt TR, Bell JT, Heijmans BT, Mill J, Relton CL. **Genomic and phenotypic insights from an atlas of genetic effects on DNA methylation**. *Nature Genetics* (2021) **53** 1311-1321. DOI: 10.1038/s41588-021-00923-x
21. Ongen H, Brown AA, Delaneau O, Panousis NI, Nica AC, Dermitzakis ET. **Estimating the causal tissues for complex traits and diseases**. *Nature Genetics* (2017) **49** 1676-1683. DOI: 10.1038/ng.3981
22. Porcu E, Rüeger S, Lepik K, Santoni FA, Reymond A, Kutalik Z. **Mendelian randomization integrating GWAS and eqtl data reveals genetic determinants of complex and clinical traits**. *Nature Communications* (2019) **10**. DOI: 10.1038/s41467-019-10936-0
23. Porcu E, Gilardi F, Darrous L, Yengo L, Bararpour N, Gasser M, Marques-Vidal P, Froguel P, Waeber G, Thomas A, Kutalik Z. **Triangulating evidence from longitudinal and mendelian randomization studies of metabolomic biomarkers for type 2 diabetes**. *Scientific Reports* (2021a) **11**. DOI: 10.1038/s41598-021-85684-7
24. Porcu E, Sadler MC, Lepik K, Auwerx C, Wood AR, Weihs A, Sleiman MSB, Ribeiro DM, Bandinelli S, Tanaka T, Nauck M, Völker U, Delaneau O, Metspalu A, Teumer A, Frayling T, Santoni FA, Reymond A, Kutalik Z. **Differentially expressed genes reflect disease-induced rather than disease-causing changes in the transcriptome**. *Nature Communications* (2021b) **12**. DOI: 10.1038/s41467-021-25805-y
25. Porcu E. *Software Heritage* (2022)
26. Qian L, Fan Y, Gao F, Zhao B, Yan B, Wang W, Yang J, Ma X. **Genetically determined levels of serum metabolites and risk of neuroticism: a Mendelian randomization study**. *The International Journal of Neuropsychopharmacology* (2021) **24** 32-39. DOI: 10.1093/ijnp/pyaa062
27. Reynolds LM, Dutta R, Seeds MC, Lake KN, Hallmark B, Mathias RA, Howard TD, Chilton FH. **Fads genetic and metabolomic analyses identify the ∆5 desaturase (FADS1) step as a critical control point in the formation of biologically important lipids**. *Scientific Reports* (2020) **10**. DOI: 10.1038/s41598-020-71948-1
28. Sabatine MS, Liu E, Morrow DA, Heller E, McCarroll R, Wiegand R, Berriz GF, Roth FP, Gerszten RE. **Metabolomic identification of novel biomarkers of myocardial ischemia**. *Circulation* (2005) **112** 3868-3875. DOI: 10.1161/CIRCULATIONAHA.105.569137
29. Sadler MC, Auwerx C, Porcu E, Kutalik Z. **Quantifying the role of transcript levels in mediating DNA methylation effects on complex traits and diseases**. *Nat Commun* (2022) **13**. DOI: 10.1038/s41467-022-35196-3
30. Shah SH, Bain JR, Muehlbauer MJ, Stevens RD, Crosslin DR, Haynes C, Dungan J, Newby LK, Hauser ER, Ginsburg GS, Newgard CB, Kraus WE. **Association of a peripheral blood metabolic profile with coronary artery disease and risk of subsequent cardiovascular events**. *Circulation. Cardiovascular Genetics* (2010) **3** 207-214. DOI: 10.1161/CIRCGENETICS.109.852814
31. Shin SY, Fauman EB, Petersen AK, Krumsiek J, Santos R, Huang J, Arnold M, Erte I, Forgetta V, Yang TP, Walter K, Menni C, Chen L, Vasquez L, Valdes AM, Hyde CL, Wang V, Ziemek D, Roberts P, Xi L, Grundberg E, Waldenberger M, Richards JB, Mohney RP, Milburn MV, John SL, Trimmer J, Theis FJ, Overington JP, Suhre K, Brosnan MJ, Gieger C, Kastenmüller G, Spector TD, Soranzo N. **An atlas of genetic influences on human blood metabolites**. *Nature Genetics* (2014) **46** 543-550. DOI: 10.1038/ng.2982
32. Smith CJ, Sinnott-Armstrong N, Cichońska A, Julkunen H, Fauman EB, Würtz P, Pritchard JK. **Integrative analysis of metabolite GWAS illuminates the molecular basis of pleiotropy and genetic correlation**. *eLife* (2022) **11**. DOI: 10.7554/eLife.79348
33. Sun BB, Maranville JC, Peters JE, Stacey D, Staley JR, Blackshaw J, Burgess S, Jiang T, Paige E, Surendran P, Oliver-Williams C, Kamat MA, Prins BP, Wilcox SK, Zimmerman ES, Chi A, Bansal N, Spain SL, Wood AM, Morrell NW, Bradley JR, Janjic N, Roberts DJ, Ouwehand WH, Todd JA, Soranzo N, Suhre K, Paul DS, Fox CS, Plenge RM, Danesh J, Runz H, Butterworth AS. **Genomic atlas of the human plasma proteome**. *Nature* (2018) **558** 73-79. DOI: 10.1038/s41586-018-0175-2
34. Szeri F, Lundkvist S, Donnelly S, Engelke UFH, Rhee K, Williams CJ, Sundberg JP, Wevers RA, Tomlinson RE, Jansen RS, van de Wetering K. **The membrane protein ANKH is crucial for bone mechanical performance by mediating cellular export of citrate and ATP**. *PLOS Genetics* (2020) **16**. DOI: 10.1371/journal.pgen.1008884
35. Taylor DL, Jackson AU, Narisu N, Hemani G, Erdos MR, Chines PS, Swift A, Idol J, Didion JP, Welch RP, Kinnunen L, Saramies J, Lakka TA, Laakso M, Tuomilehto J, Parker SCJ, Koistinen HA, Davey Smith G, Boehnke M, Scott LJ, Birney E, Collins FS. **Integrative analysis of gene expression, DNA methylation, physiological traits, and genetic variation in human skeletal muscle**. *PNAS* (2019) **116** 10883-10888. DOI: 10.1073/pnas.1814263116
36. Vandiedonck C. **Genetic association of molecular traits: a help to identify causative variants in complex diseases**. *Clinical Genetics* (2018) **93** 520-532. DOI: 10.1111/cge.13187
37. Verbanck M, Chen C-Y, Neale B, Do R. **Detection of widespread horizontal pleiotropy in causal relationships inferred from Mendelian randomization between complex traits and diseases**. *Nature Genetics* (2018) **50** 693-698. DOI: 10.1038/s41588-018-0099-7
38. Visscher PM, Wray NR, Zhang Q, Sklar P, McCarthy MI, Brown MA, Yang J. **10 years of GWAS discovery: biology, function, and translation**. *American Journal of Human Genetics* (2017) **101** 5-22. DOI: 10.1016/j.ajhg.2017.06.005
39. Võsa U, Claringbould A, Westra H-J, Bonder MJ, Deelen P, Zeng B, Kirsten H, Saha A, Kreuzhuber R, Yazar S, Brugge H, Oelen R, de Vries DH, van der Wijst MGP, Kasela S, Pervjakova N, Alves I, Favé M-J, Agbessi M, Christiansen MW, Jansen R, Seppälä I, Tong L, Teumer A, Schramm K, Hemani G, Verlouw J, Yaghootkar H, Sönmez Flitman R, Brown A, Kukushkina V, Kalnapenkis A, Rüeger S, Porcu E, Kronberg J, Kettunen J, Lee B, Zhang F, Qi T, Hernandez JA, Arindrarto W, Beutner F, Dmitrieva J, Elansary M, Fairfax BP, Georges M, Heijmans BT, Hewitt AW, Kähönen M, Kim Y, Knight JC, Kovacs P, Krohn K, Li S, Loeffler M, Marigorta UM, Mei H, Momozawa Y, Müller-Nurasyid M, Nauck M, Nivard MG, Penninx BWJH, Pritchard JK, Raitakari OT, Rotzschke O, Slagboom EP, Stehouwer CDA, Stumvoll M, Sullivan P, ’t Hoen PAC, Thiery J, Tönjes A, van Dongen J, van Iterson M, Veldink JH, Völker U, Warmerdam R, Wijmenga C, Swertz M, Andiappan A, Montgomery GW, Ripatti S, Perola M, Kutalik Z, Dermitzakis E, Bergmann S, Frayling T, van Meurs J, Prokisch H, Ahsan H, Pierce BL, Lehtimäki T, Boomsma DI, Psaty BM, Gharib SA, Awadalla P, Milani L, Ouwehand WH, Downes K, Stegle O, Battle A, Visscher PM, Yang J, Scholz M, Powell J, Gibson G, Esko T, Franke L. **Large-Scale cis- and trans-eqtl analyses identify thousands of genetic loci and polygenic scores that regulate blood gene expression**. *Nature Genetics* (2021) **53** 1300-1310. DOI: 10.1038/s41588-021-00913-z
40. Wang TJ, Larson MG, Vasan RS, Cheng S, Rhee EP, McCabe E, Lewis GD, Fox CS, Jacques PF, Fernandez C, O’Donnell CJ, Carr SA, Mootha VK, Florez JC, Souza A, Melander O, Clish CB, Gerszten RE. **Metabolite profiles and the risk of developing diabetes**. *Nature Medicine* (2011) **17** 448-453. DOI: 10.1038/nm.2307
41. Wang-Sattler R, Yu Z, Herder C, Messias AC, Floegel A, He Y, Heim K, Campillos M, Holzapfel C, Thorand B, Grallert H, Xu T, Bader E, Huth C, Mittelstrass K, Döring A, Meisinger C, Gieger C, Prehn C, Roemisch-Margl W, Carstensen M, Xie L, Yamanaka-Okumura H, Xing G, Ceglarek U, Thiery J, Giani G, Lickert H, Lin X, Li Y, Boeing H, Joost H-G, de Angelis MH, Rathmann W, Suhre K, Prokisch H, Peters A, Meitinger T, Roden M, Wichmann H-E, Pischon T, Adamski J, Illig T. **Novel biomarkers for pre-diabetes identified by metabolomics**. *Molecular Systems Biology* (2012) **8**. DOI: 10.1038/msb.2012.43
42. Wei Q, Xiao X, Fogle P, Dong Z. **Changes in metabolic profiles during acute kidney injury and recovery following ischemia/reperfusion**. *PLOS ONE* (2014) **9**. DOI: 10.1371/journal.pone.0106647
43. Williams CJ. **The role of ANKH in pathologic mineralization of cartilage**. *Current Opinion in Rheumatology* (2016) **28** 145-151. DOI: 10.1097/BOR.0000000000000247
44. Wishart DS, Guo A, Oler E, Wang F, Anjum A, Peters H, Dizon R, Sayeeda Z, Tian S, Lee BL, Berjanskii M, Mah R, Yamamoto M, Jovel J, Torres-Calzada C, Hiebert-Giesbrecht M, Lui VW, Varshavi D, Varshavi D, Allen D, Arndt D, Khetarpal N, Sivakumaran A, Harford K, Sanford S, Yee K, Cao X, Budinski Z, Liigand J, Zhang L, Zheng J, Mandal R, Karu N, Dambrova M, Schiöth HB, Greiner R, Gautam V. **HMDB 5.0: the human metabolome database for 2022**. *Nucleic Acids Research* (2022) **50** D622-D631. DOI: 10.1093/nar/gkab1062
45. Wu Y, Zeng J, Zhang F, Zhu Z, Qi T, Zheng Z, Lloyd-Jones LR, Marioni RE, Martin NG, Montgomery GW, Deary IJ, Wray NR, Visscher PM, McRae AF, Yang J. **Integrative analysis of omics summary data reveals putative mechanisms underlying complex traits**. *Nature Communications* (2018) **9**. DOI: 10.1038/s41467-018-03371-0
46. Yin X, Bose D, Kwon A, Hanks SC, Jackson AU, Stringham HM, Welch R, Oravilahti A, Fernandes Silva L, Locke AE, Fuchsberger C, Service SK, Erdos MR, Bonnycastle LL, Kuusisto J, Stitziel NO, Hall IM, Morrison J, Ripatti S, Palotie A, Freimer NB, Collins FS, Mohlke KL, Scott LJ, Fauman EB, Burant C, Boehnke M, Laakso M, Wen X. **Integrating transcriptomics, metabolomics, and GWAS helps reveal molecular mechanisms for metabolite levels and disease risk**. *American Journal of Human Genetics* (2022) **109** 1727-1741. DOI: 10.1016/j.ajhg.2022.08.007
47. Zhu Z, Zhang F, Hu H, Bakshi A, Robinson MR, Powell JE, Montgomery GW, Goddard ME, Wray NR, Visscher PM, Yang J. **Integration of summary data from GWAS and eQTL studies predicts complex trait gene targets**. *Nature Genetics* (2016) **48** 481-487. DOI: 10.1038/ng.3538
|
---
title: Orai3 and Orai1 mediate CRAC channel function and metabolic reprogramming in
B cells
authors:
- Scott M Emrich
- Ryan E Yoast
- Xuexin Zhang
- Adam J Fike
- Yin-Hu Wang
- Kristen N Bricker
- Anthony Y Tao
- Ping Xin
- Vonn Walter
- Martin T Johnson
- Trayambak Pathak
- Adam C Straub
- Stefan Feske
- Ziaur SM Rahman
- Mohamed Trebak
journal: eLife
year: 2023
pmcid: PMC9998091
doi: 10.7554/eLife.84708
license: CC BY 4.0
---
# Orai3 and Orai1 mediate CRAC channel function and metabolic reprogramming in B cells
## Abstract
The essential role of store-operated Ca2+ entry (SOCE) through Ca2+ release-activated Ca2+ (CRAC) channels in T cells is well established. In contrast, the contribution of individual Orai isoforms to SOCE and their downstream signaling functions in B cells are poorly understood. Here, we demonstrate changes in the expression of Orai isoforms in response to B cell activation. We show that both Orai3 and Orai1 mediate native CRAC channels in B cells. The combined loss of Orai1 and Orai3, but not Orai3 alone, impairs SOCE, proliferation and survival, nuclear factor of activated T cells (NFAT) activation, mitochondrial respiration, glycolysis, and the metabolic reprogramming of primary B cells in response to antigenic stimulation. Nevertheless, the combined deletion of Orai1 and Orai3 in B cells did not compromise humoral immunity to influenza A virus infection in mice, suggesting that other in vivo co-stimulatory signals can overcome the requirement of BCR-mediated CRAC channel function in B cells. Our results shed important new light on the physiological roles of Orai1 and Orai3 proteins in SOCE and the effector functions of B lymphocytes.
## Introduction
Calcium (Ca2+) is an essential regulator of immune cell function (Prakriya and Lewis, 2015; Trebak and Kinet, 2019). Crosslinking of immunoreceptors like the T cell receptor (TCR) or B cell receptor (BCR) triggers a robust elevation in intracellular Ca2+ concentrations through the release of endoplasmic reticulum (ER) Ca2+ stores and a concomitant influx of Ca2+ from the extracellular space (Baba and Kurosaki, 2016; King and Freedman, 2009). In lymphocytes, Ca2+ entry across the plasma membrane (PM) is predominately achieved through CRAC channels, which constitute the ubiquitous SOCE (Prakriya and Lewis, 2015; Trebak and Kinet, 2019; Trebak and Putney, 2017; Vaeth et al., 2020). Stimulation of immunoreceptors coupled to phospholipase C isoforms results in the production of the secondary messenger inositol-1,4,5-trisphosphate (IP3), which triggers the release of ER Ca2+ through the activation of ER-resident IP3 receptors. A reduction in ER Ca2+ concentrations is sensed by stromal interaction molecule 1 (STIM1) and its homolog STIM2, resulting in a conformational change and clustering in ER-PM junctions where STIM molecules interact with and activate PM hexameric Orai channels (Orai1-3) to mediate SOCE (Lunz et al., 2019). Ca2+ entry through CRAC channels regulates immune cell function through a host of Ca2+-sensitive transcription factors including the NFAT, nuclear factor κB (NF-κB), c-Myc, and mTORC1 (Berry et al., 2018; Trebak and Kinet, 2019; Vaeth and Feske, 2018; Vaeth et al., 2020; Vaeth et al., 2017a). The coordination of these master transcriptional regulators is indispensable for innate and adaptive immune cell effector function including entry into the cell cycle, clonal expansion, cytokine secretion, differentiation, and antibody production (Shaw and Feske, 2012; Trebak and Kinet, 2019; Vaeth et al., 2020).
Orai1 has long been established as a central component of the native CRAC channel in all cell types studied. Although all three Orai isoforms are ubiquitously expressed across tissue types and can form functional CRAC channels when ectopically expressed, only recently have Orai2 and Orai3 emerged as regulators of native CRAC channel function (Emrich et al., 2022b; Tsvilovskyy et al., 2018; Vaeth et al., 2017b; Yoast et al., 2020a). Patients with inherited loss-of-function (LoF) mutations in Orai1 (e.g. R91W mutation) develop a CRAC channelopathy with symptoms including combined immunodeficiency, ectodermal dysplasia, muscular hypotonia, and autoimmunity (Feske et al., 2006; Lian et al., 2018; McCarl et al., 2009). While these patients display relatively normal frequencies of most immune cell populations, they are highly susceptible to reoccurring viral, bacterial, and fungal infections due to defects in T cell expansion, cytokine secretion, and metabolism (Vaeth et al., 2020). There are currently no reported patient mutations in Orai2 or *Orai3* genes and the role these channels play within the immune system had largely been unclear. Interestingly, recent studies utilizing global Orai2 knockout mice have demonstrated that loss of Orai2 leads to enhanced SOCE and corresponding CRAC currents in bone marrow-derived macrophages, dendritic cells, T cells, enamel cells, and mast cells (Eckstein et al., 2019; Tsvilovskyy et al., 2018; Vaeth et al., 2017b), suggesting that Orai2 is a negative regulator of CRAC channel activity. Combined knockout in mice of both Orai1 and Orai2 in T cells led to a near ablation of SOCE and impaired humoral immunity, while ectopic expression of pore-dead mutants of Orai1 (E106Q) or Orai2 (E80Q) into individual Orai1-/- or Orai2-/- T cells blocked native SOCE (Vaeth et al., 2017b). Similarly, the generation of HEK293 cell lines lacking each Orai isoform individually and in combination resulted in altered Ca2+ oscillation profiles, CRAC currents, and NFAT isoform activation (Emrich et al., 2022b; Emrich et al., 2021; Yoast et al., 2020a; Yoast et al., 2020b). These data provide evidence that Orai2 and Orai3 exert negative regulatory effects on native CRAC channels possibly by forming heteromeric CRAC channels with Orai1.
In contrast to T cells, much less is known regarding the role of SOCE in B lymphocytes. Early landmark work investigating B cell-specific STIM knockout mice established that STIM1 mediates the vast majority of SOCE in B cells, while only the combined deletion of both STIM1 and STIM2 resulted in substantial impairments in B cell survival, proliferation, and NFAT-dependent IL-10 secretion (Matsumoto et al., 2011). Unexpectedly, mice with STIM1/STIM2-deficient B cells show normal humoral immune responses to immunization with both T cell-dependent and independent antigens. These and other studies suggest that the severe immunodeficiency observed in CRAC channelopathy patients is due to impaired T cell responses (Matsumoto et al., 2011; Vaeth et al., 2016). Investigation of Orai1-/- mice on the mixed Institute for Cancer Research (ICR) background and Orai1R93W knock-in mice (the equivalent of human R91W mutation) demonstrated that SOCE is significantly attenuated in B cells (Gwack et al., 2008; McCarl et al., 2010). However, SOCE in B cells is only partially reduced with the loss of Orai1, suggesting that Orai2 and/or Orai3 mediate the remaining SOCE in B cell populations. A recent study showed that the deletion of Orai3 in mice does not affect B cell function and humoral immune responses (Wang et al., 2022), suggesting other Orai homologs compensate for the loss of Orai3. Thus, the contribution of each Orai isoform to native CRAC channel function and downstream signaling in B cells has remained obscure.
In this study, we investigated the contributions of Orai isoforms to SOCE and its downstream signaling in B cells through multiple CRISPR/Cas9 knockout B cell lines and novel B cell-specific Orai knockout mice. Our findings demonstrate that the expression of each Orai isoform is dynamically regulated in response to B cell activation and that the magnitude of SOCE is unique among effector B cell populations. We show that the deletion of Orai1 alone does not alter BCR-evoked cytosolic Ca2+ oscillations, proliferation, and development. Unexpectedly, we show that Orai3 is involved in regulating B cell SOCE and downstream signaling functions. Deletion of both Orai1 and Orai3 strongly inhibits SOCE and hampers mitochondrial metabolism in B cells. Transcriptome and metabolomic analysis uncovered key signaling pathways that are regulated by SOCE and the phosphatase calcineurin for the efficient transition of B cells from a quiescent to a metabolically active state. Surprisingly, humoral immunity to influenza A virus infection of mice with B cell-specific deletion of both Orai1 and Orai3 was unaltered, suggesting that alternative Ca2+-independent signaling pathways activated in vivo through co-stimulatory receptors can overcome the loss of CRAC channel activity in B cells. Our data elucidate the role of SOCE in B cell functions and show that CRAC channel function in B lymphocytes is mediated by both Orai1 and Orai3. This knowledge is important for the potential targeting of CRAC channels in specific lymphocyte subsets in immune and inflammatory diseases.
## Orai channel isoforms are dynamically regulated in response to B cell activation
Throughout their lifespan, B lymphocytes must undergo dramatic periods of metabolic adaptation whereby naïve, metabolically quiescent cells prepare to clonally expand and differentiate into effector populations (Akkaya and Pierce, 2019; Boothby and Rickert, 2017). Induction of these diverse metabolic programs is driven by downstream BCR-mediated Ca2+ signals in combination with Ca2+-independent costimulatory signals including CD40 activation and/or stimulation with various toll-like receptor (TLR) ligands (Akkaya et al., 2018; Baba and Kurosaki, 2016; Berry et al., 2020). Thus, using Seahorse assays we first evaluated how activation of wild-type mouse primary splenic B cells (isolated as described in methods) regulates oxygen consumption rates (OCR) and extracellular acidification rates (ECAR) as indicators of oxidative phosphorylation (OXPHOS) and glycolysis, respectively (Figure 1A and B). Naïve primary B cells from mouse spleens were isolated by negative selection and stimulated for 24 hr under five different conditions as follows: [1] control unstimulated cells (Unstim), [2] stimulation with anti-IgM antibodies (IgM) to activate the BCR, [3] stimulation with anti-CD40 (CD40), [4] stimulation with lipopolysaccharides (LPS), and [5] co-stimulation with anti-IgM and anti-CD40. B cell stimulation for 24 hr led to a robust increase in basal OCR, except for B cells stimulated with anti-CD40 alone where the increase in basal OCR was not statistically significant (Figure 1A and C). Figure 1B shows that B cells stimulated with either anti-IgM, anti-IgM +anti-CD40, or LPS become highly energetic and upregulate both OXPHOS and glycolytic pathways. However, B cells stimulated with anti-CD40 alone remain metabolically quiescent like control non-stimulated B cells (Figure 1B). We used the protonophore trifluoromethoxy carbonylcyanide phenylhydrazone (FCCP) to dissipate the mitochondrial membrane potential and calculate the maximal respiratory capacity of B cells. Consistent with results obtained for basal OCR, stimulated B cells showed a robust increase in maximal respiratory capacity except for cells stimulated with anti-CD40 alone (Figure 1D). Both total mitochondrial content and membrane potential measured with MitoTracker Green and TMRE staining, respectively, were substantially increased following anti-IgM, anti-IgM +anti-CD40, or LPS stimulation while both parameters were only slightly increased with stimulation with anti-CD40 alone (Figure 1E and F). While previous studies suggested that Orai1 regulates the majority of SOCE in primary B cells (Gwack et al., 2008; McCarl et al., 2010), the expression of different Orai isoforms both at rest and upon B cell activation is unknown. Utilizing the same five experimental conditions described above, we analyzed the mRNA expression of Orai1, Orai2, and Orai3 following 24 hr of stimulation. Stimulation with either anti-IgM or anti-IgM +anti-CD40 dramatically increased Orai1 expression, while the magnitude of this change following anti-CD40 or LPS stimulation was smaller and did not reach statistical significance (Figure 1G). Interestingly, stimulation conditions that strongly increased mitochondrial respiration (anti-IgM, anti-IgM +anti-CD40, LPS) all significantly reduced expression of Orai2 while stimulation with anti-CD40 alone had no significant effect on Orai2 expression (Figure 1H). The expression of Orai3 has significantly increased only with anti-IgM +anti-CD40 or LPS stimulation, but not with anti-IgM or anti-CD40 alone (Figure 1I).
**Figure 1.:** *B cell activation dynamically regulates Orai channel expression.(A) Measurement of oxygen consumption rate (OCR) in primary B lymphocytes following 24 hr stimulation with anti-IgM (20 μg/mL), anti-CD40 (10 μg/mL), anti-IgM +anti-CD40, or LPS (10 μg/mL) using the Seahorse Mito Stress Test (n=3 biological replicates). (B) Energy map of maximal OCR and extracellular acidification rate (ECAR) following the addition of the protonophore carbonyl cyanide p-trifluoromethoxyphenylhydrazone (FCCP). (C, D) Quantification of basal (C) and maximal (D) respiration from Seahorse traces in (A) (One-way ANOVA with multiple comparisons to Unstimulated). (E, F) Measurement of (E) total mitochondrial content with the fluorescent dye MitoTracker Green and (F) mitochondrial membrane potential with TMRE following 24 hr stimulation (n=3 biological replicates; One-way ANOVA with multiple comparisons to Unstimulated). (G–I) Quantitative RT-PCR of (G) Orai1, (H) Orai2, and (I) Orai3 mRNA following 24 hr of stimulation with the stimuli indicated (n=3 biological replicates; one-way ANOVA with multiple comparisons to Unstimulated). All scatter plots and Seahorse traces are presented as mean ± SEM. For all figures, *p<0.05; **p<0.01; ***p<0.001; ****p<0.0001; ns, not significant.
Figure 1—source data 1.Source data for Figure 1.*
## Both Orai3 and Orai1 mediate SOCE in A20 B lymphoblasts
Robust activation of primary B cells with anti-IgM/anti-CD40 co-stimulation resulted in upregulation of both Orai1 and Orai3 mRNA with downregulation of Orai2 (Figure 1G–I). We, therefore, reasoned that B cells may sustain long-term cytosolic Ca2+ signals through Orai1 and/or Orai3, while downregulation of Orai2 might relieve the inhibition of native CRAC channels, as recently demonstrated for activated T cells (Vaeth et al., 2017b). To gain insights into the molecular composition of CRAC channels in B cells, we first developed an in vitro system utilizing the mouse A20 B lymphoblast cell line to generate single and double Orai1 and Orai3 knockout cell lines with CRISPR/Cas9 technology. Two guide RNA (gRNA) sequences were utilized to cut at the beginning and end of the coding region of mouse Orai1 and Orai3 (mOrai1 and mOrai3), effectively excising the entirety of the genomic DNA for each *Orai* gene (Figure 2A). *We* generated multiple A20 clones that were lacking Orai1 and Orai3 individually and in combination and these knockout clones were validated through genomic DNA sequencing (*Source data* 1) and the absence of Orai1 or Orai3 transcripts by qPCR (Figure 2B–D). We measured SOCE in A20 cells after passive store depletion with thapsigargin, an inhibitor of the sarcoplasmic/endoplasmic reticulum ATPase (SERCA). A20 cells lacking Orai1 demonstrated a large reduction in maximal SOCE by ~$62\%$, while in cells lacking Orai3 SOCE was reduced by ~$28\%$ (Figure 2E and F). Importantly, the combined deletion of both Orai1 and Orai3 caused a near abrogation of SOCE by ~$91\%$ (Figure 2E and F). Furthermore, measurements of cytosolic Ca2+ oscillations induced by anti-IgG stimulation demonstrated that the oscillation frequency was substantially reduced with combined Orai1/Orai3 knockout, while only slightly reduced with the loss of either Orai1 or Orai3 individually (Figure 2G–K). These data suggest that both Orai3 and Orai1 were involved in optimal CRAC channel function in A20 B lymphoblasts. These data are consistent with the function of Orai channels in other cell types. Indeed, we recently utilized a series of single, double, and triple Orai CRISPR/Cas9 knockout HEK293 cell lines and showed that SOCE, but not Orai1, is required for agonist-induced Ca2+ oscillations (Yoast et al., 2020b). Under conditions of physiological agonist stimulation that causes modest ER store depletion while eliciting Ca2+ oscillations, Orai2 and Orai3 were sufficient to maintain cytosolic Ca2+ oscillations, while having relatively minor contributions to SOCE induced by maximal store depletion (Yoast et al., 2020b).
**Figure 2.:** *Orai1 and Orai3 mediate the bulk of store-operated Ca2+ entry (SOCE) in A20 B lymphoblasts.(A) Cartoon schematic of the two gRNA CRISPR strategies we used to excise mouse Orai1 and Orai3 genes. (B–D) Quantitative RT-PCR of (B) Orai1, (C) Orai2, and (D) Orai3 mRNA in A20 Orai CRISPR clones (n=3 biological replicates). (E) Measurement of SOCE with Fura2 upon store depletion with 2 µM thapsigargin in 0 mM Ca2+ followed by re-addition of 2 mM Ca2+ to the external bath solution. (F) Quantification of peak SOCE in (E) (from left to right n=99, 100, 89, and 98 cells; Kruskal-Wallis test with multiple comparisons to WT A20). (G–J) Representative Ca2+ oscillation traces from 5 cells/condition measured with Fura2 upon stimulation with 10 μg/mL anti-IgG antibodies at 60 s (indicated by arrows) in the presence of 2 mM external Ca2+. (K) Quantification of total oscillations in 9 min from (G–J) (from left to right n=76, 79, 79, and 78 cells; Kruskal-Wallis test with multiple comparisons to WT A20). All scatter plots are presented as mean ± SEM. For all figures, *p<0.05; **p<0.01; ****p<0.0001; ns, not significant.
Figure 2—source data 1.Source data for Figure 2.*
## Orai1 is dispensable for cytosolic Ca2+ oscillations in primary B cells
To determine the contribution of Orai1-mediated Ca2+ signals to primary B cell function, we generated B cell-specific Orai1 knockout (Orai1fl/fl Mb1-Cre/+) mice (Figure 3). Compared to Orai1fl/fl controls, the average surface expression of Orai1 was significantly reduced on B220+ B cells from Orai1fl/fl Mb1-Cre/+ mice by~$70\%$ (Figure 3A and B). The residual signal of ~$30\%$ above FMO (Figure 3A) is likely the result of the non-specific binding of the Orai1 antibody by its variable region to other Orai homologs (Orai2, Orai3), which partially share the peptide epitope recognized by the anti-Orai1 antibody. This residual Orai1 antibody signal is consistent with previous studies showing that Cre-mediated deletion of Orai1fl/fl alleles is very efficient in other cell types including T cells (Kaufmann et al., 2016), and neuronal progenitor cells (Somasundaram et al., 2014). Indeed, other lines of evidence indicate that Orai1 is deleted in B cells of Orai1fl/fl Mb1-Cre/+ mice: our qPCR data show that B cells from Orai1fl/fl Mb1-Cre/+ mice showed a near abrogation of Orai1 mRNA (to ~$3\%$ of control) compared to Mb1-Cre/+ control mice (Figure 3C) without any significant change in Orai2 (Figure 3D) or Orai3 (Figure 3E) mRNA expression. Further, Ca2+ measurements demonstrated that SOCE is reduced by ~$69\%$ in B cells isolated from Orai1fl/fl Mb1-Cre/+ mice (Figure 3F and G). We measured cytosolic Ca2+ oscillations in response to anti-IgM stimulation in the presence of 1 mM extracellular Ca2+. In agreement with previous reports and consistent with our data with A20 B lymphoblasts (Figure 2G and H), Orai1-deficient B cells display a comparable frequency of agonist-induced Ca2+ oscillations as control B cells (Figure 3H–J). Thus, while Orai1 knockout significantly decreased SOCE in primary B cells, it is dispensable for maintaining Ca2+ oscillations in response to physiological agonist stimulation. In control Mb1-Cre/+ mice, Orai1 expression was comparable between B220+ B cells and CD8+ T cells, with CD4+ T cells showing a slightly lower signal (Figure 3—figure supplement 1A, B). After 24 hour stimulation of control B cells under the five conditions described above, Orai1-deficient B cells showed no apparent defects in activation as the expression of MHC-II was comparable to control B cells (Figure 3—figure supplement 1C–E), while expression of CD86 was slightly reduced with anti-IgM stimulation or anti-IgM +anti-CD40 co-stimulation (Figure 3—figure supplement 1F–H).
**Figure 3.:** *Orai1 is dispensable for BCR-induced Ca2+ oscillations in primary B cells.(A) Representative flow cytometry histogram of B cells isolated from Orai1fl/fl and Orai1fl/fl Mb1-Cre/+ mice. Splenocytes from naïve Orai1fl/fl and Orai1fl/fl Mb1-Cre/+ mice were fixed, permeabilized, and stained with rabbit anti-Orai1 polyclonal antibody (YZ6856, epitope: human ORAI1#275–291 intra-cellular C-terminal, cross-reacts with the mouse). The numbers inside the panel represent the mean fluorescence intensity (MFI) for the Orai1 antibody staining for each sample. (B) Quantification of Orai1 MFI minus fluorescence minus one (FMO) in B cells is shown. (n=4 biological replicates; unpaired T-test). (C–E) Quantitative RT-PCR of (C) Orai1, (D) Orai2, and (E) Orai3 mRNA in isolated B cells (n=6 biological replicates for each; Mann-Whitney test). (F) Measurement of SOCE with Fura2 upon store depletion with 2 µM thapsigargin in 0 mM Ca2+ followed by re-addition of 1 mM Ca2+ to the external bath solution. (G) Quantification of peak store-operated Ca2+ entry (SOCE) in (F) n=169 and 178 cells; Mann-Whitney test. (H–I) Representative Ca2+ oscillation traces from 5 cells/condition measured with Fura2 upon stimulation with 20 μg/mL anti-IgM antibodies at 1 min (indicated by arrows) in the presence of 1 mM external Ca2+. (J) Quantification of total oscillations in 9 min from (I, J) (n=97 and 111 cells; Mann-Whitney test). All scatter plots are presented as mean ± SEM. For all figures, **p<0.01; ****p<0.0001; ns, not significant.
Figure 3—source data 1.Source Data for Figure 3.*
## Loss of Orai1 does not alter B cell development
While patients and mouse models with loss of function (LoF) mutations in Orai1 present with severe immunodeficiency due to impaired T cell function, the development of most immune cell populations is largely unaltered, suggesting SOCE is dispensable for initial immune cell selection and development (Gwack et al., 2008; Lacruz and Feske, 2015; McCarl et al., 2010). We evaluated B cell development within the bone marrow and spleen of Orai1fl/fl Mb1-Cre/+ mice (Figure 4). In agreement with previous reports that investigated B cell development in global Orai1 deficient mice, the development of early B cell progenitors in fractions A-F within the bone marrow were unaltered in Orai1fl/fl Mb1-Cre/+ mice (Figure 4A–C). Similarly, we observed no significant differences in peripheral transitional type 1 (T1), T2, and T3 immature B cells in the spleen (Figure 4D and F). Mature B cell populations in the spleen are predominately follicular B cells with a smaller fraction of marginal zone B cells, and these ratios were largely comparable between control and Orai1fl/fl Mb1-Cre/+ mice (Figure 4E and F).
**Figure 4.:** *Orai1 is dispensable for B cell development.(A–B) Flow cytometric analysis of bone marrow populations for B cell fractions A (B220+CD43+HSA−BP-1−), B (B220+CD43+HSA+BP-1−), C (B220+CD43+HSA+BP-1+), D (B220+CD43−IgM−CD93+), E (B220+CD43−IgM+CD93+), and F (B220+CD43−IgM+CD93−). (C) Quantification of bone marrow populations in (A, B) (n=7 and six biological replicates; Mann-Whitney test). (D) Flow cytometric analysis of isolated populations in the spleen for B cell developmental stages T1 (B220+AA4.1+CD23−IgM+), T2 (B220+AA4.1+CD23+IgM+), and T3 (B220+AA4.1+CD23+IgM−). (E) Flow cytometric analysis of isolated populations in the spleen for marginal zone (MZ) B cells (B220+CD93−CD23−IgM+) and follicular (FO) B cells (B220+CD93−CD23+IgM+). (F) Quantification of splenic populations in (D, E) (n=7 and six biological replicates; Mann-Whitney test). All scatter plots are presented as mean ± SEM.
Figure 4—source data 1.Source data for Figure 4.*
## Both Orai3 and Orai1 synergistically contribute to SOCE in primary B cells
To determine whether Orai3 regulates Ca2+ signals and function of primary B cells, we generated B cell-specific Orai3 knockout mice (Orai3fl/fl Mb1-Cre/+) and B cell-specific double Orai1/Orai3 knockout mice (Orai1/Orai3fl/fl Mb1-Cre/+). Primary B cells isolated from spleens of Orai1, Orai3, and Orai1/Orai3 knockout mice showed near complete ablation of their respective Orai isoform mRNA (to ~0.3–$4\%$ of control) with no compensatory changes in Orai2 mRNA expression (Figure 5A). As documented above (Figure 3F and G), depletion of ER Ca2+ stores with thapsigargin demonstrated that SOCE was significantly reduced in Orai1 knockout B cells (by ~$69\%$), and this remaining Ca2+ entry was further reduced in Orai1/Orai3 double knockout B cells (by ~$83\%$; Figure 5B and C). However, SOCE in single Orai3 knockout B cells was ~$102\%$ of control, which was not significantly different from control B cells.
**Figure 5.:** *Orai1 and Orai3 synergistically mediate store-operated Ca2+ entry (SOCE) in primary B cells.(A) Quantitative RT-PCR of Orai1, Orai2, and Orai3 mRNA in negatively isolated B cells from B cell-specific Orai knockout mice (n=3 biological replicates per genotype). (B) Measurement of SOCE in naïve B cells with Fura2 upon store depletion with 2 µM thapsigargin in 0 mM Ca2+ followed by re-addition of 1 mM Ca2+ to the external bath solution. Subsequently, SOCE was inhibited with the addition of 50 µM 2-APB at 13 min followed by 5 µM Gd3+ at 18 min. (C) Quantification of peak SOCE in (B) (from left to right n=200, 200, 199, and 149 cells; Kruskal-Wallis test with multiple comparisons to Mb1-Cre/+). (D) Quantification of the rate of 2-APB inhibition from 13 to 18 min. (E) Quantitative RT-PCR of Orai1, Orai2, and Orai3 mRNA in negatively isolated B cells from wild-type and Orai2-/- mice (n=3 and six biological replicates). (F) Measurement of SOCE in naïve B cells with Fura2 from wild-type and Orai2-/- mice. (G) Quantification of peak SOCE in (F) (n=147 and 240 cells; Mann-Whitney test). All scatter plots are presented as mean ± SEM. For all figures, ***p<0.001; ****p<0.0001; ns, not significant.
Figure 5—source data 1.Source data for Figure 5.*
Orai channel isoforms demonstrate distinct pharmacological profiles and sensitivities to various CRAC channel modifiers (Bird and Putney, 2018; Zhang et al., 2020). One of the most extensively utilized CRAC channel modifiers is 2-aminoethyl diphenyl borate (2-APB) (Prakriya and Lewis, 2001), which at high (25–50 µM) concentrations strongly inhibits Orai1, partially inhibits Orai2, and potentiates Orai3 channel activity (DeHaven et al., 2008; Zhang et al., 2008; Zhang et al., 2020). After allowing Ca2+ entry to the plateau, 50 µM 2-APB was added, followed by the addition of 5 µM gadolinium (Gd3+) which potently blocks all Orai isoforms (Yoast et al., 2020b; Zhang et al., 2020). In wild-type B cells, 2-APB led to a gradual inhibition of SOCE over the course of 5 min, which completely returned to baseline following the addition of Gd3+. Interestingly, the remaining SOCE in Orai1 knockout B cells showed essentially no inhibition by 2-APB but was strongly inhibited by Gd3+. The lack of effect of 2-APB is likely due to the residual SOCE mediated by the remaining Orai3 isoform, which is resistant to inhibition by 2-APB. Furthermore, SOCE in Orai3 knockout B cells was inhibited at a significantly faster rate by 2-APB compared to wild-type B cells and was not further inhibited by Gd3+ (Figure 5B and D). The small amount of SOCE remaining in Orai1/Orai3 double knockout cells was reduced to baseline following the addition of 2-APB. This remaining SOCE in Orai1/Orai3 double knockout B cells prompted us to measure SOCE in B cells isolated from global Orai2 knockout mice (Orai2-/-). B cells from Orai2-/- mice showed complete loss of Orai2 mRNA with no apparent compensation in Orai1 or Orai3 mRNA (Figure 5E). SOCE stimulated by thapsigargin in B cells from Orai2-/- mice was enhanced by comparison to B cells from wild-type littermate controls (Figure 5F and G), suggesting that, as was shown in T cells (Vaeth et al., 2017b), Orai2 is also a negative regulator of SOCE in B cells. Thus, these experiments indicate that the remaining SOCE in B cells from Orai3/Orai1 double knockout mice is most likely mediated by Orai2.
Given the modulation of Orai channel isoforms in response to B cell activation, we performed Ca2+ imaging recordings, similar to those in Figure 5, on primary B cells that were first activated for 48 hr with anti-IgM +anti-CD40 (Figure 5—figure supplement 1A–C). While similar trends were observed in experiments with naïve primary B cells, several exceptions were notable. SOCE in Orai3 knockout B cells was reduced compared to control cells, like data in A20 Orai knockout cell lines (Figure 5—figure supplement 1A, B). Differences in inhibition by 50 µM 2-APB also became less pronounced between control and Orai3 knockout cells (Figure 5—figure supplement 1A–C). Collectively, these data demonstrate that Orai3 contributes to SOCE in activated B cells.
## SOCE is an essential regulator of NFAT activation in B cells
Ca2+ entry through CRAC channels is critical for the activation of multiple NFAT isoforms (Prakriya and Lewis, 2015; Trebak and Kinet, 2019; Vaeth and Feske, 2018). NFAT nuclear translocation mostly requires Ca2+ entry through native Orai1, while native Orai2 and Orai3 caused marginal NFAT activation in cells lacking Orai1 despite mediating a Ca2+ signal (Yoast et al., 2020b). To determine the role of Orai1 in regulating NFAT1 activation in B cells, primary B cells from control Mb1-Cre/+ and Orai1fl/fl Mb1-Cre/+ mice were stimulated with anti-IgM antibodies and native NFAT1 nuclear translocation was evaluated using ImageStream analysis (Figure 6A–C). In unstimulated B cells, colocalization of endogenous NFAT1 with nuclear DAPI staining was relatively low, and this colocalization increased three-fold following anti-IgM stimulation (Figure 6A–C). However, this anti-IgM mediated NFAT1 translocation was significantly reduced in B cells from Orai1fl/fl Mb1-Cre/+ mice (Figure 6B and C).
**Figure 6.:** *Orai1 is a regulator of nuclear factor of activated T cells (NFAT) activation in naïve and activated B cells.(A) Representative Imagestream images following intracellular staining for NFAT1 and DAPI in naïve B cells from Mb1-Cre/+ mice before and after 20 μg/mL anti-IgM stimulation for 15 min. Merge image indicates similarity score co-localization between NFAT1/DAPI. (B) Histograms of NFAT1/DAPI similarity scores before (black trace) and after (red trace) anti-IgM stimulation in naïve B cells from Mb1-Cre/+ and Orai1fl/fl Mb1-Cre/+ mice. (C) Quantification of similarity scores following anti-IgM stimulation in (B) (n=3 biological replicates for each; unpaired T-test). (D) Western blot analysis of NFAT1 and α-tubulin in naïve B cells isolated from Mb1-Cre/+ and Orai1fl/fl Mb1-Cre/+ mice. B cells were left unstimulated or treated with 2 µM thapsigargin (Tg) for 15 min before harvesting. (E) Quantification of NFAT1 dephosphorylation in (D) (n=4 biological replicates for each; Mann-Whitney test). (F) Western blot analysis of NFAT1 and α-tubulin in B cells stimulated for 48 hr with anti-IgM +anti-CD40. (G) Quantification of NFAT1 dephosphorylation in (F) (n=3 biological replicates for each; Mann-Whitney test). All scatter plots are presented as mean ± SEM. For all figures, *p<0.05; **p<0.01; ****p<0.0001.
Figure 6—source data 1.Source data for Figure 6 including labeled blots for NFAT1 and α-tubulin from Figure 6D and Figure 6F.
Figure 6—source data 2.Panel D and F-Raw unedited uncropped blots for NFAT1 and α-tubulin from Figure 6.*
We used another complimentary biochemical protocol to assess the nuclear translocation of native NFAT1 in response to stimulation with thapsigargin. After B cell stimulation, proteins were harvested and processed for Western blotting using a specific NFAT1 antibody (Figure 6D–G). In unstimulated samples, NFAT1 appears as a single band corresponding to its highly phosphorylated, non-activated state (Figure 6D and F). Stimulation of naïve B cells with thapsigargin, which completely empties ER stores and maximally activates SOCE, led to a complete shift of the phosphorylated single band into lower molecular weight bands (Figure 6D and E, left) and this shift in NFAT1 molecular weight was inhibited by $43.6\%$ in Orai1-deficient B cells (Figure 6D and E, right). Given the dynamic expression of each Orai isoform following B cell activation (Figure 1G–I), we also evaluated NFAT1 activation in B cells stimulated for 48 hr with anti-IgM +anti-CD40. While thapsigargin stimulation also led to a complete shift in NFAT1 molecular weight in activated control B cells, this shift was reduced by $78\%$ in activated B cells from Orai1fl/fl Mb1-Cre/+ mice (Figure 6F and G). Similar results were obtained when we analyzed NFAT2 isoform dephosphorylation in naïve B cells (Figure 6—figure supplement 1A). Furthermore, when naïve B cells were stimulated with anti-IgM (instead of thapsigargin), we observed similar trends of NFAT1 and NFAT2 dephosphorylation, although this dephosphorylation was not as robust as with thapsigargin (Figure 6—figure supplement 1A). Nevertheless, this anti-IgM mediated dephosphorylation of NFAT1 and NFAT2 was inhibited in B cells from Orai1fl/fl Mb1-Cre/+ mice (Figure 6—figure supplement 1A). Our findings that B cells from Orai1fl/fl Mb1-Cre/+ mice have maintained BCR-induced Ca2+ oscillations with defects in NFAT activation is in agreement with previous studies in HEK293 cells showing that NFAT activation depends mostly on native Orai1 (Yoast et al., 2020b). In these studies, intact Ca2+ oscillations in Orai1 knockout cells were mediated by Orai$\frac{2}{3}$, yet NFAT nuclear translocation was largely reduced (Yoast et al., 2020b).
We also evaluated NFAT1 activation in response to thapsigargin in both naïve B cells (Figure 6—figure supplement 1B) and activated B cells (48 hr with anti-IgM +anti-CD40; Figure 6—figure supplement 1C) isolated from either Orai3fl/fl Mb1-Cre/+ or Orai1/Orai3fl/fl Mb1-Cre/+ mice. Loss of Orai3 alone from naïve B cells had no effect on NFAT1 activation, while naïve B cells isolated from double Orai1/Orai3 knockout cells showed reductions in NFAT1 activation comparable to B cell isolated from single Orai1 knockout mice (Figure 6—figure supplement 1B). Importantly, this impairment of NFAT1 activation was further exacerbated in activated B cells isolated from Orai1 knockout and Orai1/Orai3 double knockout mice (Figure 6—figure supplement 1C). Collectively, these data reveal that Orai1 plays a more prominent role in the activation of NFAT isoforms in activated B cells compared to naïve cells.
## The combined deletion of Orai3 and Orai1 inhibits B cell proliferation
BCR-induced Ca2+ signals that are sustained through SOCE are critical for the activation of gene programs that regulate proliferation and apoptosis (Berry et al., 2020; Matsumoto et al., 2011). Interestingly, suboptimal proliferation and survival of B cells in response to BCR stimulation can largely be rescued through the addition of secondary co-stimulatory signals (e.g. CD40 or TLR stimulation) (Berry et al., 2020; Matsumoto et al., 2011; Tang et al., 2017). To understand how Ca2+ signals downstream of Orai isoforms regulate B cell expansion, primary B cells were labeled with carboxyfluorescein diacetate succinimidyl ester (CFSE) and cell divisions tracked in response to multiple conditions of stimulation (Figure 7A). BCR stimulation with anti-IgM alone resulted in multiple rounds of cell division in most wild-type cells ($87.7\%$) with $28.6\%$ of viable cells at 72 hr post-stimulation (Figure 7B and C, black). B cell viability at 72 hr was dramatically increased to $67.4\%$ when cells were co-stimulated with anti-IgM +anti-CD40, with an increase in the number of cells undergoing several cycles of cell division. Following anti-IgM stimulation Orai1-deficient B cells showed a similar percentage of proliferating cells ($83.1\%$) as controls, with a moderate reduction in cell viability to $20.9\%$ (Figure 7B and C, red). While anti-IgM mediated proliferation of Orai3-deficient B cells was comparable to controls ($90.3\%$), their viability was higher, at $34.8\%$ (Figure 7B and C, blue). The combined deletion of Orai1 and Orai3 resulted in a substantial reduction in the percentage of both viable ($10\%$) and proliferating ($60.8\%$) cells in response to anti-IgM stimulation (Figure 7B and C, teal). These defects in survival and proliferation of Orai1/Orai3-deficient B cells were partially rescued when cells were co-stimulated with anti-IgM +anti-CD40, albeit to a lesser extent than control and single Orai knockout B cells. Similar percentages of viability and proliferation were observed in all experimental groups when B cells were stimulated with either anti-CD40 or LPS, suggesting that activation of the toll-like receptor 4 (TLR4) signaling pathway is able to compensate for the loss of Orai1/Orai3. We also analyzed the proliferation and survival of B cells isolated from Orai2-/- mice and their wildtype littermates exposed to the same stimuli. We found only marginal or no effects of Orai2 knockout on B cell proliferation and survival (Figure 7—figure supplement 1A, B). Furthermore, we performed RNA sequencing on B cells isolated from control Mb1-Cre/+ and Orai3/Orai1fl/fl Mb1-Cre/+ mice after stimulation with anti-IgM for 24 hr (*Source data* 2) and differential expression analysis was performed with edgeR (*Source data* 3). Based on the edgeR output, GSEA software was applied to perform, we performed pathway analyses using gene set enrichment analysis (GSEA) (*Source data* 4). In agreement with our proliferation and survival data, GSEA showed significant downregulation of pathways that govern survival and cell cycle progression in Orai1/Orai3-deficient B cells (Figure 7D and E). Together, these findings reveal that only loss of both Orai1 and Orai3 affects B cell survival and proliferation in response to anti-IgM stimulation while single knockout of either Orai1 or Orai3 had marginal effects.
**Figure 7.:** *Orai1 and Orai3 regulate B cell proliferation and survival.(A) Measurement of B cell proliferation by tracking cabroxyfluorescein diacetate succinimidyl ester (CFSE) dilution. B lymphocytes from control, Orai1, Orai3, and Orai1/Orai3 knockout mice were loaded with CFSE (3 μM) and stimulated with anti-IgM (20 μg/mL), anti-CD40 (10 μg/mL), anti-IgM +anti-CD40, or LPS (10 μg/mL). CFSE dilution was determined 72 hr after stimulation for all conditions. (B) Quantification of the percentage of proliferating cells for each condition in (A) (from left to right n=9, 8, 4, and 4 biological replicates; one-way ANOVA with multiple comparisons to Mb1-Cre/+). CFSE dilution gate is drawn relative to unstimulated controls. (C) Quantification of the percentage of viable cells for each condition in (A) as determined by a Live/Dead viability dye (from left to right n=9, 8, 4, and 4 biological replicates; one-way ANOVA with multiple comparisons to Mb1-Cre/+). (D) Top KEGG pathways showing differential expression from RNA-sequencing analysis of B cells from Mb1-Cre/+ vs Orai1/Orai3fl/fl Mb1-Cre/+ mice stimulated for 24 hr with anti-IgM (20 μg/mL) (n=3 biological replicates for each). (E) Gene set enrichment analysis (GSEA) plots show enrichment statistics (ES) in the Hallmark G2M Checkpoint, E2F Targets, and Cell Cycle gene sets. Large positive ES values suggest activation of these pathways. Normalized enrichment score (NES) values are used to assess statistical significance, and the results for these gene sets are highly significant. All scatter plots are presented as mean ± SEM. For all figures, **p<0.01; ***p<0.001; ****p<0.0001; ns, not significant.
Figure 7—source data 1.Source data for Figure 7.*
## The combined deletion of Orai3 and Orai1 inhibits B cell metabolism
Metabolic reprogramming results in dramatically enhanced OXPHOS and remodeling of the mitochondrial network (Akkaya et al., 2018; Waters et al., 2018). Indeed, GSEA of B cells stimulated for 24 hr with anti-IgM identified the Oxidative *Phosphorylation* gene set as one of the most highly enriched pathways relative to unstimulated controls (Figure 8A). Additionally, anti-IgM stimulated B cells to demonstrate a significant increase in the total number of mitochondria per cell compared to naïve B cells as determined by transmission electron microscopy (TEM; Figure 8B and C). However, this increase in mitochondrial mass triggered by B cell activation was not affected by the depletion of Orai1 from B cells, as determined by Mitotracker Green staining (Figure 8D). We measured OCR and ECAR in B cells isolated from control and B cell-specific Orai knockout mice after B cell stimulation with anti-IgM for 24 hr (Figure 8E). Anti-IgM stimulation of wild-type B cells led to a robust increase in both OCR and ECAR (Figure 8F). Furthermore, Anti-IgM stimulation led to an increase in both basal and maximal respiration of wild-type B cells (Figure 8E–H). The increase in OCR and ECAR following BCR activation was significantly blunted in Orai1-deficient B cells, and further reduced in Orai1/Orai3-deficient B cells (Figure 8E–H). Interestingly, GSEA analysis showed that anti-IgM mediated enrichment of the Oxidative *Phosphorylation* gene set and Myc targets gene set were specifically inhibited in B cells from Orai1/Orai3fl/fl Mb1-Cre/+ mice, but not in B cells from Orai1fl/fl Mb1-Cre/+ mice (Figure 8—source data 1). However, the glycolysis gene set was inhibited in B cells from both Orai1/Orai3fl/fl Mb1-Cre/+ mice and Orai1fl/fl Mb1-Cre/+ mice (Figure 8—source data 1). Loss of Orai3 alone only partially reduced glycolytic flux and basal respiration but did not affect maximal respiration (Figure 8E–H). Previous research has shown that loss of SOCE in the chicken DT40 B cell line impaired mitochondrial metabolism by reducing CREB-mediated expression of the mitochondrial calcium uniporter (MCU) (Shanmughapriya et al., 2015). We observed no differences in CREB phosphorylation on Ser133 (a surrogate for CREB activation) or MCU expression in primary Orai1-deficient B cells from Orai1fl/fl Mb1-Cre/+ mice (Figure 8—figure supplement 1A). Likewise, we observed no differences in MCU expression in mouse A20 B cells lacking either Orai1, Orai3, or both Orai1 and Orai3 (Figure 8—figure supplement 1B). Both naïve and activated (24 hr with anti-IgM) B cells from Orai1fl/fl Mb1-Cre/+ mice showed no obvious changes in the expression of different components of the electron transport chain (Figure 8—figure supplement 1C, D).
**Figure 8.:** *Orai1 and Orai3 regulate B cell mitochondrial respiration.(A) Gene set enrichment analysis (GSEA) of the KEGG Oxidative Phosphorylation gene set in B cells stimulated for 24 hr with anti-IgM (20 μg/mL) relative to unstimulated controls. (B) Representative transmission electron microscopy (TEM) images of B cells from Mb1-Cre/+ mice. Shown are naïve, unstimulated B cells (top) and B cells stimulated for 24 hr with anti-IgM (bottom). (C) Quantification of total mitochondria per cell in unstimulated B cells and B cells stimulated for 24 hr with anti-IgM (n=25 for each; Mann-Whitney test). (D) Measurement of total mitochondrial content with MitoTracker Green in B cells from Mb1-Cre/+ and Orai1fl/fl Mb1-Cre/+ mice following 24 hr stimulation (n=3 biological replicates for each; Mann-Whitney test). (E) Measurement of oxygen consumption rate (OCR) in primary B lymphocytes following 24 hr stimulation with anti-IgM (20 μg/mL) using the Seahorse Mito Stress Test. (F) Energy map of maximal oxygen consumption rates (OCR) and extracellular acidification rates (ECAR) following carbonyl cyanide p-trifluoromethoxyphenylhydrazone (FCCP) addition. (G, H) Quantification of basal (G) and maximal (H) respiration from Seahorse traces in (E) (n=4 biological replicates for each genotype; one-way ANOVA with multiple comparisons to Mb1-Cre/+). All scatter plots and Seahorse traces are presented as mean ± SEM. For all figures, **p<0.01; ****p<0.0001; ns, not significant.
Figure 8—source data 1.Source data for GSEA summary statistics for Figure 8.
Figure 8—source data 2.Source data for Figure 8.*
To further investigate how Orai-mediated Ca2+ signaling regulates B cell metabolism, we profiled polar metabolites utilizing liquid chromatography followed by mass spectrometry in control and Orai1/Orai3-deficient B cells (Figure 9). Isolated B cells from each genotype were either unstimulated or stimulated for 24 hr with anti-IgM alone, anti-IgM +anti-CD40, anti-IgM +the calcineurin inhibitor FK506, or anti-IgM +the CRAC channel inhibitor GSK-7975A. We utilized GSK-7975A as we recently demonstrated that this compound inhibited all Orai isoforms compared to differential effects on Orai isoforms with other common SOCE inhibitors like Synta66 (Zhang et al., 2020). Stimulation of control B cells with anti-IgM led to a significant increase of glycolytic and TCA cycle metabolites along with most non-essential amino acids. This effect was further enhanced with anti-IgM +anti-CD40 co-stimulation (Figure 9A; Figure 9—figure supplement 1A). Inclusion of FK506 or GSK-7975A strongly blunted the effects of anti-IgM activation as overall metabolic profiles remained like those of unstimulated cells (Figure 10A). Importantly, this upregulation in polar metabolites upon B cell activation was significantly reduced in Orai1/Orai3-deficient B cells with either anti-IgM stimulation or anti-IgM +anti-CD40 co-stimulation (Figure 9B and C; Figure 9—figure supplement 1B). Of note, upregulation of most polar metabolites from Orai1/Orai3 knockout B cells co-stimulated with anti-IgM +anti-CD40 typically reached levels comparable to those of control B cells stimulated with anti-IgM alone. The inhibitory effects of FK506 and GSK-7975A on the metabolite status of wildtype B cells were comparable to those of Orai1/Orai3-deficient B cells (Figure 9A, Figure 9—figure supplement 1A). Collectively, these results reveal that SOCE mediated through Orai1/Orai3 channels contributes to the metabolic reprogramming of B lymphocytes.
**Figure 9.:** *Orai1/Orai3-mediated SOCE-calcineurin-NFAT pathway regulates B cell metabolism.(A) Analysis of polar metabolites in B cells from Mb1-Cre/+ mice utilizing liquid chromatography followed by mass spectrometry. B cells were either unstimulated or stimulated for 24 hr with anti-IgM, anti-IgM +anti-CD40, anti-IgM with 1 µM FK506, or anti-IgM with 10 µM GSK-7975A. (B, C). Heat maps of statistically significant polar metabolites in B cells from Mb1-Cre/+ and Orai1/Orai3fl/fl Mb1-Cre/+ mice following (B) 24 hr anti-IgM stimulation or (C) 24 hr anti-IgM +anti-CD40 stimulation. (n=3 biological replicates for each condition).
Figure 9—source data 1.Source data for Figure 9.* **Figure 10.:** *Deletion of Orai1 and Orai3 in B cells does not compromise immunity to influenza A virus (IAV).(A) Experimental outline. Littermate controls and Orai1fl/flOrai3fl/fl Mb1-Cre/+ mice have infected intranasally with 1x105 TCID50 of the x31 H3N2 strain of influenza A virus. Serum was collected on days 9 and 14, and mice were sacrificed on day 14 for analysis. (B) Representative H&E stains of lung sections. Scale bar: 500 µm. (C) Alveolar volume fraction of 10 mice per cohort. (D) Representative flow cytometry plots of B cells isolated from mediastinal lymph nodes (med LN). (E) Summary of the frequencies (%) and total cell numbers of B220–CD138+ plasma cells and Fas+GL-7+ GC B cells shown in panel D from 10 mice per cohort. (F) Representative flow cytometry plots of plasma cells isolated from the bone marrow of five IAV-infected mice per cohort. (G) Representative flow cytometry plots (left) and summary (right) of the frequencies of class-switched IgG1+ GC B cells in med LN of five IAV-infected mice per cohort. (H) IAV-specific IgM and IgG levels in the serum of 10 mice per cohort were measured on days 9 and 14. Panels B-H show the results of two independent experiments. Statistical analysis by unpaired Student’s t-test: ***p<0.001, **p<0.01, *p<0.05.
Figure 10—source data 1.Source data for Figure 10.*
## Immunity to influenza A virus is intact in mice with B cell-specific deletion of Orai1 and Orai3
Given the important role of Orai1 and Orai3 for multiple B cell functions in vitro, we hypothesized that deletion of Orai1 and Orai3 affects the B cell cytokine profile and compromises immune responses to influenza A virus (IAV) infection. Profiling of B cell cytokines and cytokine receptors by RNA-sequencing of B cells from Orai1fl/flOrai3fl/fl Mb1-Cre/+ and littermate control mice stimulated ex vivo with anti-IgM showed no obvious differences between the two groups (Figure 10—figure supplement 1C–D). To evaluate the role of Orai1/Orai3 for B cell function and humoral immunity in vivo, we infected Orai1fl/flOrai3fl/fl Mb1-Cre/+ mice and Orai1fl/flOrai3fl/fl littermate controls intranasally with a single dose of the laboratory strain A/HK/x31 (Hkx31, H3N2) (Thomas et al., 2006) and analyzed immune responses 14 days later (Figure 10A). All infected WT and Orai$\frac{1}{3}$-deficient mice survived and experienced a similar reduction in body weight over the two week-course of infection (not shown). Lung histology at day 14 post-infection (p.i.) showed comparable pulmonary inflammation and alveolar volume in Orai$\frac{1}{3}$-deficient and littermate control mice (Figure 10B and C). To directly investigate the effects of Orai1 and Orai3 on humoral immunity during IAV infection, we analyzed total B cells, germinal center (GC) B cells, and plasma cells in mice at day 14 p.i. The frequencies and numbers of B220–CD138+ plasma cells and B220+Fas+GL-7+ GC B cells were similar in the mediastinal lymph nodes of WT and Orai$\frac{1}{3}$-deficient mice (Figure 10D and E). Moreover, the numbers of CD138+ plasma cells in the bone marrow of infected Orai1fl/flOrai3fl/fl Mb1-Cre/+ mice were comparable to those in control littermate mice (Figure 10F). We next investigated if deletion of Orai1 and Orai3 in B cells affects their ability to induce class switch recombination. The percentages of class-switched IgG1+ cells (and those of IgM–IgG– non-switched B cells) among GC B cells, however, were comparable in control and Orai$\frac{1}{3}$-deficient mice (Figure 10G). Consistent with these results, we found normal levels of IAV-specific IgM and class-switched IgG antibodies in the serum (Figure 10H). Taken together, our data show that B cell-specific deletion of Orai1 and Orai3 does not significantly impair immune responses to influenza A virus infection.
## Discussion
Given the essential role of B lymphocytes in driving humoral immunity against foreign pathogens, a comprehensive understanding of the molecular pathways that govern their development, differentiation, and effector functions is critical for future targeted therapies. One of the earliest signaling events upon crosslinking of the BCR is a biphasic increase in intracellular Ca2+ concentrations (Baba and Kurosaki, 2016). Early landmark studies established that multiple Ca2+ dependent transcription factors display unique activation requirements by relying on either ER Ca2+ release (e.g. JNK, NF-κB) or sustained Ca2+ signals driven by SOCE (NFAT) (Dolmetsch et al., 1997; Healy et al., 1998). While the SOCE-calcineurin-NFAT pathway is well established in the context of B cell effector function, recent reports have shed light on the mechanisms by which SOCE also regulates NF-κB activation and its downstream target genes (Berry et al., 2020; Berry et al., 2018). These two BCR-activated signaling pathways are sustained by Ca2+ entry through CRAC channels and synergize with one another to activate a series of Ca2+-regulated checkpoints that determine B cell survival, entry into the cell cycle, and proliferation (Akkaya et al., 2018; Berry et al., 2020). Unlike recent findings that have established Orai1 and Orai2 as the major Orai isoforms mediating SOCE in T cells (Vaeth et al., 2017b), the relative contributions of Orai isoforms to the native CRAC channel in B cells remained, until now, unclear.
Our results herein establish that B cell activation through BCR stimulation alone or co-stimulation with secondary signals like CD40 or TLR ligands significantly enhanced metabolic activity. Concurrently, these stimulation conditions drive dynamic changes in the expression of each Orai isoform. Interestingly, we observe that robust B cell activation with BCR and CD40 co-stimulation results in upregulation of both Orai1 and Orai3, and downregulation of Orai2. However, we do not see any apparent increase in SOCE in activated B cells versus naive B cells, suggesting that Orai1 and Orai3 upregulation in activated B cells helps maintain the magnitude of SOCE, commensurate with the enhanced volume of activated B cells. While Orai1 has previously been shown to contribute to the majority of SOCE in B cells under conditions of maximal ER Ca2+ depletion (Gwack et al., 2008; McCarl et al., 2010), we show using B-cell specific Orai1 knockout mice that Orai1 is dispensable for maintaining cytosolic Ca2+ oscillations in response to BCR crosslinking. In agreement with our previous results in HEK293 cells (Yoast et al., 2020b), we also show that Orai1 is an essential regulator of NFAT1 and NFAT2 isoforms in B cells and that its role becomes more prominent for NFAT induction in activated B cells compared to naïve unstimulated B cells. *By* generating CRISPR/Cas9 B cell lines and B-cell specific knockout mice lacking Orai1 or Orai3 individually and in combination, we found that both Orai1 and Orai3 contribute to the native CRAC channels in B cells. However, the oligomeric state of native CRAC channels in B cells and whether Orai1 and Orai3 form homomeric or heteromeric assemblies or both in primary B cells remains an open question.
The combined loss of Orai1 and Orai3, but not either isoform alone, led to a significant reduction in both B cell proliferation and survival. The original study by Gwack et al showed a partial inhibition of proliferation in B cells from Orai1-/- mice. While we observe only a marginal inhibition of proliferation in B cells from Orai1fl/fl Mb1-Cre/+ mice and a moderate reduction in the viability of these cells. The reason for the difference between our results on B cell proliferation and those of Gwack et al., 2008 is unknown. However, one potential explanation could be that in B cells from Orai1-/- mice in Gwack et al., 2008 Orai1 is deleted from early B cell development in mice on a mixed ICR background (note that Orai1-/- mice on the C57BL/6 background show perinatal lethality) whereas in our Orai1fl/fl Mb1-Cre/+ mice, Orai1 is deleted later in B cell development in mice that are on a pure C57Bl/6 background. Another potential explanation could be the complete deletion of the *Orai1* gene in the studies of Gwack et al., 2008 vs the $97\%$ reduction in mRNA expression we observe in B cells from Orai1fl/fl Mb1-Cre/+ mice.
We show that SOCE is important for the metabolic reprogramming of B cells by regulating mitochondrial metabolism and the flux of polar metabolites in response to B cell activation. This shift in metabolic profiles in response to B cell activation could be neutralized by inhibition of either calcineurin or CRAC channels with FK506 or GSK-7975A, respectively, suggesting that this metabolic flux is driven through SOCE and NFAT-dependent mechanisms. Our data demonstrate that both Orai1 and Orai3 contribute to CRAC channel activity and to shaping cytosolic Ca2+ signaling in B cells. While our data has provided evidence for the contribution of Orai3 to CRAC channels in naïve B cells, this contribution was only apparent within the context of double Orai1 and Orai3 knockout. Our data with B cell-specific Orai3 knockout is consistent with a recent study reporting a lack of SOCE inhibition in B cells, T cells, and macrophages from Orai3 global knockout mice (Wang et al., 2022). Our data and the study of Wang et al., 2022, showed that Orai3 is highly expressed in B cells compared to other immune cells like T cells. Nevertheless, Orai3 knockout, on its own, has no measurable contribution to SOCE in naïve B cells while moderately reducing SOCE in activated B cells. This reduced SOCE in activated B cells from Orai3 knockout mice is consistent with the reduced basal respiration of B cells stimulated with anti-IgM for 24 hr. This could also explain why the double knockout of Orai1 and Orai3, which inhibits SOCE more dramatically, has a larger effect on mitochondrial respiration and metabolomic reprogramming of B cells. Alternatively, it is possible that Orai3 couples to alternative cytosolic signaling pathways that synergize with SOCE-mediated Ca2+ influx. Orai3 could mediate its effects on mitochondrial respiration through the store-independent Ca2+ influx pathway activated by arachidonic acid or its metabolite, leukotrieneC4 (Thompson et al., 2013; Zhang et al., 2015; Zhang et al., 2013; Zhang et al., 2018; Zhang et al., 2014). Earlier studies investigating human effector T cells demonstrated that Orai3 becomes upregulated in response to oxidative stress, which may act as a potential mechanism to maintain a threshold of Ca2+ influx and T cell function in various inflammatory conditions (Bogeski et al., 2010). In support of this model, new findings have demonstrated that Orai3 expression is increased in CD4+ T cells from patients with rheumatoid arthritis and psoriatic arthritis and that silencing of Orai3 reduces tissue inflammation in a human synovium adoptive transfer model (Ye et al., 2021). Interestingly, these effects were proposed to be through the store-independent function of Orai3 (Thompson et al., 2013; Zhang et al., 2015; Zhang et al., 2013; Zhang et al., 2018; Zhang et al., 2014). However, a recent study showed that Orai3 deficient mice have no altered susceptibility in a model of collagen-induced arthritis (Wang et al., 2022), challenging a role for Orai3, on its own, in the immune system.
The situation in T cells is different as the combined loss of Orai1 and Orai2 in murine T cells led to a near complete ablation of SOCE and substantially impaired T cell function (Vaeth et al., 2017b). Whether Orai3 contributes to the small, residual amount of SOCE in Orai1/Orai2-deficient T cells and/or mediates store-independent signaling functions in murine T and B cells is unknown. Furthermore, differences in Orai channel contributions are apparent between mice and humans, as SOCE and CRAC currents are completely inhibited in T cells from patients with LoF mutations in Orai1, while SOCE is only partially inhibited in Orai1-deficient murine T cells (Feske et al., 2006; Lian et al., 2018; McCarl et al., 2009; Vig et al., 2008). Curiously, B cells from Orai1 LoF mutation patients still retain residual SOCE, suggesting that another Orai isoform may contribute to SOCE in human B cells than in T cells (Feske et al., 2001), in agreement with our data herein on mouse B cells. Thus, the composition of native CRAC channels among lymphocytes appears cell-type and context-specific.
While SOCE is substantially reduced in Orai1/Orai3-deficient B cells, defects in their ability to proliferate and survive in response to antigenic stimulation are not as severe as the phenotype observed in STIM1/STIM2-deficient B cells (Berry et al., 2020; Matsumoto et al., 2011). Our data establish additivity between Orai1 and Orai3 channels in controlling B cell survival and proliferation, but we cannot exclude a potential role for Orai2 in the regulation of SOCE in B cells. Quite the opposite, we show that Orai2 is functional in B cells by documenting that B cells from Orai2-/- mice have enhanced SOCE. This suggests that Orai2 acts as a negative regulator of SOCE in B cells as was shown for T cells (Vaeth et al., 2017b), mast cells (Tsvilovskyy et al., 2018), and HEK293 cells (Yoast et al., 2020b), and that the remaining SOCE activity in B cells from Orai1/Orai3 double knockout mice is likely mediated by Orai2. This is consistent with our results showing that the SOCE inhibitor GSK-7975A inhibits the upregulation of polar metabolites by B cells, in response to anti-IgM +anti-CD40 stimulation, to a greater extent than Orai1/Orai3 double knockout. Therefore, the development of Orai triple knockout mice and additional Orai knockout pairs is needed to further clarify this issue. Similarly, how SOCE and the composition of the native CRAC channel vary among different B cell subsets and during pathological conditions is completely unknown. Indeed, earlier studies have demonstrated that effector populations like germinal center B cells display unique metabolic and Ca2+ signaling requirements (Khalil et al., 2012; Luo et al., 2018). Our data also suggest that the expression of each Orai isoform is dynamic in response to B cell activation, similar to the case of naïve vs effector T cells (Vaeth et al., 2017b).
Considering the clear effects of B cell-specific deletion of Orai1 and Orai3 on SOCE, NFAT activation, metabolic reprogramming, survival, and proliferation, it was surprising that humoral immunity to influenza A virus (IAV) infection was unaffected in Orai1fl/flOrai3fl/fl Mb1-Cre/+ mice. This raises the question regarding the role of Orai1, Orai3, and SOCE in general for B cell function and humoral immunity in vivo. Our results are, however, consistent with those from B cell-specific deletion of Stim1 and *Stim2* genes, which was shown to result in abolished SOCE causing impaired expression of two key anti-apoptotic genes and blunted activation of the mTORC1 and c-Myc metabolic signaling pathways, resulting in decreased B cell survival and proliferation (Berry et al., 2020). While this study did not investigate the effects of Stim1/Stim2 deletion on B cell function in vivo, Baba and colleagues had reported normal B cell development as well as T-dependent and T-independent antibody responses following immunization of mice with B cell-specific deletion of Stim1 and *Stim2* genes (Matsumoto et al., 2011). This was surprising, because Stim1/Stim2 deletion almost completely abolished thapsigargin and anti-IgM induced SOCE, and strongly impaired B cell survival and proliferation in vitro (Matsumoto et al., 2011). A potential explanation for the normal humoral immune responses in Orai1fl/flOrai3fl/fl Mb1-Cre/+ and Stim1fl/flStim2fl/fl Mb1-Cre/+ mice is provided by the fact that, in contrast to anti-IgM stimulation, stimulation of Stim1/Stim2-deficient B cells via CD40 or TLR4 results in normal B cell survival and proliferation (Matsumoto et al., 2011). When combined with anti-IgM stimulation, CD40 was able to rescue the proliferation defect of Stim1/Stim2-deficient B cells. A similar rescue of defective B cell proliferation by CD40 and TLR9 agonists in Stim1/Stim2-deficient B cells was reported by Berry at al. ( Berry et al., 2020). Our data with Orai1/Orai3-deficient B cells show only a partial rescue of B cell proliferation by anti-IgM and anti-CD40 co-stimulation compared to anti-IgM alone. We did, however, observe normal B cell survival and proliferation in Orai1/Orai3-deficient B cells stimulated with the TLR4 agonist LPS. Collectively, these data demonstrate that co-stimulatory pathways such as CD40, TLR4, and TLR9 can bypass the requirement for BCR-induced SOCE to induce B cell activation. IAV is a dsRNA virus that is detected by TLR7 in plasmacytoid dendritic cells and B cells (Iwasaki and Pillai, 2014). TLR7 is widely expressed in B cells and antibody responses to IAV were shown to depend on TLR7 (Geeraedts et al., 2008). Although we did not investigate if TLR7 stimulation with synthetic agonists such as imiquimod can rescue the proliferation and survival defects of Orai1/Orai3-deficient B cells in vitro, we speculate that it would, given the fact that TLR7 and TLR4, whose activation by LPS rescues the function of Stim1/Stim2 and Orai1/Orai3-deficient B cells, signal through the same Myd88-IRAK4-TRAF6 pathway to activate NF-κB and other transcription factors (Duan et al., 2022). Activation of TLR7 and potentially other costimulatory pathways, in Orai1/Orai3-deficient B cells is, therefore, a likely explanation for their normal humoral immune response to IAV in vivo. Although Ca2+ and SOCE are clearly important for B cell function, B cells have developed other Ca2+-independent means to ensure their proper activation. Future studies with Orai1fl/flOrai3fl/fl Mb1-Cre/+ mice might reveal indispensable in vivo contributions of SOCE to B cell populations such as memory B cells or to immune disease in other in vivo models such as the experimental autoimmune encephalomyelitis (EAE) mouse model. Taken together, our results uncover additive functions of Orai3 and Orai1 in B cell Ca2+ signaling that regulates NFAT activation, metabolism, survival, and proliferation of B cells.
## Mice
All animal experiments were carried out in compliance with the Institutional Animal Care and use Committee (IACUC) guidelines of the Pennsylvania State University College of Medicine. All mice were housed under specific pathogen-free conditions and experiments were performed in accordance with protocols approved by the IACUC of the Pennsylvania State University College of Medicine. Mb1-Cre mice (Hobeika et al., 2006) were obtained from The Jackson Laboratory (strain#: 020505; https://www.jax.org/strain/020505). Orai1fl/fl mice (Ahuja et al., 2017) were obtained from Dr. Paul Worley (Johns Hopkins University). Orai3fl/fl mice were generated by our laboratory through the MMRC at the University of California Davis. A trapping cassette was generated including ‘SA-βgeo-pA’ (splice acceptor-beta-geo-polyA) flanked by Flp-recombinase target ‘FRT’ sites, followed by a critical Orai3 coding exon flanked by Cre-recombinase target ‘loxP’ sites. This cassette was inserted within an intron upstream of the Orai3 critical exon, where it tags the *Orai3* gene with the lacZ reporter. This creates a constitutive null Orai3 mutation in the target *Orai3* gene through efficient splicing to the reporter cassette resulting in the truncation of the endogenous transcript. Mice carrying this allele were bred with FLP deleter C57BL/6 N mice to generate the Orai3fl/fl mouse. All experiments were performed with 8–12 week-old age and sex-matched mice.
## Cell culture
Parental A20 cells were purchased from ATCC (Catalog # TIB-208) and cultured in RPMI 1640 with L-glutamine supplemented with $10\%$ fetal bovine serum and 1 X Antibiotic-Antimycotic and were routinely tested for lack of mycoplasma contamination (Emrich et al., 2022a). Naïve splenic B lymphocytes were purified by negative selection using the EasySep Mouse B Cell Isolation Kit (STEMCELL Technologies). Primary B cells from each transgenic mouse line were cultured in RPMI 1640 with L-glutamine supplemented with $10\%$ fetal bovine serum, 1 X GlutaMAX, 1 X Penicillin-Streptomycin Solution, sodium pyruvate (1 mM), 2-ME (5 uM), and HEPES (10 mM). Primary lymphocytes were stimulated with F(ab')₂ Fragment Goat Anti-Mouse IgM antibody (Jackson ImmunoResearch, 20 µg/mL), anti-mouse CD40 antibody (BioXCell, 10 µg/mL), or LPS (Sigma-Aldrich, 10 µg/mL). All cell lines and primary lymphocytes were cultured at 37 °C in a humidified incubator with $5\%$ CO2.
## Generation of A20 Orai CRISPR/Cas9 knockout cells
For the generation of A20 Orai knockout clones, we used a similar strategy to our previous studies (Emrich et al., 2021; Yoast et al., 2021; Yoast et al., 2020b; Zhang et al., 2019). Briefly, two gRNAs targeting the beginning and end of the mouse Orai1 or Orai3 coding region were cloned into two fluorescent vectors (pSpCas9(BB)–2A-GFP and pU6-(BbsI)_CBh-Cas9-T2AmCherry; Addgene). For mOrai1 knockout, the gRNA sequences are the following: mOrai1n: 5’-GCCTTCGGATCCGGTGCGTC-3’; mOrai1c: 5’-CACAGGCCGTCCTCCGGACT-3’. For mOrai3 knockout, the gRNA sequences are the following: mOrai3n: 5’- GCGTCCGTAACTGTTCCCGC-3’; mOrai3c: 5- GAAGGAGGTCTGTCGATCCC-3’. A20 cells were electroporated with both N- and C-terminal gRNA combinations with an Amaxa Nucleofector II and single cells with high GFP and mCherry expression were sorted at one cell per well into 96-well plates 24 hr after transfection using a FACS Aria SORP Cell Sorter. Individual clones were obtained and genomic DNA was tested using primers targeting the N- and C-terminal gRNA cut sites to resolve wild-type and knockout PCR products. Knockout PCR products were cloned into pSC-B-amp/kan with the StrataClone Blunt PCR Cloning Kit (Agilent) and Sanger sequenced to determine the exact deletion. Knockout clones were also confirmed for the absence of mRNA with qRT-PCR and functionally through Ca2+ imaging experiments.
## Fluorescence imaging
A20 cell lines and primary B cells were seeded onto poly-L-lysine (Sigma-Aldrich) coated coverslips. Coverslips were mounted in Attofluor cell chambers (Thermo Scientific) and loaded with 2 µM Fura-2AM (Molecular Probes) in a HEPES-buffered saline solution (HBSS) containing 120 mM NaCl, 5.4 mM KCl, 0.8 mM MgCl2, 1 mM CaCl2, 20 mM HEPES, 10 mM D-glucose, at pH 7.4 for 30 min at room temperature. Following Fura-2 loading, cells were washed three times with HBSS and mounted on a Leica DMi8 fluorescence microscope. Fura-2 fluorescence was measured every 2 s by excitation at 340 nm and 380 nm using a fast shutter wheel and the emission at 510 nm was collected through a 40 X fluorescence objective. Fluorescence data was collected from individual cells on a pixel-by-pixel basis and processed using Leica Application Suite X. All cytosolic Ca2+ concentrations are presented as the ratio of F340/F380.
## Flow cytometry
Spleens were processed into single-cell suspensions and stained using the following antibodies: B220-BV605 (RA3-6B2, Biolegend), CD86-PE/Cy5 (GL-1, Biolegend), MHC-II-PE/Cy7 (M$\frac{5}{114.15.2}$, Biolegend), Orai1 rabbit polyclonal (Gwack et al., 2008; McCarl et al., 2009), CD3e-PE (145–2 C11, BD Biosciences), CD8a-V500 (53–6.7, BD Biosciences), CD4-BB700 (RM4-5, BD Biosciences), APC–anti-CD24 (HSA; M$\frac{1}{69}$), FITC–anti-CD23 (B3B4), PE–anti-IgM (eB121-15F9, eBioscience), APC–anti-CD93 (AA4.1, eBioscience), PE–Cy5-streptavidin (Biolegend), and Pacific blue–anti-B220 (RA3-6B2, Biolegend). All staining was performed in FACS buffer (DPBS, $2\%$ FBS, 1 mM EDTA) for 30 min at 4 °C. Prior to surface staining, all cells were stained with eBioscience Fixable Viability Dye eFluor 780 (Thermo Fisher) and anti-CD$\frac{16}{32}$ Fc block (2.4G2, Tonbo Biosciences). Measurement of mitochondrial content was performed using MitoTracker Green (Molecular Probes) and mitochondrial membrane potential was performed using TMRE (Molecular Probes). For all CFSE dilution experiments, primary B lymphocytes were stained with 3 µM CFSE (C1157, Thermo Fischer) in PBS with $5\%$ FBS for 5 min at room temperature, followed by the addition of FBS and two washes in PBS before resuspension in complete RPMI media. CFSE dilution of labeled cells was assessed 72 hr after plating and stimulation. All flow cytometry data were collected on a BD LSR II flow cytometer using FACSDiva software (BD Biosciences) and analyzed with FlowJo 9.9.6 software (Tree Star).
## Analysis of NFAT nuclear translocation
Primary B cells were purified using the EasySep Mouse B Cell Isolation Kit (STEMCELL Technologies) and stimulated with anti-IgM (20 µg/mL) for 15 min at room temperature in complete RPMI media. Cells were fixed and permeabilized using the Foxp3 /Transcription Factor Staining Buffer Set (eBioscience) following the manufacturer’s instructions. Cells were stained in 1 X permeabilization buffer for 1 hr at room temperature with NFAT1-Alexa Fluor 488 (D43B1, Cell Signaling Technology). Nuclei were stained with DAPI prior to acquisition on an Amnis ImageStream X Mark II Imaging Flow Cytometer and analyzed with IDEAS software using the ‘‘Nuclear Localization’’ feature (EMD Millipore).
## Western blot analysis
A20 and primary B lymphocytes were harvested from the culture and lysed for 10 min in RIPA buffer (150 mM NaCl, $1.0\%$ IGEPAL CA-630, $0.5\%$ sodium deoxycholate, $0.1\%$ SDS, 50 mM Tris, pH 8.0; Sigma) containing 1 X Halt protease/phosphatase inhibitors (Thermo Scientific). Following lysis, samples were clarified by centrifugation at 15,000 × g for 10 min at 4°C. Supernatants were collected and protein concentration was determined using the Pierce Rapid Gold BCA Protein Assay Kit (Thermo Scientific). Equal concentrations of protein extract were loaded into 4–$12\%$ NuPAGE BisTris gels (Life Technologies) and transferred to PVDF membranes utilizing the Transblot Turbo Transfer System (Bio-Rad). Membranes were blocked for 1 hr at room temperature in Odyssey Blocking Buffer in TBS (LI-COR) and incubated overnight at 4°C with primary antibody. The following antibodies and dilutions were used: MCU (1:2000; 14997 S, Cell Signaling Technology), GAPDH (1:5000; MAB374, Sigma), Total OXPHOS Rodent Antibody Cocktail (1:1000, ab110413, Abcam), NFAT1 (1:1000, 4389 S, Cell Signaling Technology), NFAT2 (1:1000, 8032 S, Cell Signaling Technology), α-Tubulin (1:5000, 3873 S, Cell Signaling Technology), phospho-CREB (Ser133, 1:1000, 9198 S, Cell Signaling Technology), and CREB (1:1000, 9104 S, Cell Signaling Technology). Membranes were washed with TBST and incubated for 1 hr at room temperature with the following secondary antibodies: IRDye 680RD goat anti-mouse (1:10,000 LI-COR) or IRDye 800RD donkey anti-rabbit (1:10,000 LI-COR). Membranes were imaged on an Odyssey CLx Imaging System (LI-COR) and analysis was performed in Image Studio Lite version 5.2 (LI-COR) and ImageJ.
## Quantitative RT-PCR
Total mRNA was isolated from A20 cells and primary B lymphocytes using an RNeasy Mini Kit (Qiagen) following the manufacturer’s instructions. RNA concentrations were measured using a NanoDrop 2000 (Thermo Scientific) and 1 µg of DNAse I treated RNA was used with the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems). cDNA was amplified on a QuantStudio 3 Real-Time PCR System (Applied Biosystems) using PowerUp SYBR Green Master Mix (Applied Biosystems). PCR amplification was performed by initial activation for 2 min at 50°C, followed by a 95°C 2 min melt step. The initial melt steps were then followed by 40 cycles of 95°C for 15 s, 60°C for 15 s, and 72°C for 30 s. Data were analyzed with the instrument software v1.3.1 (Applied Biosystems) and analysis of each target was carried out using the comparative Ct method.
## Seahorse extracellular flux analysis
Oxygen consumption rates (OCR) and extracellular acidification rates (ECAR) were measured using an XFe24 Extracellular Flux Analyzer (Seahorse Bioscience). A20 cells (0.8 × 106 per well) and primary B lymphocytes (2 × 106 per well) were resuspended in XF DMEM pH 7.4 media supplemented with 1 mM pyruvate, 2 mM glutamine, and 10 mM glucose (Mito Stress Test) and plated in poly-L-lysine coated microchamber wells. For the Mito Stress Test, 1.5 μM oligomycin, 2 μM FCCP, and 0.5 μM antimycin/rotenone were utilized. Data were analyzed using the Agilent Seahorse Wave Software and normalized to total protein context per well using the Pierce Rapid Gold BCA Protein Assay Kit (Thermo Scientific).
## Transmission electron microscopy
Primary B cells (unstimulated or 24 hr IgM stimulated) were seeded onto poly-L-lysine coated cell culture dishes and fixed with $1\%$ glutaraldehyde in 0.1 M sodium phosphate buffer, pH 7.3. After fixation, the cells were washed with 100 mM Tris (pH 7.2) and 160 mM sucrose for 30 mins. The cells were washed twice with phosphate buffer (150 mM NaCl, 5 mM KCl, 10 mM Na3PO4, pH 7.3) for 30 min, followed by treatment with $1\%$ OsO4 in 140 mM Na3PO4 (pH 7.3) for 1 hr. The cells were washed twice with water and stained with saturated uranyl acetate for 1 hr, dehydrated in ethanol, and embedded in Epon (Electron Microscopy Sciences, Hatfield, PA). Roughly 60 nm sections were cut and stained with uranyl acetate and lead nitrate. The stained grids were analyzed using a Philips CM-12 electron microscope (FEI; Eindhoven, The Netherlands) and photographed with a Gatan Erlangshen ES1000W digital camera (Model 785, 4 k 3 2.7 k; Gatan, Pleasanton, CA).
## RNA sequencing and differential expression analysis
Total mRNA was isolated from primary B lymphocytes using an RNeasy Mini Kit (Qiagen) following the manufacturer’s instructions. RNA concentrations were quantitated using a NanoDrop 2000 (Thermo Scientific) and library preparation was performed by Novogene. A total amount of 1 µg RNA per sample was used as input material for the RNA sample preparations. Sequencing libraries were generated using NEBNext Ultra TM RNA Library Prep Kit for Illumina (NEB, USA) following the manufacturer’s recommendations, and index codes were added to attribute sequences to each sample. Briefly, mRNA was purified from total RNA using poly-T oligo-attached magnetic beads. Fragmentation was carried out using divalent cations under elevated temperature in NEBNext First-strand Synthesis Reaction Buffer (5 X). First-strand cDNA was synthesized using random hexamer primer and M-MuLV Reverse Transcriptase (RNase H-). Second-strand cDNA synthesis was subsequently performed using DNA Polymerase I and RNase H. Remaining overhangs were converted into blunt ends via exonuclease/polymerase activities. After adenylation of 3’ ends of DNA fragments, NEBNext Adaptor with hairpin loop structure was ligated to prepare for hybridization. To select cDNA fragments of preferentially 150~200 bp in length, the library fragments were purified with AMPure XP system (Beckman Coulter, Beverly, USA). Then 3 µl USER Enzyme (NEB, USA) was used with size-selected, adaptor-ligated cDNA at 37°C for 15 min followed by 5 min at 95°C before PCR. Then PCR was performed with Phusion High-Fidelity DNA polymerase, Universal PCR primers, and Index (X) Primer. Lastly, PCR products were purified (AMPure XP system) and library quality was assessed on the Agilent Bioanalyzer 2100 system. The clustering of the index-coded samples was performed on a cBot Cluster Generation System using PE Cluster Kit cBot-HS (Illumina) according to the manufacturer’s instructions. After cluster generation, the library preparations were sequenced on an Illumina NovaSeq 6000 Platform (Illumina, San Diego, CA, USA) using a paired-end 150 run (2 × 150 bases).
## Differential expression analysis
*Ensembl* gene identifiers were converted to gene symbols, and identifiers with duplicated or missing gene symbols were removed from the analysis. Exploratory analyses were performed on the gene-level read count data, and lowly expressed genes were removed from the data set. The EDASeq R package (Risso et al., 2011) was used to create a SeqExpressionSet object based on the read counts, then upper quantile normalization was applied. The RUVSeq R package (Risso et al., 2011) was then applied using $k = 1$ and a set of 14 mouse housekeeping genes identified by Ho and Patrizi, 2021 to identify factors of unwanted variation that were included as a covariate in the differential expression analysis performed with edgeR (McCarthy et al., 2012; Robinson et al., 2010). Differentially expressed genes were chosen based on a false discovery rate threshold of q<0.05. R 4.0.5 was used for all analyses (https://www.R-project.org).
## Gene set enrichment analysis (GSEA)
Based on the edgeR output, GSEA software was applied to perform pathway analyses (Mootha et al., 2003; Subramanian et al., 2005). The edgeR output includes likelihood ratio (LR) test statistics, p-values, and false discovery rate q-values, as well as log fold change (logFC) values based on the expression values of each gene in the comparison groups of interest. Because the LR statistics are non-negative, their values alone cannot distinguish up- and down-regulated genes, as is required for a GSEA pre-ranked analysis. Thus, we computed a signed version of the LR statistic that was defined to be the product of LR statistic times and the sign of the logFC, thereby enabling us to rank the genes according to both the statistical significance and the direction of the expression differences. A GSEA pre-ranked analysis was performed using the ranked genes and gene sets available at the Molecular Signature Database (https://www.gsea-msigdb.org/gsea/msigdb/index.jsp). After the biomaRt R package (Durinck et al., 2005; Durinck et al., 2009) was used to convert mouse gene symbols to human gene symbols. Statistical significance was assessed at the q<0.05 level.
## Metabolite extraction
A metabolite extraction was carried out on each sample based on a previously described method (Pacold et al., 2016). An approximate cell count (5 × 106 cells for all samples) of the samples was used to scale the metabolite extraction to a ratio of 1 × 106 cells/1 mL extraction solution [$80\%$ LCMS grade methanol (aq) with 500 nM labeled amino acid internal standard (Cambridge Isotope Laboratories, Inc, Cat No. MSK-A2-1.2)]. The lysis was carried out in two steps as follows. First, 1 mL of freezing extraction solution was added to the tubes containing each pellet. That suspension was transferred to tubes along with zirconium disruption beads (0.5 mm, RPI) and homogenized for 5 min at 4 °C in a BeadBlaster with a 30 s on/30 s off pattern. The resulting lysate was then diluted to a fixed concentration of 1 × 106 cells/1 mL in a new tube with disruption beads in that same buffer and then re-homogenized in the same way as above. The homogenate was centrifuged at 21,000 × g for 3 min, and 450 μL of the supernatant volume was transferred to a 1.5 mL microfuge tube for speed vacuum concentration, no heating. The dry extracts were resolubilized in 50 μL of LCMS grade water, sonicated in a water bath for 2 min, centrifuged as above, and transferred to a glass insert for analysis.
## LC-MS/MS methodology
Samples were subjected to an LC-MS/MS analysis to detect and quantify known peaks. The LC column was a Millipore ZIC-pHILIC (2.1 × 150 mm, 5 μm) coupled to a Dionex Ultimate 3000 system and the column oven temperature was set to 25 °C for the gradient elution. A flow rate of 100 μL/min was used with the following buffers; (A) 10 mM ammonium carbonate in water, pH 9.0, and (B) neat acetonitrile. The gradient profile was as follows; 80–$20\%$ B (0–30 min), 20–$80\%$ B (30–31 min), 80–$80\%$ B (31–42 min). Injection volume was set to 2 μL for all analyses (42 min total run time per injection). MS analyses were carried out by coupling the LC system to a Thermo Q Exactive HF mass spectrometer operating in heated electrospray ionization mode (HESI). Method duration was 30 min with a polarity switching data-dependent Top five method for both positive and negative modes. Spray voltage for both positive and negative modes was 3.5 kV and the capillary temperature was set to 320 °C with a sheath gas rate of 35, aux gas of 10, and max spray current of 100 μA. The full MS scan for both polarities utilized 120,000 resolution with an AGC target of 3 × 106 and a maximum IT of 100ms, and the scan range was from 67 to 1000 m/z. Tandem MS spectra for both positive and negative modes used a resolution of 15,000, AGC target of 1 × 105, maximum IT of 50 ms, isolation window of 0.4 m/z, isolation offset of 0.1 m/z, fixed first mass of 50 m/z, and 3-way multiplexed normalized collision energies (nCE) of 10, 35, 80. The minimum AGC target was 1 × 104 with an intensity threshold of 2 × 105. All data were acquired in profile mode.
## Data analysis
Metabolomics data were processed with an in-house pipeline for statistical analyses and plots were generated using a variety of custom Python code and R libraries including: heatmap, MetaboAnalystR, and manhattanly. Peak height intensities were extracted based on the established accurate mass and retention time for each metabolite as adapted from the Whitehead Institute and verified with authentic standards and/or high-resolution MS/MS manually curated against the NIST14MS/MS and METLIN spectral libraries. The theoretical m/z of the metabolite molecular ion was used with a±10 ppm mass tolerance window, and a±0.2 min peak apex retention time tolerance within the expected elution window (1–2 min). To account for sample-to-sample variance in the estimated cell counts, a sum-normalization step was carried out on a per-column (sample) basis. Detected metabolite intensities in a given sample were summed, and a percentage intensity was calculated for each metabolite (custom Rscript available from NYU Metabolomics Laboratory at https://med.nyu.edu/research/scientific-cores-shared-resources/metabolomics-laboratory, contact the Laboratory Director Dr. Drew Jones at: [email protected]). The median mass accuracy vs the theoretical m/z for the library was −0.7 ppm ($$n = 90$$ detected metabolites). Median retention time range (time between earliest and latest eluting sample for a given metabolite) was 0.23 min (30 min LCMS method). A signal-to-noise ratio (S/N) of 3 X was used compared to blank controls throughout the sequence to report detection, with a floor of 10,000 (arbitrary units). Labeled amino acid internal standards in each sample were used to assess instrument performance (median CV=$5\%$).
## Influenza A virus (IAV) infection
Orai1/Orai3fl/fl Mb1-Cre/+ and Orai1/Orai3fl/fl (control) mice were anesthetized with isoflurane and infected intranasally (i.n.) with 105 TCID50 of the laboratory strain A/HK/x31 (x31-IAV) of the influenza A virus subtype H3N2. Lungs were isolated for histology. Mediastinal lymph nodes and bone marrow were used to prepare single-cell suspensions followed by flow cytometric analysis. Serum was harvested for analyzing virus-specific antibody titers.
## Histology
Lungs were fixed with $4\%$ paraformaldehyde in PBS, embedded in paraffin, and cut at 5 μm. Sample slides were stained with hematoxylin and eosin (H&E) using standard methods. Images were acquired using an SCN400 slide scanner (Leica), viewed with Omero Slidepath (Open Microscope Environment). The total alveolar volume fraction was determined using the following procedure: (i) regions representing empty space were acquired by setting a color threshold in ImageJ (Schneider et al., 2012); (ii) to calculate the total alveolar volume fraction, the area of empty space was divided by the total lung area using MATLAB (v2018a).
## ELISA
To measure X31-specific antibodies, half-area ELISA plates (Corning, Cat#3690) were coated with heat-inactivated x31-IAV overnight at 4 °C. ELISA plates were blocked with $20\%$ FBS in PBS at 37 °C for 1 hr followed by washing three times with wash buffer ($0.05\%$ v/v Tween-20 in PBS). Plates were incubated with gradient dilutions of serum from control and Orai1/Orai3fl/fl Mb1-Cre/+ mice for 1 hr at 37 °C followed by incubation with alkaline phosphatase (AP) goat anti-rabbit IgG secondary antibody (SouthernBiotech, Cat#1030–04), goat anti-rabbit IgM (SouthernBiotech, Cat#1021–04) at room temperature for 2 hr. After the addition of substrate solution, absorption was measured at 405 nm using a Flexstation three plate reader (Molecular Devices).
## Cell preparation of mediastinal lymph nodes and bone marrow
Single-cell suspensions of mediastinal lymph nodes (medLNs) and bone marrow were grounded and passed through 70 µm cell strainer (BD, 22-363-548). Cells were treated with ACK buffer for 3 min and then washed and spun at 800 g for 5 min, resuspended in RPMI medium plus $2\%$ fetal bovine serum, and stained with antibodies as described below. Cells isolated from medLNs and bone marrow were counted with trypan blue, washed, and prepared for flow cytometry analysis in PBS containing $2\%$ FBS and 2 mM EDTA. For surface staining, cells were stained with fluorescently labeled antibodies at 4°C for 15 min in the dark, followed by Live/Dead Blue (Invitrogen, Cat# L23105) staining following the manufacturer’s instructions. Samples were acquired on an LSR Fortessa (BD Biosciences) and analyzed using FlowJo software (TreeStar, versions 9.3.2 and 10.5.3.). The list of antibodies used is as follows: B220- BV510 (RA3-6B2, Biolegend), CD138-PE (281–2, Biolegend), CD95-BV421 (Jo2, BD Bioscience), GL-7-Alexa-Fluor647 (GL7, Biolegend), CD4-APC-Cy7 (GK1.5, Biolegend), and CD38-PE-Cy7 (T10, Biolegend).
## Statistics
All statistical tests were conducted using GraphPad Prism 9 and data was presented as mean ± SEM. When comparing two groups the Student’s t-test was used. If greater than two groups were compared, then a One-way analysis of variance was used. For all results with normally distributed data, parametric statistical tests were used. When data were not normally distributed, non-parametric statistical tests were used. Data normality was determined using Prism 9.0. Biological replicates were defined as primary cells isolated from an individual mouse from their respective genotype. Statistically significant differences between groups were identified within the figures where *, **, ***, and **** indicating p-values of < 0.05, < 0.01, < 0.001, and <0.0001, respectively.
## Materials availability statement
All materials generated from this study including mice strains are available upon request from the lead PI (email: [email protected]).
## Funding Information
This paper was supported by the following grants:
## Data availability
All antibodies, cell lines, chemicals, mice strains, and sequences of primers and gRNA used in the study are listed in the key resources table. Source data files have been provided.
## References
1. Ahuja M, Schwartz DM, Tandon M, Son A, Zeng M, Swaim W, Eckhaus M, Hoffman V, Cui Y, Xiao B, Worley PF, Muallem S. **Orai1-Mediated antimicrobial secretion from pancreatic acini shapes the gut microbiome and regulates gut innate immunity**. *Cell Metabolism* (2017) **25** 635-646. DOI: 10.1016/j.cmet.2017.02.007
2. Akkaya M, Traba J, Roesler AS, Miozzo P, Akkaya B, Theall BP, Sohn H, Pena M, Smelkinson M, Kabat J, Dahlstrom E, Dorward DW, Skinner J, Sack MN, Pierce SK. **Second signals rescue B cells from activation-induced mitochondrial dysfunction and death**. *Nature Immunology* (2018) **19** 871-884. DOI: 10.1038/s41590-018-0156-5
3. Akkaya M, Pierce SK. **From zero to sixty and back to zero again: The metabolic life of B cells**. *Current Opinion in Immunology* (2019) **57** 1-7. DOI: 10.1016/j.coi.2018.09.019
4. Baba Y, Kurosaki T. **Role of calcium signaling in B cell activation and biology**. *Current Topics in Microbiology and Immunology* (2016) **393** 143-174. DOI: 10.1007/82_2015_477
5. Berry CT, May MJ, Freedman BD. **STIM- and Orai-mediated calcium entry controls NF-κB activity and function in lymphocytes**. *Cell Calcium* (2018) **74** 131-143. DOI: 10.1016/j.ceca.2018.07.003
6. Berry CT, Liu X, Myles A, Nandi S, Chen YH, Hershberg U, Brodsky IE, Cancro MP, Lengner CJ, May MJ, Freedman BD. **BCR-induced ca2+ signals dynamically tune survival, metabolic reprogramming, and proliferation of naive B cells**. *Cell Reports* (2020) **31**. DOI: 10.1016/j.celrep.2020.03.038
7. Bird GS, Putney JW, J.A. Kozak, J.W. Putney. *In: Calcium entry channels in non-excitable cells* (2018) 311-324. DOI: 10.1201/9781315152592
8. Bogeski I, Kummerow C, Al-Ansary D, Schwarz EC, Koehler R, Kozai D, Takahashi N, Peinelt C, Griesemer D, Bozem M, Mori Y, Hoth M, Niemeyer BA. **Differential redox regulation of Orai ion channels: A mechanism to tune cellular calcium signaling**. *Science Signaling* (2010) **3**. DOI: 10.1126/scisignal.2000672
9. Boothby M, Rickert RC. **Metabolic regulation of the immune humoral response**. *Immunity* (2017) **46** 743-755. DOI: 10.1016/j.immuni.2017.04.009
10. DeHaven WI, Smyth JT, Boyles RR, Bird GS, Putney JW. **Complex actions of 2-aminoethyldiphenyl borate on store-operated calcium entry**. *The Journal of Biological Chemistry* (2008) **283** 19265-19273. DOI: 10.1074/jbc.M801535200
11. Dolmetsch RE, Lewis RS, Goodnow CC, Healy JI. **Differential activation of transcription factors induced by Ca2+ response amplitude and duration**. *Nature* (1997) **386** 855-858. DOI: 10.1038/386855a0
12. Duan T, Du Y, Xing C, Wang HY, Wang RF. **Toll-like receptor signaling and its role in cell-mediated immunity**. *Frontiers in Immunology* (2022) **13**. DOI: 10.3389/fimmu.2022.812774
13. Durinck S, Moreau Y, Kasprzyk A, Davis S, De Moor B, Brazma A, Huber W. **BioMart and bioconductor: A powerful link between biological databases and microarray data analysis**. *Bioinformatics* (2005) **21** 3439-3440. DOI: 10.1093/bioinformatics/bti525
14. Durinck S, Spellman PT, Birney E, Huber W. **Mapping identifiers for the integration of genomic datasets with the R/bioconductor package biomart**. *Nature Protocols* (2009) **4** 1184-1191. DOI: 10.1038/nprot.2009.97
15. Eckstein M, Vaeth M, Aulestia FJ, Costiniti V, Kassam SN, Bromage TG, Pedersen P, Issekutz T, Idaghdour Y, Moursi AM, Feske S, Lacruz RS. **Differential regulation of Ca2+ influx by Orai channels mediates enamel mineralization**. *Science Signaling* (2019) **12**. DOI: 10.1126/scisignal.aav4663
16. Emrich SM, Yoast RE, Xin P, Arige V, Wagner LE, Hempel N, Gill DL, Sneyd J, Yule DI, Trebak M. **Omnitemporal choreographies of all five STIM/orai and ip3rs underlie the complexity of mammalian ca2+ signaling**. *Cell Reports* (2021) **34**. DOI: 10.1016/j.celrep.2021.108760
17. Emrich SM, Yoast RE, Fike AJ, Bricker KN, Xin P, Zhang X, Rahman ZSM, Trebak M. **The mitochondrial sodium/calcium exchanger NCLX (slc8b1) in B lymphocytes**. *Cell Calcium* (2022a) **108**. DOI: 10.1016/j.ceca.2022.102667
18. Emrich SM, Yoast RE, Trebak M. **Physiological functions of CRAC channels**. *Annual Review of Physiology* (2022b) **84** 355-379. DOI: 10.1146/annurev-physiol-052521-013426
19. Feske S, Giltnane J, Dolmetsch R, Staudt LM, Rao A. **Gene regulation mediated by calcium signals in T lymphocytes**. *Nature Immunology* (2001) **2** 316-324. DOI: 10.1038/86318
20. Feske S, Gwack Y, Prakriya M, Srikanth S, Puppel SH, Tanasa B, Hogan PG, Lewis RS, Daly M, Rao A. **A mutation in Orai1 causes immune deficiency by abrogating CRAC channel function**. *Nature* (2006) **441** 179-185. DOI: 10.1038/nature04702
21. Gammons J, Trebak M, Mancarella S. **Cardiac-specific deletion of orai3 leads to severe dilated cardiomyopathy and heart failure in mice**. *Journal of the American Heart Association* (2021) **10**. DOI: 10.1161/JAHA.120.019486
22. Geeraedts F, Goutagny N, Hornung V, Severa M, de Haan A, Pool J, Wilschut J, Fitzgerald KA, Huckriede A. **Superior immunogenicity of inactivated whole virus H5N1 influenza vaccine is primarily controlled by Toll-like receptor signalling**. *PLOS Pathogens* (2008) **4**. DOI: 10.1371/journal.ppat.1000138
23. Gwack Y, Srikanth S, Oh-Hora M, Hogan PG, Lamperti ED, Yamashita M, Gelinas C, Neems DS, Sasaki Y, Feske S, Prakriya M, Rajewsky K, Rao A. **Hair loss and defective T- and B-cell function in mice lacking Orai1**. *Molecular and Cellular Biology* (2008) **28** 5209-5222. DOI: 10.1128/MCB.00360-08
24. Healy JI, Dolmetsch RE, Lewis RS, Goodnow CC. **Quantitative and qualitative control of antigen receptor signalling in tolerant B lymphocytes**. *Novartis Foundation* (1998) **215** 137-144. DOI: 10.1002/9780470515525.ch10
25. Ho KH, Patrizi A. **Assessment of common housekeeping genes as reference for gene expression studies using RT-qpcr in mouse choroid plexus**. *Scientific Reports* (2021) **11**. DOI: 10.1038/s41598-021-82800-5
26. Hobeika E, Thiemann S, Storch B, Jumaa H, Nielsen PJ, Pelanda R, Reth M. **Testing gene function early in the B cell lineage in mb1-cre mice**. *PNAS* (2006) **103** 13789-13794. DOI: 10.1073/pnas.0605944103
27. Iwasaki A, Pillai PS. **Innate immunity to influenza virus infection**. *Nature Reviews. Immunology* (2014) **14** 315-328. DOI: 10.1038/nri3665
28. Kaufmann U, Shaw PJ, Kozhaya L, Subramanian R, Gaida K, Unutmaz D, McBride HJ, Feske S. **Selective orai1 inhibition ameliorates autoimmune central nervous system inflammation by suppressing effector but not regulatory T cell function**. *Journal of Immunology* (2016) **196** 573-585. DOI: 10.4049/jimmunol.1501406
29. Khalil AM, Cambier JC, Shlomchik MJ. **B cell receptor signal transduction in the GC is short-circuited by high phosphatase activity**. *Science* (2012) **336** 1178-1181. DOI: 10.1126/science.1213368
30. King LB, Freedman BD. **B-Lymphocyte calcium influx**. *Immunological Reviews* (2009) **231** 265-277. DOI: 10.1111/j.1600-065X.2009.00822.x
31. Lacruz RS, Feske S. **Diseases caused by mutations in ORAI1 and STIM1**. *Annals of the New York Academy of Sciences* (2015) **1356** 45-79. DOI: 10.1111/nyas.12938
32. Lian J, Cuk M, Kahlfuss S, Kozhaya L, Vaeth M, Rieux-Laucat F, Picard C, Benson MJ, Jakovcevic A, Bilic K, Martinac I, Stathopulos P, Kacskovics I, Vraetz T, Speckmann C, Ehl S, Issekutz T, Unutmaz D, Feske S. **Orai1 mutations abolishing store-operated ca2+ entry cause anhidrotic ectodermal dysplasia with immunodeficiency**. *The Journal of Allergy and Clinical Immunology* (2018) **142** 1297-1310. DOI: 10.1016/j.jaci.2017.10.031
33. Lunz V, Romanin C, Frischauf I. **Stim1 activation of Orai1**. *Cell Calcium* (2019) **77** 29-38. DOI: 10.1016/j.ceca.2018.11.009
34. Luo W, Weisel F, Shlomchik MJ. **B cell receptor and CD40 signaling are rewired for synergistic induction of the c-myc transcription factor in germinal center B cells**. *Immunity* (2018) **48** 313-326. DOI: 10.1016/j.immuni.2018.01.008
35. Matsumoto M, Fujii Y, Baba A, Hikida M, Kurosaki T, Baba Y. **The calcium sensors STIM1 and STIM2 control B cell regulatory function through interleukin-10 production**. *Immunity* (2011) **34** 703-714. DOI: 10.1016/j.immuni.2011.03.016
36. McCarl C-A, Picard C, Khalil S, Kawasaki T, Röther J, Papolos A, Kutok J, Hivroz C, LeDeist F, Plogmann K, Ehl S, Notheis G, Albert MH, Belohradsky BH, Kirschner J, Rao A, Fischer A, Feske S. **Orai1 deficiency and lack of store-operated Ca2+ entry cause immunodeficiency, myopathy, and ectodermal dysplasia**. *Journal of Allergy and Clinical Immunology* (2009) **124** 1311-1318. DOI: 10.1016/j.jaci.2009.10.007
37. McCarl C-A, Khalil S, Ma J, Oh-hora M, Yamashita M, Roether J, Kawasaki T, Jairaman A, Sasaki Y, Prakriya M, Feske S. **Store-Operated Ca2+ entry through Orai1 is critical for T cell-mediated autoimmunity and allograft rejection**. *The Journal of Immunology* (2010) **185** 5845-5858. DOI: 10.4049/jimmunol.1001796
38. McCarthy DJ, Chen Y, Smyth GK. **Differential expression analysis of multifactor RNA-seq experiments with respect to biological variation**. *Nucleic Acids Research* (2012) **40** 4288-4297. DOI: 10.1093/nar/gks042
39. Mootha VK, Lindgren CM, Eriksson K-F, Subramanian A, Sihag S, Lehar J, Puigserver P, Carlsson E, Ridderstråle M, Laurila E, Houstis N, Daly MJ, Patterson N, Mesirov JP, Golub TR, Tamayo P, Spiegelman B, Lander ES, Hirschhorn JN, Altshuler D, Groop LC. **PGC-1alpha-responsive genes involved in oxidative phosphorylation are coordinately downregulated in human diabetes**. *Nature Genetics* (2003) **34** 267-273. DOI: 10.1038/ng1180
40. Pacold ME, Brimacombe KR, Chan SH, Rohde JM, Lewis CA, Swier LJYM, Possemato R, Chen WW, Sullivan LB, Fiske BP, Cho S, Freinkman E, Birsoy K, Abu-Remaileh M, Shaul YD, Liu CM, Zhou M, Koh MJ, Chung H, Davidson SM, Luengo A, Wang AQ, Xu X, Yasgar A, Liu L, Rai G, Westover KD, Vander Heiden MG, Shen M, Gray NS, Boxer MB, Sabatini DM. **A Phgdh inhibitor reveals coordination of serine synthesis and one-carbon unit fate**. *Nature Chemical Biology* (2016) **12** 452-458. DOI: 10.1038/nchembio.2070
41. Prakriya M, Lewis RS. **Potentiation and inhibition of Ca (2+) release-activated Ca (2+) channels by 2-aminoethyldiphenyl borate (2-APB) occurs independently of IP (3) receptors**. *The Journal of Physiology* (2001) **536** 3-19. DOI: 10.1111/j.1469-7793.2001.t01-1-00003.x
42. Prakriya M, Lewis RS. **Store-Operated calcium channels**. *Physiological Reviews* (2015) **95** 1383-1436. DOI: 10.1152/physrev.00020.2014
43. Risso D, Schwartz K, Sherlock G, Dudoit S. **Gc-Content normalization for RNA-Seq data**. *BMC Bioinformatics* (2011) **12**. DOI: 10.1186/1471-2105-12-480
44. Robinson MD, McCarthy DJ, Smyth GK. **EdgeR: a bioconductor package for differential expression analysis of digital gene expression data**. *Bioinformatics* (2010) **26** 139-140. DOI: 10.1093/bioinformatics/btp616
45. Schneider CA, Rasband WS, Eliceiri KW. **Nih image to imagej: 25 years of image analysis**. *Nature Methods* (2012) **9** 671-675. DOI: 10.1038/nmeth.2089
46. Shanmughapriya S, Rajan S, Hoffman NE, Zhang X, Guo S, Kolesar JE, Hines KJ, Ragheb J, Jog NR, Caricchio R, Baba Y, Zhou Y, Kaufman BA, Cheung JY, Kurosaki T, Gill DL, Madesh M. **Ca2+ signals regulate mitochondrial metabolism by stimulating CREB-mediated expression of the mitochondrial Ca2+ uniporter gene MCU**. *Science Signaling* (2015) **8**. DOI: 10.1126/scisignal.2005673
47. Shaw PJ, Feske S. **Regulation of lymphocyte function by Orai and STIM proteins in infection and autoimmunity**. *The Journal of Physiology* (2012) **590** 4157-4167. DOI: 10.1113/jphysiol.2012.233221
48. Somasundaram A, Shum AK, McBride HJ, Kessler JA, Feske S, Miller RJ, Prakriya M. **Store-Operated CRAC channels regulate gene expression and proliferation in neural progenitor cells**. *The Journal of Neuroscience* (2014) **34** 9107-9123. DOI: 10.1523/JNEUROSCI.0263-14.2014
49. Subramanian A, Tamayo P, Mootha VK, Mukherjee S, Ebert BL, Gillette MA, Paulovich A, Pomeroy SL, Golub TR, Lander ES, Mesirov JP. **Gene set enrichment analysis: A knowledge-based approach for interpreting genome-wide expression profiles**. *PNAS* (2005) **102** 15545-15550. DOI: 10.1073/pnas.0506580102
50. Tang H, Wang H, Lin Q, Fan F, Zhang F, Peng X, Fang X, Liu J, Ouyang K. **Loss of IP3 receptor–mediated Ca2+ release in mouse B cells results in abnormal B cell development and function**. *The Journal of Immunology* (2017) **199** 570-580. DOI: 10.4049/jimmunol.1700109
51. Thomas PG, Brown SA, Yue W, So J, Webby RJ, Doherty PC. **An unexpected antibody response to an engineered influenza virus modifies CD8+ T cell responses**. *PNAS* (2006) **103** 2764-2769. DOI: 10.1073/pnas.0511185103
52. Thompson JL, Mignen O, Shuttleworth TJ. **The Arc channel -- an endogenous store-independent Orai channel**. *Current Topics in Membranes* (2013) **71** 125-148. DOI: 10.1016/B978-0-12-407870-3.00006-8
53. Trebak M, Putney JW. **Orai calcium channels**. *Physiology* (2017) **32** 332-342. DOI: 10.1152/physiol.00011.2017
54. Trebak M, Kinet JP. **Calcium signalling in T cells**. *Nature Reviews. Immunology* (2019) **19** 154-169. DOI: 10.1038/s41577-018-0110-7
55. Tsvilovskyy V, Solís-López A, Schumacher D, Medert R, Roers A, Kriebs U, Freichel M. **Deletion of Orai2 augments endogenous CRAC currents and degranulation in mast cells leading to enhanced anaphylaxis**. *Cell Calcium* (2018) **71** 24-33. DOI: 10.1016/j.ceca.2017.11.004
56. Vaeth M, Eckstein M, Shaw PJ, Kozhaya L, Yang J, Berberich-Siebelt F, Clancy R, Unutmaz D, Feske S. **Store-Operated Ca (2+) entry in follicular T cells controls humoral immune responses and autoimmunity**. *Immunity* (2016) **44** 1350-1364. DOI: 10.1016/j.immuni.2016.04.013
57. Vaeth M, Maus M, Klein-Hessling S, Freinkman E, Yang J, Eckstein M, Cameron S, Turvey SE, Serfling E, Berberich-Siebelt F, Possemato R, Feske S. **Store-operated ca2+ entry controls clonal expansion of T cells through metabolic reprogramming**. *Immunity* (2017a) **47** 664-679. DOI: 10.1016/j.immuni.2017.09.003
58. Vaeth M, Yang J, Yamashita M, Zee I, Eckstein M, Knosp C, Kaufmann U, Karoly Jani P, Lacruz RS, Flockerzi V, Kacskovics I, Prakriya M, Feske S. **Orai2 modulates store-operated calcium entry and T cell-mediated immunity**. *Nature Communications* (2017b) **8**. DOI: 10.1038/ncomms14714
59. Vaeth M, Feske S. **Nfat control of immune function: New frontiers for an abiding trooper**. *F1000Research* (2018) **7**. DOI: 10.12688/f1000research.13426.1
60. Vaeth M, Kahlfuss S, Feske S. **Crac channels and calcium signaling in T cell-mediated immunity**. *Trends in Immunology* (2020) **41** 878-901. DOI: 10.1016/j.it.2020.06.012
61. Vig M, DeHaven WI, Bird GS, Billingsley JM, Wang H, Rao PE, Hutchings AB, Jouvin MH, Putney JW, Kinet JP. **Defective mast cell effector functions in mice lacking the CRACM1 pore subunit of store-operated calcium release-activated calcium channels**. *Nature Immunology* (2008) **9** 89-96. DOI: 10.1038/ni1550
62. Wang L, Noyer L, Wang YH, Tao AY, Li W, Zhu J, Saavedra P, Hoda ST, Yang J, Feske S. **Orai3 is dispensable for store-operated ca2+ entry and immune responses by lymphocytes and macrophages**. *The Journal of General Physiology* (2022) **154**. DOI: 10.1085/jgp.202213104
63. Waters LR, Ahsan FM, Wolf DM, Shirihai O, Teitell MA. **Initial B cell activation induces metabolic reprogramming and mitochondrial remodeling**. *IScience* (2018) **5** 99-109. DOI: 10.1016/j.isci.2018.07.005
64. Ye Z, Shen Y, Jin K, Qiu J, Hu B, Jadhav RR, Sheth K, Weyand CM, Goronzy JJ. **Arachidonic acid-regulated calcium signaling in T cells from patients with rheumatoid arthritis promotes synovial inflammation**. *Nature Communications* (2021) **12**. DOI: 10.1038/s41467-021-21242-z
65. Yoast RE, Emrich SM, Trebak M. **The anatomy of native CRAC channel (S)**. *Current Opinion in Physiology* (2020a) **17** 89-95. DOI: 10.1016/j.cophys.2020.07.012
66. Yoast RE, Emrich SM, Zhang X, Xin P, Johnson MT, Fike AJ, Walter V, Hempel N, Yule DI, Sneyd J, Gill DL, Trebak M. **The native orai channel trio underlies the diversity of ca2+ signaling events**. *Nature Communications* (2020b) **11**. DOI: 10.1038/s41467-020-16232-6
67. Yoast RE, Emrich SM, Zhang X, Xin P, Arige V, Pathak T, Benson JC, Johnson MT, Abdelnaby AE, Lakomski N, Hempel N, Han JM, Dupont G, Yule DI, Sneyd J, Trebak M. **The mitochondrial ca2+ uniporter is a central regulator of interorganellar ca2+ transfer and NFAT activation**. *Journal of Biological Chemistry* (2021) **297**. DOI: 10.1016/j.jbc.2021.101174
68. Zhang SL, Kozak JA, Jiang W, Yeromin AV, Chen J, Yu Y, Penna A, Shen W, Chi V, Cahalan MD. **Store-Dependent and -independent modes regulating Ca2+ release-activated Ca2+ channel activity of human Orai1 and Orai3**. *The Journal of Biological Chemistry* (2008) **283** 17662-17671. DOI: 10.1074/jbc.M801536200
69. Zhang X, González-Cobos JC, Schindl R, Muik M, Ruhle B, Motiani RK, Bisaillon JM, Zhang W, Fahrner M, Barroso M, Matrougui K, Romanin C, Trebak M. **Mechanisms of STIM1 activation of store-independent leukotriene C4-regulated ca2+ channels**. *Molecular and Cellular Biology* (2013) **33** 3715-3723. DOI: 10.1128/MCB.00554-13
70. Zhang X, Zhang W, González-Cobos JC, Jardin I, Romanin C, Matrougui K, Trebak M. **Complex role of STIM1 in the activation of store-independent orai1/3 channels**. *The Journal of General Physiology* (2014) **143** 345-359. DOI: 10.1085/jgp.201311084
71. Zhang W, Zhang X, González-Cobos JC, Stolwijk JA, Matrougui K, Trebak M. **Leukotriene-C4 synthase, a critical enzyme in the activation of store-independent Orai1/Orai3 channels, is required for neointimal hyperplasia**. *The Journal of Biological Chemistry* (2015) **290** 5015-5027. DOI: 10.1074/jbc.M114.625822
72. Zhang X, Gueguinou M, Trebak M, J.A. Kozak, J.W. Putney. *In: Calcium entry channels in non-excitable cells* (2018) 197-214. DOI: 10.1201/9781315152592-11
73. Zhang X, Pathak T, Yoast R, Emrich S, Xin P, Nwokonko RM, Johnson M, Wu S, Delierneux C, Gueguinou M, Hempel N, Putney JW, Gill DL, Trebak M. **A calcium/camp signaling loop at the orai1 mouth drives channel inactivation to shape NFAT induction**. *Nature Communications* (2019) **10**. DOI: 10.1038/s41467-019-09593-0
74. Zhang X, Xin P, Yoast RE, Emrich SM, Johnson MT, Pathak T, Benson JC, Azimi I, Gill DL, Monteith GR, Trebak M. **Distinct pharmacological profiles of orai1, orai2, and orai3 channels**. *Cell Calcium* (2020) **91**. DOI: 10.1016/j.ceca.2020.102281
|
---
title: 'Intermittent Fasting: Benefits, Side Effects, Quality of Life, and Knowledge
of the Saudi Population'
journal: Cureus
year: 2023
pmcid: PMC9998115
doi: 10.7759/cureus.34722
license: CC BY 3.0
---
# Intermittent Fasting: Benefits, Side Effects, Quality of Life, and Knowledge of the Saudi Population
## Abstract
Introduction Intermittent fasting (IF) is an eating pattern that alternates between periods of fasting and eating. IF has shown many benefits for people who are obese and are trying to lose weight and attain a healthy lifestyle. The aim of our study was to evaluate the efficacy of IF and how it can be used as a daily lifestyle as well as to measure the knowledge of the IF diet among the Saudi population about its benefits, side effects, and life quality.
Method *For this* retrospective cross-sectional study, data about the common side effects, benefits, and the measurement of the quality of life were collected by a survey distributed using Google Forms. Microsoft Excel was used for the data analysis, with the data and results being mainly expressed as numbers and percentages.
Results Among the 147 individuals practicing the IF plan who were surveyed, male participants were more than females ($53.7\%$ vs $46.3\%$). The highest percentage of respondents were in the age group 18-35 years old, and 88 individuals ($59.9\%$) had a high body mass index (BMI). Duration of fasting varied from less than a month to three months in $70.8\%$, and $71.4\%$ of participants had undertaken IF several times. Side effects were headache ($61.3\%$), lethargy ($68\%$), mood swings ($57.8\%$), and lastly dizziness and polyuria ($55.8\%$ and $46.2\%$, respectively). Slightly more females expressed happiness than males ($86.8\%$ vs $83.6\%$).
Conclusion The IF diet is an efficient dietary plan for those aiming at a weight loss journey over a short duration, ranging from less than a month up to three months. Minimal side effects were found during fasting, being of different intensities, which did not need surgical or medical treatment. All in all, most of our respondents were pleased with their experience and saw excellent weight loss results using the IF diet.
## Introduction
Fasting is a practice followed by our ancestors for thousands of years; it has several aspects and is of different types such as for religious, cultural, or health reasons. Muslims fast during the month of Ramadan for a specific period during the day, which is similar to intermittent fasting (IF) [1]. This type of diet is based on an eating pattern that alternates between periods of fasting and eating. IF on the other hand mainly focuses on the timing of the meal, yet the quality of the food is also considered a factor, since losing weight is the primary goal. It commonly consists of a daily fast for 16 hours, a 24-hour fast on alternate days, or a fast two days per week on non-consecutive days. These types of diets aren't only used for obese people trying to lose weight, but also for people with cardiovascular risk factors and type 2 diabetes risk factors [2].
Obesity is a major medical problem that can lead to other significant risks such as metabolic diseases, mainly type 2 diabetes, hypertension, and stroke, and gastrointestinal tract diseases such as the fatty liver. According to WHO, obesity has nearly tripled between 1975 and 2016, with $39\%$ of adults aged 18 years and over being overweight in 2016, and $13\%$ obese [3]. Preventable obesity is still considered a major problem around the world.
IF provides many benefits for the obese and those trying to lose weight and attain a healthy lifestyle. It can reduce body fat and inflammation, and improve glucose metabolism [4]. Indeed, although IF is used for weight loss, ADF (alternate day fasting), which is a type of IF, has shown benefits in non-obese people also by lowering triacylglycerol, C reactive protein, and leptin, while increasing low-density lipoprotein particle size and adiponectin concentrations [5].
As much as IF is filled with benefits, there are mild side effects that can happen during fasting, which do not generally require medical or surgical treatment. Possible side effects can include dizziness, nausea, insomnia, headache, weakness, etc. [ 6]. Another study has shown low blood sugar (hypoglycemia) to be a side effect of IF [7].
The aim of our cross-sectional study was to observe the effectiveness of IF throughout the Saudi population. Furthermore, we wanted to evaluate behavioral characteristics and experiences of the IF diet from people's viewpoints on the side effects, benefits, and life quality.
## Materials and methods
Study design and period A retrospective cross-sectional study was carried out among individuals who had or are currently following the IF. In the four-week preparation period, we created our study title, reviewed the literature to gather information on typical side effects and how to measure the quality of life, and created a questionnaire. In the following four weeks, we completed our research, which included looking for the target group on social media, creating a Google Form for the survey, and testing the form by using a small sample of the findings to make sure they presented as we planned. We collected and examined our data for a further four weeks before writing our report.
Inclusion and exclusion criteria The study includes people who have ever practiced or are currently practicing IF and who can speak the Arabic language. The target population is in Saudi Arabia in all regions of the kingdom. The initial assessment questions excluded participants who had never engaged in IF. The sample size was calculated automatically using Google Forms. A total of 300 people were registered in the Google Form; 147 of them had practiced IF, and 153 were excluded.
Data collection tool We developed a questionnaire in Arabic to collect data about IF attitudes and quality of life (questionnaire in the Appendices section). The questionnaire was developed to meet the health-related quality of life (HRQOL) measures, which were translated into Arabic to be suitable for the participant's understanding of the question in our survey. The Centers for Disease Control and Prevention (CDC) HRQOL-4 measures had acceptable test-retest reliability and strong internal validity, which has been used by the CDC and its partners, for tracking population health status and HRQOL measures in states and communities [8]. The standard four-item set of healthy days’ core questions was developed by the CDC. The questionnaire consists of the following four main parts: (i) sociodemographic data, (ii) side effects assessment, (iii) participants' attitude toward IF assessment, and (iv) participants' quality of life assessment.
Data collection technique We conducted the survey in an electronic self-assessment format using Google Forms. We targeted IF groups on social media in Saudi Arabia for a time interval of four weeks through messages and direct contact with the group administrator and ensured that only people who practiced IF participated in the survey. One of the first questions in the survey was “Have you tried Intermittent fasting?” If the answer was no, then the participant is excluded from the survey. The data obtained from the survey were reviewed and automatically copied into a personal computer.
Data entry and analysis *The data* collected using the questionnaire was rearranged in Microsoft *Excel data* sheets. The data was mainly expressed as numbers and percentages. We used a chi-square test to evaluate participants’ perceived happiness, which also indicated the p-value for statistical significance. We also used a pie chart that displays the change in the participants’ body weight following their adoption of IF and to seek the perceived happiness of the participants' IF experience.
Ethical considerations This study was approved by the Bioethics Committee for Scientific and Medical Research at the University of Jeddah (Approval number UJ-REC-069). Individual consent was required prior to data collection, and it was stated on the questionnaire's front page that completing it signified consent to participate in the study. All information was kept private and was solely utilized for scientific studies. Furthermore, we made sure that parental consent was given to participants less than 18 years old.
## Results
A total of 147 respondents who practiced the IF diet plan participated to explore their IF experience and to determine its consequences on their overall health.
Table 1 shows the respondent characteristics of gender, age, and body mass index (BMI). Male participants were more than the females ($53.7\%$ vs $46.3\%$), and $75.5\%$ were in the age group of 18-35 years. According to the BMI ranges they were derived from CDC, $38.1\%$ of the participants were within the normal weight category, with $35.4\%$ being overweight, and $23.1\%$ obese.
**Table 1**
| Characteristics | Characteristics.1 | Number | Percentage |
| --- | --- | --- | --- |
| Gender | Male | 79.0 | 53.70% |
| Gender | Female | 68.0 | 46.30% |
| Age categories | <18 years | 4.0 | 2.40% |
| Age categories | 18-35 years | 111.0 | 75.50% |
| Age categories | 36-55 years | 27.0 | 18.40% |
| Age categories | 56-70 years | 5.0 | 3.40% |
| Age categories | > 70 years | | |
| BMI categories | Within Normal (18.5-24.9) | 56.0 | 38.1% |
| BMI categories | Overweight (25-29.9) | 52.0 | 35.40% |
| BMI categories | Obesity (30-39.9) | 34.0 | 23.1% |
| BMI categories | Morbid Obesity (≥40) | 2.0 | 1.40% |
| BMI categories | underweight (<18.5) | 3.0 | 2.00% |
As seen in Table 2, most of the respondent individuals ($70.8\%$) practiced the IF diet for a duration ranging from less than a month to three months, while $29.3\%$ of the individuals practiced the diet from more than three months to six months. According to the frequency, $28.6\%$ were for the first time experiencing IF, and for the second time was $29.2\%$. Lastly, $34\%$ have experienced IF more than three times. During the IF diet plan, close to two-thirds of the respondents ($67.3\%$) were performing physical exercise and $32.7\%$ didn't perform any physical exercise.
**Table 2**
| Practices during Intermittent fasting | Practices during Intermittent fasting.1 | Number | Percentage |
| --- | --- | --- | --- |
| Duration | <1 month | 47 | 32% |
| Duration | 1 month | 36 | 24.50% |
| Duration | > 1 month-3 months | 21 | 14.30% |
| Duration | >3 month-6 months | 43 | 29.20% |
| Frequency | Once | 42 | 28.60% |
| Frequency | Twice | 43 | 29.20% |
| Frequency | 3 times | 12 | 8.20% |
| Frequency | >3 times | 50 | 34% |
| Regular Performance of Physical Exercise | Yes | 99 | 67.30% |
| Regular Performance of Physical Exercise | No | 48 | 32.70% |
Figure 1 displays the weight changes in a sample who practiced IF throughout their lifetime so far, of whom 67 respondents ($45.6\%$) reported weight loss varying from 1 kg to less than 3 kg. Of these, 49 respondents ($33.3\%$) lost 3-5 kg and 19 ($12.9\%$) respondents lost 5-10 kg; four respondents ($2.7\%$) reported losing 10-15 kg, whereas only two respondents ($1.4\%$) reported losing more than 15 kg. Lastly, six respondents ($4.1\%$) reported weight gain.
**Figure 1:** *Change in weight as a response to intermittent fasting*
We asked responders about any symptoms they may have felt while following the IF diet allowed researchers to gauge the overall effectiveness of the diet, particularly within the first month (Table 3). As shown, the most frequently occurring symptoms were headache (mild $36.1\%$, moderate $17.7\%$, and severe $7.5\%$), lethargy (mild $27.9\%$, moderate $23.1\%$, and severe $17\%$), mood swings (mild $31.3\%$, moderate $16.3\%$, and severe $10.2\%$), and dizziness with a total of 55.8, and lastly, polyuria among $46.2\%$ respondents.
**Table 3**
| Symptoms | Mild | Moderate | Severe | Not Sure | Not at All |
| --- | --- | --- | --- | --- | --- |
| Headache | 53 (36.1%) | 26 (17.7%) | 11 (7.5%) | 17 (11.6%) | 40 (27.2%) |
| Mood swings | 46 (31.3%) | 24 (16.3%) | 15 (10.2%) | 19 (12.9%) | 43 (29.3%) |
| Palpitations | 24 (16.3%) | 18 (12.2%) | 11 (7.5%) | 16 (10.9%) | 78 (53.1%) |
| Fever | 10 (6.8%) | 4 (2.7%) | 4 (2.7%) | 10 (6.8%) | 119 (81%) |
| Flu | 18 (12.2%) | 3 (2%) | 7 (4.8%) | 11 (7.5%) | 108 (73.5%) |
| Low Blood Sugar | 18 (12.2%) | 15 (10.2%) | 9 (6.1%) | 26 (17.7%) | 79 (53.7%) |
| Lethargy | 41 (27.9%) | 34 (23.1%) | 25 (17%) | 4 (2.7%) | 43 (29.3%) |
| Constipation | 29 (19.7%) | 19 (12.9%) | 9 (6.1%) | 15 (10.2%) | 75 (51%) |
| Dizziness | 41 (27.9%) | 25 (17%) | 16 (10.9%) | 9 (6.1%) | 56 (38.1%) |
| Vomiting | 16 (10.9%) | 6 (4.1%) | 4 (2.7%) | 8 (5.4%) | 113 (76.9%) |
| Dehydration | 27 (18.4%) | 18 (12.2%) | 5 (3.4%) | 8 (5.4%) | 89 (60.5%) |
| Bloating | 17 (11.6%) | 10 (6.8%) | 9 (6.1%) | 15 (10.2%) | 96 (65.3%) |
| Polyuria | 29 (19.7%) | 25 (17%) | 14 (9.5%) | 15 (10.2%) | 64 (43.5%) |
Figure 2 demonstrates a pie-chart of perceived happiness for the IF experience of our respondents, with $85\%$ of them pleased about adopting the IF diet and $15\%$ of them weren't.
**Figure 2:** *Perceived happiness regarding intermittent fasting*
According to gender in Table 4, females were happier than males ($86.8\%$ vs $83.6\%$), and when it came to the age groups we found that the most perceived happiness was in the age group 36-55 years old ($92.6\%$). In the BMI category happiness perception among obese respondents had the highest percentage ($88.2\%$). These differences showed no statistical significance ($p \leq 0.05$).
**Table 4**
| Variables | Variables.1 | Yes | Yes.1 | No | No.1 | X^2 | P-Value |
| --- | --- | --- | --- | --- | --- | --- | --- |
| Variables | Variables | Number | % | Number | % | X^2 | P-Value |
| Gender | Male | 66 | 83.6% | 13 | 16.4% | 0.297 | 0.585 |
| Gender | Female | 59 | 86.8% | 9 | 13.2% | 0.297 | 0.585 |
| Age | <18 Years | 3 | 75% | 1 | 25% | 2.074 | 0.149 |
| Age | 18-35 Years | 91 | 82% | 20 | 18 | 2.074 | 0.149 |
| Age | 36-55 Years | 25 | 92.6% | 2 | 7.4% | 2.074 | 0.149 |
| Age | 56-70 Years | 4 | 80% | 1 | 20% | 2.074 | 0.149 |
| Age | > 70 Years | | | | | 2.074 | 0.149 |
| BMI categories | Within Normal | 49 | 87.5% | 7 | 12.5% | 3.489 | 0.061 |
| BMI categories | Overweight | 43 | 82.7% | 9 | 17.3% | 3.489 | 0.061 |
| BMI categories | Obesity | 30 | 88.2% | 4 | 11.7% | 3.489 | 0.061 |
| BMI categories | Morbid Obesity | 1 | 50% | 1 | 50% | 3.489 | 0.061 |
| BMI categories | Underweight | 2 | 66.6% | 1 | 33.3% | 3.489 | 0.061 |
## Discussion
IF has shown an increase in popularity as a method for weight loss and health optimization, and previous research has found promising results for its utility [10]. The current study collected data on IF to perceive its efficacy, side effects, and benefits, as well as the measurement of the resulting quality of life among the Saudi population adopting IF.
The findings of our study showed that the number of males was slightly higher than the females by $7.4\%$, while another study has shown that female participants tend to be a majority, and also practice IF longer than males [1]. When it comes to age category, most of our respondents were between the age of 18 and 35, which makes sense since this age group is adolescents/adults who have greater concern about their weight than any other age category. The relationship between weight loss and young adults was studied by Jessica Gokee LaRose et al., who stated that people <35 years old have a higher chance of gaining weight than older adults [11].
Most of our respondents had a high BMI; a combination of overweight ($35.4\%$), obese ($23.1\%$), and morbidly obese ($1.4\%$). Among the respondents, $59.9\%$ had a high BMI and $38.1\%$ were within the normal weight, with the remaining $2\%$ underweight. So, most of these participants practiced IF to lose weight and reduce their BMI number. A study by Maha H Alhussain et al. in Saudi Arabia compared the usage of IF for different purposes and found that almost half of the respondents ($47.59\%$) were trying to lose weight [12].
When it came to the duration, people who practiced the IF diet for less than a month and up to three months constituted $70.8\%$, while those who did it for a long term (>3-6 months) were $29.3\%$. Frequency-wise, the percentages were relatively similar to one another; the people who practiced it once were $28.6\%$, twice $29.3\%$, and thrice were only $8.2\%$ making it the least, whereas those adopting IF more than three times were $34\%$, accounting for the largest percentage.
Moreover, our statistics show that $67.3\%$ of the people who adopted the IF diet also performed physical exercise, whereas only $32.7\%$ practiced IF without any physical exercise. People felt that exercising while fasting isn’t a great option since you aren’t eating, yet a study has shown that exercising while you are on your fast increases adipose tissue lipolysis and peripheral fat oxidation, both of which act as energy producers for the body [13]. Alexandra Ferreira Vieira found that those who did moderate aerobic workouts during the fasting state showed a significant increase in fat oxidation [14].
A total of $94.5\%$ of our respondents have lost weight (ranging from 1 kg up to 10 kg), and $4.1\%$ had a significant decrease in weight (greater than or equal to 10 kg). Only six of our participants gained weight during their IF, which might be due to the quality of the food they chose to eat. Lora E. Burke et al. carried out a systematic review of 22 studies with self-monitoring; 15 out of the 22 studies involved dietary self-monitoring. Within these 22 studies, they found out there was more weight loss among people who self-monitored than among the less frequent self-monitoring group [15].
In spite of the fact that IF may be used by many people to cleanse their bodies or for weight loss, yet the diet does come with side effects, which you may experience within the first month of your fast. The most frequent side effects were headache, dizziness, polyuria, mood swings, and lastly lethargy. All of these symptoms had different intensity levels, from mild to severe. Headache is a common side effect seen in fasting in general, which is mainly due to hypoglycemia, and it is characterized as a diffuse and non-pulsatile headache. The headache is of mild to moderate intensity, it happens during the fasting period, and when you fast at least for 8 hours [16].
Lastly, $85\%$ of respondents were happy with their results, and only $15\%$ weren’t happy with their IF experience. As mentioned before, this may be due to their nutritional lifestyle routine or because they did not achieve the desired results.
The study has some limitations. First, it was distributed over social media groups with a special interest in the IF diet, thereby possibly exaggerating the positive attitude toward this diet. Second, the study's retrospective questions may have been influenced by how participants overall felt and how they remembered their symptoms.
## Conclusions
The majority of our participants tried the IF diet ranging from a month up to three months, suggesting that this type of diet is a short-term solution for people trying to lose weight. Almost all of the participants lost weight. Side effects were reported in our results, with a variation of intensities (mild, moderate, severe), particularly within the first month. All in all, the majority of our respondents were pleased with their experience and saw excellent weight loss results with the IF diet.
## References
1. Alnasser A, Almutairi M. **Considering intermittent fasting among Saudis: insights into practices**. *BMC Public Health* (2022) **22** 592. PMID: 35346130
2. Welton S, Minty R, O'Driscoll T, Willms H, Poirier D, Madden S, Kelly L. **Intermittent fasting and weight loss: systematic review**. *Can Fam Physician* (2020) **66** 117-125. PMID: 32060194
3. **World Health Organization, "Health Topics"**. (2022) **9** 2021
4. Malinowski B, Zalewska K, Węsierska A. **Intermittent fasting in cardiovascular disorders - an overview**. *Nutrients* (2019) **11** 673. PMID: 30897855
5. Varady KA, Bhutani S, Klempel MC. **Alternate day fasting for weight loss in normal weight and overweight subjects: a randomized controlled trial**. *Nutr J* (2013) **12** 146. PMID: 24215592
6. Grajower MM, Horne BD. **Clinical management of intermittent fasting in patients with diabetes mellitus**. *Nutrients* (2019) **11** 873. PMID: 31003482
7. Vasim I, Majeed CN, DeBoer MD. **Intermittent fasting and metabolic health**. *Nutrients* (2022) **14** 631. PMID: 35276989
8. **Measuring Healthy Days: Population Assessment of Health-related Quality of Life**. (2023)
9. **Defining Adult Overweight & Obesity**. (2023)
10. Wilhelmi de Toledo F, Grundler F, Sirtori CR, Ruscica M. **Unravelling the health effects of fasting: a long road from obesity treatment to healthy life span increase and improved cognition**. *Ann Med* (2020) **52** 147-161. PMID: 32519900
11. LaRose JG, Leahey TM, Hill JO, Wing RR. **Differences in motivations and weight loss behaviors in young adults and older adults in the National Weight Control Registry**. *Obesity (Silver Spring)* (2013) **21** 449-453. PMID: 23404944
12. Alhussain MH, Almarri DM, Arzoo S. **Study on the effect of intermittent fasting on body mass index, physical activity and sleep in adults**. *J Clin Diagn Res* (2021) **15** 0-7
13. Zouhal H, Saeidi A, Salhi A. **Exercise training and fasting: current insights**. *Open Access J Sports Med* (2020) **11** 1-28. PMID: 32021500
14. Vieira AF, Costa RR, Macedo RC, Coconcelli L, Kruel LF. **Effects of aerobic exercise performed in fasted v. fed state on fat and carbohydrate metabolism in adults: a systematic review and meta-analysis**. *Br J Nutr* (2016) **116** 1153-1164. PMID: 27609363
15. Burke LE, Wang J, Sevick MA. **Self-monitoring in weight loss: a systematic review of the literature**. *J Am Diet Assoc* (2011) **111** 92-102. PMID: 21185970
16. Soares AA, de Vasconcelos CAC, Silva-Néto RP. **Headaches and food abstinence: a review**. *J Clin Case Stu* (2018) **3**
|
---
title: 'Comparison of maternal and fetal health outcomes in the pandemic period of
covid-19 with the same last year duration in health centers of second largest city
of Iran: A population-based cohort study'
authors:
- Neda Davaryari
- Saeed Davaryar
- Adele Azarshab
- Mohammad Moein Vakilzadeh
- Veda Vakili
- Zahra Moazzeni
journal: Heliyon
year: 2023
pmcid: PMC9998123
doi: 10.1016/j.heliyon.2023.e14439
license: CC BY 4.0
---
# Comparison of maternal and fetal health outcomes in the pandemic period of covid-19 with the same last year duration in health centers of second largest city of Iran: A population-based cohort study
## Abstract
### Objective
The exact link between COVID-19 pandemic and different adverse outcomes of pregnancy remains unclear. Plus, large-scale research is lacking. In the present study, we aimed to compare the maternal and fetal health outcomes during the COVID-19 pandemic with the same last year duration in Iran.
### Design
Two retrospective cohorts (pre-COVID-19 and during COVID-19) were studied. The pre-COVID-19 cohort include pregnant women who had given birth between January 1, 2019 and December 31, 2019. The COVID-19 cohort, who had given birth between January 1, 2020 and December 31, 2020. The characteristics of pregnant women before COVID-19 and during COVID-19 pandemic were compared with Fisher's exact test. Univariate and multivariate log-binomial regression models were used to determine the risk ratios of the impacts of the COVID-19 pandemic on adverse pregnancy outcomes.
### Results
Among 128968 women showed that women who had given birth during the pandemic were more likely to be of young age, lower rates of alcohol consumption and smoking, lower weight gain, and higher rates of using synthetic milk for feeding neonates ($P \leq 0.05$). Also, the risks of preterm labor were high (cOR $95\%$ CI, 1.13 to 1.31; $p \leq 0.01$) and the risk of caesarian were low (cOR $95\%$ CI, 0.95 0.92 to 0.98; $p \leq 0.01$) among pregnant women who gave birth during the COVID-19 pandemic compared with those who gave birth before the pandemic.
### Conclusions
In summary, we found that during the COVID-19 pandemic there were the higher risks of preterm labor and lower risk of caesarean among pregnant women.
## Introduction
At the end of 2019, the discovery of a new coronavirus, severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2), was reported and linked to a series of pneumonia cases in Wuhan, China. The virus spread throughout the country rapidly, with several other countries following soon afterwards. The World Health Organization (WHO) named the disease Coronavirus disease 2019 (COVID-19), and declared it a global health emergency in March 2020 [1,2].
The risk of COVID-19 infection in pregnant women appears to be the same as in general population; however, it is speculated that immunological changes associated with pregnancy might leave the mother more susceptible to viral infections and their complications [3]. Indeed, some studies have demonstrated that COVID-19 infections are more severe in the final months of pregnancy [4,5]. In addition, a number of studies have indicated higher rates of adverse pregnancy outcomes in women with COVID-19, including preterm birth, cesarean section, and perinatal death [4,6,7]. COVID-19 pandemic has also seen numerous lockdown measures implemented occasionally in many countries, which could potentially disrupt maternal and neonatal health services and lead to adverse health outcomes [8].
As such, the exact link between COVID-19 pandemic and different adverse outcomes of pregnancy remain unclear, and more studies are needed to better elucidate the impacts of COVID-19 pandemic on pregnancy outcomes. Plus, large-scale research is lacking. In the present study, we aimed to compare the maternal and fetal health outcomes during the COVID-19 pandemic with the same last year duration in Iran.
## Study settings
In Iran, maternity care is state-funded and easily accessible for everyone. Pregnant women are closely followed by health care providers in designated health centers, run under the supervision of the regional medical university, for the duration of pregnancy and a few weeks after giving birth. Their data, including past medical histories, pregnancy status and pregnancy outcomes are recorded in detail by health care staff in an electronic archive. This study was approved by the Ethics Committee of Mashhad University of Medical Sciences (IR.MUMS.REC.1399.160).
## Study design and population
Two retrospective cohorts (pre-COVID-19 and during COVID-19) were studied. The pre-COVID-19 cohort consisted of 63618 pregnant women who had given birth between January 1, 2019 and December 31, 2019 (approximately one year before the beginning of the pandemic). There were 65350 pregnant women in the COVID-19 cohort, who had given birth between January 1, 2020 and December 31, 2020 (approximately one year after the beginning of the pandemic).
## Data collection
Data were collected from the electronic information system of health centers of Mashhad University of Medical Sciences. These data included related demographic characteristics (age, education, and occupation), past medical & social histories, pregnancy relate data (gravidity, parity, history of previous delivery problems), and health status (body mass index (BMI) before pregnancy, gestational weight gain (GWG)). Collected maternal outcomes of pregnancy were gestational diabetes, gestational hypertension, delivery mode, hospitalization, blood transfusion, and postpartum preeclampsia. Evaluated fetal pregnancy outcomes included preterm birth, low birth weight and macrosomia.
## Statistical analyses
The characteristics of pregnant women before COVID-19 and during COVID-19 pandemic were compared with Fisher's exact test. Univariate and multivariate log-binomial regression models were used to determine the crude risk ratios (cRRs) and adjusted risk ratios (aRRs) of the effects of the COVID-19 pandemic on poor pregnancy outcomes. Sensitivity analysis was done by fitting various models to survey the robustness of the estimation. Three models were fitted: the first model (A) was unadjusted; the second model (B) was adjusted for demographic characteristics (age, education, and occupation) and the third model (C) was based on model B, with further adjustments for pregnancy conditions (gravidity, parity, history of abortion), BMI before pregnancy, GWG, past medical and social history, history of low birth weight delivery, history of stillbirth, and history of delivery with congenital disorder. All analyses were performed using Statistical Package for the Social Sciences (SPSS) software version 26. A p value < 0.05 was considered statistically significant.
## Results
A total of 128968 pregnant women were included in this study, with a mean age of 30.56 (±6.71, SD). $71.9\%$ were unemployed and $74.4\%$ had a bachelor's degree or less. Characteristics and pregnancy histories of the study population are provided in Table 1 and Table 2 respectively. Pregnant women who experienced the COVID-19 pandemic were more likely to be younger ($30.1\%$ compared to $25.3\%$ in the pre-COVID-19 group), employed ($28.7\%$ compared to $25.9\%$ in the pre-COVID-19 group), and to have received a primary school education or less ($21.4\%$ compared to $20.8\%$ in the pre-COVID-19 group) when compared to women in the pre-pandemic group. Also, pregnant women during the COVID-19 pandemic were more likely to have a smaller number of gravidities, but a higher number of parities, compared with women in the pre-COVID-19 pandemic group. Regarding the pregnancy history, compared with women in the pre-COVID-19 pandemic group, pregnant women during the COVID-19 pandemic were more likely to have a positive history of low-birth-weight delivery ($5.6\%$ vs $6.1\%$, respectively), but were less likely to have a history of delivery with congenital disorder ($0.9\%$ vs $0.7\%$, respectively). There were not significant differences between cases in the pre-COVID-19 pandemic group and cases during the COVID-19 regarding the history of Diabetes Miletus and hypertension (all $p \leq 0.05$), but the rates of other chronic diseases were higher in COVID-19 pandemic group ($5.3\%$ vs $6.2\%$). Evaluating the social history of women showed that the rate of alcohol consumption and smoking were both significantly lower in pregnant women during the COVID-19 pandemic, compared with women in the pre-COVID-19 pandemic group ($p \leq 0.01$, and $$p \leq 0.002$$ respectively), but there were no significant differences in rates of addiction ($$p \leq 0.279$$). Weight gain during pregnancy was lower during the COVID-19 pandemic compared with women in the pre-COVID-19 pandemic group ($63.8\%$ vs $65.5\%$). Feeding neonates with synthetic milk was more common during COVID-19 pandemic, compared with pre-COVID-19 pandemic group ($$p \leq 0.001$$). Also, there was a significant difference between BMI index before pregnancy, There were no significant differences in other characteristics between the two groups (all $p \leq 0.05$).Table 1Characteristics of 128968 pregnant women before and during the COVID-19 pandemic. Table 1ItemsN/mean (SD)Pre-COVID-19 (N, %; mean, SD)COVID-19 (N, %; mean, SD)P valueMaternal age (years)30.56 (6.71)31.09 (6.59)30.04 (6.78)<0.01Maternal age (years)≤242598911012 (17.3)14977 (22.9)<0.0125–356731533445 (52.6)33870 (51.8)≥353566419161 (30.1)16503 (25.3)OccupationUnemployed94,75647521 (74.1)47235 (71.3)<0.01employed3561116576 (25.9)19035 (28.7)EducationPrimary school or less2778813481 (20.8)14307 (21.4)0.001Junior high school2989814692 (22.7)15206 (22.7)Senior high school4033320161 (31.1)20172 (30.1)Undergraduate or above3372216488 (25.4)17234 (25.8)Past medical historychronic disease69703252 (5.3)3718 (6.2)<0.01Diabetes Miletus624286 (0.5)56061 (0.6)0.213ahypertension605288 (0.5)318 (0.6)0.684aSocial historyAlcohol consumption9075 (0.1)15 (0.0)<0.01aAddiction16889 (0.2)79 (0.1)0.281aSmoker1126601 (1.1)525 (0.9)0.002aGestational weight gainAppropriate6282929721 (65.5)33108 (63.8)<0.01Inappropriate3439315647 (34.5)18746 (36.2)BMI before pregnancy, kg/m2Underweight (18.5)70583312 (6.2)3746 (6.7)0.002Normal (18.5–24.9)5196825429 (47.6)26539 (47.7)Overweight (25–29.9)3401016827 (31.5)17183 (30.9)Obese [30]160317847 (14.7)8184 (14.7)aFisher exact test. Table 2Pregnancy history of 128968 pregnant women before and during the COVID-19 pandemic. Table 2ItemsN/mean (SD)Pre-COVID-19 (N, %; mean, SD)COVID-19 (N, %; mean, SD)P valueGravidity2.52 (1.37)2.54 (1.36)2.51 (1.37)0.001Gravidity12740312867 (23.6)14536 (25.5)<0.0123629218043 (33.2)18249 [32]≥34776723501 (43.2)24266 (42.5)parity2.07 (1.13)2.05 (1.11)2.08 (1.14)<0.01Parity14609723117 (35.7)22980 (34.5)<0.0124742924182 (36.3)23247 (35.9)≥33774918311 (28.3)19438 (29.2)Positive History forLow-birth-weight delivery48952310 (5.6)2585 (6.1)0.001aabortion3225515886 (24.5)16369 (24.5)0.856Stillbirth1250634 (1.6)616 (1.5)0.459adelivery with congenital disorder642356 (0.9)286 (0.7)0.003aGestational weight gainAppropriate6282929721 (65.5)33108 (63.8)<0.01Inappropriate3439315647 (34.5)18746 (36.2)Child feedingMothers milk12529861950 [97]63348 (96.7)0.001Mothers and synthetic milk33741616 (2.5)1758 (2.7)Synthetic milk968301 (0.5)397 (0.6)aFisher exact test.
Assessing the pregnancy outcomes demonstrated that the prevalence of caesarean sections was lower during the COVID-19 pandemic period compared with women prior to the pandemic ($42.0\%$ vs $41.3\%$; $$p \leq 0.007$$). During the COVID-19 pandemic period, the occurrence of post-partum preeclampsia was greater compared to women before the pandemic, with a difference of $0.2\%$ ($1.1\%$ versus $0.9\%$), which was found to be statistically significant ($$p \leq 0.045$$). However, there were no significant differences in other pregnancy outcomes between the two periods. ( $p \leq 0.05$, Table 3). The prevalence of preterm birth and macrosomia were higher during the COVID-19 pandemic period compared with cases prior to the pandemic ($4.2\%$ vs $4.7\%$, $p \leq 0.01$; and $4.3\%$ vs $4.6\%$, respectively; $$p \leq 0.008$$). However, prevalence of low birth weight was not significantly different during the COVID-19 pandemic compared with the pre-pandemic period ($$p \leq 0.868$$).Table 3Pregnancy outcomes before and during the COVID-19 pandemic. Table 3outcomesPre-COVID-19 (N, %; mean, SD)COVID-19 (N, %; mean, SD)P valueAdverse maternal outcomesGestational diabetes1707 (3.8)1911 (4.0)0.180aGestational hypertension607 (0.9)594 (0.9)0.354aCaesarean section*25771 [42]24649 (41.3)0.007hospitalization5893 (9.1)5963 (8.9)0.256aBlood transfusion77 (0.1)89 (0.1)0.485aPostpartum preeclampsia574 (0.9)622 (1.1)0.045aAdverse fetal outcomesPreterm birth2695 (4.2)3142 (4.7)<0.01aLow birth weight4934 (7.6)5077 (7.6)0.868aMacrosomia2784 (4.3)3058 (4.6)0.015aaFisher exact test.
In our log-binomial regression models (Table 4), although risk for Caesarean section in univariate model was 0.97 times ($95\%$ CI, 0.92 to 0.98; $$p \leq 0.003$$) during the COVID-19 pandemic compared with pre-COVID-19, but after adjusting for all confounding factors, the risk was not significantly difference ($$p \leq 0.467$$). The risk of macrosomia during the COVID-19 pandemic compared with pre-COVID-19 cases was increased by 1.08 times ($95\%$ CI, 1.01 to 1.16; $$p \leq 0.023$$) when adjusting for age, occupation and education (Table 3). The risk of premature birth was increased by 1.2 times in all three models ($95\%$ CI, 1.10 to 1.31; $p \leq 0.01$).Table 4The influence of the COVID-19 pandemic on pregnancy outcomes. Table 4OutcomesModel AModel BModel CcOR ($95\%$ CI)P valueaOR ($95\%$ CI)P valueaOR ($95\%$ CI)P valueAdverse maternal outcomesGestational diabetes0.97 (0.89–1.05)0.4751.04 (0.95–1.12)0.3441.02 (0.94–1.11)0.545Gestational hypertension0.88 (0.75–1.02)0.1090.94 (0.81–1.10)0.5030.88 (0.75–1.03)0.115Caesarean section0.95 (0.92–0.98)0.0030.99 (0.96–1.02)0.9030.98 (0.95–1.02)0.467hospitalization1.00 (0.95–1.05)0.9000.98 (0.93–1.04)0.6080.97 (0.92–1.03)0.393Blood transfusion1.41 (0.91–2.19)0.1211.43 (0.92–2.23)0.1061.41 (0.90–2.19)0.129Postpartum preeclampsia0.96 (0.82–1.12)0.6310.99 (0.84–1.16)0.9260.99 (0.84–1.16)0.908Adverse fetal outcomesPreterm birth1.13 (1.13–1.31)<0.011.15 (1.15–1.33)<0.011.20 (1.10–1.31)<0.01Low birth weight1.02 (0.96–1.08)0.3681.04 (0.98–1.10)0.1931.03 (0.97–1.09)0.193Macrosomia1.07 (0.99–1.14)0.0571.08 (1.01–1.16)0.0231.06 (0.99–1.14)0.088Model A: a univariate model without controlling for any confounding factors. Model B: controls for demographic characteristics (age, occupation and education).Model C: based on model B, supplemented to control for gravidity, parity, history of abortion, BMI before pregnancy, gestational weight gain, history of chronic disease, History of Low-birth-weight delivery, History of Stillbirth, History of delivery with congenital disorder, history of Diabetes Miletus, history of hypertension, Alcohol consumption, Addiction and smoking).aOR, adjusted odds ratio; BMI, body mass index; cOR, crude odds ratio.
## Summary of the findings
Results of this cohort study on 128968 women with a focus on secondary impacts of the COVID-19 pandemic on pregnancy outcomes showed that women who had given birth during the pandemic were more likely to be of young age, employed, less educated, have a smaller number of gravidity and higher number of parity, have a positive history of Low-birth-weight, less likely to have a history of delivery with congenital disorder, have higher rates of other chronic diseases, lower rates of alcohol consumption and smoking, lower weight gain, and higher rates of using synthetic milk for feeding neonates. There were the higher risks of preterm labor ($13\%$ up to $15\%$) and lower risk of caesarian ($5\%$), among pregnant women who gave birth during the COVID-19 pandemic compared with those who gave birth before the pandemic.
## Strengths and limitations
The strengths of this study include its cohort-study design, large sample size, and use of a variety variable to detect the impacts of the COVID-19 pandemic on pregnancy outcomes. Moreover, log-binomial regression models were utilized to assess the effect of either a policy intervention or a natural alteration. This multicenter cohort study was carried out in a major city of Iran. Nevertheless, there were certain limitations to our study. Firstly, it was a retrospective investigation. Moreover, we lacked access to data beyond the scope of pregnancy outcomes.
## Comparison with other studies
Although there are conflicting reports of pregnancy outcomes during the COVID-19, but to our knowledge, this is the initial multicenter cohort study in Iran that concentrates on the secondary consequences of the COVID-19 pandemic on pregnancy outcomes. Some studies have demonstrated higher rates of gestational diabetes, hypertension, premature rupture of membrane, and admission to intensive care unit during the pandemic, compared with pre-COVID-19 period [[9], [10], [11], [12]]; while other studies have not demonstrated any such changes. A systematic review and meta-analysis by Chmielewska et al. did not find any significant increase in the prevalence of gestational diabetes, and hypertensive disorders of pregnancy during COVID-19 pandemic [13]. Yang et al. conducted a study where they reported a decrease in the unadjusted likelihood of preterm birth during the COVID-19 pandemic period in comparison to the pre-pandemic period [14]. A recent study by Du et al. found that there was a heightened chance of premature rupture of membranes and fetal distress during the COVID-19 pandemic. However, no significant connections were detected between the pandemic and other pregnancy outcomes [15].
One of the conflicting issues is the impact of COVID-19 pandemic on rates of preterm birth (PTB). In a systematic review and meta-analysis conducted by Yang et al., which analyzed the impact of COVID-19 on pregnancy and neonatal outcomes across 43 studies comprising of 986,466 women during the pandemic period and 8,716,000 women in the pre-pandemic period, it was found that the unadjusted odds of preterm birth (PTB) were lower during the pandemic period as compared to the pre-pandemic period (pooled unadjusted odds ratio [OR] 0.95, $95\%$ confidence interval [CI] 0.93–0.98). However, when adjusted estimates were taken into consideration, which were reported by five of the studies with different factors adjusted, the pooled analysis did not reveal any significant differences in the odds of PTB during the pandemic period [14]. In the present study unadjusted odds ratio of preterm birth during COVID-19 pandemic was 1.13 ($95\%$ CI 1.13–1.31); also, the risk of premature birth was increased significantly even in adjusted models ($95\%$ CI, 1.10 to 1.31; $p \leq 0.01$). Also, the evaluation of the Yang review in birthweight of studies, shows that There was no difference in the odds of low birth weight, very low birthweight, or extremely low birth weight during COVID-19 pandemic [14]. In our study, we evaluate birth weight in two categories including macrosomia and low birth weight. Significant odds ratio was noted for macrosomia when adjusted by mother's age, occupation, and education. But other crude or adjusted models of odds ratio for low birth weight and macrosomia were not different between pre COVID-19 period compare with during COVID-19 period.
In retrospective study on 2503 pregnant women in a hospital in Tehran, Iran; Ranjbar et al. showed that the rate of preterm birth and low births weight had decreased during COVID-19 pandemic period [16]. Although their study was performed just in single center and did not evaluate the odds ratio of the pregnancy outcomes, but it was the only similar study that had been performed in the same region as of the current study.
In Du M et al. study, the prevalence of caesarean sections among pregnant women during the COVID-19-pandemic was increased. Also, they found that there was a greater proportion of women aged ≥35 years in the COVID-19 cohort [15]. In the present study, the crude odds ratio of caesarean sections had decreased only in unadjusted model had significant (cOR 0.95, $95\%$ CI 0.92–0.98; $$p \leq 0.003$$) particularly those in the early stages of pregnancy, were studied during the COVID-19 pandemic period. This might be related to the recent incentive policies of childbearing in Iran.
Ceulemans et al. study on impact of the COVID-19 pandemic on breastfeeding, shows that COVID-19 pandemic has not affected breastfeeding practices, nor breastfeeding cessation [17]. In our study, feeding neonates with synthetic milk was more common during COVID-19 pandemic compared with pre-COVID-19 pandemic group.
Lebel and her collogues study on prevalence of Alcohol and substance use in pregnancy during the COVID-19 pandemic demonstrated that alcohol and substance use among pregnant Women during the pandemic was lower or comparable with overall rates of previous estimates in other samples [18]. Although in our study the number of pregnant women who drank alcohol or smoked were low both before and during COVID-19 pandemic based on self-reports, but the rates of alcohol drinking and smoking were significantly lower during pandemic ($p \leq 0.01$, and $$p \leq 0.002$$ respectively). Also, in our study there were no significant differences in addiction before and during the COVID-19 pandemic ($$p \leq 0.281$$).
In a systematic review and meta-analysis of Chmielewska and coauthors on 40 studies, the results showed no significant effects of COVID-19 pandemic on prevalence of maternal gestational diabetes, and hypertensive disorders of pregnancy [13]. In the present study, crude and adjusted odds ratio of both gestational diabetes and gestational hypertension did not significantly differ during COVID-19 pandemic compared with pre COVID-19 pandemic ($p \leq 0.05$).
The pandemic process negatively affected the daily routine and lead to sedentary life style. The sedentary lifestyle associated with obesity and get weight. In addition to, the studies indicated that obesity independently increased the risk of fetal loss [19]. However, in our study the stillbirth rates were shown as similar between the periods of study. It could be explained by the short period duration of study which could demonstrated the associated results later.
## Conclusions
In summary, we found that there were more pregnant women who had given birth during the pandemic were more likely to have lower weight gain, lower rates of alcohol consumption and smoking, and. Also, there were the higher risks of preterm labor and lower risk of caesarian among pregnant women who gave birth during the COVID-19 pandemic compared with those who gave birth before the pandemic.
## Author contribution statement
Zahra Moazzeni: Conceived and designed the experiments; Contributed reagents, materials, analysis tools or data.
Neda Davaryari: Conceived and designed the experiments; Performed the experiments; Wrote the paper.
Saeed Davaryar: Performed the experiments; Analyzed and interpreted the data; Wrote the paper.
Adele Azarshab: Performed the experiments; Wrote the paper.
Mohammad Moein Vakilzadeh: Analyzed and interpreted the data; Wrote the paper.
Veda Vakili: Contributed reagents, materials, analysis tools or data; Wrote the paper.
## Funding statement
Neda Davaryari was supported by Mashhad University of Medical Sciences, Mashhad, Iran [code: 990124].
## Data availability statement
Data will be made available on request.
## Declaration of interest's statement
The authors declare no competing interests.
## References
1. Organization W.H.. (2020)
2. Huang C.. **Clinical features of patients infected with 2019 novel coronavirus in Wuhan, China**. *Lancet* (2020) **395** 497-506. PMID: 31986264
3. Akhtar H.. **COVID-19 (SARS-CoV-2) infection in pregnancy: a systematic review**. *Gynecol. Obstet. Invest.* (2020) **85** 295-306. PMID: 32728006
4. Knight M.. **Characteristics and outcomes of pregnant women admitted to hospital with confirmed SARS-CoV-2 infection in UK: national population based cohort study**. *BMJ* (2020) **369** m2107. PMID: 32513659
5. Mullins E.. **Coronavirus in pregnancy and delivery: rapid review**. *Ultrasound Obstet. Gynecol.* (2020) **55** 586-592. PMID: 32180292
6. Chen H.. **Clinical characteristics and intrauterine vertical transmission potential of COVID-19 infection in nine pregnant women: a retrospective review of medical records**. *Lancet* (2020) **395** 809-815. PMID: 32151335
7. Allotey J.. **Clinical manifestations, risk factors, and maternal and perinatal outcomes of coronavirus disease 2019 in pregnancy: living systematic review and meta-analysis**. *BMJ* (2020) **370** m3320. PMID: 32873575
8. Kc A.. **Effect of the COVID-19 pandemic response on intrapartum care, stillbirth, and neonatal mortality outcomes in Nepal: a prospective observational study**. *Lancet Global Health* (2020) **8** e1273-e1281. PMID: 32791117
9. Justman N.. **Lockdown with a price: the impact of the COVID-19 pandemic on prenatal care and perinatal outcomes in a tertiary care center**. *Isr. Med. Assoc. J.* (2020) **22** 533-537. PMID: 33236549
10. Gu X.X.. **How to prevent in-hospital COVID-19 infection and reassure women about the safety of pregnancy: experience from an obstetric center in China**. *J. Int. Med. Res.* (2020) **48**
11. Kugelman N.. **Changes in the obstetrical emergency department profile during the COVID-19 pandemic**. *J. Matern. Fetal Neonatal Med.* (2020) 1-7. PMID: 29804488
12. Goyal M.. **The effect of the COVID-19 pandemic on maternal health due to delay in seeking health care: experience from a tertiary center**. *Int. J. Gynaecol. Obstet.* (2021) **152** 231-235. PMID: 33128794
13. Chmielewska B.. **Effects of the COVID-19 pandemic on maternal and perinatal outcomes: a systematic review and meta-analysis**. *Lancet Global Health* (2021) **9** e759-e772. PMID: 33811827
14. Yang J.. **COVID‐19 pandemic and population‐level pregnancy and neonatal outcomes in general population: a living systematic review and meta‐analysis (Update# 2: november 20, 2021)**. *Acta Obstet. Gynecol. Scand.* (2022) **101** 273-292. PMID: 35088409
15. Du M.. **Association between the COVID-19 pandemic and the risk for adverse pregnancy outcomes: a cohort study**. *BMJ Open* (2021) **11** e047900
16. Ranjbar F.. **Changes in pregnancy outcomes during the COVID-19 lockdown in Iran**. *BMC Pregnancy Childbirth* (2021) **21** 1-6. PMID: 33388035
17. Ceulemans M.. **SARS-CoV-2 infections and impact of the COVID-19 pandemic in pregnancy and breastfeeding: results from an observational study in primary care in Belgium**. *Int. J. Environ. Res. Publ. Health* (2020) **17** 6766
18. Kar P.. **Alcohol and substance use in pregnancy during the COVID-19 pandemic**. *Drug Alcohol Depend.* (2021) **225**
19. Woolner A.M., Bhattacharya S.. **Obesity and stillbirth**. *Best Pract. Res. Clin. Obstet. Gynaecol.* (2015) **29** 415-426. PMID: 25457855
|
---
title: Utilization of Hypolipidemic Drugs, Patterns, and Factors Affecting Dyslipidemia
Among Type 2 Diabetes Mellitus at a Tertiary Care Teaching Hospital in South India
journal: Cureus
year: 2023
pmcid: PMC9998133
doi: 10.7759/cureus.34748
license: CC BY 3.0
---
# Utilization of Hypolipidemic Drugs, Patterns, and Factors Affecting Dyslipidemia Among Type 2 Diabetes Mellitus at a Tertiary Care Teaching Hospital in South India
## Abstract
Background The prevalence of dyslipidemia is higher in type 2 diabetes mellitus (T2DM) and hypolipidemic drugs like statins are effective for the primary and secondary prevention of cardiovascular events. Most of the patients with type 2 diabetes have a mixed type of dyslipidemia. This study aimed to evaluate the utilization of hypolipidemic drugs, patterns, and factors affecting dyslipidemia in T2DM participants.
Methods This cross-sectional observational study was approved by the institutional ethics committee (IEC) of the Vydehi Institute of Medical Sciences and Research Center. It was conducted for a period of one year from July 2021 to June 2022. Participants with T2DM visiting the Department of General Medicine and Endocrinology were enrolled after obtaining informed consent. Demographic details, medication history, and laboratory data were recorded in case report form and statistical measures were applied.
Results Out of 237 participants enrolled in the study, the predominance ($$n = 133$$, $56\%$) was males. The mean age of the study population was 47.92±9.17 years, and the mean duration of diabetes was 6.8±5.3 years. Out of the total participants, 164 ($69\%$) had deranged lipid profiles. Out of them, 129 ($78.65\%$) were on hypolipidemic drugs. Regarding drug utilization, 122 ($94.6\%$) received statins either rosuvastatin ($54\%$) or atorvastatin ($40\%$). In the deranged lipid profiles pattern, $24\%$ [58] participants had one abnormal lipid parameter and the majority $70\%$ [166] had combined lipid profile abnormality. Factors like increased BMI were significantly associated with dyslipidemia ($$p \leq 0.004$$). Utilization of hypolipidemic drugs was also significantly associated with the control of dyslipidemia ($p \leq 0.001$). It was observed that participants who were not on lipid-lowering drugs had 5.38 times more chance of dyslipidemia (OR=5.38; CI=2.82-10.28; $p \leq 0.001$).
Conclusion A high prevalence of dyslipidemia was observed among diabetic patients. Statins were the most prescribed drug in the study. BMI and lack of pharmacotherapy were found to have a statistically significant association with dyslipidemia in diabetic patients.
## Introduction
Diabetic dyslipidemia is characterized by elevated plasma triglyceride levels, low high-density lipoprotein cholesterol (HDL-C) levels, and an increase in several small dense forms of low-density lipoprotein (LDL) particles which are highly atherogenic [1]. According to ICMR guidelines for the management of type 2 diabetes mellitus (T2DM), targets for lipid control in individuals are total cholesterol less than 200 mg/dL, HDL cholesterol more than 40 mg/dL for men, and more than 50 mg/dL for women, LDL cholesterol less than 100 mg/dL, triglycerides less than 150 mg/dL, respectively [2]. Dyslipidemia and hypertension are definite risk factors for atherosclerotic cardiovascular disease (ASCVD). ASCVD is also defined as coronary heart disease (CHD), cerebrovascular disease, or peripheral arterial disease and is the leading cause of morbidity and mortality in diabetics [3].
Besides antidiabetic drugs, lifestyle modification and hypolipidemic drugs are effective [3,4]. Statin is one such medication that has demonstrated efficacy in extensive studies and is advised in current guidelines for individuals with T2DM. In both primary and secondary prevention, statin showed a consistent reduction in mortality and major adverse cardiovascular events by (Risk ratio=0.80; $95\%$ CI=0.77-0.83) for every one mmol/L (38.7 mg/dL) decrease in LDL-C [5]. American Diabetes Association (ADA) published medical care standards, including recommendations for starting lipid-lowering therapy and prescribing moderate-intensity statins for those with no additional risk factors. High-intensity statins are recommended for those with either CVD risk factors or overt CVD with consideration of 10-year ASCVD risk [3]. The American College of Cardiology/American Heart Association (ACC/AHA) clinical practice guidelines advise prescribing moderate-intensity statins to patients aged 40 to 75 who have T2DM and those with LDL-C more than 70 mg/dL without considering 10-year ASCVD risk [6]. According to ADA high-intensity statin therapy like atorvastatin (40-80mg) and rosuvastatin (20-40mg) lowers LDL cholesterol by $50\%$ and moderate-intensity statin therapy like atorvastatin (10-20mg), rosuvastatin (5-10mg), simvastatin (20-40mg), pravastatin (40-80mg), lovastatin (40mg), fluvastatin extended-release (80mg), pitavastatin (1-4mg) lowers LDL cholesterol by $30\%$-$49\%$ [3].
Previous studies on diabetic dyslipidemia in India found that dyslipidemia was more in females and the age group 45-54 years. They also concluded that the majority $44.2\%$ of participants suffer from a mixed type of dyslipidemia [7,8]. They also found that in India the prevalence of dyslipidemia is about $86\%$ [8]. Only a few studies were conducted in the southern part of India. This current study focuses on statin utilization in people with diabetic dyslipidemia.
## Materials and methods
The present study was a cross-sectional observational study. The study was conducted by the Department of Pharmacology in collaboration with the Department of Endocrinology and the Department of General Medicine at Vydehi Institute of Medical Sciences and Research Center, Bengaluru. The duration of the study was one year from July 2021 to June 2022. The study was conducted after receiving approval from the institutional ethics committee, under the registration number VIEC/PG/APP/$\frac{015}{2020}$-21. Participants were screened for inclusion and exclusion criteria. Both male and female patients with type 2 diabetes irrespective of lipid-lowering therapy between 18 and 60 years attending the OPD and consenting to the study were included. The exclusion criteria included type 1 diabetes mellitus patients, gestational diabetes mellitus, and inpatients. The study objectives and process were explained to the participants or their relatives in their language. Participants were asked to read and sign an informed consent form. The demographic details, medical history, personal history, medication history, and laboratory data of all participants included in the study were recorded on Case Report Form (CRF). Participants were divided into two groups based on lipid profile level. Group one was participants with high lipid levels than the normal range, and group two was participants with low lipid levels than the normal range. Drug utilization was evaluated. Several factors like age, gender, body mass index (BMI), duration of disease, comorbidities, drug utilization of hypolipidemic drugs, family history, and HbA1c were compared between the two groups. Finally, all details from the CRF form were recorded on an excel sheet and statistical measures were applied. The following biological references were considered low lipid levels. Total cholesterol less than 200 mg/dL, triglycerides less than 150 mg/dL, HDL cholesterol more than 40 mg/dL, LDL cholesterol less than 100 mg/dL, VLDL between 2 and 30 mg/dL. Any one or more than one lipid parameter higher than the normal lipid range was considered high lipid levels. The sample size was calculated considering a confidence interval of $95\%$, precision of $5\%$, and power of study at $80\%$.
Data collected were entered into a Microsoft office excel sheet. Demographic details were subjected to descriptive statistical analysis and were expressed as mean SD and percentages. Drug utilization patterns were analyzed and expressed in frequencies and percentages. The categorical variable was compared using the chi-square test and odds ratio. Association between all categorical variables and lipid profile was observed and a P-value of less than or equal to 0.05 was considered statistically significant.
## Results
Demographic details The total number of participants in the study was 237. Out of 237 participants, $56\%$ [133] were males. Out of the males, $72\%$ [96] were found to have high lipid levels. The mean age (± SD) of the participants was 47.91 ± 9.17 years. Most participants were in the age group of more than 30 years. The mean BMI (± SD) was 25.27 ± 4.08 kg/m2. Most participants had a family history of T2DM. The mean duration (± SD) of the disease was 6.8 ± 5.3 years. Most participants had a duration of disease of less than six years. And in the same category, $75\%$ [91] were found to have dyslipidemia. The mean HbA1c (± SD) was 8.9 ± $2.35\%$. The results are presented in Table 1.
**Table 1**
| Variables | Category | Mean ± SD | n (%) | Dyslipidemia | Dyslipidemia.1 |
| --- | --- | --- | --- | --- | --- |
| Variables | Category | Mean ± SD | n (%) | High lipid levels n (%) | Low lipid levels n (%) |
| Sex | Male | - | 133 (56) | 96 (72) | 37 (28) |
| Sex | Female | - | 104 (44) | 68 (65) | 36 (35) |
| Age | < 30 years | 47.91 ± 9.17 | 7 (3) | 7 (100) | 0 (0) |
| Age | > 30 years | 47.91 ± 9.17 | 230 (97) | 157 (68) | 73 (32) |
| Occupation | Merchants | - | 51 (22) | 40 (78) | 11 (22) |
| Occupation | Farmer | - | 20 (8) | 15 (75) | 5 (25) |
| Occupation | Employee | - | 67 (28) | 42 (63) | 25 (37) |
| Occupation | Homemakers | - | 99 (42) | 67 (68) | 32 (32) |
| BMI | < 24.9 kg/m2 | 25.27 ± 4.08 | 106 (45) | 62 (58) | 44 (42) |
| BMI | 25-29.9 kg/m2 | 25.27 ± 4.08 | 102 (43) | 81 (79) | 21 (21) |
| BMI | ≥ 30 kg/m2 | 25.27 ± 4.08 | 29 (12) | 21 (72) | 8 (28) |
| Family history of diabetes mellitus | Yes | - | 137 (58) | 92 (67) | 45 (33) |
| Family history of diabetes mellitus | No | - | 100 (42) | 72 (72) | 28 (28) |
| Duration of diabetes mellitus | < 6 years | 6.84 ± 5.30 | 122 (51) | 91 (75) | 31 (25) |
| Duration of diabetes mellitus | 6-10 years | 6.84 ± 5.30 | 67 (28) | 44 (66) | 23 (34) |
| Duration of diabetes mellitus | > 10 years | 6.84 ± 5.30 | 48 (21) | 29 (60) | 19 (40) |
| HbA1c | ≤ 7% | 8.9 ± 2.3 | 61 (26) | 43 (70) | 18 (30) |
| HbA1c | >7% | 8.9 ± 2.3 | 176 (74) | 121 (69) | 55 (31) |
| Hypertension | Yes | - | 100 (42) | 71 (71) | 29 (29) |
| Hypertension | No | - | 137 (58) | 93 (68) | 44 (32) |
| Hypolipidemic drug utilization | Yes | - | 129 (54) | 108 (84) | 21 (16) |
| Hypolipidemic drug utilization | No | - | 108 (46) | 56 (52) | 52 (48) |
Hypolipidemic drugs utilization Out of 237 participants, $69\%$ [164] participants were found to have high lipid levels. Out of them, $78.65\%$ [129] received hypolipidemic drugs. A total of 129 generic drugs were prescribed. Most of the participants $94.6\%$ [122] received HMG-CoA reductase inhibitors, statins. In fixed dose combination (FDC), rosuvastatin and fenofibrate was the common prescription. The drug utilization pattern is presented in Figure 1.
**Figure 1:** *Frequency of hypolipidemic drugs prescribedFDC= Fixed Dose Combination, PPARs= Peroxisome proliferator-activated receptors*
Patterns of dyslipidemia Regarding dyslipidemia, high lipid level was observed in $69\%$ [164] participants, isolated in $24\%$ [58] of participants and combined abnormality in $70\%$ [166]. After analyzing dyslipidemia components low level of HDL-C was found in $35\%$ [82], hypertriglyceridemia in $44\%$ [104], elevated level of LDL in $60\%$ [142], and hypercholesterolemia in $26\%$ [61] participants, respectively. The details are presented in Table 3.
**Table 2**
| Lipid abnormality | Lipid abnormality.1 | n (%) |
| --- | --- | --- |
| Isolated | Low HDL | 18 (7.59) |
| | High TC | 0 (0) |
| | High LDL | 31 (13) |
| | High TG | 9 (3.79) |
| Combined | Low HDL & High LDL | 15 (6.3) |
| | Low HDL & High TG | 14 (5.9) |
| | Low HDL, High LDL & High TG | 21 (8.8) |
| | Low HDL, High LDL & High TC | 1 (0.42) |
| | Low HDL & High TC | 14 (5.9) |
| | High LDL & High TG | 15 (6.3) |
| | High TG & High TC | 1 (0.42) |
| | High LDL, High TC & High TG | 32 (13.5) |
| | Low HDL, High LDL, High TC & High TG | 13 (5.4) |
| Combined & Isolated | Low HDL | 82 (35) |
| | High TG | 104 (44) |
| | High LDL | 142 (60) |
| | High TC | 61 (26) |
Factors affecting dyslipidemia The factors like gender, age, occupation, hypertension, family history of diabetes, and duration of disease, HbA1c had no statistically significant association with dyslipidemia. The participants with hypertension had 1.2 times more chance of having dyslipidemia (OR=1.2; CI=0.60-2.38; $$p \leq 0.59$$). The participants with a family history of T2DM had 1.17 times more chance of having dyslipidemia (OR=1.17; CI=0.62-2.20; $$p \leq 0.62$$). Regarding BMI and dyslipidemia statistical significance was seen ($$p \leq 0.004$$). It was seen that participants in the overweight and obese category had 1.07 times more chance of having dyslipidemia than those in the normal weight category (OR=1.07; CI=0.98-1.16; $$p \leq 0.92$$). Regarding drug utilization and dyslipidemia statistical significance was observed ($p \leq 0.001$). It was observed that participants who were not on lipid-lowering medication had 5.38 times more chance of having a derranged lipid profile (OR=5.38; CI=2.82-10.28; $p \leq 0.001$). The results are presented in Tables 3, 4.
## Discussion
The purpose of this study was to evaluate the prevalence of dyslipidemia and the utilization of hypolipidemic drugs use among participants suffering from T2DM. In this study, $56\%$ [133] of the participants were males, and high lipid levels were present in $72\%$ [96] of males. Comparable results were observed in a prior study by Bekele et al. where $53.6\%$ [120] participants were males and $70\%$ [84] had high lipid levels [9]. In this study, $97\%$ [230] of participants were older than 30 years of age, and $68\%$ [157] of them had high lipid levels. Comparable results were observed by Bekele et al. where the majority of participants $66.5\%$ [149] were older than 30 years and out of that $43.8\%$ [98] had dyslipidemia. Kassahun Haile et al. where most participants $86.3\%$ [214] were older than 30 years out of which $72.4\%$ [155] had dyslipidemia [9,10]. The mean BMI in this study was 25.27 ± 4.08 kg/m2. The majority of participants $45\%$ [106] had a BMI of less than 24.9 kg/m2 and $58\%$ [62] of them had high lipid levels. Equivalent results were observed by Bekele et al. where $58\%$ [130] participants were in a normal BMI category and $29.5\%$ [66] had dyslipidemia [9]. These results contrasted with the study by Kebede et al. where $56\%$ [183] were in the overweight category out of which $59\%$ [113] had dyslipidemia [11]. This may be because due to different ethnicity. The majority of participants $51\%$ [122] had a duration of disease of fewer than six years and among them $75\%$ [91] had dyslipidemia. Similar results were observed by Hyassat et al. [ 12].
In the present study, $54\%$ [129] of participants received hypolipidemic medications. Out of which $94.6\%$ [122] of participants received statins, and atorvastatin accounted for $40\%$ [52] prescriptions. Like Jayaram's study [13], Gupta et al. also found statins were prescribed in the majority of participants, and atorvastatin was mostly prescribed medication [14]. Similar results were observed in Patel's [15] and Adhikari's studies where atorvastatin was the most prescribed medication [16].
In this study, $69\%$ [164] of the individuals had elevated lipid levels. Among them, $24\%$ [58] of participants had one abnormal lipid parameter and $70\%$ [166] had several abnormal lipid parameters. Borle et al. discovered that $28\%$ [14] of participants exhibited isolated lipid profile abnormalities, while $58\%$ [29] had combined and mixed lipid profile abnormalities [8]. According to Hyassat et al., $62.48\%$ [493] of participants had combined and mixed deranged lipid profiles and $27.8\%$ [220] had isolated lipid profile abnormality [12]. A study by Sarfraz et al. also observed that $34\%$ [68] participants had isolated and $66\%$ [132] had combined and mixed dyslipidemia [17]. In the present study, low level of HDL-C was found in $35\%$ [82], hypertriglyceridemia in $44\%$ [104], elevated levels of LDL in $60\%$ [142] and hypercholesterolemia were found in $26\%$ [61], respectively. Comparable results were observed by Dayakar et al. [ 18].
Gender, age, occupation, hypertension, family history of diabetes, and disease duration exhibited no statistically significant association with the incidence of dyslipidemia in the present study. Also, no statistical significance was found between hypertension and abnormal lipid profiles ($$p \leq 0.60$$). Participants with hypertension had a 1.2 times greater likelihood of having high lipid levels (OR=1.2; CI=0.60-2.38; $$p \leq 0.59$$). In contrast to a study conducted by Bekele et al. where hypertension had 1.3 times more chance of having dyslipidemia (AOR=1.331; CI=0.436-4.062; $$p \leq 0.016$$), and Haile et al. found that participants with hypertension had 2.65 times more chance of having dyslipidemia (AOR-2.65; CI-1.4-4.9; $$p \leq 0.01$$) [9,10].
A family history of type 2 diabetes in this study did not statistically significantly correlate with abnormal lipid profiles ($$p \leq 0.42$$). The odds of having abnormal lipid profiles were also 1.17 times higher in persons with a family history (OR=1.17; CI=0.62-2.20; $$p \leq 0.62$$). Bekele et al. reported similar findings, finding that participants with a family history had a 2.1 times greater likelihood of having high lipid levels (AOR=2.1; CI=0.454-9.877; $$p \leq 0.339$$) [9]. HbA1c exhibited no statistically significant connection with abnormal lipid profiles in this study ($$p \leq 0.79$$). Participants with poor glycemic control were more likely to have high lipid levels (OR=1.00; CI=0.89-1.14; $$p \leq 0.98$$). Kebede et al. found that patients with poor glycemic control were 1.3 times more likely to have abnormal lipid profiles (OR=1.3; CI=0.8-1.9; $p \leq 0.05$) [11].
BMI was shown to be strongly linked with abnormal lipid profiles in this study ($$p \leq 0.004$$). It was also observed that those who were overweight or obese had 1.07 times greater likelihood of having high lipid levels (OR=1.07; CI=0.98-1.16; $$p \leq 0.92$$). A similar substantial correlation was observed by Kebede et al., where obese participants had 3.5 times the risk of having disordered lipid profiles (OR=3.5; CI=1.6-7.9; p 0.001). Ahmmed et al. found that participants who were overweight or obese had a 2.08 times greater likelihood of having high lipid levels (OR=2.08; CI=1.73-2.23; p 0.001) [11,19].
In the present study, drug use was found to have a statistically significant relationship with dyslipidemia (p 0.001). Participants who were not using cholesterol-lowering medication had a 5.38 times greater likelihood of having abnormal lipid profiles than those who were (OR=5.38; CI=2.82-10.28; p 0.001). There is insufficient evidence or previous research to support the findings.
Limitations of the study *In this* study the sample size was small. Participants in this study belonged to a small geographical area and were from a single centre. High-risk participants from the cardiology department were not included. Confounding factors such as poor glycemic control and concurrent primary hyperlipidemias that act independently of T2DM were not evaluated.
## Conclusions
In this study, a high prevalence of increased lipid levels was observed among diabetic patients. Also, poor T2DM control was found to contribute significantly to high lipid levels, with other confounders such as primary hyperlipaemias playing a role in the overall process. This study has highlighted the need for a collaborative approach between healthcare providers and patients toward the holistic goal of the management of T2DM and associated comorbidities.
## References
1. Stefanović A, Zeljković A, Vekić J, Spasojević-Kalimanovska V, Jelić-Ivanović Z, Spasić S. **Dyslipidemia in type 2 diabetes mellitus**. *Arh Farm (Belgr* (2019) **69** 338-348
2. ICMR ICMR. **ICMR guidelines for management of type 2 diabetes**. *Indian Counc Med Res* (2018) **1** 82
3. Standards of Medical Care in Diabetes. **American standards of medical care in Diabetes American standards of medical care in Diabetes American Diabetes Association**. *Diabetes Care* (2020) **2022** 45
4. Matthias AT, Kaushalya J, Somathilake G, Garusinghe C. **Utilization of statins in patients with type 2 diabetes mellitus: the practice in a lower middle income South Asian country**. *Int J Diabetes Dev Ctries* (2022) **80** 1-7
5. Khan SU, Michos ED. **Cardiovascular mortality after intensive LDL-Cholesterol lowering: Does baseline LDL-Cholesterol really matter?**. *Am J Prev Cardiol* (2020) **1** 100013. PMID: 34327454
6. Whelton PK, Carey RM, Aronow WS. **2017 ACC/AHA/AAPA/ABC/ACPM/AGS/APHA/ash/ASPC/NMA/PCNA guideline for the Prevention, Detection, Evaluation, and management of high blood pressure in adults: executive summary: a report of the American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines**. *Circulation* (2018) **138** 0
7. Yadav D, Mishra M, Tiwari A, Bisen PS, Goswamy HM, Prasad GB. **Prevalence of dyslipidemia and hypertension in Indian type 2 diabetic patients with metabolic syndrome and its clinical significance**. *Osong Public Health Res Perspect* (2014) **5** 169-175. PMID: 25180150
8. Borle A, Chhari N, Gupta G, Bathma V. **Study of prevalence and pattern of dyslipidaemia in type 2 diabetes mellitus patients attending rural health training centre of medical college in Bhopal, Madhya Pradesh, India**. *Int J Community Med Public Heal* (2016) **3** 140-144
9. Bekele S, Yohannes T, Mohammed AE. **Dyslipidemia and associated factors among diabetic patients attending durame general hospital in southern nations, nationalities, and people’s region**. *Diabetes Metab Syndr Obes* (2017) **10** 265-271. PMID: 28790859
10. Haile K, Timerga A. **Dyslipidemia and its associated risk factors among adult type-2 diabetic patients at Jimma University Medical Center, Jimma, Southwest Ethiopia**. *Diabetes Metab Syndr Obes* (2020) **13** 4589-4597. PMID: 33273834
11. Kebede WM, Gizachew KD, Mulu GB. **Prevalence and risk factors of dyslipidemia among type 2 diabetes patients at a referral hospital, north eastern Ethiopia**. *Ethiop J Health Sci* (2021) **31** 1267-1276. PMID: 35392337
12. Hyassat D, Al-Saeksaek S, Naji D. **Dyslipidemia among patients with type 2 diabetes in Jordan: prevalence, pattern, and associated factors**. *Front Public Health* (2022) **10** 1002466. PMID: 36424970
13. Nitya S, Kiruthika S, Meenakshi R, Suriya H, Yuvarajan S. **Drug utilization pattern of hypolipidemic drugs among geriatric diabetic patients for primary prevention of CVD: a prospective observational study**. *J Family Med Prim Care* (2022) **11** 1325-1333
14. Gupta R, Lodha S, Sharma KK. **Evaluation of statin prescriptions in type 2 diabetes: India Heart Watch-2**. *BMJ Open Diabetes Res Care* (2016) **4** 0
15. Patel K, Joshi H, Khandhedia C, Shah H, Shah K, Patel V. **Study of drug utilization, morbidity pattern and cost of hypolipidemic agents in a tertiary care hospital**. *Int J Basic Clin Pharmacol* (2013) **2** 470
16. Santusta Adhikari, Sunil Shrestha, Rajani Shakya NK. **Prevalence of chronic complications and drug utilization pattern of type II diabetes mellitus**. *Mathews J Diabetes Obes* (2017) **1** 6
17. Sarfraz M, Sajid S, Ashraf MA. **Prevalence and pattern of dyslipidemia in hyperglycemic patients and its associated factors among Pakistani population**. *Saudi J Biol Sci* (2016) **23** 761-766. PMID: 27872574
18. Dayakar E, Sree CS, Sanjay E. **Study on the prevalence of dyslipidemia in type 2 diabetes mellitus**. *Int J Adv Med* (2019) **6** 786
19. Ahmmed MS, Shuvo S Das, Paul DK. **Prevalence of dyslipidemia and associated risk factors among newly diagnosed type-2 diabetes mellitus (T2DM) patients in Kushtia, Bangladesh**. *PLOS Glob Public Heal* (2021) **1** 3
|
---
title: Red cell distribution width and mortality in older patients with frailty in
the emergency department
authors:
- Janne Alakare
- Kirsi Kemp
- Timo Strandberg
- Maaret Castrén
- Jukka Tolonen
- Veli-Pekka Harjola
journal: BMC Emergency Medicine
year: 2023
pmcid: PMC9998144
doi: 10.1186/s12873-023-00801-1
license: CC BY 4.0
---
# Red cell distribution width and mortality in older patients with frailty in the emergency department
## Abstract
### Background
The red cell distribution width (RDW) reflects the degree of heterogeneity of red blood cells. Elevated RDW is associated both with frailty and with increased mortality in hospital-admitted patients. In this study we evaluate whether high RDW values are associated with mortality in older emergency department (ED) patients with frailty, and if the association is independent of the degree of frailty.
### Methods
We included ED patients with the following criteria: ≥ 75 years of age, Clinical Frailty Scale (CFS) score of 4 to 8, and RDW % measured within 48 h of ED admission. Patients were allocated to six classes by their RDW value: ≤ $13\%$, $14\%$, $15\%$, $16\%$, $17\%$, and ≥ $18\%$. The outcome was death within 30 days of ED admission. Crude and adjusted odds ratios (OR) with $95\%$ confidence intervals (CI) for a one-class increase in RDW for 30-day mortality were calculated via binary logistic regression analysis. Age, gender and CFS score were considered as potential confounders.
### Results
A total of 1407 patients ($61.2\%$ female), were included. The median age was 85 with an inter-quartile range (IQR) of 80–89, median CFS score 6 (IQR: 5–7), and median RDW 14 (IQR: 13–16). Of the included patients, $71.9\%$ were admitted to hospital wards. A total of 85 patients ($6.0\%$) died during the 30-day follow-up. Mortality rate was associated with RDW increase (p for trend <.001). Crude OR for a one-class increase in RDW for 30-day mortality was 1.32 ($95\%$ CI: 1.17–1.50, $p \leq .001$). When adjusted for age, gender and CFS-score, OR of mortality for one-class RDW increase was still 1.32 ($95\%$ CI: 1.16–1.50, $p \leq .001$).
### Conclusion
Higher RDW values had a significant association with increased 30-day mortality risk in frail older adults in the ED, and this risk was independent of degree of frailty. RDW is a readily available biomarker for most ED patients. It might be beneficial to include it in risk stratification of older frail ED patients to identify those who could benefit from further diagnostic assessment, targeted interventions, and care planning.
## Background
Red cell distribution width (RDW) is a measure reflecting the degree of heterogeneity of red blood cell size. RDW is calculated by dividing the standard deviation of red blood cell volumes by the mean corpuscular volume (MCV), usually expressed as a percentage value. RDW has traditionally been used for differential diagnosis of anaemia, but, subsequently, RDW elevation has been found to associate with higher short- and long-term increased mortality, both in the general population and in patients with many specific conditions [1–5], such as infections and sepsis [6–8], liver cirrhosis [9, 10], diabetic ketoacidosis [11], trauma [12, 13], acute pancreatitis [14], cardiac diseases [8, 15–19], pulmonary embolism [20], COVID-19, and acute respiratory failure [21–23]. In hospitals, high RDW has been shown to predict poor prognosis among general patients [24], among surgical patients [25], and among patients with critical illness [26, 27]. Older patients in emergency departments (EDs) and hospital wards have an increased risk of mortality if their RDW is elevated [8, 12, 28–30]. Although several mechanisms for this association have been presented [1, 2, 31], defined mechanisms for the association of elevated RDW and increased mortality have not yet been established.
Frailty syndrome, an ageing-related state of vulnerability due to decline in physiological reserves and functions [32], is usually defined either as a clinical phenotype or by calculating accumulated deficits such as diseases, physical and cognitive impairments, psychosocial risks, and geriatric syndromes [33–35]. Frailty is an independent predictor of mortality in patients admitted to emergency departments and hospital wards [36–38]. Elevated RDW has been shown to associate with frailty besides increased mortality of older ED patients [39–41].
Because frailty is related with both increased mortality of older ED patients and elevated RDW, frailty may be a confounder explaining increased mortality of older patients with elevated RDW. We studied whether elevated RDW is a risk predictor in older patients with frailty in the ED, and how frailty stage affects the association between elevated RDW and mortality.
## Methods
This study is a secondary analysis of an observational cohort study in frail older ED patients that was performed in an ED of a teaching hospital in Finland. In the primary study we included patients who were ≥ 75 years of age, had a score between 4 to 9 on the 9-point Clinical Frailty Scale (CFS) [34], and were registered residents of the hospital’s service area. ED visit data were collected between December 11th, 2018 and June 7th, 2019. The included patients were followed up from electronic health records. Methods for the primary study have been described in detail in our previous article [42].
The clinical laboratory service of the ED routinely gives RDW values (% value as integer) for all blood counts tested. Besides the clinical laboratory service, the ED has point-of-care testing equipment available, which does not provide RDW values. Point-of care testing is typically preferred, if more extensive laboratory testing is not anticipated based on patient’s chief complaint or condition. For the secondary analysis conducted here, those patient visits from the primary study who had the CFS score 4–8 and had RDW tested 0–48 h after ED admission were included. If more than one blood count was drawn from a patient within 48 h of ED admission, the result of the first laboratory test was used for the analysis. Patients who had a CFS score of 9 were excluded because such patients are defined as having a short life expectancy < 6 months, but otherwise not living with severe frailty.
Nonparametric baseline data were presented with interquartile ranges (IQR). The outcome measure was 30-day mortality. Patients were allocated to six classes based on their RDW value: ≤ $13\%$, $14\%$, $15\%$, $16\%$, $17\%$, and ≥ $18\%$. We used same cut-off values as a recent study to enable comparison of our results in frail ED patients to general older adult ED patient population [42]. Mortality rate was calculated for each class. The Cochran–Armitage test for trend was used to test the statistical significance of the trend of increasing mortality with higher RDW values.
Crude and adjusted ORs with $95\%$ confidence intervals (CI) of a one-class increase in RDW for 30-day mortality were calculated. Univariate and multivariate models of binary logistic regression analysis were used for crude and adjusted ORs, respectively. Age, sex, and CFS score were considered as potential confounders and were included in the analysis.
As a sensitivity analysis to assess if categorisation of the RDW values has impact on the results, we performed a regression analysis with RDW as continuous variable. We also performed a sensitivity analysis with haemoglobin as a potential confounder, because haemoglobin level is directly related to red blood cells, like RDW is, and may be associated with mortality.
From clinical perspective, we were interested whether RDW is independent of vital parameters. The National Early Warning Score 2 (NEWS2), a widely used prognostic score based on common vital signs, was included in the baseline data for our previous study [42]. We performed an additional testing by adjusting with the NEWS2 besides other potential confounders used in the regression analysis.
A p value of < 0.05 was considered statistically significant. GraphPad Prism software, version 9.4.1 (Graphpad Software LCC) was used for the Cochran–Armitage test. SPSS software, version 28 (IBM) was used for all other statistical analyses.
The primary study which this secondary analysis was based on, was registered at ClinicalTrials.gov on December 20th, 2018, identifier NCT03783234.
## Results
A total of 1407 ($61.2\%$ female) patient visits were included after excluding 294 visits for patients who either had no blood count drawn within 48 h of ED admission or had only point-of-care blood count testing without RDW-values, as well as four cases for patients with dual peak RDW values (due to previous red cell transfusions), and seven cases for patients with a CFS score of 9. Patient characteristics for the analytical sample are presented in Table 1: median age was 85 (IQR: 80–89), median CFS was 6 (IQR: 5–7), and median RDW % was 14 (IQR: 13–16, range: 12–28). Distribution of the RDW % values in the analytical sample are presented in Fig. 1. Of the included patients, 1011 ($71.9\%$) were admitted to hospital wards. Table 1Patient characteristicsN1407Agemedian (IQR)85 (80–89)Femalen (%)861 (61.2)Hospital admissionn (%)1011 (71.9)CFSmedian (IQR)6 (5–7) CFS: 4n (%)282 (20.0) CFS: 5–6718 (51.0) CFS: 7–8407 (28.9)RDW %Median (IQR)14 (13–16) RDW ≤ $13\%$n (%)412 (29.3) RDW $14\%$358 (25.4) RDW $15\%$262 (18.6) RDW $16\%$149 (10.6) RDW $17\%$80 (5.7) RDW ≥ $18\%$146 (10.4)Abbreviations: IQR Interquartile range, CFS Clinical Frailty Scale, RDW Red cell distribution widthFig. 1Distribution of RDW % in included patients. Normal distribution marked by black curve. Abbreviation: RDW, red cell distribution width Follow-up data for 30-day mortality were available for all ED visits. A total of 85 of 1407 ($6.0\%$) of included patients with RDW value available died during the 30-day follow-up. Within 30 days of ED admission, mortality rates were as follows: $\frac{9}{412}$ ($2.2\%$) of patients in the RDW ≤ $13\%$ group, $\frac{19}{358}$ ($5.3\%$) in the RDW $14\%$ group, $\frac{21}{262}$ ($8.0\%$) in the RDW $15\%$ group, $\frac{14}{149}$ ($9.4\%$) in the RDW $16\%$ group, $\frac{5}{80}$ ($6.3\%$) in the RDW $17\%$ group, and $\frac{17}{146}$ ($11.6\%$) in the RDW ≥ 18 group. Mortality rate was significantly higher with an increase in RDW (p for trend < 0.001). Mortality rates are presented in Fig. 2. For comparison, 30-day mortality of excluded patients who had no RDW value available was $\frac{8}{298}$ ($2.7\%$).Fig. 230-day mortality rates in each red cell distribution width category. The Cochran–Armitage test for trend was used to test the statistical significance of the trend. Abbreviation: RDW, red cell distribution width Crude OR of a one-class increase in RDW for 30-day mortality was 1.32 ($95\%$ CI: 1.17–1.50, $p \leq 0.001$). When adjusted for age, sex and CFS score, OR of a one-class increase was still 1.32 ($95\%$ CI: 1.16–1.50, $p \leq 0.001$). Crude and adjusted odds ratios are presented in Table 2.Table 2Crude and adjusted odds ratios for 30-day mortalityOR, crude ($95\%$ CI)p for crude OROR, adjusted ($95\%$ CI)p for adjusted ORRDW a1.32 (1.17–1.50) <.0011.32 (1.16–1.50) <.001CFS b1.47 (1.22–1.78) <.0011.43 (1.18–1.73) <.001Age c1.03 (1.00–1.07).0761.04 (1.00–1.08).082Female0.73 (0.47–1.14).1690.67 (0.42–1.05).083Binary logistic regression was used for the analysis. RDW class, CFS, age, and sex were included in the analysis for adjusted odds ratiosAbbreviations: RDW red cell distribution width, CFS Clinical Frailty Scale, OR odds ratio, CI confidence intervalaone-class increasebone-point increasecone-year increase In the sensitivity analysis with RDW as a continuous variable the significance of the results were not changed. Crude and adjusted ORs of $1\%$-unit increase of RDW for 30-day mortality were: 1.15 ($95\%$ CI: 1.06–1.24, $p \leq 0.001$), and 1.15 ($95\%$ CI: 1.07–1.25, $p \leq 0.001$), respectively. The absolute OR values were expectedly lower as scale increased from 6 categorical steps to 16 steps in %-units (range of RDW, 12–28). When haemoglobin level was added as a potential confounder, the adjusted OR of one-class increase in RDW was 1.34 ($95\%$ CI: 1.17–1.54, $p \leq 0.001$), without significant change in results.
When NEWS2 was added as a confounder, the adjusted OR of one-class increase of RDW for 30-day mortality was slightly lower than without it, but still significant: 1.27 ($95\%$: 1.11–1.47, $p \leq 0.001$).
## Discussion
In this study, increasing RDW was associated with higher 30-day mortality in frail older ED patients. The association remained significant when adjusted for age, gender and CFS score.
This study shows that the association of higher RDW value and increased mortality applies to the frail older population in an acute care setting. The association is independent of CFS score. The mortality rate increase is similar to those rates shown in a recent large cohort study of general hospital-admitted older patients [29], supporting the hypothesis that RDW is independent of frailty as a risk predictor. A small notch in mortality rate was noted in the group of patients with RDW of $17\%$. However, since the total trend was statistically significant, we interpret this dip to be variation due to limited sample size.
Many mechanisms, both short- and long-term, have been suggested for the association of elevated RDW and increased mortality. Impaired erythropoiesis and shortened red cell survival due to organ dysfunction, metabolic imbalances, and inflammatory reactions may cause alterations in red cell volumes. Oxidative stress and suppression of the erythrocyte lineage due to alterations in neutrophil and thrombocyte production during inflammation in acute conditions are potential contributors. Other possible causes of higher RDW include poor nutrition and erythrocyte fragmentation [1, 2, 31]. In addition, direct causality of high RDW and impaired intravascular haemodynamics, especially with vascular pathologies, has been presented [31]. Telomere shortening may be a link for poor outcomes in older vulnerable patients, as this is associated with both MCV variation and ageing-related all-cause mortality [43, 44]. Association of RDW increase with mortality was slightly lower, when NEWS2 was added as a potential confounder, which supports that elevated RDW is reflecting both short-, and long-term clinical deterioration.
RDW is a readily available biomarker for most ED patients. In clinical practice, RDW may be overlooked as a marker when clinical state and risks are assessed. Including RDW in patient assessment could lead to better high-risk feature identification, better targeting of further diagnostic work-up, effective interventions, and individualized advanced care planning.
Current risk-assessment methods, including ED triage systems, have limited performance, especially in older adults [45–47]. Machine-learning systems are promising tools for objective and more accurate risk assessment in emergency care, and may help in identifying patients who would benefit from targeted interventions [48–50]. In this study we have considered RDW as a general predictor. As higher RDW predicts poor outcomes in numerous different conditions, RDW may be associated with other predictive biomarkers in different specific conditions. Independence of RDW in multivariable predictive models could be studied preferably with machine learning methods with large data sets, as many predictive variables, including haemoglobin, white blood cell, and platelet counts, may have nonlinear associations.
Kim et al. stated in their article that RDW value should be included in risk stratification strategies for hospitalized older patients [29]. Based on earlier studies and our results, we agree with those authors and suggest that RDW should also be included in risk stratification of frail patients in emergency departments and hospital wards. Older patients often have non-specific presentations in the ED, and vital signs are less reliable for detecting early clinical deterioration in older patients [51–53]. Therefore, older patients, with or without frailty, could be one patient group that would benefit in particular from more comprehensive deep-learning risk-assessment methods. It may be favourable to include RDW among other variables when such artificial intelligence models are studied.
The strengths of this study include the prospectively collected patient data from a clinical setting, that is representative for the frail older ED patient population. Frailty status was assessed systematically with the CFS during ED admission. Baseline and outcome data were available for all patients included.
The study has some limitations. The analysis of this study was based on data collected in a previous prospective study. In this study, $82\%$ of patients who met the eligibility criteria, had an RDW value available. The included patients had higher mortality than patients who had no laboratory testing, or only point-of-care testing available. Chief complaints or acute disease severity were not included in the analyses, but we assume that those patients without blood tests taken were more likely visiting the ED for simple, low-acuity complaints, and therefore the results may not be representative for low-acuity patients.
## Conclusion
Higher RDW values were significantly associated with increased 30-day mortality in frail older adults in the ED. In this study, RDW was independent of frailty state as a risk predictor. RDW is a readily available parameter for most ED patients who have laboratory tests. It might be beneficial to include RDW in risk stratification of older frail ED patients in order to identify patients at high risk of adverse outcomes who could benefit from further diagnostic assessment, targeted interventions, and care planning.
## References
1. Salvagno GL, Sanchis-Gomar F, Picanza A, Lippi G. **Red blood cell distribution width: a simple parameter with multiple clinical applications**. *Crit Rev Clin Lab Sci* (2015) **52** 86-105. DOI: 10.3109/10408363.2014.992064
2. Pan J, Borné Y, Engström G. **The relationship between red cell distribution width and all-cause and cause-specific mortality in a general population**. *Sci Rep* (2019) **9** 16208. DOI: 10.1038/s41598-019-52708-2
3. Patel KV, Ferrucci L, Ershler WB, Longo DL, Guralnik JM. **Red blood cell distribution width and the risk of death in middle-aged and older adults**. *Arch Intern Med* (2009) **169** 515-523. DOI: 10.1001/archinternmed.2009.11
4. Perlstein TS, Weuve J, Pfeffer MA, Beckman JA. **Red blood cell distribution width and mortality risk in a community-based prospective cohort**. *Arch Intern Med* (2009) **169** 588-594. DOI: 10.1001/archinternmed.2009.55
5. Pilling LC, Atkins JL, Kuchel GA, Ferrucci L, Melzer D. **Red cell distribution width and common disease onsets in 240,477 healthy volunteers followed for up to 9 years**. *PLoS One* (2018) **13** e0203504. DOI: 10.1371/journal.pone.0203504
6. Kim SY, Woo SH, Lee WJ, Kim DH, Seol SH, Lee JY. **The qSOFA score combined with the initial red cell distribution width as a useful predictor of 30 day mortality among older adults with infection in an emergency department**. *Aging Clin Exp Res* (2021) **33** 1619-1625. DOI: 10.1007/s40520-020-01738-2
7. Lee JH, Chung HJ, Kim K, Jo YH, Rhee JE, Kim YJ. **Red cell distribution width as a prognostic marker in patients with community-acquired pneumonia**. *Am J Emerg Med* (2013) **31** 72-79. DOI: 10.1016/j.ajem.2012.06.004
8. Nakashima K, Ohgami E, Kato K, Yoshitomi S, Maruyama T, Harada M. **Prognostic significance of red cell distribution width in hospitalized older patients with heart failure or infection**. *Geriatr Gerontol Int* (2019) **19** 988-992. DOI: 10.1111/ggi.13755
9. Abdel-Samiee M, Ahmed SamyKohla M, FathyRagab M, Ghanem HS, KamelAbouGabal A, Abdelsameea E. **Red cell distribution width as a predictor of outcome in hospitalized cirrhotic patients**. *Eur J Gastroenterol Hepatol* (2021) **33** e978-e85. DOI: 10.1097/MEG.0000000000002337
10. Turcato G, Campagnaro T, Bonora A, Vignola N, Salvagno GL, Cervellin G. **Red blood cell distribution width independently predicts 1-month mortality in acute decompensation of cirrhotic patients admitted to emergency department**. *Eur J Gastroenterol Hepatol* (2018) **30** 33-38. DOI: 10.1097/MEG.0000000000000993
11. Dai H, Su X, Li H, Zhu L. **Association between red blood cell distribution width and mortality in diabetic ketoacidosis**. *J Int Med Res* (2020) **48** 300060520911494. DOI: 10.1177/0300060520911494
12. Marom O, Paz I, Topaz G, Ohana N, Yaacobi E. **Red cell distribution width-A mortality predictor in older adults with proximal femoral fracture**. *Arch Gerontol Geriatr* (2022) **100** 104623. DOI: 10.1016/j.archger.2022.104623
13. Sakai Y, Wakao N, Matsui H, Watanabe T, Iida H, Katsumi A. **Elevated red blood cell distribution width is associated with poor outcome in osteoporotic vertebral fracture**. *J Bone Miner Metab* (2021) **39** 1048-1057. DOI: 10.1007/s00774-021-01242-1
14. Şenol K, Saylam B, Kocaay F, Tez M. **Red cell distribution width as a predictor of mortality in acute pancreatitis**. *Am J Emerg Med* (2013) **31** 687-689. DOI: 10.1016/j.ajem.2012.12.015
15. Peng Y, Guan X, Wang J, Ma J. **Red cell distribution width is correlated with all-cause mortality of patients in the coronary care unit**. *J Int Med Res* (2020) **48** 300060520941317. DOI: 10.1177/0300060520941317
16. Huang S, Zhou Q, Guo N, Zhang Z, Luo L, Luo Y. **Association between red blood cell distribution width and in-hospital mortality in acute myocardial infarction**. *Medicine (Baltimore)* (2021) **100** e25404. DOI: 10.1097/MD.0000000000025404
17. Wang Z, Chen X, He S. **Prognostic value of red blood cell distribution width for mortality in patients with hypertrophic cardiomyopathy**. *Clin Biochem* (2021) **96** 19-25. DOI: 10.1016/j.clinbiochem.2021.07.002
18. Zhang X, Wang Y, Chen N, Liu Y, Xiao J, Lin Z. **Red cell distribution width is associated with short-term mortality in critically ill patients with heart failure**. *ESC Heart Fail* (2022). DOI: 10.1002/ehf2.14023
19. Szekely Y, Finkelstein A, Bazan S, Halkin A, Abbas Younis M, Erez J. **Red blood cell distribution width as a prognostic factor in patients undergoing transcatheter aortic valve implantation**. *J Cardiol* (2019) **74** 212-216. DOI: 10.1016/j.jjcc.2019.04.005
20. Zhou XY, Chen HL, Ni SS. **Red cell distribution width in predicting 30-day mortality in patients with pulmonary embolism**. *J Crit Care* (2017) **37** 197-201. DOI: 10.1016/j.jcrc.2016.09.024
21. Zhang W, Wang Y, Wang J, Wang S. **Association between red blood cell distribution width and long-term mortality in acute respiratory failure patients**. *Sci Rep* (2020) **10** 21185. DOI: 10.1038/s41598-020-78321-2
22. Banon T, Wortsman J, Ben Moshe S, Gazit S, Peretz A, Ben Tov A. **Evaluating red blood cell distribution width from community blood tests as a predictor of hospitalization and mortality in adults with SARS-CoV-2: a cohort study**. *Ann Med* (2021) **53** 1410-1418. DOI: 10.1080/07853890.2021.1968484
23. Lorente L, Martín MM, Argueso M, Solé-Violán J, Perez A, Marcos YRJA. **Association between red blood cell distribution width and mortality of COVID-19 patients**. *Anaesth Crit Care Pain Med* (2021) **40** 100777. DOI: 10.1016/j.accpm.2020.10.013
24. Hunziker S, Stevens J, Howell MD. **Red cell distribution width and mortality in newly hospitalized patients**. *Am J Med* (2012) **125** 283-291. DOI: 10.1016/j.amjmed.2011.08.021
25. Abdullah HR, Sim YE, Sim YT, Ang AL, Chan YH, Richards T. **Preoperative Red Cell Distribution Width and 30-day mortality in older patients undergoing non-cardiac surgery: a retrospective cohort observational study**. *Sci Rep* (2018) **8** 6226. DOI: 10.1038/s41598-018-24556-z
26. Bazick HS, Chang D, Mahadevappa K, Gibbons FK, Christopher KB. **Red cell distribution width and all-cause mortality in critically ill patients**. *Crit Care Med* (2011) **39** 1913-1921. DOI: 10.1097/CCM.0b013e31821b85c6
27. Jia L, Cui S, Yang J, Jia Q, Hao L, Jia R. **Red blood cell distribution width predicts long-term mortality in critically ill patients with acute kidney injury: a retrospective database study**. *Sci Rep* (2020) **10** 4563. DOI: 10.1038/s41598-020-61516-y
28. Kim SH, Yeon JH, Park KN, Oh SH, Choi SP, Kim YM. **The association of Red cell distribution width and in-hospital mortality in older adults admitted to the emergency department**. *Scand J Trauma Resusc Emerg Med* (2016) **24** 81. DOI: 10.1186/s13049-016-0274-8
29. Kim KM, Nerlekar R, Tranah GJ, Browner WS, Cummings SR. **Higher red cell distribution width and poorer hospitalization-related outcomes in elderly patients**. *J Am Geriatr Soc* (2022) **70** 2354-2362. DOI: 10.1111/jgs.17819
30. Wang AY, Ma HP, Kao WF, Tsai SH, Chang CK. **Red blood cell distribution width is associated with mortality in elderly patients with sepsis**. *Am J Emerg Med* (2018) **36** 949-953. DOI: 10.1016/j.ajem.2017.10.056
31. Ananthaseshan S, Bojakowski K, Sacharczuk M, Poznanski P, Skiba DS, PrahlWittberg L. **Red blood cell distribution width is associated with increased interactions of blood cells with vascular wall**. *Sci Rep* (2022) **12** 13676. DOI: 10.1038/s41598-022-17847-z
32. 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
33. Fried LP, Tangen CM, Walston J, Newman AB, Hirsch C, Gottdiener J. **Frailty in older adults: evidence for a phenotype**. *J Gerontol A Biol Sci Med Sci* (2001) **56** M146-M156. DOI: 10.1093/gerona/56.3.m146
34. Rockwood K, Song X, MacKnight C, Bergman H, Hogan DB, McDowell I. **A global clinical measure of fitness and frailty in elderly people**. *CMAJ* (2005) **173** 489-495. DOI: 10.1503/cmaj.050051
35. Xue QL. **The frailty syndrome: definition and natural history**. *Clin Geriatr Med* (2011) **27** 1-15. DOI: 10.1016/j.cger.2010.08.009
36. Romero-Ortuno R, Wallis S, Biram R, Keevil V. **Clinical frailty adds to acute illness severity in predicting mortality in hospitalized older adults: An observational study**. *Eur J Intern Med* (2016) **35** 24-34. DOI: 10.1016/j.ejim.2016.08.033
37. Rueegg M, Nissen SK, Brabrand M, Kaeppeli T, Dreher T, Carpenter CR. **The clinical frailty scale predicts 1-year mortality in emergency department patients aged 65 years and older**. *Acad Emerg Med* (2022) **29** 572-580. DOI: 10.1111/acem.14460
38. Lee JH, Park YS, Kim MJ, Shin HJ, Roh YH, Kim JH. **Clinical Frailty Scale as a predictor of short-term mortality: a systematic review and meta-analysis of studies on diagnostic test accuracy**. *Acad Emerg Med* (2022). DOI: 10.1111/acem.14493
39. Li CM, Chao CT, Chen SI, Han DS, Huang KC. **Elevated red cell distribution width is independently associated with a higher frailty risk among 2,932 community-dwelling older adults**. *Front Med (Lausanne)* (2020) **7** 470. DOI: 10.3389/fmed.2020.00470
40. Kim KM, Lui LY, Browner WS, Cauley JA, Ensrud KE, Kado DM. **Association Between Variation in Red Cell Size and Multiple Aging-Related Outcomes**. *J Gerontol A Biol Sci Med Sci* (2021) **76** 1288-1294. DOI: 10.1093/gerona/glaa217
41. Li Q, Chen X, Han B. **Red blood cell distribution width is associated with frailty in older inpatients in China: Sex differences in a cross-sectional study**. *Exp Gerontol.* (2021) **150** 111392. DOI: 10.1016/j.exger.2021.111392
42. Kemp K, Alakare J, Harjola VP, Strandberg T, Tolonen J, Lehtonen L. **National Early Warning Score 2 (NEWS2) and 3-level triage scale as risk predictors in frail older adults in the emergency department**. *BMC Emerg Med* (2020) **20** 83. DOI: 10.1186/s12873-020-00379-y
43. Kozlitina J, Garcia CK. **Red blood cell size is inversely associated with leukocyte telomere length in a large multi-ethnic population**. *PLoS One* (2012) **7** e51046. DOI: 10.1371/journal.pone.0051046
44. Wang Q, Zhan Y, Pedersen NL, Fang F, Hägg S. **Telomere length and all-cause mortality: a meta-analysis**. *Ageing Res Rev* (2018) **48** 11-20. DOI: 10.1016/j.arr.2018.09.002
45. Zachariasse JM, van der Hagen V, Seiger N, Mackway-Jones K, van Veen M, Moll HA. **Performance of triage systems in emergency care: a systematic review and meta-analysis**. *BMJ Open* (2019) **9** e026471. DOI: 10.1136/bmjopen-2018-026471
46. Brouns SHA, Mignot-Evers L, Derkx F, Lambooij SL, Dieleman JP, Haak HR. **Performance of the Manchester triage system in older emergency department patients: a retrospective cohort study**. *BMC Emerg Med* (2019) **19** 3. DOI: 10.1186/s12873-018-0217-y
47. Grossmann FF, Zumbrunn T, Frauchiger A, Delport K, Bingisser R, Nickel CH. **At risk of undertriage? Testing the performance and accuracy of the emergency severity index in older emergency department patients**. *Ann Emerg Med* (2012) **60** 317-25.e3. DOI: 10.1016/j.annemergmed.2011.12.013
48. Grant K, McParland A, Mehta S, Ackery AD. **Artificial Intelligence in Emergency Medicine: Surmountable Barriers With Revolutionary Potential**. *Ann Emerg Med* (2020) **75** 721-726. DOI: 10.1016/j.annemergmed.2019.12.024
49. Kirubarajan A, Taher A, Khan S, Masood S. **Artificial intelligence in emergency medicine: A scoping review**. *J Am Coll Emerg Physicians Open* (2020) **1** 1691-1702. DOI: 10.1002/emp2.12277
50. Ehrlich H, McKenney M, Elkbuli A. **The niche of artificial intelligence in trauma and emergency medicine**. *Am J Emerg Med* (2021) **45** 669-670. DOI: 10.1016/j.ajem.2020.10.050
51. Nemec M, Koller MT, Nickel CH, Maile S, Winterhalder C, Karrer C. **Patients presenting to the emergency department with non-specific complaints: the Basel Non-specific Complaints (BANC) study**. *Acad Emerg Med* (2010) **17** 284-292. DOI: 10.1111/j.1553-2712.2009.00658.x
52. Nissen SK, Candel BGJ, Nickel CH, de Jonge E, Ryg J, Bogh SB. **The Impact of Age on Predictive Performance of National Early Warning Score at Arrival to Emergency Departments: Development and External Validation**. *Ann Emerg Med* (2022) **79** 354-363. DOI: 10.1016/j.annemergmed.2021.09.434
53. Pirneskoski J, Lääperi M, Kuisma M, Olkkola KT, Nurmi J. **Ability of prehospital NEWS to predict 1-day and 7-day mortality is reduced in older adult patients**. *Emerg Med J* (2021) **38** 913-918. DOI: 10.1136/emermed-2019-209400
|
---
title: Efficient Breast Cancer Diagnosis from Complex Mammographic Images Using Deep
Convolutional Neural Network
authors:
- Hameedur Rahman
- Tanvir Fatima Naik Bukht
- Rozilawati Ahmad
- Ahmad Almadhor
- Abdul Rehman Javed
journal: Computational Intelligence and Neuroscience
year: 2023
pmcid: PMC9998154
doi: 10.1155/2023/7717712
license: CC BY 4.0
---
# Efficient Breast Cancer Diagnosis from Complex Mammographic Images Using Deep Convolutional Neural Network
## Abstract
Medical image analysis places a significant focus on breast cancer, which poses a significant threat to women's health and contributes to many fatalities. An early and precise diagnosis of breast cancer through digital mammograms can significantly improve the accuracy of disease detection. Computer-aided diagnosis (CAD) systems must analyze the medical imagery and perform detection, segmentation, and classification processes to assist radiologists with accurately detecting breast lesions. However, early-stage mammography cancer detection is certainly difficult. The deep convolutional neural network has demonstrated exceptional results and is considered a highly effective tool in the field. This study proposes a computational framework for diagnosing breast cancer using a ResNet-50 convolutional neural network to classify mammogram images. To train and classify the INbreast dataset into benign or malignant categories, the framework utilizes transfer learning from the pretrained ResNet-50 CNN on ImageNet. The results revealed that the proposed framework achieved an outstanding classification accuracy of $93\%$, surpassing other models trained on the same dataset. This novel approach facilitates early diagnosis and classification of malignant and benign breast cancer, potentially saving lives and resources. These outcomes highlight that deep convolutional neural network algorithms can be trained to achieve highly accurate results in various mammograms, along with the capacity to enhance medical tools by reducing the error rate in screening mammograms.
## 1. Introduction
The predominant cause of cancer-related deaths among women globally is breast cancer [1, 2]. According to a report by the World Health Organization's (WHO) cancer research institute, the International Agency for Research on Cancer, in 2018 globally, 17.1 million breast cancer cases were reported. The number of cases is predicted to increase to double the amount by 2025 [3]. Breast cancer is a highly invasive tumor that primarily affects women [4]. The high death rate among women makes it the second deadliest malignancy after lung cancer [5, 6]. A study by the National Institute of Cancer in China found 1.67 million breast cancer and 522,000 deaths cases from 2008 to 2012 [7].
Despite extensive efforts from medical professionals and researchers, a definitive method for treating breast cancer has yet to be established and reliable evidence for its prevention remains elusive [8–11]. Some components of breast cancer tissues are highly malignant and pose a severe danger to patients' lives as they can spread to other vital organs [12–15]. The growth of mammary cells can lead to tumors in women. Tumors are classified as benign or malignant based on the area, size, and location, using the BI-RAD scores [16, 17]. Benign tumors are not life-threatening and can be treated through medication to prevent further growth [17, 18]. Malignant tumors, on the other hand, can spread to other parts of the body via the lymphatic system or blood, making them much more dangerous [19–22]. This uncontrolled cell proliferation in the breast leads to the formation of malignant tumors, which can only be treated through surgery or radiation therapy [23, 24].
Early detection of breast cancer is crucial for accurate diagnosis and analysis, and many researchers are turning to biomedical imaging to aid specialist radiologists. Various methods such as MRI, mammography, and ultrasound are utilized to identify breast carcinoma [25, 26]. However, the large volume of images challenges radiologists in identifying potential cancerous areas. Therefore, an efficient automated method is needed, and computer-aided diagnostic (CAD) systems are being utilized in aiding radiologists in detecting cancerous breast tumors [27, 28].
Increasingly, deep learning techniques are being applied to medical imaging to develop automated computer-aided diagnosis (CAD) systems [29–35]. Deep learning is considered the most effective method for detecting and classifying medical images [29, 30]. With these techniques, the mammogram image's significant low to high-level hierarchical features can be directly extracted, making deep learning the most reliable medical imaging method [29]. Several CAD systems based on deep learning have been developed for breast lesions detection, which outperforms traditional systems [36]. Accurate detection of breast lesions is crucial for improving breast cancer diagnosis [29, 31]. However, detecting these lesions can be challenging due to their varying texture, shape, position, and size. Deep learning and image processing methods have been proposed to overcome the limitations of conventional technology, which cannot perform automated identification [29]. The final stage in the CAD model is the classification of breast lesions into benign or malignant, which is important in assessing the correctness of the diagnostic [30].
Currently employed methods for detecting breast cancer are slow, costly, and require extra efforts to run the radiology equipment. Accurately detecting breast cancer automatically from an image processing perspective is not easy. Hence, early diagnosis and proper treatment are deemed crucial. Therefore, an efficient screening system and automation are necessary for breast cancer detection due to the following reasons [12]: incorrect diagnoses and predictions, tumors appearing in low contrast areas, unreliable human diagnoses, overburdening of radiologists, human error in diagnosis, need for large training data to avoid overfitting in deep learning algorithms, high computational complexity, and longer processing time for accurate tumor identification.
A novel breast cancer detecting system is proposed with an improved architecture that integrates deep convolutional neural network (DCNN) and breast mammogram images to address previous drawbacks mentioned above. The proposed system intends to divide breast tumors into benign and malignant categories. The system's performance is evaluated and compared with existing classification systems using a public mammographic image dataset named INbreast. The new system includes transfer learning to fine-tune the pretrained DCNN and detailed results from experiments on the INbreast dataset. The system's performance is evaluated using the following metrics: AUC, specificity, accuracy, sensitivity, and F-1 score.
The rest of the paper is organized as follows. Section 2 presents the related work. Section 3 provides the proposed approach for breast cancer detection. Section 4 presents experimental analysis, results and comparison with existing work. Section 5 concludes this paper and presents future work.
## 2. Related Work
Breast cancer diagnosis in modern medical procedures often involves using mammography images [37]. A summary of recently developed systems for breast cancer diagnosis using mammogram images is presented in this section.
Structured support vector machine (SSVM) and conditional random field (CRF) are two structured prediction techniques proposed in [38] to classify mass mammograms. Both approaches used potential functions based on deep convolution and belief networks. The results demonstrated that the CRF method outperformed the SSVM method in training and inference time. Authors in [29] utilized four-fold cross-validation on X-ray mammograms from the INbreast dataset to estimate a full-resolution convolutional network (FrCN). It resulted in an F1 score of $99.24\%$, an accuracy of $95.96\%$, and a Matthews correlation coefficient (MCC) of $98.96\%$. In another study [39], the BDR-CNN-GCN approach was proposed by combining a graph-convolutional network (GCN) with a basic 8-layer CNN that includes batch normalization and dropout layers. The final BDR-CNN-GCN model was formed by integrating the two-layer GCN with the CNN. This method was tested using the MIAS dataset, and successful results were obtained with a $96.10\%$ accuracy level.
Authors in [40] proposed modifying the YOLOv5 network for identifying and classifying breast cancers, with the algorithm run using specific parameter values. The modified YOLOv5 was compared with a faster RCNN and YOLOv3, achieving an accuracy of $96.50\%$ and an MCC value of $93.50\%$. The diverse features (DFeBCD) method was proposed by [41], which classified mammograms into two categories normal and abnormal. They used two classifiers, an emotion learning-inspired integrated classifier (ELiEC) and SVM, with the IRMA mammography dataset. The ELiEC classifier outperformed SVM, achieving an accuracy rate of $80.30\%$. In [30], a deep-CNN model that utilized transfer learning (TL) was introduced to prevent overfitting when working with small datasets. DDSM, MIAS, BCDR, and INbreast were used to assess its performance. INbreast dataset achieved an accuracy of $95.5\%$, the DDSM dataset achieved an accuracy level of $97.35\%$, and the BCDR database achieved a $96.67\%$ accuracy level.
Authors in [42] for extracting features from breast mammograms utilized lifting wavelet transform (LWT). Feature vectors' size was reduced using linear discriminant analysis (LDA) and principal component analysis (PCA). The classification was performed using the moth flame optimization and extreme learning machine (ELM) approach with MIAS and DDSM and datasets, achieving accuracy of $95.80\%$ and $98.76\%$, respectively. In addition, researchers have also trained the CNN Inception-v3 model on 316 images, resulting in a sensitivity of 0.88, specificity of 0.87, and an AUC of 0.946 [43]. Furthermore, in [44], a CNN and TL classification method was proposed to evaluate the performance of eight fine-tuned pretrained models. Authors in [45] presented a hybrid classification model using Mobilenet, ResNet50, and Alexnet with an accuracy level of $95.6\%$. In [46], four different CNN architectures (VGG19, InceptionV3, ResNet50, and VGG16) were utilized for model training using 5000 images, while prediction models were evaluated on 1007 images.
Authors in [47] utilized alpha, geostatistics, and diversity analyses forms in their proposed breast cancer detection method. They employed the SVM classifier on MIAS and DDSM databases, which resulted in a detection accuracy level of $96.30\%$. The SVM classifier and gray level co-occurrence matrix (GLCM) were employed by [48] for detecting breast cancer abnormalities in the MIAS data set. Their method achieved an accuracy of $93.88\%$ and surpassed the performance of the k-nearest neighbour (kNN) algorithm. Authors in [49] used AlexNet and SVM to enhance classification accuracy with data augmentation techniques. The method achieved $71.01\%$ accuracy, which increased to $87.2\%$ with SVM and was evaluated on DDSM and CBIS-DDSM datasets.
A DenseNet deep learning framework extracted image features and classified cancerous and benign cells by feeding them into a fully connected (FC) layer. The effectiveness of this technique was evaluated by adjusting the hyperparameters [50]. An algorithm named DICNN was developed by Irfan et al. [ 51], which uses a dilated semantic segmentation network and morphological operation. Combining these feature vectors with SVM classification yielded an accuracy of $98.9\%$.
Although prior breast cancer detection and classification systems have improved information extraction, several issues still need attention, such as low contrast in tumor location, high memory complexity, long processing time, and the need for a large amount of training data for deep learning approaches. In response to these problems, we propose a new approach to breast cancer detection and classification, which will be discussed in detail in the following section.
## 3. Methodology
In this section, the processes used for implementing our proposed scheme are described in depth. The system consists of the following steps: [1] image enhancement, [2] image segmentation, [3] feature extraction and the selection, and [4] feature classification. The proposed system is illustrated in Figure 1.
## 3.1. Dataset
This study used a digital breast X-ray database named INbreast to implement the proposed CAD approach. The INbreast dataset is a public database that contains more recent FFDM images. It typically has an image size of 3328 × 4084 pixels. It contains 115 patients' cases along with 410 mammograms with both craniocaudal (CC) view and a mediolateral oblique (MLO) view. Of these 115 patients, 90 had mammograms taken of both breasts, totaling 360 images, while the other 25 had only two mammograms taken each. In total, 410 mammograms were produced from 115 patients, including cases of normal, benign, and malignant breasts. 107 cases with breast lesions were used from the MLO and CC views for evaluation purposes.
## 3.2. Convolutional Neural Network
This subsection will examine the fundamental structure of all convolutional neural network (CNN) architectures. CNNs are deep neural networks used for image recognition and classification. In recent years, CNNs have become a crucial tool in image analysis, especially for identifying faces, text, and medical imaging. CNNs have a long history of success in image classification and segmentation, first developed in 1989. CNNs replicate the human brain's visual information processing by incorporating layers of “neurons” that only respond to their local surroundings. These networks can understand the topological aspects of an image through a combination of convolutional, pooling, and fully connected (FC) layers. The architecture of a CNN is shown in Figure 2.
## 3.2.1. Convolutional Layers
The convolutional layers are assembled into feature maps based on local connections and weight distribution principles. A filter bank, a group of weights, connects neurons in a feature map to corresponding local regions in the preceding layer. Each feature map uses a different filter bank, and all the units in the map share the same filter row. This weight distribution and local connection help reduce the number of parameters by utilizing the close relationship between neighboring pixels and location-independent image features. The output of the weights is then sent to an activation function, such as ReLU or Sigmoid. This activation function enables the nonlinear transformation of the input data, which is necessary for the following processing stages.
## 3.2.2. Pooling Layer
As illustrated in Figure 2, the pooling layer follows the convolution layer and uses subsampling to integrate the features from the convolutional layer into a single layer semantically. This layer's primary objective is to decrease the size of the image by combining pixels into one value while preserving its features. In this layer, typical operations include max as well as main pooling.
## 3.2.3. Fully Connected Layer
The last layer in CNN is the dense classification layer, which is responsible for determining the category of input data based on extracted features from CNN. The number of units in the FC layer is the same as the number of different classifications (categories).
## 3.3. Proposed Workflow
This section provides the proposed workflow for breast cancer diagnosis using a deep convolutional neural network.
## 3.3.1. Image Enhancement
Image enhancement refers to increasing contrast and suppressing noise in mammogram images to assist radiologists in detecting breast abnormalities. Various image enhancement methods exist, including adaptive contrast enhancement (AHE). AHE improves the local contrast and reveals more image details, making it a helpful technique for enhancing both natural and medical images [52]. However, it may also result in considerable noise. In this paper, we utilized the contrast-limited adaptive histogram equalization (CLAHE) technique, a form of AHE, to enhance image contrast [52]. A drawback of AHE is that it can over-enhance the images due to the integration process [49]. To mitigate this issue, CLAHE is used as it limits the local histogram by setting a clip level, thus controlling contrast enhancement. Figure 3 illustrates an image enhanced by the CLAHE algorithm.
Furthermore, CLAHE algorithm steps are given as follows [53]:Split image into equal-sized contextual regions. Apply histogram equalization to all contextual regions. Limit the histogram to the level of the clip. Reallocate the clipped values in the histogram. Obtain enhanced pixel value through histogram integration.
## 3.4. Image Segmentation
Image segmentation involves dividing an image into regions with similar characteristics and features. Segmentation aims to simplify the image for easier analysis [54]. Popular image segmentation techniques include edge detection, partial differential equation (PDE), fuzzy theory, artificial neural network (ANN), region-based segmentation, and thresholding.
## 3.4.1. Thresholding Method
One of the simplest image segmentation methods is the thresholding method [55, 56]. The pixels of the image are split according to their intensity level. The global threshold is the most commonly used thresholding technique [57]. It is accomplished by setting a threshold value (T) constant throughout the image. The output image is derived from the original image based on the threshold value.
## 3.4.2. Region-Based Segmentation Methods
It is a simple approach compared to other methods, as it involves dividing an image into different sections based on predetermined. Compared to others, it is a straightforward method because it entails separating an image into different sections based on predetermined criteria [58]. There are two primary kinds of region-based segmentation: [1] region splitting and merging and [2] region growing. Region growing allows the removal of a region from an image using defined criteria, such as intensity. It involves selecting a starting seed point. It is important to note that unlike region growing, region splitting and merging work on the entire image [59].
In the present study extracting the region of interest (ROI) involves using both thresholding and region-based techniques. The tumor in the INbreast dataset samples cites moreira2012inbreast is labeled by a white bounding box, as shown in Figure 4. For extracting ROI, the tumor region is first determined by setting a threshold value based on the white color pixels in the image. The threshold for all images is determined to be 80 after several attempts, independent of tumor size. After identifying the greatest area inside this threshold within the image, the tumor is automatically cropped. Figure 4 shows ROI extracted using threshold and region-based methods.
The method for extracting ROI can be summarized in four steps:Thresholding the grayscale mammogram image to create a binary image. Labelling and counting the binary image objects, then retaining only the largest one, which is the tumor, as defined by the white bounding box. Assign the largest area within the threshold value to “1” and the rest a value of “0.”Multiply binary image with original mammogram image for obtaining final ROI without including other parts of breast or artifacts.
## 3.4.3. Feature Extraction and Selection
Numerous methods exist for feature extraction. Due to their exceptional performance, deep convolutional neural networks (DCNN) garnered significant interest in recent years. Consequently, the DCNN is utilized in this paper.
## 3.4.4. Deep Convolutional Neural Network
The success of DCNN in image classification and analysis has been documented in various studies [60, 61]. Convolutional neural networks (CNNs) are composed of multiple trainable stages that culminate in a supervised classifier and feature maps [62]. Three primary types of layers are employed to build CNN structures: convolutional, pooling, and fully connected (FC) layers [63]. The ResNet50 CNN classification model categorizes breast cancer as benign or malignant in this work.
## 3.4.5. Feature Learning through Transfer Learning
Machine learning has various feature learning methods (FL), allowing a system to automatically identify the representations required for feature detection, prediction, or classification from a preprocessed dataset [64]. This implies that the machine can learn and use the features to perform tasks such as classification or prediction. In deep learning, FL can be accomplished by constructing a complete CNN to train and test image datasets or adjust a pretrained CNN for classification or prediction on a new image dataset, referred to as transfer learning.
In deep learning, transfer learning (TL) is a widely-used technique that enables the utilization of a pretrained network for new prediction or classification tasks. This is achieved by adjusting the parameters of the pretrained network with randomly initialized weights for the new task. TL typically results in faster training than starting from scratch and is considered an optimization that saves time and improves performance, as stated in [65]. For this purpose, transfer learning is utilized to fine-tune ResNet50 CNN. This involves using pretrained weights from the ImageNet dataset [66] for retraining after preprocessing the collected dataset. The network parameters and hyperparameters are optimized during this process.
## 3.4.6. Classification
The features are taken from ResNet-50 and processed via a fully connected (FC) layer with a $40\%$ dropout rate to avoid overfitting [67, 68]. This layer is then activated with the rectifying function, ReLU. All negative values are set to zero in the input matrix, while other remains unchanged. The use of ReLU leads to faster and more reliable convergence than a sigmoid activation function during training deep networks [69]. The output layer comprises a sigmoid function (binary classifier) to provide class probabilities. The sigmoid function normalizes the input into two outcomes, i.e., malignant vs. benign [70].
## 4. Evaluation and Results
The proposed deep convolutional neural network for mammogram imaging undergoes examination and validation in this section. Information about benchmark datasets, assessment metrics, and comparisons to other leading techniques are also covered.
## 4.1. Image Acquisition Process
The proposed system's performance is evaluated using digitized mammogram images from the INbreast dataset [71]. The database is used to demonstrate the efficiency and reliability of the proposed method for identifying breast cancer. INbreast dataset includes 336 mammogram images, with 269 abnormal and 69 normal images, where 220 are benign and 49 malignant cases. Tables 1 and 2 show the distribution of mammography images.
## 4.2. Metrics of Performance
The purpose of cross-validation is to improve efficiency, validate performance, and assess the results from the dataset. To assess the classification efficiency of the proposed method, multiple metrics are utilized such as confusion matrix, accuracy, sensitivity, specificity, error rate, F1 score, and area under the curve (AUC). All these metrics act as benchmark values for comparing the proposed method against previous algorithms [72]. These measurements are defined as follows.
## 4.2.1. Confusion Matrix
Confusion matrix represents the performance of a classifier in the form of a table. In ML, it is also known as an error matrix. The image regions are labeled positive or negative based on the data type. The classifier's decision can be correct (true) or incorrect (false). This results in four outcomes: true positive (TP), true negative (TN), false positive (FP), and false negative (FN). Correct decisions are represented along the diagonal of the confusion matrix.
## 4.2.2. Accuracy
Accuracy characterizes the suitable labeled images for benign, normal, and malignant mammograms. The accuracy of the process is computed as follows:[1]Accuracy=TP+TNTP+FP+TN+FN.
TP accurately represents positive examples; TN addresses classified negative examples; TN means incorrectly classified examples as accurately classified; and FN indicates accurately classified examples as the wrong sample.
## 4.2.3. Specificity
The chances that the test will correctly recognize the patient who has the disease is shown in the following equation:[2]Specificity=TNTN+FP.
## 4.2.4. Sensitivity
The chance that the test will correctly recognize a patient with the disease is shown in equation:[3]Sensitivity=TPTP+FN.
## 4.2.5. F1 Score
It is a weighted average of precision and recall used for assessing the classifier's performance. It considers both false positives and negatives in its calculation, as shown in the following equation:[4]F1score=2∗Precision∗RecallPrecision+Recall.
## 4.2.6. Area Under the Curve (AUC)
AUC is the classifier's ability to distinguish between benign, normal, and malignant mammograms.
## 5. Results and Discussion
For this study, a subset is taken from the INbreast dataset, and each sample is increased to four images. During the experiment, $60\%$ images were used for training, and the remaining $40\%$ were used for testing. The samples were first subjected to enhancement and segmentation according to the procedures described in the “Methodology” section. Afterward, features were extracted from the samples using a CNN. Finally, all the samples were classified using ResNet-50.
The proposed DCNN method categorizes mammogram images of breast tumors into benign or malignant. A dataset named INbreast is used for experimentation. Table 3 displays the classification accuracy achieved by the proposed ResNet-50 method across the INbreast database. From the INbreast dataset, 132 benign and 29 malignant image samples were selected for training, and 20 malignant and 88 benign for testing. The resulting accuracy is $93\%$. The proposed ResNet-50 approach is also compared quantitatively with previously existing algorithms. The study's results revealed that the presented approach outperformed these algorithms with high accuracy, specificity, F1 score values, and sensitivity.
As shown in Table 3, the proposed approach demonstrated improved results on the INbreast database with an accuracy of $93.0\%$, specificity of $93.86\%$, and sensitivity of $93.83\%$. It outperforms other methods in terms of accuracy. Although the accuracy achieved by [16] is slightly higher at $91.0\%$, the proposed approach still exhibits the best performance compared to the other methods. Compared to existing methods, the proposed approach enhances breast cancer detection and classification performance. It can potentially be used for real-time evaluation and to support radiologists in automating the analysis of mammogram images. However, performance may vary when the same method is applied to different datasets due to factors such as background noise, lighting conditions, occlusion, overfitting, and the nature of the method.
The performance of the presented approach is also evaluated using the confusion matrix and ROC curves. Figure 5 illustrates the confusion matrix on the INbreast data set. AUC, a crucial statistical metric in the ROC curve, is computed using the INbreast data set. Metric in the ROC curve is calculated INbreast data set. ROC curves were constructed based on true positive rate (sensitivity) and false positive rate (1-specificity) rates, controlled by the threshold of the obtained probability maps. Figure 6 shows the ROC curve graph.
Table 4 presents our proposed system's results of breast cancer detection. The proposed approach achieved an F1 score and AUC of $93.03\%$ and $93.02\%$, respectively, on the INbreast database.
In recent years, breast cancer detection and classification applications have gained widespread use in the medical field, making the diagnostic process more accurate [76, 77]. The goal of the proposed method is to enhance clinical diagnosis by enhancing the detection of breast cancer. The opinions of two medical specialists were gathered based on the accuracy level generated by our proposed algorithm. These experts expressed their appreciation for the improved results of ResNet-50 compared to other approaches. To sum it up, the proposed approach enhances performance compared to other methods and can be utilized for real-time evaluations along with helping radiologists automate the evaluation of mammograms.
## 6. Conclusion
The proposed system aimed to detect malignant breast masses and classify benign and malignant tissues in mammograms. A novel computer-aided detection (CAD) system is proposed, which involves thresholding and region-based segmentation techniques. A region-based method with a threshold of 80 determines the largest area included in this threshold. A deep convolutional neural network (DCNN) is utilized during feature extraction. Specifically, the ResNet-50 is retrained to classify the mammograms into two classes (malignant or benign), and its parameters were modified to classify breast mammograms. The proposed approach is applied to the INbreast database to evaluate its performance of the proposed approach. The proposed method achieved an accuracy of $93.0\%$, specificity of $93.86\%$, AUC of $93.02\%$, a sensitivity of $93.83\%$, and an F1 score of $93.03\%$, which are extremely satisfying results. The proposed method surpasses the detection and classification of mammograms, delivering more precise results and improved visual outcomes compared to other systems. The proposed system efficiently detects and classifies malignant breast masses with reduced computation time and produced satisfactory results. Alternative networks, such as deep convolutional networks (VGG) and AlexNet architecture, will be proposed for future development. In the future, we intend to extend this work by collecting large datasets on breast cancer in different age intervals to detect cancer in its early stages.
## Data Availability
The (Breast Cancer Diagnosis) data used to support the findings of this study are included within the article.
## Conflicts of Interest
The authors declare that they have no conflicts of interest.
## References
1. Xian M., Zhang Y., Cheng H.-Da, Xu F., Zhang B., Ding J.. **Automatic breast ultrasound image segmentation: a survey**. (2018) **79** 340-355. DOI: 10.1016/j.patcog.2018.02.012
2. Xu C., Gu Z., Liu J.. **Adenosquamous carcinoma of the breast: a population-based study**. (2021) **28** 848-858. DOI: 10.1007/s12282-021-01222-3
3. Gardezi S. J. S., Elazab A., Lei B., Wang T.. **Breast cancer detection and diagnosis using mammographic data: systematic review**. (2019) **21**. DOI: 10.2196/14464
4. Whang J. S., Baker S. R., Patel R., Luk L., Castro A.. **The causes of medical malpractice suits against radiologists in the United States**. (2013) **266** 548-554. DOI: 10.1148/radiol.12111119
5. Adam F.. **Breast cancer: symptoms, causes, and treatment**. (2022)
6. El-Chami D., Al Haddad M., Abi-Habib R., El-Sibai M.. **Recombinant anthrax lethal toxin inhibits cell motility and invasion in breast cancer cells through the dysregulation of rho gtpases**. (2021) **21** p. 163. DOI: 10.3892/ol.2020.12424
7. Chen W., Zheng R., Zhang S.. **Cancer incidence and mortality in China in 2013: an analysis based on urbanization level**. (2017) **29** 1-10. DOI: 10.21147/j.issn.1000-9604.2017.01.01
8. Al-Antari M. A., Han S.-M., Kim T.-S.. **Evaluation of deep learning detection and classification towards computer-aided diagnosis of breast lesions in digital x-ray mammograms**. (2020) **196**. DOI: 10.1016/j.cmpb.2020.105584
9. Nader R., Tannoury E., Rizk T., Ghanem H., Luvizotto Junior E.. **Atezolizumab-induced encephalitis in a patient with metastatic breast cancer: a case report and review of neurological adverse events associated with checkpoint inhibitors**. (2021) **11**. DOI: 10.4322/acr.2021.261
10. Mansour N., Bodman-Smith K., Khnayzer R. S., Daher C. F.. **A photoactivatable ru (ii) complex bearing 2, 9-diphenyl-1, 10-phenanthroline: a potent chemotherapeutic drug inducing apoptosis in triple negative human breast adenocarcinoma cells**. (2021) **336**. DOI: 10.1016/j.cbi.2020.109317
11. Younes M., Ammoury C., Haykal T., Nasr L., Sarkis R., Rizk S.. **The selective anti-proliferative and pro-apoptotic effect of a. cherimola on mda-mb-231 breast cancer cell line**. (2020) **20** p. 343. DOI: 10.1186/s12906-020-03120-1
12. Punitha S., Al-Turjman F., Stephan T.. **An automated breast cancer diagnosis using feature selection and parameter optimization in ann**. (2021) **90**. DOI: 10.1016/j.compeleceng.2020.106958
13. Mao Ye-J., Lim H.-J., Ni M., Yan W. H., Wong D. W. C., Cheung J. C. W.. **Breast tumour classification using ultrasound elastography with machine learning: a systematic scoping review**. (Jan 2022) **14** p. 367. DOI: 10.3390/cancers14020367
14. Nasser H. A., Assaf S., Aouad G., Mouawad Y., Elamely S.. **Breast manifestations of type i diabetes mellitus**. (2020) **26** 2079-2080. DOI: 10.1111/tbj.13938
15. Al Saud N., Jabbour S., Kechichian E.. **A systematic review of zosteriform rash in breast cancer patients: an objective proof of flap reinnervation and a management algorithm**. (2018) **81** 456-461. DOI: 10.1097/sap.0000000000001505
16. Dhungel N., Carneiro G., Bradley A. P.. **A deep learning approach for the analysis of masses in mammograms with minimal user intervention**. (2017) **37** 114-128. DOI: 10.1016/j.media.2017.01.009
17. Byra M.. **Breast mass classification with transfer learning based on scaling of deep representations**. (2021) **69**. DOI: 10.1016/j.bspc.2021.102828
18. Hitti E., Hadid D., Melki J., Kaddoura R., Alameddine M.. **Mobile device use among emergency department healthcare professionals: prevalence, utilization and attitudes**. (2021) **11** p. 1917. DOI: 10.1038/s41598-021-81278-5
19. Shen Li, Margolies L. R., Rothstein J. H., Fluder E., McBride R., Sieh W.. **Deep learning to improve breast cancer detection on screening mammography**. (Aug 2019) **9**. DOI: 10.1038/s41598-019-48995-4
20. Jansson D., Dieriks V. B., Rustenhoven J.. **Cardiac glycosides target barrier inflammation of the vasculature, meninges and choroid plexus**. (Feb 2021) **4** p. 260. DOI: 10.1038/s42003-021-01787-x
21. El Zarif O., Haraty R. A.. **Toward information preservation in healthcare systems**. (2020) 163-185
22. Salloum G., Tekli J.. **Automated and personalized nutrition health assessment, recommendation, and progress evaluation using fuzzy reasoning**. (2021) **151**. DOI: 10.1016/j.ijhcs.2021.102610
23. Song B.-Il. **A machine learning-based radiomics model for the prediction of axillary lymph-node metastasis in breast cancer**. (2021) **28**
24. Mukand N. H., Ko N. Y., Nabulsi N. A.. **The association between physical health-related quality of life, physical functioning, and risk of contralateral breast cancer among older women**. (2021) **29**. DOI: 10.1007/s12282-021-01309-x
25. Sadad T., Munir A., Saba T., Hussain A.. **Fuzzy c-means and region growing based classification of tumor from mammograms using hybrid texture feature**. (2018) **29** 34-45. DOI: 10.1016/j.jocs.2018.09.015
26. Jubeen M., Rahman H., Rahman A. U.. **An automatic breast cancer diagnostic system based on mammographic images using convolutional neural network classifier**. (2022) **4** 77-86. DOI: 10.56979/401/2022/93
27. Qader Zeebaree D., Mohsin Abdulazeez A., Asaad Zebari D., Haron H., Nuzly Abdull Hamed H.. **Multi-level fusion in ultrasound for cancer detection based on uniform lbp features**. (2021) **66** 3363-3382. DOI: 10.32604/cmc.2021.013314
28. Maqsood S., Damaševičius R., Shah F. M., MaskeliūNas R.. **Detection of macula and recognition of aged-related macular degeneration in retinal fundus images**. (2021) **40** 957-987. DOI: 10.31577/cai_2021_5_957
29. Al-Antari M. A., Al-Masni M. A., Choi M.-T., Han S.-M., Kim T.-S.. **A fully integrated computer-aided diagnosis system for digital x-ray mammograms via deep learning detection, segmentation, and classification**. (2018) **117** 44-54. DOI: 10.1016/j.ijmedinf.2018.06.003
30. Chougrad H., Zouaki H., Alheyane O.. **Deep convolutional neural networks for breast cancer screening**. (2018) **157** 19-30. DOI: 10.1016/j.cmpb.2018.01.011
31. Kooi T., Litjens G., van Ginneken B.. **Large scale deep learning for computer aided detection of mammographic lesions**. (2017) **35** 303-312. DOI: 10.1016/j.media.2016.07.007
32. Maqsood S., Javed U.. **Multi-modal medical image fusion based on two-scale image decomposition and sparse representation**. (2020) **57**. DOI: 10.1016/j.bspc.2019.101810
33. Muzammil S. R., Maqsood S., Haider S., Damaševičius R.. **Csid: a novel multimodal image fusion algorithm for enhanced clinical diagnosis**. (2020) **10** p. 904. DOI: 10.3390/diagnostics10110904
34. Maqsood S., Javed U., Riaz M. M., Muzammil M., Muhammad F., Kim S.. **Multiscale image matting based multi-focus image fusion technique**. (2020) **9** p. 472. DOI: 10.3390/electronics9030472
35. Jabeen K., Khan M. A., Alhaisoni M.. **Breast cancer classification from ultrasound images using probability-based optimal deep learning feature fusion**. (2022) **22** p. 807. DOI: 10.3390/s22030807
36. Yassin N. I. R., Omran S., El Houby E. M., Allam H., Allam H.. **Machine learning techniques for breast cancer computer aided diagnosis using different image modalities: a systematic review**. (2018) **156** 25-45. DOI: 10.1016/j.cmpb.2017.12.012
37. Rezaei Z.. **A review on image-based approaches for breast cancer detection, segmentation, and classification**. (2021) **182**. DOI: 10.1016/j.eswa.2021.115204
38. Dhungel N., Carneiro G., Bradley A. P.. **Deep learning and structured prediction for the segmentation of mass in mammograms**. 605-612
39. Zhang Yu-D., Satapathy S. C., Guttery D. S., Górriz J. M., Wang S. H.. **Improved breast cancer classification through combining graph convolutional network and convolutional neural network**. (2021) **58**. DOI: 10.1016/j.ipm.2020.102439
40. Suh Y. J., Jung J., Cho B.-J.. **Automated breast cancer detection in digital mammograms of various densities via deep learning**. (2020) **10** p. 211. DOI: 10.3390/jpm10040211
41. Chouhan N., Khan A., Shah J. Z., Hussnain M., Khan M. W.. **Deep convolutional neural network and emotional learning based breast cancer detection using digital mammography**. (2021) **132**. DOI: 10.1016/j.compbiomed.2021.104318
42. Muduli D., Dash R., Majhi B.. **Automated breast cancer detection in digital mammograms: a moth flame optimization based elm approach**. (2020) **59**. DOI: 10.1016/j.bspc.2020.101912
43. Wang Yi, Choi E. J., Choi Y., Zhang H., Jin G. Y., Ko S. B.. **Breast cancer classification in automated breast ultrasound using multiview convolutional neural network with transfer learning**. (2020) **46** 1119-1132. DOI: 10.1016/j.ultrasmedbio.2020.01.001
44. Masud M., Eldin Rashed A. E., Hossain M. S.. **Convolutional neural network-based models for diagnosis of breast cancer**. (2020) **34** 11383-11394. DOI: 10.1007/s00521-020-05394-5
45. Eroğlu Y., Yildirim M., Çinar A.. **Convolutional neural networks based classification of breast ultrasonography images by hybrid method with respect to benign, malignant, and normal using mrmr**. (2021) **133**. DOI: 10.1016/j.compbiomed.2021.104407
46. Zhang H., Han L., Chen Ke, Peng Y., Lin J.. **Diagnostic efficiency of the breast ultrasound computer-aided prediction model based on convolutional neural network in breast cancer**. (2020) **33** 1218-1223. DOI: 10.1007/s10278-020-00357-7
47. Braz Junior G., da Rocha S. V., de Almeida J. D. S., de Paiva A. C., Silva A. C., Gattass M.. **Breast cancer detection in mammography using spatial diversity, geostatistics, and concave geometry**. (2019) **78** 13005-13031. DOI: 10.1007/s11042-018-6259-z
48. Harefa J., Alexander A., Pratiwi M.. **Comparison classifier: support vector machine (svm) and k-nearest neighbor (k-nn) in digital mammogram images**. (2017) **2** 35-40
49. Ragab D. A., Sharkas M., Marshall S., Ren J.. **Breast cancer detection using deep convolutional neural networks and support vector machines**. (2019) **7** p. e6201. DOI: 10.7717/peerj.6201
50. Kousalya K., Saranya T.. **Improved the detection and classification of breast cancer using hyper parameter tuning**. (2021). DOI: 10.1016/j.matpr.2021.03.707
51. Irfan R., Almazroi A. A., Rauf H. T., Damasevicius R., Nasr E. A., Abdelgawad A. E.. **Dilated semantic segmentation for breast ultrasonic lesion detection using parallel feature fusion**. (2021) **11** p. 1212. DOI: 10.3390/diagnostics11071212
52. Pisano E. D., Zong S., Hemminger B. M.. **Contrast limited adaptive histogram equalization image processing to improve the detection of simulated spiculations in dense mammograms**. (1998) **11** 193-200. DOI: 10.1007/bf03178082
53. Sahakyan A., Sarukhanyan H.. **Segmentation of the breast region in digital mammograms and detection of masses**. (2012) **3**. DOI: 10.14569/ijacsa.2012.030218
54. Abid I., Almakdi S., Rahman H.. **A convolutional neural network for skin lesion segmentation using double u-net architecture**. (2022) **33** 1407-1421. DOI: 10.32604/iasc.2022.023753
55. Ahmad R., Khalid A., Rahman H.. **Brain tumor detection using image segmentation and classification**. (2020) **1** 59-65
56. Abbas M., Arshad M., Rahman H.. **Detection of breast cancer using neural networks**. (2020) **1** 75-88
57. Kaur D., Kaur Y.. **Various image segmentation techniques: a review**. (2014) **3** 809-814
58. Khan W.. **Image segmentation techniques: a survey**. (2014) **1** 166-170. DOI: 10.12720/joig.1.4.166-170
59. Kaur M., Goyal P.. **A review on region based segmentation**. (2015) **4** 3194-3197
60. Han J., Zhang D., Hu X.. **Background prior-based salient object detection via deep reconstruction residual**. (2014) **25** 1309-1321
61. Zabalza J., Ren J., Zheng J.. **Novel segmented stacked autoencoder for effective dimensionality reduction and feature extraction in hyperspectral imaging**. (2016) **185** 1-10. DOI: 10.1016/j.neucom.2015.11.044
62. LeCun Y., Kavukcuoglu K., Farabet C.. **Convolutional networks and applications in vision**. 253-256
63. Alexandre Spanhol F., Oliveira L. S., Petitjean C., Laurent H.. **Breast cancer histopathological image classification using convolutional neural networks**. 2560-2567
64. Bengio Y., Courville A., Vincent P.. **Representation learning: a review and new perspectives**. (2013) **35** 1798-1828. DOI: 10.1109/tpami.2013.50
65. Raúl de la F. L.. **Wild Data - Transfer Learning - Deep Learning - Blog**. (2018)
66. He K., Zhang X., Ren S., Sun J.. **Deep residual learning for image recognition**. 770-778
67. Srivastava N., Hinton G., Krizhevsky A., Sutskever I., Salakhutdinov R.. **Dropout: a simple way to prevent neural networks from overfitting**. (2014) **15** 1929-1958
68. Kuppusamy P., Kumari N. M. J., Alghamdi W. Y.. **Job scheduling problem in fog-cloud-based environment using reinforced social spider optimization**. (2022) **11** p. 99. DOI: 10.1186/s13677-022-00380-9
69. Brownlee J.. **A gentle introduction to the rectified linear unit (relu) for deep learning neural networks**. (Apr 2019)
70. Al-Haija Q. A., Manasra G. F.. **Development of breast cancer detection model using transfer learning of residual neural network (resnet-50)**. (2020) **1** 30-39. DOI: 10.15864/ajse.1304
71. Moreira I. C., Amaral I., Domingues I., Cardoso A., Cardoso M. J., Cardoso J. S.. **Inbreast: toward a full-field digital mammographic database**. (2012) **19** 236-248. DOI: 10.1016/j.acra.2011.09.014
72. Dhasarathan C., Hasan M. K., Islam S.. **Covid-19 health data analysis and personal data preserving: a homomorphic privacy enforcement approach**. (2023) **199** 87-97. DOI: 10.1016/j.comcom.2022.12.004
73. Carneiro G., Nascimento J., Bradley A. P.. **Automated analysis of unregistered multi-view mammograms with deep learning**. (2017) **36** 2355-2365. DOI: 10.1109/tmi.2017.2751523
74. Zhang H., Wu R., Yuan T.. **De-ada**. (2020) **539** 461-486. DOI: 10.1016/j.ins.2020.05.080
75. Pezeshki H., Rastgarpour M., Sharifi A., Yazdani S.. **Extraction of spiculated parts of mammogram tumors to improve accuracy of classification**. (2019) **78** 19979-20003. DOI: 10.1007/s11042-019-7185-4
76. Aleem J., Qureshi P. A. A. A., Babar N., Sultan A., Rehman A. U. **Metastatic choriocarcinoma of the breast: a rare entity**. (2022) **14**. DOI: 10.7759/cureus.22417
77. Shaban M., Raza S. E. A., Hassan M.. **A digital score of tumour-associated stroma infiltrating lymphocytes predicts survival in head and neck squamous cell carcinoma**. (2022) **256** 174-185. DOI: 10.1002/path.5819
|
---
title: Electroacupuncture Zusanli (ST36) Relieves Somatic Pain in Colitis Rats by
Inhibiting Dorsal Root Ganglion Sympathetic-Sensory Coupling and Neurogenic Inflammation
authors:
- Yi-li Wang
- Hai-yan Zhu
- Xi-qian Lv
- Xing-ying Ren
- Ying-chun Peng
- Jin-yu Qu
- Xue-fang Shen
- Ran Sun
- Meng-lu Xiao
- Hong Zhang
- Zhao-hui Chen
- Peng Cong
journal: Neural Plasticity
year: 2023
pmcid: PMC9998159
doi: 10.1155/2023/9303419
license: CC BY 4.0
---
# Electroacupuncture Zusanli (ST36) Relieves Somatic Pain in Colitis Rats by Inhibiting Dorsal Root Ganglion Sympathetic-Sensory Coupling and Neurogenic Inflammation
## Abstract
Referred somatic pain triggered by hyperalgesia is common in patients with inflammatory bowel disease (IBD). It was reported that sprouting of sympathetic nerve fibers into the dorsal root ganglion (DGR) and neurogenic inflammation were related to neuropathic pain, the excitability of neurons, and afferents. The purpose of the study was to explore the potential and mechanism of electroacupuncture (EA) at Zusanli (ST36) for the intervention of colon inflammation and hyperalgesia. Sprague-Dawley (SD) was randomly divided into four groups, including control, model, EA, and sham-EA. Our results showed EA treatment significantly attenuated dextran sulfate sodium- (DSS-) induced colorectal lesions and inflammatory cytokine secretion, such as TNF-α, IL-1β, PGE2, and IL-6. EA also inhibited mechanical and thermal pain hypersensitivities of colitis rats. Importantly, EA effectively abrogated the promotion effect of DSS on ipsilateral lumbar 6 (L6) DRG sympathetic-sensory coupling, manifested as the sprouting of tyrosine hydroxylase- (TH-) positive sympathetic fibers into sensory neurons and colocalization of and calcitonin gene-related peptide (CGRP). Furthermore, EA at Zusanli (ST36) activated neurogenic inflammation, characterized by decreased expression of substance P (SP), hyaluronic acid (HA), bradykinin (BK), and prostacyclin (PGI2) in colitis rat skin tissues corresponding to the L6 DRG. Mechanically, EA treatment reduced the activation of the TRPV1/CGRP, ERK, and TLR4 signaling pathways in L6 DRG of colitis rats. Taken together, we presumed that EA treatment improved colon inflammation and hyperalgesia, potentially by suppressing the sprouting of sympathetic nerve fibers into the L6 DGR and neurogenic inflammation via deactivating the TRPV1/CGRP, ERK, and TLR4 signaling pathways.
## 1. Introduction
Inflammatory bowel disease (IBD) is a chronic, life-threatening inflammatory disease affecting the gastroenterological system, which includes Crohn's disease (CD) and ulcerative colitis (UC) [1, 2]. The etiology and the pathogenesis of inflammatory bowel diseases (IBD) are still not completely understood [3]. Studies have confirmed that hyperalgesia caused by IBD was closely related to the hypersensitivity response triggered by sympathetic-sensory coupling in the skin-dorsal root ganglion (DRG) [4, 5]. A previous study illustrated that sympathetic nerve sprouting in DRG was observed in a model of colitis [6]. On the other hand, neurogenic inflammation was the important mechanism of referred pain caused by visceral lesions on body surfaces [7]. The activation and release of mediators, such as substance P (SP) from unmyelinated afferent nerve endings caused venule vasodilation and increased permeability, which led to inflammatory reactions such as plasma extravasation and edema [8, 9]. Thus, regulation of sympathetic-sensory coupling and neurogenic inflammation provides a new idea for the treatment of visceral hyperalgesia.
It has been well established in traditional Chinese medicine that acupoints which have a diagnostic and curative effect on diseases are an important position for the correlation between the meridian and visceral organs [10, 11]. The acupoints are the paresthesias in corresponding parts of the body surface through neurogenic involved responses in the pathological process of the body [12]. Recently, study reported that neurogenic inflammatory sites were found on the dorsal trunk cutaneous of the liver injury rat model, which was matched with locations of acupoints [13]. Meanwhile, electroacupuncture at neurogenic spots reduced bile duct ligation-induced liver injury [13]. Similarly, our previous study found that colitis rats exhibited secondary hyperalgesia, accompanied by sensitization phenomenon in the Zusanli (ST36) acupoints. Electroacupuncture on Zusanli (ST36) acupoints significantly reduced colon lesions and relieved somatic referred pain of colitis rats [14]. However, the underlying mechanism of electroacupuncture at Zusanli (ST36) relieving colitis and pain hypersensitivity has not been studied.
In this study, we explored whether electroacupuncture at Zusanli (ST36) inhibits DRG sympathetic-sensory coupling and surface neurogenic inflammatory response, thereby reducing colon lesions and eliminating referred pain in a colitis rat model. Furthermore, we also investigated the molecular mechanisms of electroacupuncture at Zusanli (ST36) in the treatment of colitis by regulating the DRG sympathetic-sensory coupling and neurogenic inflammation.
## 2.1. Animal Treatment
All experiments were approved according to the Ethics Committee the Institutional Animal Welfare and Use Committee of the Institute of Acupuncture-Moxibustion, China Academy of Chinese Medicine (no. 20170313). 32 Sprague-Dawley (SD) male rats (SPF grade, 12 weeks, 180-200 g) were purchased from Chengdu Dossy Experimental Animals Co., Ltd. (Chengdu, Sichuan; SCXY (Chuan) 2020-034). The feeding environment was 23 ± 1°C, relative humidity 50 ± $5\%$, and light/darkness for 12 h circulation. SD rats are allowed to eat and drink freely. The SD rats were randomly divided into 4 groups ($$n = 8$$), namely, the control group, colitis model group, Zusanli electroacupuncture (Zusanli-EA) group, and sham electroacupuncture (sham-EA) group. For the colitis model group, rats were gavaged with $5\%$ (w/v) dextran sulfate sodium (DSS) saline solution (MP Biomedicals, Santa Ana, California, USA) for 4 days (50 mL/d) as previously described [15, 16]. The status of rats was monitored using the disease activity index (DAI) [17]. Meanwhile, the control group rats were gavaged an equal volume of saline solution. For the Zusanli-EA group, electroacupuncture was immediately performed under isoflurane inhalation anesthesia after modeling. Rats received electroacupuncture treatment at Zusanli acupoint (ST36, bilateral) using 1.0-inch filigree needles (0.25 mm × 13 mm, Huatuo Brand, depth of about 7 mm). An electroacupuncture treatment device (G6805-2A, Huatuo Brand) was from Suzhou Medical Appliance Factory, China. The electroacupuncture parameter is a dilute wave with $\frac{2}{100}$ Hz, the intensity of 1 mA, and performed for 15 minutes, once a day, for 21 consecutive days as previously described [18]. For the sham-EA group, rats were anesthetized by inhalation of 3–$4\%$ isoflurane and then received sham electroacupuncture with a pragmatic placebo needle on sham acupoints. The neurological function and DAI score of rats were tested per week. At the end of the 3-week administration, all rats were anesthetized with $1\%$ sodium pentobarbital (50 mg/kg) and euthanized. The colon tissues, serum, ipsilateral lumbar 6 (L6) dorsal root ganglia (DRG), and nearby skin were removed and kept at -80°C for subsequent analysis. The flow of subjects through the experimental procedure is described in Figure 1(a).
## 2.2. Measurement of Thermal Sensitivity
A BME-410C thermal stimulation meter (Tianjin Berne Technology Co., Ltd., Tianjin, China) was used to detect thermal withdrawal latency (TWL) of rat hind paws according to published methods [19]. Briefly, the rats were placed in a hot plate instrument (52°C), and the paw-withdrawal latency is defined as the time since the foot touches the hot plate instrument until the hind paw licking (s).
## 2.3. Measurement of Mechanical Sensitivity
A BME-404 electrical mechanical analgesia tester (Institute of Biomedical Engineering Chinese Academy of Medical Sciences, Tianjin, China) was used to measure the mechanical withdrawal threshold (MWT) of the rat hind paw. Stainless steel filaments (0.6 mm in diameter) were employed to stimulate the plantar surface of the left hind paw with pressure. When a retracted paw response occurs, the force (g) was automatically recorded.
## 2.4. Hematoxylin and Eosin (H&E) Stain
The colon tissues were collected and fixed in $4\%$ paraformaldehyde overnight, processed, and embedded in paraffin. The tissue sections were stained with hematoxylin and eosin (H&E) to observe the degree of the lesion and inflammatory cell infiltration under 10x and 400x magnification optical microscope (Olympus BH2, Tokyo, Japan).
## 2.5. Transmission Electron Microscopy (TEM)
The pathological changes in colon tissues were observed by TEM. Briefly, colon tissues were fixed with $3\%$ glutaraldehyde for 15 min and postfixed with $1\%$ osmium tetroxide for 2 h at 4°C. The colon tissues were then incubated with propanone for 2 h and embedded with Ep812 resin. The blocks were sliced with a Leica EM UC7, and sample sections were stained with uranium acetate-lead citrate. A JEM-1400Flash transmission electron microscopy (JEOL; Tokyo, Japan) was used to examine.
## 2.6. Immunohistochemistry (IHC) Stain
Paraffin blocks were sectioned at 4 μm. IHC stain was performed to detect the 5-hydroxytryptamine (5-HT) expression in skin tissue corresponding to the L6 DRG of mice according to instructions of the IHC protocol. 5-HT antibody was purchased from Beijing Zhongshan Golden Bridge Biotechnology Co., Ltd. (BS-1126R, Beijing, China; $\frac{1}{100}$). IHC images were captured using a digital trinocular camera microscope camera system (BA400Digital, Motic Instruments, Inc., Baltimore, MD, USA). Image quantification was performed using Halo software (Indica Labs; Albuquerque, NM).
## 2.7. Immunofluorescence (IF) Stain
Paraffin sections of L6 DRG were dewaxed and hydrated. The sections were incubated in QuickBlock™ Blocking Buffer (Beyotime Biotechnology, Jiangsu, China, P0260) for 30 min at room temperature. Then, the sections were incubated with calcitonin gene-related peptide (CGRP) antibody (bs-0791R, Beijing Bosen Biological Technology Co., Ltd. Beijing, China, $\frac{1}{100}$), tyrosine hydroxylase (TH) antibody (ab129991, Abcam; $\frac{1}{100}$), or NeuN (ab129991, Abcam; $\frac{1}{100}$) antibody at 4°C overnight and washed 3 times with phosphate-buffered saline (PBS, ZLI-9062, Beijing Zhongshan Golden Bridge Biotechnology Co., Ltd., Beijing, China). Then, DAPI (ZLI-9557; Beijing Zhongshan Golden Bridge Biotechnology Co., Ltd., Beijing, China) was added dropwise into the sections for 5 min. The staining was observed under a fluorescence microscope Olyvia (Olympus, Tokyo, Japan) at 100x and 400x magnifications. Image processing was conducted with Image J (National Institutes of Health, Bethesda, MA, USA).
## 2.8. Western Blot Analysis
L6 DRG tissues were treated using RIPA buffer (Signaling Technology, Inc.). The concentration of protein was determined by a BCA kit (Sigma-Aldrich; Merck KGaA). Total protein (30 μg/sample) was separated via $10\%$ SDS-PAGE. And then, the separated proteins are transferred to nitrocellulose membranes. The membranes were blocked with $5\%$ nonfat dried milk overnight at 4°C and incubated with corresponding protein antibodies. Then, the membranes were washed with Tris-buffered saline/$0.1\%$ Tween (TBST) and incubated for 1.5 hours with an HRP Goat anti-Rabbit IgG. The bands were visualized using the ECL system (Affinity Biosciences, Cincinnati, Ohio, USA), and β-actin was used as an internal control. The net optical density was measured using Quantity One software (Bio-Rad). Antibody information used for Western blot analysis is shown in Table 1.
## 2.9. Detection of the Levels of Inflammatory Cytokines in Serum and Pain-Causing Substances in Skin Tissues
The levels of interleukin (IL)-1β, IL-6, tumor necrosis factor (TNF)-α, prostaglandin E2 (PGE2), and IL-10 in rat serum and substance P (SP), hyaluronic acid (HA), bradykinin (BK), and prostacyclin (PGI2) in rat skin tissues (corresponding to the L6 DRG) were examined by quantizing enzyme-linked immunosorbent assay (ELISA) kit (ZhuoCai Biological Technology, China) based on the manufacturer's instructions. The absorbance of wells was measured with a microplate reader (SpectraMAX Plus384, USA) at 450 nm wavelength to calculate the sample concentration.
## 2.10. Statistical Analysis
The data were represented as means ± standard deviation. Statistical analysis was performed using SPSS 20.0 (IBM Corp.). One-way analysis of variance (ANOVA) with Tukey's post hoc test and Student's t-test was used for statistical analysis. Differences with a $P \leq 0.05$ were considered to indicate statistically significant.
## 3.1. EA Attenuated Colitis Severity and Somatic Hyperalgesia in DSS-Induced Colitis Rats
DAI score was undertaken weekly to evaluate the progression of colitis. As compared with the control, the DAI score was significantly increased on day 7 after DSS treatment (Figure 1(b)). EA treatment continuously reduced the DAI score. On days 11 and 14, the DAI score was significantly decreased compared with the sham-EA group (Figure 1(b)). Meanwhile, colitis rats displayed lower TWL and MWT on days 11 and 14, which were dramatically corrected by EA treatment (Figures 1(c) and 1(d)). These results indicated that the therapeutic effect of EA at Zusanli in the early stage of colitis in rats is significant. Furthermore, the results of the H&E stain showed that model group colonic mucosa was severely damaged, with a high number of infiltrating inflammatory cells in comparison to the control group (Figure 2(a)). It was found that the EA group exhibited significantly reduced mucosal injury and infiltration of inflammatory cells compared with the sham-EA group (Figure 2(a)). Meanwhile, we also used TEM to observe the pathological changes of colon tissues. As compared with the control, colitis rats exhibited severe colonic injury marked by severe damage appeared including transmural infiltration of inflammatory cells, necrosis, and destruction of crypts, massive transmural inflammatory cell infiltration, and thickening of the colonic wall (Figure 2(b)). EA treatment significantly improved DSS-induced damage to colon tissues compared with the sham-EA group (Figure 2(b)). These data suggested that EA treatment alleviated DSS-induced colonic injury and hypersensitivity in colitis rats.
## 3.2. EA Attenuated Sympathetic-Sensory Coupling of L6 DRG in DSS-Induced Colitis Rats
Compared with the control group, the distribution of TH-positive sympathetic fibers in the DRG was increased at 21 d after DSS induction (Figures 3(a) and 3(b)). Meanwhile, the TH-positive sympathetic fibers wrapped around sensory neurons to form sympathetic-sensory coupling (Figures 3(a) and 3(b)). EA treatment decreased the expression of TH and inhibited sympathetic fibers from wrapping around sensory neurons (Figure 3(a)). In addition, sensory nerve fibers were labeled with CGRP, which was a factor in neurogenic inflammatory responses. As shown in Figure 3(b), the density of CGRP-positive sensory nerve fibers and TH-positive sympathetic fibers were both increased in L6 DRG of colitis rats (Figures 3(c) and 3(d)). Compared with the sham-EA group, EA treatment suppressed sympathetic nerve sprouting in L6 DRG of colitis rats (Figures 3(c) and 3(d)). These findings confirmed that DSS induction promoted the sprouting of sympathetic nerve fibers which corresponded to the increased sympathetic nerve activity. EA intervention ameliorated these changes.
## 3.3. EA Decreased Neurogenic Inflammatory on Body Surfaces in DSS-Induced Colitis Rats
To evaluate DSS-associated neurogenic inflammatory on body surfaces, we performed an ELISA assay. Three weeks after DSS induction, the expression levels of neurogenic inflammatory response-related inflammatory and pain-causing substances SP, HA, BK, and PGI2 in skin tissues were significantly increased compared with the control group (Figure 4(a)). EA treatment decreased the expression of SP, HA, BK, and PGI2 compared with the sham-EA group (Figure 4(a)). Meanwhile, the secretion of proinflammatory factors IL-1β, IL-6, TNF-α, and PGE2 was enhanced, as well as the levels of anti-inflammatory factor IL-10 were reduced in the serum of colitis rats, which was all reversed by EA treatment (Figure 4(b)). IHC stain showed that DSS induction promoted neurotransmitter 5-HT expression in skin tissues, which was also eliminated by EA treatment (Figures 4(c) and 4(d)). Meanwhile, TH was reported to be a key enzyme that regulates neurotransmitters in nerve cells [20]. The increased TH expression induced by DSS was dramatically blocked by EA treatment in L6 DRG of colitis rats (Figures 4(e) and 4(f)).
## 3.4. EA Inhibited TRPV1/CGRP, ERK, and TLR4 Signaling Pathways in DSS-Induced Colitis Rats
To elucidate the underlying cellular mechanism of EA stimulation, we analyzed whether EA at ST36 augmented TRPV1/CGRP, ERK, and TLR4 signaling pathway activation, which were known to be a crucial role in the neurogenic inflammatory. We found that DSS induction resulted in the activation of TRPV1/CGRP, ERK, and TLR4 signaling pathways, characterized by the increased expression of TRPV1, CGRP, MEK, p-MEK, CREB, p-CREB, TLR4, IRF3, p-IRF3, P-65, and p-P65. EA stimulation significantly inhibited this process, whereas EA at nonacupoint did not produce a significant improvement (Figures 5(a)–5(j)).
## 4. Discussion
It is well known that referred somatic pain was caused by activation of primary nociceptive afferents by visceral lesions stimuli [21]. Referred somatic pain was accompanied by secondary hyperalgesia, reflex muscle spasms, deep tenderness, and autonomic hyperactivity [22, 23]. Studies have found that the sympathetic nervous system was closely related to hyperalgesia due to visceral lesions [24–26]. Sympathetic nerve sprouts in the DRG by coupling around neurons to form sympathetic-sensory coupling were found in animal models of pathological conditions. Referred somatic pain and peak sympathetic sprouting were observed in the neuropathic pain model of the cuff and spared nerve injury (SNI) in the sciatic territory [27]. A previous study found that nerve growth factor (NGF) released from the sprouted sympathetic fibers in the synovial membrane and upper dermis contributed to the pain-related behavior associated with arthritis [28]. It has been shown that jaw pain due to myocardial ischemia could be explained by the convergence of cardiac visceral afferent fibers with spinothalamic tract (STT) neurons [29]. In a rat model of trinitrobenzene sulfonic acid- (TNBS-) induced colitis, sympathetic nerve fiber sprout was found in the DRG of the lumbosacral segment (L6, S1), manifested by tyrosine hydroxylase- (TH-) positive nerve fibers increased [30]. Dextran sulfate sodium (DSS) treatment caused mechanical hypersensitivity in the abdominal and facial skin of colitis mice by increasing TRPA1 expression in cultured DRG neurons and selectively enhanced currents evoked by the TRPA1 agonist [31]. Importantly, in the current study, we detected that DSS induced colon tissue lesions and somatic hyperalgesia in rats. Moreover, DSS treatment increased sprouting of sympathetic fibers in L6 DRG into the sensory ganglia.
On the other hand, neurogenic inflammation was an important part of the pathogenesis of referred somatic pain. Primary afferent nociceptive neurons released neuropeptides to the periphery, leading to mast cell degranulation and the release of biologically active substances that produce pain and/or inflammation such as CGRP, SP, 5-HT, HA, BK, and PGI2, which induced neurogenic inflammation characterized by vasodilatation, protein extravasation, and leukocyte migration [32, 33]. In the inflammatory pain model, PGI2 was involved in pain transmission at the spinal cord [34]. 5-HT participated in the mediation of joint pain in experimental arthritis by exciting and sensitizing the medial articular afferent nerve [35]. A previous report indicated that the blockade of receptor channels such as TRPV1 and TRPA1 on nociceptive sensory neurons was shown to attenuate experimental colitis by suppressing the release of GRP and SP [36]. Furthermore, HA sensitized the nociceptor TRPV1 in mouse nociceptive dorsal root ganglion neurons and was known to contribute to relieving visceral hypersensitivity, symptoms, and abdominal pain in IBD patients [37]. Our results demonstrated that DSS colitis upregulated SP, HA, BK, PGI2, and 5-HT expression in the skin and increased TRPV1, TRPA1, and TH in L6 DRG, implying that DSS promoted surface neurogenic inflammation evoked by sympathetic-sensory coupling in skin-DRG.
EA, as a traditional therapeutic method, has been used to treat IBD and hyperalgesia in China. EA treatment at Zusanli (ST36) attenuated the macroscopic damage and the myeloperoxidase activity of colonic samples [38]. Furthermore, EA at Zusanli (ST36) and Guanyuan (CV4) activated microglia in hippocampus CA1 and CA3 regions of DSS-induced colitis mice [39]. Interestingly, EA at Zusanli (ST36) and Shangjuxu (ST37) significantly reduced the severity of colonic inflammation, as well as the visceral hypersensitivity and referral hind paw hyperalgesia in colitis rats by increasing [40]. EA at Zusanli (ST36) eliminated the expression and activation of mast cells and improved visceral hypersensitivity in experimental colitis [41]. The mechanism may be via inhibiting of NGF/TrkA/TRPV1 peripheral afferent pathway triggered by the mast cells [41]. We found that EA at Zusanli (ST36) improved the pathological state of the colon tissues and referred somatic pain in a colitis rat model, which was related to the inhibiting of sensory-sympathetic coupling in L6 DRG and neurogenic inflammation in the skin.
As is well known, TLR4 initiated downstream genes such as NF-KB and IRF3 and activated the expression of inflammatory factors such as IL-1β, IL-6, and TNF-a, thereby inducing inflammatory responses and pain-related hypersensitivity. Increased expression of TLR4, p-p65, TNF-α, and IL-1β in (L4/L5) DRGs was observed in a postoperative pain model [42]. Importantly, TNBS treatment enhanced TLR4 and TRPV1 coexpression in primary afferents including the trigeminal sensory neurons and DGR neurons of colitis mice [43]. Additionally, suppressing the synthesis of ERK in DRG has proven effective to alleviate hyperalgesia. The deactivation of the MEK/ERK pathway in the DRG of chronic constriction injury rats alleviated neuropathic pain development [44]. H2O2-induced hyperalgesia was related to increased phosphorylation of ERK in neurons of DRG [45]. A recent study found that in a rat model of colitis, the activation of ERK5 mediated BDNF upregulation in the DRG primary afferent neurons [46]. In this research, we demonstrated that DSS induction activated TRPV1/CGRP, ERK, and TLR4 signaling pathways, which were significantly offset by EA at Zusanli (ST36).
In conclusion, EA at Zusanli (ST36) relieved hyperalgesia induced by colitis via the inhibition of surface neurogenic inflammation and sympathetic sprouting into the DRG, which were mediated by TRPV1/CGRP, ERK, and TLR4 signaling pathway deactivation. EA at Zusanli (ST36) may be an effective treatment for referred somatic pain in UC patients.
## Data Availability
The datasets used or analyzed during the current study are available from the corresponding author upon reasonable request.
## Conflicts of Interest
The authors declare that they have no competing interests.
## Authors' Contributions
YL W, HY Z, and P C conceived and designed the experiments. YL W, HY Z, XQ L, XY R, YC P, JY Q, XF S, and R S performed the experiments. YL W, HY Z, ML X, H Z, and ZH C analyzed the data. P C contributed to the reagents and materials. YL W and HY Z wrote the manuscript. All authors were substantially involved in the research, acquisition of data, analysis, and manuscript preparation and have read and approved the final submitted manuscript. YL W and HY Z contributed equally to this work and shared first authorship.
## References
1. Sairenji T., Collins K. L., Evans D. V.. **An update on inflammatory bowel disease**. (2017) **44** 673-692. DOI: 10.1016/j.pop.2017.07.010
2. Guan Q.. **A comprehensive review and update on the pathogenesis of inflammatory bowel disease**. (2019) **2019** 16. DOI: 10.1155/2019/7247238
3. Ghouri Y. A., Tahan V., Shen B.. **Secondary causes of inflammatory bowel diseases**. (2020) **26** 3998-4017. DOI: 10.3748/wjg.v26.i28.3998
4. Ceuleers H., Van Spaendonk H., Hanning N.. **Visceral hypersensitivity in inflammatory bowel diseases and irritable bowel syndrome: the role of proteases**. (2016) **22** 10275-10286. DOI: 10.3748/wjg.v22.i47.10275
5. Straub R. H., Grum F., Strauch U.. **Anti-inflammatory role of sympathetic nerves in chronic intestinal inflammation**. (2008) **57** 911-921. DOI: 10.1136/gut.2007.125401
6. Shen S., Tiwari N., Madar J., Mehta P., Qiao L. Y.. **Beta 2-adrenergic receptor mediates noradrenergic action to induce cyclic adenosine monophosphate response element-binding protein phosphorylation in satellite glial cells of dorsal root ganglia to regulate visceral hypersensitivity**. (2022) **163** 180-192. DOI: 10.1097/j.pain.0000000000002330
7. Matsuda M., Huh Y., Ji R. R.. **Roles of inflammation, neurogenic inflammation, and neuroinflammation in pain**. (2019) **33** 131-139. DOI: 10.1007/s00540-018-2579-4
8. Mayer E. A., Raybould H., Koelbel C.. **Neuropeptides, inflammation, and motility**. (1988) **33** 71s-77s. DOI: 10.1007/bf01538134
9. Kim D. H., Ryu Y., Hahm D. H.. **Acupuncture points can be identified as cutaneous neurogenic inflammatory spots**. (2017) **7**. DOI: 10.1038/s41598-017-14359-z
10. Gong Y., Li N., Lv Z.. **The neuro-immune microenvironment of acupoints-initiation of acupuncture effectiveness**. (2020) **108** 189-198. DOI: 10.1002/jlb.3ab0420-361rr
11. Xing J. J., Zeng B. Y., Li J., Zhuang Y., Liang F. R.. **Acupuncture point specificity**. (2013) **111** 49-65. DOI: 10.1016/b978-0-12-411545-3.00003-1
12. Li F., He T., Xu Q.. **What is the acupoint? A preliminary review of acupoints**. (2015) **16** 1905-1915. DOI: 10.1111/pme.12761
13. Yi Y. J., Kim D. H., Chang S., Ryu Y., Kim S. C., Kim H. Y.. **Electroacupuncture at neurogenic spots in referred pain areas attenuates hepatic damages in bile duct-ligated rats**. (2021) **22** p. 1974. DOI: 10.3390/ijms22041974
14. Wang Y. L., Su Y. S., He W., Jing X. H.. **Electroacupuncture relieved visceral and referred hindpaw hypersensitivity in colitis rats by inhibiting tyrosine hydroxylase expression in the sixth lumbar dorsal root ganglia**. (2019) **77, article 101957**. DOI: 10.1016/j.npep.2019.101957
15. Kurt-Jones E. A., Cao L., Sandor F.. **Trefoil family factor 2 is expressed in murine gastric and immune cells and controls both gastrointestinal inflammation and systemic immune responses**. (2007) **75** 471-480. DOI: 10.1128/iai.02039-05
16. Olivera-Castillo L., Grant G., Kantún-Moreno N.. **A glycosaminoglycan-rich fraction from sea cucumber**. (2020) **12** p. 1698. DOI: 10.3390/nu12061698
17. Anzola A., González R., Gámez-Belmonte R.. **miR-146a regulates the crosstalk between intestinal epithelial cells, microbial components and inflammatory stimuli**. (2018) **8** p. 17350. DOI: 10.1038/s41598-018-35338-y
18. Jiang H., Zhang X., Wang Y.. **Mechanisms underlying the antidepressant response of acupuncture via PKA/CREB signaling pathway**. (2017) **2017** 11. DOI: 10.1155/2017/4135164
19. Wang L., Chopp M., Szalad A.. **Sildenafil ameliorates long term peripheral neuropathy in type II diabetic mice**. (2015) **10**. DOI: 10.1371/journal.pone.0118134
20. Daubner S. C., Le T., Wang S.. **Tyrosine hydroxylase and regulation of dopamine synthesis**. (2011) **508** 1-12. DOI: 10.1016/j.abb.2010.12.017
21. Orhurhu V. J., Roberts J. S., Ly N., Cohen S. P.. (2022)
22. Myers D. E.. **The receptive field for visceral pain referred orofacially by the vagus nerves**. (2021) **34** 24-29. DOI: 10.1002/ca.23604
23. Gil D. W., Wang J., Gu C., Donello J. E., Cabrera S., Al-Chaer E. D.. **Role of sympathetic nervous system in rat model of chronic visceral pain**. (2016) **28** 423-431. DOI: 10.1111/nmo.12742
24. Raja S. N.. **Role of the sympathetic nervous system in acute pain and inflammation**. (1995) **27** 241-246. DOI: 10.3109/07853899509031966
25. Lázaro-Navas I., Lorenzo-Sánchez-Aguilera C., Pecos-Martín D.. **Immediate effects of dry needling on the autonomic nervous system and mechanical hyperalgesia: a randomized controlled trial**. (2021) **18**. DOI: 10.3390/ijerph18116018
26. Giamberardino M. A.. **Referred muscle pain/hyperalgesia and central sensitisation**. (2003) **35** 85-88. DOI: 10.1080/16501960310010205
27. Nascimento F. P., Magnussen C., Yousefpour N., Ribeiro-da-Silva A.. **Sympathetic fibre sprouting in the skin contributes to pain-related behaviour in spared nerve injury and cuff models of neuropathic pain**. (2015) **11** p. 59. DOI: 10.1186/s12990-015-0062-x
28. Longo G., Osikowicz M., Ribeiro-da-Silva A.. **Sympathetic fiber sprouting in inflamed joints and adjacent skin contributes to pain-related behavior in arthritis**. (2013) **33** 10066-10074. DOI: 10.1523/jneurosci.5784-12.2013
29. Kikuta S., Dalip D., Loukas M., Iwanaga J., Tubbs R. S.. **Jaw pain and myocardial ischemia: a review of potential neuroanatomical pathways**. (2019) **32** 476-479. DOI: 10.1002/ca.23367
30. Xia C. M., Colomb D. G., Akbarali H. I., Qiao L. Y.. **Prolonged sympathetic innervation of sensory neurons in rat thoracolumbar dorsal root ganglia during chronic colitis**. (2011) **23** p. 801-e339. DOI: 10.1111/j.1365-2982.2011.01728.x
31. Jain P., Materazzi S., De Logu F.. **Transient receptor potential ankyrin 1 contributes to somatic pain hypersensitivity in experimental colitis**. (2020) **10** p. 8632. DOI: 10.1038/s41598-020-65618-5
32. Sorkin L. S., Eddinger K. A., Woller S. A., Yaksh T. L.. **Origins of antidromic activity in sensory afferent fibers and neurogenic inflammation**. (2018) **40** 237-247. DOI: 10.1007/s00281-017-0669-2
33. Kulka M., Sheen C. H., Tancowny B. P., Grammer L. C., Schleimer R. P.. **Neuropeptides activate human mast cell degranulation and chemokine production**. (2008) **123** 398-410. DOI: 10.1111/j.1365-2567.2007.02705.x
34. Doi Y., Minami T., Nishizawa M., Mabuchi T., Mori H., Ito S.. **Central nociceptive role of prostacyclin (IP) receptor induced by peripheral inflammation**. (2002) **13** 93-96. DOI: 10.1097/00001756-200201210-00022
35. Herbert M. K., Schmidt R. F.. **Activation of normal and inflamed fine articular afferent units by serotonin**. (1992) **50** 79-88. DOI: 10.1016/0304-3959(92)90115-r
36. Engel M. A., Becker C., Reeh P. W., Neurath M. F.. **Role of sensory neurons in colitis: increasing evidence for a neuroimmune link in the gut**. (2011) **17** 1030-1033. DOI: 10.1002/ibd.21422
37. Wouters M. M., Balemans D., Van Wanrooy S.. **Histamine receptor H1-mediated sensitization of TRPV1 mediates visceral hypersensitivity and symptoms in patients with irritable bowel syndrome**. (2016) **150** 875-887.e9. DOI: 10.1053/j.gastro.2015.12.034
38. Kang J. W., Kim T. W., La J. H.. **Electroacupuncture ameliorates experimental colitis induced by acetic acid in rat**. (2004) **5** 189-195. DOI: 10.4142/jvs.2004.5.3.189
39. Zhang N., Zhang Q., Xie L.. **Electroacupuncture and moxibustion regulate hippocampus glia and mitochondria activation in DSS-induced colitis mice**. (2020) **2020** 11. DOI: 10.1155/2020/2530253
40. Lv P. R., Su Y. S., He W.. **Electroacupuncture alleviated referral hindpaw hyperalgesia via suppressing spinal long-term potentiation (LTP) in TNBS-induced colitis rats**. (2019) **2019** 11. DOI: 10.1155/2019/2098083
41. Chen Y., Cheng J., Zhang Y., Chen J. D. Z., Seralu F. M.. **Electroacupuncture at ST36 relieves visceral hypersensitivity via the NGF/TrkA/TRPV1 peripheral afferent pathway in a rodent model of post-inflammation rectal hypersensitivity**. (2021) **14** 325-339. DOI: 10.2147/jir.s285146
42. Xing F., Zhang W., Wen J.. **TLR4/NF-**. (2018) **14**. DOI: 10.1177/1744806918807050
43. Wu Y., Wang Y., Wang J.. **TLR4 mediates upregulation and sensitization of TRPV1 in primary afferent neurons in 2,4,6-trinitrobenzene sulfate-induced colitis**. (2019) **15**. DOI: 10.1177/1744806919830018
44. Sun W., Zhang L., Li R.. **Overexpression of miR-206 ameliorates chronic constriction injury-induced neuropathic pain in rats via the MEK/ERK pathway by targeting brain-derived neurotrophic factor**. (2017) **646** 68-74. DOI: 10.1016/j.neulet.2016.12.047
45. Singh A. K., Vinayak M.. **Activation of ERK signalling by Src family kinases (SFKs) in DRG neurons contributes to hydrogen peroxide (H**. (2017) **51** 838-850. DOI: 10.1080/10715762.2017.1382691
46. Yu S. J., Grider J. R., Gulick M. A., Xia C. M., Shen S., Qiao L. Y.. **Up-regulation of brain-derived neurotrophic factor is regulated by extracellular signal-regulated protein kinase 5 and by nerve growth factor retrograde signaling in colonic afferent neurons in colitis**. (2012) **238** 209-217. DOI: 10.1016/j.expneurol.2012.08.007
|
---
title: Dietary Level of the Omega-3 Fatty Acids EPA and DHA Influence the Flesh Pigmentation
in Atlantic Salmon
authors:
- T. Ytrestøyl
- M. Bou
- C. Dimitriou
- G. M. Berge
- T.-K. Østbye
- B. Ruyter
journal: Aquaculture Nutrition
year: 2023
pmcid: PMC9998164
doi: 10.1155/2023/5528942
license: CC BY 4.0
---
# Dietary Level of the Omega-3 Fatty Acids EPA and DHA Influence the Flesh Pigmentation in Atlantic Salmon
## Abstract
Atlantic salmon with a start weight of 53 g were fed diets with different levels of EPA and DHA or a diet with 1: 1 EPA+DHA ($0\%$, $1.0\%$, and $2.0\%$ of the diet). At 400 g, all fish groups were mixed and equally distributed in new tanks and fed three diets with $0.2\%$, $1.0\%$, or $1.7\%$ of EPA+DHA. At 1200 g, the fish were transferred to seawater pens where they were fed the same three diets until they reached a slaughter size of 3.5 kg. The fillet concentration of astaxanthin and its metabolite idoxanthin was analysed before transfer to seawater pens at 1200 g and at slaughter. The fatty acid composition in the fillet was also analysed at the same time points. Salmon fed low levels of EPA and DHA had lower fillet astaxanthin concentration and higher metabolic conversion of astaxanthin to idoxanthin compared to salmon fed higher dietary levels of EPA and/or DHA. DHA had a more positive effect on fillet astaxanthin concentrations than EPA. There were positive correlations between fillet DHA, EPA, sum N-3 fatty acids, and fillet astaxanthin concentration. A negative correlation was found between the concentration of N-6 fatty acids in the fillet and the astaxanthin concentration.
## 1. Introduction
Feeds for Norwegian farmed salmon have gone through major changes in composition in the last decades, from essentially a marine based diet in the early 1990s to a diet with $70\%$ plant ingredients today [1, 2]. One consequence of this shift in ingredients is reduced levels of the healthy N-3 fatty acids eicosapentaenoic acid (EPA) and docosahexaenoic acid (DHA) in the diet. Deficiency of these important fatty acids may have serious consequences for the salmon, and requirements for N-3 polyunsaturated fatty acids (PUFA) have been reported to range from 0.5 to $2.7\%$ of the diet, depending on the species, life stage, and experimental conditions ([3–9]). However, it is important that studies on requirements not only assess fish growth and survival but also other important criteria such as fish health and fillet quality. The pink flesh color in salmon is important for consumer acceptance. The color intensity is determined by the deposition of keto-carotenoids such as astaxanthin in the muscle, but the basal biological mechanisms that regulate the deposition of astaxanthin in the salmon muscle have not been studied in detail. In particular, the effects of interactions between astaxanthin and other dietary components such as N-3 fatty acids, vitamin A, and other antioxidants on pigmentation are needed. Most of the existing studies are done on diet formulations with a higher content of marine ingredients than what is used in commercial farming today. Some studies report no significant effects of replacing fish oil with various plant oils on astaxanthin deposition in the muscle [10–14]. Other studies have found a positive effect of increasing the levels of N-3 PUFA in feed for Atlantic salmon on sensory-evaluated color tone [15] and fillet astaxanthin concentration [16]. Long-term studies with Atlantic salmon have shown a negative effect on flesh astaxanthin concentrations of feeding diets with low levels of N-3 fatty acids [17]. Thus, high PUFA oils may have a positive effect on astaxanthin utilization. The mechanisms are not known, and most of the studies mentioned above only measured flesh color and not digestibility or metabolism of astaxanthin, both of which may affect fillet astaxanthin deposition.
The muscle retention of astaxanthin varies with season and life stage but is typically between 5–$10\%$ whereas the digestibility of astaxanthin is often around $40\%$ [16, 18, 19]. Thus, most of the digested astaxanthin is metabolized and/or excreted, possibly through several different pathways. Carotenoids are potent antioxidants due to their chemical structures with a long carbon chain of unconjugated double bounds quenching singlet oxygen and free radicals. Carotenoids like astaxanthin are also metabolized into diverse metabolites and cleavage products [20] that may have important biological functions of which many are linked to lipid metabolism [21]. Astaxanthin is a precursor of vitamin A in salmon and other fish [22–24], and a key gene in the regulation of this pathway in mammals is β,β-carotene-15,15′-monooxygenase (BCO1) [25, 26]. The expression of this enzyme is regulated by several dietary factors in mammals, including the content of EPA and DHA, N-6 fatty acids, vitamin A, and carotenoid content [27, 28]. Salmonids also have a reductive metabolism of ketocarotenoids [24, 29, 30]. Idoxanthin (3,3′,4′-trihydroxy-β,β-carotene-4-one) is the first metabolite in this reductive pathway of astaxanthin, and the concentration of idoxanthin and other reductive metabolites are influenced by life stage, genotype, and environmental conditions [16, 19, 30–32].
In the present study, diets with different concentrations of EPA and DHA or combinations of EPA and DHA were fed to Atlantic salmon from 53 g up to 3.5 kg, and the effects on fillet concentration of astaxanthin and its metabolite idoxanthin were measured. The fatty acid composition of the fillet was also measured at 1.2 and 3.5 kg and correlations were made between fillet astaxanthin, idoxanthin, and fatty acid compositions. A transcriptome analysis of the intestine was performed at 3.5 kg of genes related to oxidative stress and inflammation and carotenoid metabolism.
## 2.1. Experimental Design
Individually tagged (PIT-tags, passive integrated transponder; Biosonic) Atlantic salmon (S. Salar) with a mean initial weight of 52.8 g and adapted to seawater were fed two dietary levels of EPA, DHA, or a 1: 1 mixture of EPA and DHA ($1.0\%$ and $2.0\%$ of the diet in each dietary group) in duplicate tanks. One group was fed a diet without EPA and DHA ($0\%$, control). This resulted in a total of 7 diets, with the same content of protein (46.6–$47.0\%$), fat (24.6–$25.9\%$), energy (22.1–22.6 MJ/kg), and astaxanthin (50 mg/kg) that the fish were fed until they were 400 g (period 1). The basal test diet from 52.8 to 400 g was without fish meal and fish oil but carefully formulated to meet the nutritional requirements for amino acids. EPA and DHA were added in the form of concentrates. The formulation and chemical composition of the diets and a detailed description of the experimental conditions are given in Bou et al., [ 3]. The fish were kept under continuous light (L:D 24: 0) in indoor seawater tanks (1 m3) with flow-through seawater (33 g L−1 salinity). Temperature and oxygen were measured daily. The temperature varied between 6.3°C and 13.8°C (mean temperature 10.0°C), and the oxygen saturation level was over $85\%$. From 400 g, fish from all diet groups were evenly distributed into 9 tanks (3m3) with flow-through seawater on land (period 2A) and fed experimental diets containing three different levels of EPA + DHA (0.2, 1.0, and $1.7\%$) in triplicate until they reached approximately 1200 g. Feeds were produced as 4-, 5-, and 7-mm pellets according to fish size. The diets in the seawater period 2A in tanks on land were different in their formulation. The $0.2\%$ EPA + DHA diet was fishmeal- and fish oil-free, and the main oil source was a mixture of poultry, rapeseed, and linseed oil (50: 30: 20; v:v:v), and the main protein source was poultry meal. The $1.0\%$ EPA + DHA diet was also fishmeal-free, but the main source of EPA and DHA was fish oil. The composition of the $1.7\%$ EPA+DHA diet resembled a commercial diet with respect to the content of fishmeal and fish oil. The diets contained around $32.8\%$ fat, $36.5\%$ protein, and 50 mg astaxanthin per kg. All diets were produced by Biomar. The feeding trials in period 1 from 50 to 400 g and period 2A from 400 to 1200 g were conducted at the Nofima Research Station in Sunndalsøra, Norway. When the fish were 1200 g, they were transported by truck to sea cages at LetSea Aquaculture Research Station in Dønna, Norway (period 2B, pens in seawater). The fish were fed diets with the same levels of EPA and DHA ($0.2\%$, $1.0\%$, and $1.7\%$) as in tanks on land until they reached a weight of 3.5 kg. Diet concentrations of fatty acids in diets fed until 400 g (period 1), sea water period in tanks on land (period 2A), and sea cage period 2B are given in Table 1. A detailed description of diet formulation and fatty acid composition is found in Bou et al., [ 4]. During period 2B in sea cages, the fish were treated for sea lice 4 times (25 April, 10 July, 20 August, and 6 November) by providing chemical bath treatment with azamethiphos (0.2 mL/m3 for 35 min; Trident Vet, Neptune Pharma) and deltamethrin (0.3 mL/m3 for 30 min; ALPHA MAX, PHARMAQ). Further details on feeds and experimental conditions in seawater are found in Bou el al., [ 4].
## 2.2. Sampling and Analysis of Carotenoids
The effects of dietary EPA and DHA concentrations on flesh pigmentation were evaluated when the fish were 1200 g and 3.5 kg. At 1.2 and 3.5 kg, the fillet (right side) was sampled from 30 fish per diet treatment and stored on −80°C for later analysis of carotenoid and fatty acid composition as described below and in Bou et al., [ 3]. Fish were killed by an overdose of the anesthetic metacain (MS-222; 0.08 g/l), and individual weights and lengths were recorded. The skin and bone of the fillet were removed before the fillet was homogenized, and the carotenoids extracted using a 1: 1: 3 mixture of distilled water, methanol (containing 500 ppm butylated hydroxytoluene), and chloroform as described by Bjerkeng et al., [ 33]. A Spherisorb S5CN-4800 nitrile column (Hichrom Ltd., Theale, Berkshire, UK) was used to determine the amount of astaxanthin and idoxanthin in the samples using a mobile phase with $20\%$ (v/v) acetone in n-hexane+ (HPLC system I). An external standard of all-E-astaxanthin (Hoffmann-La Roche, Basel, Switzerland) with known concentration was prepared to establish a response line, and sample concentrations were calculated using peak areas from the chromatograms. The concentration of astaxanthin standards were determined spectrophotometrically in n-hexane containing $4.5\%$ (v/v) CHCl3. Standards of idoxanthin 3′, 4′-cis and trans glycolic isomers were prepared according to Aas et al. [ 29]. The retention times (RT) for the 3′, 4′-cis and 3′, 4′-trans glycolic isomers of idoxanthin were ca. 7.6 and 9.3 min, respectively.
## 2.3. Fillet Colour Measurements
In instrumental tristimulus colour analysis (CIE L∗a∗b∗), CIE [1986] was performed on the fillets when the fish was 1200 g (30 fillets per diet treatment). The measurements were done on the dorsal muscle posterior to the dorsal fin, under the adipose fin, and in the tail on each fillet, using a Minolta Chroma Meter CR-300 (Minolta, Osaka, Japan). Measurements were made directly on the fillets, and the measuring head was rotated 90o between duplicate measurements per position, and means of six recordings per fish were used for data analysis. Visual color assessment using a Roche SalmoFan (Hoffmann-La Roche, Basel, Switzerland) was done on of the same fillets at the same positions. The color was evaluated using a scale between 20 and 34 (20 represents a low degree of pigmentation and 34 represents a highly pigmented fillet).
## 2.4. RNA Isolation, cDNA Synthesis, and Quantification of Transcript Levels by Real-Time Quantitative PCR
Total RNA was isolated from the intestine at the final sampling (3.5 kg) using a PureLink Pro 96 RNA Purification Kit (Invitrogen), according to the manufacturer's instructions. RNA was treated with PureLink DNaseI (Invitrogen) to remove any contaminating DNA. RNA concentration was measured using a NanoDrop® ND-1000 spectrophotometer (NanoDrop Technologies). Reverse transcription of 1 μg total RNA into complementary DNA (cDNA) was carried out using a TaqMan® Reverse Transcription Reagents kit (Applied Biosystems) according to the manufacturer's protocol in a 20-μl reaction volume.
PCR primers (Table 2) were designed using Vector NTI (Invitrogen) and synthesized by Invitrogen. The efficiency was checked from 10-fold serial dilutions of cDNA for each primer pair. Real-time PCR was performed in a LightCycler 480 Instrument (Roche Applied Science). The PCR master mix consisted of 0.5 μl forward and 0.5 μl reverse primer (0.5 μM final concentrations), 4 μl of a 1: 10 dilution of cDNA and 5 μl LightCycler 480 SYBR® Green I Master (Roche Applied Science). All samples were analysed in duplicate with a non-template control for each gene. The reaction was performed by incubating the samples at 95°C for 5 min, forty-five cycles of 95°C for 15 s and 60°C for 15 s, and 72°C for 15 s for denaturation, annealing, and extension, respectively. The specificity of PCR amplification was confirmed by melting curve analysis (95°C for 5 s and 65°C for 1 min, and a continuous temperature ramp (0.11°C/s) from 65 to 97°C). Eukaryotic translation initiation factor 3 (etif3), RNA polymerase 2 (rpol2), and elongation factor 1-alpha (ef1a) were evaluated as reference genes, and it was found that the latter was the most stable. Relative expression levels of mRNA transcripts were calculated using the -ΔΔCt method using ef1a as the reference gene [34].
## 2.5. Statistical Analysis
Diet effects on fillet astaxanthin and idoxanthin concentration and color were analysed by ANOVA in SAS jmp with diet in period 1 (50–400 g) and period 2 (400–3500 g) as fixed factors. P values <0.05 were considered significant differences and $P \leq 0.1$ were considered a trend. Linear correlations were made between fillet content of astaxanthin, idoxanthin, content of EPA, DHA, sum N-3, N-6, and saturated (N-0) fatty acids. Correlation analyses were done in SAS jmp. Principal component analysis (PCA) was performed using muscle concentrations of astaxanthin, idoxanthin, and fatty acid composition (content of EPA, DHA, sum N-0, N-3, N-6, and N-9) as input variables using the software Unscrambler® X, version 10.3 (CAMO).
## 3.1. Fatty Acid Composition of the Fillet
The fatty acid composition of the fillets at 1.2 and 3.5 kg are shown in Tables 3 and 4, respectively. The fatty acid composition of the fillet reflected the dietary fatty acid composition and showed a 2.2-fold increase in percentage of EPA, a 1.7-fold increase in percentage of DHA and a 1.6-fold decrease in percentage of N-6 fatty acids as EPA + DHA in the diets increased from $0.2\%$ to $1.7\%$ in the diets. When looking into prediet effects in these two sizes of fish, fish fed increasing dietary levels of EPA and DHA in period 1 had significantly higher muscle levels of these fatty acids when they reached 1.2 kg (Table S1), although they had been fed the same diets since they were 400 g. The effects of the prediets were still significantly present at the final sampling, when the fish reached 3.5 kg, with fish fed higher content of EPA+DHA during the early life stage showing higher levels of these FA regardless of the main diet received during period 2 (Table S2).
## 3.2. Carotenoid Concentrations and Fillet Color
When the salmon was 1200 g, the major effects on fillet astaxanthin and idoxanthin concentrations were due to the diet fed during period 2B (400–1200 g) ($P \leq 0.0001$, Figure 1). However, there were also some significant effects of the diet used in period 1 (50–400 g) on muscle carotenoid concentration ($P \leq 0.05$). Fish that had been fed a diet without EPA and DHA from 50 to 400 g ($0\%$) had lower astaxanthin concentration in the fillet, compared to fish fed diets containing EPA and DHA (Figure 1(a)). Fish fed diets with DHA in period 1 had the highest astaxanthin concentrations. There was no significant effect of dietary content of EPA and DHA in period 1 on fillet idoxanthin concentration at 1200 g in fish fed $1\%$ EPA+DHA or $1.7\%$ EPA+DHA in period 2B (Figure 1(b)). In salmon fed diets with low EPA and DHA concentration in period 2 there was a tendency for higher idoxanthin concentrations in fillets of salmon fed low levels of EPA and DHA in period 1 ($$P \leq 0.09$$, Figure 1(b)). However, the major effect on fillet astaxanthin concentration at 1200 g was the diet concentration of EPA and DHA in period 2B (400–1200 g). Salmon fed the diet with $0.2\%$ EPA+DHA (low) had the lowest fillet astaxanthin concentration with a mean of 1.38 mg/kg when all diet groups from period 1 are pooled (Figure 2(a)). Salmon fed the $1.0\%$ and the $1.7\%$ diets in period 2A had 1.79 and 2.75 mg astaxanthin per kg, respectively, when all diets in period 1 were pooled. The differences in fillet astaxanthin were also reflected in fillet visual color measured by SalmoFan scores and Minolta chroma, its redness and yellowness (Figures 2(b) and 2(c)). Fillets from salmon fed diets containing $0.2\%$ EPA and DHA from 400 to 1200 g were less colored than fillets from fish fed the diets containing $1.0\%$ or $1.7\%$ EPA and DHA ($P \leq 0.0001$). However, the diet provided to the fish during period 1 (50–400 g) did not have a significant effect on fillet color of 1200 g salmon measured as Salmofan score, chroma, redness, and yellowness assessed by Minolta. There was a significant effect of diet in period 2A on the concentration of idoxanthin in the fillet ($P \leq 0.0001$). Salmon fed the commercial control diet ($1.7\%$ EPA+DHA) had lower idoxanthin concentration compared to salmon fed the 0.2 and $1.0\%$ diets (Figure 2(a)). Idoxanthin amounted $14\%$ of the total carotenoid content in fillet of salmon fed the $1.7\%$ control diet (all diets in period 1 pooled) whereas the idoxanthin concentration was around $30\%$ of total carotenoids in the 0.2 and $1.0\%$ diets (pooled mean for all diets period 1). There were significant negative correlations between fillet concentrations of EPA, DHA, sum N-9 fatty acids, and % idoxanthin of total carotenoids ($P \leq 0.0001$, R2 = 0.27, 0.25 and 0.25, respectively).
Correlations between fillet astaxanthin and fatty acid composition showed a positive correlation between fillet DHA and EPA concentration and astaxanthin concentration ($P \leq 0.0001$, R2 = 0.43 and 0.46, respectively) and a negative correlation with N-6 fatty acids ($P \leq 0.01$, R2 = 0.13). PCA analysis showed no clear pattern related to diet in period 1. The diet in period 2 was responsible for the differences observed (Figure 3).
There was no significant effect of diet in period 1 on fish bodyweight at 400 g. There was however an effect of diet in period 2A on the bodyweight; the fish fed the commercial control diet with a $1.7\%$ of EPA+DHA were 1.3 kg; the fish fed with a$1\%$ of EPA+DHA diet were 1.2 kg; and the fish fed with a low EPA + DHA diet were 1.1 kg when the fish were transferred to pens in seawater ($P \leq 0.0001$). Linear regressions between fish weight and fillet astaxanthin and idoxanthin at 1200 g were made to test whether fish weight had an effect on fillet color. For the 0.2 and $1.0\%$ diet, there was no significant relationship between fish weight and fillet color, astaxanthin, or idoxanthin concentration. For the commercial control diet ($1.7\%$ EPA+DHA) there was a significant positive correlation between bodyweight and fillet astaxanthin, chroma, redness and yellowness ($P \leq 0.01$), with bodyweight explaining $9\%$ of the total variation in fillet colour parameters and $12\%$ of the variation in fillet astaxanthin concentration.
There were no significant differences in bodyweight between diets at the final sampling when the salmon weighed around 3.5 kg. There was however a clear positive effect of a higher dietary concentration of omega-3 fatty acids on fillet astaxanthin concentration. The effect of diet in period 1 on fillet astaxanthin or idoxanthin concentrations was no longer significant. Data from the dietary groups in period 1 is therefore pooled in Figure 2(d). The fillet astaxanthin concentration was significantly affected by diet fatty acid composition in period 2 A and B, between 400 and 3.5 kg ($P \leq 0.001$). Salmon fed the $0.2\%$ diet had the lowest fillet astaxanthin concentration and fish fed the commercial control diet with a $1.7\%$ EPA+DHA had the highest fillet astaxanthin concentration. The concentration of idoxanthin in the fillet at 3.5 kg was also affected by diet (Figure 2(d)). As in the 1200 g fish, it was the highest in the $1\%$ diet and lowest in the commercial control diet ($P \leq 0.001$). There was a significant linear correlation between fillet content of EPA and DHA ($P \leq 0.0001$, $y = 0.81$x + 0.75, R2 = 0.64). Positive correlations between fillet concentrations of astaxanthin and content of EPA and DHA were found ($P \leq 0.0001$) explaining around $25\%$ of the variation in fillet astaxanthin. There were also positive correlations between sum of N-3 and N-0 fatty acids and fillet astaxanthin concentration ($P \leq 0.0001$, R2 = 0.26 and 0.28, respectively) whereas a weaker negative correlation with sum N-6 fatty acids in fillet was found ($P \leq 0.001$, R2 = 0.11). For idoxanthin in fillet (% of total carotenoids) weak negative correlations were found with concentration of DHA in fillet ($P \leq 0.05$, R2 = 0.06) and sum N-0 in fillet ($P \leq 0.01$, R2 = 0.09).
## 3.3. Gene Expression in Intestine
The expression of the β-carotene oxygenases bmco1, bmco1-like, and bmco2-c in the intestine was measured at the final sampling at 3.5 kg. There were no significant effects of diet in periods 1 or 2 on the expression on the three carotenoid oxygenase genes (Table 5). The diet treatments in period 1 is therefore not shown, but instead pooled within diet treatment in period 2 (Figures 4(a)–4(c)). Overall, salmon fed $1\%$ shoved the numerically the highest expression of all three genes in the intestine, but the differences were not statistically significant due to large variation between individuals. At the final sampling, expression of genes involved in cellular defense mechanisms against oxidative stress and inflammation were also measured in the intestine. The expression of cyclooxygenase 2 (cox2), a gene involved in synthesis of proinflammatory eicosanoids, was elevated in fish fed the $1\%$ diet compared to the two other diets (Figure 4(d)). There was also an effect of diet from 50 to 400 g, where increasing levels of EPA and DHA in early life stages down regulated the expression of cox2 (Figure S1). The expression of the gene coding for nuclear factor erythroid 2-related factor 2 (nrf-2) involved in regulation of genes protecting against oxidative damage was upregulated in fish fed $1.0\%$ EPA+DHA compared to in salmon fed diets with a 0.2 and $1.7\%$ of EPA+DHA (Figure 4(e)). Reduced expression of the transcription factor NF-κB was found in diets with a 0.2 and $1.0\%$ EPA + DHA compared to the salmon fed with a $1.7\%$ diet (Figure 4(f)).
## 4. Discussion
The reduction in dietary concentration of the important omega-3 fatty acids EPA and DHA in later years due to the replacement of marine ingredients with plant ingredients may have consequences for salmon health and quality. Requirements have often been assessed based on growth effects in trials of short duration under experimental conditions without stress, and many of the trials have been done on juveniles in freshwater. However, requirements cannot be based only on growth effects, but also have to consider fish health, welfare, and quality [3, 4]. Requirements for large salmon under challenging conditions in seawater are also needed to ensure fish health and the quality of the final product. Bou et al. [ 4] found that $1\%$ EPA+DHA in the Atlantic salmon diet, a level previously regarded as sufficient, was too low to maintain fish health under demanding environmental conditions in sea cages. Low EPA+DHA resulted in increased mortality during delousing at high water temperatures, increased fat content in the liver, intestine, and viscera, reduced intervertebral space, and caused mid-intestinal hyper-vacuolization. EPA and DHA in the different tissue membrane phospholipids were typically replaced by pro-inflammatory N-6 fatty acids. In the present study, negative effects of low levels of EPA and DHA on flesh pigmentation and a positive correlation between muscle concentrations of astaxanthin and N-3 fatty acids were found, whereas a negative correlation between astaxanthin concentration and muscle content of N-6 fatty acids was observed. Long chain omega-3 fatty acids may improve pigmentation through several mechanisms. Fatty acid composition affects the solubility and transfer of carotenoids into the aqueous phase of carotenoids in triglyceride emulsions [35]. Differences in the transfer efficiency of astaxanthin into intestinal micelles may explain the effects of some dietary oils observed on astaxanthin deposition. Saturated fatty acids have been shown to be negative for the digestibility of astaxanthin at low temperatures [36], but not at higher temperatures. Studies on interactions between PUFAs and temperature on astaxanthin digestibility in vivo are however, lacking in salmon, and the digestibility was not measured in the present study.
In the present study, the diet raw material composition was identical until the fish were 400 g, but from 400 g until 3.5 kg, the diets differed in the content of fish oil and fish meal [4]. The effects seen in diet in period 1 can thus be attributed to the differences in EPA and DHA concentration, but in period 2, the effects could also potentially be a result of the different fatty acid composition, in particular of the N-6 fatty acid content which was higher in the $0.2\%$ diet and lowest in the $1.7\%$ diet [4]. The fatty acid concentration of the fillet reflected that of the diet, but the fatty acid concentration at slaughter was still significantly affected by diet EPA and DHA content in Period 1 from 50 to 400 g. Although there was no longer a significant effect of diet in period 1 on fillet astaxanthin at slaughter, a stronger positive correlation between fillet astaxanthin and EPA and DHA content was found at slaughter than the negative correlation found between astaxanthin and N-6 fatty acids in the fillet at slaughter. Some other studies support the findings in the present study. In a study on Atlantic salmon, Waagbø et al. [ 17] replaced $64\%$ of the dietary fish oil with vegetable oil and reduced the fishmeal content in a full-scale seawater production. Reduced growth and feed efficiency were observed with decreasing fishmeal inclusion levels. When the low marine diets were boosted with a South American omega-3 fish oil three months prior to slaughter, fillet color and astaxanthin content improved significantly. The average astaxanthin concentration was 5.0 mg kg−1 in Norwegian Quality Cut (NQC) samples from all fish groups taken after one year of feeding, with no difference between the dietary groups. After an additional 5 months, the average astaxanthin content in NQC samples was still 5.5 mg kg−1 for fish groups fed the control low-marine diets, but 6.2 mg kg−1 in the fish groups fed the omega-3- boosting diets. However, there were no significant differences between diets in visual color measured by SalmoFan and Minolta. Some earlier studies have also shown positive effects of high-PUFA oils on flesh pigmentation. Atlantic salmon fed diets with high Peruvian PUFA oil deposited $13\%$ more carotenoids in the fillet than fish fed diets supplemented with herring oil, and a positive linear relationship was found between final fillet idoxanthin concentration and total saturated fatty acids in supplement oils [16]. Fillets of Atlantic salmon fed diets containing $29\%$ Peruvian, high-PUFA fish oil contained $41\%$ more carotenoids (canthaxanthin and astaxanthin combined) than fillets of salmon fed a diet with $29\%$ soybean oil ([37, 38]). Since most of the studies that have examined the effect of diet fatty acid composition on fillet color have replaced fish oils or fish meal with different plant oils and protein sources, it is not possible to say if the effect on pigmentation is due to a higher content of PUFA in the diet or a lower content of N-6 fatty acids.
The N-6 fatty acids are known to promote the formation of pro-inflammatory and pro-aggregatory eicosanoids, whereas the N-3 fatty acids have the opposite effects. Carotenoids such as astaxanthin are potent antioxidants due to their chemical structure with a long carbon chain of unconjugated double bounds quenching singlet oxygen and free radicals [21, 39]. Higher oxidative stress levels in the salmon could thus result in more astaxanthin being oxidized in the salmon and lead to reduced astaxanthin levels in the muscle. Low EPA+DHA levels also increased fat deposition in the liver, intestine, and viscera and caused mid-intestinal hyper-vacuolization in the 0.2 and $1\%$ EPA+DHA groups [4]. There were also indications of a higher oxidative stress level in the intestine of salmon fed the $1.0\%$ diet. The transcription factor nrf2 (NF-E2-related factor 2) is a potent transcriptional activator and plays a central role in inducible expression of many cytoprotective genes in response to oxidative stress [40, 41]. The expression of nrf-2 in the intestine was higher in fish fed $1.0\%$ EPA+DHA than in fish fed $0.2\%$ EPA+DHA, while in fish fed, the control diet had intermediate expression. Salmon fed the diet with a $1\%$ of EPA+DHA also had the highest expression of cyclooxygenase 2 (cox2), a gene involved in the synthesis of proinflammatory eicosanoids. The nuclear factor-kappa B (NF-κB) is a transcription factor that plays a key role in the control of genes involved in inflammation, cell proliferation, and apoptosis and is activated in response to inflammatory stimuli and environmental stressors [42]. Reduced expression of nf-κb was found in the intestine of salmon fed diets with 0.2 and $1.0\%$ EPA+DHA compared to the control diet with $1.7\%$ EPA and DHA which could suggest less response in these groups. However, the regulation of expression nf-κb in mammals is shown to be quite complex and also mediated through feed-back regulation [43]. Little information is available on the regulation of this pathway in salmon, but a negative feedback mechanism could explain lower expression of this gene in salmon fed diets with lower levels of EPA and DHA.
Carotenoids are transformed into diverse metabolites and cleavage products that may have important biological functions including regulation of expression of genes involved in cell metabolism and defense against oxidative damage [20, 25]. In mammals, carotenoids have been shown to lower blood concentrations of inflammation markers and pro-inflammatory cytokines, improve insulin sensitivity, and reduce obesity [44, 45]. Many of these biological effects of carotenoids in mammals are linked to the activity of carotenoid oxygenases that cleaves β-carotene as the first step in retinol synthesis [26]. Two distinct forms of this enzyme are found in mammals: beta-carotene 15,15′-monooxygenase (BCMO1), and beta-carotene-9,10-dioxygenase 2 (BCDO2). The Atlantic salmon beta-carotene oxygenase gene family contains 5 members, three bco2, and two bcmo1 paralogs (bcmo1 and bcmo1 like) that have tissue-specific expression, with the highest expression in the liver and intestine and lowest in the muscle [46]. The regulation of the carotenoid oxygenases has been studied in mammals, but little is so far known for fish. The presence of a PPARγ response element in the upstream of BCMO1 promoter, may explain the connection between carotenoid and lipid metabolism in mammals [47], but if this is also the case in salmon is not known. Dietary factors, including type and amount of dietary fat have been shown to influence the activity of BCMO1 in mammals [28]. Unsaturated N-3 fatty acids were shown to enhance β,β-carotene 15,15′-dioxygenase activity in rat intestine [27, 48]. However, in the present study, there were not found any clear effects of dietary content of N-3 fatty acids on gene expression of carotenoid oxidases in the intestine. Overall, the $1\%$ diet shoved the highest expression of all three genes, but the differences were not statistically significant due to the large variation between individuals.
## 5. Conclusions
Low levels of EPA and DHA in the diet were negative for salmon flesh pigmentation, both in a tank environment and in seawater pens. The fillet concentration in all treatments was quite low for a salmon of 3.5 kg. The fish in this trial had several delousing operations during the seawater period that led to elevated mortality when the water temperature was high. There is currently little information on how these essential nutrients interact and how stress may affect the dietary requirements necessary for optimal health and quality of the salmon. Handling in connection with delousing have become more frequent in commercial salmon farming in recent years, and it has also become more challenging to obtain the sufficient flesh color. Although not providing a casual explanation, the present study shows a connection between low dietary EPA and DHA levels and reduced flesh color in salmon.
## Data Availability
The data used to support the findings of this study are included within the article.
## Ethical Approval
The feeding trial followed the Norwegian guidelines for research animals and was approved by the Norwegian Food Safety Authority.
## Conflicts of Interest
There are no conflicts of interests for this manuscript.
## Authors' Contributions
The design of the trial was performed by BR and GMB. Writing and statistical analysis of data were performed by TY, MB, and BR. Analysis of gene expression was performed by TKØ and MB. Measurements of visual flesh color were worked by CD. Analysis of fatty acids was in charged by MB and CD. All the authors have read and approved the final version of the manuscript.
## References
1. Aas T. S., Åsgård T., Ytrestøyl T.. **Utilization of feed resources in the production of rainbow trout (**. (2022.0) **26, article 101317**. DOI: 10.1016/j.aqrep.2022.101317
2. Ytrestøyl T., Aas T. S., Åsgård T.. **Utilisation of feed resources in production of Atlantic salmon (**. (2015.0) **448** 365-374. DOI: 10.1016/j.aquaculture.2015.06.023
3. Bou M., Berge G. M., Baeverfjord G.. **Requirements of n-3 very long-chain PUFA in Atlantic salmon (**. (2017.0) **117** 30-47. DOI: 10.1017/S0007114516004396
4. Bou M., Berge G. M., Baeverfjord G., Sigholt T., Østbye T.-K., Ruyter B.. **Low levels of very-long-chain n-3 PUFA in Atlantic salmon (**. (2017.0) **6, article e32**. DOI: 10.1017/jns.2017.28
5. Glencross B. D.. **Exploring the nutritional demand for essential fatty acids by aquaculture species**. (2009.0) **1** 71-124. DOI: 10.1111/j.1753-5131.2009.01006.x
6. Lutfi E., Berge G. M., Bæverfjord G.. **Increasing dietary levels of the omega-3 long-chain polyunsaturated fatty acids, EPA and DHA, improves the growth, welfare, robustness, and fillet quality of Atlantic salmon in sea cages**. (2023.0) **129** 10-28. DOI: 10.1017/S0007114522000642
7. Qian C., Hart B., Colombo S. M.. **Re-evaluating the dietary requirement of EPA and DHA for Atlantic salmon in freshwater**. (2020.0) **518, article 734870**. DOI: 10.1016/j.aquaculture.2019.734870
8. Rosenlund G., Torstensen B., Stubhaug I., Usman N., Sissener N.. **Atlantic salmon require long-chain**. (2016.0) **5** p. e19. DOI: 10.1017/jns.2016.10
9. Ruyter B., Røsjø C., Einen O., Thomassen M.. **Essential fatty acids in Atlantic salmon: time course of changes in fatty acid composition of liver, blood and carcass induced by a diet deficient in n-3 and n-6 fatty acids**. (2000.0) **6** 109-117. DOI: 10.1046/j.1365-2095.2000.00136.x
10. Bell J. G., McEvoy J., Tocher D. R., McGhee F., Campbell P. J., Sargent J. R.. **Replacement of fish oil with rapeseed oil in diets of Atlantic salmon (**. (2001.0) **131** 1535-1543. DOI: 10.1093/jn/131.5.1535
11. Bell J. G., Henderson R. J., Tocher D. R.. **Substituting fish oil with crude palm oil in the diet of Atlantic salmon (**. (2002.0) **132** 222-230. DOI: 10.1093/jn/132.2.222
12. Bell J. G., Tocher D. R., Henderson R. J., Dick J. R., Crampton V. O.. **Altered fatty acid compositions in Atlantic salmon (**. (2003.0) **133** 2793-2801. DOI: 10.1093/jn/133.9.2793
13. Rosenlund G., Obach A., Sandberg M. G., Standal H., Tveit K.. **Effect of alternative lipid sources on long-term growth performance and quality of Atlantic salmon (**. (2001.0) **32** 323-328. DOI: 10.1046/j.1355-557x.2001.00025.x
14. Torstensen B. E., Bell J. G., Rosenlund G.. **Tailoring of Atlantic salmon (**. (2005.0) **53** 10166-10178. DOI: 10.1021/jf051308i
15. Waagbø R., Sandnes K., Torrissen O. J., Sandvin A., Lie Ø.. **Chemical and sensory evaluation of fillets from Atlantic salmon (**. (1993.0) **46** 361-366. DOI: 10.1016/0308-8146(93)90005-Z
16. Bjerkeng B., Hatlen B., Wathne E.. **Deposition of astaxanthin in fillets of Atlantic salmon (**. (1999.0) **180** 307-319. DOI: 10.1016/S0044-8486(99)00206-9
17. Waagbø R., Berntssen M. H. G., Danielsen T.. **Feeding Atlantic salmon diets with plant ingredients during the seawater phase – a full-scale net production of marine protein with focus on biological performance, welfare, product quality and safety**. (2013.0) **19** 598-618. DOI: 10.1111/anu.12010
18. Wathne E., Bjerkeng B., Storebakken T., Vassvik V., Odland A. B.. **Pigmentation of Atlantic salmon (**. (1998.0) **159** 217-231. DOI: 10.1016/S0044-8486(97)00218-4
19. Ytrestøyl T., Struksnæs G., Rørvik K.-A., Bjerkeng B.. **Utilisation of astaxanthin in Atlantic salmon from seawater transfer to slaughter**. p. 1
20. Amengual J., Lobo G. P., Golczak M.. **A mitochondrial enzyme degrades carotenoids and protects against oxidative stress**. (2011.0) **25** 948-959. DOI: 10.1096/fj.10-173906
21. Fakhri S., Abbaszadeh F., Dargahi L., Jorjani M.. **Astaxanthin: a mechanistic review on its biological activities and health benefits**. (2018.0) **136** 1-20. DOI: 10.1016/j.phrs.2018.08.012
22. Guillou A., Choubert G., Storebakken T., de la Noüet J., Kaushik S.. **Bioconversion pathway of astaxanthin into retinol**. (1989.0) **94** 481-485. DOI: 10.1016/0305-0491(89)90185-5
23. Moren M., Naess T., Hamre K.. **Conversion of**. (2002.0) **27** 71-80. DOI: 10.1023/B:FISH.0000021819.46235.12
24. Schiedt K., Leuenberger F. J., Vecchi M., Glinz E.. **Absorption, retention, and metabolic transformations of carotenoids in rainbow trout, salmon and chicken**. (1985.0) **57** 685-692. DOI: 10.1351/pac198557050685
25. Lobo G. B., Amengual J., Li H. N. M.. (2010.0) **285** 27891-27899. DOI: 10.1074/jbc.M110.132571
26. Lobo G. P., Amengual J., Palczewski G., Babino D., von Lintig J.. **Mammalian carotenoid-oxygenases: key players for carotenoid function and homeostasis**. (2012.0) **1821** 78-87. DOI: 10.1016/j.bbalip.2011.04.010
27. During A., Nagao A., Terao J.. (1998.0) **128** 1614-1619. DOI: 10.1093/jn/128.10.1614
28. Lietz G., Lange J., Rimbach G.. **Molecular and dietary regulation of**. (2010.0) **502** 8-16. DOI: 10.1016/j.abb.2010.06.032
29. Aas G. H., Bjerkeng B., Hatlen B., Storebakken T.. **Idoxanthin, a major carotenoid in the flesh of Arctic charr (**. (1997.0) **150** 135-142. DOI: 10.1016/S0044-8486(96)01458-5
30. Schiedt K., Foss P., Storebakken T., Liaaen-Jensen S.. **Metabolism of carotenoids in salmonids--I. idoxanthin, a metabolite of astaxanthin in the flesh of atlantic salmon (**. (1989.0) **92** 277-281. DOI: 10.1016/0305-0491(89)90278-2
31. Bjerkeng B., Hatlen B., Jobling M.. **Astaxanthin and its metabolites idoxanthin and crustaxanthin in flesh, skin, and gonads of sexually immature and maturing Arctic charr (**. (2000.0) **125** 395-404. DOI: 10.1016/S0305-0491(99)00183-2
32. Ytrestøyl T., Coral-Hinostroza G. N., Hatlen B., Robb D. H. F., Bjerkeng B.. **Carotenoid and lipid content in muscle of Atlantic salmon,**. (2004.0) **138** 29-40. DOI: 10.1016/j.cbpc.2004.01.011
33. Bjerkeng B., Refstie S., Fjalestad K. T., Storebakken T., Rødbotten M., Roem A.. **Quality parameters of the flesh of Atlantic salmon (**. (1997.0) **157** 297-309. DOI: 10.1016/S0044-8486(97)00162-2
34. Livak K. J., Schmittgen T. D.. **Analysis of relative gene expression data using real-time quantitative PCR and the 2**. (2001.0) **25** 402-408. DOI: 10.1006/meth.2001.1262
35. Borel P., Grolier P., Armand M.. **Carotenoids in biological emulsions: solubility, surface-to-core distribution, and release from lipid droplets**. (1996.0) **37** 250-261. DOI: 10.1016/S0022-2275(20)37613-6
36. Sigholt T., Berge G. M., Ruyter B., Åsgård T.. **Graded levels of palm oil in diets for Atlantic salmon (**. p. 1
37. Regost C., Jakobsen J. V., Rørå A. M. B.. **Flesh quality of raw and smoked fillets of Atlantic salmon as influenced by dietary oil sources and frozen storage**. (2004.0) **37** 259-271. DOI: 10.1016/j.foodres.2003.12.003
38. Rørå A. M. B., Birkeland S., Hultmann L., Rustad T., Skåra T., Bjerkeng B.. **Quality characteristics of farmed Atlantic salmon (**. (2005.0) **38** 201-211. DOI: 10.1016/j.lwt.2004.06.008
39. Elliott R.. **Mechanisms of genomic and non-genomic actions of carotenoids**. (2005.0) **1740** 147-154. DOI: 10.1016/j.bbadis.2004.12.009
40. Itoh K., Chiba T., Takahashi S.. **An Nrf2/small Maf heterodimer mediates the induction of phase II detoxifying enzyme genes through antioxidant response elements**. (1997.0) **236** 313-322. DOI: 10.1006/bbrc.1997.6943
41. Motohashi H., Yamamoto M.. **Nrf2-Keap1 defines a physiologically important stress response mechanism**. (2004.0) **10** 549-557. DOI: 10.1016/j.molmed.2004.09.003
42. De Bosscher K., Vanden Berghe W., Haegeman G.. **Cross-talk between nuclear receptors and nuclear factor**. (2006.0) **25** 6868-6886. DOI: 10.1038/sj.onc.1209935
43. Oeckinghaus A., Ghosh S.. **The NF-kappaB family of transcription factors and its regulation**. (2009.0) **1**. DOI: 10.1101/cshperspect.a000034
44. Bhuvaneswari S., Arunkumar E., Viswanathan P., Anuradha C. V.. **Astaxanthin restricts weight gain, promotes insulin sensitivity and curtails fatty liver disease in mice fed a obesity-promoting diet**. (2010.0) **45** 1406-1414. DOI: 10.1016/j.procbio.2010.05.016
45. Gammone M. A., Riccioni G., D'Orazio N.. **Carotenoids: potential allies of cardiovascular health?**. (2015.0) **59**. DOI: 10.3402/fnr.v59.26762
46. Helgeland H., Sodeland M., Zoric N.. **Genomic and functional gene studies suggest a key role of**. (2019.0) **9**. DOI: 10.1038/s41598-019-56438-3
47. Boulanger A., McLemore P., Copeland N. G.. **Identification of beta-carotene 15,15′-monooxygenase as a peroxisome proliferator-activated receptor target gene**. (2003.0) **17** 1304-1306. DOI: 10.1096/fj.02-0690fje
48. Raju M., Lakshminarayana R., Krishnakantha T. P., Baskaran V.. **Micellar oleic and eicosapentaenoic acid but not linoleic acid influences the**. (2006.0) **288** 7-15. DOI: 10.1007/s11010-005-9091-5
|
---
title: 'Women’s experiences and views on early breastfeeding during the COVID-19 pandemic
in Norway: quantitative and qualitative findings from the IMAgiNE EURO study'
authors:
- Eline Skirnisdottir Vik
- Sigrun Kongslien
- Ingvild Hersoug Nedberg
- Ilaria Mariani
- Emanuelle Pessa Valente
- Benedetta Covi
- Marzia Lazzerini
journal: International Breastfeeding Journal
year: 2023
pmcid: PMC9998246
doi: 10.1186/s13006-023-00553-5
license: CC BY 4.0
---
# Women’s experiences and views on early breastfeeding during the COVID-19 pandemic in Norway: quantitative and qualitative findings from the IMAgiNE EURO study
## Abstract
### Background
Little is known about women’s experience of care and views on early breastfeeding during the COVID-19 pandemic in Norway.
### Methods
Women ($$n = 2922$$) who gave birth in a facility in Norway between March 2020 and June 2021 were invited to answer an online questionnaire based on World Health Organization (WHO) Standard-based quality measures, exploring their experiences of care and views on early breastfeeding during the COVID-19 pandemic. To examine associations between year of birth [2020, 2021] and early breastfeeding-related factors, we estimated odds ratios (ORs) with $95\%$ confidence intervals (CIs) using multiple logistic regression. Qualitative data were analysed using Systematic Text Condensation.
### Results
Compared to the first year of the pandemic [2020], women who gave birth in 2021 reported higher odds of experiencing adequate breastfeeding support (adjOR 1.79; $95\%$ CI 1.35, 2.38), immediate attention from healthcare providers when needed (adjOR 1.89; $95\%$ CI 1.49, 2.39), clear communication from healthcare providers (adjOR 1.76; $95\%$ CI 1.39, 2.22), being allowed companion of choice (adjOR 1.47; $95\%$ CI 1.21, 1.79), adequate visiting hours for partner (adjOR 1.35; $95\%$ CI 1.09, 1.68), adequate number of healthcare providers (adjOR 1.24; $95\%$ CI 1.02, 1.52), and adequate professionalism of the healthcare providers (adjOR 1.65; $95\%$ CI 1.32, 2.08). Compared to 2020, in 2021 we found no difference in skin-to-skin contact, early breastfeeding, exclusive breastfeeding at discharge, adequate number of women per room, or women’s satisfaction. In their comments, women described understaffed postnatal wards, early discharge and highlighted the importance of breastfeeding support, and concerns about long-term consequences such as postpartum depression.
### Conclusions
In the second year of the pandemic, WHO Standard-based quality measures related to breastfeeding improved for women giving birth in Norway compared to the first year of the pandemic. Women’s general satisfaction with care during COVID-19 did however not improve significantly from 2020 to 2021. Compared to pre-pandemic data, our findings suggest an initial decrease in exclusive breastfeeding at discharge during the COVID-19 pandemic in Norway with little difference comparing 2020 versus 2021. Our findings should alert researchers, policy makers and clinicians in postnatal care services to improve future practices.
### Supplementary Information
The online version contains supplementary material available at 10.1186/s13006-023-00553-5.
## Background
The COVID-19 pandemic has affected maternal and perinatal healthcare worldwide [1, 2]. New and continuously shifting government regulations had a large impact on the organization of hospital care, including restrictions that may have affected breastfeeding [3, 4]. Both the World Health Organization (WHO) and the International Confederation of Midwives (ICM) called the attention of the international community to women’s and newborn’s rights during the COVID-19 pandemic [5, 6]. However, a reduction in exclusive breastfeeding initiation during the first wave of the pandemic was observed [3, 7]. Further, European hospitals have differed in their interpretation of guidance around breastfeeding policies during the pandemic, with some advising formula feeding although no national or international guidelines recommended discontinuation of breastfeeding [8].
Breastfed children are less likely to die of infections [9], less likely to become overweight [10] and are less prone to diabetes and allergy later in life [11], while their intelligence quotients can be higher [12, 13]. Women who breastfeed have a reduced risk of a range of physical and emotional health problems, such as ovarian and breast cancers, postpartum depression and maternal excess of stress [14]. The multiple benefits of breastfeeding likely outweigh possible risk factors related to SARS-CoV-2 infection and should therefore be encouraged [15, 16]. In addition, while SARS-CoV-2 transmission from mother to baby via breastmilk seems to be unlikely, milk produced by infected mothers has been found to be a source of anti-SARS-CoV-2 IgA and IgG valuable for the baby’s health [16].
This study is part of the IMAgiNE EURO project, a multi-country survey conducted in 18 countries in the WHO European Region to collect views of women on the quality of maternal and newborn healthcare during the COVID-19 pandemic [17]. Similar to what occurred in other European countries, in order to prevent the spread of COVID-19, since 12 March 2020 the Norwegian government introduced strict social distancing regulations with major impacts on society [18]. During the COVID-19 pandemic in Norway, an increased use of formula, a reduction in exclusive breastfeeding and reduction in the average length of breastfeeding were documented [4]. However, little is known about women’s experience of care and views on early breastfeeding during the COVID-19 pandemic in Norway. More evidence on this topic may be of interest for researchers, policy makers and clinicians and support further improvements in the quality of maternal and newborn care.
The aim of this study was to investigate quality of care at facility level, and women’s experiences and views on early breastfeeding practices during different phases of the COVID-19 pandemic in Norway.
## Study design
Details of the multi-country IMAgiNE EURO project are reported elsewhere [17, 19]. In the current mixed-method cross-sectional study, both quantitative and qualitative data were analysed to investigate women’ experiences and views on early breastfeeding during the pandemic. To strengthen the reporting of quantitative data, a STROBE (Strengthening the reporting of observational studies in epidemiology) checklist was used (see Additional File 1). Likewise, we used the COREQ Checklist (COnsolidated criteria for REporting Qualitative research Checklist) in reporting our qualitative data (see Additional File 2).
## Setting
In Norway, approximately 53,000 babies are born each year [20]. Maternity care is part of the public healthcare system and is built on the principle of free and equal access for all regardless of factors such as ethnicity or social background [21, 22]. Prior to the COVID-19 pandemic, Norway reported one of the highest rates of exclusive breastfeeding at 6 months of age in Europe [23]. Breastfeeding is promoted in a national action plan launched in 2017 for a healthier diet [24] and national surveys published in 2020 showed that $78\%$ of babies in Norway were breastfed at 6 months [25] and $48\%$ at 12 months [26]. By National law, parents in Norway are entitled to 49 weeks paid parental leave [27]. Additionally, a nursing mother returning to work is entitled to 30 minutes time off, which may be taken twice daily or as a reduction in working hours by up to 1 hour per day, to promote breastfeeding [28].
During the COVID-19 pandemic, in Norway the risk of maternal hospitalization due to COVID-19 disease in pregnancy was low [29]. However, maternity wards changed their practices, companion of choice often encountered restrictions in participation of care [30], and women were on average discharged from hospital earlier than in preceding years [20]. Further, during the pandemic women who gave birth in Norway have described feelings of loneliness and isolation in relation to antenatal care, when arriving at the hospital for labor, in cases of induction of labor, and at the postnatal ward [30]. This could result in a lack of maternal support which is crucial for the establishment of exclusive breastfeeding, especially during the hospital stay in the early postpartum period [3]. Women remained isolated from their social network after discharge due to strict social distancing regulations [18]. While several restrictions remained in healthcare services in Norway in 2021, the first year of the pandemic [2020] was characterized by more uncertainties both for new families and healthcare workers.
## Participants
Data were collected with a voluntary anonymous online survey, open to women ≥18 years-of-age who gave birth at a facility in the WHO European Region between 1 March 2020 and 30 June 2021 ($$n = 34$$,391). Data were cleaned as previously reported according to standard operating procedures [17]. Briefly, suspected duplicates and cases missing $20\%$ or more answers on the 40 key quality measures and five key socio-demographic variables (i.e., date of birth, age, education, parity, whether the women gave birth in the same country where she was born) were excluded. The current study includes responses given by women who gave birth in Norway in the same period ($$n = 2922$$). For the purposes of this study, twin or multiple births ($$n = 31$$), and infants admitted to the Neonatal Intensive Care Unit (NICU) or Special Care Baby Unit (SCBU) ($$n = 291$$) were excluded from the analysis. Further exclusion of cases can be seen in Fig. 1.Fig. 1Flowchart of the derivation of the study sample. Women who gave birth in Europe during the COVID-19 pandemic (March 2020 to July 2021) 1Percentage of missing data for each woman was calculated over mandatory questions ($$n = 45$$)
## Data collection
Data were collected with an online questionnaire (see Additional File 3). The questionnaire was developed and validated by an international team of experts [17]. The questionnaire was built on WHO standards of improving quality of maternal and newborn care (QMNC) ($$n = 30$$) [31] and also included questions relevant due to the COVID-19 pandemic ($$n = 10$$). The wording on education levels was agreed among partners during a Delphi exercise including 26 experts from 11 countries of the WHO European Region [17, 19]. The last question in the questionnaire was an open-ended question with no word limit to collect input and suggestions from mothers. The open-ended question was phrased as follows: Do you have any suggestions to improve quality of care provided at the facility where you gave birth or to improve this questionnaire?
The online survey was made available in 22 languages and actively promoted though social media (i.e., Facebook groups, such as pregnancy due date groups, migrant groups in Norway and a group for breastfeeding mothers) in Norway by researcher ESV. Project partners in the WHO European Region promoted the survey through social media, organizational websites, local networks and Non-Governmental Organizations by project partners [17]. Women participated in the study in their preferred language regardless of which country they gave birth in.
Data were collected through an online survey using REDCap 8.5.21 (© 2021 Vanderbilt University).
## Data analysis
Both quantitative and qualitative data were analysed. In the current study, the analysis of quantitative data included 13 out of the original 40 quality measures in the full survey. All variables ($$n = 3$$) explicitly describing breastfeeding and infant feeding were based on WHO’s standards for improving QMNC in health facilities [31]. In addition, qualitative data were extracted from the final open-ended question collecting women’s suggestions on how to improve quality of care. Both quantitative and qualitative variables are described in the following text.
## Analysis of quantitative data
We used descriptive statistics to summarize quantitative data, reported as frequencies and percentages. Differences of sample characteristics by year of birth [2020, 2021] were tested with a Chi-square test or a Fisher exact test. To examine associations between year of birth and early breastfeeding-related factors ($$n = 13$$), we estimated crude and adjusted odds ratios (adjORs) with $95\%$ confidence intervals (CIs) using logistic regression. The main exposure was year of birth [2020, 2021]. Outcome variables were dichotomized variables related to early breastfeeding and included the following: opportunity to have skin-to-skin within the first hour after birth (yes, no); early breastfeeding within the first hour after giving birth when applicable (yes, no); adequate breastfeeding support (yes, no); exclusive breastfeeding at discharge from hospital (yes, no); immediate attention by healthcare providers when needed (yes, no); clear communication from healthcare providers (yes, no); allowed companion of choice (yes, no); adequate number of women per room (yes, no); adequate visiting hours for companion of choice (yes, no); adequate number of healthcare providers (yes, no); adequate professionalism of the healthcare providers (yes, no); reduction in QMNC due to COVID-19 (yes, no); and reduction in their general satisfaction due to COVID-19 pandemic (yes, no). When the option “partially” was available, such answers were categorized as “no”. Other variables included in the model were: women born in Norway (yes, no, missing); answered the survey in other language than Norwegian (yes, no); age range (18–24, 25–30, 31–35, 36–39, ≥ 40); educational level (None, Elementary school, Junior High school, High School, University degree, Postgraduate degree / Master / Doctorate or higher); parity (1, > 1); birth mode (vaginal spontaneous, instrumental vaginal birth, Cesarean section).
A two-tailed P-value < 0.05 was considered statistically significant. Statistical analyses were performed using Stata version 14 (Stata Corporation, College Station, TX, USA) and R version 4.1.1 (R Foundation for Statistical Computing, Vienna, Austria).
## Analysis of qualitative data
Systematic Text Condensation (STC) was used to analyse data related to breastfeeding from the open-ended question. Comments including variations on the words breastfeeding and milk / formula, suckle / latch, and / or breast were included in the qualitative analysis. STC was conducted using the following four step method: 1) We became familiar with the data and identified preliminary themes (i.e., partner, information, understaffed ward, consequences of poor breastfeeding support, positive descriptions of breastfeeding support, and poor competence); 2) Meaning units describing women’s experiences and views on early experiences with breastfeeding during the COVID-19 pandemic in Norway were sorted into code groups (i.e., postnatal wards, partner’s role, and consequences of poor breastfeeding support); 3) Code groups were sorted into subgroups, and meaning units in each subgroup were summarized and condensed; and 4) The condensates from Step 3 formed the basis for our final analytic text [32]. In line with the method and guidelines for reporting qualitative research, we used direct quotations to elucidate the findings [32]. Initially, qualitative data were analysed by one researcher (ESV) using NVivo (NVivo Release 1.6.1 [4830], 1999-2022 QSR International). In Step 3, preliminary results were discussed with two more researcher, SK and IHN, who both had access to the raw data. ESV, SK and IHN read and agreed on the final analytic text.
## Results
The study sample included answers from 2922 births, 2425 ($83.0\%$) in 2020 and 497 ($17.0\%$) in 2021 were included (Fig. 1). Characteristics of the study sample are shown in Table 1. Both in 2020 and 2021, few women had an immigrant background ($$n = 204$$) or answered the survey in a different language than Norwegian ($$n = 19$$). In both 2020 and 2021, most women were 25 to 35 years old; nearly four out of five had either a university or postgraduate degree; the number of instrumental births was similar ($12.4\%$ vs. $11.5\%$). When compared to 2021, the numbers of first-time mothers were statistically significant higher in 2020 ($58.4\%$ vs. $53.1\%$, $$P \leq 0.030$$); Cesarean sections were slightly but not statistically significant more frequent ($14.3\%$ vs. $11.3\%$); and the number of women born outside Norway and vaginal spontaneous births were slightly but not statistically significantly lower ($7.2\%$ vs. 5.8 and $73.3\%$ vs. $77.3\%$, respectively).Table 1Characteristics of the study sampleYear of birth2020 $$n = 24252021$$ $$n = 497$$ P-value Women born in Norway Yes2249 (92.7)468 (94.2)0.258 No175 (7.2)29 (5.8)0.271 Missing1 (0.0)0 (0.0)> 0.99 Answered the survey in a language other than Norwegian Yes17 (0.7)2 (0.4)0.654 No2408 (99.3)495 (99.6)0.654 Age range 18–24147 (6.1)32 (6.4)0.750 25–301059 (43.7)237 (47.7)0.101 31–35910 (37.5)184 (37.0)0.833 36–39249 (10.3)37 (7.4)0.054 ≥ 4060 (2.5)7 (1.4)0.148Educational levela None1 (0.0)0 (0.0)> 0.99 Elementary school0 (0.0)0 (0.0)– Junior High school37 (1.5)7 (1.4)0.845 High School498 (20.5)101 (20.3)0.914 University degree1244 (51.3)259 (52.1)0.741 Postgraduate degree / Master / Doctorate or higher645 (26.6)130 (26.2)0.839 Parity 11416 (58.4)264 (53.1)0.030 > 11009 (41.6)233 (46.9)0.030 Birth mode Vaginal spontaneous1778 (73.3)384 (77.3)0.068 Instrumental vaginal birth301 (12.4)57 (11.5)0.559 Cesarean section346 (14.3)56 (11.3)0.077 aWording on education levels was agreed among partners during the Delphi exercise Associations between year of birth [2020, 2021] and factors related to early breastfeeding ($$n = 13$$ factors) are shown in Table 2. When compared to the first year of the pandemic [2020], women who gave birth in 2021 reported higher odds of adequate breastfeeding support (adjOR 1.79; $95\%$ CI 1.35, 2.38); immediate attention when needed (adjOR 1.89; $95\%$ CI 1.49, 2.39); clear communication from healthcare providers (adjOR 1.76; $95\%$ CI 1.39, 2.22); allowed companion of choice (adjOR 1.47; $95\%$ CI 1.21, 1.79); adequate visiting hours for partner and / or relatives (adjOR 1.35; $95\%$ CI 1.09, 1.68); adequate number of healthcare providers (adjOR 1.24; $95\%$ CI 1.02, 1.52); and adequate professionalism of the healthcare providers (adjOR 1.65; $95\%$ CI 1.32, 2.08). When compared to the first year of the pandemic [2020], women who gave birth in 2021 reported lower odds of reduction in QMNC due to COVID-19 pandemic (adjOR 0.73; $95\%$ CI 0.60, 0.89). We found no difference by year of birth; in having skin-to-skin contact with the baby in the first hour after giving birth; early breastfeeding; exclusive breastfeeding at discharge; adequate number of women per room; or reduction in their general satisfaction due to COVID-19.Table 2Results of the logistic regression analysisa Factors related to early breastfeedingn cases (yes)n women (total)Crude OR ($95\%$ CI)Adjusted ORb ($95\%$ CI) Opportunity to have skin-to-skin contact in the first hour after giving birth (when applicable, i.e., in absence of maternal or neonatal health problems) 2020214822701.001.00 20214604801.31 (0.81, 2.12)1.07 (0.59, 1.92) Early breastfeeding (when applicable, i.e., in absence of maternal or neonatal health problems) 2020196822331.001.00 20214304691.48 (1.04, 2.11)1.32 (0.88, 1.96) Adequate breastfeeding support 2020190224251.001.00 20214334971.86 (1.40, 2.46)1.79 (1.35, 2.38) Exclusive breastfeeding at discharge 2020186824251.001.00 20214024971.26 (0.99, 1.61)1.14 (0.89, 1.46) Immediate attention when needed 2020163624251.001.00 20213974971.91 (1.51, 2.42)1.89 (1.49, 2.39) Clear communication from healthcare providers 2020165724251.001.00 20213954971.79 (1.42, 2.27)1.76 (1.39, 2.22) Allowed companion of choice 202096224251.001.00 20212464971.49 (1.23, 1.71)1.47 (1.21, 1.79) Adequate number of women per room 2020181224251.001.00 20213564970.85 (0.69, 1.06)0.84 (0.67, 1.04) Adequate visiting hours for partner and / or relatives 202055224251.001.00 20211414971.34 (1.08, 1.67)1.35 (1.09, 1.68) Adequate number of healthcare providers 202083024251.001.00 20211984971.27 (1.04, 1.55)1.24 (1.02, 1.52) Adequate professionalism of the healthcare providers 2020159424251.001.00 20213804971.69 (1.35, 2.12)1.65 (1.32, 2.08) Reduction in the QMNC due to COVID-19 2020163124251.001.00 20212944970.71 (0.58, 0.86)0.73 (0.60, 0.89) Reduction in their general satisfaction due to COVID-19 (among women who reported a reduction in the QMNC due to COVID-19) 2020144216311.001.00 20212642941.15 (0.77, 1.73)1.13 (0.75, 1.70) aOR are calculated with 2020 year of birth as reference category bAnalyses were adjusted for the following variables: born in Norway (yes, no, missing), responded to a non-Norwegian survey (yes, no), age range (18–24, 25–30, 31–35, 36–39, ≥ 40), educational level (None, Elementary school, Junior high school, High school, University degree, Postgraduate degree / Master / Doctorate or higher), parity (1, > 1), birth mode (Vaginal spontaneous, Instrumental vaginal birth, Cesarean section). Wording on education levels agreed among partners during the Delphi exerciseAbbreviations: COVID-19 *Corona virus* disease of 2019, QMNC Quality of maternal and newborn care
## Findings on qualitative data
Of the 2922 women who responded, 1021 ($34.9\%$) provided a free-text comment. Of the 1021 free-text comments, 88 comments ($8.6\%$) were directly related to breastfeeding. There were eight non-Norwegian language comments (i.e., four comments in English, three in Swedish and one in German; none of which were related to breastfeeding). The following themes were identified during analysis of the open-ended question: 1) Understaffed postnatal wards, 2) Early discharge and a lack of professional support, 3) The importance of breastfeeding support from companion of choice, and 4) Long term consequences.
## Understaffed postnatal wards
Women in the study described understaffed postnatal wards. Positive characterizations of healthcare providers included words such as being nice and skilled, often followed by descriptions of how the wards were understaffed and unavailable to the women. One woman who pointed out that understaffed postnatal wards may be a common problem, put it like this: “The breastfeeding guidance was extremely poor and divergent... A large part of the staff was nice, but it was very clear that they were understaffed. For all I know, understaffing [of postnatal wards] is a common problem.” ( Norwegian woman No. 1746).
## Early discharge and a lack of professional support
Among negative narratives, several mentioned that they would have liked to stay for longer at the postnatal ward, to learn how to breastfeed and feel safe before being discharged. Some related early discharge due to COVID-19 restrictions and explained how they had to be discharged early to lessen the risk of spreading the virus. Others disclosed voluntary early discharge due to COVID-19 restrictions or a lack of help at the hospital. Most birth narratives contained positive language, while descriptions of the immediate postpartum period tended to be more negative in nature. Women in the study highlighted a need for improved breastfeeding support and continuity of care. Most women described intentions to breastfeed as a means of providing optimal nutrition and to bond with the newborn. Some women described being advised to watch videos or use the internet for breastfeeding support, while others described healthcare providers as stressed or heavy-handed when asking for breastfeeding support. Women reported feeling like a burden, being forgotten, or feeling like bad mothers. Some described how they felt pushed to breastfeed by healthcare providers and not being offered help if they for different reasons preferred to abstain from breastfeeding. Others described how healthcare providers fed the baby with formula without information or consent. Some women questioned the healthcare providers competence in breastfeeding and the quality of handovers between hospital shifts. One woman who initially planned to breastfeed, who ended up not breastfeeding, described it like this: “The staff often said, ‘I’ll get someone to help you’, but it never happened. It happened several times that I had to wait more than half an hour to get help. With a screaming newborn that I was unable to breastfeed, it felt absolutely horrific. The staff would come, but then had to run again halfway through a sentence.” ( Norwegian woman No. 552).
## The importance of breastfeeding support from companion of choice
While some women described how partners were allowed to stay at the postnatal ward, other women described how their companion of choice was not allowed to visit her and baby at the postnatal ward due to COVID-19 restrictions. Some women discharged themselves early to be reunited with their partner. Having a partner present during the postnatal period was identified as a supportive factor for breastfeeding. This was reported to have positive impacts on the emotional well-being of the mother, as well as providing practical assistance and reducing the workload on healthcare providers. One woman who preferred her partner to stay at the postnatal ward put it like this: “It was a great burden that my partner was not allowed to be there. Having children and breastfeeding is a family project, and not something I had volunteered to do alone.” ( Norwegian woman No. 3135).
Others were pleased with the calm atmosphere that followed the strict visitor restrictions, including restrictions related to the partner. A calm atmosphere was described as positive for breastfeeding.
## Long term consequences
Women expressed how family centered care was lacking, with possible negative consequences for breastfeeding, bonding, and becoming a family. Further, fearing postpartum depression was mentioned by several, and some stated that their experiences made them fear having more children in the future. One woman who believes she suffers from a postpartum depression put it in the following way: “... I experienced the first days of my daughter’s life as very traumatic. Due to COVID-19 restrictions, we also did not receive sufficient help and guidance (including breastfeeding support). Due to the lack of help, I have struggled with what I believe is a postpartum depression.” ( Norwegian woman No. 506).
## Discussion
When compared to women who gave birth in Norway during the first year of the COVID-19 pandemic, women who gave birth the following year were more likely to experience adequate breastfeeding support; immediate attention when needed; clear communication from healthcare providers; being allowed a companion of choice; adequate visiting hours for partner and / or relatives; adequate number of healthcare providers; and adequate professionalism of the healthcare providers. Compared to 2020, in 2021, the women also experienced lower odds of reduction in QMNC due to COVID-19 pandemic. When comparing factors related to breastfeeding in 2020 vs. 2021, we found no difference in the opportunity to have skin-to-skin contact with the baby in the first hour after giving birth; early breastfeeding; exclusive breastfeeding at discharge; adequate number of women per room; or reduction in their general satisfaction due to COVID-19. In comments related to breastfeeding (2020 and 2021), women described understaffed postnatal wards; early discharge and highlighted the importance of breastfeeding support from healthcare providers and companion of choice; and concerns about long-term consequences such as postpartum depression.
Many factors contribute to successful breastfeeding in the early postpartum period, and we found a range of breastfeeding-related factors were improved for women giving birth in Norway in 2021, compared to the first year of the COVID-19 pandemic [2020]. One UK study from the COVID-19 pandemic reported that face-to-face breastfeeding support was reduced during the pandemic and some women struggled to get breastfeeding support, while others found strict regulations positive because of increased time at home, less pressure and fewer visitors [33]. In the current study, answers to the open-ended question revealed some women were pleased with the calm atmosphere that followed the strict visitor restrictions. Our findings are in line with the UK study and suggest that the COVID-19 pandemic affected women’s breastfeeding experiences differently.
When comparing data from 2020 with data from 2021, we found little difference in early skin-to-skin contact, early breastfeeding within the first hour after giving birth or exclusive breastfeeding at discharge. Women’s satisfaction with the number of women per room stayed constant over the study period and women’s general satisfaction with care due to COVID-19 did not improve significantly from 2020 to 2021. One Italian study including 204 mothers and babies in the early stage of the COVID-19 pandemic (9 March to 8 May 2020) found a decrease in exclusively breastfeeding in the studied population [3]. Consistent with our findings, one study including 821 women who gave birth in Norway in the spring of 2020 found great reliance on breast-milk substitutes, which may imply that fewer women in Norway were exclusively breastfeeding during the initial phase of the COVID-19 pandemic [4]. Findings of a quantitative study including 3642 women giving birth in Norway during the pandemic adds support that one in three women experienced being discharged early due to COVID-19 related factors [34]. In our data, one in four women ($23.2\%$) who underwent labor in Norway in the study period reported not exclusively breastfeeding at discharge [17]. In the current study, when comparing data from 2020 with data from 2021, we found no difference in early breastfeeding or exclusive breastfeeding at discharge. In 2020, a nation-wide Norwegian report showed that $97\%$ of babies born in Norway in 2018 were breastfed before postpartum discharge [25]. To our knowledge, updated national data on exclusive breastfeeding at discharge has not been published. However, early discharge may be one explanation for why fewer women exclusively breastfeed their babies early in the pandemic [4]. The lack of improvement in exclusive breastfeeding during the study period should alert policy makers in postnatal care services to implement specific quality improvement actions. To better understand the reasons for a lack of improvement in exclusive breastfeeding at discharge and no change in women’s general satisfaction with care due to COVID-19 from 2020 to 2021, future studies with other designs are needed.
In comments related to breastfeeding, the open-ended question in our survey gave information on understaffed postnatal wards, the importance of breastfeeding support from healthcare professionals and companion of choice, and a concern for long term consequences, such as postpartum depression, due to insufficient breastfeeding support during the pandemic. Our findings related to understaffed postnatal wards and the importance of partner are supported by a qualitative study exploring women’s experiences with giving birth in Norway during the pandemic [30]. Further, studies from Norway and the UK support the concern for the occurrence of postpartum depression, as they found an increase in maternal depression and anxiety postpartum, during the COVID-19 pandemic [34, 35]. A Norwegian national report on parental experience of QMNC published before the COVID-19 pandemic [2018] found that new parents in Norway were well satisfied with the care given on labor wards, however, postnatal care scored lower than other areas [36]. Findings in the Norwegian report suggest that our results related to QMNC in Norway during the first year of the Covid-19 pandemic cannot be attributed to the pandemic alone and must therefore be interpreted with caution. The qualitative findings in the current study support the concerns arising from the quantitative data, such as those related to women experiencing inadequate breastfeeding support, lack of attention when needed, not being allowed companion of choice, or a low number of healthcare providers.
## Strengths and limitations
It may be seen as a limitation to the study that changes in local and national COVID-19 regulations over time were not accounted for. Because we only included data from women who gave birth during the pandemic, comparison with pre-pandemic data must be made with caution. It may be seen as a strength that the study includes both quantitative and qualitative data (i.e., triangulation or mixed-methods), an approach which provides a more comprehensive picture of the results than either method could do alone [37]. The qualitative data is not suitable for quantification, and comparison between the free-text responses given in 2020 vs. 2021 is therefore not included. The study used standard procedures and indicators, and allowed for future rounds of data collection, and comparison over time and settings. Open-ended questions can provide crucial information that closed-ended questions cannot deliver [38, 39]. Women themselves chose whether to answer the open-ended question or not, thus these questions were not subject to systematic measurements [38, 39]. Therefore, we did not analyse the open-ended questions for 2020 and 2021 separately. Due to self-administration, open-ended questions may cause selection bias in those responding [39]; women who were satisfied with breastfeeding support may be less likely to provide comments related to breastfeeding. The results from the open-ended questions should therefore be interpreted with caution. We acknowledge that the online survey lacked important information on the sample, such as more information on maternal and newborn clinical characteristics which may be relevant for the interpretation of the results [17]. Caution is necessary when comparing the current study’s $7.0\%$ response rate for migrant women with national data indicating that $28.9\%$ of women who gave birth in Norway in 2020 and 2021 were born outside the country [20]. Women who experienced vaginal birth, planned or emergency Cesarean sections were all included in analysis, however, experiences related to early breastfeeding may differ between these groups due to several factors. Causality cannot be drawn from this cross-sectional study [40].
## Recommendation for research and for policies
Our study provides critical information for researchers, policy makers and clinicians on the need for continuous surveillance of national breastfeeding rates and for improving postnatal care services and breastfeeding support in Norway and similar settings. This study highlights the importance of promoting continuity of care and evidence-based interventions, such as inclusion of companion of choice in postnatal wards. To improve women’s general satisfaction with postpartum care, adequate staffing for breastfeeding support must be made available to all new mothers.
## Conclusions
In the second year of the COVID-19 pandemic, several but not all breastfeeding-related factors improved for women giving birth in Norway compared to the first year of the pandemic. Women’s general satisfaction with care during COVID-19 and rates of exclusive breastfeeding did however not improve significantly from 2020 to 2021. The findings should alert researchers, policy makers and clinicians in postnatal care services to improve future practices.
## Supplementary Information
Additional file 1.Additional file 2.Additional file 3.
## References
1. Vivilaki VG, Asimaki E. **Respectful midwifery care during the COVID-19 pandemic**. *Eur J Midwifery* (2020.0) **4** 1-2. DOI: 10.18332/ejm/120070
2. Chmielewska B, Barratt I, Townsend R, Kalafat E, van der Meulen J, Gurol-Urganci I, O'Brien P, Morris E, Draycott T, Thangaratinam S. **Effects of the COVID-19 pandemic on maternal and perinatal outcomes: a systematic review and meta-analysis**. *Lancet Glob Health* (2021.0) **9** E759-E772. DOI: 10.1016/s2214-109x(21)00079-6
3. Latorre G, Martinelli D, Guida P, Masi E, De Benedictis R, Maggio L. **Impact of COVID-19 pandemic lockdown on exclusive breastfeeding in non-infected mothers**. *Int Breastfeed J* (2021.0) **16** 36. DOI: 10.1186/s13006-021-00382-4
4. 4.Stette AHH, Thorsteinsen C, Henriksen L. COVID-19: initiation of breastfeeding during a global pandemic. Sykepleien Forskning. 2021;16(85747). 10.4220/sykepleienf.2021.85747.
5. 5.Women’s Rights in Childbirth Must be Upheld During the Coronavirus Pandemic [https://www.internationalmidwives.org/assets/files/news-files/2020/03/icm-statement_upholding-womens-rights-during-covid19-5e83ae2ebfe59.pdf].
6. 6.COVID-19 and breastfeeding; Position paper [https://www.euro.who.int/__data/assets/pdf_file/0010/437788/breastfeeding-COVID-19.pdf].
7. Zanardo V, Tortora D, Guerrini P, Garani G, Severino L, Soldera G, Straface G. **Infant feeding initiation practices in the context of COVID-19 lockdown**. *Early Hum Dev* (2021.0) **152** 105286. DOI: 10.1016/j.earlhumdev.2020.105286
8. Merewood A, Davanzo R, Haas-Kogan M, Vertecchi G, Gizzi C, Mosca F, Burnham L, Moretti C. **Breastfeeding supportive practices in European hospitals during the COVID-19 pandemic**. *J Matern Fetal Neonatal Med* (2022.0) **35** 8514-8520. DOI: 10.1080/14767058.2021.1986482
9. Sankar MJ, Sinha B, Chowdhury R, Bhandari N, Taneja S, Martines J, Bahl R. **Optimal breastfeeding practices and infant and child mortality: a systematic review and meta-analysis**. *Acta Paediatr* (2015.0) **104** 3-13. DOI: 10.1111/apa.13147
10. Armstrong J, Reilly JJ. **Breastfeeding and lowering the risk of childhood obesity**. *Lancet* (2002.0) **359** 2003-2004. DOI: 10.1016/s0140-6736(02)08837-2
11. Sadauskaitė-Kuehne V, Ludvigsson J, Padaiga Ž, Jašinskienė E, Samuelsson U. **Longer breastfeeding is an independent protective factor against development of type 1 diabetes mellitus in childhood**. *Diabetes Metab Res Rev* (2004.0) **20** 150-157. DOI: 10.1002/dmrr.425
12. 12.Horta BL, Loret de Mola C, Victora CG. Breastfeeding and intelligence: a systematic review and meta-analysis. Acta Paediatr 2015, 104:14-19. 10.1111/apa.13139.
13. Pereyra-Elías R, Quigley MA, Carson C. **To what extent does confounding explain the association between breastfeeding duration and cognitive development up to age 14? Findings from the UK millennium cohort study**. *PLoS One* (2022.0) **17** e0267326. DOI: 10.1371/journal.pone.0267326
14. Ciampo LAD, Ciampo IRLD. **Breastfeeding and the benefits of lactation for women’s health**. *Revista Brasileira de Ginecologia e Obstetrícia* (2018.0) **40** 354-359. DOI: 10.1055/s-0038-1657766
15. Lubbe W, Botha E, Niela-Vilen H, Reimers P. **Breastfeeding during the COVID-19 pandemic–a literature review for clinical practice**. *Int Breastfeed J* (2020.0) **15** 82. DOI: 10.1186/s13006-020-00319-3
16. Pace RM, Williams JE, Järvinen KM, Belfort MB, Pace CD, Lackey KA, Gogel AC, Nguyen-Contant P, Kanagaiah P, Fitzgerald T. **Characterization of SARS-CoV-2 RNA, antibodies, and neutralizing capacity in milk produced by women with COVID-19**. *MBio* (2021.0) **12** e03192-e03120. DOI: 10.1128/mBio.03192-20
17. Lazzerini M, Covi B, Mariani I, Drglin Z, Arendt M, Nedberg IH, Elden H, Costa R, Drandić D, Radetić J. **Quality of facility-based maternal and newborn care around the time of childbirth during the COVID-19 pandemic: online survey investigating maternal perspectives in 12 countries of the WHO European region**. *Lancet Regional Health-Europe* (2022.0) **13** 100268. DOI: 10.1016/j.lanepe.2021.100268
18. 18.Timeline: News from Norwegian Ministries about the Coronavirus disease Covid-19 [https://www.regjeringen.no/en/topics/koronavirus-covid-19/timeline-for-news-from-norwegian-ministries-about-the-coronavirus-disease-covid-19/id2692402/] 10.1787/e55a6b9a-en
19. Lazzerini M, Mariani I, TRdM e L, Felici E, Martelossi S, Lubrano R, Lucarelli A, Trobia GL, Cogo P, Peri F. **WHO standards-based tools to measure service providers’ and service users’ views on the quality of hospital child care: development and validation in Italy**. *BMJ Open* (2022.0) **12** e052115. DOI: 10.1136/bmjopen-2021-052115
20. 20.Medisinsk fødselsregister - statistikkbank [Medical Birth Registry - statistics] [http://statistikkbank.fhi.no/mfr/].
21. 21.Healthcare for persons without legal residence in Norway [https://www.helsenorge.no/en/foreigners-in-norway/healthcare-for-persons-without-legal-residence/#pregnant-women%E2%80%99s-rights-to-health-services].
22. 22.Mossialos E, Wenzl M, Osborn R, Sarnak D. International profiles of health care systems: Canadian Agency for Drugs and Technologies in health; 2016. 2015. 10.15868/socialsector.25100.
23. Theurich MA, Davanzo R, Busck-Rasmussen M, Díaz-Gómez NM, Brennan C, Kylberg E, Bærug A, McHugh L, Weikert C, Abraham K. **Breastfeeding rates and programs in Europe: a survey of 11 national breastfeeding committees and representatives**. *J Pediatr Gastroenterol Nutr* (2019.0) **68** 400-407. DOI: 10.1097/mpg.0000000000002234
24. 24.Norwegian Directorate of Health. Nasjonal handlingsplan for bedre kosthold (2017-2021) [Norwegian National Action Plan for a Healthier Diet (2017-2021)]. In.; 2017.
25. Myhre J, Andersen L, Kristiansen A. *Spekost 3 Landsomfattende undersøkelse av kostholdet blant spedbarn i Norge, 6 måneder [Spekost 3 Nationwide survey of the diet among infants in Norway, 6 months]* (2020.0)
26. 26.Norwegian Institute of Public Health. Spekost 3 Landsomfattende undersøkelse av kostholdet blant spedbarn i Norge, 12 måneder [Spekost 3 Nationwide survey of the diet among infants in Norway, 12 months]. In.; 2020.
27. 27.Lov om folketrygd (folketrygdloven) [National insurance act] [https://lovdata.no/dokument/NL/lov/1997-02-28-19/].
28. 28.Ministry of Labour and Social InclusionAct relating to working environment, working hours and employment protection, etc2006(Working Environment Act). In. *Act relating to working environment, working hours and employment protection, etc* (2006.0)
29. Engjom H, Aabakke AJ, Klungsøyr K, Svanvik T, Äyräs O, Jonasdottir E, Thurn L, Jones E, Pettersson K, Nyfløt LT. **COVID-19 in pregnancy–characteristics and outcomes of pregnant women admitted to hospital because of SARS-CoV-2 infection in the Nordic countries**. *Acta Obstet Gynecol Scand* (2021.0) **100** 1611-1619. DOI: 10.1111/aogs.14160
30. Eri TS, Downe S, Vedeler C, Blix E, Nilsen ABV. *Giving birth and becoming a parent during the COVID-19 pandemic: a qualitative analysis of 827 women’s responses to an online survey in Norway* (2021.0)
31. 31.World Health Organization. Standards for improving quality of maternal and newborn care in health facilities. 2016.
32. Malterud K. **Systematic text condensation: a strategy for qualitative analysis**. *Scandinavian J Public Health* (2012.0) **40** 795-805. DOI: 10.1177/1403494812465030
33. Brown A, Shenker N. **Experiences of breastfeeding during COVID-19: lessons for future practical and emotional support**. *Matern Child Nutr* (2021.0) **17** e13088. DOI: 10.1111/mcn.13088
34. 34.Eberhard-Gran M, Engelsen LY, Al-Zirqi I, Vangen S. Depressive symptoms and experiences of birthing mothers during COVID-19 pandemic. Tidsskrift for Den norske legeforening. 2022. 10.4045/tidsskr.21.0450.
35. Fallon V, Davies SM, Silverio SA, Jackson L, De Pascalis L, Harrold JA. **Psychosocial experiences of postnatal women during the COVID-19 pandemic. A UK-wide study of prevalence rates and risk factors for clinically relevant depression and anxiety**. *J Psychiatr Res* (2021.0) **136** 157-166. DOI: 10.1016/j.jpsychires.2021.01.048
36. Holmboe O, Sjetne I. *Brukeres erfaringer med fødsels- og barselomsorgen i 2017. Nasjonale resultater. [User experiences of birth and postnatal care in 2017. National results]* (2018.0)
37. Heale R, Forbes D. **Understanding triangulation in research**. *Evidence-Based Nurs* (2013.0) **16** 98. DOI: 10.1136/eb-2013-101494
38. Singer E, Couper MP. **Some methodological uses of responses to open questions and other verbatim comments in quantitative surveys Methods, data, analyses**. *A J Quantitative Methods Surv Methodol (mda)* (2017.0) **11** 115-134
39. Reja U, Manfreda KL, Hlebec V, Vehovar V. **Open-ended vs. close-ended questions in web questionnaires**. *Develop Appl Stat* (2003.0) **19** 159-177
40. Spector PE. **Do not cross me: optimizing the use of cross-sectional designs**. *J Bus Psychol* (2019.0) **34** 125-137. DOI: 10.1007/s10869-018-09613-8
|
---
title: 'Effects of eating together online on autonomic nervous system functions: a
randomized, open-label, controlled preliminary study among healthy volunteers'
authors:
- Hideaki Hasuo
- Nahoko Kusaka
- Mutsuo Sano
- Kenji Kanbara
- Tomoki Kitawaki
- Hiroko Sakuma
- Tomoya Sakazaki
- Kohei Yoshida
- Hisaharu Shizuma
- Hideo Araki
- Motoyuki Suzuki
- Satoshi Nishiguchi
- Masaki Shuzo
- Gaku Masuda
- Kei Shimonishi
- Kazuaki Kondo
- Hirotada Ueda
- Yuichi Nakamura
journal: BioPsychoSocial Medicine
year: 2023
pmcid: PMC9998259
doi: 10.1186/s13030-023-00263-8
license: CC BY 4.0
---
# Effects of eating together online on autonomic nervous system functions: a randomized, open-label, controlled preliminary study among healthy volunteers
## Abstract
### Background
Eating alone has been significantly associated with psychological distress. However, there is no research that evaluates the effects or relation of eating together online to autonomic nervous system functions.
### Methods
This is a randomized, open-label, controlled, pilot study conducted among healthy volunteers. Participants were randomized into either an eating together online group or an eating-alone group. The effect of eating together on autonomic nervous functions was evaluated and compared with that of the control (eating alone). The primary endpoint was the change in the standard deviation of the normal-to-normal interval (SDNN) scores among heart rate variabilities (HRV) before and after eating. Physiological synchrony was investigated based on changes in the SDNN scores.
### Results
A total of 31 women and 25 men (mean age, 36.6 [SD = 9.9] years) were included in the study. In the comparison between the aforementioned groups, two-way analysis of variance revealed interactions between time and group on SDNN scores. SDNN scores in the eating together online group increased in the first and second halves of eating time (F[1,216], $P \leq 0.001$ and F[1,216], $$P \leq 0.022$$). Moreover, high correlations were observed in the changes in each pair before and during the first half of eating time as well as before and during the second half of eating time ($r = 0.642$, $$P \leq 0.013$$ and $r = 0.579$, $$P \leq 0.030$$). These were statistically significantly higher than those in the eating-alone group ($$P \leq 0.005$$ and $$P \leq 0.040$$).
### Conclusions
The experience of eating together online increased HRV during eating. Variations in pairs were correlated and may have induced physiological synchrony.
### Trial registration
The University Hospital Medical Information Network Clinical Trials Registry, UMIN000045161. Registered September 1, 2021. https://center6.umin.ac.jp/cgi-open-bin/icdr/ctr_view.cgi?recptno=R000051592.
## Background
Diet is important not only for nutritional and health aspects but also because it constitutes an essential part of daily social interactions [1]. Dietary environment can affect health from various biopsychosocial aspects. Solitary eating has been related to the development of depressive symptoms, increased mortality, and/or reduced diet quality and intake [2–5]. Recently, social isolation owing to the COVID-19 pandemic has resulted in increased solitary eating, which has been noted to be associated with psychological distress [6, 7]. Additionally, a related cohort study showed that a greater degree of unhappiness was associated with a greater proportion of eating alone [1].
Eating together has been reported to increase food intake and to improve taste through social associations with other people [8, 9]. The social facilitation of eating is defined as the promotion of an individual’s activities, such as an increase in food intake, by the presence of other people while eating [9]. Eating together is also influenced by social modeling [10], in which one’s eating behavior influences others and vice versa. As a factor affected by social interactions with other people, the feeling of relaxation enhanced by communication while eating together is important [8, 10]. Apparently, eating together increases subjective wellbeing and provides a sense of relaxation [11, 12].
In recent years, eating together online has become popular following the evolution of video conversation technologies. This phenomenon, termed digital commensality, can circumvent environmental constraints to increase the maintenance and enhancement of health from the biopsychosocial aspect of those who have been eating alone [13]. Moreover, it has been reported that eating together online can be perceived by participants as “just alone but together,” with increased food intake and reduced loneliness [14]. The results of this report suggest that eating together, even online, may stimulate social interactions. However, this has not been specifically demonstrated. Furthermore, no study has investigated the effects of eating together online on feelings of relaxation, energy or loneliness on autonomic nervous functions, which is an objective evaluation of relaxation. This study thus aimed to address these research gaps and hypothesized that eating together online affects autonomic nervous function by social interaction through a variety of factors, including feelings of relaxation, energy, and loneliness.
It has been reported that relaxation, such as by hypnosis, or intense loneliness reduces heart rate variability (HRV) at rest; reduced autonomic function can be predicted by HRV [15, 16]. It has also been reported that hand gripping between a patient with cancer and their family caregiver positively affects each other’s HRV [17]. The association or interdependency of physiological activities between two people is referred to as physiological synchrony [18]. Quantitation using maximal cross-correlation or cross-correlation with local slopes has been reported to be effective for the assessment of physiological synchrony using HRV [19, 20]. However, no method has been established yet. We thus further hypothesized that people eating together online would favorably affect each other’s autonomic functions, which would further provoke physiological synchrony.
## Objective
This study aims to evaluate among healthy volunteers the effect on HRV of eating together online in comparison with persons eating alone.
## Study design
This is a randomized, open-label, controlled, preliminary study conducted among healthy volunteers who worked at Kansai Medical University in Osaka, Japan. The study was approved by the Medical Ethics Committee of Kansai Medical University (reference number: 2021167) and was performed in accordance with the Declaration of Helsinki (as revised in 2013). Written informed consent was obtained from all study participants before the commencement of the study procedure. The study was registered with the University Hospital Medical Information Network Clinical Trials Registry (approval number: UMIN000045161) on September 1, 2021. This study was conducted from January to April 2022.
## Study participants
The study participants were healthy volunteers, defined as “normal” persons who had no significant medical conditions or histories and no difficulty in their daily lives. They were employees at Kansai Medical University who responded to our post on the volunteer recruitment bulletin board at the university. The exclusion criteria included [1] currently taking medication or seeking medical care and [2] having neurological or mental disorders such as cognitive dysfunction and being unable to communicate. Participants were excluded if they met the diagnostic criteria for neurological or mental disorders according to the fifth edition of the Diagnostic and Statistical Manual of Mental Disorders [21], confirmed by two psychosomatic physicians.
## Study procedures
Figure 1 summarizes the study regimen. Participants were randomized into either an eating together online (pairing) group or a control (eating alone) group by a computer using the minimization method at a 1:1 ratio. Each participant was informed of their allocated group after randomization. The participants were not allowed to change their groups during the study period. Concurrently, the investigators participating in the study were also informed of their designated groups. The study staff in charge of the statistical analysis concealed the results of the randomization. The participants’ names were also kept anonymous. The data were collected in interview rooms by the clinicians responsible for the study. Each investigator interviewed and assessed the participants at the beginning of the study. Fig. 1Flowchart of the study procedures. UMACL, UWIST Mood Adjective Checklist; HRV, Heart rate variability The study participants completed a self-report questionnaire, the UWIST Mood Adjective Checklist (UMACL), at the beginning and end of the study period. Each study subject attached a special electrode pad connected to the HRV measurement device (myBeat WHS-1; Union Tool Co., Tokyo, Japan) to their chest. The participants ate a snack for 10 min and took breaks for 5 min each before and after eating. The 10 min eating time was divided into first and second halves. The data evaluated using the HRV were continuously recorded for 20 min. The HRV scores for each session were calculated based on the mean of the 5-min HRV record. We used HRV analysis software (Kubios HRV version 3.1; Kubios Oy, Kuopio, Finland), which is highly reliable for short-term recording [22].
The snacks were commercial cookies. The participants eating together online ate the snacks while looking at their partner on the computer screen, during which conversation was allowed. The participants in the eating-alone group ate the snack while looking at an offline black screen, during which conversation was not allowed; however, soliloquy was allowed. During the 20-min eating time, including snack time and the short breaks, participants remained in their allocated eating rooms.
## UWIST mood adjective checklist
The Japanese version of the UMACL was used to evaluate the mood of comfort [23]. The original checklist, developed by Matthews et al., was created based on dimension theory, making it possible to assess arousal levels [24]. The scale has two subscales that can be used to evaluate energetic arousal (10 items; vigorous vs. tired: coefficient α = 0.79) and tense arousal (10 items; nervous vs. relaxed: coefficient α = 0.76) [25]. High energetic arousal represents active and happy, whereas low tense arousal represents calm and quiet. Participants were asked to respond on a 4-point Likert scale. In a previous study, the mean energetic arousal was 24.4 (standard deviation [SD]: 0.5) for males and 24.4 (SD: 0.4) for females, and the mean tense arousal was 18.5 (SD: 0.4) for males and 17.56 (SD: 0.3) for females [25].
## Heart rate variability and standard deviation of the normal-to-normal interval
HRV, the fluctuation of heartbeat intervals measured using an electrocardiogram, is used to evaluate autonomic nerve activities [26, 27]. HRV tends to be lower in a person with anxiety or depression. However, it is relative rather than absolute; therefore, it is not directly compared among individuals.
Standard deviation of the normal- to- normal interval (SDNN) is the quantification of HRV to further compare it among individuals. Particularly, SDNN is the standard deviation of the R-R intervals of the heartbeat in a certain time duration and is obtained via time-domain analysis. SDNN was used to evaluate cardiovascular compatibility. SDNN includes all the different types of variations and represents total variability [28]. It assesses the flexibility of the autonomic nervous system and the balance of sympathetic and parasympathetic nervous systems, with an increase in SDNN reflecting the stability of these systems [16, 29]. The grand mean of SDNN scores among resting adults is 50 mseconds [29].
## Endpoints
The primary endpoint of this study was the change in the SDNN score before and during eating. The key secondary endpoints were the amount of cookie intake, change in UMACL score, and correlation coefficient of pairs in change in SDNN score.
## Sample size estimation
This is a preliminary study conducted among healthy volunteers to evaluate the effects of eating together online on autonomic nervous system functions. To the best of our knowledge, no similarly designed studies have been conducted previously. Therefore, we recruited as many study volunteers as possible.
## Statistical analysis
Continuous data are summarized as means with SD, and discrete data are presented as the number of subjects (n) and their frequencies (%), as appropriate. Pearson’s chi-square test was used to evaluate discrete data, including age, sex, and mutual relationships. An unpaired two-sided t-test was used to compare mean age. Changes in UMACL scores (before and after eating) were analyzed using one-way repeated-measures analysis of variance (ANOVA). To compare the change in the UMACL and SDNN scores between the two groups, two-way repeated measures ANOVA with fixed effects of time points and groups was used. Moreover, the variable effect of subjects was used to examine the time course changes of these scores. In ANOVA, multiple comparisons were corrected using Bonferroni’s method. Lastly, after calculating the correlation coefficients of pairs in the change in SDNN scores, we performed Fisher’s z-transformation on the correlation coefficients followed by the z-test statistic.
The last UMACL and SDNN scores of participants who withdrew from the study before completion were used for analysis. A significance level of alpha < 0.05 was used for statistical analysis. Statistical analyses were conducted using SPSS version 18.0 J for Macintosh (SPSS Inc., Chicago, IL, USA). Only the z-test statistic was calculated based on web links and references without using SPSS [30, 31].
## Clinical demographic characteristics
A total of 56 healthy volunteers were randomized into either the eating together online group ($$n = 28$$) or the eating-alone group ($$n = 28$$) or 28 pairs ($100.0\%$) and completed the study. Table 1 presents the demographic and clinical characteristics of the participants. The mean age was 36.6 years (SD: 10.1), and 25 were male and 31 female. The mutual relationships of the participants included 47 work colleagues and nine friends. There were no group differences in age, sex, or mutual relationship. Table 1Demographic and mutual relationships of the study participantsEating together online groupEating-alone group($$n = 28$$)($$n = 28$$) P-valueAge (year), mean (SD)37.1(10.3)36.1(9.9)0.711Sex, n (%) Male11(39.3)14(50.0)0.296 Female17(60.7)14(50.0)Mutual relationships Work colleague23(82.1)24(85.7)0.500 Friend5(17.9)4(14.3) SD Standard deviation
## Changes in SDNN scores and between-group comparisons
Figure 2 shows the changes in the SDNN scores between the pre-eating break and during eating or the break after eating. During the pre-eating break, the SDNN scores were 32.8 (SD: 10.8) and 32.2 (SD: 12.7) for the eating together online and eating-alone groups, respectively. The SDNN scores for the eating together online group were significantly higher during eating than during the pre-eating break, rather than thereafter (first half of eating; $P \leq 0.001$, second half of eating; $P \leq 0.001$, rest after eating; $$P \leq 0.071$$). The SDNN scores for the eating-alone group were higher during and after eating than during the pre-eating break. However, the difference was insignificant (first half of eating; $$P \leq 0.708$$, second half of eating; $$P \leq 0.093$$, rest after eating; $$P \leq 0.556$$).Fig. 2Changes in SDNN between the pre-eating break and during or break after eating. SDNN, standard deviation of the normal-to-normal interval A two-way ANOVA showed an interaction between time and group in the change in SDNN score (F[3,216], $$P \leq 0.037$$). The change in SDNN score of the eating together online group was significantly higher than that of the eating-alone group before and during the first half of eating, and before as well as during the second half of eating (F[1,216], $P \leq 0.001$ and F[1,216], $$P \leq 0.022$$). However, the difference did not differ for before and after the breaks (F[1,216)]; $$P \leq 0.287$$).
## Amount of cookie intake
The amount of cookie intake was 4.2 (SD: 2.6) for the online eating together group and 3.5 (SD: 2.5) for the eating-alone group ($$P \leq 0.310$$).
## Changes and between-group comparisons of scores
Table 2 shows the change in the UMACL scores and comparisons between the groups. The mean energetic arousal scores at the beginning of the study were 30.0 (SD: 1.2) for male and 31.3 (SD: 0.5) for female participants, while those for mean tense arousal were 26.9 (SD: 0.5) and 26.3 (SD: 0.3), respectively. The energetic arousal score was higher after eating in the eating together online group and differed between the groups ($$P \leq 0.036$$). The tense arousal score was lower after eating in both the groups, with no difference between the groups ($$P \leq 0.199$$).Table 2Changes and between-group comparisons of UMACL scoresEating together online groupEating-alone group P-valueBefore eatingAfter eating P-valueBefore eatingAfter eating P-valuemeanSDmeanSDmeanSDmeanSDEnergetic Arousal score30.53.732.43.40.00130.73.429.93.70.2840.036Tense Arousal score26.81.524.61.70.00126.71.225.91.50.0340.199 SD Standard deviation, UMACL UWIST Mood Adjective Checklist
## The correlation coefficients of pairs in changes in SDNN scores
Figure 3 shows a plot of changes in the SDNN scores for each pair. Comparing the correlation coefficients of both groups revealed that the eating together online group had higher correlations of pairs in changes both before and during the first half of eating and before and during the second half of eating; these differed significantly compared with the eating-alone group ($$P \leq 0.005$$ and $$P \leq 0.040$$). Before the break and during the first half of eating in the eating together online group, 10 out of 14 pairs showed positive changes in the SDNN score. However, two pairs showed negative changes. Before the break and during the second half of eating in the eating together online group, 13 out of the 14 pairs showed positive changes in the SDNN score. However, one pair showed negative changes. Nevertheless, the two groups differed insignificantly before and after the breaks ($$P \leq 0.405$$).Fig. 3The correlation coefficients of pairs (r) in changes in SDNN scores. SDNN, standard deviation of the normal-to-normal interval
## Discussion
To the best of our knowledge, this is the first study to examine the effects of eating together online on reciprocal HRV using objective parameters.
The first important finding in this study is that eating together online significantly increased HRV during eating, compared with eating alone, implying that eating together online enhanced autonomic functions during eating [26, 27]. This is a favorable outcome for people who are not able to eat together, such as those who are admitted to facilities or hospitals or physically separated from their families. During the recent COVID-19 pandemic, hospitalized patients with cancer have been prohibited from receiving visitors. Consequently, more than half of them reported feeling lonely [32]; further encouragement of eating together online is desirable.
The UMACL reported that eating together online resulted in significantly higher energetic arousal postprandial than pre-snacking, indicating that vigorous arousal may affect autonomic function. Tense arousal decreased in both groups, and there was no significant difference. The mean tense arousal score, a subscale of UMACL, in this study was significantly higher than that in a previous study [25], and it is possible that significant differences between before and after the study were difficult to obtain. SDNN assesses the flexibility of the autonomic nervous system and the balance of sympathetic and parasympathetic nervous systems [16, 29]; hence, it is likely to be affected by various factors such as energetic arousal, tense arousal, and loneliness. Another study reported that significant loneliness reduces HRV [16]. Therefore, our result regarding increased HRV following eating together online may indicate alleviated loneliness associated with the presence of other people. Apparently, autonomic signals, such as electrocardiogram and electrodermal activity, were synchronized with each other because the strangers were in the same space without direct communication [33]. Future studies using self-administered questionnaires of loneliness will be useful for investigating whether visual presence in the same space, even online, reduces loneliness.
The second important finding of this study is that the pre- and intra-snack correlations of SDNN changes when eating as a pair online were statistically significantly stronger compared with those eating alone. This suggests that eating together online may positively affect HRV through social interaction with other people. Notably, during the second half of eating in the eating together online group, the SDNN changes were synchronized in all pairs, including one pair whose SDNN scores were both negative. This suggests that the physiological synchrony of eating together online may have increased as the diet progressed. Methods to assess physiological synchrony in eating together have not yet been established. Therefore, we implemented a new method to visually clarify that physiological synchrony of change in SDNN occurs in each pair only in the eating together online group.
Although the design of our study does not identify the cause of this effect on social interactions, energetic arousal may have contributed to the outcome of the UMACL.Apparently, pairing with a person with a poor relationship is more likely to increase physiological synchrony as assessed by HRV than pairing with a friend [19]. The report stated that this may be because if relationships with others were poor, participants tried to increase affinity to establish social affiliation. We believe that increasing affinity may also increase energy arousal. More than $80\%$ of the participants in this study were work colleagues; however, poor relationships may have influenced social interactions and energetic arousal.
Lastly, unlike previous studies [8, 9], there was no difference in ingestion between the two groups in our study. Previous studies state that the increased ingestion for eating together online groups is attributed to choosing and eating more types of food owing to social interactions with other people [8]. However, this study was one type of cookie only.
This study has two limitations. The first is that in the eating-alone group, the element of conversation was missing in the assessment of its effects on autonomic nervous system functioning during eating. Future studies creating an eating together online group that does not talk would be able to reduce this bias. Second, the results of the study cannot be generalized because it was conducted among relatively young, healthy volunteers who were work colleagues employed by the same institution. Sociodemographic characteristics such as being prone to eating alone, male sex, older age, and unemployment are listed [1]. Work colleagues were particularly difficult to assess because of their varying degrees of relationship. Finally, no structured interviews were conducted to screen participants for neurological or mental disorders such as depression and anxiety.
## Conclusions
The experience of eating together online increased HRV during eating. Variations in pairs were correlated and may have induced physiological synchrony.
## References
1. Yiengprugsawan V, Banwell C, Takeda W, Dixon J, Seubsman SA, Sleigh AC. **Health, happiness and eating together: what can a large Thai cohort study tell us?**. *Glob J Health Sci* (2015.0) **7** 270-277. DOI: 10.5539/gjhs.v7n4p270
2. Wang X, Shen W, Wang C, Zhang X, Xiao Y, He F. **Association between eating alone and depressive symptom in elders: a cross-sectional study**. *BMC Geriatr* (2016.0) **16** 19. DOI: 10.1186/s12877-016-0197-2
3. Tani Y, Kondo N, Noma H, Miyaguni Y, Saito M, Kondo K. **Eating alone yet living with others is associated with mortality in older men: the JAGES cohort survey**. *J Gerontol B Psychol Sci Soc Sci* (2018.0) **73** 1330-1334. DOI: 10.1093/geronb/gbw211
4. Chae W, Ju YJ, Shin J, Jang SI, Park EC. **Association between eating behaviour and diet quality: eating alone vs. eating with others**. *Nutr J.* (2018.0) **17** 117. DOI: 10.1186/s12937-018-0424-0
5. Hughes G, Bennett KM, Hetherington MM. **Old and alone: barriers to healthy eating in older men living on their own**. *Appetite* (2004.0) **43** 269-276. DOI: 10.1016/j.appet.2004.06.002
6. Kubo Y, Ishimaru T, Hino A, Nagata M, Ikegami K, Tateishi S. **A cross-sectional study of the association between frequency of telecommuting and unhealthy dietary habits among Japanese workers during the COVID-19 pandemic**. *J Occup Health.* (2021.0) **63** e12281. DOI: 10.1002/1348-9585.12281
7. Konno Y, Nagata M, Hino A, Tateishi S, Tsuji M, Ogami A. **Association between loneliness and psychological distress: a cross-sectional study among Japanese workers during the COVID-19 pandemic**. *Prev Med Rep.* (2021.0) **24** 101621. DOI: 10.1016/j.pmedr.2021.101621
8. Ruddock HK, Brunstrom JM, Vartanian LR, Higgs S. **A systematic review and meta-analysis of the social facilitation of eating**. *Am J Clin Nutr* (2019.0) **110** 842-861. DOI: 10.1093/ajcn/nqz155
9. Hetherington MM, Anderson AS, Norton GN, Newson L. **Situational effects on meal intake: a comparison of eating alone and eating with others**. *Physiol Behav* (2006.0) **88** 498-505. DOI: 10.1016/j.physbeh.2006.04.025
10. Cruwys T, Bevelander KE, Hermans RC. **Social modeling of eating: a review of when and why social influence affects food intake and choice**. *Appetite* (2015.0) **86** 3-18. DOI: 10.1016/j.appet.2014.08.035
11. Ishikawa M, Takemi Y, Yokoyama T, Kusama K, Fukuda Y, Nakaya T. **“Eating together” is associated with food behaviors and demographic factors of older Japanese people who live alone**. *J Nutr Health Aging* (2017.0) **21** 662-672. DOI: 10.1007/s12603-016-0805-z
12. Sommer W, Stürmer B, Shmuilovich O, Martin-Loeches M, Schacht A. **How about lunch? consequences of the meal context on cognition and emotion**. *PLoS One.* (2013.0) **8** e70314. DOI: 10.1371/journal.pone.0070314
13. Spence C, Mancini M, Huisman G. **Digital commensality: eating and drinking in the company of technology**. *Front Psychol* (2019.0) **10** 2252. DOI: 10.3389/fpsyg.2019.02252
14. Wang C, Peng Y, Qiu L, Wan X. **Cloud-based commensality: enjoy the company of co-diners without social facilitation of eating**. *Front Psychol.* (2021.0) **12** 758966. DOI: 10.3389/fpsyg.2021.758966
15. Hasuo H, Kanbara K, Mizuno Y, Nishiyama J, Fukunaga M, Yunoki N. **A family caregiver’s relaxation enhances the gastric motility function of the patient: a crossover study**. *Biopsychosoc Med* (2015.0) **9** 21. DOI: 10.1186/s13030-015-0048-y
16. Roddick CM, Chen FS. **Effects of chronic and state loneliness on heart rate variability in women**. *Ann Behav Med* (2021.0) **55** 460-475. DOI: 10.1093/abm/kaaa065
17. Sakuma H, Hasuo H, Fukunaga M. **Effect of handholding on heart rate variability in both patients with cancer and their family caregivers: a randomized crossover study**. *Biopsychosoc Med* (2021.0) **15** 14. DOI: 10.1186/s13030-021-00217-y
18. Palumbo RV, Marraccini ME, Weyandt LL, Wilder-Smith O, McGee HA, Liu S. **Interpersonal autonomic physiology: a systematic review of the literature**. *Pers Soc Psychol Rev* (2017.0) **21** 99-141. DOI: 10.1177/1088868316628405
19. Bizzego A, Azhari A, Campostrini N, Truzzi A, Ng LY, Gabrieli G. **Strangers, friends, and lovers show different physiological synchrony in different emotional states**. *Behav Sci (Basel)* (2019.0) **10** 11. DOI: 10.3390/bs10010011
20. Tschacher W, Meier D. **Physiological synchrony in psychotherapy sessions**. *Psychother Res* (2020.0) **30** 558-573. DOI: 10.1080/10503307.2019.1612114
21. 21.American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders 5th ed: DSM-5. Washington, DC: American Psychiatric Association; 2013.
22. Tarvainen MP, Niskanen JP, Lipponen JA, Ranta-Aho PO, Karjalainen PA. **Kubios HRV–heart rate variability analysis software**. *Comput Methods Programs Biomed* (2014.0) **113** 210-220. DOI: 10.1016/j.cmpb.2013.07.024
23. Shirasawa S, Ishida T, Hakoda Y, Haraguchi M. **The effects of energetic arousal on memory search**. *Jpn J Psychon Sci* (1999.0) **17** 93-99
24. Matthews G, Jones DM, Chamberlain AG. **Refining the measurement of mood: the UWIST mood adjective checklist**. *Br J Psychol* (1990.0) **81** 17-42. DOI: 10.1111/j.2044-8295.1990.tb02343.x
25. Matthews G, Joyner L, Gilliland K, Campbell SE, Falconer S, Huggins J. **Validation of a comprehensive stress state questionnaire: towards a state big three**. *Pers Psychol Eur* (1999.0) **7** 335-350
26. Berntson GG, Bigger JT, Eckberg DL, Grossman P, Kaufmann PG, Malik M. **Heart rate variability: origins, methods, and interpretive caveats**. *Psychophysiology* (1997.0) **34** 623-648. DOI: 10.1111/j.1469-8986.1997.tb02140.x
27. Kim HG, Cheon EJ, Bai DS, Lee YH, Koo BH. **Stress and heart rate variability: a meta-analysis and review of the literature**. *Psychiatry Investig* (2018.0) **15** 235-245. DOI: 10.30773/pi.2017.08.17
28. Strüven A, Holzapfel C, Stremmel C, Brunner S. **Obesity, nutrition and heart rate variability**. *Int J Mol Sci* (2021.0) **22** 4215. DOI: 10.3390/ijms22084215
29. Nunan D, Sandercock GR, Brodie DA. **A quantitative systematic review of normal values for short-term heart rate variability in healthy adults**. *Pacing Clin Electrophysiol* (2010.0) **33** 1407-1417. DOI: 10.1111/j.1540-8159.2010.02841.x
30. 30.Preacher KJ. Calculation for the test of the difference between two independent correlation coefficients [Computer software]. 2002. http://quantpsy.org. Accessed 30 Dec 2022.
31. Cohen J, Cohen P, West SG, Aiken LS. *Applied multiple regression/correlation analysis for the behavioral sciences* (1983.0)
32. Miaskowski C, Paul SM, Snowberg K, Abbott M, Borno HT, Chang SM. **Loneliness and symptom burden in oncology patients during the COVID-19 pandemic**. *Cancer* (2021.0) **127** 3246-3253. DOI: 10.1002/cncr.33603
33. Golland Y, Arzouan Y, Levit-Binnun N. **The mere co-presence: synchronization of autonomic signals and emotional responses across co-present individuals not engaged in direct interaction**. *PLoS One.* (2015.0) **10** e0125804. DOI: 10.1371/journal.pone.0125804
|
---
title: 'Motivating non-physician health workers to reduce the behavioral risk factors
of non-communicable diseases in the community: a field trial study'
authors:
- Mehran Asadi-Aliabadi
- Seyed M. Karimi
- Fariba Mirbaha-Hashemi
- Arash Tehrani-Banihashemi
- Leila Janani
- Ebrahim Babaee
- Marzieh Nojomi
- Maziar Moradi-Lakeh
journal: Archives of Public Health
year: 2023
pmcid: PMC9998263
doi: 10.1186/s13690-023-01047-w
license: CC BY 4.0
---
# Motivating non-physician health workers to reduce the behavioral risk factors of non-communicable diseases in the community: a field trial study
## Abstract
### Background
Non-communicable diseases behavioral risk factors can be improved if effective interventions are designed considering the health system’s capabilities and local resources. This study evaluated the effectiveness of interventions that aimed at increasing non-physician community health workers’ motivation in reducing non-communicable diseases behavioral risk factors in the community.
### Methods
A randomized field trial study was conducted in 32 community health centers in 4 Iranian districts after a baseline population survey on the status of NCDs of 30–70-year-old individuals ($$n = 1225$$). The interventions were performed to improve insufficient physical activity, insufficient fruit consumption, insufficient vegetable consumption, high salt intake, and tobacco use. Four intervention packages were implemented in 24 community health centers; the other 8 centers were used as control groups. The non-physician community health workers performed the interventions. The packages additively included goal-setting, evidence-based education, operational planning, and incentive payments. A second survey was conducted 1 year after the start of the interventions to identify the effects on an independent random sample of 30–70-year-old individuals ($$n = 1221$$). Difference-in-difference method was used to quantify the interventions’ effects.
### Results
The average age of participants in both surveys was about 49 years. Also, about half of the participants were female, and about $43\%$ were illiterate or had a primary school education. The interventions had statistically significant effects only on decreasing the prevalence of insufficient physical activity. The package with all the intervention components decreased the odds of insufficient physical activity to 0.24 ($95\%$ CI, 0.08, 0.72). The package with operational planning but no performance-based financing did not change the odds of insufficient physical activity.
### Conclusions
This study highlighted the importance of components, design, and implementation details of interventions intended to reduce NCDs behavioral risk factors. Some risk factors, such as insufficient physical activity, seem more easily modifiable with limited low-cost interventions in a one-year horizon. However, risk factors related to healthy food consumption and tobacco use need more extensive interventions.
### Trial registration
This trial was registered on the Iranian Registry of Clinical Trials (IRCT20081205001488N2) on 3 June 2018 (https://en.irct.ir/trial/774).
### Supplementary Information
The online version contains supplementary material available at 10.1186/s13690-023-01047-w.
## Background
Non-communicable diseases (NCDs) are the leading cause of death worldwide, as they accounted for $74\%$ of total deaths and $85\%$ of premature deaths in low- and middle-income countries (LMICs) in 2019 [1, 2]. In 2015, the United Nations (UN) acknowledged that the increase in the burden of NCDs is a major threat to its Sustainable Development Goals (SDGs) for the twenty-first century and targeted to reduce NDCs’ premature deaths to one-third by 2030 [3]. In Iran, NCD mortality increased from 49 to $82\%$ from 1990 to 2017 [4]. In 2017, the loss of 7.0 million years of life was attributed to NDCs, indicating a $98\%$ increase from 1990. Also, 15.0 million years were lived with disability due to NDCs, showing a $48\%$ increase from 1990 [5, 6].
Structural social changes such as rapid urbanization, globalization, and population aging have accelerated the prevalence of NCDs [7]. Changes in social structures are usually followed by lifestyle changes and increased prevalence of behavioral risk factors such as unhealthy diet, physical inactivity, tobacco smoking, and alcohol use [8]. Behavioral risk factors precede the development of metabolic risk factors (e.g., raised blood pressure, overweight/obesity, and increased blood cholesterol) and can then further develop into NCDs [9]. According to the World Health Organization (WHO), 1.6 million preventable deaths per year are caused by physical inactivity, 7.2 million by tobacco use, 4.1 million by excess salt/sodium intake, and 3.9 million by inadequate fruit and vegetable consumption [10, 11]. These estimates indicate the importance of behavioral risk factors as preventable contributors to the development of NCDs [10].
In 2019, $16.5\%$ of Iran’s deaths were attributed to an unhealthy diet, $14.1\%$ to tobacco use, $4.4\%$ to inadequate physical activity, and $1.2\%$ to excess salt intake. The corresponding attributable DALYs to the risk factors were 7.5, 9.2, 1.9, and $0.6\%$, respectively [1]. Therefore, effective investments in curtailing NCDs risk factors can save and improve lives and will be economically justifiable if at-risk people are identified in the early stages [12].
Many cost-effective intervention methods to control NCDs have been introduced worldwide. However, the main challenge is implementing them in LMICs, which typically face a limited skilled workforce, financial resources, and community participation [13, 14]. Efforts to control the increasing rate of NCDs in Iran intensified in 2014 by launching an adaptation of the WHO’s Package of Essential, IraPEN [15]. However, the mismatch of typical training provided to health care workers with the extent of NCDs in the community, insufficient documentation to determine the existing NCDs’ status, inconsistency between training and practice, and instability of financial resources have obstructed the successful implementation of the package [16, 17].
In this study, a field trial was devised and carried out to address the above shortcomings. A set of intervention packages were designed to encourage reducing NCDs behavioral risk factors and implemented by non-physician community health workers (NPHWs) in a group of randomly selected community health centers (CHCs)—locally called Health Houses—in four Iranian districts. The NPHWs were informed about the status of NCDs behavioral risk factors in their catchment areas and encouraged to meet a set of goals to reduce the risk factors. Then, they randomly received a combination of different interventions: evidence-based training, an action plan to reduce the risk factors based on the NCDs status in their catchment areas, and incentive payments based on their performance in achieving the goals. The trial was conducted to understand the effects of these interventions on NCDs risk factors in the studied population.
## Study sample
Four districts (equivalent to counties) were selected for the field trial. One of the districts was the no-intervention district, and the other three were intervention districts. In each selected district, four urban and four rural CHCs were selected. In each of the 32 selected CHCs, a baseline survey was conducted on 30–70-year-old residents of its catchment area to understand the existing status of NCDs risk factors. The survey was administered from June to September 2018 using a Persian translation of an adapted WHO stepwise approach to surveillance (STEPS) questionnaires [18]. Then, four different intervention packages were randomly assigned to the selected urban or rural CHCs in the intervention districts. In any intervention district, one urban and one rural CHC received intervention package A, one urban and one rural CHC received intervention package B, and so on (Fig. 1). The intervention period was 12 months, after which the second survey was conducted on the same age population in the 32 CHCs to assess the impacts of the interventions from September to November 2019.Fig. 1CONSORT flow diagram. Note: Analysed based on population in first and second surveys. CHCs, Community Health Centers; UCHCs, Urban Community Health Centers; RCHCs, Rural Community Health Centers. a These three universities are located in three provinces of Tehran, Semnan, and Bushehr in Iran. One of the universities’ health care systems’ key tasks is providing primary health services to the covered population. b Shahriar, Dashtestan, and Damghan. c Garmsar For each survey, a random sample of the population in the catchment area of each studied CHC was drawn. The sample size for each of the surveys was set to be 320 in each district; it was stratified by urban/rural CHCs, sex and age groups (30–39, 40–49, 50–59 and 60–69 years); 4 urban and 4 rural CHCs from each district were selected (totally, 12 urban and 12 rural in the intervention districts, in addition to the 4 urban and 4 rural in the “no intervention” district). So, the planned total sample size for each round of survey was 1280 [(12 + 12 + 4 + 4) * 40]. If we look at this number based on the intervention packages, the sample size was 240 for each of the intervention packages 1 to 4, and 320 for the no-intervention group [(4* 240) + 320 = 1280]. The sample size of 240 for each intervention group was enough to find a one-third decrease in the prevalence of $34\%$ (based on a primary estimate for prevalence of physical inactivity) with an Alpha of 0.05 and a power of $80\%$. We used a higher sample size for the no-intervention group (320 versus 240) both to increase the power and to use a similar sampling protocole in all districts. As explained in the CONSORT flow diagram (Fig. 1), the final sample size in the 1st and 2nd surveys were 1226 and 1221 individuals, respectively. Since the cluster sample sizes were not proportional to the catchment areas’ population, sampling weights were used in all analyses. All interviews were in-person. If a household was not available at the first reach, it was contacted by the research team up to three times on three successive days to perform the interview. In each household, only one male and one female member from each of the following age group strata were interviewed: 30–39, 40–49, 50–59, and 60–70 years. If more than one male or female from an age group were living in a household, one of them was randomly selected and interviewed. A necessary inclusion criterion was informed consent by interviewees.
While the sampling method was similar in both surveys, the selected participants were not necessarily the same. Although the survey participants were expected to be among the target groups of the interventions, they were not necessarily the ones who received the intervention directly. According to the country’s Integrated Health Record System—called the SIB system locally [19]—on average, $20\%$ of the catchment area population visit their local CHC in any given quarter. The country’s NPHWs’ reach, however, was expected to be much more than CHC’s direct utilization rate because, in addition to those who refer to them, they are responsible for improving the health of the entire population assigned to them. Based on how CHCs are structured and organized in the country, if individuals in a catchment area do not visit CHCs and demand health care, the NPHWs must reach out to them, encourage visits, or at least monitor their health remotely over the phone to make sure they are ceceiving health care somewhere else (such as the provate sector). This active follow-up method, especially in rural areas, has led to significant health improvements in the country, for example, remarkable decreases in maternal and childhood mortality and communicable diseases over several decades (Barzegar and Djazayeri 1981 [20], Rahbar and Ahmadi 2015 [21], and Keshvari et al. 2016 [22]). This study was built on the same infrastructures and intended to extend the NPHWs’ experiences regarding maternal and child health and communicable disease to non-communicable diseases.
Based on the approved protocol, the trial was planned to continue for 24 months, with a third survey at the end of the study. However, it was terminated prematurely after 12 months to comply with the country’s COVID-19 social distancing protocols.
The selected districts were Shahriar (population = 744,210), Dashtestan (population = 252,047), Damghan (population = 94,190), and Garmsar (population = 77,421) (Additional file 1). The districts’ populations are based on the country’s 2016 census [23]. A simple randomization method was used to select four urban and four rural CHCs in each district and to assign the intervention packages to the selected CHCs. Detailed explanations of the inclusion criteria for the districts, CHCs, and participants were explained in the protocol [24]. NPHWs implemented four intervention packages after receiving extensive training. Physicians were not the target group of this trial because they were undergoing a separate incentive payment scheme [25]. The CONSORT checklist can be found in Additional file 2 [26].
## The interventions and intervention packages
An intervention package in this study included the first, the first two, the first three, or all of the following interventions:The first intervention (target-setting): Short-term targets (e.g., decrease in tobacco use and salt consumption) were set based on the preliminary results obtained from the baseline survey and the national goals to control NCDs behavioral risk factors. The national goals were to reduce insufficient physical activity, insufficient fruit and vegetable, salt intake, and tobacco use by 20, 30, 30, and $30\%$, respectively, until 2025 [27, 28]. Specific quarterly and yearly targets for NCDs risk factors are reported in Additional file 3. Meetings were held with the NPHWs of the selected CHCs, and they were informed about the status of NCDs behavioral risk factors in their catchment area population and the national goals to reduce them. Also, the proposed targets were presented to them, and they were encouraged to work through achieving them. The research team did not go beyond providing information on the status of NCDs in the catchment areas and national NDC goals, setting the goals mentioned above, and encouraging the NPHWs to achieve them. The second intervention (evidence-based education): The research team set up a 16-hour workshop for the NPHWs’ of the CHCs that received this intervention. The workshops were merely informational, during which the adverse health effects of overconsumption of salt, underconsumption of fruit and vegetable, insufficient physical activity, and tobacco use were extensively discussed. The trainees were provided with the related informational brochures as well. In addition, the team used Disease Control Priorities, 3rd edition [28, 29], and the Iranian version of the WHO package of essential NCDs (PEN) [30] to prepare a review summary of the effectiveness of the interventions aimed at decreasing the risk of NCDs in LMICs. The review also included success stories in other countries and methods of selecting, planning, and implementing cost-effective interventions. For the review. The third intervention (operational planning): The research team coordinated with NPHWs and the local health experts to collaboratively devise operational plans for the selected CHCs in a 12-hour workshop. The major component of the NPHWs’ action plans was periodic (biweekly or monthly) educational sessions for the covered population on the causes and detrimental health effects of NDCs and practical methods to decrease the risk of NDCs (such as increasing physical activities, adjusting the diet, reading food products’ nutrition label, decreasing the consumption of salt, canned and fast food, and the use of tobacco products). Healthy lifestyle (especially in regards to movement and diet) and smoking reduction counseling was also offered to the CHC visitors. Other action plan items were organizing weekly public walking events, setting up group activities (such as painting, reading, and board games), and coordinating with government-owned sports facilities to provide free hours to the public (specifically, three two-hour sessions a week). The focus of the action plan was different from one CHC to another based on the finding of the baseline survey at the CHC’s catchment area. For example, if the body mass index was particularly high in a catchment area, more frequent educational sessions were set on the risk of obesity and the importance of physical activity, and more public walking events were organized by the NPHWs. To support the action plans’ execution, the team allocated a supportive budget for the devised operational plans. The maximum supportive budget was 60 million Rials—equivalent to 556 United States dollars, USD, based on the exchange rate of 107,832 Rial/USD at the time of study [31]. The budget could be spent on purchasing equipment for the CHC (for example, digital blood pressure sphygmomanometers, body weight scales, and height measuring devices) and materials for educational sessions, group sports, and non-sport events. The fourth intervention (performance-based financing or PBF): NPHWs of the selected CHCs received incentive payments. The payments were calculated at the CHC level, and paid to all NPHWs of that CHC per the pre-defined targets every 3 months. The average level of achievement of each center to its 3-month targets for eight different NCD behavioral and metabolic risk factors was quantified. The CHCs were then classified based on the percentage of their achievements into one of the following four groups: < 25, 25–49.99, 50–62.49, and $62.5\%$ or more. These groups, respectively, received no incentive, one-third, two-thirds, and full incentive. The full monthly incentive was $10\%$ of the average monthly salary of a typical NPHW in the studied districts, which was determined to be approximately 25 million Rial (or 232 USD). Therefore, the maximum monthly incentive payment was approximately 23 USD. No payment was delayed because they were made directly to the NPHWs’ bank account immediately after each assessment.
Intervention package A included only the first intervention, goal setting. Intervention package B included the first two interventions: goal setting and evidence-based training. Intervention package C included the first two interventions plus an action plan. In intervention package D, PBF was added to other interventions (Table 1). CHCs that received the intervention packages A, B, C, and D were also called intervention groups A, B, C, and D, respectively, in this study. The no-intervention district (Garmsar district) received neither of the interventions. Table 1Assignment of interventions to CHCs inside each of the three treatment districtsIntervention PackageUrban/Rural SeparationIntervention:Target-SettingEvidence-Based EducationOperational PlanningPerformance-Based FinancingA1 Rural, 1 UrbanYesNoNoNoB1 Rural, 1 UrbanYesYesNoNoC1 Rural, 1 UrbanYesYesYesNoD1 Rural, 1 UrbanYesYesYesYes Every two to four weeks, the implementation status of the interventions was reviewed and checked by the district and province supervisors selected by the research team. Also, reports on the interventions were received by the research team every quarter. The reports contained performance reviews per the set goals. Moreover, they included detailed accounts of the activities conducted at the CHCs that received an action plan. The key components of the reports were activities’ dates and type of the activity and the number of participants.
## Statistical analysis
This study’s objective was to compare NCDs behavioral risk factors before and after the interventions and identify effective interventions. NCDs behavioral risk factors analyzed in this study were zero-one indicators of insufficient physical activity, insufficient fruit consumption, insufficient vegetable consumption, high salt intake, and tobacco use.
Not meeting the WHO recommendations on physical activity (Metabolic Equivalent of Task, MET, less than 600 METs per week) was defined as insufficient physical activity [32]. The WHO’s recommendations were used to determine insufficient fruit (less than two medium-sized fruits, such as two medium apples or half a cup of nuts, in the last 24 hours) and vegetable consumption (less than three cups of raw leafy vegetables or one and a half cups of cooked or chopped vegetables in the last 24 hours) [33–35]. A person was identified as a high salt consumer if the person always or often added salt or salt additives to the food [36]. Current tobacco smoking was defined as the use of any tobacco products, including cigarettes, cigars, or pipes, on a daily or non-daily basis in the last 30 days [37].
The prevalence of each NCDs risk factor in the baseline and second surveys was calculated in populations assigned with each intervention package. Then, the difference in the prevalence rates between the two surveys was calculated. For the more formal analysis of the effect of the designed intervention packages, the difference-in-difference (DID) design was employed. The following equation shows the linear specification of the DID design: 1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${Y}_{ic t}=\alpha +\beta {IntPackage}_{ic}+\gamma {Post}_{it}+\rho \left({IntPackage}_{ic}\times {Post}_{it}\right)+\theta {CHC}_{it}+\delta {X}_{ic t}+{\varepsilon}_{ic t}$$\end{document}Yict=α+βIntPackageic+γPostit+ρIntPackageic×Postit+θCHCit+δXict+εictwhere i indicates a surveyed individual, c indicates the community health center to which the individual is affiliated, and t indicates the survey year. The dependent variable, Y, is a binary variable that indicates one of the NCDs behavioral risk factors for the individual. The variable IntPackage is a categorical variable with five values (0, 1, 2, 3, and 4), indicating the intervention package assigned to the CHC where the individual receives health services (Table 1). The value 0 was assigned to individuals surveyed in the no-intervention district. Therefore, four β s were estimated. Estimations of βA, βB, βC, and βD provide an adjusted comparison of the average level of dependent variable at the catchment areas that received intervention packages A, B, C, and D, respectively, to that in the non-intervention district. The variable Post indicates the survey year. It takes the value 0 if the individual was surveyed before the implementation of the interventions [2018] and 1 if surveyed after the trial [2019]. The variable CHC is a community health center indicator (1, 2, …, 32), as any surveyed individual is affiliated with a specific CHC. This variable accounts for the influence of all unobservable/unmeasurable CHC-specific confounders that might not change over the study period (e.g., health care resources in the community, attitudes towards using modern medicine versus traditional practices, average distance from the CHCs, and overall weather patterns). The variable X is a vector of socioeconomic factors including age, sex, marital status (in three categories: never married, married, divorced or widowed), the level of education (in four categories: illiterate or primary, secondary, high school, and some college), labor market status (in six categories: public wage and salary job, private wage and salary job, self-employed, homemaker, retired, and unemployed), health insurance status, and homeownership status. The coefficients of interest in this specification are ρs (i.e., ρA, ρB, ρC, and ρD) which show the effect of the intervention packages versus no intervention among those surveyed after the intervention.
Given the binary nature of the outcome variables in this study, logistic models were used in fitting Eq. 1. Odds ratios were calculated, representing the change in the odds of the dependent variable being equal to 1 due to one unit change in either of the terms on the right-hand side of Eq. 1. Equation 2 is the logistic transformation of Eq. 1.
2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathit{\ln}\left(\frac{p_{Y_{ic t}}}{1-{p}_{Y_{ic t}}}\right)=\alpha +\beta {IntPackage}_{ic}+\gamma {Post}_{it}+\rho \left({IntPackage}_{ic}\times {Post}_{it}\right)+\theta {CHC}_{it}+\delta {X}_{ic t}+{\epsilon}_{ic t}$$\end{document}lnpYict1-pYict=α+βIntPackageic+γPostit+ρIntPackageic×Postit+θCHCit+δXict+ϵictwhere \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${p}_{Y_{ict}}$$\end{document}pYict is the probability of the dependent variable Yict being reported as 1.
To account for the possibility that the NCDs behavioral risk factors (the Ys) may not be independently distributed within the population covered by each community health center, hence the estimated standard error be artificially low, standard errors were clustered at the CHC level [38]. Also, a sampling weight was assigned to the surveyed individuals. Sampling weights were calculated as the inverse of the ratio of sampled individuals characterized by sex and age group relative to the number of individuals in the respective sex and age group (namely, 30–39, 40–49, 50–59, and 60–70) in the population. For a specific individual, the ratio was the multiplication of two shares: [1] the share of surveyed individuals of the same sex and age group in the corresponding catchment area’s population, and [2] the share of individuals of the same sex and age group in the corresponding district in the country’s population. The shares were separately calculated for urban and rural areas.
The estimations were done with and without adjusting for the socioeconomic factors (X) to assess the extent of any potential observable bias in the selection of CHCs and the assignment of intervention packages. The statistical package used for analyses was STATA 14.0 (Stata, Inc., College Station, Texas).
## Ethical issues/statement
This study has been approved by the national committee on ethics in medical research (code: IR.NIMAD.REC.1396.084) as well as our institutional review board (code: IR.IUMS.REC.1395.1057613). Written informed consent has been obtained from study participants.
## Results
A total of 2446 people aged 30–70 years participated in the two surveys, 1225 individuals in the first and 1221 in the second survey. The socioeconomic characteristics of survey participants in both rounds were strikingly similar. For example, the mean age and sex ratio of the participants in the two surveys were very similar (about 49 years and about $50\%$ women). Approximately $43\%$ had illiterate or primary school education, while about $11\%$ had a college education. About $5\%$ were employed, 24–$27\%$ were self-employed, 93–$95\%$ were insured, and 84–$85\%$ were homeowners (Table 2).Table 2Demographic and economic characteristics of participants in each group in the two surveysSocioeconomic FactorsTotaln = 2446Intervention Package/GroupAn = 445Bn = 463Cn = 465Dn = 450Nonen = 623Survey 1n = 1225Survey 2n = 1221Survey 1n = 225Survey 2n = 220Survey 1n = 232Survey 2n = 231Survey 1n = 239Survey 2n = 226Survey 1n = 224Survey 2n = 226Survey1n = 305Survey 2n = 318Mean Age (Standard Deviation)49.3 (0.33)49.4 (0.32)49.5 (0.76)49.6 (0.74)49.3 (0.73)49.3 (0.76)49.9 (0.73)49.1 (0.74)48.9 (0.78)49.1 (0.73)48.8 (0.65)49.7 (0.63)p-value0.8310.968Sex: Male (%)611 (49.9)606 (49.6)112 (49.8)107 (48.6)117 (50.4)117 (50.6)113 (47.3)107 (47.4)110 (49.1)117 (51.8)159 (52.1)158 (49.7) Female (%)614 (50.1)615 (50.4)113 (50.2)113 (51.4)115 (49.6)114 (49.4)126 (52.7)119 (52.6)114 (50.9)109 (48.2)146 (47.9)160 (50.3)p-value0.8530.899Education: Illiterate or Primary School (%)497 (43.5)517 (42.5)105 (50.0)122 (55.5)109 (49.1)116 (50.4)117 (52.5)121 (53.8)88 (47.3)83 (36.7)78 (25.8)75 (23.7) Secondary School (%)158 (13.8)196 (16.1)19 (9.0)35 (15.9)39 (17.6)39 (17.0)20 (9.0)30 (13.3)27 (14.5)44 (19.5)53 (17.6)48 (15.1) High School (%)365 (31.9)368 (30.2)65 (31.0)43 (19.6)58 (26.1)64 (27.8)68 (30.5)52 (23.1)45 (24.2)70 (31.0)129 (42.7)139 (43.9) Some College (%)123 (10.8)137 (11.2)21 (10.0)2 (9.0)16 (7.2)11 (4.8)18 (8.0)22 (9.8)26 (14.0)29 (12.8)42 (13.9)55 (17.3)p-value0.001>0.001>Marital Status: Never Married (%)82 (7.0)84 (6.9)12 (5.7)7 (3.2)9 (4.0)13 (5.7)18 (8.1)12 (5.4)19 (9.0)22 (9.7)24 (7.9)30 (9.4) Married (%)996 (85.4)1039 (85.4)187 (89.5)198 (90.4)190 (86.0)194 (84.7)187 (84.2)198 (88.4)176 (83.0)185 (81.9)256 (84.8)264 (83.0) Divorced/Widowed (%)88 (7.6)93 (7.7)10 (4.8)14 (6.4)22 (10.0)22 (9.6)17 (7.7)14 (6.2)17 (8.0)19 (8.4)22 (7.3)24 (7.6)p-value0.4560.581Job: Public Wage and Salary (%)101 (8.7)75 (6.2)13 (6.3)8 (3.6)18 (8.1)14 (6.1)16 (7.2)15 (6.8)20 (9.4)13 (5.8)34 (11.4)25 (7.9) Private Wage and Salary (%)104 (9.0)92 (7.6)21 (10.1)31 (14.0)21 (9.5)15 (6.5)28 (12.7)21 (9.6)12 (5.7)18 (8.1)22 (7.4)7 (2.2) Self-Employed (%)316 (27.2)292 (24.1)53 (25.5)46 (20.8)54 (24.3)56 (24.6)32 (14.5)49 (22.3)86 (40.6)58 (26.0)91 (30.5)83 (26.1) Homemaker (%)491 (42.3)563 (46.5)89 (42.8)103 (47.1)98 (44.1)110 (47.8)110 (49.8)105 (47.7)73 (34.4)102 (45.7)121 (40.6)143 (45.0) Retired (%)91 (7.8)132 (10.9)20 (9.6)24 (10.9)19 (8.6)24 (10.4)22 (10.0)19 (8.6)12 (5.7)25 (11.2)18 (6.0)40 (12.6) Unemployed (%)58 (5.0)57 (4.7)12 (5.8)8 (3.6)12 (5.4)11 (4.8)13 (5.9)11 (5.0)9 (4.25)7 (3.1)12 (4.0)20 (6.2)p-value0.0310.307Health Insurance: Insured (%)1035 (92.6)1152 (94.7)203 (96.2)209 (95.4)209 (94.6)205 (90.3)171 (96.1)210 (92.9)164 (78.8)215 (95.1)288 (96.0)313 (98.4) Uninsured (%)83 (7.4)64 (5.3)8 (3.8)10 (4.6)12 (5.4)22 (9.7)7 (3.9)16 (7.1)44 (21.2)11 (4.9)12 (4.0)5 (1.6)p-value0.001>0.001Homeownership: Yes (%)898 (83.8)1027 (85.1)183 (87.6)171 (79.2)170 (86.3)209 (92.9)164 (92.7)192 (86.1)135 (70.7)175 (77.4)246 (82.8)280 (88.3) No (%)173 (16.2)180 (14.9)26 (12.4)45 (20.8)27 (13.7)16 (7.1)13 (7.3)31 (13.9)56 (29.3)51 (22.6)51 (17.2)37 (11.7)p-value0.001>0.001> There was no significant difference in the average age and sex of survey participants from catchment areas of no-intervention and intervention districts. However, the share of participants with primary or no education in the non-intervention district (25.8 and $23.7\%$ in the first and second surveys) was much lower than that in the intervention districts (between 36.5 and $55.5\%$). Consequently, there were more participants with a high school or a college education in the non-intervention than in intervention districts. In addition, the share of high school-educated participants who were potentially exposed to intervention package A decreased from $31.0\%$ in survey 1 to $19.6\%$ in survey 2, an $11.4\%$ decrease; the share of illiterate or primary school-educated participants who were potentially exposed to intervention package D decreased from $47.3\%$ in survey 1 to $36.7\%$ in survey 2, an $11.6\%$ decrease. Most surveyed individuals were married, regardless of the intervention packages assigned to their pertinent CHCs: the share of married participants was between 81.9 and $90.4\%$ across intervention groups and surveys. Also, the majority of participants were either homemakers (between 34.4 and $49.8\%$) or self-employed (between 14.5 and $40.6\%$), had health insurance (between 78.8 and $98.4\%$), and were homeowners (between 70.7 and $92.9\%$) (Table 2). Given the observed variations in the characteristics, they were used in adjusting the estimated effects of the intervention packages.
The crude comparison of the levels of NCDs behavioral risk factors in each intervention group before and after the interventions showed the largest decrease in insufficient physical activity in intervention group D with a $29\%$ decrease ($95\%$ Confidence Interval: $20\%$, $38\%$), then in intervention groups B and A with $25\%$ ($95\%$ CI: $17\%$, $34\%$) and $15\%$ ($95\%$ CI: $6\%$, $24\%$) decreases, respectively. The observed level of physical activity did not change in the control group during the study period.
A large and consistent increase in fruit consumption was also observed in all intervention and non-intervention groups. We suspect the increase in fruit consumption was largely attributable to the strikingly lower relative prices for fruits during the study period (Additional file 4). No consistent and statistically significant pattern of increase or decrease in fruit and vegetable consumption, high salt intake, and tobacco use was observed in the studied groups (Table 3).Table 3The difference in NCDs behavioral risk factors between the two surveysNCDs Risk FactorsIntervention *Package a* / GroupFirst Survey (%)Second Survey (%)Difference (%) ($95\%$ Confidence Interval)Insufficient Physical ActivityA4025−15 (−24, −6)B4924−25 (−34, −17)C4140−1 (−11, 8)D5324−29 (−38, −20)None39401 (−7, 8)Insufficient Fruit ConsumptionA6657−9 (−19, 0)B5953−6 (−1, 3)C6851−17 (−27, −8)D7259−13 (−22, −4)None7356−17 (−25, −10)Insufficient Vegetable ConsumptionA32342 (−6, 11)B33396 (− 2, 15)C3736− 1 (− 11, 9)D40477 (− 2, 16)None3837− 1 (−9, 7)High Salt IntakeA1513−2 (−9, 5)B19223 (−4, 11)C51510 [4, 16]D19223 (−5, 11)None1513−2 (− 8, 3)Current Tobacco UseA16204 (− 3, 11)B2019− 1 (− 8, 7)C16226 (− 2, 12)D14140.3 (− 6, 7)None1811−7 (− 13, 1)NCDs Non-communicable diseasesa Intervention package A included target-setting. Intervention package B included A plus evidence-based education. Intervention package C included B plus operational planning. Intervention package D included plus performance-based financing The results of the statistical analyses largely confirmed the results of the crude comparisons. Using the non-intervention district as the reference group and employing a DID research design (Eq. 2), no improvement in fruit and vegetable consumption, high salt intake, and tobacco use was estimated. However, fairly consistent improvements in the level of physical activity in all intervention groups were observed compared to the non-intervention group. Among the intervention packages, B and D resulted in decreasing insufficient physical activity. Specifically, the unadjusted estimations showed that the odds of reporting insufficient physical activity decreased to 0.32 ($95\%$ CI: 0.11, 0.88) among the surveyed individuals covered by CHCs that received the intervention package B, and to 0.28 ($95\%$ CI: 0.10, 0.75) among those covered by CHCs that received the intervention package D. These results were confirmed after adjusting for socioeconomic factors: 0.27 ($95\%$ CI: 0.09, 0.85) decrease in the odds of insufficient physical activity in intervention group B, 0.24 ($95\%$ CI: 0.08, 0.72) in intervention group D (Table 4). In other words, the package with all intervention components (which added the provision of an action plan and incentive payments to the previous two) decreased the likelihood of insufficient physical activity by $76\%$ ($95\%$ CI: $28\%$, $92\%$).Table 4The estimated effects of intervention packages on the NCDs behavioral risk factorsNCDs Risk FactorIntervention Package aUnadjustedAdjusted for Socioeconomic FactorsOdds Ratio ($95\%$ CI)p-valueOdds Ratio ($95\%$ CI)p-valueInsufficient Physical ActivityA0.49 (0.20, 1.22)0.120.56 (0.21, 1.51)0.25B0.32 (0.11, 0.88)0.020.27 (0.09, 0.85)0.02C0.91 (0.25, 3.40)0.890.81 (0.17, 3.76)0.78D0.28 (0.10, 0.75)0.010.24 (0.08, 0.72)0.01NoneReference GroupInsufficient Fruit ConsumptionA1.44 (0.35, 5.95)0.611.52 (0.27, 8.62)0.63B1.70 (0.45, 6.35)0.431.94 (0.38, 9.84)0.42C1.04 (0.23, 4.72)0.950.99 (0.13, 7.32)0.99D1.19 (0.35, 4.03)0.781.42 (0.31, 6.61)0.64NoneReference GroupInsufficient Vegetable ConsumptionA1.17 (0.37, 3.71)0.791.13 (0.26, 4.89)0.86B1.39 (0.53, 3.62)0.511.71 (0.64, 4.54)0.28C1.00 (0.34, 2.91)0.990.70 (0.19, 2.62)0.59D1.39 (0.44, 4.42)0.571.71 (0.46, 6.43)0.42NoneReference GroupHigh Salt IntakeA1.01 (0.33, 3.11)0.981.04 (0.32, 3.38)0.94B1.45 (0.51, 4.12)0.481.23 (0.41, 3.70)0.71C3.92 (1.05, 14.63)0.044.35 (0.80, 23.47)0.08D1.43 (0.37, 5.56)0.611.38 (0.36, 5.33)0.64NoneReference GroupCurrent Tobacco UseA2.37 (1.09, 5.18)0.031.87 (0.83, 4.23)0.13B1.73 (0.65, 4.59)0.271.47 (0.50, 4.29)0.48C2.54 (0.84, 7.66)0.091.55 (0.56, 4.29)0.40D1.82 (0.81, 4.12)0.151.24 (0.44, 3.54)0.68NoneReference GroupBoldface indicates statistical significance ($p \leq 0.05$)NCDs Non-communicable diseasesa Intervention package A included target-setting. Intervention package B included A plus evidence-based education. Intervention package C included B plus operational planning. Intervention package D included plus performance-based financing
## Discussion
The frontline CHCs in rural areas were piloted in the 1970s and expanded rapidly to a nationwide network in the early 1980s. Each CHC has at least one female and one male community health worker. Most of them are NPHWs, usually selected from local people and trained in a specific practical program designed by the country’s Ministry of Health. Their focus has been providing primary health care services, including health education, promotion of food supply and proper nutrition, advocacy for and monitoring of water safety and basic sanitation, maternal and child health, immunization according to the national immunization program, prevention and control of locally important diseases, treatment of common diseases and injuries, and provision of essential medications [20]. In the past two decades, they have been more involved in managing NCDs risk factors, and there is indirect evidence of the success of their role in this area [39]). The functions of the urban CHCs are close to their rural counterparts. Each urban CHC has at least one NPHW for every 2500 individuals in the catchment area. The average coverage of rural and urban CHCs is currently 1200 and 12,500 individuals, respectively. In either case, urban or rural, people covered by a CHC cannot use the services of other CHCs. In rural areas, especially in remote areas, utilization of CHC services is usually high. In urban areas and especially in large cities, there are several alternatives, usually more specialized, in public and private sectors for all kinds of health services but with a higher out-of-pocket cost. Also, active follow-up of the target population for health services in urban centers is not as usual and orderly as in rural centers [40].
This trial used the aforementioned community health foundation to examine the effectiveness of four interventional packages designed to improve four major NCDs behavioral risk factors, namely, insufficient physical activity, insufficient fruit and vegetable consumption, high salt consumption, and current tobacco use. NPHWs implemented the interventions at a randomly selected number of CHCs in four Iranian districts. The most basic intervention package included target-setting for NPHWs. Evidence-based education, operational planning, and PBF for NPHWs were added to target-setting in other packages. Improvements were observed in only one NCDs behavioral risk factor, insufficient physical activity. The most effective intervention package included all four interventions. In this study, incentives were paid to enhance the achievement of public health goals. The finding is in line with the result of previous studies that have shown increasing the motivation of NPHWs can pave the way for achieving predetermined goals in areas where access to physicians is limited [41, 42].
Our findings accord with some review studies which shown that interventions in primary health care can significantly improve physical activity compared to other NCDs behavioral risk factors, even 6 to 12 months after intervention [43, 44]. A randomized controlled trial (RCT) in the United Kingdom showed that motivational interviewing of primary care patients by NPHWs improved physical activity levels assessed 12 months after the intervention [45]. Other RCTs also showed interventions with multiple lifestyle improvement components can increase physical activity [46–48].
This study’s measured improvements in physical activity can be attributed to the provision of regular physical activity programs by NPHWs to the CHC visitors, NPHWs’ active involvement in physical activity initiatives, and encouraging local councils to waive recreational areas’ and sports facilities’ usage fees for group activities supported by CHCs. The finding that the addition of PBF to other interventions made the results stronger indicates the importance of incentive payments and their role in promoting NPHWs’ efforts. A similar study in San Diego, California, USA, included direct involvement of health workers in public exercises and group walking over 12 months and measured a noticeable increase in physical activity in the community [49, 50]. Other studies have also shown the effectiveness of community-based physical activity interventions [51, 52].
An irregularity was found in the results for physical activity. Adding more components to the intervention package was expected to make it more effective. Accordingly, adding evidence-based training to goal-setting made the effect larger and statistically significant. However, adding an action plan to the intervention package without PBF eradicated the effect of goal-setting and evidence-based training. Adding PBF to the package that included goal-setting, evidence-based training, and an action plan, on the other hand, significantly increased the magnitude of the effect. Therefore, an action plan without incentive payment appeared to be ineffective. This irregularity may be an important result of this study. Anecdotal evidence collected by the authors showed that some NPHWs of CHCs who received the intervention package with and without incentive payments (packages C and D) might be communicating during the study period, although the research team set up workshops for different interventions at different weeks and advised district health care authorities to keep the intervention assignments to the CHCs in their district confidential. Hence, no incentive payment to NPHWs who implemented intervention package C might have acted as a disincentive, perhaps because they were expected to do the same amount of work as NPHWs who implemented intervention package D.
Comparing the effects on physical activity of intervention packages B and D, we found no discernable difference: the adjusted odds ratios for intervention packages B and D were 0.27 ($95\%$ CI: 0.09, 0.85) and 0.24 ($95\%$ CI: 0.08, 0.72), respectively. This is another unexpected result, indicating that a package with embedded financial incentives (D) did not improve the results over a package without the incentives (B). This finding highlights the importance of providing evidence-based training to the participating NPHWs.
The goal of increasing fruit and vegetable consumption may be achievable by spending more time and holding counseling sessions, as other studies have shown [53, 54]. On the other hand, failure to achieve this goal might be due to the high inflation rate in the country during the study period, especially the sharp increase in food, fruit, and vegetable prices because of a new round of international economic sanctions on the country from 2018 [55]. Virtually, the average cost of food in 2019 was estimated to be 3.6 times more than that in 2017 [56]. Factors influencing failure to achieve the desired effect of interventions in the consumption of fruit and vegetable have been reported before. Poor nutritional knowledge [57], the role of media in advertising food with low nutritional value [58], and the high relative price of fruits [59] are posed as the main factors.
Measuring no improvement in reducing high salt intake in this study is consistent with the contemporary consumption habits in the country and the challenges of changing them. For example, a WHO report indicated that in 2012, salt consumption in Iran was twice the recommended level [60, 61]. Evidence has shown that changing behavior and modifying consumption patterns require comprehensive and extensive programs [62]. Other studies have shown that changing the habit of salt consumption is less probable only with educational and information campaigns [63, 64]. One potential challenge in reducing salt intake in rural Iran may be the presence of traditional healers who disseminate rumors mixed with religious stories about the benefits of using salt, salt stone, and sea salt, even for people with hypertension [65].
No measured effect of this study’s interventions on tobacco use was expected as influencing tobacco use habits faces easy access and low price challenges. The smoking economics literature shows that policies that increase the price of tobacco products (through increased sales taxes, for example) are most influential in decreasing tobacco use [66–70]. Such macro-level policies were out of this study’s scope.
This study has several limitations. First, only short-term effects were measured because individuals in the intervention CHCs were not followed up. The trial was initially designed for 24 months [24]. Nonetheless, in compliance with limitations imposed by the emergence of the COVID-19 pandemic and the required social distancing policies, the study was stopped prematurely. Similar studies with longer periods of intervention might be necessary to measure longer-term effects on NCDs behavioral risk factors. Second, the survey participants were not necessarily the ones who were directly affected by the interventions. Although the influence of a typical CHC’s activities in the country is expected to go beyond direct visits, the measured effects may only reflect the lower bounds of the actual effects on the treated individuals. Third, participating NPHWs’ relocation to serve in other CHCs was out of the researchers’ control. Nonetheless, no relocation took place between the selected intervention CHCs. The research team traced all relocations monthly and provided the replacing NPHWs with the same training. Fourth, implementing simultaneous national research projects, such as the High Blood Pressure Campaign, was not under the researchers’ control [71]. However, such factors are expected to affect all intervention and non-intervention groups similarly. The fifth is a general limitation of studies that target NPHWs. Although we used training workshops to mprove NPHWs’ knowledge about non-communicable diseases, and developing their skills in applying the trial’s designed interventions, we do not believe these efforts would replace formal academic or occupational training or employing more skilled health workers. This might be one of the reasons for not measuring an effect on most behavioral risk factors in this study.
## Conclusion
Paying incentives to NPHWs, along with other interventions, could be considered a useful means of improving physical activity in the community. This study could not be continued because of COVID-19. Longer studies are needed to identify the long-term effects of such interventions.
## Supplementary Information
Additional file 1. Districts of intervention and no intervention in IRPONT study. Additional file 2. CONSORT checklist. Additional file 3. Specific quarterly and yearly targets for NCDs risk factors. Additional file 4. Supplementary analysis of fruits/ vegetables consumption Additional file 5. IRPONT collaborators.
## References
1. 1.GBD. Global burden of disease study 2019: Institute for Health Metrics and Evaluation (IHME), Seattle, WA, United States; 2020. [Available from: http://ghdx.healthdata.org/gbd-results-tool]
2. 2.WHOGuidance note on the integration of noncommunicable diseases into the united nations development assistance framework2019World Health Organisation, Geneva, Switzerland. *Guidance note on the integration of noncommunicable diseases into the united nations development assistance framework* (2019.0)
3. 3.WHOWorld health statistics 2016: monitoring health for the SDGs sustainable development goals2016World Health Organization, Geneva, Switzerland. *World health statistics 2016: monitoring health for the SDGs sustainable development goals* (2016.0)
4. 4.GBDGlobal burden of disease study 20192019Institute for Health Metrics and Evaluation (IHME), Seattle, WA, United States. *Global burden of disease study 2019* (2019.0)
5. 5.GBDGlobal burden of diseases 2017: Institute for Health Metrics and Evaluation2019University of Washington, Seattle, WA, United States. *Global burden of diseases 2017: Institute for Health Metrics and Evaluation* (2019.0)
6. Kyu HH, Abate D, Abate KH, Abay SM, Abbafati C, Abbasi N. **Global, regional, and national disability-adjusted life-years (DALYs) for 359 diseases and injuries and healthy life expectancy (HALE) for 195 countries and territories, 1990–2017: a systematic analysis for the Global Burden of Disease Study 2017**. *Lancet* (2018.0) **392** 1859-1922. DOI: 10.1016/S0140-6736(18)32335-3
7. Khorrami Z, Etemad K, Yarahmadi S, Khodakarim S, Kameli M, Hezaveh AM. **Urbanization and noncommunicable disease (NCD) risk factors: WHO STEPwise Iranian NCD risk factors surveillance in 2011**. *East Mediterr Health J* (2017.0) **23** 469. DOI: 10.26719/2017.23.7.469
8. 8.WHOGlobal status report on noncommunicable diseases 20102011GenevaWorld Health Organization Regional Office for the Western. *Global status report on noncommunicable diseases 2010* (2011.0)
9. 9.WHOWHO STEPS surveillance manual: STEPwise approach to non communicable disease risk factors surveillance2017GenevaWorld Health Organization. *WHO STEPS surveillance manual: STEPwise approach to non communicable disease risk factors surveillance* (2017.0)
10. 10.WHONon communicable diseases in World Health Organization Geneva2018. *Non communicable diseases in World Health Organization Geneva* (2018.0)
11. 11.WHOIncreasing fruit and vegetable consumption to reduce the risk of noncommunicable diseases2019World Health Organization, Geneva, Switzerland. *Increasing fruit and vegetable consumption to reduce the risk of noncommunicable diseases* (2019.0)
12. 12.WHOManagement of noncommunicable diseases2019World Health Organization, Geneva, Switzerland. *Management of noncommunicable diseases* (2019.0)
13. 13.WHOA prioritized research agenda for prevention and control of noncommunicable diseases in 20112011World Health Organization, Geneva, Switzerland. *A prioritized research agenda for prevention and control of noncommunicable diseases in 2011* (2011.0)
14. Asadi-Aliabadi M, Karimi SM, Tehrani-Banihashemi A, Mirbaha-Hashemi F, Janani L, Babaee E. **Effectiveness of pay for performance to non-physician health care providers: a systematic review**. *Health Policy* (2022.0) **126** 592-602. DOI: 10.1016/j.healthpol.2022.03.007
15. 15.WHOIslamic Republic of Iran on a fast-track to beating noncommunicable diseases2017. *Islamic Republic of Iran on a fast-track to beating noncommunicable diseases* (2017.0)
16. Davari M, Haycox A, Walley T. **Health care challenges in Iran**. *Iran J Public Health* (2005.0) **34** 30-31
17. Etemad K, Heidari A, Panahi M, Lotfi M, Fallah F, Sadeghi S. **A challenges in implementing package of essential noncommunicable diseases interventions in Iran’s healthcare system**. *J Health Res Commun* (2016.0) **2** 32-43
18. 18.Djalalinia S, Modirian M, Sheidaei A, Yoosefi M, Zokaiee H, Damirchilu B, et al. Protocol design for large scale cross–sectional studies of surveillance of risk factors of non–communicable diseases in Iran: STEPs 2016. Arch Iran Med. 2017;20(9):608-16.
19. 19.Bitaraf S, Janani L, Hajebi A, Motevalian SA. Information system success of the Iranian integrated health record system based on the clinical information system success model. MJIRI. 2022;36:25.
20. Barzegar MA, Djazayery A. **Evaluation of rural primary health care services in Iran: report on vital statistics in West Azarbaijan**. *Am J Public Health* (1981.0) **71** 739-742. DOI: 10.2105/AJPH.71.7.739
21. 21.Rahbar M, Ahmadi M. Lessons learnt from the model of instructional system for training community health workers in rural health houses of Iran. Iran Red Crescent Med J. 2015;17(2):e2145.
22. 22.Keshvari M, Mohammadi E, Farajzadegan Z, Zargham-Boroujeni A. Experience of Behvarzes (Iranian primary healthcare providers) from giving primary health services in health houses. J Educ Health Promot. 2016;5:7.
23. 23.SCIPopulation and housing censuses2016Statistical Center of Iran, Tehran, Iran. *Population and housing censuses* (2016.0)
24. Asadi-Aliabadi M, Tehrani-Banihashemi A, Mirbaha-Hashemi F, Janani L, Babaee E, Karimi SM. **Evaluating the impact of results-based motivating system on noncommunicable diseases risk factors in Iran: study protocol for a field trial**. *Med J Islam Repub Iran* (2021.0) **35** 66. PMID: 34277503
25. 25.ParliamentThe fifth development plan act: the parliament of Iran2010. *The fifth development plan act: the parliament of Iran* (2010.0)
26. 26.Campbell MK, Piaggio G, Elbourne DR, Altman DG. Consort 2010 statement: extension to cluster randomised trials. Bmj. 2012;345:e5661.
27. 27.MOHMENational action plan for non-communicable diseases prevention and control in Iran 2015–20252015Iranian National Committee for NCDs Prevention and Control, Ministry of Health and Medical Education, Tehran, Iran. *National action plan for non-communicable diseases prevention and control in Iran 2015–2025* (2015.0)
28. Gelband H, Jha P, Sankaranarayanan R, Gauvreau C, Horton S. *Cancer: disease control priorities* (2015.0)
29. Prabhakaran D, Anand S, Watkins D, Gaziano T, Wu Y. **Cardiovascular, respiratory, and related disorders: key messages on essential interventions to address its burden in low-and middle-income countries**. *Disease control priorities. 5* (2015.0)
30. 30.WHOWHO package of essential noncommunicable (PEN) disease interventions for primary health care2020. *WHO package of essential noncommunicable (PEN) disease interventions for primary health care* (2020.0)
31. 31.CBIMonthly economic indicators and statistics2019Central bank of Islamic Republic of Iran, Tehran, Iran. *Monthly economic indicators and statistics* (2019.0)
32. Ainsworth BE, Haskell WL, Whitt MC, Irwin ML, Swartz AM, Strath SJ. **Compendium of physical activities: an update of activity codes and MET intensities**. *Med Sci Sports Exerc* (2000.0) **32** S498-S504. DOI: 10.1097/00005768-200009001-00009
33. Agudo A. *Measuring intake of fruit and vegetables [electronic resource] / Antonio Agudo* (2005.0)
34. 34.MOHMEA set of basic interventions for non-communicable diseases in the primary health care system of Iran (IraPEN for physician)2016. *A set of basic interventions for non-communicable diseases in the primary health care system of Iran (IraPEN for physician)* (2016.0)
35. 35.WHOFood-based dietary guidelines in the WHO European region World Health Organization2003CopenhagenWHO Regional Office for Europe. *Food-based dietary guidelines in the WHO European region World Health Organization* (2003.0)
36. 36.WHOWHO STEPS, noncommunicable disease risk factor survey2016. *WHO STEPS, noncommunicable disease risk factor survey* (2016.0)
37. 37.WHO. Prevalence of current (daily or occasional) tobacco smoking among adults (15 years and older) (percentage) [Available from: https://www.who.int/data/gho/indicator-metadata-registry/imr-details/128]
38. Bertrand M, Duflo E, Mullainathan S. **How much should we trust differences-in-differences estimates?**. *Q J Econ* (2004.0) **119** 249-275. DOI: 10.1162/003355304772839588
39. Farzadfar F, Murray CJ, Gakidou E, Bossert T, Namdaritabar H, Alikhani S. **Effectiveness of diabetes and hypertension management by rural primary health-care workers (Behvarz workers) in Iran: a nationally representative observational study**. *Lancet.* (2012.0) **379** 47-54. DOI: 10.1016/S0140-6736(11)61349-4
40. Tabrizi JS, Pourasghar F, Nikjoo RG. **Status of Iran’s primary health care system in terms of health systems control knobs: a review article**. *Iran J Public Health* (2017.0) **46** 1156. PMID: 29026780
41. Jung DH, Palta M, Smith M, Oliver TR, DuGoff EH. **Differences in receipt of three preventive health care services by race/ethnicity in medicare advantage plans: tracking the impact of pay for performance, 2010 and 2013**. *Prev Chronic Dis* (2016.0) **13** E125-E. DOI: 10.5888/pcd13.160047
42. Doran T, Fullwood C, Kontopantelis E, Reeves D. **Effect of financial incentives on inequalities in the delivery of primary clinical care in England: analysis of clinical activity indicators for the quality and outcomes framework**. *Lancet* (2008.0) **372** 728-736. DOI: 10.1016/S0140-6736(08)61123-X
43. Hayes L, McParlin C, Azevedo LB, Jones D, Newham J, Olajide J. **The effectiveness of smoking cessation, alcohol reduction, diet and physical activity interventions in improving maternal and infant health outcomes: a systematic review of meta-analyses**. *Nutrients.* (2021.0) **13** 1036. DOI: 10.3390/nu13031036
44. Murray JM, Brennan SF, French DP, Patterson CC, Kee F, Hunter RF. **Effectiveness of physical activity interventions in achieving behaviour change maintenance in young and middle aged adults: a systematic review and meta-analysis**. *Soc Sci Med* (2017.0) **192** 125-133. DOI: 10.1016/j.socscimed.2017.09.021
45. Hardcastle SJ, Taylor AH, Bailey MP, Harley RA, Hagger MS. **Effectiveness of a motivational interviewing intervention on weight loss, physical activity and cardiovascular disease risk factors: a randomised controlled trial with a 12-month post-intervention follow-up**. *Int J Behav Nutr Phys Act* (2013.0) **10** 1-16. DOI: 10.1186/1479-5868-10-40
46. 46.Pulling Kuhn A, Stoepker P, Dauenhauer B, Carson RL. A systematic review of multi-component comprehensive school physical activity program (CSPAP) interventions. Am J Health Promot. 2021;35(8):1129-49.
47. Rezola-Pardo C, Rodriguez-Larrad A, Gomez-Diaz J, Lozano-Real G, Mugica-Errazquin I, Patiño MJ. **Comparison between multicomponent exercise and walking interventions in long-term nursing homes: a randomized controlled trial**. *Gerontologist* (2020.0) **60** 1364-1373. DOI: 10.1093/geront/gnz177
48. Simons-Morton DG, Calfas KJ, Oldenburg B, Burton NW. **Effects of interventions in health care settings on physical activity or cardiorespiratory fitness**. *Am J Prev Med* (1998.0) **15** 413-430. DOI: 10.1016/S0749-3797(98)00078-6
49. Kassavou A, Turner A, French DP. **Do interventions to promote walking in groups increase physical activity? A meta-analysis**. *Int J Behav Nutr Phys Act* (2013.0) **10** 1-12. DOI: 10.1186/1479-5868-10-18
50. Lai AY, Lam EY, Fabrizo C, Lee DP, Wan AN, Tsang JS. **A community-based lifestyle-integrated physical activity intervention to enhance physical activity, positive family communication, and perceived health in deprived families: a cluster randomized controlled trial**. *Public Health Front* (2020.0) **8** 434. DOI: 10.3389/fpubh.2020.00434
51. Bopp M, Fallon E. **Community-based interventions to promote increased physical activity**. *Appl Health Econ Health Policy* (2008.0) **6** 173-187. DOI: 10.1007/BF03256132
52. Israel BA, Schulz AJ, Parker EA, Becker AB. **Review of community-based research: assessing partnership approaches to improve public health**. *Annu Rev Public Health* (1998.0) **19** 173-202. DOI: 10.1146/annurev.publhealth.19.1.173
53. Krist AH, Davidson KW, Mangione CM, Barry MJ, Cabana M, Caughey AB. **Behavioral counseling interventions to promote a healthy diet and physical activity for cardiovascular disease prevention in adults with cardiovascular risk factors: US Preventive Services Task Force recommendation statement**. *JAMA.* (2020.0) **324** 2069-2075. DOI: 10.1001/jama.2020.21749
54. Resnicow K, Jackson A, Wang T, De AK, McCarty F, Dudley WN. **A motivational interviewing intervention to increase fruit and vegetable intake through black churches: results of the eat for life trial**. *Am J Public Health* (2001.0) **91** 1686-1693. DOI: 10.2105/AJPH.91.10.1686
55. Khatibi SR, Karimi SM, Moradi-Lakeh M, Kermani M, Motevalian SA. **Fossil energy price and outdoor air pollution: predictions from a QUAIDS model**. *Biofuel Res J* (2020.0) **7** 1205-1216. DOI: 10.18331/BRJ2020.7.3.4
56. 56.Hejazi J, Emamgholipour S. The effects of the re-imposition of US sanctions on food security in Iran. Int J Health Policy Manag. 2022;11(5):651-7.
57. Salehi L, Eftekhar H, Mohammad K, Tavafian SS, Jazayery A, Montazeri A. **Consumption of fruit and vegetables among elderly people: a cross sectional study from Iran**. *Nutr J* (2010.0) **9** 1-9. DOI: 10.1186/1475-2891-9-2
58. Pomerleau J. *Effectiveness of interventions and programmes promoting fruit and vegetable intake [electronic resource]* (2005.0)
59. Van Duyn MAS, Kristal AR, Dodd K, Campbell MK, Subar AF, Stables G. **Association of awareness, intrapersonal and interpersonal factors, and stage of dietary change with fruit and vegetable consumption: a national survey**. *Am J Health Promot* (2001.0) **16** 69-78. DOI: 10.4278/0890-1171-16.2.69
60. Rezaei S, Mahmoudi Z, Sheidaei A, Aryan Z, Mahmoudi N, Gohari K. **Salt intake among Iranian population: the first national report on salt intake in Iran**. *J Hypertens* (2018.0) **36** 2380-2389. DOI: 10.1097/HJH.0000000000001836
61. Noohi F, Sarrafzadegan N, Khosravi A, Andalib E. **The first Iranian recommendations on prevention, evaluation and management of high blood pressure**. *ARYA Atheroscler* (2012.0) **8** 97-118. PMID: 23362408
62. Trieu K, McMahon E, Santos JA, Bauman A, Jolly K-A, Bolam B. **Review of behaviour change interventions to reduce population salt intake**. *Int J Behav Nutr Phys Act* (2017.0) **14** 1-15. DOI: 10.1186/s12966-017-0467-1
63. Cotter J, Cotter MJ, Oliveira P, Cunha P, Polónia J. **Salt intake in children 10–12 years old and its modification by active working practices in a school garden**. *J Hypertens* (2013.0) **31** 1966-1971. DOI: 10.1097/HJH.0b013e328363572f
64. Veroff DR, Sullivan LA, Shoptaw E, Venator B, Ochoa-Arvelo T, Baxter JR. **Improving self-care for heart failure for seniors: the impact of video and written education and decision aids**. *Popul Health Manag* (2012.0) **15** 37-45. DOI: 10.1089/pop.2011.0019
65. Mokhtari M, Vahid H. **Salt and its role in health and disease prevention from the perspectives of Iranian medicine and modern medicine**. *Iran J Med Sci* (2016.0) **41** S58. PMID: 27840524
66. Doubeni CA, Li W, Fouayzi H, DiFranza JR. **Perceived accessibility of cigarettes among youth: a prospective cohort study**. *Am J Prev Med* (2009.0) **36** 239-242. DOI: 10.1016/j.amepre.2008.11.006
67. Homaie Rad E, Pulok MH, Rezaei S, Reihanian A. **Quality and quantity of price elasticity of cigarette in Iran**. *Int J Health Plann Manage* (2021.0) **36** 60-70. DOI: 10.1002/hpm.3062
68. Mayne SL, Gordon-Larsen P, Schreiner PJ, Widome R, Jacobs DR, Kershaw KN. **Longitudinal associations of cigarette prices with smoking cessation: the coronary artery risk development in young adults study**. *Nicotine Tob Res* (2019.0) **21** 678-685. DOI: 10.1093/ntr/nty109
69. Mistry R, Pednekar MS, Gupta PC, Raghunathan TE, Appikatla S, Puntambekar N. **Longitudinal study of adolescent tobacco use and tobacco control policies in India**. *BMC Public Health* (2018.0) **18** 1-8. DOI: 10.1186/s12889-018-5727-8
70. Poorolajal J, Mohammadi Y, Mahmoodi A. **Challenges of tobacco control program in Iran**. *Arch Iran Med* (2017.0) **20** 0
71. 71.Maftoon F, Aghasi M, Rafiee BM. A national campaign on ‘high blood pressure control’in Iran: a rapid evaluation. Payesh Health Monit. 2020;19(6):711-21.
|
---
title: 'Prevalence, incidence, and treatment of anaemia in patients with non-dialysis-dependent
chronic kidney disease: findings from a retrospective real-world study in Italy'
authors:
- Roberto Minutolo
- Giuseppe Grandaliano
- Paolo Di Rienzo
- Robert Snijder
- Luca Degli Esposti
- Valentina Perrone
- Lora Todorova
journal: Journal of Nephrology
year: 2022
pmcid: PMC9998309
doi: 10.1007/s40620-022-01475-x
license: CC BY 4.0
---
# Prevalence, incidence, and treatment of anaemia in patients with non-dialysis-dependent chronic kidney disease: findings from a retrospective real-world study in Italy
## Abstract
### Background
Limited data are available on the epidemiology and clinical management of anaemia in patients with non-dialysis-dependent chronic kidney disease (NDD-CKD).
### Methods
This retrospective observational study was based on records from databases of five Local Health Units across Italy. Adults with reported NDD-CKD stage 3a–5 between 1 January 2014 and 31 December 2016 were identified. Annual prevalence and incidence of anaemia (age- and sex-standardised) and clinical management (erythropoiesis-stimulating agents [ESAs], intravenous [IV] iron, and blood transfusions) were evaluated. Eligibility for ESAs was defined by ≥ 2 records of Hb < 10 g/dL, or < 11 g/dL over 6 months.
### Results
Overall, 101,143 individuals with NDD-CKD (3a–5) recorded between 2014 and 2016 were identified, of whom 40,020 ($39.6\%$) were anaemic. Prevalence of anaemia was $33.8\%$ in 2016 and incidence of anaemia was stable (11.4–$12.4\%$) from 2014 to 2016. Prevalence and incidence of anaemia increased with CKD stage. Among eligible patients, $12.8\%$ with Hb < 11 g/dL and $15.5\%$ with Hb < 10 g/dL received ESAs, and the proportion treated increased with CKD stage. Among ESA-treated patients with at least 2 years of follow up, $18.4\%$ and $19.3\%$ received IV iron in the Hb < 11 and < 10 g/dL groups, respectively, and $16.5\%$ and $19.4\%$ received blood transfusions. Corresponding proportions for the overall anaemic cohort were $9.0\%$ and $11.3\%$, respectively.
### Conclusions
Anaemia is a significant issue in patients with NDD-CKD. Low rates of ESA treatment indicate a potential treatment gap and suggest that anaemia may not be adequately controlled in many patients.
### Supplementary Information
The online version contains supplementary material available at 10.1007/s40620-022-01475-x.
## Introduction
Chronic kidney disease (CKD) has an estimated global prevalence of 11–$13\%$ and is a growing public health issue worldwide [1]. Data from the 2008–2012 Cardiovascular risk profile in Renal patients of the Italian Health Examination Survey (CARHES), which examined Italian residents aged 35–79 years, showed a CKD prevalence of about $7\%$, with early stages (1 and 2) accounting for nearly $60\%$ of cases [2].
Anaemia is a common complication in patients with CKD, and its prevalence increases with advancing CKD stage [3, 4]. The main cause of anaemia associated with CKD is reduced production of erythropoietin by the failing kidneys as a result of impaired oxygen-sensing mechanisms in CKD [5]. Anaemia is associated with CKD progression and other adverse outcomes, including major adverse cardiovascular events, hospitalisation, and all-cause mortality [3].
Iron supplementation and erythropoiesis-stimulating agents (ESAs) are the mainstays of treatment for anaemia in CKD [6–8]. Recent data from the prospective, multinational Chronic Kidney Disease Outcomes and Practice Patterns Study (CKDopps) on the management of anaemia in patients with non-dialysis-dependent CKD (NDD-CKD) under nephrologist care in Brazil, France, Germany, and the USA, highlighted that anaemia monitoring and treatment are suboptimal, with many patients untreated despite guideline-based indications to treat [9, 10]. With respect to Italy, data on the prevalence and management of anaemia of NDD-CKD are limited. In a prospective cohort study of adults with NDD-CKD stage 3–5 attending 25 renal clinics in Italy in 2003, the prevalence of mild anaemia was $41.3\%$ at the first visit [11]. Another prospective study in 19 outpatient renal clinics across Italy reported an unexpectedly high prevalence of anaemia (severe 18.0–$19.3\%$, mild 43.2–$44.0\%$) among patients with NDD-CKD over a 6-month period, which was linked to clinical inertia in initiating anaemia treatment [12].
Therefore, this observational study was designed to evaluate the prevalence of NDD-CKD (stage 3a–5), in addition to the prevalence and incidence of anaemia of NDD-CKD (stage 3a–5), and the therapeutic management of this complication in Italian clinical practice.
## Materials and methods
This was a retrospective cohort study involving adult patients (aged ≥ 18 years) with reported NDD-CKD stage 3a–5 between 1 January 2014 and 31 December 2016 (the inclusion period). Records were extracted from administrative (demographic, hospitalisation, pharmaceutical, and outpatient specialist services) and laboratory databases of five Local Health Units (LHUs) across Italy (full details provided in Online Resource 1). The Ethics Committee of each participating LHU was notified of this study and provided approval (reference numbers: Prot. N AslVC.Farm.19.02, Prot. 96/SegCE/2019, Prot. N. 0,219,118, Prot. N. 03, and Prot. CEUR-2020-Os-029 [Online Resource 1]).
NDD-CKD (3a–5) and respective stages were defined as detailed in Online Resource 2. Anaemia was defined as a haemoglobin (Hb) value below the cut-off specified by Kidney Disease Improving Global Outcomes (KDIGO) guidelines (< 13 g/dL in males and < 12 g/dL in females) [7]. Patients were grouped into three cohorts: NDD-CKD (3a–5) cohort, anaemia of NDD-CKD (3a–5) cohort, and anaemia of NDD-CKD (3a–5) ESA-treated cohort (referred to hereafter as NDD-CKD, anaemic, and ESA-treated cohorts, respectively). The three patient cohorts had distinct index dates during the study period (Fig. 1). Patients were excluded if 1) they were on dialysis at or before the CKD index date; 2) they had a cancer diagnosis ≤ 5 years before the CKD index date; 3) the characterisation (‘look-back’) period was < 1 year; 4) the anaemia index date was ≥ 1 month before the CKD index date (incident anaemic patients); or 5) the ESA treatment index date was ≥ 1 month prior to the anaemia index date (incident ESA-treated patients). For each cohort, the index date marked the beginning of the follow-up period (Fig. 1). An identification period of ≥ 1 year before the respective index dates was used to characterise patients in terms of comorbidities (see Online Resource 2 for details) and previous treatment for the management of anaemia. The follow-up period was 2 years after the respective index date, or from index date until death, dialysis, transplant, or exiting the database, whichever occurred first. Fig. 1Study design. aFor the NDD−CKD cohort (CKD index date), date of first diagnosis of CKD stage 3a–5 or earliest record of eGFR <60 mL/min/1.73 m2 within the inclusion period; for the anaemic cohort (anaemia of CKD index date): date of earliest second of two consecutive tests (within 3 months of each other) reporting Hb <13 g/dL (males) or <12 g/dL (females) within the inclusion period; for the ESA−treated cohort (ESA treatment index date), date of first ESA prescription within the inclusion period. bOr until death, dialysis, transplant, or exiting the database (whichever occurred first). CKD chronic kidney disease; eGFR estimated glomerular filtration rate; ESA erythropoiesis−stimulating agent; Hb haemoglobin; NDD−CKD non−dialysis dependent chronic kidney disease The objectives of this study were to: 1) estimate the annual prevalence of NDD-CKD; 2) estimate the annual prevalence and incidence of anaemia among patients with NDD-CKD; and 3) evaluate the management of anaemia in patients with NDD-CKD, including treatment with ESAs (frequency and types of ESAs used), intravenous iron, and blood transfusions.
Incident and prevalent patients within the three cohorts are defined in Online Resource 3. Results were age- and sex-standardised according to *Italian census* data from 1 January of the year following the inclusion date. Prevalence was defined as the presence of NDD-CKD or anaemia of NDD-CKD at 31 December of the year. Incident cases were defined as newly diagnosed NDD-CKD or anaemia of NDD-CKD during the year of interest. Eligibility for ESA treatment was defined using two Hb thresholds: ≥ 2 records of Hb < 10 g/dL over 6 months; and ≥ 2 records of Hb < 11 g/dL over 6 months. The Hb < 10 g/dL threshold is recommended by international guidelines [7], while the Hb < 11 g/dL threshold is supported by the Italian Society of Nephrology [13] and defined in the Italian Therapeutic Plan for ESA prescription [14, 15]. Patients with a record of ESA prescription and ≥ 2 records of Hb < 10 or < 11 g/dL over 6 months were categorised as the ESA-treated cohort (referred to hereafter as Hb < 10 g/dL ESA-treated cohort and Hb < 11 g/dL ESA-treated cohort). ESAs were sub-categorised as short-acting (epoetin alfa/beta) or long-acting (darbepoetin alfa, methoxy polyethylene glycol-epoetin beta). Autoimmune diseases were defined by ICD-9-CM codes 696, 714, 720, 555, and 556 (see Online Resource 2 for further details). ICD-9-CM 250.0x (diabetes mellitus without mention of complication) and ATC A10 (drugs used in diabetes) codes were used as a proxy for diabetes.
Data analysis was primarily descriptive. Statistical significance was determined for binary variables using the Cochran–Armitage test, continuous variables using linear regression, and categorical variables using a Cochran–Mantel–Haenszel test. Analyses were reported by CKD stage as well as for the overall sample. All analyses were performed using STATA SE, version 12.0 (StataCorp LLC, College Station, TX, USA). End-stage renal disease (ESRD) risk assessment was determined using the Fine and Gray proportional sub-distribution hazards model, with death as a competing risk factor. Risk of death was determined using the Cox proportional hazards model. Covariates for each included: CKD stage, age, sex, cardiovascular disease, hypertension, diabetes mellitus, and autoimmune diseases.
## Disposition of patients
The five LHU databases included records for just over 1.5 million inhabitants, which corresponds to about $3\%$ of the entire adult population of Italy [16]. A total of 101,143 individuals with NDD-CKD were included in the analysis, of whom 40,020 ($39.6\%$) were anaemic (Fig. 2).Fig. 2Patient flowchart. aAnaemia of NDD-CKD (3a–5) cohort. bAnaemia of NDD-CKD (3a–5) ESA-treated cohort (eligibility criterion for ESA treatment: ≥ 2 records of Hb < 11 g/dL over a 6-month period). cAnaemia of NDD-CKD (3a–5) ESA-treated cohort (eligibility criterion for ESA treatment: ≥ 2 records of Hb < 10 g/dL over a 6-month period). ESA erythropoiesis-stimulating agent; Hb haemoglobin; LHU local health unit; NDD-CKD non-dialysis dependent chronic kidney disease
## Patient demographics and characteristics at index date
Over $70\%$ of NDD-CKD patients had CKD stage 3a (Table 1). Mean age was 76.1 years and the majority of patients (> $90\%$) were aged ≥ 60 years. Approximately half of male patients were 60–79 years of age ($50.7\%$), and approximately half of female patients were ≥ 80 years of age ($50.2\%$) (Online Resource 4). No clear trends were observed for mean age across the CKD stages in the anaemic and ESA-treated cohorts (Online Resource 5 and 6). The proportion of males ranged between 39 and $48\%$ across the CKD stages in the NDD-CKD cohort (Table 1). Of those patients with NDD-CKD, $43.3\%$ of males and $37.0\%$ of females were included in the anaemic cohort (Online Resource 4).Table 1Patient demographics for the NDD-CKD (3a–5) cohort at index dateCKD stageOverallp valued3a3b45N71,63820,32071891996101,143–Age, mean (SD)74.7 (12.4)79.7 (11.2)80.0 (12.1)75.1 (14.5)76.1 (12.4) < 0.001Age in years, n (%): 18–39948 (1.3)139 (0.7)68 (0.9)40 (2.0)1,195 (1.2) < 0.001 40–596948 (9.7)1036 (5.1)439 (6.1)245 (12.3)8668 (8.6) 60–7935,632 (49.7)7203 (35.4)2284 (31.8)788 (39.5)45,907 (45.4) ≥ 8028,100 (39.2)11,941 (58.8)4397 (61.2)920 (46.1)45,358 (44.8)Male, n (%)29,608 (41.3)7939 (39.1)2845 (39.6)954 (47.8)41,346 (40.9) < 0.001Included in anaemic cohorta, n (%)23,626 (33.0)10,181 (50.1)4682 (65.1)1531 (76.7)40,020 (39.6)–Included in Hb < 11 g/dL ESA-treated cohortb, n956 (1.3)851 (4.2)946 (13.2)485 (24.3)3238 (3.2)–Included in Hb < 10 g/dL ESA-treated cohortc, n832 (1.2)715 (3.5)771 (10.7)424 (21.2)2742 (2.7)–Anaemia was defined as a Hb value below the cut-off specified by Kidney Disease Improving Global Outcomes (KDIGO) guidelines (< 13 g/dL in males and < 12 g/dL in females) [7]CKD chronic kidney disease; ESA erythropoiesis-stimulating agent; Hb haemoglobin; NDD-CKD non-dialysis dependent chronic kidney disease; SD standard deviationaAnaemia of NDD-CKD (3a–5) cohortbAnaemia of NDD-CKD (3a–5) ESA-treated cohort (eligibility criterion for ESA treatment: ≥ 2 records of Hb < 11 g/dL over a 6-month period)cAnaemia of NDD-CKD (3a–5) ESA-treated cohort (eligibility criterion for ESA treatment: ≥ 2 records of Hb < 10 g/dL over a 6-month period)dResult of a statistical test to identify the existence of a trend in the ordered CKD stages *In* general, rates of comorbidities tended to be slightly higher in the anaemic cohort than in the NDD-CKD cohort (Table 2 and Online Resource 7 and 8). Cardiovascular disease, hypertension, diabetes mellitus, autoimmune diseases, and myelodysplastic syndrome were detected in $23.5\%$, $89.9\%$, $32.4\%$, $2.1\%$, and $0.8\%$ of patients with anaemia, respectively, without any apparent trends across the CKD stages (Table 2). The proportions of patients in the anaemic cohort who had received previous treatments for anaemia (ESAs, iron, blood transfusion) showed an upward trend with CKD progression (Table 2). Clinical characteristics of patients with anaemia of NDD-CKD stratified by sex can be found in Online Resource 9.Table 2Clinical characteristics of patients with anaemia of NDD-CKD (3a–5) at index dateCKD stageOverallp valuea3a3b45N23,62610,1814682153140,020–CV disease, n (%)5321 (22.5)2548 (25.0)1216 (26.0)328 (21.4)9413 (23.5) < 0.001Hypertension, n (%)20,827 (88.2)9457 (92.9)4322 (92.3)1366 (89.2)35,972 (89.9) < 0.001Diabetes mellitus, n (%)7341 (31.1)3451 (33.9)1660 (35.5)502 (32.8)12,954 (32.4) < 0.001Autoimmune diseases, n (%)513 (2.2)211 (2.1)97 (2.1)30 (2.0)851 (2.1)0.886Myelodysplastic syndrome, n (%)196 (0.8)84 (0.8)49 (1.0)9 (0.6)338 (0.8)0.416Previous ESA treatment, n (%)389 (1.6)410 (4.0)567 (12.1)396 (25.9)1762 (4.4) < 0.001Previous iron therapy (IV or oral), n (%)211 (0.9)175 (1.7)200 (4.3)136 (8.9)722 (1.8) < 0.001Previous blood transfusion, n (%)51 (0.2)32 (0.3)38 (0.8)10 (0.7)131 (0.3) < 0.001CKD chronic kidney disease; CV cardiovascular; ESA erythropoiesis-stimulating agent; IV intravenousaResult of a statistical test to identify the existence of a trend in the ordered CKD stages
## Prevalence of NDD-CKD
In 2016, the standardised prevalence of NDD-CKD was $5.6\%$ (Table 3). CKD stage 3a was the most common ($4.2\%$), while the prevalence of stages 3b–5 ranged from 0.1–$1.0\%$. Overall, there was a numerically higher proportion of female ($6.3\%$) than male ($4.7\%$) patients with NDD-CKD in 2016.Table 3Prevalence of NDD-CKD (3a–5) and anaemia of NDD-CKD (3a–5) in 2016CKD stageOverall3a3b45Prevalence of NDD-CKD (3a–5) in the LHU patient population 2016Numerator (adult patients diagnosed)62,68315,2214486123583,625 2016Denominator (adult LHU patient population)1,507,3911,507,3911,507,3911,507,3911,507,391 2016Prevalence$4.2\%$$1.0\%$$0.3\%$$0.1\%$$5.5\%$ 2016Standardised prevalencea$4.2\%$$1.0\%$$0.3\%$$0.1\%$$5.6\%$ 2016Standardised prevalence, male$3.5\%$$0.8\%$$0.3\%$$0.1\%$$4.7\%$ 2016Standardised prevalence, female$4.7\%$$1.2\%$$0.3\%$$0.1\%$$6.3\%$Prevalence of anaemia among patients with NDD-CKD (3a–5) 2016Numerator (patients diagnosed with anaemia)17,6816779282697328,259 2016Denominator (NDD-CKD patients by CKD stage)62,68315,2214486123583,625 2016Prevalence$28.2\%$$44.5\%$$63.0\%$$78.8\%$$33.8\%$ 2016Standardised prevalencea$28.2\%$$44.6\%$$63.1\%$$78.9\%$$33.8\%$ 2016Standardised prevalence, male$31.2\%$$48.5\%$$66.7\%$$81.1\%$$37.0\%$ 2016Standardised prevalence, female$26.2\%$$42.2\%$$60.6\%$$76.5\%$$31.7\%$Bold italic text emphasizes the key data discussed in the body of the text rather than specific differencesCKD chronic kidney disease; LHU Local Health Unit; NDD-CKD non-dialysis dependent chronic kidney diseaseaPrevalence at 31 December of each year was age- and sex-standardised using census data from 1 January of the following year
## Prevalence and incidence of anaemia of NDD-CKD
Standardised prevalence of anaemia among patients with NDD-CKD in 2016 was $33.8\%$ (Table 3) and increased with CKD stage, rising from $28.2\%$ among patients with stage 3a to $78.9\%$ among those with stage 5. There was a numerically higher proportion of male ($37.0\%$) than female ($31.7\%$) patients with anaemia of NDD-CKD. Overall, the standardised incidence of anaemia of NDD-CKD was $12.4\%$, $12.3\%$, and $11.4\%$ in 2014, 2015, and 2016, respectively, and increased with CKD stage within each year (Table 4). For all years, overall standardised incidence of anaemia of CKD was numerically higher for males than females. Table 4Annual incidence of anaemia among patients with NDD-CKD (3a–5) (2014–2016)YearIncidence of anaemiaCKD stageOverall3a3b452014Numerator (patients newly diagnosed with anaemia)a4174212597427775502014Denominator (NDD-CKD patients by CKD stage)42,35213,1904472111861,1322014Incidence$9.9\%$$16.1\%$$21.8\%$$24.8\%$$12.4\%$2014Standardised incidenceb$9.9\%$$16.1\%$$21.8\%$$24.8\%$$12.4\%$2014Standardised incidence, male$11.2\%$$17.9\%$$23.4\%$$23.6\%$$13.7\%$2014Standardised incidence, female$9.0\%$$15.1\%$$20.8\%$$25.9\%$$11.5\%$2015Numerator (patients newly diagnosed with anaemia)a5077207880724882102015Denominator (NDD-CKD patients by CKD stage)50,13112,339350388666,8592015Incidence$10.1\%$$16.8\%$$23.0\%$$28.0\%$$12.3\%$2015Standardised incidenceb$10.1\%$$16.9\%$$23.1\%$$28.0\%$$12.3\%$2015Standardised incidence, male$11.1\%$$18.3\%$$24.3\%$$26.6\%$$13.3\%$2015Standardised incidence, female$9.5\%$$16.0\%$$22.3\%$$29.6\%$$11.7\%$2016Numerator (patients newly diagnosed with anaemia)a4901182370222576512016Denominator (NDD-CKD patients by CKD stage)52,94510,788261168767,0312016Incidence$9.3\%$$16.9\%$$26.9\%$$32.8\%$$11.4\%$2016Standardised incidenceb$9.3\%$$16.9\%$$26.9\%$$32.8\%$$11.4\%$2016Standardised incidence, male$10.3\%$$18.6\%$$29.6\%$$30.1\%$$12.5\%$2016Standardised incidence, female$8.6\%$$15.9\%$$25.1\%$$36.1\%$$10.7\%$p valuec < 0.0010.174 < 0.0010.001 < 0.001Bold italic text emphasizes the key data discussed in the body of the text rather than specific differencesCKD chronic kidney disease; NDD-CKD non-dialysis dependent chronic kidney diseaseaIncident cases were defined as newly diagnosed NDD-CKD during the year of interestbAnnual incidence was age- and sex-standardised using census data from 1 January of the following yearcp value refers to the comparison of standardised incidence value among different years
## Eligibility for ESAs and ESA treatment
Using the Hb < 11 g/dL criterion to determine ESA eligibility, 25,360 patients were considered eligible for ESA treatment, of whom 3238 were prescribed ESAs. Using the Hb < 10 g/dL criterion, 17,703 patients were considered eligible for ESA treatment, and 2742 were prescribed ESAs (Fig. 2). Over the whole study period (2014–2016), the overall proportion of patients with anaemia of NDD-CKD who were eligible for ESA treatment increased with CKD stage, regardless of the eligibility criterion used, or sex (Table 5; Online Resource 10). The proportion of ESA-eligible patients was $63.4\%$ using the Hb < 11 g/dL threshold ($57.6\%$ for male and $68.0\%$ for female patients), and $44.2\%$ using the Hb < 10 g/dL threshold ($40.7\%$ for male and $47.1\%$ for female patients). There was an upward trend across the CKD stages for the proportion of eligible patients who were treated with ESAs. The proportion of eligible patients treated with ESAs was $12.8\%$ when the Hb threshold was < 11 g/dL, ranging from $6.7\%$ for patients in stage 3a to $39.8\%$ for those in stage 5. Using the stricter eligibility criterion of Hb < 10 g/dL, a slightly higher proportion of eligible patients were treated with ESAs overall ($15.5\%$) and for each CKD stage. Table 5Proportion of patients with anaemia of NDD-CKD (3a–5) eligible for ESA treatment, and treated with ESAs (overall study period, 2014–2016)CKD stageOverallp valuea3a3b45Number of patients with anaemia of NDD-CKD (3a–5)23,62610,1814682153140,020–Eligibility criterion ≥ 2 records of Hb < 11 g/dL over a 6-month period Eligible for ESA, n (%)14,203 (60.1)6592 (64.7)3347 (71.5)1218 (79.6)25,360 (63.4) < 0.001 Eligible patients treated with ESAs, n/N (%)$\frac{956}{14}$,203 (6.7)$\frac{851}{6592}$ (12.9)$\frac{946}{3347}$ (28.3)$\frac{485}{1218}$ (39.8)$\frac{3238}{25}$,360 (12.8) < 0.001Eligibility criterion ≥ 2 records of Hb < 10 g/dL over a 6-month period Eligible for ESA, n (%)9675 (41.0)4612 (45.3)2442 (52.2)974 (63.6)17,703 (44.2) < 0.001 Eligible patients treated with ESAs, n/N (%)$\frac{832}{9675}$ (8.6)$\frac{715}{4612}$ (15.5)$\frac{771}{2442}$ (31.6)$\frac{424}{974}$ (43.5)$\frac{2742}{17}$,703 (15.5) < 0.001CKD chronic kidney disease; ESA erythropoiesis-stimulating agent; Hb haemoglobin; NDD-CKD non-dialysis dependent chronic kidney diseaseaResult of statistical test to identify the existence of a trend in the ordered CKD stages Overall, $59.5\%$ ($\frac{1927}{3238}$) of patients in the Hb < 11 g/dL ESA-treated cohort received short-acting ESAs and $49.3\%$ ($\frac{1597}{3238}$) received long-acting ESAs. The corresponding proportions for the Hb < 10 g/dL ESA-treated cohort were $62.0\%$ ($\frac{1699}{2742}$) and $47.4\%$ ($\frac{1300}{2742}$), respectively. Some patients switched between the two ESA types, receiving both long- and short-acting ESAs during the study period. In the Hb < 11 g/dL ESA-treated cohort, the proportion of patients receiving short-acting ESAs decreased with CKD stage from $69.6\%$ to $54.0\%$, in favour of long-acting ESAs (the proportion of patients increased from $36.1\%$ for stage 3a to $63.7\%$ for stage 5). A similar pattern was observed for the Hb < 10 g/dL ESA-treated cohort.
## Iron supplementation and blood transfusions (patients with at least 2 years of follow up)
Intravenous (IV) iron infusions were received by $9.0\%$ of patients in the anaemic cohort, and $18.4\%$ and $19.3\%$ of patients in the Hb < 11 g/dL ESA-treated and Hb < 10 g/dL ESA-treated cohorts, respectively (Table 6). Data for patients who received oral iron during the follow-up period were not consistently available in the study databases, and thus were not analysed here. Blood transfusions were received by $11.3\%$, $16.5\%$, and $19.4\%$ of patients in the anaemic cohort, Hb < 11 g/dL ESA-treated cohort, and Hb < 10 g/dL ESA-treated cohort, respectively. Table 6Proportions of patients in the anaemia of NDD-CKD (3a–5) and ESA-treated cohorts with at least 2 years of follow up who received intravenous iron supplementation and blood transfusionsCKD stageOverallp valuee3a3b45Intravenous irona, n/N (%) Anaemic cohortb$\frac{1564}{16}$,509 (9.5)$\frac{472}{6336}$ (7.4)$\frac{196}{2528}$ (7.8)$\frac{108}{658}$ (16.4)$\frac{2340}{26}$,031 (9.0) < 0.001 Hb < 11 g/dL ESA-treated cohortc$\frac{127}{590}$ (21.5)$\frac{67}{547}$ (12.2)$\frac{82}{515}$ (15.9)$\frac{64}{192}$ (33.3)$\frac{340}{1844}$ (18.4) < 0.001 Hb < 10 g/dL ESA-treated cohortd$\frac{113}{495}$ (22.8)$\frac{56}{439}$ (12.8)$\frac{68}{392}$ (17.3)$\frac{49}{155}$ (31.6)$\frac{286}{1481}$ (19.3) < 0.001Blood transfusion, n/N (%) Anaemic cohortb$\frac{1803}{16}$,509 (10.9)$\frac{728}{6336}$ (11.5)$\frac{334}{2528}$ (13.2)$\frac{81}{658}$ (12.3)$\frac{2946}{26}$,031 (11.3)0.026 Hb < 11 g/dL ESA-treated cohortc$\frac{123}{590}$ (20.8)$\frac{83}{547}$ (15.2)$\frac{79}{515}$ (15.3)$\frac{19}{192}$ (9.9)$\frac{304}{1844}$ (16.5)0.014 Hb < 10 g/dL ESA-treated cohortd$\frac{118}{495}$ (23.8)$\frac{79}{439}$ (18.0)$\frac{72}{392}$ (18.4)$\frac{19}{155}$ (12.3)$\frac{288}{1481}$ (19.4)0.054Percentages calculated using the proportion of patients with at least 2 years of follow up in each cohortAnaemia was defined as a Hb value below the cut-off specified by Kidney Disease Improving Global Outcomes (KDIGO) guidelines (< 13 g/dL in males and < 12 g/dL in females) [7]CKD, chronic kidney disease; ESA, erythropoiesis-stimulating agent; Hb, haemoglobin; NDD-CKD, non-dialysis dependent chronic kidney diseaseaNo data were available for oral iron therapybAnaemia of NDD-CKD (3a–5) cohortcAnaemia of NDD-CKD (3a–5) ESA-treated cohort (eligibility criterion for ESA treatment: ≥ 2 records of Hb < 11 g/dL over a 6-month period)dAnaemia of NDD-CKD (3a–5) ESA-treated cohort (eligibility criterion for ESA treatment: ≥ 2 records of Hb < 10 g/dL over a 6-month period)e Result of statistical test to identify the existence of a trend in the ordered CKD stages
## Risk of end stage renal disease and death
The sub-distribution model showed that CKD stage, age, sex, hypertension, diabetes mellitus, and anaemia had a significant effect on the development of ESRD in patients with CKD ($p \leq 0.001$ for all; Online Resource 11). In particular, anaemia was associated with a fourfold higher risk of ESRD (sub-distribution hazard ratio 4.00, confidence interval [CI]: 2.75–5.81; $p \leq 0.001$). Similarly, we found an increased risk of death in CKD patients with anaemia (hazard ratio 2.24, CI: 2.17–2.31; $p \leq 0.001$) compared to patients without anaemia.
## Discussion
To date, this is the largest real-world study on anaemia of NDD-CKD in Italy. Starting with records for approximately 1.5 million inhabitants from a pool of five areas geographically distributed across Italy, 101,143 patients with NDD-CKD were identified. Anaemia of NDD-CKD was characterised in terms of epidemiology, demographics, clinical characteristics, and management.
The overall standardised prevalence of NDD-CKD in 2016 was $5.6\%$. Previous Italian studies describing prevalence rates for CKD have not reported data on anaemia, or focused on NDD-CKD, and used different patient inclusion criteria; therefore, those findings are not directly comparable with the results from this research [2, 17]. In this study, the standardised prevalence rate in 2016 for anaemia among patients with NDD-CKD was $33.8\%$, which differs from rates (41–$62\%$) reported by other Italian studies enrolling patients regularly followed in renal clinics [11, 12]. Different prevalence rates reported by studies are expected given the variations in study populations and methodology. Patients with NDD-CKD were predominantly female in the present study. This is consistent with previous studies, which report that the prevalence of CKD was higher in female patients, despite increased severity and risk of progression of CKD in male patients [18]. In the present study, the prevalence of anaemia among patients with NDD-CKD was numerically higher in males than females, which conflicts with findings from previous studies that females were more likely to develop anaemia of CKD than males [3, 19]. This may be a consequence of the higher Hb cut-off value in male patients compared to females (< 13 g/dL versus < 12 g/dL) for the definition of anaemia in the present study. As a result, the Hb cut-off value might be attained at an earlier stage of disease in male than female patients [20]. Furthermore, comorbidities with the potential to induce anaemia (e.g., CV disease and diabetes) were reported in higher proportions of male patients than female patients in this study. Standardised prevalence and incidence of anaemia among patients with NDD-CKD increased through CKD stages, suggesting a greater need for anaemia management in patients with more advanced CKD. Trends in the standardized incidence of anaemia among patients with NDD-CKD from 2014 to 2016, overall and by CKD stage, were variable; particularly for stage 3a where anaemia was less frequent.
Anaemia is known to become more common as CKD progresses, and may be associated with increased risks of cardiovascular disease, hospitalisation, cognitive impairment, and death [6]. In the current study, hypertension was the most frequent comorbidity, with rates of $83.8\%$ to $96.4\%$ across the three cohorts, and without a clear relationship with CKD stage. These observations are in line with other studies in patients with NDD-CKD, in which about 80–$87\%$ of patients had hypertension [21, 22]. Similarly, no trend across CKD stages was observed for the prevalence of diabetes mellitus, a clinical condition more frequently associated with new-onset anaemia [23]. The present study also showed that a diagnosis of anaemia in patients with CKD was associated with a higher risk of ESRD, and of death, independently of risk factors related to adverse outcomes in this patient population [11, 22, 24, 25].
Among anaemic patients, we applied two different cut-offs of Hb concentration (≥ 2 records of either Hb < 10 or < 11 g/dL over 6 months) in estimating ESA eligibility to reflect international guidelines and clinical practice in Italy, respectively. This ensured that the data would be relevant worldwide and not just in Italy. Requiring ≥ 2 records over 6 months rather than a single record reduced the chance that a reduction in Hb may have had a temporary cause such as inflammation, infection, bleeding, or surgery [26]. Based on the two eligibility criteria, $44.2\%$ (Hb < 10 g/dL) and $63.4\%$ (Hb < 11 g/dL) of anaemic patients were eligible for ESA treatment. Regardless of the Hb cut-off, the proportion of eligible patients actually treated with ESAs was low (≤ $15.5\%$), although the proportion treated increased across the CKD stages (from < $10\%$ in stage 3a to approximately $40\%$ in stage 5). These results suggest that patients who required treatment for anaemia were inadequately treated when ESAs could have been prescribed. Therapeutic inertia has been reported previously in the context of nephrology care, with $34\%$ of NDD-CKD patients not receiving ESAs despite having anaemia over a 6-month observation period [12]. This phenomenon has been confirmed recently by the CKDopps analysis, which showed that a large proportion of patients with anaemia of NDD-CKD did not receive anaemia medication within 1 year [10]. However, it would be an oversimplification to conclude from the present study that anaemia was undertreated. Indeed, international guidelines recommend that the decision to treat with ESAs is not based on Hb level alone, but should also consider symptoms related to anaemia, prior cardiovascular history (e.g., stroke), active or past history of malignancy, rate of decrease of Hb concentration, prior response to iron therapy, the risk for transfusion, and risks related to ESA therapy [7, 26]. Therefore, the large proportion of untreated patients may reflect the need for healthcare professionals to understand the underlying causes of anaemia before prescribing ESAs, while also reducing the burden of cost of inappropriate prescriptions on the National Health Service. In addition, this large untreated patient population may rely on a sub-optimal healthcare service, which may not refer all patients to the appropriate specialist, such as a nephrologist, to receive treatment. Of note, the present analysis was based on ESAs prescribed to patients in the outpatient setting, and ESA use during hospitalisation was not captured. However, ESAs prescribed during hospitalisation continue to be given after discharge, and therefore the absence of hospital records of ESA use is unlikely to have affected the study findings.
The proportion of anaemic patients who received IV iron infusions in the present study ($9.0\%$) was generally consistent with data reported in other studies. A study using data from the CKDopps found that IV iron was received by $12\%$, $9\%$, $33\%$, and $8\%$ of patients with anaemia of CKD stages 3a–5 in Brazil, France, Germany, and the USA, respectively [9]. Another study conducted in Italy found that IV iron was prescribed to approximately $3\%$ of NDD-CKD patients receiving iron supplements [12]. A challenge remains in the estimation of the proportion of patients receiving oral iron therapy. Oral iron supplements are likely to be the predominant form of iron therapy in patients with NDD-CKD, but their over-the-counter availability makes their use more difficult to track. As oral iron data were not readily available in the databases used for the current study, these data were not analysed for the follow-up period.
An original finding of this study was the rate of blood transfusions in the NDD-CKD population. Studies from the USA have reported a rising prevalence of blood transfusions in the non-dialysis population, which has been linked with a concomitant and marked reduction in the use of ESAs [27]. This greater use of blood transfusions than ESAs or IV iron is believed to have been triggered by policy changes, including drug reimbursement, and additional FDA ‘black box’ warnings related to the potential adverse effects of ESAs [27]. Indeed, in one US study, the proportion of NDD-CKD patients aged 66–85 years with anaemia who received blood transfusions ($22.2\%$) was found to be almost double the proportion observed in anaemic patients ($11.3\%$) in the present study [27]. This finding could be explained by differences between the US and Italian healthcare systems. Of note, in this study the use of blood transfusions decreased with advancing CKD stage in parallel with increasing ESA prescriptions. This trend was independent of the severity of anaemia (either Hb < 10 g/dL or < 11 g/dL).
The ‘real-world’ setting and large sample size are major strengths of this study. Using population-based data, a ‘snapshot’ has been provided of the management of anaemia in more than 100,000 patients with NDD-CKD in Italy, a population for whom these epidemiologic data were lacking. This study, however, has limitations. We cannot exclude that other haemoglobinopathies (in particular thalassaemia) have been classified as renal anaemia in our study, although the advanced age of our study population strongly limits the potential impact of thalassaemia on our estimates. Also, the inclusion of patients with anaemia from non-CKD-related causes in this study may have resulted in a slight overestimation of the number of patients eligible for ESAs, and the occurrence of clinical inertia. Frequently, the use of large administrative databases to investigate a specific clinical question can be associated with limited suitable data, thus reducing cohort size [28]. The databases captured only direct medical healthcare resource use reimbursed by the Italian National Health Service. In patients without ICD-9-CM codes for CKD, only one creatinine value was used to define the presence of CKD, whereas international guidelines require at least two pathological creatinine values or other markers of kidney damage more than 3 months apart [29]. This discrepancy may have resulted in a partial overestimation of CKD cases. A proxy (e.g., specific drugs, hospitalisations) was used for comorbidities, such as diabetes, which may have introduced error in estimating their prevalence.
The length of look-back period can also impact estimates of prevalence and incidence: a shorter look-back period can potentially lead to the overestimation of an incidence rate due to the misclassification of prevalent and recurrent cases as incident cases [30]. The minimum look-back period for inclusion in this study was 1 year but could have been up to 8 years depending on the individual patient’s inclusion date; however, the length of look-back period was not analysed for this study. Other limitations include the evaluation of drug treatments based on prescriptions only, and therefore ESA prescriptions could have been related to conditions other than anaemia of CKD. Patients with a cancer diagnosis were excluded from the study in order to reduce this potential bias, as ESAs could be prescribed to these patients and therefore could have influenced the results for CKD and anaemia. Eligibility for ESA treatment was based on Hb levels alone. As other decision criteria for ESA treatment recommended by the guidelines, such as symptoms of anaemia and prior cardiovascular history, were not recorded in the LHU databases, it is not possible to comment on the extent to which patients with anaemia may have been undertreated [7, 26]. The iron supplementation data were limited to IV iron. Lastly, it is unknown to what extent the data for Italy from the current study may be generalised to other countries given the differences in healthcare systems.
In conclusion, this study demonstrated that anaemia is a significant issue in patients with NDD-CKD, and the incidence increases with CKD stage. The proportion of patients eligible for ESA treatment who actually received ESAs was relatively low, indicating a potential treatment gap, and suggesting that anaemia may not be adequately controlled. There appears to be an unmet need to remedy the apparent clinical inertia and improve the diagnosis and treatment of anaemia of NDD-CKD in Italy. However, this study also highlights the careful choices made by healthcare professionals when prescribing ESAs, accounting for both the underlying conditions of the patients, and the cost of inappropriate ESA prescription to the Italian National Health Service.
## Supplementary Information
Below is the link to the electronic supplementary material. Supplementary file1 (DOCX 87 kb)
## References
1. Hill NR, Fatoba ST, Oke JL. **Global prevalence of chronic kidney disease - a systematic review and meta-analysis**. *PLoS ONE* (2016.0) **11** e0158765. DOI: 10.1371/journal.pone.0158765
2. De Nicola L, Donfrancesco C, Minutolo R. **Prevalence and cardiovascular risk profile of chronic kidney disease in Italy: results of the 2008–12 National Health Examination Survey**. *Nephrol Dial Transpl* (2015.0) **30** 806-814. DOI: 10.1093/ndt/gfu383
3. Palaka E, Grandy S, van Haalen H. **The impact of CKD anaemia on patients: incidence, risk factors, and clinical outcomes: a systematic literature review**. *Int J Nephrol* (2020.0) **2020** 7692376. DOI: 10.1155/2020/7692376
4. Fishbane S, Ross DW, Hong S. **Anemia in non-dialysis-dependent CKD: to treat or not to treat?**. *Am J Kidney Dis* (2019.0) **73** 297-299. DOI: 10.1053/j.ajkd.2018.11.006
5. Del Vecchio L, Locatelli F. **Investigational hypoxia-inducible factor prolyl hydroxylase inhibitors (HIF-PHI) for the treatment of anemia associated with chronic kidney disease**. *Expert Opin Investig Drugs* (2018.0) **27** 613-621. DOI: 10.1080/13543784.2018.1493455
6. Babitt JL, Lin HY. **Mechanisms of anemia in CKD**. *J Am Soc Nephrol* (2012.0) **23** 1631-1634. DOI: 10.1681/ASN.2011111078
7. **KDIGO clinical practice guideline for anemia in chronic kidney disease**. *Kidney Int Suppl* (2012.0) **2** 279-335. DOI: 10.1038/kisup.2012.40
8. Mikhail A, Brown C, Williams JA. **Renal association clinical practice guideline on anaemia of chronic kidney disease**. *BMC Nephrol* (2017.0) **18** 1-29. DOI: 10.1186/s12882-017-0688-1
9. Wong MMY, Tu C, Li Y. **Anemia and iron deficiency among chronic kidney disease Stages 3–5ND patients in the chronic kidney disease outcomes and practice patterns study: often unmeasured, variably treated**. *Clin Kidney J* (2020.0) **13** 613-624. DOI: 10.1093/CKJ/SFZ091
10. Lopes MB, Tu C, Zee J. **A real-world longitudinal study of anemia management in non-dialysis-dependent chronic kidney disease patients: a multinational analysis of CKDopps**. *Sci Rep* (2021.0) **11** 1784. DOI: 10.1038/s41598-020-79254-6
11. De Nicola L, Minutolo R, Chiodini P. **Prevalence and prognosis of mild anemia in non-dialysis chronic kidney disease: a prospective cohort study in outpatient renal clinics**. *Am J Nephrol* (2010.0) **32** 533-540. DOI: 10.1159/000321468
12. Minutolo R, Locatelli F, Gallieni M. **Anaemia management in non-dialysis chronic kidney disease (CKD) patients: a multicentre prospective study in renal clinics**. *Nephrol Dial Transpl* (2013.0) **28** 3035-3045. DOI: 10.1093/ndt/gft338
13. 13.Società Italiana di Nefrologia (2012) Linea Guida: identificazione, prevenzione e gestione della Malattia Renale Cronica nell’adulto. https://documenti.sinitaly.org/wp-content/uploads/sites/7/2017/03/LG_MRC2012.pdf. Accessed 15 Nov 2021
14. 14.Agenzia Italiana del Farmaco Piano terapeutico aifa per prescrizione ssn di eritropoietine. https://www.aifa.gov.it/documents/20142/516919/piano_terapeutico_eritropoietine__ex_nota_12__allegato_1_0.pdf. Accessed 15 Nov 2021
15. 15.Gazzetta Ufficiale Della Republica Italiana, Anno 160°, Numero 92; Giovedì, 18 aprile 2019. https://www.gazzettaufficiale.it/eli/gu/2019/04/18/92/sg/pdf. Accessed 15 Nov 2021
16. 16.Varrella S (2020) Population in Italy 2019, by age group. https://www.statista.com/statistics/789270/population-in-italy-by-age-group/. Accessed 22 Jul 2021
17. Gambaro G, Yabarek T, Graziani MS. **Prevalence of CKD in Northeastern Italy: results of the INCIPE study and comparison with NHANES**. *Clin J Am Soc Nephrol* (2010.0) **5** 1946-1953. DOI: 10.2215/CJN.02400310
18. Goldberg I, Krause I. **The role of gender in chronic kidney disease**. *Eur Med J* (2016.0) **1** 58-64. DOI: 10.33590/emj/10312319
19. Shiferaw WS, Akalu TY, Aynalem YA. **Risk factors for anemia in patients with chronic renal failure: a systematic review and meta-analysis**. *Ethiop J Health Sci* (2020.0) **30** 829-842. DOI: 10.4314/ejhs.v30i5.23
20. Minutolo R, Provenzano M, Chiodini P. **New-onset anemia and associated risk of ESKD and death in non-dialysis CKD patients: a multicohort observational study**. *Clin Kidney J* (2022.0) **15** 1120-1128. DOI: 10.1093/ckj/sfac004
21. Cozzolino M, Bolasco P, Ronco C. **Clinical management of chronic kidney disease patients in Italy: results from the IRIDE study**. *Nephron* (2018.0) **140** 39-47. DOI: 10.1159/000490769
22. De Nicola L, Chiodini P, Zoccali C. **Prognosis of CKD patients receiving outpatient nephrology care in italy**. *Clin J Am Soc Nephrol* (2011.0) **6** 2421-2428. DOI: 10.2215/CJN.01180211
23. El-Achkar TM, Ohmit SE, McCullough PA. **Higher prevalence of anemia with diabetes mellitus in moderate kidney insufficiency: the Kidney Early Evaluation Program**. *Kidney Int* (2005.0) **67** 1483-1488. DOI: 10.1111/j.1523-1755.2005.00226.x
24. Choi HS, Han K-D, Jung J-H. **The risk of end-stage renal disease in systemic lupus erythematosus**. *Medicine (Baltimore)* (2019.0) **98** e16420. DOI: 10.1097/MD.0000000000016420
25. Kato S, Chmielewski M, Honda H. **Aspects of immune dysfunction in end-stage renal disease**. *Clin J Am Soc Nephrol* (2008.0) **3** 1526-1533. DOI: 10.2215/CJN.00950208
26. Locatelli F, Bárány P, Covic A. **Kidney disease: improving global outcomes guidelines on anaemia management in chronic kidney disease: a European Renal Best Practice position statement**. *Nephrol Dial Transpl* (2013.0) **28** 1346-1359. DOI: 10.1093/ndt/gft033
27. St Peter WL, Guo H, Kabadi S. **Prevalence, treatment patterns, and healthcare resource utilization in Medicare and commercially insured non-dialysis-dependent chronic kidney disease patients with and without anemia in the United States**. *BMC Nephrol* (2018.0) **19** 67. DOI: 10.1186/s12882-018-0861-1
28. Adibuzzaman M, DeLaurentis P, Hill J, Benneyworth BD. **Big data in healthcare the promises, challenges and opportunities from a research perspective: a case study with a model database**. *AMIA Annu Symp Proc* (2018.0) **2017** 384-392. PMID: 29854102
29. **KDIGO 2012 clinical practice guideline for the evaluation and management of chronic kidney disease**. *Kidney Int Suppl* (2013.0) **3** 4477-4483
30. Kim M, Chae KH, Chung YJ. **The effect of the look-back period for estimating incidence using administrative data**. *BMC Health Serv Res* (2020.0) **20** 166. DOI: 10.1186/s12913-020-5016-y
|
---
title: 'Maternal factors associated with iron deficiency without anaemia in early
pregnancy: ECLIPSES study'
authors:
- Lucía Iglesias-Vázquez
- Mercedes Gimeno
- Pilar Coronel
- Ida Henriette Caspersen
- Josep Basora
- Victoria Arija
journal: Annals of Hematology
year: 2023
pmcid: PMC9998312
doi: 10.1007/s00277-023-05123-7
license: CC BY 4.0
---
# Maternal factors associated with iron deficiency without anaemia in early pregnancy: ECLIPSES study
## Abstract
Several population-specific genetic, sociodemographic, and maternal lifestyle factors are related to iron status in early pregnancy, and their identification would allow preventive actions to be taken. The study aimed to identify maternal factors associated with iron deficiency (ID) in early pregnancy in non-anaemic pregnant women from a European Mediterranean country. Cross-sectional study using the initial population of the ECLIPSES study performed in non-anaemic pregnant women before gestational week 12. Serum ferritin (SF) and haemoglobin concentrations were measured to evaluate iron status, and ID was defined as SF < 15 µg/L. Several sociodemographic and lifestyle data were recorded and used as covariates in the multivariate-adjusted regression models. Out of the 791 participants, $13.9\%$ had ID in early pregnancy. Underweight (OR 3.70, $95\%$CI 1.22, 15.53) and parity (1 child: OR 2.03, $95\%$CI 1.06, 3.88; ≥ 2 children: OR 6.96, $95\%$CI 3.09, 15.69) increased the odds of ID, while a high intake of total meat (≥ 108.57 g/day: OR 0.37, $95\%$CI 0.15, 0.87), red/processed meat (≥ 74.29 g/day: OR 0.70, $95\%$CI 0.35, 0.98), protein (≥ 65.05 g/day: OR 0.85, $95\%$CI 0.30, 0.99), and dietary iron (≥ 8.58 mg/day: OR 0.58, $95\%$CI 0.35, 0.94) protected against it. Smoking was also associated with a reduction in ID odds (OR 0.34, $95\%$CI 0.12, 0.99). Baseline BMI, parity, smoking, and diet are associated with ID in early pregnancy in non-anaemic women. Pregnancy planning policies should focus on women at higher risk of ID, such as those who are underweight, multiparous, or following vegetarian diets. This clinical trial was registered at www.clinicaltrialsregister.eu as EudraCT number 2012–005,480-28 and at www.clinicaltrials.gov with identification number NCT03196882.
### Supplementary Information
The online version contains supplementary material available at 10.1007/s00277-023-05123-7.
## Introduction
Maternal iron status during pregnancy is a public health concern given the high prevalence of anaemia and iron deficiency (ID) during the gestational period. Estimates indicate that around $25\%$ of pregnant women worldwide suffer from anaemia, mainly caused by ID [1]. Anaemia poses significant problems for maternal and child health [2–4], so haemoglobin levels are routinely monitored to detect it preventively. However, there is a high percentage of pregnant women presenting ID without anaemia who are not diagnosed and monitored in daily clinical practice, as serum ferritin (SF) is not routinely measured [5, 6]. In Europe, 10–$33\%$ of pregnant women have ID [7], such as in our study where we have previously observed $14\%$ of participants with ID in early pregnancy [8]. It is therefore of great importance among childbearing women to maintain an adequate iron balance, given the role of iron in multiple physiological processes that take place during pregnancy. Wide evidence supports that prenatal iron imbalances also in absence of anaemia can be detrimental to mother and child. In this regard, previous studies have found that not only anaemia but also ID during the gestational period is associated with preeclampsia, premature births, and even miscarriages [4, 9], as well as physical and cognitive developmental delays in children in postnatal life [3, 10–13].
It is worth mentioning that maternal iron levels in early pregnancy by themselves strongly influence the progression of iron stores during gestation and a woman’s iron status at the end of pregnancy. Indeed, in the ECLIPSES study, we already found a positive association of iron-related biomarker concentrations between the first and third trimester of gestation [8]. Previous studies reached similar findings when assessing the correlation of iron status between early and late pregnancy [14, 15].
Several studies in developed countries have identified some maternal factors that influence the initial iron status of pregnant women, such as sociodemographic, genetic, and lifestyle characteristics [16–19]. However, not all factors that are related have always been analysed together. Many of the risk factors for ID are population-specific, and it is therefore important to disentangle which maternal characteristics place women at risk for ID in each population group. In this regard, few studies have assessed factors associated with iron status in non-anaemic pregnant women.
This study aimed to identify maternal factors associated with ID in early pregnancy in a sample of non-anaemic pregnant women, although with a moderate prevalence of ID, from a European Mediterranean country.
## Study design and population
The present work included the baseline population of the ECLIPSES study, a community-based randomized controlled trial (RCT) conducted in the province of Tarragona (Catalonia, Spain) [20], before starting the intervention. Briefly, the participants were recruited by midwives in their primary care centres during the routine obstetrical visits prior to gestational week (GW) 12. The main inclusion criteria were over 18 years old and not having anaemia (haemoglobin [Hb] ≥ 110 g/L). Women who had taken > 10 mg iron daily during the 3 months prior to GW12 were excluded.
## Outcome
The study outcome was the iron status of women in early pregnancy and its associated factors. For this, concentrations of iron-related biomarkers (Hb and SF) were measured. Since having anaemia was an exclusion criterion, only ID defined as SF < 15 µg/L according to the WHO guidelines was considered [21].
## Data collection
The research staff recorded the sociodemographic and lifestyle data of the participants during a personal interview using specific questionnaires, including maternal age, baseline body mass index (BMI), smoking habit, ethnicity, parity, pregnancy planning, and use of hormonal contraceptives. The educational level and occupational status of women and their partners were also registered. The family’s socioeconomic status (SES) was calculated from the sociodemographic data of participants and their partners, including educational level and occupational status. Dietary assessment was done using a short food frequency questionnaire (FFQ) validated in our population [22]. Food groups assessed included total meat, red and processed meat, fish, fruits, vegetables, legumes, and dairy products as grams per day (g/day). From this information, energy intake (kcal/day) and nutrients (g/day or mg/day) were calculated using the REGAL (Répertoire Général des Aliments) food composition table [23], complemented by a Spanish food composition table [24]. As for the nutrient intake, protein, fibre, vitamin C, calcium, and dietary iron were assessed. Detailed information is available in Aparicio et al. [ 25]. Information from the FFQ allowed us to calculate the percentage of adherence to the Mediterranean diet, considered a high–quality dietary pattern [25, 26]. Extended information on data collection can be found elsewhere [8, 20].
Blood samples were taken on GW12 to perform blood and genetic tests. Haematological parameters (Hb and MCV), some specific biochemical markers (SF and C-reactive protein [CRP]), and genetic mutations of the HFE gene (C282Y, H63D and S65C) were performed.
## Statistical analyses
Continuous variables (mean and SD) were described using Student’s t-test and ANOVA test, while the chi-squared test was used for categorical variables (percentages). Natural logarithm (Ln) transformation was applied to normalize the distribution of SF, increasing the validity of analyses, and using the median and interquartile ranges (IQR). Multivariate regression models (multiple linear regressions and logistic regressions) were used to assess the effect of different prenatal predictors on maternal iron status in early pregnancy. The regression models were adjusted for a wide range of potential confounders, described in the bivariate analyses, including age (< 25 years, 25–35 years, and > 35 years), baseline BMI (underweight, BMI < 18.5; normal weight, BMI 18.5–24.9; overweight, BMI 25–29.9; and obesity, BMI ≥ 30), smoking habit (yes or not), SES (low or middle-high), ethnicity (Caucasian, Latin American, Arab, and Black), parity (primiparous, 1 child, or ≥ 2 children), pregnancy planning (yes or no), use of hormonal contraceptives (yes or no), HFE genotype (WT/WT, C282Y/WT, H63D carrier, and S65C carrier), and dietary intake expressed as quartiles. Daily dietary consumption of food groups (total meat, red and processed meat, fish, fruits, vegetables, legumes, and dairy products) and nutrients (protein, fibre, vitamin C, calcium, and iron) were separately included in the regression models to avoid over-adjustment. Daily energy intake as kcal was included in both models. Given that SF can raise in infectious or inflammatory processes, the regression model for SF concentration was additionally adjusted for CRP levels. All statistical analyses were performed using SPSS (version 25.0 for Windows; SPSS Inc., Chicago, IL, USA) and statistical significance was set at $p \leq 0.05.$
## Results
The study included 791 pregnant women, with a median age of 31 (17–46) years and a median gestational age of 12 weeks at the assessment. Sociodemographic and lifestyle characteristics were presented in Table 1. Regarding the body mass index (BMI), near of $60\%$ of participants had normal weight and more than $40\%$ had excess weight, including overweight and obesity. Most of them were Caucasian ($80.4\%$) and belonged to a middle SES ($67\%$), almost $18\%$ were smokers at the recruitment, and $18.3\%$ reported having used hormonal contraceptives before getting pregnant. For $50.4\%$ of the participants, this was their first pregnancy, and $80\%$ had planned to become pregnant. As for the HFE genotype, $33.1\%$ had some mutation, the H63D/WT ($26.1\%$) and C282Y/WT ($3.5\%$) being the most represented genotypes. The less represented genotypes were considered together, leaving the following categories: “WT/WT,” “C282Y/WT,” “H63D carrier” (including C282Y/H63D, H63D/ H63D), and “S65C carrier” (including H63D/S65C, S65C/S65C). In relation to educational level, almost $30\%$ accounted for higher or vocational education, while less than $5\%$ reported having unfinished primary school. Median dietary iron intake was 8 mg/day, and adherence to the Mediterranean diet was high in almost all the participants in the study. A strong statistically significant correlation (Spearman rho 0.915, $p \leq 0.001$) was found between energy (kcal) and iron intake. Compared to women of normal weight, underweight women reported lower intakes of energy (1827.96 and 1724.57 kcal/day, respectively) and iron (7.97 and 7.40 mg/day, respectively), although the difference was not statistically significant (data not shown).Table 1Baseline characteristics of the population in the study ($$n = 791$$)Age, years31 [17–46]Maternal ethnic origin < 2514.7 Caucasian80.4 25–3563.3 Latin American10.7 > 3522.0 Arab6.3Gestational age, weeks12 [8–12] Black2.0 Baseline BMI25.05 (4.50) Asian0.6 Underweight1.6Education Normal weight57.8 Unfinished primary school4.6 Overweight26.4 Primary school28.4 Obesity14.2 Secondary school38.3Smoking habit (yes)17.8 Higher/vocational education28.7 Use of hormonal contraceptives (yes)18.3Familiar SES Parity Low16.2 Primiparous50.4 Middle67.0 1 child37.6 High16.8 ≥ 2 children11.9Food intakePregnancy planning (yes)80.0 Total meat (g/d)92 [0–314] HFE genotype Red and processed meat (g/d)56 [0–239] WT/WT66.9 Fish (g/d)45 [0–179] H63D/WT26.1 Fruits (g/d)244 [0–929] C282Y/WT3.5 Vegetables (g/d)78 [0–441] S65C/WT1.3 Legumes (g/d)15 [0–60] C282Y/H63D1.0 Dairy products (g/d)294 [0–1530] H63D/H63D1.0Energy (kcal/d)1788 [853–4290] H63D/S65C0.2 Nutrient intake S65C/S65C0.2 Protein (g/d)56 [10–167] C282Y/ C282Y0.0 Fibre (g/d)13 [3–35] C282Y/ S65C0.0 Vitamin C (mg/d)77 [3–280] Calcium (mg/d)663 [55–2123] Iron (mg/d)8 [2–24]Data are expressed as mean (SD), median [min–max] and %BMI body mass index, WT wild type, SES socioeconomic status Concentrations of SF and Hb, as well as the percentage of women with ID in early pregnancy, were described according to genetic, sociodemographic and lifestyle characteristics (Supplementary Table 1) and by quartiles of dietary intake (Supplementary Table 2). From the overall sample, $13.9\%$ had ID in early pregnancy. Both SF and Hb concentrations increased across increasing BMI categories. In addition, smokers and primiparous women showed higher SF concentrations and a lower percentage of ID than their counterparts early in pregnancy. As for dietary intake, SF concentrations increased progressively across quartiles of total meat and iron intake, whereas a U-shaped distribution was observed for red and processed meat intake, and an inverse U-shaped for calcium intake by quartiles.
Multivariate adjusted analyses showed the effect of prenatal sociodemographic and lifestyle factors on SF and Hb concentrations, and ID at GW12 (Table 2). Parity and being underweight were negatively associated with maternal Hb and SF concentrations, whereas a high intake of total meat (≥ 108.57 g/day), and red and processed meat (≥ 74.29 g/day) increased them in early pregnancy. Additionally, young maternal age (< 25 years) reduced and smoking increased SF concentrations at GW12 but did not show any effect on Hb levels. In relation to ID, smoking, and a high intake of total meat, and red and processed meat, reduced its odds by $66\%$, $63\%$ and $30\%$, respectively. Consistent with the observed effect on SF levels, underweight and non-parous women were more likely to suffer from ID in early pregnancy compared to their peers. As for dietary intake, complementarily, when the models were adjusted for daily nutrient intake instead of food groups consumption, a high intake of protein (≥ 65.05 g/day) and iron (≥ 8.58 mg/day) was positively associated with Hb and SF concentrations in early pregnancy, as well as with a reduction in the percentage of ID.Table 2Associations between early pregnancy concentrations of haemoglobin and serum ferritin and iron deficiency, and predictor variablesHaemoglobin (g/L)Serum ferritin (µg/L)ID (SF < 15 µg/L)β ($95\%$CI)OR ($95\%$CI)Age (years)a < 25 − 0.28 (− 2.68, 2.11) − 0.24 (− 0.45, − 0.03)*1.71 (0.70, 4.18) 25–350.00 (Ref.)0.00 (Ref.)1.00 (Ref.) > 351.62 (− 0.16, 3.40) − 0.14 (− 0.30, 0.02)1.06 (0.56, 2.03)Smoking No0.00 (Ref.)0.00 (Ref.)1.00 (Ref.) Yes1.17 (− 0.94, 3.27)0.08 (0.10, 0.27)*0.34 (0.12, 0.99)*Baseline BMI (kg/m2) < 18.5 − 0.70 (− 2.59, − 0.05)* − 0.35 (− 0.90, − 0.20)*3.70 (1.22, 15.53)* 18.5–24.90.00 (Ref.)0.00 (Ref.)1.00 (Ref.) 25–29.90.89 (− 0.92, 2.71) − 0.04 (− 0.21, 0.12)0.85 (0.43, 1.67) ≥ 301.81 (− 0.44, 4.06)0.02 (− 0.19, 0.22)0.80 (0.33, 1.91)Parity Primiparous0.00 (Ref.)0.00 (Ref.)1.00 (Ref.) 1 child − 2.33 (− 3.99, − 0.68)* − 0.17 (− 0.31, − 0.02)*2.03 (1.06, 3.88)* ≥ 2 children − 2.76 (− 4.78, − 0.46)* − 0.48 (− 0.71, − 0.25)*6.96 (3.09, 15.69)*Total meat intake (g/d) < 68.570.00 (Ref.)0.00 (Ref.)1.00 (Ref.) 68.57–91.730.66 (− 1.65, 2.97)0.12 (− 0.08, 0.31)0.53 (0.25, 1.12) 91.74–108.560.06 (− 2.14, 2.26)0.12 (− 0.07, 0.30)0.47 (0.22, 1.02) ≥ 108.572.48 (0.25, 4.72)*0.20 (0.15, 0.33)*0.37 (0.15, 0.87)*Red and processed meat (g/d) < 37.140.00 (Ref.)0.00 (Ref.)1.00 (Ref.) 37.14–55.990.75 (− 1.62, 3.13)0.04 (− 0.17, 0.24)0.42 (0.14, 1.28) 56.00–74.280.30 (− 2.62, 2.03)0.01 (− 0.19, 0.21)0.55 (0.24, 1.29) ≥ 74.291.02 (0.65, 3.91)*0.16 (0.08, 0.41)*0.70 (0.35, 0.98)*Age (years)b < 25 − 0.28 (− 2.23, 2.78) − 0.24 (− 0.45, − 0.03)*1.62 (0.66, 4.01) 25–350.00 (Ref.)0.00 (Ref.)1.00 (Ref.) > 350.86 (− 1.00, 2.72) − 0.14 (− 0.30, 0.02)1.11 (0.57, 2.14)Smoking No0.00 (Ref.)0.00 (Ref.)1.00 (Ref.) Yes0.82 (− 1.37, 3.02)0.08 (0.10, 0.27)*0.33 (0.11, 0.98)*Parity Primiparous0.00 (Ref.)0.00 (Ref.)1.00 (Ref.) 1 child − 2.30 (− 4.03, − 0.58)* − 0.17 (− 0.31, − 0.02)*2.20 (1.13, 4.28)* ≥ 2 children − 1.23 (− 3.96, 1.50) − 0.48 (− 0.71, − 0.25)*7.39 (3.10, 17.59)*Protein (g/d) < 48.100.00 (Ref.)0.00 (Ref.)1.00 (Ref.) 48.10–55.920.34 (− 0.85, 1.07)0.12 (− 0.09, 0.33)0.84 (0.31, 2.27) 55.93–65.040.85 (0.52, 2.01)*0.19 (0.01, 0.37)*0.70 (0.22, 0.96)* ≥ 65.051.06 (0.87, 2.65)*0.25 (0.10, 0.31)*0.85 (0.20, 0.99)*Iron (mg/d) < 6.310.00 (Ref.)0.00 (Ref.)1.00 (Ref.) 6.31–7.670.28 (− 0.45, 1.65)0.06 (− 0.20, 0.15)0.89 (0.64, 1.98) 7.68–8.570.94 (− 0.67, 1,96)0.26 (− 0.39, 0.42)0.62 (0.50, 1.32) ≥ 8.581.13 (0.21, 2.98)*0.30 (0.12, 0.73)*0.58 (0.35, 0.94)*ID iron deficiency, SF serum ferritin, BMI body mass index*Statistically significant differences compared to the reference groupaAdjusted for age (< 25, 25–35, > 35), baseline BMI (< 18.5, 18.5–24.9, 25–29.9, ≥ 30), smoking habit (yes/no), SES (low, middle-high), ethnicity (Caucasian, Latin American, Arab, Black), parity (primiparous, 1 child, ≥ 2 children), pregnancy planning (yes/no), use of hormonal contraceptives (yes/no), HFE genotype (WT/WT, C282Y/WT, H63D carrier, S65C carrier), and daily dietary intake (kcal, total meat, red and processed meat, fish, fruits, vegetables, legumes, dairy products)bAdjusted for age (< 25, 25–35, > 35), baseline BMI (< 18.5, 18.5–24.9, 25–29.9, ≥ 30), smoking habit (yes/no), SES (low, middle-high), ethnicity (Caucasian, Latin American, Arab, Black), parity (primiparous, 1 child, ≥ 2 children), pregnancy planning (yes/no), use of hormonal contraceptives (yes/no), HFE genotype (WT/WT, C282Y/WT, H63D carrier, S65C carrier), and daily dietary intake (kcal, protein, fibre, vitamin C, calcium, iron)
## Discussion
This study contributes to the identification of some maternal sociodemographic and lifestyle factors associated with the risk of developing ID in early pregnancy in a sample of non-anaemic pregnant women from northeastern Spain, with the aim of preventing this deficiency, which is so common during pregnancy, and to avoid its negative consequences.
From all the analysed maternal biological, sociodemographic, and lifestyle conditions, the most relevant predictor of iron status in early pregnancy identified in this work was multiparity, showing almost 7 times higher odds of ID than primiparous women. That has already been repeatedly reported from around the world [16, 17, 27, 28] and the main explanation is that the high iron cost of pregnancy puts multiparous women at risk of not recovering iron stores from one pregnancy to the next one [29]. Another important predictor of ID in early pregnancy was being underweight, in accordance with previous findings [16, 18, 30, 31]. This association would be indirectly reflecting the effect of poor nutritional status, with low food and nutrient intake. Additionally, underweight women in our study reported lower energy and iron intake than those of normal weight, reinforcing the proposed argument. Other of the studied maternal factors showed a minor impact on the women’s iron status. This is the case of age which, although previous studies have associated younger age with a higher risk of ID [17, 32], being a young mother led to a decrease in SF concentration but did not influence the likelihood of ID in our case. As for the smoking habit, smokers usually show higher levels of SF than non-smokers [19] and, therefore, apparently decrease the odds of ID, as has been found in the present study. According to scientific evidence, cigarette smoking disrupts iron homeostasis inducing a systemic accumulation [19, 33, 34], which leads to the detection of an increase in iron reserves.
This study also assessed dietary intake and its relationship with iron levels. Thus, it was expected that a higher daily intake of total, red and processed meat, as well as protein and iron, would increase both Hb and SF concentrations in early pregnancy, protecting against ID. However, a concern arising from our results is the low iron intake that women reported (median: 8 mg/day), with only 8 participants ($1\%$) meeting the DRI (16 mg/day), and $35.5\%$ not even reaching the EAR (7 mg/day) that EFSA indicates for pregnant women [35]. However, according to a recent review [36], that is not an isolated problem, but the dietary iron intake of most pregnant women in *Europe is* well below the recommendations. Finally, it has to be stated that, contrary to expected, no effect was found of mutated HFE genotypes on iron-related biomarker concentrations or ID in early pregnancy. Despite HFE mutations being common in the European population, they are less frequent in Southern Europe [37, 38]. Especially, the variant with high clinical penetrance, HFE C282Y homozygous [39, 40], is absent in our study population which may preclude observing some influence of HFE mutations on women’s iron status.
A high percentage of women with ID in early pregnancy show no signs of anaemia and their low iron stores go undetected and untreated since *Hb is* often the only biomarker measured for assessing iron status in routine practice [5, 6]. It is worth mentioning that *Hb is* the biomarker that is altered the latest when iron status is assessed, as it is only altered when iron stores are already nil and erythropoietic synthesis is unable to synthesise Hb in the required amount. We must therefore consider Hb to be an ineffective biomarker for the prevention of ID during pregnancy, where a decrease in iron levels as pregnancy progresses is the norm unless preventive iron supplementation is carried out. Therefore, some women, even if they have normal Hb levels, may have a high chance of developing ID and anaemia later in pregnancy, with negative consequences for their health and that of their baby. Otherwise, measurement of SF concentrations, which is not common in clinical practice, would provide valuable information on iron stores, including incipient iron deficiency states. The concentration of SF is internationally recognized as a very robust biomarker for assessing iron reserves; when low, there is no possibility of false positives for ID and, although it is true that it may increase in presence of infectious/inflammatory processes, the analyses in our study were adjusted for CRP concentration to account for acute infections.
It must be said that deciphering which factors are associated with ID in early pregnancy does not mean that iron stores should not continue to be monitored throughout pregnancy. However, it has been observed that early iron status is highly correlated with iron status during and at the end of gestation. Therefore, it is in the early stages of pregnancy, and even periconceptionally, that effective preventive actions can be considered, such as increased promotion of healthier lifestyles from pregnancy planning services and increased monitoring of non-modifiable characteristics where appropriate. Thus, knowing which maternal conditions or characteristics may affect maternal iron stores would allow obstetricians and midwives to focus public health actions on the target population for early prevention of ID, helping women to achieve the best possible iron status.
The strengths of the present work include [1] the extensive data collection carried out, which covered many sociodemographic and lifestyle characteristics of the participants, including diet and toxic habits; [2] SF levels were adjusted for CRP concentrations in multivariate analyses allowing control for acute inflammatory processes at the time of measurements that could bias the results. However, some limitations must be considered when interpreting these findings. First, the observational design of the study may influence the external validity of the results. On the assumption that the sample includes only non-anaemic women in early pregnancy, the current findings cannot be extrapolated to other populations. Second, information on recent blood donations, which would reduce iron reserves, was not available. Third, dietary assessment using questionnaires is susceptible to misreporting bias; however, given that women with low BMI in our study did not over-report food intake, as is often the case, we believe that potential misreporting bias would not greatly influence our results. Finally, information about interpregnancy interval was not available, which could have allowed further interpretation of parity as a predictor of iron status.
## Conclusion
Since Hb concentration is very often the only biomarker of iron status used in clinical practice, pregnant women with anaemia are treated with iron supplementation, but those with ID without anaemia remain under-diagnosed and untreated, so estimating SF concentration in early pregnancy and its associated factors is of great importance for early prevention of ID. Multiparity and being underweight are strong predisposing factors of maternal ID in early pregnancy in non-anaemic women. A diet high in meat, protein, and iron reduce the likelihood of starting pregnancy with ID. Midwives and obstetricians should pay special attention to the iron status of pregnant women at high risk of ID, such as those who are underweight, multiparous or on vegetarian diets.
## Supplementary Information
Below is the link to the electronic supplementary material. Supplementary file1 (DOCX 38 kb)
## References
1. 1.World Health Organization (2015). The global prevalence of anaemia in 2011. WHO Rep 48. 10.1017/S1368980008002401
2. Smith C, Teng F, Branch E. **Maternal and perinatal morbidity and mortality associated with anemia in pregnancy**. *Obstet Gynecol* (2019.0) **134** 1234. DOI: 10.1097/AOG.0000000000003557
3. 3.Figueiredo ACMG, Gomes-Filho IS, Silva RB, et al (2018) Maternal anemia and low birth weight: a systematic review and meta-analysis. Nutrients 10. 10.3390/nu10050601
4. Rahmati S, MiladAzami PN, Sayehmiri K. **The relationship between maternal anemia during pregnancy with preterm birth: a systematic review and meta-analysis**. *J Matern Neonatal Med* (2020.0) **33** 2679-2689. DOI: 10.1080/14767058.2018.1555811
5. Teichman J, Nisenbaum R, Lausman A, Sholzberg M. **Suboptimal iron deficiency screening in pregnancy and the impact of socioeconomic status in a high-resource setting**. *Blood Adv* (2021.0) **5** 4666. DOI: 10.1182/BLOODADVANCES.2021004352
6. Al-Naseem A, Sallam A, Choudhury S, Thachil J. **Iron deficiency without anaemia: a diagnosis that matters**. *Clin Med* (2021.0) **21** 107-113. DOI: 10.7861/CLINMED.2020-0582
7. Milman N, Taylor CL, Merkel J, Brannon PM. **Iron status in pregnant women and women of reproductive age in Europe**. *Am J Clin Nutr* (2017.0) **106** 1655S-1662S. DOI: 10.3945/ajcn.117.156000
8. Iglesias Vázquez L, Arija V, Aranda N. **The effectiveness of different doses of iron supplementation and the prenatal determinants of maternal iron status in pregnant Spanish women: ECLIPSES study**. *Nutrients* (2019.0) **11** 2418. DOI: 10.3390/nu11102418
9. Lewandowska M, Sajdak S, Lubiński J. **Can serum iron concentrations in early healthy pregnancy be risk marker of pregnancy-induced hypertension?**. *Nutrients* (2019.0) **11** 1086. DOI: 10.3390/nu11051086
10. Radlowski EC, Johnson RW. **Perinatal iron deficiency and neurocognitive development**. *Front Hum Neurosci* (2013.0) **7** 585. DOI: 10.3389/fnhum.2013.00585
11. Hernández-Martínez C, Canals J, Aranda N. **Effects of iron deficiency on neonatal behavior at different stages of pregnancy**. *Early Hum Dev* (2011.0) **87** 165-169. DOI: 10.1016/j.earlhumdev.2010.12.006
12. Ribot B, Aranda N, Viteri FE. **Depleted iron stores without anaemia early in pregnancy carries increased risk of lower birthweight even when supplemented daily with moderate iron**. *Hum Reprod* (2012.0) **27** 1260-1266. DOI: 10.1093/humrep/des026
13. Vallée L. **Fer et neurodéveloppement**. *Arch Pediatr* (2017.0) **24** 5S18-5S22. DOI: 10.1016/S0929-693X(17)24005-6
14. Arija V, Ribot B, Aranda N. **Prevalence of iron deficiency states and risk of haemoconcentration during pregnancy according to initial iron stores and iron supplementation**. *Public Health Nutr* (2013.0) **16** 1371-1378. DOI: 10.1017/S1368980013000608
15. Aranda N, Ribot B, Viteri FE. **Predictors of haemoconcentration at delivery: Association with low birth weight**. *Eur J Nutr* (2013.0) **52** 1631-1639. DOI: 10.1007/s00394-012-0468-4
16. Caspersen IH, Iglesias-Vázquez L, Abel MH. **Iron status in mid-pregnancy and associations with interpregnancy interval, hormonal contraceptives, dietary factors and supplement use**. *Br J Nutr* (2021.0) **126** 1270-1280. DOI: 10.1017/S0007114521000295
17. Loy SL, Lim LM, Chan SY. **Iron status and risk factors of iron deficiency among pregnant women in Singapore: a cross-sectional study**. *BMC Public Health* (2019.0) **19** 1-10. DOI: 10.1186/S12889-019-6736-Y/TABLES/2
18. Tan J, Qi YN, He GL. **Association between maternal weight indicators and iron deficiency anemia during pregnancy: a cohort study**. *Chin Med J (Engl)* (2018.0) **131** 2566-2574. DOI: 10.4103/0366-6999.244109
19. Ghio AJ, Hilborn ED, Stonehuerner JG. **Particulate matter in cigarette smoke alters iron homeostasis to produce a biological effect**. *Am J Respir Crit Care Med* (2008.0) **178** 1130-1138. DOI: 10.1164/rccm.200802-334OC
20. Arija V, Fargas F, March G. **Adapting iron dose supplementation in pregnancy for greater effectiveness on mother and child health: protocol of the ECLIPSES randomized clinical trial**. *BMC Pregnancy Childbirth* (2014.0) **14** 33. DOI: 10.1186/1471-2393-14-33
21. 21.WHO guideline on use of ferritin concentrations to assess iron status in individuals and populations. Geneva: World Health Organization; 2020. Licence: CC BY-NC-SA 3.0 IGO.
22. Trinidad Rodríguez I, Fernández Ballart J, Cucó Pastor G. **Validation of a short questionnaire on frequency of dietary intake: reproducibility and validity**. *Nutr Hosp* (2008.0) **23** 242-252. PMID: 18560701
23. Favier JC, Ireland-Ripert J, Toque C, Feinberg M. *Répertoire Général des Aliments: Tables de Composition* (1995.0)
24. 24.Mataix J, García L, Mañas M, et al (2003) Tabla de composición de alimentos
25. 25.Aparicio E, Jardí C, Bedmar C, et al (2020) Nutrient intake during pregnancy and post-partum: ECLIPSES study. Nutrients 12. 10.3390/NU12051325
26. 26.Jardí C, Aparicio E, Bedmar C, et al (2019) Food consumption during pregnancy and post-partum. ECLIPSES study. Nutrients 11. 10.3390/NU11102447
27. Khambalia AZ, Collins CE, Roberts CL. **Iron deficiency in early pregnancy using serum ferritin and soluble transferrin receptor concentrations are associated with pregnancy and birth outcomes**. *Eur J Clin Nutr* (2016.0) **70** 358-363. DOI: 10.1038/EJCN.2015.157
28. Miller EM. **Iron status and reproduction in US women: national health and nutrition examination survey, 1999–2006**. *PLoS One* (2014.0) **9** e112216. DOI: 10.1371/journal.pone.0112216
29. Bothwell T. **Iron requirements in pregnancy and strategies to meet them**. *Am J Clin Nutr* (2000.0) **72** 257S-S264. DOI: 10.1093/ajcn/72.1.257S
30. Sumarmi S, Puspitasari N, Handajani R, Wirjatmadi B. **Underweight as a risk factor for iron depletion and iron-deficient erythropoiesis among young women in rural areas of East Java, Indonesia**. *Malays J Nutr* (2016.0) **22** 219-232
31. Agrawal S, Singh A. **Obesity or underweight—what is worse in pregnancy?**. *J Obstet Gynaecol India* (2016.0) **66** 448. DOI: 10.1007/S13224-015-0735-4
32. da Costa AG, Vargas S, Clode N, Graça LM. **Prevalence and risk factors for iron deficiency anemia and iron depletion during pregnancy: a prospective study**. *Acta Med Port* (2016.0) **29** 514-518. DOI: 10.20344/amp.6808
33. Elisia I, Lam V, Cho B. **The effect of smoking on chronic inflammation, immune function and blood cell composition**. *Sci Rep* (2020.0) **10** 1-16. DOI: 10.1038/s41598-020-76556-7
34. Zhang WZ, Butler JJ, Cloonan SM. **Smoking-induced iron dysregulation in the lung**. *Free Radic Biol Med* (2020.0) **133** 238-247. DOI: 10.1016/j.freeradbiomed.2018.07.024
35. 35.Bresson JL, Burlingame B, Dean T, et al (2015) Scientific opinion on dietary reference values for iron. EFSA Journal 13. 10.2903/j.efsa.2015.4254
36. 36.Milman N (2020) Dietary iron intake in pregnant women in Europe: a review of 24 studies from 14 countries in the period 1991–2014. J Nutr Metab 2020. 10.1155/2020/7102190
37. Merryweather-Clarke AT, Pointon JJ, Jouanolle AM. **Geography of HFE C282Y and H63D mutations**. *Genet Test* (2000.0) **4** 183-198. DOI: 10.1089/10906570050114902
38. Lucotte G, Dieterlen F. **A European allele map of the C282Y mutation of hemochromatosis: celtic versus Viking origin of the mutation?**. *Blood Cells Mol Dis* (2003.0). DOI: 10.1016/S1079-9796(03)00133-5
39. Crownover B, Covey C. **Hereditary hemochromatosis**. *Am Fam Physician* (2013.0) **87** 183-190. DOI: 10.1179/1024533213Z.000000000222
40. Grosse SD, Gurrin LC, Bertalli NA, Allen KJ. **Clinical penetrance in hereditary hemochromatosis: estimates of the cumulative incidence of severe liver disease among HFE C282Y homozygotes**. *Genet Med* (2017.0) **20** 383-389. DOI: 10.1038/gim.2017.121
|
---
title: 'Diuretic dose trajectories in dilated cardiomyopathy: prognostic implications'
authors:
- Vincenzo Nuzzi
- Antonio Cannatà
- Pierpaolo Pellicori
- Paolo Manca
- Davide Stolfo
- Caterina Gregorio
- Giulia Barbati
- Daniel I. Bromage
- Theresa McDonagh
- John G. F. Cleland
- Marco Merlo
- Gianfranco Sinagra
journal: Clinical Research in Cardiology
year: 2022
pmcid: PMC9998319
doi: 10.1007/s00392-022-02126-8
license: CC BY 4.0
---
# Diuretic dose trajectories in dilated cardiomyopathy: prognostic implications
## Abstract
### Background
For patients with heart failure, prescription of loop diuretics (LD) and of higher doses are associated with an adverse prognosis. We investigated LD dose trajectories and their associations with outcomes in patients with dilated cardiomyopathy (DCM).
### Methods
Associations between outcomes and both furosemide-equivalent dose (FED) at enrolment and change in FED in the subsequent 24 months were evaluated. According to FED trajectory, patients were classified as (i) dose↑ (FED increase by ≥ $50\%$ or newly initiated); (ii) dose↓ (FED decrease by ≥ $50\%$); (iii) stable dose (change in FED by < $50\%$); and (iv) never-users. The primary outcome was all-cause-death/heart transplantation/ventricular-assist-device/heart failure hospitalization. The secondary outcome was all-cause-death/heart transplantation/ventricular-assist-device.
### Results
Of 1,131 patients enrolled, 738 ($65\%$) were prescribed LD at baseline. Baseline FED was independently associated with outcome (HR per 20 mg increase: 1.12 [$95\%$ CI 1.04–1.22], $$p \leq 0.003$$). Of the 908 with information on FED within 24 months from enrolment, $31\%$ were never-users; $29\%$ were dose↓; $26\%$ were stable dose and $14\%$ were dose↑. In adjusted models, compared to never-users, stable dose had a higher risk of the primary outcome (HR 2.42 [$95\%$ CI 1.19–4.93], $$p \leq 0.015$$), while dose↑ had the worst prognosis (HR 2.76 [$95\%$ CI 1.27–6.03], $$p \leq 0.011$$). Results were similar for the secondary outcome. Compared to patients who remained on LD, discontinuation of LD (143, $24\%$) was associated with an improved outcome (HR 0.43 [$95\%$ CI 0.28–0.65], $p \leq 0.001$).
### Conclusions
In patients with DCM, LD use and increasing FED are powerful markers of adverse outcomes. Patients who never receive LD have an excellent prognosis.
### Graphical abstract
Among 1131 DCM patients $65\%$ received loop diuretics at enrolment (upper left side). The bar chart on the upper right side shows the categorization in never-users/ dose↓/stable dose/ dose↑ over 24 months of follow-up. At the bottom is reported on the left side of each panel (observation period) the trajectory of LD dose in the four groups (left panel) and in patients who have their LD suspended vs those who continue LD (right panel) in the first two years. On the right side of each panel is shown the incidence of primary outcomes during the subsequent follow-up in the subgroups (outcome assessment)
### Supplementary Information
The online version contains supplementary material available at 10.1007/s00392-022-02126-8.
## Introduction
Loop diuretics (LD) are a cornerstone of treatment for heart failure (HF) to manage symptoms and signs of congestion, but their effect on long-term outcomes is controversial [1–3]. Observational studies suggest that higher doses of LD are associated with worse outcomes, perhaps because LD use and dose reflect the severity of congestion and cardiac dysfunction [4, 5]. However, LD might also accelerate disease progression by reducing the chance to up-titrate neurohormonal drugs, progressively worsen renal function and induce hypovolemia with consequent iatrogenic hypotension [6, 7].
So far, only a randomized trial including only 40 patients evaluated the effects of LD reduction in stable outpatients with chronic HF [8]. Current guidelines recommend using the lowest dose of LD to control signs and symptoms of congestion [1, 9], but the management of congestion is difficult [10]. Diuretic requirements may decline as disease-modifying therapies are introduced [11, 12] or might increase if the underlying disease progresses or complications, such as atrial fibrillation (AF), develop. In a sub-analysis of the CHAMPION trial, it has been proven that the most frequent therapy adjustment in chronic HF regards the LD dose, more frequently in terms of increased dose, suggesting that selected patients may require LD therapy intensification. Clinical experience and randomized trials suggest that discontinuing LD in carefully selected patients is feasible, safe and well tolerated [13]. On the other hand, suspending LD therapy can also cause a rapid and substantial increase in congestion in other patients [14]. Moreover, under-utilization of LD may lead to persistent congestion which might have adverse effects on cardiac remodelling and prognosis [9, 10].
Dilated cardiomyopathy (DCM) is a common cause of left ventricular ejection fraction (LVEF) impairment leading to HF with reduced (HFrEF) or mildly reduced LVEF. Recovery of LVEF is more common in patients with DCM than HFrEF due to other causes. [ 15]. LD therapy may no longer be required if cardiac function and congestion improve substantially, although the clinical course of DCM may be influenced by several other factors, such as the development of left bundle branch block (LBBB), AF and renal dysfunction and may also be truncated by sudden death [11]. Regular follow-up, with effective implementation of medical and device therapy, generally improves cardiac function and survival [11]. However, the trajectories of LD prescription in DCM were not previously explored and, therefore, the clinical associations and the prognostic implications of LD dose changes are still unknown. The longitudinal evolution of clinical and instrumental findings, alongside the LD dose prescription changes, might provide relevant insights into the pathophysiology linking diuretics and outcomes. In the present study, we investigated the long-term trajectories in the dose of LD in patients with DCM, and their associations with cardiac function and prognosis.
## Study population
Consecutive outpatients with DCM, enrolled in the Trieste Heart Muscle Disease Registry between January 1st 1990 and March 7th 2019 who had a complete clinical and echocardiographic evaluation at baseline, including information on LD dose, were included in this study [16, 17]. Patients who had at least one additional clinical assessment performed within 24 (± 4) months after enrolment were considered for further analyses on the diuretic trajectory. Patients with significant coronary artery disease (> $50\%$ stenosis of an epicardial coronary artery, ruled out by coronary angiography or computed tomography), a history of significant systemic hypertension (i.e. blood pressure > $\frac{140}{90}$ mmHg), primary valve disease, active myocarditis, high alcohol intake (≥ 80 g/day [≥ 46 units per week], for at least 5 years), tachycardia-induced or peripartum cardiomyopathy, congenital heart disease or history of any advanced extra-cardiac disease (i.e. terminal cancer) associated with poor short-term prognosis were excluded [18].
Daily furosemide equivalent dose (FED) was calculated by multiplying doses of torasemide by 4 and doses of bumetanide by 40 [1, 19]. The FED was recorded at each visit, whenever possible. All beta-blockers and renin–angiotensin system inhibitors were converted into bisoprolol- and ramipril-equivalent doses, respectively (Supplementary Table S1).
Patients were classified according to the LD dose trajectory observed in the first 24 months of follow-up as follows: (i) dose↑, if LD dose was increased by at least $50\%$ compared to baseline or the patient was initiated on an LD; (ii) dose↓, if LD dose decreased by at least $50\%$ or if LD was withdrawn; (iii) never-users, if patients never received LD therapy; and (iv) stable dose, all remaining patients. For patients with multiple visits during the first 2 years, the last visit available before the 2 years of follow-up was considered to define the LD dose trajectory. Twelve patients went through both an increase and a decrease of at least $50\%$ in LD dose within the first 24 months and were not included in the analysis on follow-up diuretic trajectory. The study was approved by the institutional ethical boards and complied with the Declaration of Helsinki. All the patients provided written informed consent.
Altogether 1156 patients with DCM were enrolled; of these, 25 did not have information on the use of LD at baseline and were excluded from the present analysis (consort diagram for this study is reported in Supplementary Fig. S1). This analysis included 1131 patients. At enrolment, 738 ($65\%$) patients were receiving a LD and the median FED was 25 (IQR 25–50) mg/day; 508 ($45\%$) received 1–40 mg of FED daily, 171 ($15\%$) received 41–80 mg of FED/day and 59 ($5\%$) received higher doses (Supplementary Table S2). Compared to never-users, those taking doses of LD > 80 mg/day were older, more likely to be men and to have AF and more severe symptoms (New York Heart Association class III/IV). They also had a lower LVEF, a larger left atrium (LA), were more likely to have a restrictive filling pattern (RFP) and were more likely to have moderate or severe MR.
## Echocardiography and genetic test
Echocardiograms were recorded on digital media storage at the institutions’ echocardiographic core laboratories and analysed offline by experienced echocardiographers, blinded to patient clinical information and outcomes, according to the latest available recommendations [20]. Details are reported in the supplementary text.
A subgroup of patients underwent next-generation sequencing, as previously described [20]. We classified patients in Titin-truncating variant-related DCM vs other forms of DCM, on the basis of the specific natural history of Titin-truncating variant-related DCMs [21, 22]. Additional details are provided as supplementary material.
## Study outcomes
The primary outcome was a composite of all-cause mortality, heart transplantation, durable ventricular-assist device implantation and HF hospitalization (ACD/HTx/VAD/HFH). The secondary outcome was a harder outcome including only all-cause mortality, heart transplantation and durable ventricular-assist device implantation (ACD/HTx/VAD). Information about outcomes was obtained from official reports and hospital discharge letters; for patients coming from other regions, information regarding outcomes were collected by direct contact with patients, their families or general practitioners, from the regional healthcare data warehouse and death registers. The trend of clinical and echocardiographic parameters in the four groups of patients divided according to the LD temporal trajectories was analysed in order to study the association between LD dose change and longitudinal functional parameters at follow-up (i.e. LVEF, left atrium dimension, New York Heart Association class, mitral regurgitation (MR), heart rate, beta-blockers and renin–angiotensin system inhibitors dose). No patient was lost to follow-up concerning outcomes. For this analysis, the last day of follow-up was the 7th of April 2021.
## Statistical analysis
Descriptive statistics are reported as mean and standard deviation, median and interquartile range [IQR], or counts and percentages, as appropriate. Comparisons of continuous variables were performed with cross-sectional comparisons between groups by the one-way ANOVA test, or the non-parametric Mann–Whitney U-test, when appropriate. The chi-square or Fisher exact tests were used for the comparisons of categorical variables. The trajectory of FED during the follow-up was illustrated using a smoothed conditional means analysis with penalized cubic regression splines. Most modifications of FED occurred within 24 months (Fig. 1) and consequently, the last available FED assessment in this period was used to define LD trajectory. Survival curves for the outcome measure were estimated using the Kaplan–Meier estimator and they were compared using the Log-Rank Test. Univariate and multivariable Cox regression models were fitted for the primary and secondary outcomes. For analyses of the relationship between baseline FED and outcomes, the baseline values for other variables were included in the multivariable models and the time was measured from enrolment. For analyses of the association between LD trajectories and outcomes the LD dose trajectory was treated as a time-depending variable; baseline was set at the 24th month of follow-up. Multivariable models were also taken from this assessment and the results were adjusted for variables recorded at 24 months visit. The same method was used for the assessment of the association between LD withdrawal and outcome. When there was evidence of non-linear effects between a continuous predictor and the primary outcome, the former was modelled using restricted cubic spline analyses with degrees of freedom. The variables considered in the multivariable model for adjustment were derived from univariate analysis. Variables with a statistical significance < 0.05 on univariate analysis were included in multivariable models. Baseline variables associated with diuretic withdrawal at 24-month follow-up were identified by univariate and multivariable Cox regression models. In the sensitivity analysis for the outcome, 19 patients with non-fatal events (i.e. HFH) that occurred between baseline and 2-year follow-up were excluded. The Cumulative Incidence Function was used to show the probability of LD withdrawal during follow-up, taking into account the competing risk of death. All statistical analyses were performed with IBM-SPSS (New York) version 25 and R statistical package version 3.6.2 (R Foundation, Vienna, Austria), with libraries “survival”, “ggplot2” and “splines”. Fig. 1Association of furosemide-equivalent dose at enrolment and primary outcome. Reference value was set at a FED of 20 mg/day (vertical dotted line). HR for the primary outcome is adjusted for age, sex, enrolment period, systolic blood pressure, AF, NYHA III or IV, LVEF, LVEDVI, LAESD, moderate–severe MR, restrictive filling pattern, MRA and ICD implantation
## Prognostic role of baseline loop diuretic dose
Higher baseline FED, even after adjusting for potential confounders, was associated with the primary outcome in a linear fashion (HR per 20 mg increase: 1.12 [$95\%$ CI 1.04–1.22], $$p \leq 0.003$$) (Fig. 1).
## Follow-up evolution of loop diuretic prescription
During the first 2 years of follow-up (23 [14–26] months), 908 ($80\%$) patients had at least one other clinical assessment with information on LD dose. Among them, 129 ($14\%$) patients were classified as dose↑, 234 ($26\%$) as stable dose, 263 ($29\%$) as dose↓ and 282 ($31\%$) as never-users. Among dose↑, 99 ($77\%$) had their LD dose increased by at least $50\%$, and 30 ($23\%$) were newly initiated on LD therapy (Fig. 2A). Compared to other patients, never-users were younger, were less likely to have severe symptoms, AF or LBBB, had a higher LVEF, a smaller LA, and were less likely to have moderate–severe MR or an RFP (Table 1).Fig. 2Changes, or lack of, in prescriptions of loop diuretics during the first 24 months of follow-up (908 patients) (A). Smoothed mean analysis showing the mean FED evolution during a long-term follow-up according to the trajectory in the first 24 months in the overall population (B). In green are represented never-users patients, in blue dose↓ patients, in brown stable dose patients and in red dose↑ patients; in grey are depicted the confidence intervals. The dotted lines represent the 24-month follow-up timeTable 1Baseline characteristics of the study population with at least a second assessment of loop diuretic dose within 24 months with which to define diuretic trajectory (follow-up population)908 patientsNDose↑ (129, $14\%$)NStable dose (234, $26\%$)NDose↓ (263, $29\%$)NNever-users (282, $31\%$)p-valueClinical evaluationAge, years12953 (39–63)23454 (46–63)26353 (44–62)28244 (34–56)0.002Men, no. (%) 12987 [67]234157 [67]263186 [71]282204 [72]0.547Disease duration, months1207 (1–48)2155 (1–19)2433 (1–14)22812 (3–44) < 0.001Enrollment decade, no (%) 1990–200012948 [37]23466 [28]26391 [35]282102 [36]0.507 2000–201043 [33]85 [36]92 [35]90 [32] 2010–202038 [30]83 [36]80 [30]90 [32] Heart rate, bpm12271 (60–85)22474 (65–86)25175 (64–90)27469 (60–76)0.063AF, no. % 12510 [8]21637 [17]24633 [13]25914 [5]0.004SBP, mmHg125120 (110–136)231120 (110–140)257120 (115–140)275120 (115–140)0.298NYHA III or IV, no. % 12343 [35]22472 [32]24083 [35]2720 [0] < 0.001 LBBB, no. % 12654 [43]23285 [37]26070 [27]27869 [25] < 0.001Creatinine, mmol/l11295 (82–106)20390 (80–106)24491 (80–106)20988 (77–97)0.035Sodium, mEq/l81140 (138–142)147139 (138–142)186140 (138–141)168141 (138–142)0.055Genetic testTTNtv346 [18]6713 [19]7018 [26]10115 [15]0.360Genetic negative/other variants3428 [82]6754 [81]7052 [74]10186 [85]–EchocardiographyLVEF, %12929 (22–34)23428 (23–34)26328 (22–34) 28238 (32–44)0.001LVEDVI, ml/m212997 (78–123)23491 (78–117)26393 (76–118)28277 (68–93) < 0.001LAESD, mm11743 (38–48)24842 (37–48)21743 (38–49)26237 (32–42) < 0.001Moderate or severe MR, no. (%) 12655 [44]233111 [50]255118 [46]26635 [13] < 0.001RFP, no. (%) 10235 [34]20459 [29]17057 [34]24021 [9] < 0.001Medications and device therapyACE-I or ARB or ARNI, no. (%) 129128 [99]234232 [99]263262 [100]282257 [91] < 0.001Ramipril dose equivalent, mg1275 (2.5–10)2325 (2.5–10)2625 (3.75–10)2745 (2.5–7.3)0.031Beta-blockers, no. (%) 129112 [87]234217 [93]263247 [94]282248 [88]0.026Bisoprolol dose equivalent, mg992.5 (1.25–6.25)1972.5 (2.5–5)2283.75 (2.5–7.5)2333.75 (2.5–7.5)0.080MRA, no. (%) 12963 [49]234153 [65]263164 [62]28230 [11] < 0.001Ivabradine, no. % 1294 [3]23410 [4]2635 [2]2825 [2]0.271Diuretics, no. (%) 12999 [77]234234 [100]263263 [100]2820 [0] < 0.001Furosemide dose equivalent, mg12925 (0–25)23425 (25–50)26325 (25–50)2820 (0–0) < 0.001CRT, no. % 12924 [19]23448 [21]26327 [10]28227 [10] < 0.001ICD, no. % 12948 [37]23481 [35]26373 [28]28271 [25]0.025AF atrial fibrillation, SBP systolic blood pressure, NYHA New York Heart Association, LBBB left bundle branch block, TTNtv Titin truncating variant, LVEF left ventricular ejection fraction, LVEDVI left ventricular end-diastolic volume index, LAESD left atrial end-systolic diameter, MR mitral regurgitation, RFP restrictive filling pattern, ACE-i angiotensin-converting enzyme–inhibitors, ARB angiotensin receptor blockers, ARNI angiotensin receptor neprilysin inhibitors, MRA mineralocorticoid receptors antagonists, CRT cardiac resynchronization therapy, ICD implantable cardioverter-defibrillatorp-values are estimated by the χ2 test for categorical variables; continuous variables are estimated by Student’s t-testIn bold are reported variables with significant differences among groups
## Loop diuretic dose trajectories, longitudinal echocardiographic assessment and genetic information
Substantial changes in FED were more frequent in the first 2 years of follow-up (Fig. 2B). On multivariable analysis, baseline factors associated with a reduction in FED were higher LVEF (HR per $1\%$ 1.02 [$95\%$ CI 1.01–1.04], $$p \leq 0.006$$) and lower BMI (HR per kg/m2 0.97 [$95\%$ CI 0.94–0.99], $$p \leq 0.022$$) (Supplementary Table S3).
Of the 908 patients with at least two clinical assessments, 751 ($83\%$) had an echocardiography paired with the LD dose reassessment. Substantial improvement in LVEF was observed for all patient subgroups; however, the proportion of patients who normalized their LVEF (e.g. LVEF ≥ $50\%$) during the first 2 years of follow-up was greater in dose↓ and never-users ($27\%$ and $23\%$, respectively) compared to dose↑ and stable dose ($5\%$ and $10\%$, respectively). Importantly, only dose↓ had a smaller LA compared to the initial echocardiographic assessment, which was associated with a reduction in moderate-severe MR (that declined from 35 to $10\%$). In contrast, the prevalence of moderate–severe MR at follow-up remained ≥ $40\%$ for those who intensified LD therapy. Few patients classified as dose↓ or stable dose had severe symptoms on follow-up assessment; on the other hand, $41\%$ of those with dose↑ remained severely symptomatic. Overall, doses of other treatments for HF increased during follow-up, except for beta-blockers among the dose↑, with a consequent higher heart rate compared to never users (Fig. 3).Fig. 3Baseline and 24-month follow-up assessment of the main echocardiographic characteristics, NYHA functional class and medical therapy in the four groups. 751 patients with available information on the diuretic trajectory and a paired echocardiographic at follow-up are included. p-values for the repeated measures are calculated with Mann–Whitney U-test. Ramipril equivalent dose includes the conversion of angiotensin converter enzyme inhibitors, angiotensin receptor antagonists and angiotensin neprilysin inhibitor. * p-value vs never-users < 0.05; °p-value vs dose < ↓0.05; ^p-value vs stable dose < 0.05 Finally, we observed that, despite a similar dose of FED at enrolment for patients with and without TTNtv, the LD dose was reduced over the first 2 years of follow-up only for those with TTNtv (Supplementary Fig. S2).
## Association of loop diuretics trajectories with outcome
Over a median follow-up of 122 (IQR 62–195) months, 412 ($36\%$) patients met the study primary endpoint (248 deaths, 95 heart transplantations, 14 VAD, 115 HFH). Patients who had FED increased during follow-up had the worst outcome, while the risk was lowest for never-users ($p \leq 0.001$) (Fig. 4). The results were consistent setting the FED re-evaluation at 6- and 12-month follow-up (Supplementary Fig. S3). In adjusted models, compared to never-users patients (reference), stable dose patients had a more than the two-fold greater risk of ACD/HTx/VAD/HFH (HR 2.42 [$95\%$ CI 1.19–4.93], $$p \leq 0.015$$), while dose↑ had the poorest outcome (HR 2.76 [$95\%$ CI 1.27–6.03], $$p \leq 0.011$$) (Table 2). Similar results were observed for the secondary composite outcome (Supplementary Fig. S4). In sensitivity analysis, associations of LD trajectories with outcomes did not change when patients with non-fatal events (i.e. HFH, 20 patients) prior to the follow-up evaluation were excluded (Supplementary Fig. S5A, B).Fig. 4Kaplan–*Meier analysis* for the primary outcome according to the loop diuretics trajectory during the first 24 months. On the left side (observation period) is shown the trajectory of the FED in the four groups in the first 2 years. On the right side is shown the incidence of primary outcome during the subsequent follow-up in the subgroups (outcome assessment). HFH occurring before the follow-up visit was not considered. In green are represented never-users patients, in blue dose↓ patients, in brown stable dose patients and in red dose↑ patients. In the table are shown the unadjusted (left table) and adjusted (right table) HR, CI and p-values. Details regarding the multivariable model are reported in the textTable 2Time-depending univariate and multivariable analysis for primary outcome (ACD/HTx/VAD/HFH) considering the trajectory of diuretic dose over 24 monthsUnivariate, HR ($95\%$ CI), p-valueMultivariable, HR ($95\%$ CI), p-valueClinical evaluationAge, per year1.020 (1.012–1.028), $p \leq 0.0011.003$ (0.988–1.019), $$p \leq 0.670$$BMI per Kg/m20.991 (0.966–1.016), $$p \leq 0.462$$Men1.426 (1.111–1.832), $$p \leq 0.0051.950$$ (1.134–3.352), $$p \leq 0.016$$Disease duration per month1.002 (0.998–1.005), $$p \leq 0.326$$Enrollment decade (compared to 1990–2000)ReferenceReference 2000–20100.713 (0.547–0.929), $$p \leq 0.0120.984$$ (0.518–1.871), $$p \leq 0.961$$ 2010–20200.466 (0.306–0.710), $p \leq 0.0010.429$ (0.199–0.922), $$p \leq 0.030$$Heart rate per bpm2.889 (1.939–4.303), $p \leq 0.0011.000$ (0.977–1.025), $$p \leq 0.972$$AF1.825 (1.362–2.446), $p \leq 0.0011.287$ (0.536–3.087), $$p \leq 0.572$$SBP per mmHg0.992 (0.984–0.999), $$p \leq 0.0330.998$$ (0.987–1.009), $$p \leq 0.742$$NYHA III or IV4.387 (3.214–5.989), $p \leq 0.0011.136$ (0.328–3.930), $$p \leq 0.840$$LBBB1.184 (0.943–1.488), $$p \leq 0.146$$Creatinine per mmol/l1.001 (0.999–1.002), $$p \leq 0.222$$Diuretics trajectory (compared to never-users group)ReferenceReference Dose↓1.639 (1.118–2.263), $$p \leq 0.0031.554$$ (0.878–2.750), $$p \leq 0.130$$ Stable dose2.885 (2.114–3.936), $p \leq 0.0012.423$ (1.190–4.932), $$p \leq 0.015$$ Dose↑4.060 (2.905–5.675), $p \leq 0.0012.764$ (1.266–6.031), $$p \leq 0.011$$EchocardiographyLVEF per %0.944 (0.933–0.955), $p \leq 0.0010.971$ (0.950–0.992), $$p \leq 0.008$$IVS per mm0.980 (0.927–1.035), $$p \leq 0.468$$LAESD per mm1.069 (1.051–1.087), $p \leq 0.0011.045$ (1.011–1.080), $$p \leq 0.008$$Moderate or severe MR2.443 (1.861–3.208), $p \leq 0.0012.254$ (1.427–3.559), $p \leq 0.001$RFP2.757 (1.858–4.091), $p \leq 0.0011.602$ (0.872–2.944), $$p \leq 0.129$$Medications and device therapy ACE-i/ARB/ARNI1.185 (0.663–2.117), $$p \leq 0.567$$ Ramipril/ARNI dose equivalent per mg1.009 (0.989–1.028), $$p \leq 0.387$$ Beta-blockers0.986 (0.651–1.492), $$p \leq 0.945$$ Bisoprolol dose equivalent per mg0.978 (0.948–1.009), $$p \leq 0.163$$ MRA1.846 (1.422–2.396), $p \leq 0.0011.011$ (0.564–1.810), $$p \leq 0.972$$ Ivabradine1.874 (0.826–4.253), $$p \leq 0.133$$ Furosemide dose equivalent per 20 mg1.127 (1.106–1.150), $p \leq 0.0010.980$ (0.871–1.103), $$p \leq 0.705$$ CRT1.126 (0.838–1.513), $$p \leq 0.432$$ ICD1.290 (1.031–1.614), $$p \leq 0.0261.045$$ (0.692–1.578), $$p \leq 0.833$$All the variables are measured at 24 monthsBM body mass index, AF atrial fibrillation, SBP systolic blood pressure, NYHA New York Heart Association, LBBB left bundle branch block, LVEF left ventricular ejection fraction, IVS interventricular septum, LAESD left atrial end-systolic diameter, MR mitral regurgitation, RFP restrictive filling pattern, ACE-i angiotensin-converting enzyme–inhibitors, ARB angiotensin receptor blockers, ARNI angiotensin receptor neprilysin inhibitors, MRA mineralocorticoid receptors antagonists, CRT cardiac resynchronization therapy, ICD implantable cardioverter-defibrillatorIn bold are reported variables with significant association with the outcome at multivariable analysis
## Loop diuretics therapy withdrawal
Of the 608 patients who were taking LD at baseline, 143 ($24\%$) had their LD stopped within 24 months; a further 49 ($8\%$) patients had LD withdrawn during longer follow-up (Supplementary Fig. S6). Compared to those who had their LD withdrawn, those who continued LD were older, had a lower LVEF, a larger LA and were more likely to have moderate–severe MR, RFP and AF (Supplementary Table S4). On multivariable analysis, younger age, absence of moderate–severe MR and higher doses of beta-blockers were independently associated with stopping LD therapy within 24 months (Supplementary Table S5). LD therapy withdrawal within 24 months was associated with a reduced risk of the primary (HR 0.37 [$95\%$ CI 0.29–0.48], $p \leq 0.001$) and secondary (HR 0.34 [$95\%$ CI 0.26–0.45], $p \leq 0.001$) outcome (Supplementary Fig. S7).
## Discussion
In a large cohort of patients with DCM managed at a national tertiary-care center, use, higher daily dose and increasing the dose of LD all identified a greater risk of adverse outcomes. For patients who were not prescribed or discontinued LD, the prognosis was excellent (Graphic abstract). Changes in LD dose were more frequent during the 24 months and these modifications were associated with consistent changes in cardiac structure and function. For the subset of patients with genetic sequencing, only those with TTNtv-truncating variant-related DCM reduced their doses of diuretics during follow-up, perhaps suggesting a greater response to anti-HF therapy.
The principal therapeutic goals of managing DCM are prolonging life, control of symptoms, maintenance or improvement in quality of life and reduction in disability and morbidity. These goals can be achieved by implementing guideline-recommended pharmacological, device therapy, diuretics, and advices on life style [1, 18]. In our cohort, more than one-third of patients were not receiving LD at enrolment, which is perhaps a slightly higher proportion than other registries that enrolled outpatients with more diverse etiologies of HF [4, 11]. Several reasons may explain these differences. For instance, family screening programs for DCM patients might have allowed early diagnosis and initiation of HF treatment in our patients, when LV dysfunction was still asymptomatic; alternatively, a diagnosis of DCM may have been triggered by an arrhythmic event rather than symptoms or signs of congestion. Finally, patients with DCM might be characterized by a lower propensity to develop fluid overload compared to other forms of HF [23].
The association between LD use and higher doses with poor prognosis in ambulatory patients with HF is well known [4, 5], even among those with mild symptoms. In a post-hoc analysis of the EMPHASIS-HF trial, the use of LD was more strongly associated with prognosis than a history of HF hospitalization or plasma natriuretic peptide concentrations [24]. An association between the use of LD and adverse cardiovascular outcomes has been reported even in the absence of a diagnosis of HF for patients with type II diabetes and for those with AF. It is possible that the diagnosis of HF is often missed because signs and symptoms of congestion are masked by the use of LD [25].
The relationship between changes in LD treatment and prognosis in patients with HF has received less attention. Using the European Society of Cardiology Heart Failure Long-Term (ESC-HF-LT) Registry, Kapelios and colleagues showed that $16\%$ of patients with all-etiologies HF had their LD dose increased and only $8\%$ had it decreased; while in our DCM cohort, up to $29\%$ of patients underwent LD down-titration [26]. They also found that LD dose escalation was associated with higher mortality but those with a reduction in LD dose had a similar outcome to those maintained on a stable dose. However, these authors used different criteria to define dose↓ or dose↑ (i.e. any change from baseline, respectively) and only included patients who were receiving an LD at enrolment.
The ReBIC-1 randomized trial demonstrated the feasibility of LD withdrawal in carefully selected outpatients with HFrEF, but did not provide information regarding long-term outcomes [13]. While neurohormonal antagonists should be continued for DCM lifelong to prevent clinical deterioration, [27], our findings suggest that when LD can be withdrawn without worsening congestion the subsequent risk of adverse cardiovascular events is low. Indeed, in this retrospective DCM population, when the treating physician considered it safe to reduce or even withdraw LD, the subsequent risk of adverse outcomes was significantly reduced. On the other hand, patients that may benefit from LD dose reduction have to be carefully selected, as in $40\%$ of our cohort the LD dose was increased or maintained stable to avoid congestion and increased filling pressure.
The most likely explanation for the association between LD and prognosis is that patients with more severe cardiac and renal dysfunction also have more evidence of congestion, requiring treatment with LD to control symptoms and signs. Our results, showing the more favourable evolution of clinical and echocardiographic data in patients reducing their LD dose, support this hypothesis. Thus, LD are associated with a worse prognosis, which might be much worse for such patients if they were not used [5]. However, it is also possible that LD impair renal function and cause hypotension that reduces the ability to titrate other treatments for HF [6].
The analyses on the longitudinal trend of clinical and echocardiographic data according to the LD dose trajectories represent a novel result. We observed that patients who did not receive LD had better LV function and smaller ventricular and atrial volumes. Conversely, patients in whom LD were increased were less likely to show favourable cardiac reverse remodelling at follow-up. A dilated and dysfunctional LV will increase the severity of MR and LA dilation; eventually, this will lead to pulmonary and systemic venous hypertension and congestion requiring LD therapy. Further prospective studies are required to clarify if clinical and echocardiographic evolution might guide LD therapy in patients with DCM.
Interestingly, for the subset of patients with genetic characterization, reduction in LD appeared more common in those with TTNtv-related DCM, perhaps suggesting a better response to HF treatment [28]. This finding should be confirmed in other cohorts.
Currently, prognostic stratification for DCM is mostly based on the severity of LV systolic function on imaging, assessment of fibrosis by cardiac MRI, biomarkers, cardiopulmonary exercise testing or genetic studies [23, 29, 30]. Information on the use and dose of LD is readily available to healthcare providers. Our study demonstrates that LD use and dose are powerful prognostic markers of clinical outcomes; patients with DCM who do not take LD have an excellent prognosis and can be reassured. They should avoid interventions that are associated with risk or substantial costs and should be enrolled less in trials investigating the effects of new treatments on morbidity and mortality because there is little prospect that they will benefit but they might experience serious adverse effects. Patients prescribed with LD, especially at higher or increasing doses, are at much higher risk and require further investigation to identify remediable causes and should be included in randomized trials of new interventions. Altogether, these results support the notion that LD should be used as a pharmaco-epidemiological marker of underlying congestion, cardiac dysfunction, and adverse prognosis in patients with DCM.
## Limitations
The long enrolment period may have introduced a bias, as a treatment for HF has improved in the past few decades. However, in our centre, all patients with DMC and LV dysfunction received comprehensive treatment with anti-neurohormonal therapies since 1990 [31], and we adjusted for the enrolment period in our multivariable models. We report all-cause mortality; the incidence of non-cardiovascular death is low in younger patients with DCM [16]. We did not attempt to investigate possible differences between different types of LD because $99\%$ of patients enrolled in this study were treated with furosemide. The FED change cut-off chosen to classify patients was defined according to the clinical significance of an increase of at least $50\%$ in FED. Information on natriuretic peptides, genetic testing, cardiac magnetic resonance and clinical congestion was not systematically available. Finally, the retrospective nature allowed us to find prognostic associations as we could not exclude the presence of potential unmeasured confounders.
## Conclusion
For patients with DCM, LD use and increasing dose are associated with a poor prognosis and adverse cardiac remodelling. Conversely, in selected patients not receiving LD or being able to stop them is associated with a greater likelihood that ventricular function will recover and a good prognosis. LD use and dose should be included in prognostic models for DCM and considered as inclusion criteria in clinical trials when their objective is to reduce morbidity and mortality.
## Supplementary Information
Below is the link to the electronic supplementary material. Supplementary file1 (DOCX 813 KB)
## References
1. McDonagh TA, Metra M, Adamo M, Gardner RS, Baumbach A, Böhm M. **ESC Guidelines for the diagnosis and treatment of acute and chronic heart failure: developed by the Task Force for the diagnosis and treatment of acute and chronic heart failure of the European Society of Cardiology (ESC). With the special contribution of the Heart Failure Association (HFA) of the ESC**. *Eur J Heart Fail* (2022) **24** 4-131. DOI: 10.1002/ejhf.2333
2. Brunner-La Rocca HP, Linssen GC, Smeele FJ, van Drimmelen AA, Schaafsma HJ, Westendorp PH. **Contemporary drug treatment of chronic heart failure with reduced ejection fraction: the CHECK-HF Registry**. *JACC Heart Fail* (2019) **7** 13-21. DOI: 10.1016/j.jchf.2018.10.010
3. Täger T, Fröhlich H, Seiz M, Katus HA, Frankenstein L. **READY: relative efficacy of LDs in patients with chronic systolic heart failure-a systematic review and network meta-analysis of randomised trials**. *Heart Fail Rev* (2019) **24** 461-472. DOI: 10.1007/s10741-019-09771-8
4. Damman K, Kjekshus J, Wikstrand J, Cleland JG, Komajda M, Wedel H. **Loop diuretics, renal function and clinical outcome in patients with heart failure and reduced ejection fraction**. *Eur J Heart Fail* (2016) **18** 328-336. DOI: 10.1002/ejhf.462
5. Pellicori P, Cleland JG, Zhang J, Kallvikbacka-Bennett A, Urbinati A, Shah P. **Cardiac dysfunction, congestion and loop diuretics: their relationship to prognosis in heart failure**. *Cardiovasc Drugs Ther* (2016) **30** 599-609. DOI: 10.1007/s10557-016-6697-7
6. Martens P, Verbrugge FH, Nijst P, Dupont M, Mullens W. **Changes in loop diuretic dose and outcome after cardiac resynchronization therapy in patients with heart failure and reduced left ventricular ejection fractions**. *Am J Cardiol* (2017) **120** 267-273. DOI: 10.1016/j.amjcard.2017.04.021
7. Maggioni AP, Dahlström U, Filippatos G, Chioncel O, Crespo Leiro M, Drozdz J, Fruhwald F, Gullestad L, Logeart D, Fabbri G, Urso R, Metra M, Parissis J, Persson H, Ponikowski P, Rauchhaus M, Voors AA, Nielsen OW, Zannad F, Tavazzi L. **EURObservational Research Programme: regional differences and 1-year follow-up results of the Heart Failure Pilot Survey (ESC-HF Pilot)**. *Eur J Heart Fail* (2013) **15** 808-817. DOI: 10.1093/eurjhf/hft050
8. Kapelios CJ, Kaldara E, Ntalianis A, Nana E, Pantsios C, Repasos E. **Lowering furosemide dose in stable chronic heart failure patients with reduced ejection fraction is not accompanied by decompensation: a randomized study**. *Int J Cardiol* (2014) **177** 690-692. DOI: 10.1016/j.ijcard.2014.09.156
9. Mullens W, Damman K, Harjola VP, Mebazaa A, Brunner-La Rocca HP, Martens P. **The use of diuretics in heart failure with congestion—a position statement from the Heart Failure Association of the European Society of Cardiology**. *Eur J Heart Fail* (2019) **21** 137-155. DOI: 10.1002/ejhf.1369
10. Cleland JGF, Pellicori P. **To master heart failure, first master congestion**. *Lancet* (2021) **398** 935-936. DOI: 10.1016/S0140-6736(21)01914-0
11. Merlo M, Cannatà A, Gobbo M, Stolfo D, Elliott PM, Sinagra G. **Evolving concepts in dilated cardiomyopathy**. *Eur J Heart Fail* (2018) **20** 228-239. DOI: 10.1002/ejhf.1103
12. Costanzo MR, Stevenson LW, Adamson PB, Desai AS, Heywood JT, Bourge RC. **Interventions linked to decreased heart failure hospitalizations during ambulatory pulmonary artery pressure monitoring**. *JACC Heart Fail* (2016) **4** 333-344. DOI: 10.1016/j.jchf.2015.11.011
13. Rohde LE, Rover MM, Figueiredo Neto JA, Danzmann LC, Bertoldi EG, Simões MV. **Short-term diuretic withdrawal in stable outpatients with mild heart failure and no fluid retention receiving optimal therapy: a double-blind, multicentre, randomized trial**. *Eur Heart J* (2019) **40** 3605-3612. DOI: 10.1093/eurheartj/ehz554
14. Dauw J, Martens P, Tersalvi G, Schouteden J, Deferm S, Gruwez H, De Moor B. **Diuretic response and effects of diuretic omission in ambulatory heart failure patients on chronic low-dose loop diuretic therapy**. *Eur J Heart Fail* (2021) **23** 1110-1119. DOI: 10.1002/ejhf.2145
15. Sutton MG, Plappert T, Hilpisch KE, Abraham WT, Hayes DL, Chinchoy E. **Sustained reverse left ventricular structural remodelling with cardiac resynchronization at one year is a function of etiology: quantitative Doppler echocardiographic evidence from the Multicenter InSync Randomized Clinical Evaluation (MIRACLE)**. *Circulation* (2006) **113** 266-272. DOI: 10.1161/CIRCULATIONAHA.104.520817
16. Merlo M, Cannatà A, Pio Loco C, Stolfo D, Barbati G, Artico J. **Contemporary survival trends and aetiological characterization in non-ischaemic dilated cardiomyopathy**. *Eur J Heart Fail* (2020) **22** 1111-1121. DOI: 10.1002/ejhf.1914
17. Nuzzi V, Cannatà A, Manca P, Castrichini M, Barbati G, Aleksova A. **Atrial fibrillation in dilated cardiomyopathy: outcome prediction from an observational registry**. *Int J Cardiol* (2021) **15** 140-147. DOI: 10.1016/j.ijcard.2020.08.062
18. Pinto YM, Elliott PM, Arbustini E, Adler Y, Anastasakis A, Böhm M. **Proposal for a revised definition of dilated cardiomyopathy, hypokinetic non-dilated cardiomyopathy, and its implications for clinical practice: a position statement of the ESC working group on myocardial and pericardial diseases**. *Eur Heart J* (2016) **37** 1850-1858. DOI: 10.1093/eurheartj/ehv727
19. Vargo DL, Kramer WG, Black PK, Smith WB, Serpas T, Brater DC. **Bioavailability, pharmacokinetics, and pharmacodynamics of torsemide and furosemide in patients with congestive heart failure**. *Clin Pharmacol Ther* (1995) **57** 601-609. DOI: 10.1016/0009-9236(95)90222-8
20. Lang RM, Badano LP, Mor-Avi V, Afilalo J, Armstrong A, Ernande L. **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
21. Cannatà A, Merlo M, Dal Ferro M, Barbati G, Manca P, Paldino A. **Association of titin variations with late-onset dilated cardiomyopathy**. *JAMA Cardiol.* (2022). DOI: 10.1001/jamacardio.2021.5890
22. Manca P, Cannatà A, Nuzzi V, Bromage DI, Varrà GG, Rossi M. **Prevalence and evolution of right ventricular dysfunction among different genetic backgrounds in dilated cardiomyopathy**. *Can J Cardiol* (2021) **37** 1743-1750. DOI: 10.1016/j.cjca.2021.06.024
23. Manca P, Nuzzi V, Cannatà A, Merlo M, Sinagra G. **Contemporary etiology and prognosis of dilated non-ischemic cardiomyopathy published online ahead of print, 2021 Aug 2]**. *Minerva Cardiol Angiol.* (2021) **10** 10. DOI: 10.23736/S2724-5683.21.05736-7
24. Pellicori P, Fitchett D, Kosiborod MN, Ofstad AP, Seman L, Zinman B. **Use of diuretics and outcomes in patients with type 2 diabetes: findings from the EMPA-REG OUTCOME trial**. *Eur J Heart Fail* (2021) **23** 1085-1093. DOI: 10.1002/ejhf.2220
25. Cleland JG, Shelton R, Nikitin N, Ford S, Frison L, Grind M. **Prevalence of markers of heart failure in patients with atrial fibrillation and the effects of ximelagatran compared to warfarin on the incidence of morbid and fatal events: a report from the SPORTIF III and V trials**. *Eur J Heart Fail* (2007) **9** 730-739. DOI: 10.1016/j.ejheart.2007.01.013
26. Kapelios CJ, Laroche C, Crespo-Leiro MG, Anker SD, Coats AJS, Díaz-Molina B. **Association between LD dose changes and outcomes in chronic heart failure: observations from the ESC-EORP Heart Failure Long-Term Registry**. *Eur J Heart Fail* (2020) **22** 1424-1437. DOI: 10.1002/ejhf.1796
27. Halliday BP, Wassall R, Lota AS, Khalique Z, Gregson J, Newsome S. **Withdrawal of pharmacological treatment for heart failure in patients with recovered dilated cardiomyopathy (TRED-HF): an open-label, pilot, randomised trial**. *Lancet* (2019) **393** 61-73. DOI: 10.1016/S0140-6736(18)32484-X
28. Dal Ferro M, Stolfo D, Altinier A, Gigli M, Perrieri M, Ramani F. **Association between mutation status and left ventricular reverse remodelling in dilated cardiomyopathy**. *Heart* (2017) **103** 1704-1710. DOI: 10.1136/heartjnl-2016-311017
29. Stelken AM, Younis LT, Jennison SH, Miller DD, Miller LW, Shaw LJ. **Prognostic value of cardiopulmonary exercise testing using percent achieved of predicted peak oxygen uptake for patients with ischemic and dilated cardiomyopathy**. *J Am Coll Cardiol* (1996) **27** 345-352. DOI: 10.1016/0735-1097(95)00464-5
30. Masci PG, Doulaptsis C, Bertella E, Del Torto A, Symons R. **Incremental prognostic value of myocardial fibrosis in patients with non-ischemic cardiomyopathy without congestive heart failure**. *Circ Heart Fail* (2014) **7** 448-456. DOI: 10.1161/CIRCHEARTFAILURE.113.000996
31. Pinamonti B, Di Lenarda A, Sinagra G, Camerini F. **Restrictive left ventricular filling pattern in dilated cardiomyopathy assessed by Doppler echocardiography: clinical, echocardiographic and hemodynamic correlations and prognostic implications. Heart Muscle Disease Study Group**. *J Am Coll Cardiol* (1993) **22** 808-815. DOI: 10.1016/0735-1097(93)90195-7
|
---
title: Genome-wide identification, expression profile and evolutionary relationships
of TPS genes in the neotropical fruit tree species Psidium cattleyanum
authors:
- Drielli Canal
- Frank Lino Guzman Escudero
- Luiza Alves Mendes
- Marcia Flores da Silva Ferreira
- Andreia Carina Turchetto-Zolet
journal: Scientific Reports
year: 2023
pmcid: PMC9998390
doi: 10.1038/s41598-023-31061-5
license: CC BY 4.0
---
# Genome-wide identification, expression profile and evolutionary relationships of TPS genes in the neotropical fruit tree species Psidium cattleyanum
## Abstract
Terpenoids are essential for plant growth, development, defense, and adaptation mechanisms. Psidium cattleyanum (Myrtaceae) is a fleshy fruit tree species endemics from Atlantic Forest, known for its pleasant fragrance and sweet taste, attributed to terpenoids in its leaves and fruits. In this study, we conducted genome-wide identification, evolutionary and expression analyses of the terpene synthase gene (TPS) family in P. cattleyanum red guava (var. cattleyanum), and yellow guava (var. lucidum Hort.) morphotypes. We identified 32 full-length TPS in red guava (RedTPS) and 30 in yellow guava (YlwTPS). We showed different expression patterns of TPS paralogous in the two morphotypes, suggesting the existence of distinct gene regulation mechanisms and their influence on the final essential oil content in both morphotypes. Moreover, the oil profile of red guava was dominated by 1,8-cineole and linalool and yellow guava was enriched in α-pinene, coincident in proportion to TPS-b1 genes, which encode enzymes that produce cyclic monoterpenes, suggesting a lineage-specific subfamily expansion of this family. Finally, we identified amino acid residues near the catalytic center and functional areas under positive selection. Our findings provide valuable insights into the terpene biosynthesis in a Neotropical Myrtaceae species and their potential involvement in adaptation mechanisms.
## Introduction
Psidium cattleyanum Sabine (Myrtaceae), commonly known as araçá, cattley guava, strawberry guava, and cherry guava, is a fleshy fruit belonging to the Neotropical Myrteae tribe (Myrtaceae). The species is native to the Atlantic Forest, where it has readily adapted to a variety of climates, is associated with wet forests across the tropics1, occurs in areas under stress conditions2,3, and is considered among the worst invasive species4,5.
The genus *Psidium is* rich in essential oils6,7, stored in the leaf secretory cavities8–10, and traditionally used for extraction, with inexpensive resources and potential uses in the pharmaceutical and medicine industries2,11. These essential oils regulate environmental processes and ecological interactions between organisms, such as defense against herbivores and pathogens11,12, protection against abiotic environments13,14 and attraction of pollinators, especially in neotropical species with fleshy berries that serve as a food source15,16.
Psidium cattleyanum species is divided into two morphotypes. The red guava (P. cattleyanum Sabine var. cattleianum) and yellow guava (P. cattleyanum Sabine var. lucidum Hort). The ripe fruits of red and yellow guava present red and yellow epicarps, respectively17. They also exhibit differences in antioxidant activity and phenolic content18, leaf morphology, and phytochemistry, size, and habit19–21. Previous studies have also detected considerably different oil profiles of yellow and red guava, which was attributed to differences in isolation techniques or the area of collection22–25. However, the genetic and evolutionary factors that can induce modifications in the secondary metabolism of the plant in the two morphotypes of P. cattleyanum remain largely unknown. Therefore, the study of genes of the biosynthetic pathway of these compounds in this species is highly relevant.
Because Myrtaceae species exhibit the highest concentrations and functional versatility of foliar terpenes among plants, significant efforts have been made to investigate the molecular mechanisms determining the structural diversity of terpene synthase (TPS) genes in this family. However, to the best of our knowledge, there are still no studies of these genes in Neotropical Myrtaceae species. The TPS family catalyzes the cyclization and rearrangement of geranyl diphosphate (GPP) or its cis-isomer neryl diphosphate (NPP) into monoterpenes (C10) and trans-geranyl diphosphate (GGPP) into diterpenes (C20) in the plastidic 2C-metil-D-eritritol-4-fosfato (MEP) pathway. In addition, farnesyl diphosphate (FPP) is converted into sesquiterpenes (C15) and triterpenes (C30) via the mevalonate (MVA) pathway in the cytosol, endoplasmic reticulum, and peroxisomes26–29. TPS controls not only the terpene chemodiversity present in plants but is also responsible for the unique composition of each taxon30.
Recent studies have revealed that, among those with dried capsular fruits, the species of the Eucalypteae tribe, including Eucalyptus grandis, E. globulus, and Corymbia citriodora, contain the largest number of complete TPS genes reported in eudicotyledons (70, 69, and 89 complete genes, respectively). This is due to the key role of terpenes in defense over their long lifespans26,29,31. Terpene synthase genes have also been identified in *Melaleuca alternifolia* and Leptospermum scoparium, with 37 and 49 putative TPS genes, respectively32,33. The oil profile patterns in foliar terpenes across this species, which are common in forest woodlands, are the monoterpenes α-pinene and 1,8-cineole. Instead, fleshy-fruited species from Myrtaceae family have low foliar 1,8-cineole concentrations, with a greater diversity of abundant foliar sesquiterpenes34.
In the present study, we aimed to conduct a comprehensive genome-wide analysis of TPS genes in P. cattleyanum to gain insights into the underlying mechanisms responsible for the differences in terpenoid biosynthesis and in the essential oil profiles in two morphotypes. Based on genomic and transcriptomic data, we identified the TPS gene repertoire and revealed its expression pattern in two P. cathleyanum morphotypes. We also examined the expansion and diversification of the TPS gene family among the Myrtaceae species. Finally, we investigated key amino acids using positive selection analysis to understand their effects on product specificity and consequently explain the chemical variability of the essential oil compounds. Our findings provide a foundation for deciphering TPS biosynthesis in P. cattleyanum and diversification of the two morphotypes. This knowledge will contribute to further studies on natural populations and the evolution of the Myrteae tribe, providing evidence of the successful distribution and adaptation of these species.
## Genome-wide identification of putative terpene synthases
We performed a genome-wide sequence homology search to identify the complete repertoire of TPS genes across the *Psidium cattleyanum* morphotype genomes. The genomes were assembled separately for comparison. Based on the conservation of hidden Markov model (HMM) profiles and BLAST searches, we identified 110 loci in the red genome (RedTPS) (Supplementary Table S1) and 106 loci in the yellow genome (YlwTPS) (Supplementary Table S2). Three RedTPS and seven YlwTPS sequences were excluded from further analysis due to the presence of premature stop codons, lack of both C and N terminal domains, presence of less than three exons, and one gene that presented 38 exons likely to be pseudogenes, partial genes, or assembly errors (Supplementary Table S3). In YlwTPS, 28 lost the C-terminal domain (PF03936), 45 lost the N-terminal domain (PF01397), and 33 of them contained two domains. In RedTPS, 27 lost the PF03936 domain, 49 lost the PF01397 domain, and 34 of them contained two domains (Supplementary Tables S1, S2). Of the remaining TPS gene models, only 32 RedTPS and 30 YlwTPS were classified as full-length putative loci coding genes (Supplementary Tables S1, S2). The number of TPS may be underrepresented due to incomplete sequences or atypical gene structures obtained and in part due to draft genome assembly.
To identify putative orthologs between the two morphotypes, we created a sequence percentage identity matrix (Supplementary Table S4), and genes containing the top hits are shown (Supplementary Table S5). Only four TPS partial genes had identical sequences in the two genomes (Pca_red_g91813 and Pca_ylw_g91315; Pca_red_g71488 and Pca_ylw_g60043; Pca_red_g21900 and Pca_ylw_g23854; Pca_red_g46186 and Pca_ylw_g20612). In addition, sequence identity among TPS genes between the two morphotypes was considerably lower, with only nine full length genes having greater than $90\%$ amino acid identity (Supplementary Table S5). However, comparing partial and full genes, only 23 showed > $90\%$ identity. The number increased when comparing only partial genes, where 29 genes showed > $90\%$ identity (Supplementary Table S5).
Most TPS genes of subfamilies TPS-a, TPS-b, and TPS-g contained six to nine exons (Fig. 1a), with exceptions (Supplementary Tables S1, S2). Genes from the remaining subfamilies, TPS-c, TPS-e, and TPS-f, contained 7–14 exons (Fig. 2A). Moreover, only one full YlwTPS (Pca_ylw_g56204) and four RedTTPS (Pca_red_g44464, Pca_red_g43593, Pca_red_g28651, and Pca_red_g25997) lacked the highly conserved aspartate-rich motif “DDXXD” (Supplementary Tables S1, S2). The TPS-c subfamily is present in land plants and is characterized by the “DXDD” motif but not the “DDXXD” motif in their proteins, which was detected in only one RedTPS and two full YlwTPS26. The second motif in the C-terminal domain, “NSE/DTE”, is less conserved in TPS and presents the variation “(L,I) × (D,N,G)D(F,I,L) × (S,T,G,A)xxxE”. Figure 1Phylogeny and gene structure of TPS from secondary metabolism. ( a) Conserved domains in TPS genes and their consensus sequences from P. cattleyanum. ( b) Phylogenetic tree of the Tps-a, Tps-b and Tps-g subfamilies from P. cattleyanum genome and characterized representative TPS from other Myrtaceae species. This tree was constructed through maximum likelihood analysis comparing the red and yellow morphotypes (Pca_red and Pca_ylw), C. citriodora subsp. variegata (Cci), E. grandis (Egr), E. globulus (Egl), M. alternifolia (Mal) and A. thaliana (Ath). Functional characterized terpene synthases are written in bold. Bootstrap values supported by < $60\%$ are noted by number. A few species from TPS-c clade were used as the outgroup. Figure 2Phylogeny and gene structure of the TPS from primary metabolism. ( a) Conserved domains in TPS genes and their consensus sequences from P. cattleyanum. ( b) Phylogenetic tree of the Tps-c, Tps-e and Tps-f subfamilies from P. cattleyanum genome and representative TPS from other Myrtaceae species. This tree was constructed through maximum likelihood methods comparing the red and yellow morphotypes (Pca_red and Pca_ylw), C. citriodora subsp. variegata (Cci), E. grandis (Egr), E. globulus (Egl), M. alternifolia (Mal) and A. thaliana (Ath). Functional characterized terpene synthases are written in bold. Bootstrap values supported by > $60\%$ are noted by number. A few species from TPS-a clade were used as the outgroup.
In the clade corresponding to the TPS-b subfamily monoterpene synthases, using different algorithm predictors, we found that only five full RedTPS and three full YlwTPS have an N-terminal transit peptide required for plastidial targeting (Supplementary Table S6).
We identified seven RedTPS and five YlwTPS with the N-terminal domain containing an “RRX8W" motif. In addition to these motifs, there is a highly conserved arginine-rich “RXR” motif. The TPS-g (Pca_ylw_g32591; Pca_red_g25997) subfamily is closely related to TPS-b; however, it lacks the conserved “R(R)X8W” motif in its encoded proteins, and its members may function in producing acyclic mono-, sesqui-, and diterpene products26.
## Molecular evolutionary analysis
To accurately classify the members of the P. cattleyanum TPS gene family based on sequence relatedness as well as functional assessments, we first collected 164 sequences of full-length TPS genes (containing the two TPS domains and having sequence lengths greater than 200 amino acids) from previous studies of species functionally characterized A. thaliana and E. grandis (Myrtaceae family) (Supplementary Fig. S2).
The topology of the phylogenetic tree allowed us to divide TPSs into subfamilies belonging to secondary metabolism, clustered with subfamily TPS-a, which produces sesquiterpenes (C15) with 14 RedTPS and 12 YlwTPS (Table 1, Fig. 3), and TPS-b, which encodes enzymes that produce monoterpenes with 14 RedTPS and 14 YlwTPS. Only one TPS-g gene was found in each morphotype, which predominantly produced acyclic mono-, sesqui-, and diterpenes (Table 2, Fig. 1b). In the cluster representing primary metabolism, a single gene, TPS-c, which produces diterpenes (C20), was found in P. cattleyanum red morphotype, while two were found in the yellow morphotype (Table 2; Fig. 2B). In the TPS-e/f subfamily, which produces mono-, sesqui-, and diterpenes, a single gene was found in the yellow morphotype, whereas two were found in the red morphotype. Our analysis including other Myrtaceae TPS genes showed that all TPS proteins identified in this study clustered into monophyletic-specific clades related to the subfamilies. The TPS-a and TPS-b subfamilies were the most expanded, accounting for approximately $80\%$ of the total TPS full length genes identified (Fig. 3).Table 1Chemical constituents of leaf oil from red and yellow morphotypes of Psidium cattleyanum. NCompoundaRIbContent (%)cClassificationdRedYellow1α-Pinene93010.035.4MH2β-Pinene972–3.2MH3β–Myrcene991–9.5MH41,8-Cineole102859.522.4MO5β-Ocimene10392.9–MH6γ-Terpinene10584.1–MH7Linalool11009.63.7MO8α-Terpineol11895.62.7MO9β-Caryophyllene14142.52.7SH10Nerolidol15632.4–SO11Caryophyllene oxide1579–6.6SO12Viridiflorol1598–2.2SO13Aromadendrene epoxide1633–2.5SOTotal96.690.9aMajor compounds listed in the elution order using Rtx®-5MS column.bRI: Retention index determined by the normalization of retention times with respect to an n-alkane mixture (C7–C40)87–89.cCompounds with a relative area of > $2\%$ were identified.dTerpenic classification: oxygenated monoterpene (MO), hydrogenated sesquiterpene (SH), oxygenated sesquiterpene (SO).Figure 3Proportion of TPS gene subfamilies found in Myrtaceae species. The number of genes in each subfamily relative to the total number of genes indicates the proportion of TPS genes. Psidium cattleyanum had the highest proportion of TPS-b1 genes (~ $40\%$).Table 2Numbers of TPS genes in Myrtaceae species. SpeciesTotal TPS gene modelsPutative full lengthFull lengthabcde/fghPsidium cattleyanum red11032141410210Psidium cattleyanum yellow10630121420110Corymbia citriodora12784383310390Melaleuca alternifolia3737141410440Eucalyptus globulus14310344372010100Eucalyptus grandis17270381510880Leptospermum scoparium49239710420 Site model selection analyses indicate sites that evolve under positive selection fit the data significantly better than the respective null models (M8 vs. M7: LRT = 14.46, df = 2, $$p \leq 0.001$$), however, the posterior probability was low ($p \leq 0.55$) (Supplementary Table S9). Therefore, positive selection may only occur during specific stages of evolution or in particular branches, we tested a branch-specific model to detect positive selection in the three clades formed in the TPS-b subfamily, which were fixed as foreground branches. Clade 1 contained only TPS-b1 genes from Eucalyptus and Psidium species. Clade 2 contained only TPSb-1 genes from Psidium species. A third clade contained some genes from Populus, Vitis, and Eucalyptus which were grouped with three pinene synthase genes from Psidium and classified as the only TPS-b2 genes. The one-ratio branch model indicated an overall purifying selection for TPS evolution (ω mean values smaller than 1.0). We also investigated selective pressure using the branch site model according to the likelihood ratio tests (LRT) and comparisons of clade 1 ($$p \leq 0.26$$), clade 2 ($p \leq 0.05$), and clade 3 ($p \leq 0.05$), indicating that some sites were statistically significant (Supplementary Table S12). However, only five residues were strongly identified to be under positive selection in clade 2, located in the N-terminal portion of TPSb-1 genes of Psidium, including residue 121, with an aspartate (D) and the alteration to a leucine (L) in the foreground branches, and residue 124 with the most commonly found lysine (K), arginine (R), or tryptofan (W) and its alteration to alanine (A) in foreground branches. We also detected residues 222 with a cysteine (C) and alteration to valine (V) or leucine (L) in the foreground branches, and site 279 with a threonine (T) or isoleucine (I) that presented an alteration to cysteine (C) or tyrosine (Y) in the foreground branches, around “RDR” and “DDXXD” motifs in the C-terminal portion (Fig. 4). Clades 1 and 3 show the residuals with weak signs of positive selection. Figure 4Positive selected sites in TPS-b1 branch including only *Psidium* genes. ( a) The pinene synthase sequence Pca_red_g24428 representing clade 2 as foreground clade. ( b) The linalool synthase sequence Pca_red_g28382 represents clade 2 as foreground clade. Amino acids that were identified on positive selection (red circles) are demonstrated on the protein sequence of these representative species corresponding to the sites in each alignment presented on Supplementary Table S12. Also, the representation of mainly motifs of the entrance of the active site (yellow square, circles, and triangle) represented for the “DDXXD”, “NSE/DTE” and “RXR” domains.
## Global and differential expression analysis associated with the terpene biosynthesis
To gain more insight into the TPS biosynthetic pathway, global and differential expression profiles were evaluated on TPS genes from RNA samples extracted from its leaves and compared in two morphotypes. After library construction, illumina sequencing, and assembly, approximately 84 and 86 million paired end reads already cleaned were generated for yellow and red morphotypes, respectively.
Looking at total gene expression across the two morphotypes, approximately $30\%$ of TPS genes were expressed in leaves (transcripts anchored in 35 genes in the red genome and transcripts anchored in 30 genes in the yellow genome). Genes that showed some expression patterns fell into five clades (Supplementary Tables S7, S8). We found 17 full-length and 18 partial TPS genes with evidence of expression in the red genome and 13 full-length and 18 partial TPS genes in the yellow genome.
A heat map showing differential gene expression using DESeq2 based on |log2Fold Change |≥ 1 and FDR < 0.05 in the red and yellow morphotypes, with two biological replicates in leaves, is shown in Fig. 4. As the two genomes were assembled separately and belonged to the same species, two heatmaps were generated, anchoring all transcripts in the red genome (Fig. 5A) and all transcripts in the yellow genome (Fig. 5B). Therefore, statistical analysis can be performed and then compared. Among these, 19 gene sequences were upregulated in the red morphotype, with only 10 full TPS genes (Fig. 5C; Supplementary Tables S7, S8). In the yellow morphotype, 32 TPS genes were upregulated, but only 14 were full TPS. A total of 12 TPS genes showed the same expression pattern between the two transcriptome comparisons and > $90\%$ of identity, indicating that the same gene was found in the different genome assemblies. Figure 5Differential expression of terpene synthase genes of P. cattleyanum leaves. ( a) Red morphotype genome. ( b) Yellow morphotype genome. The blue colored cells indicate the Log2 transformed FPKM (unit of fragments per kb of exon per million mapped reads) with no expression and value zero in this tissue (down-regulated), and red color indicates a higher percentage of total expression for a given gene (up-regulated). Squares represent full-length genes and black dots represent partial genes. ( c) Veen Diagram representing the up-regulated unique transcripts common between the morphotypes.
## Terpenoid profling in Psidium cattleyanum leaves
The leaves of *Psidium cattleyanum* were examined for chemical compositions of the volatile terpene compounds, to investigate the genetic influence on the chemical variations of the oil content between the two morphotypes. The content of each terpenoid was calculated as a percentage of the total essential oil using gas chromatography with a flame ionization detector (GC-FID) and gas chromatography coupled to mass spectrometry (GC–MS) approaches. Thirteen compounds were identified, and the most abundant monoterpenes in both morphotypes were 1,8-cineole, α-pinene, linalool, and α-terpineol (Table 1; Supplementary Fig. S1A). Although these compounds were commonly found, they showed significant quantitative variation. For example, the α-pinene showed a large difference of $35.4\%$ in yellow and only $10.0\%$ in red morphotype; 1,8-cineole showed a difference of $59.5\%$ in the red and $22.4\%$ in yellow morphotype; whereas linalool showed a difference of $9.6\%$ and $3.7\%$ in the red and yellow morphotypes, respectively.
In addition to quantitative variations, the plants used also showed qualitative variations in the chemical composition of their essential oils. The hydrogenated monoterpenes β-ocimene ($2.9\%$) and γ-terpinene ($4.1\%$) were observed only in red morphotype essential oil, and the oxygenated sesquiterpene nerolidol ($2.4\%$) (Supplementary Fig. S1B). The hydrogenated monoterpenes β-pinene ($3.2\%$) and β-myrcene ($9.5\%$) were observed only in yellow morphotype, and the oxygenated sesquiterpenes caryophyllene oxide ($6.6\%$), aromadendrene epoxide ($2.5\%$), and viridiflorol ($2.2\%$) (Supplementary Fig. S1C).
## Discussion
The Myrtaceae family is recognized for its great potential to produce volatile oils of economic interest35. The identification of photochemical profiles of some species combined with genomic studies, revealing a high diversity of TPS genes that control the synthesis pathways of these compounds and are responsible for the various biological activities of essential oils28,29,31,36.
In this study, the TPS family has been characterized in Psidium cattleyanum, a fleshy-fruited species from the Myrtaceae family, for the first time at the genomic and transcriptomic levels. It reveals a low number of putative functional full-length TPS genes (32 RedTPS and 30 YlwTPS) required for this species associated with wet forests across the neotropics, when compared with the woody-fruited species (Table 2) from open forest and woodland, such as Eucalypteae tribe, including *Eucalyptus grandis* (70 full length TPS), *Eucalyptus globulus* (103 full length TPS), and *Corymbia citriodora* (84 full length TPS), all species with the diversity center in the Asia and Oceania37. These species are predicted to defend their leaves much more strongly. Moreover, the relatively long lifespan of eucalyptus (well over 200 years)33 compared to Psidium (approximately 40 years)38, may drive further gene diversification as the need to adapt to long-term environmental changes. These results imply that evolutionary forces have acted differently upon lineages since they diverged from their most recent common ancestor more than 70 million years ago1,39.
*Partial* genes might be considered non-functional, even though some of their incomplete sequences could have resulted from poor sequencing techniques. Still, the redundancy of TPS genes has been observed in many other plants, e.g., in grape (Vitis vinifera) there are 152 TPS-like genes, but only 62 full length TPS, with two domain structures40 where tandem duplication rates for both domains (~ $90\%$) are the main mechanisms for family expansion. In E. grandis there were 70 full-length TPS, but seven had only the PF01397 domain where gene losses were mostly related to tandem duplications ($71.4\%$) and less related to segmental duplication ($3.9\%$) events, and 22 TPS with only the PF03936 domain more related to tandem duplication ($71.7\%$) and fewer segmental duplication events ($4.3\%$)41. We observed the same pattern in Psidium cattleyanum, where 28 RedTPS and 27 YlwTPS had only the PF01397 domain and 48 RedTPS and 45 YlwTPS had only the PF03936 domain (Supplementary Table S9). These data suggest that domain loss has been a common event in plants during the evolution of the TPS gene family, with the loss of the PF01397 domain being more frequent in the Myrtaceae family and plants in general than the loss of the PF03936 domain41. The functionality of these single domain-containing TPS is not yet known, but more investigation on regulatory mechanisms, expansion history, and evolutionary advantage of the domains separately should provide a comprehensive view of the impact of partial genes in the diversification of TPS in plants26.
Transcriptome examination revealed that out of 32 full-length RedTPS, 10 genes were upregulated in the red morphotype. Among the 30 full-length YlwTPS, only 14 genes were upregulated in the yellow morphotype. This demonstrates that the differential expression patterns in the two morphotypes can also contribute to the final terpene content in the leaves (Fig. 5). The high abundance of transcripts in this study (FPKM, Fragments Per Kilobase of exon per Million reads) from the TPS-a and TPS-b1 subfamilies in the transcriptome indicated their involvement in the formation of mono and sesquiterpenoid volatiles in leaves.
A comparison of the essential oil composition revealed the presence of oxygenated monoterpenes on leaves of P. cattleyanum, where the major compound was α-pinene ($35.4\%$) in yellow morphotype and the 1,8-cineole (59,$5\%$) and linalool ($9.6\%$) in the red morphotype. This variation in the essential oil of P. cattleyanum morphotypes have also been previously described in native species in southern Brazil24,25. In cultivated plants of P. cattleyanum in different parts of the world, previous studies have identified the chemical composition with β-caryophyllene, a hydrocarbon sesquiterpene, as the main component7,23,25,42–45, which was also found in smaller amounts in both morphotypes in this work.
The variations found between the two morphotypes in this study reflect a genetic and evolutionary origin. The identification of chemotype phenotypes (qualitative variability in foliar essential oil composition) within a single species has already been reported among different varieties or ecotypes of other species46–49, mainly when a significant shift in the relative concentrations involved more similar compounds, such as cineole and pinene50. The yellow morphotype tends to be found at slightly lower elevations than the red morphotype17,21, this could reflect environmental adaptation48 and also in the terpenes plasticity51.
In the TPS-a subfamily that encodes only sesqui-TPSs found in both eudicot and monocot plants40, phylogenetic analysis revealed two YlwTPS (Pca_ylw_g29958 and Pca_ylw_g20359) closely related to RtTPS3 (AXY92168)52 and in the same branch of the gene EgranTPS038 (Euc_Eucgr_J01451) of E. grandis. In addition, four RedTPS (Pca_red_g40189, Pca_red_g58727, Pca_red_g34404, and Pca_red_g61229) were found in the same branch as RtTPS4 (AXY92169)52; both belong to a branch of the betacaryophylene synthase (BS) (Fig. 6).Figure 6A schematic view of putative terpene synthase genes involved in α-pinene, 1,8-cineol, linalool and β-caryophyllene biosynthesis in P. cattleyanum red and yellow morphotypes. DXP 1-deoxylulose 5-phosphate, DMAPP dimethylallyl diphosphate, IPP isopentenil diphosphate, GPPS geranyl diphosphate synthase, GPP geranyl diphosphate, FPPS farnesyl diphosphate synthase, FPP farnesyl diphosphate, ER endoplasmic reticulum, MEP methyl erythritol 4-phosphate pathway, MVA mevalonate pathway.
The monophyletic TPS-b subfamily is divided into two groups. The TPS-b1 clade contains putative cyclic monoterpene synthases, with transit peptides positioned upstream of the “RRX8W” motif and therefore has a high probability of localizing in the plastids29. The subfamily had the highest number of full-length genes ($40\%$) as a proportion of the total number of TPS genes compared to *Melaleuca alternifolia* ($32.4\%$) and *Populus trichocarpa* ($31.2\%$) (Fig. 3). The high proportions of the TPS-b1 subfamily could be indicative of rapid ongoing evolution and lineage-specific gene family expansion of this subfamily in warm subtropical habitats, particularly for protection from damage caused by rapid temperature fluctuations53,54. Some terpenes can act by selecting the defense of antimicrobial secondary metabolites such as cyclic monoterpenes32. This suggests that subfamilies of TPS-b1 expansion might be related to species or ecotype diversification, enabling quick adaptation in response to environmental changes.
The other TPS-b subfamily contains putative isoprene/ocimene (C5, C10) synthases, described as TPS-b232 and has few genes in *Psidium cattleyanum* ($6.6\%$ in yellow and $3.1\%$ in red morphotype) as a proportion of the total number of TPS genes compared with *Eucalyptus globulus* ($9.4\%$) and *Melaleuca alternifolia* ($5.4\%$) (Fig. 3)29,32,55. However, when including genes functionally characterized in the TPS-b2 clade, the relationships among *Psidium* genes were not entirely congruent because we detected these three genes (Pca_ylw_g54543, Pca_red_g66899, and Pca_ylw_g73225) positioned in the same clade as *Rhodomyrtus tomentosa* RtTPS1 (AXY92166), characterized as pinene synthase (PS), a cyclic terpene52, despite the high support (bootstrap value of 97) in the same branch of acyclic EglobTPS106, functionally characterized as isoprene synthase29 (Fig. 1b). Including more TPS from Myrtaceae species from neotropics and functionally characterizing the *Psidium* genes should clarify their role and division within the clade TPS-b.
Two other genes (Pca_ylw_g32667 and Pca_red_g24428) clustered together with RtTPS2 (AXY92167) and EpTPS1 (MK873024) in the TPS-b1 branch and were related to pinene synthase (Fig. 6). Moreover, four genes (Pca_red_g44464, Pca_red_g28382; Pca_ylw_g3537, Pca_ylw_g40677) were in the same branch as EpTPS2 and EpTPS3 from *Eucalyptus polybractea* belonging to the CS TPS-b1 clade56, and one gene from E. grandis monoterpene synthase (XP_010046521), which has similarity ($95\%$ amino acid identity) to CS that produces 1,8-cineole an in vitro assay using GPP as substrate57. Analysis of transcript abundance showed that the gene Pca_red_g28382 was highly expressed in leaves. The dominant product of many characterized CS enzymes is 1,8-cineole; however, they also produce small amounts of limonene, β-myrcene, sabinene, β-pinene, α-pinene, and α-terpineol. This group of compounds synthesized by CS is known as the ‘cineole cassette’, which has been reported in many plants58,59. Therefore, as multiple TPS genes are often expressed in the same tissue and many of these TPS’ have overlapping ranges of products, it is not easy to identify the action of individual TPS enzymes on the profile of the terpene observed in that tissue60.
Other expressed TPS-b genes Pca_ylw_g54543, Pca_red_g66899, and Pca_ylw_g73225 were positioned in the same clade as RtTPS1 of Rhodomyrtus tomentosa, and Pca_red_g24428 grouped with RtTPS2. Previous studies have shown the in vitro activity of RtTPS1 and RtTPS2, which mainly produce (+)-α-pinene and (+)-β-pinene with GPP, whereas RtTPS1 is also active with FPP, producing β-caryophyllene, along with a smaller amount of α-humulene53. This suggests that, depending on their expression profile or subcellular location, the enzymatic products of these TPS present in leaves can contribute to the different terpene mixtures found in the essential oil. We also detected the expression of four genes (Pca_red_g44464, Pca_red_g28382; Pca_ylw_g3537, Pca_ylw_g40677) in the same branch of EpTPS2 and EpTPS3, belonging to the PS TPS-b clade.
There is a large diversity of Myrtaceae species, with α-pinene and 1,8-cineole being the dominant compounds in the leaves. The reaction cascade that leads to these two compounds includes the same carbocation intermediate, γ-terpinyl cation49. There is evidence to show that the amino acid changes induced through site-directed mutagenesis can result in a different ratio of particular terpenes produced47,61,62 and in natural systems, this might lead to different dominant compounds, such as the a-terpineol synthases of many species, which are the only characterized terpene synthases that are not 1,8-cineole synthases, but produce significant amounts of 1,8-cineole59.
The TPS-g subfamily has two subclades encoding TPS’ without the “R(R)X8W” motif, which facilitates isomerization of the geranyl cation in the linalyl cation. This subfamily is closely related to the TPS-b subfamily, and its members may function with the prevalence of acyclic monoterpene products. We also identified two genes from Psidium (Pca_red_g25997 and Pca_ylw_g32591) in the same branch as the functionally characterized PS of EgranTPS10129 (Egr_EucgrE03562; Fig. 6).
We screened TPS genes to identify the LS based on functionally characterized enzymes from other plant species. Phylogenetic analysis demonstrated that only two genes (Pca_ylw_g14698 and Pca_red_g69489) are closely related to LS from the rosids Clarkia breweri (Cbr_AAD1984), *Oenothera arizonica* (Oca_AAD1984), and *Clarkia concinna* (Cco_AAD1983). They fall into the TPS-f synthase classification, proposed to be the most ancient, and could have been due to a relatively recent common ancestor, copalyl diphosphate synthase (CPS)63, as evidenced by the sequence conservation of this region in the N-terminus of the protein (Fig. 6). In this study, LS gene expression was not observed in leaves. Monoterpene synthases of this subfamily are responsible for the conversion of GDP into the bulk of monoterpenes found in vegetative organs, whereas the subfamilies TPS-f and TPS-g are thought to be exclusively active in flowers, likely having a primary function in attracting insect pollinators36,64. In addition, other genes could be expressed when directly involved in plant defense against herbivores by attracting predators65 or by directly driving herbivores away66.
Depending on the extent to which gene function is affected, single-base substitutions may result in changes in terpene composition and profile, and if upstream pathway elements are involved, even in terpene concentrations46,67. To infer whether selection acted on the TPS-b subfamily, we used several statistical tests to compare clades on the phylogenetic tree. Codon substitution patterns with a maximum likelihood approach implementing a branch-site model indicated positive selection acting on a specific TPS-b1 branch, including some pinene and cineole synthase genes and other non-functionally assigned genes.
In particular, some positively selected sites are located in the N-terminal region, which controls substrate specificity. It is interesting to note that residue 224 contains an arginine (R) in the PS genes (Fig. 4A), whereas we observed an alteration to a tryptophan (W) residue in the CS genes (Fig. 4B). Conserved arginines close to the diphosphate moiety stabilize the evolving negative charges68. The tryptophan residue contributes to stabilization of the cation and deprotonation of the substrate69. In addition, the positively selected residues 222 and 279 were located around the aspartate-rich motif (“DDXXD”) in the C-terminal half, which is important for the coordination of divalent ion(s), water molecules, and stabilization of the active site70–72.
These results illustrate the importance of these residues to product spectrum of TPS genes, mainly in this case of PS and LS, that have the same carbocation intermediate, thereby differing in their profiles46. Future studies should investigate in detail how the active site promotes discrimination from other potential substrates. Analysis of this type of data could be used to better understand the diversity of terpene synthases and the role of different terpenes in mediating ecological interactions34.
Several biological and pharmacological activities have been reported for pinene, cineol, and linalool, including anti-inflammatory and antinociceptive properties11,73–75, anticancer61,76,77, antifungal78,79, antidiabetic80, antioxidant, antimicrobial77,81,82, antidepressive and neuroprotective77, allelopathic83, antibacterial, and insecticidal activity84,85. The high content of these compounds in the volatile oils of these species suggests that they could constitute an alternative commercial source of this compound86.
## Conclusion
In this study we identified putative TPS genes responsible for the formation of predominant essential oil compounds in Psidium cattleyanum. The chemotypic variability found in the red and yellow morphotypes confirm our hypothesis about the complex and polymorphic nature of the genes encoding the key enzymes regulating compound production and suggest adaptive genetic plasticity of the two morphotypes. The TPS-b clade has undergone substantial expansion compared to other subfamilies and includes some positively selected amino acid residues, evidence the monoterpene synthase genes are important for adaptation to Psidium at different niches. The present study provides the first insight into the genetic basis of TPS in P. cattleyanum morphotypes, gaining insights about the biodiversity in the Atlantic rainforest for further ecological genetic studies in the genus.
## Plant materials
Young leaf samples of the yellow and red morphotypes were grown on the same open ground plot (in two 5-m long rows per cultivar) at the Federal University of Rio Grande do Sul (Porto Alegre, Brazil). The plants were 20–25 years old during the sampling year [2020]. The leaves were washed with distilled water, frozen, and stored at -18 °C until extraction of volatile compounds, immediately frozen in liquid nitrogen and stored at − 80 °C for further RNA extraction.
## Chromatographic profile of the essential oils
We collected volatiles from the leaves of the two morphotypes under the same growth conditions and ambient temperature, in biological triplicates. Approximately 100 g of dry leaves from the two morphotypes, were extracted with 1000 mL of reverse osmosis water using a Clevenger apparatus87, following four hours of extraction by hydro-distillation. Samples of the essential oils extracted from the leaves were analyzed using gas chromatography with a flame ionization detector (GC-FID) (Shimadzu GC-2010 Plus) and gas chromatography coupled to mass spectrometry (GC–MS) (Shimadzu GCMS-QP2010 SE).
We conducted the analyses according to the following conditions: helium (He) as the carrier gas for both detectors, with the flow and linear speeds of 2.80 mL min−1 and 50.8 cm s−1 (GC-FID), and 1.98 mL min−1 and 50.9 cm s−1 (GC–MS), respectively; injection port temperature of 220 °C with a split ratio of 1:30; fused silica capillary column (30 m × 0.25 mm); stationary phase Rtx®-5MS (0.25 μm film thickness); oven with an initial temperature of 40 °C, maintained for 3 min, then gradually increased by 3 °C min−1 until 180 °C, where it remained for 10 min (total analysis time: 59.67 min); and FID and MS detector temperature of 240 °C and 200 °C, respectively49. The used samples were taken from the vials in 1 μL of a solution containing $3\%$ essential oil dissolved in hexane with 0.1 mol L−1 dimethylacetamide (DMA; external standard for reproducibility control).
The GC–MS analyses were performed using electron impact equipment with an impact energy of 70 eV, scanning speed of 1000, scanning interval of 0.50 fragments s−1, and fragments detected from 29 to 400 (m/z). The GC-FID analyses were carried out in a flame formed by H2 and atmospheric air at a temperature of 300 °C. Flow rates of 40 mL min−1 and 400 mL min−1 were used for H2 and air, respectively. Identification of the compounds in the essential oils was accomplished by comparing the obtained mass spectra with those available in the spectral library database (Wiley 7, NIST 05, and NIST 05 s) and retention indices (RI). To calculate the RIs, we used a mixture of saturated alkanes C7–C40 (Supelco-USA) and adjusted retention time of each compound, obtained by GC-FID. The values calculated for each compound were compared with those reported in literature88–90.
We calculated the relative percentage of each compound in the essential oil using the ratio between the integral area of the peaks and the total area of all sample constituents obtained via GC-FID analyses. The compounds with a relative area above $2\%$ were identified and considered predominant if above $10\%$.
## Terpene synthase gene identification and annotation
Initially, we used two terpene synthase-specific domains, PF01397 and PF03936, which represent respectively the N-terminal and C-terminal domains of TPS from the Pfam database (http://pfam.xfam.org/) 91, as queries to search for terpene synthase homolog genes in the P. cattleyanum yellow and red morphotypes predicted genes from their genomes (unpublished data). We analysed each morphotype separately using HMMER version 3.192. We also performed a local BLASTP search for TPS genes in the P. cattleyanum reference genome based on functionally characterized genes93,94. We created a preliminary list of putative TPS genes based on hits with a high similarity (e-value < 1e − 05).
To better understand the structural sequence features of each gene, we used the open reading frame (ORF) Finder of NCBI (http://www.ncbi.nlm.nih.gov/orffinder/) to identify the ORFs for each sequence recovered. Gene structure was determined using the Gene Structure Display Server (GSDS; http://gsds.cbi.pku.edu.cn) 95. We confirmed the presence of functional domains based on the translation of gene sequences identified in Simple Modular Architecture Research Tool (SMART)96. Moreover, several algorithms were used to predict a putative transit peptide for chloroplast targeting in the N-terminal sequence upstream of the RRX8W motif (ChloroP 1.197, TargetP v.1.0198, PCLR 0.999). To determine the sequence diversity between the two morphotypes, a complete set of pairwise comparisons of protein sequences was performed using Clustal Omega (https://www.ebi.ac.uk/Tools/msa/clustalo/).
## Phylogenetic reconstruction
In this study, we first used terpene synthase protein sequences from fully sequenced genomes of A. thaliana100 and E. grandis29, to classify the putative genes found in P. cattleyanum according to the previous classification in the subfamilies TPS-a,-b,-c,-e/f, and -g by sequence similarity26.
To examine the evolutionary history of TPS genes, a second analysis including more species (E. grandis, E. globulus, A. thaliana, P. trichocarpa, V. vinifera, C. citriodora, and M. alternifolia) was carried out. *We* generated a tree with TPS sequences related to primary metabolism (subfamilies -c, -e, and -f) with a total of 45 sequences and a second tree related to secondary metabolism (subfamilies a, b, g) including 360 sequences29,32,55.
The functionally characterized pinene (RtTPS1 and RtTPS2 accession number AXY92166 and AXY92167, respectively) and caryophyllene synthases (RtTPS3 and RtTPS4 accession numbers AXY92168 and AXY92169) from Rhodomyrtus tomentosa52, pinene synthase (EpTPS1 accession number MK873024) and 1,8-cineole synthases (EpTPS2 and EpTPS3 accession numbers MK873025 and QCQ05478) from Eucalyptus polybractea56, beta cayophyllene synthase (Eucgr. J01451) from E. grandis29, myrcene synthase from *Antirrhium majus* (AAO41727)101, two isoprene synthase genes from E. globulus (EglobTPS106), E. grandis (Eucgr. K00881)29 and five linalool synthases from *Oenothera californica* (AAD19841)63, Clarkia breweri (AAD19840), *Clarkia concinna* (AAD19839), and Fragaria x ananassa (CAD57106)102 were also included in the phylogenetic analysis to assess the homology of known TPS to *Psidium* genes.
For each dataset used to construct the trees, we first aligned the amino acid sequences of putative TPS genes using ClustalW implemented within MEGA v7.0 software package103. Due to high levels of variation and variable exon counts between taxa, we trimmed the alignment using Gblocks104 with the following parameters: smaller final blocks, gap positions within the final blocks, and less strict flanking positions. We used the maximum-likelihood method implemented in PhyML v2.4.4105 online web server106 to perform the phylogenetic analysis. The JTT + G + F was the best-fit substitution model selected with ModelGenerator for protein analyses107. The confidence values in the tree topology were assessed by running 100 bootstrap replicates. Trees were visualized using Figtree v1.4.4108.
## Molecular evolutionary analysis involving TPS-b
To understand the molecular evolution at the amino acid level and the intensity of natural selection acting on metabolism in a specific clade, we used a tree based on codon alignment produced by the maximum-likelihood method using the software EasyCodeML109. We retrieved Coding Sequencing (CDS) sequences from TPS-b genes from A. thaliana, E. grandis, P. cattleyanum, V. vinifera and P. trichocarpa species in Phytozome v11 (http://phytozome.jgi.doe.gov/; last accessed November 2020), to use in positive selection analysis. The dataset included 76 sequences and 389 amino acids from five species. We performed statistical analysis using the CodeML program in PAML version 4.9 software using the site, branch, and branch-site models110, implemented in EasyCodeML109.
Parameter estimates (ω) and likelihood scores111 were calculated for the three pairs of models. These were M0 (one-ratio, assuming a constant ω ratio for all coding sites) vs. M3 (discrete, allowed for three discrete classes of ω within the gene), M1a (nearly neutral, allowed for two classes of ω sites: negative sites with ω0 < 1 estimated from our data and neutral sites with ω1 = 1) vs. M2a (positive selection, added a third class with ω2 possibly > 1 estimated from our data), and M7 (beta, a null model in which ω was assumed to be beta-distributed among sites) vs. M8 (beta and ω, an alternative selection model that allowed an extra category of positively selected sites)112.
A series of branch models and branch site models were tested: the one-ratio model for all lineages and the two-ratio model, where the original enzyme functional evolution occurred. The branch-site model assumes that the branches in the phylogeny are divided into the foreground (the one of interest for which positive selection is expected) and background (those not expected to exhibit positive selection).
Likelihood ratio tests (LRT) were conducted to determine which model measured the statistical significance of the data. The twice the log likelihood difference between each pair of models (2ΔL) follows a chi-square distribution with the number of degrees of freedom equal to the difference in the number of free parameters, resulting in a p-value for this113. A significantly higher likelihood of the alternative model compared to the null model suggests positive selection. Positive sites with high posterior probabilities (> 0.95) were obtained using empirical Bayes analysis. If ω > 1, then there is a positive selection on some branches or sites, but the positive selection sites may occur in very short episodes or on only a few sites during the evolution of duplicated genes; ω < 1 suggests a purifying selection (selective constraints), and ω = 1 indicates neutral evolution. Finally, naive empirical Bayes (NEB) approaches were used to calculate the posterior probabilities that a site comes from the site class with ω > 1112. The selected sites and images of protein topology were predicted using Protter114.
## Transcriptome analysis
For expression analysis, we used the published RNA-Seq dataset from leaves for the yellow and red morphotypes of P. cattleyanum115. To verify the quality of reads and the presence of Illumina adaptors, we used the FastQC software (http://www.bioinformatics.babraham.ac.uk/projects/fastqc/). Based on these data, we used the Trim Galore software (http://www.bioinformatics.babraham.ac.uk/projects/trim_galore/) to eliminate read strings with a quality below 30 and adapter sequences.
Two replicates from the red morphotype and two from the yellow morphotype, corresponding to four RNAseq libraries, were aligned on the draft genome assembly of each morphotype (unpublished data) using TopHat2116. The read count tables mapped to each gene were generated using the featureCounts module of the Subread software117, from the bam anchor files generated by TopHat2. The criteria used to create the counting tables were as follows: fragments (pairs of reads) were counted instead of individual reads, pairs of reads anchored on different chromosomes or anchoring on identical chromosomes but on different strands were not considered, and neither were the reads anchored in multiple places in the genome.
We used the DESeq2 package version 1.36118 to perform statistical analysis and identify differential expression. We analyzed the counting tables using a false discovery rate (FDR) of 0.05, log2 fold change ≥ ± 1119 and separated them into a group formed by the “up-regulated” genes and another formed by the “down-regulated” genes.
As the genome of each morphotype was assembled separately and corresponded to the same evaluated species in question, we performed two independent comparative transcriptomic analyses: a comparison of morphotype red leaf against yellow leaf anchoring in red morphotype genome (i) and in yellow genome (ii). We evaluated the differential expression considering each gene found in each morphotype, and were able to detect more genes under differential gene expression (DGE), considering that some gene copies were detected only in one of the reference genomes.
## Ethical standards
The yellow and red morphotypes of *Psidium cattleyanum* were sampled originally as part of the project “Genomics and Transcriptomics Analysis of *Psidium cattleyanum* Sabine (Myrtaceae)”. The studied samples were collected in full compliance with specific federal permits issued by the approved by the Brazilian Ministry of Environment (MMA) and the Chico Mendes Institute for Biodiversity Conservation (ICMBio), and approved by the Biodiversity Information and Authorization System (SISBIO 43338-2) and National System for Governance of Genetic Heritage and Associated Traditional Knowledge (SisGen A7B0331). The studied plants are kept in an ex situ collection at the Federal University of Rio Grande do Sul (UFRGS). Exsiccates will be deposited in the ICN herbarium of UFRGS. As official authorities in Brazil reported, the species used in this study are not endangered or protected in the Rio Grande do Sul State, where the sampling occurred.
## Supplementary Information
Supplementary Figures. Supplementary Tables. The online version contains supplementary material available at 10.1038/s41598-023-31061-5.
## References
1. Biffin E. **Evolution of exceptional species richness among lineages of fleshy-fruited Myrtaceae**. *Ann. Bot.* (2010.0) **106** 79-93. DOI: 10.1093/aob/mcq088
2. Patel S. **Exotic tropical plant**. *Rev. Environ. Sci. Bio/Technol.* (2012.0) **11** 243-248. DOI: 10.1007/s11157-012-9269-8
3. Tng DY. **Characteristics of the**. *Austral Ecol.* (2016.0) **41** 344-354. DOI: 10.1111/aec.12319
4. Tassin J. **Ranking of invasive woody plant species for management on Réunion Island**. *Weed Res.* (2006.0) **46** 388-403. DOI: 10.1111/j.1365-3180.2006.00522.x
5. Enoki T, Drake DR. **Alteration of soil properties by the invasive tree**. *Plant Ecol.* (2017.0) **218** 947-955. DOI: 10.1007/s11258-017-0742-x
6. Stefanello MÉA, Pascoal AC, Salvador MJ. **Essential oils from neotropical Myrtaceae: Chemical diversity and biological properties**. *Chem. Biodivers.* (2011.0) **8** 73-94. DOI: 10.1002/cbdv.201000098
7. Vasconcelos LC. **Phytochemical analysis and effect of the essential oil of**. *Environ. Sci. Pollut. Res.* (2019.0) **26** 26216-26228. DOI: 10.1007/s11356-019-05912-6
8. Oliveira RF. **Study Post-Harvest about impact and compression mechanical in the cell quality of guava fruit (CV. Paluma)**. *Int. J. Sci.* (2014.0) **3** 30-34
9. Silva LC. **Leaf morpho-anatomical structure determines differential response among restinga species exposed to emissions from an iron ore pelletizing plant**. *Water Air Soil Pollut.* (2020.0) **231** 1-9. DOI: 10.1007/s11270-020-04533-x
10. Abrao FY. **Anatomical study of the leaves and evaluation of the chemical composition of the volatile oils from**. *Res. Soc. Dev.* (2021.0) **10** e49110615929-e49110615929. DOI: 10.33448/rsd-v10i6.15929
11. Santos Pereira E. *Food Chem.* (2018.0) **258** 95-103. DOI: 10.1016/j.foodchem.2018.03.024
12. Unsicker SB, Kunert G, Gershenzon J. **Protective perfumes: The role of vegetative volatiles in plant defense against herbivores**. *Curr. Opin. Plant Biol.* (2009.0) **12** 479-485. DOI: 10.1016/j.pbi.2009.04.001
13. Suni T. **Formation and characteristics of ions and charged aerosol particles in a native Australian Eucalypt forest**. *Atmos. Chem. Phys.* (2008.0) **8** 129-139. DOI: 10.5194/acp-8-129-2008
14. Vickers CE. **A unified mechanism of action for volatile isoprenoids in plant abiotic stress**. *Nat. Chem. Biol.* (2009.0) **5** 283-291. DOI: 10.1038/nchembio.158
15. Cseke LJ, Kaufman PB, Kirakosyan A. **The biology of essential oils in the pollination of flowers**. *Nat. Prod. Commun.* (2007.0) **2** 1934578X0700201225
16. Cordeiro GD. **Nocturnal floral scent profiles of Myrtaceae fruit crops**. *Phytochemistry* (2019.0) **162** 193-198. DOI: 10.1016/j.phytochem.2019.03.011
17. Chalannavar RK. **Chemical constituents of the essential oil from leaves of**. *J. Med. Plants Res.* (2013.0) **7** 783-789
18. Rocha LD. **Comparative anatomy study of stem bark of yellow strawberry-guava and red strawberry-guava,**. *Acta Botanica Brasilica* (2008.0) **22** 1114-1122. DOI: 10.1590/S0102-33062008000400022
19. 19.Raseira, M. D. C. B. & Raseira, A. Contribuição ao estudo do araçazeiro, Psidium cattleyanum. Pelotas: EMBRAPA-CPACT (1996).
20. 20.Sobral, M. A família Myrtaceae no Rio Grande do Sul. 1st edition. UNISINOS, São Leopoldo, Brazil (2003).
21. 21.Souza, L. P. & Sobral, M. D. G. Morfotipos do Araçazeiro, Psidium cattleianum Sabine (Myrtaceae) no Estado do Paraná. O Araçazeiro: Ecologia e Controle Biológico. FUPEF, Curitiba, 19–28 (2007).
22. Vernin G. **Analysis of the volatile compounds of**. *J. Essent. Oil Res.* (1998.0) **10** 353-362. DOI: 10.1080/10412905.1998.9700923
23. Biegelmeyer R. **Comparative analysis of the chemical composition and antioxidant activity of red (Psidium cattleianum) and yellow (**. *J. Food Sci.* (2011.0) **76** C991-C996. DOI: 10.1111/j.1750-3841.2011.02319.x
24. Egea MB. **Comparative analysis of aroma compounds and sensorial features of strawberry and lemon guavas (**. *Food Chem.* (2014.0) **164** 272-277. DOI: 10.1016/j.foodchem.2014.05.028
25. Rocha CH. **Chemical composition of the leaf oils from two morphotypes of**. *Nat. Prod. Res.* (2021.0) **35** 4094-4097. DOI: 10.1080/14786419.2020.1721490
26. Chen F. **The family of terpene synthases in plants: A mid-size family of genes for specialized metabolism that is highly diversified throughout the kingdom**. *Plant J.* (2011.0) **66** 212-229. DOI: 10.1111/j.1365-313X.2011.04520.x
27. Vranová E, Coman D, Gruissem W. **Network analysis of the MVA and MEP pathways for isoprenoid synthesis**. *Annu. Rev. Plant Biol.* (2013.0) **64** 665-700. DOI: 10.1146/annurev-arplant-050312-120116
28. 28.Webb, H., Foley, W. J. & Külheim, C. The genetic basis of foliar terpene yield: Implications for breeding and profitability of Australian essential oil crops. Plant Biotechnol. 14–1009 (2014).
29. Külheim C. **The Eucalyptus terpene synthase gene family**. *BMC Genom.* (2015.0) **16** 1-18. DOI: 10.1186/s12864-015-1598-x
30. Bohlmann J, Meyer-Gauen G, Croteau R. **Plant terpenoid synthases: Molecular biology and phylogenetic analysis**. *Proc. Natl. Acad. Sci.* (1998.0) **95** 4126-4133. DOI: 10.1073/pnas.95.8.4126
31. Myburg AA. **The genome of**. *Nature* (2014.0) **510** 356-362. DOI: 10.1038/nature13308
32. Calvert J. **Terpene synthase genes in**. *Plant Syst. Evol.* (2018.0) **304** 111-121. DOI: 10.1007/s00606-017-1454-3
33. Thrimawithana AH. **A whole genome assembly of**. *N. Z. J. Crop. Hortic. Sci.* (2019.0) **47** 233-260. DOI: 10.1080/01140671.2019.1657911
34. Padovan A. **The evolution of foliar terpene diversity in Myrtaceae**. *Phytochem. Rev.* (2014.0) **13** 695-716. DOI: 10.1007/s11101-013-9331-3
35. 35.Deschamps, C. et al. Essential oils yield and composition of Myrtaceae species from Atlantic Forest of South Brazil. In Embrapa Agroindústria de Alimentos-Resumo em anais de congresso (ALICE). Nat. Volatiles Essent. Oils4 (2017).
36. Keszei A. **Functional and evolutionary relationships between terpene synthases from Australian Myrtaceae**. *Phytochemistry* (2010.0) **71** 844-852. DOI: 10.1016/j.phytochem.2010.03.013
37. Grattapaglia D. **Progress in Myrtaceae genetics and genomics: Eucalyptus as the pivotal genus**. *Tree Genet. Genomes* (2012.0) **8** 463-508. DOI: 10.1007/s11295-012-0491-x
38. 38.Takeda, L. N. et al. Psidium guajava L.: A systematic review of the multifaceted health benefits and economic importance. Food Rev. Int. 1–31 (2022).
39. Thornhill AH. **Interpreting the modern distribution of Myrtaceae using a dated molecular phylogeny**. *Mol. Phylogenet. Evol.* (2015.0) **93** 29-43. DOI: 10.1016/j.ympev.2015.07.007
40. Martin DM. **Functional annotation, genome organization and phylogeny of the grapevine (**. *BMC Plant Biol.* (2010.0) **10** 1-22. DOI: 10.1186/1471-2229-10-226
41. Jiang SY. **A comprehensive survey on the terpene synthase gene family provides new insight into its evolutionary patterns**. *Genome Biol. Evol.* (2019.0) **11** 2078-2098. DOI: 10.1093/gbe/evz142
42. Tucker AO. **Volatile leaf oils of American myrtaceae. III.**. *J. Essent. Oil Res.* (1995.0) **7** 187-190. DOI: 10.1080/10412905.1995.9698497
43. Marin R. **Volatile components and antioxidant activity from some Myrtaceous fruits cultivated in Southern Brazil**. *Lat. Am. J. Pharm.* (2008.0) **27** 172
44. Adam F. **Aromatic plants of French Polynesia. V. Chemical composition of essential oils of leaves of**. *J. Essent. Oil Res.* (2011.0) **23** 98-101. DOI: 10.1080/10412905.2011.9700433
45. Soliman FM. **Comparative study of the volatile oil content and antimicrobial activity of**. *Bull. Fac. Pharm. Cairo Univ.* (2016.0) **54** 219-225
46. Köllner TG. **The variability of sesquiterpenes emitted from two Zea mays cultivars is controlled by allelic variation of two terpene synthase genes encoding stereoselective multiple product enzymes**. *Plant Cell* (2004.0) **16** 1115-1131. DOI: 10.1105/tpc.019877
47. Tholl D. **Two sesquiterpene synthases are responsible for the complex mixture of sesquiterpenes emitted from Arabidopsis flowers**. *Plant J.* (2005.0) **42** 757-771. DOI: 10.1111/j.1365-313X.2005.02417.x
48. Keszei A, Brubaker CL, Foley WJ. **A molecular perspective on terpene variation in Australian Myrtaceae**. *Aust. J. Bot.* (2008.0) **56** 197-213. DOI: 10.1071/BT07146
49. Souza TDS. **Essential oil of**. *Sci. Hortic.* (2017.0) **216** 38-44. DOI: 10.1016/j.scienta.2016.12.026
50. Brophy JJ. **Leaf essential oils of the genus Leptospermum (Myrtaceae) in eastern Australia, Part 6. Leptospermum polygalifolium and allies**. *Flavour Fragr. J.* (2000.0) **15** 271-277. DOI: 10.1002/1099-1026(200007/08)15:4<271::AID-FFJ910>3.0.CO;2-E
51. Schemske DW. **Is there a latitudinal gradient in the importance of biotic interactions?**. *Annu. Rev. Ecol. Evol. Syst.* (2009.0) **40** 245-269. DOI: 10.1146/annurev.ecolsys.39.110707.173430
52. He SM. **De novo transcriptome characterization of rhodomyrtus tomentosa leaves and identification of genes involved in α/β-pinene and β-caryophyllene biosynthesis**. *Front. Plant Sci.* (2018.0) **9** 1231. DOI: 10.3389/fpls.2018.01231
53. Singsaas EL, Sharkey TD. **The effects of high temperature on isoprene synthesis in oak leaves**. *Plant Cell Environ.* (2000.0) **23** 751-757. DOI: 10.1046/j.1365-3040.2000.00582.x
54. Sharkey TD, Yeh S. **Isoprene emission from plants**. *Annu. Rev. Plant Biol.* (2001.0) **52** 407-436. DOI: 10.1146/annurev.arplant.52.1.407
55. Butler JB. **Annotation of the Corymbia terpene synthase gene family shows broad conservation but dynamic evolution of physical clusters relative to Eucalyptus**. *Heredity* (2018.0) **121** 87-104. DOI: 10.1038/s41437-018-0058-1
56. Kainer D. **High marker density GWAS provides novel insights into the genomic architecture of terpene oil yield in Eucalyptus**. *New Phytol.* (2019.0) **223** 1489-1504. DOI: 10.1111/nph.15887
57. Goodger JQ. **Monoterpene synthases responsible for the terpene profile of anther glands in**. *Tree Physiol.* (2021.0) **41** 849-864. DOI: 10.1093/treephys/tpaa161
58. Raguso RA. **Phylogenetic fragrance patterns in Nicotiana sections Alatae and Suaveolentes**. *Phytochemistry* (2006.0) **67** 1931-1942. DOI: 10.1016/j.phytochem.2006.05.038
59. Fähnrich A, Neumann M, Piechulla B. **Characteristic alatoid ‘cineole cassette’monoterpene synthase present in**. *Plant Mol. Biol.* (2014.0) **85** 135-145. DOI: 10.1007/s11103-014-0176-y
60. Matarese F, Scalabrelli G, D’Onofrio C. **Analysis of the expression of terpene synthase genes in relation to aroma content in two aromatic**. *Funct. Plant Biol.* (2013.0) **40** 552-565. DOI: 10.1071/FP12326
61. Chen H. **Positive Darwinian selection is a driving force for the diversification of terpenoid biosynthesis in the genus**. *BMC Plant Biol.* (2014.0) **14** 1-12. DOI: 10.1186/s12870-014-0239-x
62. Köllner TG, Gershenzon J, Degenhardt J. **Molecular and biochemical evolution of maize terpene synthase 10, an enzyme of indirect defense**. *Phytochemistry* (2009.0) **70** 1139-1145. DOI: 10.1016/j.phytochem.2009.06.011
63. Cseke L, Dudareva N, Pichersky E. **Structure and evolution of linalool synthase**. *Mol. Biol. Evol.* (1998.0) **15** 1491-1498. DOI: 10.1093/oxfordjournals.molbev.a025876
64. Boachon B. **CYP76C1 (Cytochrome P450)-mediated linalool metabolism and the formation of volatile and soluble linalool oxides in Arabidopsis flowers: A strategy for defense against floral antagonists**. *Plant Cell* (2015.0) **27** 2972-2990. PMID: 26475865
65. Kessler A, Baldwin IT. **Defensive function of herbivore-induced plant volatile emissions in nature**. *Science* (2001.0) **291** 2141-2144. DOI: 10.1126/science.291.5511.2141
66. Moraes CM, Mescher MC, Tumlinson JH. **Caterpillar-induced nocturnal plant volatiles repel conspecific females**. *Nature* (2001.0) **410** 577-580. DOI: 10.1038/35069058
67. Prosser IM. **Cloning and functional characterisation of a cis-muuroladiene synthase from black peppermint (Mentha × piperita) and direct evidence for a chemotype unable to synthesise farnesene**. *Phytochemistry* (2006.0) **67** 1564-1571. DOI: 10.1016/j.phytochem.2005.06.012
68. Phillips MA. **cDNA isolation, functional expression, and characterization of (+)-α-pinene synthase and (−)-α-pinene synthase from loblolly pine (**. *Arch. Biochem. Biophys.* (2003.0) **411** 267-276. DOI: 10.1016/S0003-9861(02)00746-4
69. Maruyama T, Ito M, Honda G. **Molecular cloning, functional expression and characterization of (E)-β-farnesene synthase from**. *Biol. Pharm. Bull.* (2001.0) **24** 1171-1175. DOI: 10.1248/bpb.24.1171
70. Starks CM. **Structural basis for cyclic terpene biosynthesis by tobacco 5-epi-aristolochene synthase**. *Science* (1997.0) **277** 1815-1820. DOI: 10.1126/science.277.5333.1815
71. Rynkiewicz MJ, Cane DE, Christianson DW. **Structure of trichodiene synthase from**. *Proc. Natl. Acad. Sci.* (2001.0) **98** 13543-13548. DOI: 10.1073/pnas.231313098
72. Whittington DA. **Bornyl diphosphate synthase: Structure and strategy for carbocation manipulation by a terpenoid cyclase**. *Proc. Natl. Acad. Sci.* (2002.0) **99** 15375-15380. DOI: 10.1073/pnas.232591099
73. Santos FA, Rao VSN. **Antiinflammatory and antinociceptive effects of 1, 8-cineole a terpenoid oxide present in many plant essential oils**. *Phytother. Res. Int. J. Devoted Pharmacol. Toxicol. Eval. Nat. Prod. Deriv.* (2000.0) **14** 240-244
74. Karthikeyan R. **Alpha pinene modulates UVA-induced oxidative stress, DNA damage and apoptosis in human skin epidermal keratinocytes**. *Life Sci.* (2018.0) **212** 150-158. DOI: 10.1016/j.lfs.2018.10.004
75. Rufino AT. **Anti-inflammatory and chondroprotective activity of (+)-α-pinene: Structural and enantiomeric selectivity**. *J. Nat. Prod.* (2014.0) **77** 264-269. DOI: 10.1021/np400828x
76. Rodenak-Kladniew B. **1, 8-Cineole promotes G0/G1 cell cycle arrest and oxidative stress-induced senescence in HepG2 cells and sensitizes cells to anti-senescence drugs**. *Life Sci.* (2020.0) **243** 117271. DOI: 10.1016/j.lfs.2020.117271
77. Pereira I. **Linalool bioactive properties and potential applicability in drug delivery systems**. *Colloids Surf. B* (2018.0) **171** 566-578. DOI: 10.1016/j.colsurfb.2018.08.001
78. Vilela GR. **Activity of essential oil and its major compound, 1, 8-cineole, from**. *J. Stored Prod. Res.* (2009.0) **45** 108-111. DOI: 10.1016/j.jspr.2008.10.006
79. Nóbrega JR. **Antifungal action of α-pinene against**. *Nat. Prod. Res.* (2021.0) **35** 6190-6193. DOI: 10.1080/14786419.2020.1837803
80. Kim DY. **Eucalyptol ameliorates Snail1/β-catenin-dependent diabetic disjunction of renal tubular epithelial cells and tubulointerstitial fibrosis**. *Oncotarget* (2017.0) **8** 106190. DOI: 10.18632/oncotarget.22311
81. Marzoug HNB. *Molecules* (2011.0) **16** 1695-1709. DOI: 10.3390/molecules16021695
82. Bouzenna H. **Potential protective effects of alpha-pinene against cytotoxicity caused by aspirin in the IEC-6 cells**. *Biomed. Pharmacother.* (2017.0) **93** 961-968. DOI: 10.1016/j.biopha.2017.06.031
83. Romagni JG, Allen SN, Dayan FE. **Allelopathic effects of volatile cineoles on two weedy plant species**. *J. Chem. Ecol.* (2000.0) **26** 303-313. DOI: 10.1023/A:1005414216848
84. Utegenova GA. **Chemical composition and antibacterial activity of essential oils from**. *Molecules* (2018.0) **23** 1679. DOI: 10.3390/molecules23071679
85. Langsi JD. **Evaluation of the insecticidal activities of α-Pinene and 3-Carene on**. *Insects* (2020.0) **11** 540. DOI: 10.3390/insects11080540
86. Allenspach M, Steuer C. **α-Pinene: A never-ending story**. *Phytochemistry* (2021.0) **190** 112857. DOI: 10.1016/j.phytochem.2021.112857
87. 87.Brasil. Farmacopeia Brasileira. Agência Nacional de Vigilância Sanitária. Anvisa, Brasília 2, 1265–1269 (2010).
88. Adams RP. **Identification of essential oil components by gas chromatography/mass spectrometry**. *Carol Stream Allured Publ. Corp.* (2007.0) **456** 544-545
89. 89.NIST (National Institute of Standards and Technology). Standard Reference Database 69. NIST (2011).
90. 90.El-Sayed, A. M. The Pherobase: Database of Pheromones and Semiochemicals. http://www.pherobase.com, 10 out (2021).
91. Finn RD. **The Pfam protein families database: Towards a more sustainable future**. *Nucleic Acids Res.* (2016.0) **44** D279-D285. DOI: 10.1093/nar/gkv1344
92. Finn RD, Clements J, Eddy SR. **HMMER web server: Interactive sequence similarity searching**. *Nucleic Acids Res.* (2011.0) **39** W29-W37. DOI: 10.1093/nar/gkr367
93. Degenhardt J, Köllner TG, Gershenzon J. **Monoterpene and sesquiterpene synthases and the origin of terpene skeletal diversity in plants**. *Phytochemistry* (2009.0) **70** 1621-1637. DOI: 10.1016/j.phytochem.2009.07.030
94. Durairaj J. **An analysis of characterized plant sesquiterpene synthases**. *Phytochemistry* (2019.0) **158** 157-165. DOI: 10.1016/j.phytochem.2018.10.020
95. Hu B. **GSDS 2.0: An upgraded gene feature visualization server**. *Bioinformatics* (2015.0) **31** 1296-1297. DOI: 10.1093/bioinformatics/btu817
96. Letunic I, Doerks T, Bork P. **SMART 6: Recent updates and new developments**. *Nucleic Acids Res.* (2009.0) **37** D229-D232. DOI: 10.1093/nar/gkn808
97. Emanuelsson O, Nielsen H, Von Heijne G. **ChloroP, a neural network-based method for predicting chloroplast transit peptides and their cleavage sites**. *Protein Sci.* (1999.0) **8** 978-984. DOI: 10.1110/ps.8.5.978
98. Emanuelsson O, Nielsen H, Brunak S, Von Heijne G. **Predicting subcellular localization of proteins based on their N-terminal amino acid sequence**. *J. Mol. Biol.* (2000.0) **300** 1005-1016. DOI: 10.1006/jmbi.2000.3903
99. Schein AI, Kissinger JC, Ungar LH. **Chloroplast transit peptide prediction: A peek inside the black box**. *Nucleic Acids Res.* (2001.0) **29** e82-e82. DOI: 10.1093/nar/29.16.e82
100. 100.Tholl, D. & Lee, S. Terpene specialized metabolism in Arabidopsis thaliana. Arabidopsis Book Am. Soc. Plant Biol.9 (2011).
101. Dudareva N. **(E)-β-Ocimene and myrcene synthase genes of floral scent biosynthesis in snapdragon: Function and expression of three terpene synthase genes of a new terpene synthase subfamily**. *Plant Cell* (2003.0) **15** 1227-1241. DOI: 10.1105/tpc.011015
102. Huang XZ. **The terpene synthase gene family in**. *Plant Cell Environ.* (2018.0) **41** 261-274. DOI: 10.1111/pce.13088
103. Kumar S. **MEGA X: Molecular evolutionary genetics analysis across computing platforms**. *Mol. Biol. Evol.* (2018.0) **35** 1547. DOI: 10.1093/molbev/msy096
104. Castresana J. **Selection of conserved blocks from multiple alignments for their use in phylogenetic analysis**. *Mol. Biol. Evol.* (2000.0) **17** 540-552. DOI: 10.1093/oxfordjournals.molbev.a026334
105. Guindon S, Gascuel O. **A simple, fast, and accurate algorithm to estimate large phylogenies by maximum likelihood**. *Syst. Biol.* (2003.0) **52** 696-704. DOI: 10.1080/10635150390235520
106. Guindon S. **PHYML Online—A web server for fast maximum likelihood-based phylogenetic inference**. *Nucleic Acids Res.* (2005.0) **33** W557-W559. DOI: 10.1093/nar/gki352
107. Keane TM. **Assessment of methods for amino acid matrix selection and their use on empirical data shows that ad hoc assumptions for choice of matrix are not justified**. *BMC Evol. Biol.* (2006.0) **6** 1-17. DOI: 10.1186/1471-2148-6-29
108. 108.Morariu, V. et al. Automatic online tuning for fast Gaussian summation. Adv. Neural Inf. Process. Syst.21 (2008).
109. Gao F. **EasyCodeML: A visual tool for analysis of selection using CodeML**. *Ecol. Evol.* (2019.0) **9** 3891-3898. DOI: 10.1002/ece3.5015
110. Yang Z. **PAML: A program package for phylogenetic analysis by maximum likelihood**. *Comput. Appl. Biosci.* (1997.0) **13** 555-556. PMID: 9367129
111. Wong WS. **Accuracy and power of statistical methods for detecting adaptive evolution in protein coding sequences and for identifying positively selected sites**. *Genetics* (2004.0) **168** 1041-1051. DOI: 10.1534/genetics.104.031153
112. Yang Z, Wong WS, Nielsen R. **Bayes empirical Bayes inference of amino acid sites under positive selection**. *Mol. Biol. Evol.* (2005.0) **22** 1107-1118. DOI: 10.1093/molbev/msi097
113. Whelan S, Goldman N. **Distributions of statistics used for the comparison of models of sequence evolution in phylogenetics**. *Mol. Biol. Evol.* (1999.0) **16** 1292-1292. DOI: 10.1093/oxfordjournals.molbev.a026219
114. Omasits U. **Protter: Interactive protein feature visualization and integration with experimental proteomic data**. *Bioinformatics* (2014.0) **30** 884-886. DOI: 10.1093/bioinformatics/btt607
115. 115.Vetö, N. M. et al. Transcriptomics analysis of Psidium cattleyanum Sabine (Myrtaceae) unveil potential genes involved in fruit pigmentation. Genet. Mol. Biol.43 (2020).
116. Kim D. **TopHat2: Accurate alignment of transcriptomes in the presence of insertions, deletions and gene fusions**. *Genome Biol.* (2013.0) **14** 1-13. DOI: 10.1186/gb-2013-14-4-r36
117. Liao Y, Smyth GK, Shi W. **The Subread aligner: Fast, accurate and scalable read mapping by seed-and-vote**. *Nucleic Acids Res.* (2013.0) **41** e108-e108. DOI: 10.1093/nar/gkt214
118. Love MI, Huber W, Anders S. **Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2**. *Genome Biol.* (2014.0) **15** 1-21. DOI: 10.1186/s13059-014-0550-8
119. Benjamini Y, Hochberg Y. **Controlling the false discovery rate: A practical and powerful approach to multiple testing**. *J. R. Stat. Soc. Ser. B (Methodol.)* (1995.0) **57** 289-300
|
---
title: Central corneal thickness and its associations in a Russian population. The
Ural eye and Medical Study
authors:
- Mukharram M. Bikbov
- Timur R. Gilmanshin
- Rinat M. Zainullin
- Gyulli M. Kazakbaeva
- Artur F. Zaynetdinov
- Ildar F. Nuriev
- Songhomitra Panda-Jonas
- Inga I. Arslangareeva
- Ainur A. Zinnatullin
- Dilya F. Yakupova
- Ellina M. Rakhimova
- Yulia V. Uzianbaeva
- Renat I. Khikmatullin
- Nikolay A. Nikitin
- Said K. Aminev
- Svetlana R. Mukhamadieva
- Venera F. Mavlieva
- Jost B. Jonas
journal: Eye
year: 2022
pmcid: PMC9998395
doi: 10.1038/s41433-022-02026-1
license: CC BY 4.0
---
# Central corneal thickness and its associations in a Russian population. The Ural eye and Medical Study
## Abstract
### Background
To assess central corneal thickness (CCT) and its associations in a Russian population.
### Methods
The population-based Ural Eye and Medical Study included 5899 ($80.5\%$) out of 7328 eligible individuals. As part of an ophthalmological and general examination, CCT was measured by Scheimflug imaging.
### Results
The study included 5792 ($98.2\%$) participants (age:58.8 ± 10.6 years;range: 40–94 years) with available bilateral CCT measurements. Mean CCT was larger in Russians than non-Russians (549.5 ± 32.8 µm versus 539.2 ± 33.9 µm; $P \leq 0.001$). In multivariable analysis, thicker CCT was associated (regression coefficient r:0.43) with younger age (standardized regression coefficient beta:−0.09; non-standardized regression coefficient B:−0.29;$95\%$ confidence interval (CI):−0.39,−0.20; $P \leq 0.001$), male sex (beta:0.05; B:3.10; $95\%$CI:1.18,5.03; $$P \leq 0.002$$), urban region of habitation (beta:0.10; B:6.83; $95\%$CI:4.61, 9.05; $P \leq 0.001$), Russian ethnicity (beta:0.04; B:3.48; $95\%$CI:1.04, 5.91; $$P \leq 0.005$$), higher level of education (beta:0.04; B:0.97;$95\%$CI:0.29,1.66; $$P \leq 0.006$$), higher serum bilirubin concentration (beta:0.05;B:0.15; $95\%$CI:0.07,0.23;$P \leq 0.001$), lower corneal refractive power (beta:−0.09;B:11.92; $95\%$CI:−2.50,−1.35; $P \leq 0.001$), smaller anterior chamber angle (beta:−0.07;B:−0.38;$95\%$CI:−0.52,−0.24;$P \leq 0.001$), higher IOP readings (beta:0.38; B:3.47; $95\%$CI:3.21,3.73; $P \leq 0.001$), and higher rise in IOP readings by medical mydriasis (beta:0.07; B:0.88;$95\%$CI:0.54,1.22;$P \leq 0.001$). In that model, CCT was not associated with body height ($$P \leq 0.14$$), previous cataract surgery ($$P \leq 0.10$$), axial length ($$P \leq 0.18$$) or prevalence of glaucoma ($$P \leq 0.11$$). The mean inter-eye difference in CCT was 8.52 ± 13.9 µm (median:6.0;95CI:8.16,8.88). A higher inter-eye CCT difference was associated with older age (beta:0.08; B:0.11;$95\%$CI:0.07,0.15; $$P \leq 0.01$$), lower level of education (beta:−0.04;B:−0.34; $95\%$CI:−0.60,−0.08; $P \leq 0.001$) and status after cataract surgery (beta:0.04; B:2.92;$95\%$CI:1.02,4.83; $$P \leq 0.003$$).
### Introduction conclusions
In this ethnically mixed population from Russia with an age of 40+ years, mean CCT (541.7 ± 33.7 µm) was associated with parameters such as younger age, male sex, Russian ethnicity, and higher educational level. These associations may be taken into account when the dependence of IOP readings on CCT are considered. Glaucoma prevalence was unrelated to CCT.
## Introduction
Central corneal thickness (CCT) is a clinically important parameter in the diagnosis of glaucoma, since the measurement of intraocular pressure (IOP) markedly depends on CCT [1–5]. It has additionally been discussed that a thin cornea may be a structural risk factor for an increased susceptibility for glaucomatous optic nerve damage at a given IOP [6, 7]. It has therefore become clinical routine to measure CCT to correct the IOP readings for their dependence on CCT. In previous hospital-based studies and population-based investigations, CCT has been measured in various ethnic populations such as Western Europeans, East Asians including Japanese and Chinese, Mongolians, Malay, Indians and Hispanics [8–15]. None of these studies however assessed the CCT in a population from Russia, and in particular did not take into account the multi-ethnicity of the total population of Russia. In addition, in most of the previous investigations, associations of CCT with other parameters were tested for an only relatively small number of variables. We therefore conducted this study to measure the CCT in an ethnically mixed population Russia and to assess associations of CCT with a large number of other ocular parameters and systemic and medical variables.
## Methods
The Ural Eye and Medical *Study is* a population-based investigation which was performed in the Russian republic of Bashkortostan at the southwestern end of the Ural Mountains in the study period from 2015 to 2017 [16, 17]. Study regions were Ufa as capital of Bashkortostan in a distance of about 1400 km East of Moscow and a rural region in the Karmaskalinsky District in a distance of 65 km from Ufa. The republic of Bashkortostan located between the Volga River and the Ural Mountains, is with a population of 4 million people the most populous republic in Russia. Inclusion criteria for the study were living in the study regions and an age of 40 years or older. The Ethics Committee of the Academic Council of the Ufa Eye Research Institute approved the study design and confirmed that the study adhered to the Declaration of Helsinki, and all participants gave an informed written consent. Out of a total group of 7328 eligible individuals, 5899 ($80.5\%$) individuals (3319 [$56.3\%$] women) with a mean age of 59.0 ± 10.7 years (range: 40–94 years) participated in the study. The study population did not differ significantly in the gender and age distribution from the Russian population as explored in the census carried out in 2010 [16].
Using a bus, the study participants were brought from their homes to the Ufa Eye Institute where a team of about 20 trained social workers, technicians and ophthalmologists performed all examinations. As also described in detail previously, the series of examinations started with a detailed interview consisting of more than 250 standardized questions on the socioeconomic background, smoking and alcohol consumption, physical activity, diet, depression and anxiety, and known diagnosis and therapy of major diseases [17, 18]. The examinations further included anthropometry, blood pressure measurement, handgrip dynamometry, spirometry, and biochemical analysis of blood samples taken under fasting conditions. We defined arterial hypertension according to the new criteria published by the American Heart Association, and criteria for the diagnosis of diabetes mellitus were a fasting serum glucose concentration of ≥7.0 mmol/L or a self-reported history of physician-based diagnosis or therapy of diabetes mellitus.
The series of ophthalmologic examinations consisted of measurement of visual acuity including automated and subjective refractometry (Auto-2Ref/Keratometer HRK-7000A HUVITZ Co, Ltd., Gyeonggi-do, Korea), perimetry (PTS 1000 Perimeter, Optopol Technology Co., Zawercie, Poland), Scheimflug imaging of the anterior segment, slit lamp-based biomicroscopy of the anterior and posterior ocular segment, non-contact tonometry (Tonometer Kowa KT-800, Kowa Company Ltd., Hamamatsu City, Japan), re-assessment of the anterior segment and lens for the presence of pseudoexfoliation after medical mydriasis, photography of the cornea and lens (Topcon slit lamp and camera, Topcon Corp. Tokyo, Japan), optical coherence tomography (OCT) (RS-3000, NIDEK co., Ltd., Aichi Japan) of the peripapillary retinal nerve fibre layer, optic nerve head and macula, and assessment of the degree of fundus tessellation using the fundus photographs. Using an anterior segment imaging device (Pentacam HR, Typ70900, OCULUS, Optikgeräte GmbH Co., Wetzlar, Germany), we measured the CCT by Scheimflug imaging. We applied the automatic mode for the data assessment. In the case of CCT readings out of the range of expectation, the measurements were repeated. All measurements were performed by the same ophthalmologist trained and supervised in the technique. Nuclear lens opacities were differentiated into 6 grades using the classifying scheme for cataract of the Age-Related Eye Disease Study [19]. We defined the presence of nuclear cataract as a nuclear cataract grade of 3 or higher. Cortical lens opacities and posterior subcapsular opacities were graded using photographs taken by retro-illumination (Topcon slit lamp and camera, Topcon Corp. Tokyo, Japan). Using a grid, we measured the percentage area of opacity. Age-related macular degeneration (AMD) was defined as suggested by the recent Beckman Initiative for Macular Research Classification Committee [20]. Glaucoma was defined by morphological criteria as described by Foster and colleagues [21].
Inclusion criterion for the present study was the availability of CCT measurements. The data of only one randomly selected eye per individual was taken for statistical analysis. The data were statistically analyzed using a statistical software package (SPSS for Windows, version 25.0, IBM-SPSS, Chicago, IL, USA). We assessed the mean values of the parameters (expressed as mean and standard deviation or as mean and $95\%$ confidence intervals (CI)) and examined associations between CCT and other systemic parameters and ocular parameters, first in a univariable analysis, followed by a multivariable analysis. The latter included the CCT as dependent variable and as independent parameters all those variables that were associated ($P \leq 0.05$) with CCT in the univariable analyses. Out of the list of independent variables, we then dropped parameters due to collinearity with other independent variables or if they were no longer significantly associated with CCT. We calculated odds ratios (OR) and their $95\%$ CI. All P-values were two-sided and considered statistically significant when the values were less than 0.05.
## Results
Out of 5889 participants of the Ural Eye and Medical Study, the present study included 5792 ($98.2\%$) individuals (mean age: 58.8 ± 10.6 years; range: 40–94 years) with available bilateral CCT measurements. The study population consisted of 1170 ($20.2\%$) Russians, 1045 ($18.0\%$) Bashkirs, 2394 ($41.3\%$) Tartars, 579 ($10.0\%$) Chuvash, 21 ($0.4\%$) Mari, and 583 ($10.1\%$) other or undefined ethnic groups. The group of individuals with CCT measurements and participating in the present investigations as compared to the group of those without CCT readings was significantly younger (58.8 ± 10.6 years versus 67.5 ± 11.4 years; $P \leq 0.001$) and included proportionally more women ($$P \leq 0.03$$).
Mean CCT was 541.7 ± 33.7 µm in the right eyes (median: 541 µm; range: 200–779 µm) and 543.1 ± 33.7 µm in the left eyes (median: 543 µm,; range: 174–801 µm) with a significant ($P \leq 0.001$) difference between both eyes (Figs. 1,2). The mean CCT was 549.5 ± 32.8 µm (median: 548 µm; range: 397–664 µm) in the Russian group and it was 539.2 ± 33.9 µm (median: 538 µm; range: 200–779 µm) in the non-Russian group, with a significant ($P \leq 0.001$) difference between both groups. Fig. 1Central Corneal Thickness. Histogram showing the distribution of central corneal thickness in the Ural Eye and Medical Study. Fig. 2Bilateral Difference in Central Corneal Thickness. Histogram showing the distribution of the inter-eye side difference in central corneal thickness in the Ural Eye and Medical Study.
In univariable analysis, thicker CCT was associated ($P \leq 0.05$) with the systemic parameters of older age, male sex, urban region of habitation, Russian ethnicity, taller body height, heavier body weight, higher body mass index, higher socioeconomic score and higher level of education, higher prevalence of current smoking and higher number of cigarette smoking package years, higher prevalence if any alcohol consumption, higher prevalence of a history of unconsciousness, menopause and diabetes mellitus, higher serum concentration of alanine aminotransferase, aspartate aminotransferase and total bilirubin and haemoglobin, lower serum concentration of urea, lower erythrocyte sedimentation rate, higher erythrocyte count, higher percentage of segment nuclear granulocytes on total leucocytes, higher estimated glomerular filtration rate and lower prevalence of chronic kidney disease and anaemia, higher prevalence of diabetes mellitus, lower hearing loss score, and higher dynamometric hand grip force (Table 1). Thicker CCT was associated with the ocular parameters of longer axial length, lower corneal refractive power, higher corneal volume, lower anterior chamber volume and smaller anterior chamber angle degree, lower degree and prevalence of cortical cataract, higher IOP readings before and after medical mydriasis, lower difference between the IOP readings obtained before and after medical inducing mydriasis, thicker retinal thickness measured 300 µm temporal to the fovea, higher prevalence of angle-closure glaucoma and diabetic retinopathy, higher Schirmer test, and lower prevalence of reticular pseudodrusen of the macula (Table 2).Table 1Associations (univariable analysis) between central corneal thickness systemic parameters in the Ural Eye and Medical Study. ParameterIntervalStandardized regression coefficient betaNon- standardized regression coefficient B$95\%$ confidence interval of BP-ValueAge1-year intervals−0.09−0.27−0.35, −0.19<0.001GenderMen / Women−0.06−3.72−5.47, −1.97<0.001Region of habitationRural / Urban0.149.477.72, 11.2<0.001EthnicityAny other ethnicity / Russian0.1210.17.91, 12.2<0.001Body height1 cm0.070.270.17, 0.37<0.001Body weightkg0.080.190.13, 0.25<0.001Body mass indexkg/m20.050.310.14, 0.49<0.001Waist circumferencecm0.030.06−0.001, 0.130.053Hip circumferencecm0.010.03−0.04, 0.100.44Waist/hip circumference ratioRatio0.039.21−0.39, 18.80.06Socioeconomic ScoreScore0.092.001.43, 2.57<0.001Level of educationIlliteracy / Passing 5th Grade / 8th Grade / 10th Grade / 11th Grade / Graduates / Specialized Secondary Education / Post Graduates0.102.361.76, 2.96<0.001Physical activity ScoreScore0.010.03−0.09, 0.140.64Smoking, currentlyNo / Yes0.044.721.65, 6.890.001Smoking, package yearsNumber0.030.080.01, 0.150.03Alcohol consumption, anyNo / Yes0.032.170.05, 4.290.045In a week how many days do you eat fruits?Number of days0.010.09−0.35, 0.530.67In a week how many days do you eat vegetables?Number of days−0.01−0.25−0.86, 0.370.43How much salt do you consume every day?g0.010.15−0.24, 0.540.44History of cardiovascular disorders including strokeNo / Yes0.021.37−0.68, 3.420.19History of angina pectorisNo / Yes0.011.11−1.90, 4.130.47History of asthmaNo / Yes0.012.73−2.56, 8.030.31History of arthritisNo / Yes−0.02−1.38−3.32, 0.570.17History of previous bone fracturesNo / Yes0.021.60−0.38, 3.580.11History of low back painNo / Yes0.021.34−0.49, 3.170.15History of thoracic spine painNo / Yes−0.0040.32−2.47, 1.820.77History of neck painNo / Yes0.021.59−0.41, 3.600.12History of headacheNo / Yes−0.01−0.91−2.74, 0.910.33History of cancerNo / Yes0.012.47−2.68, 7.620.35History of dementiaNo / Yes−0.027.73−19.0, 3.550.18History of diarrhoeaNo / Yes−0.001−0.35−13.2, 12.50.96History of iron-deficiency anaemiaNo / Yes0.011.29−2.65, 5.230.52History of low blood pressure and hospital admittanceNo / Yes0.011.97−2.72, 6.670.41History of osteoarthritisNo / Yes0.000.04−2.31, 2.400.97History of skin diseaseNo / Yes0.022.69−1.37, 6.740.19History of thyreopathyNo / Yes0.0040.41−2.45, 3.260.78History of tumblingNo / Yes0.021.72−0.51, 3.950.13History of unconsciousnessNo / Yes0.033.200.04, 6.360.047Age of the last menstrual bleedingYears0.020.15−0.12, 0.420.28Age of last regular menstrual bleedingYears0.030.17−0.10, 0.450.22History of menopauseNo / Yes−0.05−4.52−7.60, 1.450.004History of diabetes mellitusNo / Yes0.044.361.24, 7.490.006Serum concentration of:Alanine aminotransferaseIU/L0.040.100.03, 0.180.005Aspartate aminotransferaseIU/L0.030.100.02, 0.180.01Aspartate aminotransferase-to- Alanine aminotransferase ratioRatio−0.003−0.19−1.97, 1.580.83Bilirubin, totalµmol/L0.040.110.04, 0.190.004High-density lipoproteinsmmol/L−0.01−0.52−1.54, 0.510.32Low-density lipoproteinsmmol/L−0.003−0.10−0.87, 0.670.80Cholesterolmmol/L0.010.24−0.28, 0.750.36Triglyceridesmmol/L0.031.17−0.06, 2.390.06Rheumatoid factorIU/mL−0.02−0.60−1.54, 0.340.21Erythrocyte sedimentation rateMm/min−0.6−0.19−0.27, −0.11<0.001Glucosemmol/L0,020.33−0.19, 0.850.22Ureammol/L−0.04−0.90−1.49, −0.300.003Creatinineµmol/L0.0030.005−0.03, 0.040.79Haemoglobing/L0.070.150.09, 0.21<0.001Erythrocyte count106 cells / µL0.065.523.24, 7.79<0.001Leucocyte count109 cells / L0.010.14−0.47, 0.750.65Rod-core granulocytes% of leucocytes−0.01−0.17−0.80, 0.470.61Segment nuclear granulocyte% of leucocytes0.030.150.03, 0.270.01Eosinophil granulocytes% of leucocytes−0.01−0.39−1.25, 0.470.37Lymphocytes% of leucocytes−0.01−0.06−0.20, 0.080.40Monocytes% of leucocytes−0.04−0.51−0.89, −0.130.009Diabetes mellitus, prevalenceYes/No0.044.321.59, 7.040.002Estimated glomerular filtration rate30 mL/min/1.73 m²0.050.090.04, 0.13<0.001Stage of chronic kidney disease0–5−0.04−0.16−0.28, −0.040.007AnaemiaNo / Yes−0.04−3.24−5.29, −1.190.002Blood pressure, systolicmm Hg0.010.01−0.03, 0.050.69Blood pressure, diastolicmm Hg0.020.08−0.04, 0.190.18Blood pressure, meanmm Hg0.020.05−0.02, 0.120.15Arterial hypertensionYes/No0.010.76−0.98, 2.500.39Arterial hypertension, stage0–40.010.22−0.62, 1–070.61Ankle-brachial pressure index, right0.0020.43−5.91, 6.770.89Prevalence of chronic obstructive pulmonary diseaseYes/No0.011.19−2.42, 4.790.52Hearing lossHearing loss score (0–44)−0.05−0.15−0.23, −0.060.001Depression ScoreDepression score unit (range: −4 to +15)−0.02−0.13−0.36, 0.100.27State-Trait Anxiety InventoryState-Trait Anxiety Inventory Score (range: −7 to 13)−0.03−0.23−0.48, 0.010.06Manual dynamometry, right handdekaNewton0.090.260.18, 0.34<0.001Manual dynamometry, right handdekaNewton0.090.27−0.19, 0.35<0.001Table 2Associations (univariable analysis) between central corneal thickness and ocular parameters in the Ural Eye and Medical Study. ParameterIntervalStandardized regression coefficient betaNon- standardized regression coefficient B$95\%$ confidence interval of BP-ValueRefractive error, spherical equivalentDioptres0.020.250.14, 0.230.21Refractive error, cylindrical valueDioptres0.020.75−0.37, 1.860.19Axial lengthmm0.041.320.52, 2.120.001Corneal refractive powerDioptres−0.13−2.62−3.16, −2.08<0.001Corneal volumemm30.735.995.84, 6.14<0.001Anterior chamber depthmm−0.01−0.50−2.28, 1.290.59Anterior chamber volumeµL−0.08−0.08−0.10, −0.05<0.001Anterior chamber angleDegree−0.07−0.35−0.48, −0.23<0.001Lens thicknessmm−0.01−1.09−3.24, 1.070.32Nuclear cataract degreeGrade−0.02−0.71−1.63, 0.220.14Nuclear cataract, prevalenceNo / Yes−0.01−0.80−2.71, 1.110.41Cortical cataract, degreePercentage−0.04−0.13−0.23, −0.030.008Cortical cataract, prevalenceNo / Yes−0.04−4.19−6.97, −1.420.003Subcapsular cataract, degreePercentage0.010.16−0.22, 0.550.41Subcapsular cataract, prevalenceNo / Yes0.0010.43−11.8, 12.60.95Fundus tessellation, macula regionGrade0.0010.03−1.04, 1.100.95Fundus tessellation, peripapillary regionGrade−0.006−0.19−1.10, 0.720.68Intraocular pressure, before MydriasismmHg0.342.952.74, 3.16<0.001Intraocular pressure, after MydriasismmHg0.322.722.48, 2.96<0.001Intraocular pressure, difference “after Mydriasis” minus “before mydriasis”mmHg−0.04−0.43−0.78, −0.070.02Retinal thickness (total), foveaµm0.010.01−0.01, 0.020.46Retinal thickness (total), 300 µm temporal to the foveaµm0.030.030.002, 0.050.03Retinal thickness (total), 300 µm nasal to the foveaµm0.020.02−0.01, 0.040.14Retinal nerve fibre layer thickness, peripapillaryµm0.030.05−0.003, 0.100.07Pterygium, prevalenceNo / Yes−0.03−8.31−15.1, 1.610.02Pseudoexfoliation,No / Yes−0.01−0.94−5.74, 3.860.70Glaucoma, prevalenceNo / Yes0.011.822.79, 6.430.44Glaucoma stage0–5−0.01−0.71−2.64, 1.210.47Open-angle glaucoma, prevalenceNo / Yes−0.01−1.90−7.37, 3.560.50Angle-closure glaucoma, prevalenceNo / Yes0.0310.92.38, 19.30.01Diabetic retinopathy, prevalenceNo / Yes0.049.562.52, 16.60.008Diabetic retinopathy, ETDRS gradingScale0.030.22−0.03, 0.460.08Myopic maculopathy, stage0–4−0.01−0.80−3.37, 1.780.55Age-related macular degeneration, early stage, prevalenceNo / Yes−0.01−1.56−5.26, 2.150.41Age-related macular degeneration, intermediate stage, prevalenceNo / Yes0.0020.40−4.10, 4.910.86Age-related macular degeneration, late stage, prevalenceNo / Yes−0.01−1.23−7.24, 4.790.69Age-related macular degeneration, reticular pseudodrusen, prevalenceNo / Yes−0.04−6.18−10.5, −1.870.005Age-related macular degeneration, any stage, prevalenceNo / Yes−0.01−1.41−4.47, 1.670.37Dry eye, Schirmer testmm0.030.150.02, 0.290.03Meibomian gland dysfunctionGrade 0–4−0.02−0.67−1.70, 0.360.20Visual acuity, best correctedlogMAR−0.01−1.00−3.30, 1.300.39 In the multivariable analysis, we dropped due to collinearity, the parameters of body weight (versus body mass index); variance inflation factor (VIF: 116), package years (versus current smoking, VIF: 4.47), socioeconomic index (versus level of education, VIF:7.44), stage of chronic kidney disease (versus estimated glomerular filtration rate, VIF: 5.7), prevalence of anaemia (versus serum haemoglobin concentration, VIF: 2.4), prevalence of cortical cataract (versus degree of cortical cataract, VIF: 2.2), serum concentration of alanine aminotransferase (versus serum concentration of aspartate aminotransferase, VIF: 4.6), anterior chamber volume (versus anterior chamber angle, VIF: 2.6), and IOP after mydriasis (versus IOP before mydriasis). Due to a lack of statistic al significance, we dropped body mass index ($$P \leq 0.38$$), serum concentration of aspartate aminotransferase ($$P \leq 0.93$$), hearing loss score ($$P \leq 0.88$$), erythrocyte sedimentation rate ($$P \leq 0.86$$), percentage of segment nuclear granulocytes on total leucocytes ($$P \leq 0.62$$), serum concentration of urea ($$P \leq 0.59$$) and haemoglobin ($$P \leq 0.46$$), retinal thickness 300 µm temporal to the fovea ($$P \leq 0.71$$), prevalence of diabetic retinopathy ($$P \leq 0.55$$) and reticular pseudodrusen ($$P \leq 0.80$$), body height ($$P \leq 0.93$$), history of unconsciousness ($$P \leq 0.41$$), prevalence of diabetes ($$P \leq 0.61$$), Schirmer´s test ($$P \leq 0.34$$), dynamometric hand grip force ($$P \leq 0.35$$), prevalence of angle-closure glaucoma ($$P \leq 0.12$$), consumption any alcohol ($$P \leq 0.33$$), degree of cortical cataract ($$P \leq 0.25$$), axial length ($$P \leq 0.29$$), estimated glomerular filtration rate ($$P \leq 0.10$$), and prevalence of current smoking ($$P \leq 0.14$$).
In the final model, a thicker central corneal thickness was associated (regression coefficient r: 0.43) with younger age (standardized regression coefficient beta: −0.09; non-standardized regression coefficient B: −0.29; $95\%$ confidence interval (CI): −0.39, −0.20; $P \leq 0.001$), male sex (beta: 0.05; B: 3.10; $95\%$CI: 1.18, 5.03; $$P \leq 0.002$$), urban region of habitation (beta: 0.10; B: 6.83; $95\%$CI: 4.61, 9.05; $P \leq 0.001$), Russian ethnicity (beta: 0.04; B: 3.48; $95\%$CI: 1.04, 5.91; $$P \leq 0.005$$), higher level of education (beta: 0.04; B: 0.97; $95\%$CI: 0.29, 1.66; $$P \leq 0.006$$), higher serum bilirubin concentration (beta: 0.05; B: 0.15; $95\%$CI: 0.07, 0.23; $P \leq 0.001$), lower corneal refractive power (beta: −0.09; B: 11.92; $95\%$CI: −2.50, −1.35; $P \leq 0.001$), smaller anterior chamber angle (beta: −0.07; B: −0.38; $95\%$CI: −0.52, −0.24; $P \leq 0.001$), higher IOP readings (beta: 0.38; B: 3.47; $95\%$CI: 3.21, 3.73; $P \leq 0.001$), and higher rise in IOP readings by medical mydriasis (beta: 0.07; B: 0.88; $95\%$CI: 0.54, 1.22; $P \leq 0.001$) (Table 3). If the parameter of region of habitation and serum bilirubin concentration were separately dropped from the analysis, the other results remained mostly unchanged. If the parameters of diabetes prevalence ($$P \leq 0.43$$), glucose serum concentration ($$P \leq 0.89$$), body height ($$P \leq 0.14$$), previous cataract surgery ($$P \leq 0.10$$), axial length ($$P \leq 0.18$$) or prevalence of glaucoma ($$P \leq 0.11$$) were added to the model, they were not significantly associated with CCT. If the parameter of diabetes prevalence was kept in the list of independent parameters, its associations with a thicker CCT remained to be statistically significant, if the list additionally contained the parameters of age, sex, region of habitation, ethnicity, level of education, bilirubin serum concentration, corneal refractive power, and anterior chamber angle. If IOP was further added to the model, the parameter of diabetes prevalence lost its significance with CCT ($$P \leq 0.16$$). In univariable analysis, a higher diabetes prevalence correlated with a thicker CCT ($P \leq 0.001$; beta: 0.047).Table 3Associations (multivariable analysis) between central corneal thickness and ocular parameters in the Ural Eye and Medical Study. ParameterIntervalStandardized regression coefficient betaNon- standardized regression coefficient B$95\%$ confidence interval of BP-ValueAgeYears−0.09−0.29−0.39, −0.20<0.001GenderWomen/Men0.053.101.18, 5.030.002Region of HabitationRural/Urban0.106.834.61, 9.05<0.001EthnicityNon-Russian/Russian0.043.481.04, 5.910.005Level of educationIlliteracy/Passing 5th Grade /8th Grade/10th Grade/ 11th Grade/Graduates / Specialized Secondary Education/Post Graduates0.040.970.29, 1.660.006Serum concentration of bilirubinµmol/L0.050.150.07, 0.23<0.001Corneal refractive powerDioptres−0.09−1.92−2.50, −1.35<0.001Anterior chamber angleDegree−0.07−0.38−0.52, −0.24<0.001Intraocular pressure readingsmmHg0.070.880.54, 1.22<0.001Medical mydriasis associated rise in intraocular pressuremmHg0.383.473.21, 3.73<0.001 The mean inter-eye difference in CCT was 8.52 ± 13.9 µm (median: 6.0; $95\%$CI: 8.16, 8.88) (Fig. 2). A higher inter-eye CCT difference was associated with older age (beta:0.08; B: 0.11; $95\%$CI: 0.07, 0.15; $$P \leq 0.01$$), lower level of education (beta: −0.04; B: −0.34; $95\%$CI: −0.60, −0.08; $P \leq 0.001$) and status after cataract surgery (beta: 0.04; B: 2.92; $95\%$CI: 1.02, 4.83; $$P \leq 0.003$$)
## Discussion
In our ethnically mixed population from Russia, CCT (mean: 541.7 ± 33.7 µm) increased with the systemic parameters of younger age, higher level of education and higher serum bilirubin concentration, and with the ocular parameters of lower corneal refractive power, smaller anterior chamber angle, higher IOP readings, and higher rise in IOP readings by medical mydriasis. In addition, CCT was larger in men than women, urban versus rural region of habitation, and in Russians (543.4 ± 31.6 µm) versus non-Russians (539.2 ± 33.9 µm). In that multivariable model, CCT was not associated with body height, previous cataract surgery, axial length or prevalence of glaucoma A higher inter-eye difference in CCT (mean: 8.52 ± 13.9 µm) was correlated with older age, lower level of education and status after cataract surgery.
The mean CCT of 543 µm in Russians was considerably larger than the mean CCT found in Indians (514 µm; Central India and Medical Study; 504 µm, South Indian Chennai Glaucoma Study), Japanese (521 μm, Tajimi Study), indigenous Australians (512 mm) and Afro-Americans (530 mm; Barbados Eye Study), and it is comparable with the mean CCT reported for West-Europeans (537 μm, Rotterdam Study), North Chinese (556 µm, Beijing Eye Study), and Malays (541 μm, Singapore Malay Eye Study) [8, 12–15, 22–24]. Interestingly, the CCT in the non-Russian group in our study population, including Bashkirs, Tartars, Chuvash and other ethnic groups, was significantly thinner (539.2 ± 33.9 µm; $P \leq 0.001$) than in the Russian group. The findings obtained in our study further support the dependence of CCT on the ethnic background, so that the latter should be taken into account in the diagnosis of glaucoma, if the CCT measurements are not available.
The observation made in our study that CCT was not related with axial length, confirms previous investigations on other ethnic groups [11, 14, 23]. It supports the notion that myopic axial elongation occurs predominantly in the posterior hemisphere of the globe, while the anterior ocular segment including the cornea and its diameter and thickness is not affected by the process of axial elongation [25]. In our study population thicker CCT was associated with a lower corneal refractive power (i.e., a greater radius of corneal curvature or a flatter cornea). The correlation remained to be statistically significant, if eyes with a corneal refractive power of more than 45 dioptres were excluded, so that the relationship was not due to the inclusion of eyes with a keratoconus. The association between a larger thickness and more pronounced flatness of the cornea is of clinical interest since both parameters, a thicker cornea and a flatter cornea, lead to an underestimation of the true IOP [26, 27].
As also noted in numerous previous studies, a thicker CCT was associated with higher IOP readings [1, 2, 8–10, 13–15, 22–24]. In addition to this dependence of the IOP readings on CCT, the IOP measurements depend on the corneal flatness: the flatter the cornea is, the easier it is to applanate the cornea [26, 27].
Interestingly, CCT was not related with the prevalence of glaucoma as a whole or with the prevalence of open-angle glaucoma or angle-closure glaucoma. It is partially in contrast to the results of the Ocular Hypertension Treatment Study in which a thinner cornea at baseline was associated with a higher risk of developing glaucomatous optic nerve damage [7]. In agreement with other studies, such as the Liwan Study, it supports the notion that CCT is not a risk factor for glaucomatous optic neuropathy if the dependence of the IOP measurements on CCT have been taken into account during the diagnosis of glaucomatous optic neuropathy [28–31]. It fits with the results of a histomorphometric study that CCT and thickness of the lamina cribrosa were not significantly correlated with each other [32].
The prevalence of diabetes or the glucose serum concentration did not correlate with CCT in our study population. This observation does not agree with the finding made in the Singapore Malay Eye Study, in which CCT, as measured by ultrasound pachymetry and after controlling for age and gender, was significantly higher in individuals with diabetes than in those without diabetes ($P \leq 0.001$), and in which higher CCT was associated with higher serum glucose concentration ($$P \leq 0.02$$) and higher HbA1c value ($P \leq 0.001$) [33]. As in our study, other investigations, such as the Iranian Yazd Eye Study, the Korean Namil Study and the South Indian Sankara Nethralaya Diabetic Retinopathy also did not find an association between CCT and diabetes [34–36].
In our study population, CCT was statistically independent of additional other major non-ophthalmological disease such as arterial hypertension, chronic obstructive pulmonary disease, asthma, chronic kidney disease, hepatic disorders associated with an increase in the serum concentration of transaminases, hearing loss, depression and anxiety. Our investigations extends the findings obtained in previous studies on missing relationships between CCT and non-ophthalmological parameters.
Interestingly, CCT was significantly higher in the left eyes than in the right eyes (543.1 ± 33.7 µm versus 541.7 ± 33.7) µm in our study. The reason for this inter-ocular difference has remained unclear. If it was an artefact, its effect on the study results may have been small, since the data of a randomly chosen eye per individual was taken for further statistical analysis.
When the results of our study are discussed, its limitations should be taken into account. First, the value of an epidemiological investigation is profoundly connected with the rate of participation and how much the study area and study population are representative for the region and population it aimed at. In the Ural Eye and Medical Study, the participation rate was $80.5\%$ out of the eligible population of 7328 individuals, and the present study included $98.2\%$ of these participants. There was however a significant difference in age and sex between the participants and non-participants. Second, the study areas were typical for Southern Russia with respect to its demography, geography and climate. The fraction of Russians in our study population and in the study region was lower than in North-Western Russia and Central Russia. To address that issue, we included the parameter of ethnic background into the multivariable analysis and found a significant difference in CCT between Russians and non-Russians. In a study including only *Russians this* finding would not have been detected. Third, we used Scheimflug imaging for the CCT measurements, while other investigations applied optical low-coherence reflectometer pachymetry or sonographic pachymetry for the determination of CCT. Studies showed that the CCT data obtained by Scheimflug imaging could be compared with those measured by optical low-coherence reflectometer pachymetry [37, 38]. Even if there was a systemic difference in the CCT measurements between studies due to differences in the techniques applied, such a difference might not have affected the associations between CCT and other ocular and general parameters as examined in the present study. Strengths of the Ural Eye and Medical Study were the relatively large study population and the relatively high number of ocular and systemic disorders and parameters assessed and included into the statistical analysis.
In conclusion, in a typical, ethnically mixed, population from Russia with an age of 40+ years, mean CCT (541.7 ± 33.7 µm) was associated with parameters such as younger age, male sex, Russian ethnicity, and higher educational level. These associations may be taken into account when the dependence of IOP readings on CCT are considered. Glaucoma prevalence was unrelated to CCT.
## What was known before
Central corneal thickness (CCT) is a clinically important parameter in the diagnosis of glaucoma, since the measurement of intraocular pressure markedly depends on CCT. It has additionally been discussed that a thin cornea may be a structural risk factor for an increased susceptibility for glaucomatous optic nerve damage at a given IOP.
## What this study adds
In this ethnically mixed population from Russia with an age of 40+ years, mean CCT (541.7 ± 33.7 µm) was associated with parameters such as younger age, male sex, Russian ethnicity, and higher educational level. These associations may be taken into account when the dependence of IOP readings on CCT are considered. Glaucoma prevalence was unrelated to CCT.
## References
1. Goldmann H, Schmidt T. **Über Applanationstonometrie**. *Ophthalmologica* (1957.0) **134** 221-42. DOI: 10.1159/000303213
2. Ehlers N, Bramsen T, Sperling S. **Applanation tonometry and central corneal thickness**. *Acta Ophthalmol* (1975.0) **53** 34-43. DOI: 10.1111/j.1755-3768.1975.tb01135.x
3. Brandt JD, Beiser JA, Kass MA, Gordon MO. **Central corneal thickness in the Ocular Hypertension Treatment Study (OHTS)**. *Ophthalmology* (2001.0) **108** 1779-88. DOI: 10.1016/S0161-6420(01)00760-6
4. Herndon LW, Weizer JS, Stinnett SS. **Central corneal thickness as a risk factor for advanced glaucoma damage**. *Arch Ophthalmol* (2004.0) **122** 17-21. DOI: 10.1001/archopht.122.1.17
5. Copt RP, Thomas R, Mermoud A. **Corneal thickness in ocular hypertension, primary open-angle glaucoma, and normal tension glaucoma**. *Arch Ophthalmol* (1999.0) **117** 14-16. DOI: 10.1001/archopht.117.1.14
6. Leske MC, Heijl A, Hyman L, Bengtsson B, Dong L, Yang Z. **Predictors of long-term progression in the Early Manifest Glaucoma Trial**. *Ophthalmology* (2007.0) **114** 1965-72. DOI: 10.1016/j.ophtha.2007.03.016
7. Gordon MO. **The Ocular Hypertension Treatment Study: baseline factors that predict the onset of primary open-angle glaucoma**. *Arch Ophthalmol* (2002.0) **120** 714-20. DOI: 10.1001/archopht.120.6.714
8. Wolfs RCW, Klaver CCW, Vingerling JR, GROBBEE DE, HOFMAN A, de JONG PTVM. **Distribution of central corneal thickness and its association with intraocular pressure: The Rotterdam Study**. *Am J Ophthalmol* (1997.0) **123** 767-72. DOI: 10.1016/S0002-9394(14)71125-0
9. Foster P, Baasanhu J, Alsbirk PH, Uranchimeg D, Johnson GJ. **Central corneal thickness and intraocular pressure in a Mongolian population**. *Ophthalmology* (1998.0) **105** 969-73. DOI: 10.1016/S0161-6420(98)96021-3
10. Aghaian E, Choe JE, Lin S, Stamper RL. **Central corneal thickness of Caucasians, Chinese, Hispanics, Filipinos, African Americans, and Japanese in a glaucoma clinic**. *Ophthalmology* (2004.0) **111** 2211-9. DOI: 10.1016/j.ophtha.2004.06.013
11. Fam HB, How AC, Baskaran M, Lim KL, Chan YH, Aung T. **Central corneal thickness and its relationship to myopia in Chinese adults**. *Br J Ophthalmol* (2006.0) **90** 1451-3. DOI: 10.1136/bjo.2006.101170
12. Landers JA, Billing KJ, Mills RA, Henderson TR, Craig JE. **Central corneal thickness of indigenous Australians within Central Australia**. *Am J Ophthalmol* (2007.0) **143** 360-2. DOI: 10.1016/j.ajo.2006.09.047
13. Tomidokoro A, Araie M, Iwase A. **Corneal thickness and relating factors in a population-based study in Japan: the Tajimi study**. *Am J Ophthalmol* (2007.0) **144** 152-4. DOI: 10.1016/j.ajo.2007.02.031
14. Zhang H, Xu L, Chen C, Jonas JB. **Central corneal thickness in adult Chinese. Association with ocular and general parameters. The Beijing Eye Study**. *Graef Arch Clin Exp Ophthalmol* (2008.0) **246** 587-92. DOI: 10.1007/s00417-007-0760-9
15. Su DH, Wong TY, Foster PJ, Tay WT, Saw SM, Aung T. **Central corneal thickness and its associations with ocular and systemic factors: the Singapore Malay Eye Study**. *Am J Ophthalmol* (2009.0) **147** 709-16. DOI: 10.1016/j.ajo.2008.10.013
16. 16.https://www.google.com/publicdata/explore?ds=z5567oe244g0ot_&met_y=population&idim=city_proper%3A036580%3A035310%3A035240&hl=en&dl=en. Assessed 17.8.2021.
17. Bikbov MM, Gilmanshin TR, Kazakbaeva GM, Zainullin RM, Rakhimova EM, Rusakova IA. **Prevalence of myopic maculopathy among adults in a Russian population**. *JAMA Netw Open* (2020.0) **3** e200567. DOI: 10.1001/jamanetworkopen.2020.0567
18. Bikbov MM, Zainullin RM, Gilmanshin TR, Kazakbaeva GM, Yakupova DF, Nuriev IF. **Prevalence and associated factors of pseudoexfoliation in a Russian population. The Ural Eye and Medical Study**. *Am J Ophthalmol* (2020.0) **210** 158-66. DOI: 10.1016/j.ajo.2019.10.003
19. **The Age-Related Eye Disease Study (AREDS) system for classifying cataracts from photographs: AREDS report no. 4**. *Am J Ophthalmol* (2001.0) **131** 167-75. DOI: 10.1016/S0002-9394(00)00732-7
20. Ferris FL, Wilkinson CP, Bird A, Chakravarthy U, Chew E, Csaky K. **Clinical classification of age-related macular degeneration**. *Ophthalmology* (2013.0) **120** 844-51. DOI: 10.1016/j.ophtha.2012.10.036
21. Foster PJ, Buhrmann R, Quigley HA, Johnson GJ. **The definition and classification of glaucoma in prevalence surveys**. *Br J Ophthalmol* (2002.0) **86** 238-42. DOI: 10.1136/bjo.86.2.238
22. Nangia V, Jonas JB, Sinha A, Matin A, Kulkarni M. **Central corneal thickness and its association with ocular and general parameters in Indians: The Central India Eye and Medical Study**. *Ophthalmology* (2010.0) **117** 705-10. DOI: 10.1016/j.ophtha.2009.09.003
23. Vijaya L, George R, Baskaran M, Arvind H, Raju P, Ramesh SV. **Prevalence of primary open-angle glaucoma in an urban south Indian population and comparison with a rural population. The Chennai Glaucoma Study**. *Ophthalmology* (2008.0) **115** 648-54. DOI: 10.1016/j.ophtha.2007.04.062
24. Nemesure B, Wu SY, Hennis A, Leske C. **Corneal thickness and intraocular pressure in the Barbados eye studies**. *Arch Ophthalmol* (2003.0) **121** 240-4. DOI: 10.1001/archopht.121.2.240
25. Jonas JB, Ohno-Matsui K, Jiang WJ, Panda-Jonas S. **Bruch membrane and the mechanism of myopization. A new theory**. *Retina* (2017.0) **37** 1428-40. DOI: 10.1097/IAE.0000000000001464
26. Bikbov MM, Kazakbaeva GM, Zainullin RM, Salavatova VF, Gilmanshin TR, Yakupova DF. **Intraocular pressure and its associations in a Russian population: The Ural Eye and Medical Study**. *Am J Ophthalmol* (2019.0) **204** 130-9. DOI: 10.1016/j.ajo.2019.02.030
27. Jonas JB, Nangia V, Matin A, Sinha A, Kulkarni M, Bhojwani K. **Intraocular pressure and associated factors. The Central India Eye and Medical Study**. *J Glaucoma* (2011.0) **20** 405-9. DOI: 10.1097/IJG.0b013e3181f7af9b
28. Xu L, Zhang H, Wang YX, Jonas JB. **Central corneal thickness and glaucoma in adult Chinese. The Beijing Eye Study**. *J Glaucoma* (2008.0) **17** 647-53. DOI: 10.1097/IJG.0b013e3181666582
29. Jonas JB, Stroux A, Velten I, Juenemann A, Martus P, Budde WM. **Central corneal thickness correlated with glaucoma damage and rate of progression**. *Invest Ophthalmol Vis Sci* (2005.0) **46** 1269-74. DOI: 10.1167/iovs.04-0265
30. Jonas JB, Stroux A, Oberacher-Velten IM, Kitnarong N, Juenemann A. **Central corneal thickness and development of glaucomatous optic disc hemorrhages**. *Am J Ophthalmol* (2005.0) **140** 1139-41. DOI: 10.1016/j.ajo.2005.06.056
31. Wang D, Huang W, Li Y, Zheng Y, Foster PJ, Congdon N. **Intraocular pressure, central corneal thickness, and glaucoma in chinese adults: the liwan eye study**. *Am J Ophthalmol* (2011.0) **152** 454-62. DOI: 10.1016/j.ajo.2011.03.005
32. Jonas JB, Holbach L. **Central corneal thickness and thickness of the lamina cribrosa in human eyes**. *Invest Ophthalmol Vis Sci* (2005.0) **46** 1275-9. DOI: 10.1167/iovs.04-0851
33. Su DH, Wong TY, Wong WL, Saw SM, Tan DT, Shen SY. **Diabetes, hyperglycemia, and central corneal thickness: the Singapore Malay Eye Study**. *Ophthalmology* (2008.0) **115** 964-68. DOI: 10.1016/j.ophtha.2007.08.021
34. Soleimanizad R, Nowroozzadeh MH, Ziaei H, Pakravan M, Yaseri M, Katibeh M. **The association of central corneal thickness with ocular and general parameters in a community setting: The Yazd Eye Study**. *J Ophthalmic Vis Res* (2017.0) **12** 141-50. PMID: 28540004
35. Hwang YH, Kim HK, Sohn YH. **Central corneal thickness in a Korean population: the Namil Study**. *Invest Ophthalmol Vis Sci* (2012.0) **53** 6851-5. DOI: 10.1167/iovs.12-10173
36. Sudhir RR, Raman R, Sharma T. **Changes in the corneal endothelial cell density and morphology in patients with type 2 diabetes mellitus: a population-based study, Sankara Nethralaya Diabetic Retinopathy and Molecular Genetics Study (SN-DREAMS, Report 23)**. *Cornea* (2012.0) **31** 1119-22. DOI: 10.1097/ICO.0b013e31823f8e00
37. Barkana Y, Gerber Y, Elbaz U, Schwartz S, Ken-Dror G, Avni I. **Central corneal thickness measurement with the Pentacam Scheimpflug system, optical low-coherence reflectometry pachymeter, and ultrasound pachymetry**. *J Cataract Refract Surg* (2005.0) **31** 1729-35. DOI: 10.1016/j.jcrs.2005.03.058
38. Beutelspacher SC, Serbecic N, Scheuerle AF. **Assessment of central corneal thickness using OCT, ultrasound, optical low coherence reflectometry and Scheimpflug pachymetry**. *Eur J Ophthalmol* (2011.0) **21** 132-7. DOI: 10.5301/EJO.2010.1093
|
---
title: Decreased Paneth cell α-defensins promote fibrosis in a choline-deficient L-amino
acid-defined high-fat diet-induced mouse model of nonalcoholic steatohepatitis via
disrupting intestinal microbiota
authors:
- Shunta Nakamura
- Kiminori Nakamura
- Yuki Yokoi
- Yu Shimizu
- Shuya Ohira
- Mizu Hagiwara
- Zihao Song
- Li Gan
- Tomoyasu Aizawa
- Daigo Hashimoto
- Takanori Teshima
- Andre J. Ouellette
- Tokiyoshi Ayabe
journal: Scientific Reports
year: 2023
pmcid: PMC9998432
doi: 10.1038/s41598-023-30997-y
license: CC BY 4.0
---
# Decreased Paneth cell α-defensins promote fibrosis in a choline-deficient L-amino acid-defined high-fat diet-induced mouse model of nonalcoholic steatohepatitis via disrupting intestinal microbiota
## Abstract
Nonalcoholic steatohepatitis (NASH) is a chronic liver disease characterized by fibrosis that develops from fatty liver. Disruption of intestinal microbiota homeostasis, dysbiosis, is associated with fibrosis development in NASH. An antimicrobial peptide α-defensin secreted by Paneth cells in the small intestine is known to regulate composition of the intestinal microbiota. However, involvement of α-defensin in NASH remains unknown. Here, we show that in diet-induced NASH model mice, decrease of fecal α-defensin along with dysbiosis occurs before NASH onset. When α-defensin levels in the intestinal lumen are restored by intravenous administration of R-Spondin1 to induce Paneth cell regeneration or by oral administration of α-defensins, liver fibrosis is ameliorated with dissolving dysbiosis. Furthermore, R-Spondin1 and α-defensin improved liver pathologies together with different features in the intestinal microbiota. These results indicate that decreased α-defensin secretion induces liver fibrosis through dysbiosis, further suggesting Paneth cell α-defensin as a potential therapeutic target for NASH.
## Introduction
Nonalcoholic steatohepatitis (NASH) is a progressive chronic liver disease characterized by fat accumulation in the liver, hepatocyte damage, inflammation, and subsequent fibrosis, which progresses to cirrhosis and liver cancer1. Suppressing liver fibrosis is a major challenge in the treatment of NASH, because fibrosis increases the risk of liver-related mortality2. The fibrosis progression rate among nonalcoholic fatty liver disease (NAFLD) patients is about $40\%$ and varies among individuals3. Thus, a multiple hit hypothesis has been considered in which NASH development proceeds through the complex crosstalk of varied factors not limited to the liver and that the process of NASH onset is diverse4. The progression of NASH pathology is associated with overnutrition5, yet the specific factors that lead to progression from nonalcoholic fatty liver to NASH have not been clarified.
The liver plays a central role in metabolism and directly communicates with the intestine to form the “gut-liver axis”6. The human intestine harbors forty trillion bacteria affecting a variety of host functions including energy acquisition and consumption, in which the intestinal microbiota form a symbiotic relationship with the host7. On the other hand, dysbiosis, disruption of intestinal microbiota homeostasis, is associated with many metabolic dysfunctions such as fatty liver and obesity8,9. Moreover, the composition and diversity of the intestinal microbiota are influenced by diets, and metabolites such as short-chain fatty acids (SCFA) regulate liver fatty acid syntheses of fatty acids and bile acids in the liver10. In addition, dysbiosis induces hepatic inflammation in NAFLD patients, in which alterations in bacterial metabolites and an influx of pathogen-associated molecular patterns promote chronic inflammation in the liver11.
The intestinal epithelial cells that separate the intestinal lumen from the underlying tissue are constantly exposed to food and microbes, absorbing nutrients needed for host survival and sending them to the liver while preventing invasion by pathogens12. Paneth cells, an epithelial lineage located at the base of small intestinal crypts, secrete granules rich in α-defensins, human defensin (HD) 5 in humans and cryptdins (Crps) in mice, into the intestinal lumen, and contribute to innate enteric immunity mainly by the potent microbicidal activities of α-defensins against pathogens13–17. Furthermore, α-defensins regulate composition of the intestinal microbiota by the selective bactericidal activities, killing pathogenic bacteria while being symbiotic with commensal bacteria18. The absence of activated Crps in matrix metalloproteinase 7-deficient mice increases the prevalence of Firmicutes and decreases Bacteroidetes in the small intestine, while the production of HD5 in DEFA5 transgenic mice increases Bacteroidetes and decreases Firmicutes, indicating that α-defensins affect the intestinal microbiota composition19. Previous reports showed that Paneth cell α-defensins play a role in regulating the intestinal microbiota20 and that impaired Paneth cell granule secretion is associated with diseases such as Crohn’s disease, graft-versus-host disease (GVHD), and depression21–24.
In this study, we hypothesized that abnormalities in Paneth cells and their α-defensins induce dysbiosis, resulting in damages of the liver, and lead to liver fibrosis. Here we show, in choline-deficient, L-amino acid-defined, high-fat diet (CDAHFD)-induced NASH model mice, the quantity of α-defensins secreted from Paneth cells decreased significantly before the onset of NASH, resulting in dysbiosis with decreased microbiota diversity followed by fibrosis leading to NASH. Furthermore, restoring luminal levels of α-defensins by inducing Paneth cell regeneration by treatment with R-Spondin1 (R-Spo1)25–27 or by oral administration of a mouse α-defensin, Crp4, both suppress dysbiosis and ameliorate liver fibrosis. Our results show that dysbiosis associated with decreased Paneth cell α-defensin secretion contributes to disease onset and progression in the CDAHFD model of NASH, further providing insights into Paneth cell α-defensins in “gut-liver axis” in NASH fibrosis.
## Paneth cell α-defensin secretion decreases before onset of NASH in the CDAHFD group
To investigate whether Paneth cell α-defensins are involved in the onset of NASH, we examined mice fed CDAHFD. First, histological analyses of the liver were conducted on mice fed standard diet (SD) or CDAHFD, fat accumulation and lobular inflammation were observed in the CDAHFD group at 1 week (wk), also both the area of steatosis and the number of inflammatory foci increased with wks. Hepatocellular ballooning increased in the CDAHFD group at 3 wk and continued to increase until 12 wk (Fig. 1a,b). Sirius red staining was performed to evaluate liver fibrosis, showing that Sirius red-positive fibrotic areas first appeared at 3 wk and continued to increase until 12 wk in the CDAHFD group (Fig. 1a,b). The number of inflammatory foci showed a strong positive correlation with fibrosis (Supplementary Fig. 1a). We further examined specific marker gene expressions of ER stress, oxidative stress, and impaired autophagy, known to be involved in NASH pathology4. Among ER stress markers, Pdia3 mRNA expression decreased at 1 wk and increased at 3 wk, Chop increased at 6 and 12 wk, and Perk increased at 12 wk in the CDAHFD group compared to the SD group (Fig. 1c). Expression of Nox2 mRNA, a key source of redox radicals, increased, and antioxidant enzyme Prdx6 mRNA decreased after 1 wk in the CDAHFD group. Among autophagy activating protein genes, Atg3 mRNA expression declined after 1 wk followed by decreasing expression of both Atg12 and Lc3b mRNAs from 3 wk in the CDAHFD group. In addition, F$\frac{4}{80}$ and Cd11b, markers of resident and invasive macrophages, respectively, and dendritic cell marker Cd11c mRNA levels continued to increase after 1 wk. Fibrosis related growth factor Tgfb1 increased after 3 wk in the CDAHFD group. mRNA levels for Trailr2 increased from 1 wk, and Bax increased at 12 wk (Fig. 1c). The number of terminal deoxynucleotidyl transferase–mediated dUTP nick end labeling (TUNEL)-positive liver cells in the CDAHFD group increased at 3 wk and continued to increase with progression of fibrosis up to 12 wk (Fig. 1d). The number of TUNEL-positive cells showed strong positive correlation with both the number of inflammatory foci and the area of fibrosis (Supplementary Fig. 1b). These results indicated that inflammation and apoptosis in the liver contributed to the onset and progression of fibrosis in the NASH model and that ER stress, oxidative stress, and impairment of autophagy are involved in these pathways. Figure 1CDAHFD group develop liver injury and fibrosis associated with dysregulated expression of ER stress, oxidative stress, and autophagy related genes. ( a) Representative images of hematoxylin and eosin (H&E)- and Sirius red-stained liver sections. Arrowheads indicate Sirius red-positive area. Scale bars: 50 μm. ( b) Quantification of steatosis area, lobular inflammation, hepatocellular ballooning, and Sirius red-positive area. The degree of lobular inflammation was evaluated by counting the number of inflammatory foci. ( c) Hepatic mRNA expression of ER stress, oxidative stress, autophagy, inflammation, and apoptosis related genes was analyzed by real-time PCR. ( d) Representative images of TUNEL staining of liver sections (left). Arrowheads indicate TUNEL-positive cells. Scale bars: 10 μm. Quantification of TUNEL-positive cells (right). Data are shown as mean ± SEM for $$n = 3$$–6 per group. * $P \leq 0.05$, **$P \leq 0.01$ and ***$P \leq 0.001$, by unpaired two-tailed Student’s t test.
Fecal levels of Crp1 were measured to evaluate the quantity of α-defensins secreted by Paneth cells. Crp1 is the most abundant α-defensins, in the mouse small intestine28. Fecal Crp1 levels in the CDAHFD group significantly decreased compared to the SD group before the onset of liver fibrosis, and continued for 12 wk (Fig. 2a). Furthermore, fecal Crp1 levels showed negative correlation with steatosis, inflammation, apoptosis, and fibrosis (Supplementary Fig. 2). To investigate the cause of decreased Crp1 secretion, we analyzed Crp1 mRNA expression, Paneth cell number, and whole-mount fluorescence immunostaining of Crp1 on the ileal tissue samples. No differences were observed in both Crp1 mRNA expression and Paneth cell number between the SD and CDAHFD groups (Fig. 2b,c, and Supplementary Fig. 3a,b). In contrast, fluorescent intensities of Crp1 in the CDAHFD group were diminished compared to SD after 3 wk (Fig. 2b,d).Figure 2Secretion and protein expression of Paneth cell α-defensin decrease in the CDAHFD group. ( a) Quantification of fecal levels of Crp1. Data are shown as mean ± SEM for $$n = 5$$–6 mice/group. ( b) Representative images of immunofluorescence staining of Crp1 (green) in small intestine from SD and CDAHFD group. DAPI (blue) stains the nucleus. Scale bars: 20 μm. ( c) Quantification of the number of Paneth cells and (d) Crp1 fluorescence intensity on Paneth cells shown in (b). Data are shown as mean ± SEM for $$n = 9$$–18 fields/group and quantified based on 3 fields per mouse. Each group contains 3–6 mice. * $P \leq 0.05$, **$P \leq 0.01$ and ***$P \leq 0.001$, by unpaired two-tailed Student’s t test.
## Paneth cell granule secretion is impaired in the CDAHFD group
Because α-defensins are abundant in Paneth cell granules normally, we examined the morphology of Paneth cell granules to possibly explain the decreased levels of α-defensins in Paneth cells of the CDAHFD group. The diameter of Paneth cell granules decreased after 1 wk of the CDAHFD group, and the number of Paneth cell granules also decreased after 3 wk (Fig. 3a,b). Transmission electron microscopy (TEM) images showed the abnormal morphologies of Paneth cell granules in the CDAHFD group (Supplementary Fig. 4). To further determine the molecular basis for the functional alterations in Paneth cells, RNA sequencing (RNA-seq) analysis was performed on isolated Paneth cells from both groups. One hundred and forty-five genes were upregulated in the CDAHFD group compared to the SD group, while 2,202 genes were downregulated (Fig. 3c). These differentially regulated genes were compared with gene products by referring to GO terms: Secretion (GO: 0046903) and trans-Golgi network (GO: 0005802), known to have key roles in the granule secretory pathway. Twelve upregulated genes in CDAHFD Paneth cells include ion transporter Car4, oxidative stress gene Chac1, glycosylation-related Tmem165, and secretion-related Stxbp1. Ninety-seven down regulated genes include ion transporters Cracr2a, Kcnq1, Cftr, and Kcnn4, vesicle transporters Optn, Trappc9, Vamp2, Dop1a, Rab26, Lrba, Arfrp1, and Gga3, autophagy-related Klhl20, Nlrp6, Ap4m1, and Bsn, and oxidative stress-related protein Arntl1 (Fig. 3d and Supplementary Table 1). These results indicated that expression of genes related to the granule secretory pathway is dysregulated in Paneth cells of the CDAHFD group. Figure 3Paneth cell granule secretion is impaired in the CDAHFD group. ( a) Representative images of whole mount small intestine obtained with confocal microscopy. Scale bar: 5 μm. ( b) Granule diameter and granule number of each Paneth cell per field. Data are shown as mean ± SEM for $$n = 9$$–18 fields/group and quantified based on at least 3 fields per mouse. Each group contains 3–6 mice. ( c,d) Paneth cells isolated from small intestine of mice fed with SD and CDAHFD for 3 wk were analyzed by RNA-seq. Paneth cells are pooled from 6 mice/group and data show the average of two independent experiments. ( c) Scatterplot of global gene expression profiles of SD and CDAHFD group derived from RNA-seq analysis. ( d) Heatmap showing log2 fold change of differentially expressed genes (1.5-fold increase or decrease) associated with Paneth cell functions in the CDAHFD group compared with SD group. ( e) Enteroids derived from the small intestine of SD and CDAHFD groups were stimulated by 1 μM CCh for 10 min. Percent granule secretion was calculated as percent area granule secretion. Data are shown as mean ± SEM for $$n = 40$$ crypts/group. Each group contains 4 mice. * $P \leq 0.05$, **$P \leq 0.01$ and ***$P \leq 0.001$, by unpaired two-tailed Student’s t test.
Abnormalities of ER stress, autophagy, and oxidative stress related genes are known to result in morphological changes in Paneth cell granules21,29, suggesting impairment of granule secretory function induced by CDAHFD. In addition, Paneth cells of the CDAHFD group showed dysregulated expression of genes essential for granule secretion including vesicle transporters and ion channel transporters. Thus, to evaluate Paneth cell granule secretion of the CDAHFD group directly, granule secretory responses of Paneth cells were visualized in enteroids, a three-dimensional culture system of small intestinal epithelial cells17. When carbachol (CCh), which induces Paneth cell granule secretion, was added to enteroid cultures derived from small intestinal crypts of the CDAHFD group, the quantity of granules secreted by Paneth cells was significantly lower than that of the SD group enteroids (Fig. 3e). These results indicated that Paneth cell dysfunction impairs granule secretion, resulting in diminished release of α-defensins into the intestinal lumen.
## Dysbiosis occurs along with decreased Crp1 secretion in the CDAHFD group
To discern the effect of decreased Crp1 secretion on the intestinal microbiota, we conducted 16S rRNA gene sequencing on fecal DNA. Principal coordinate analysis (PCoA) showed that the bacterial community of the CDAHFD group formed a different cluster compared to the SD group at 1 wk and continued through 12 wk (Fig. 4a). α-*Diversity analysis* showed that observed operational taxonomic units (OTUs) and Shannon index were significantly decreased at 1 wk, and positive correlation was observed between fecal Crp1 levels and the α-diversity indexes, indicating that dysbiosis occurs before the onset of NASH in the CDAHFD group (Fig. 4b,c). At the family level, the relative abundances of Eggerthellaceae, Bacteroidaceae, Prevotellaceae, Streptococcaceae, Peptostreptococcaceae, and Ruminococcaceae were significantly increased in the CDAHFD group. In contrast, Muribaculaceae, unassigned family of unassigned Bacteroidia class, unassigned Clostridiales, Clostridiales VadinBB60 group, and Erysipelotrichaceae were significantly decreased relative to the SD group (Supplementary Fig. 5a). At the genus level, a total of 28 genera showed significantly different occupancy rates between the SD and CDAHFD group during disease progression, and among them, twenty genera showed a positive or negative correlation with fecal Crp1 (Fig. 4d). To determine whether the dysbiosis relates to disease progression, correlation analyses were conducted between the relative abundance of the intestinal microbiota and NASH pathology. The relative abundance of 10 genera which increased in the CDAHFD group, including Bacteroides, Alloprevotella, GCA-900066575, Lachnoclostridium, uncultured Lachnospiraceae, unassigned Peptostreptococcaceae, Harryflintia, Ruminiclostridium 5, Ruminiclostridium 9, and UCG-009 correlated positively with NASH pathology including steatosis, inflammation, apoptosis, and fibrosis. In contrast, the 6 genera that decreased in abundance, Muribaculum, unassigned Muribacululaceae, Muribaculaceae uncultured bacterium, NK4A136 group, unassigned genus of the unassigned Clostridiales family, and uncultured Clostridiales VadinBB60 group negatively correlated with NASH pathology (Fig. 4d and Supplementary Fig. 5b). Furthermore, fecal Crp1 levels showed negative correlation with NASH pathology (Supplementary Fig. 2). These results suggested that the decrease of Crp1 secretion is associated with NASH development and progression via dysbiosis. Figure 4CDAHFD group shows dysbiosis correlated with both the quantity of α-defensin in feces and NASH pathology. ( a) PCoA plot of intestinal microbiota based on weighted UniFrac distance of SD and CDAHFD groups. Significance was computed with PERMANOVA. ( b) Observed OTUs and Shannon index in SD and CDAHFD groups. ( c) *Correlation analysis* between fecal Crp1 levels and α-diversity indexes. ( d) 28 genera that were significantly increased or decreased in the CDAHFD group compared with the SD group. Red indicates genera significantly increased in the CDAHFD group, and blue indicates genera significantly decreased in the CDAHFD group. Heatmap showing Pearson correlation coefficients between relative abundance of significantly changed genera in the CDAHFD group and fecal Crp1 levels from Fig. 2a or NASH pathology from Fig. 1b. Correlation analysis between Crp1 levels and relative abundance of individual genera was conducted using the data of 1, 3, 6, and 12 wk. Correlation analysis between NASH pathology and relative abundance of individual genera was conducted using the data of 12 wk. ( e) Serum zonulin levels in SD and CDAHFD group. ( f) *Correlation analysis* between fecal Crp1 and serum zonulin levels. ( g) Quantification of bacteria in spleen by CFUs cultured from spleen of SD and CDAHFD group at 3, 6, and 12 w under anaerobic condition. Data are shown as mean ± SEM for $$n = 5$$–6 mice/group in (a,b,c,d). Data are shown as mean ± SEM for $$n = 3$$–6 mice/group in (e,f,g). * $P \leq 0.05$, **$P \leq 0.01$ and ***$P \leq 0.001$, by unpaired two-tailed Student’s t test (b,e,g) and Pearson correlation test (c,d,f).
Because increased intestinal permeability associated with dysbiosis allows for translocation of enteric microbiota or bacterial components to the liver, contributing to NASH development, we next measured serum zonulin, a marker of intestinal permeability30. At 1 wk, the SD and CDAHFD group had similar serum zonulin levels. However, at 3 wk and continuing for 12 wk, serum zonulin levels in the CDAHFD group were increased significantly compared to SD mouse sera (Fig. 4e). To further assess whether the intestinal permeability increased, we analyzed mRNA expression of tight junction protein, and determined that the mRNA expression of claudin-1 decreased significantly in the ileum of the CDAHFD group at 3 and 6 wk (Supplementary Fig. 6a,b). Furthermore, serum zonulin showed negative correlation with fecal Crp1 levels, showing that dysbiosis induced by decreased Crp1 relates to increased intestinal permeability (Fig. 4f). In testing for bacterial translocation, bacterial colonies were detected in spleens at 6 wk in the CDAHFD group (Fig. 4g). Thus, these findings suggested that the dysbiosis associated with decreased Crp1 secretion resulted sequentially in intestinal hyperpermeability as well as bacterial translocation.
## R-Spo1 administration restores luminal α-defensin secretion, suppressing dysbiosis and reducing NASH progression
To clarify the association between the early reduction of luminal α-defensins and dysbiosis along with NASH pathology, we tested whether restoration of luminal α-defensins prevents NASH development. Since Wnt activator R-Spo1 enhances α-defensin secretion by stimulating Paneth cell differentiation from stem cells27, CDAHFD-fed mice were injected intravenously with R-Spo1 for 3 wk. Histological analyses of the liver showed that the area of steatosis, the degree of lobular inflammation and hepatocellular ballooning, and the area of fibrosis were all decreased significantly in R-Spo1-treated CDAHFD mice (CDAHFD + R-Spo1 group) compared to CDAHFD-fed mice treated with PBS (CDAHFD + PBS group) (Fig. 5a,b). R-Spo1 treatment significantly reduced hepatic mRNA expression of Nox2, F$\frac{4}{80}$, Tgfb1, Trailr2, and Bax, tended to reduce Cd11c expression in the liver, and significantly decreased the number of TUNEL-positive cells (Fig. 5c,d). Fecal Crp1 levels were measured to determine whether R-Spo1 treatment restores luminal levels of secreted Crp1. Fecal Crp1 in the CDAHFD + R-Spo1 group started to increase at 1 wk, and significantly increased after 2 wk compared to the CDAHFD + PBS group (Fig. 5e). Numbers of Paneth cells in the CDAHFD + R-Spo1 group increased significantly at 3 wk along with stem cells, indicating that R-Spo1 treatment restored fecal Crp1 levels by elevating Paneth cell numbers (Fig. 5f).Figure 5R-Spo1 restores luminal α-defensin, suppresses intestinal dysbiosis and improves NASH progression. Six-week-old C57BL/6J mice were fed CDAHFD to induce NASH and intravenously injected with R-Spo1 at a dose of 600 μg or PBS three times a week for 3 wk. ( a) Representative images of H&E- and Sirius red-stained liver sections. Scale bars: 50 μm. ( b) Quantification of steatosis area, lobular inflammation, hepatocellular ballooning, and Sirius red-positive area. The degree of lobular inflammation was evaluated by counting the number of inflammatory foci. ( c) Hepatic mRNA expression of ER stress, oxidative stress, autophagy, inflammation, and apoptosis marker genes. ( d) Representative images of TUNEL staining of liver sections and quantification of TUNEL-positive cells. Scale bars: 10 μm. ( e) Fecal Crp1 levels in CDAHFD + PBS and CDAHFD + R-Spo1 group. ( f) Representative images of immunofluorescence staining of Crp1 (green) and Olfm4 (red) in small intestine from CDAHFD + PBS and CDAHFD + R-Spo1 mice at 3 wk (left). Arrowheads indicate Paneth cells and asterisks indicate stem cells. Scale bar: 20 μm. Quantification of the number of Paneth cells assessed by expression of Crp1 and the number of stem cells assessed by expression of Olfm4 per ileal unit area (right). Data are shown as mean ± SEM for $$n = 6$$ fields/group and quantified based on at least 2 fields per mouse. Each group contains 3 mice. ( g) Relative abundance of individual genera that were significantly recovered in CDAHFD + R-Spo1 group from CDAHFD group of 28 genera shown in Fig. 4d (upper panel). Correlation analysis between fecal Crp1 levels and relative abundance of individual genera (lower panel). ( h) *Correlation analysis* between relative abundance of individual genera and NASH pathology. ( i) *Correlation analysis* between fecal Crp1 levels and NASH pathology. Data are shown as mean ± SEM for $$n = 3$$ per group in (b,c,d,e,g). * $P \leq 0.05$, **$P \leq 0.01$ and ***$P \leq 0.001$, by unpaired two-tailed Student’s t test (b,c,d,e,f,g) and Pearson correlation test (g,h,i).
Next, we assessed whether enhanced fecal Crp1 levels by R-Spo1 treatment attenuates dysbiosis. Among the families of intestinal bacteria which increased in the CDAHFD group during disease progression (Supplementary Fig. 5a), the relative abundances of Prevotellaceae and Peptostreptococcaceae were reduced by R-Spo1 treatment (Supplementary Fig. 7a). In contrast, R-Spo1 treatment reduced the proportion of the intestinal microbiota that increased during disease progression including Muribaculaceae, unassigned Clostridiales, and unassigned family of unassigned Bacteroidia (Supplementary Fig. 7a). Fecal Crp1 levels were positively correlated with the relative abundance of Muribaculacea and unassigned Clostridiales (Supplementary Fig. 7b). At the genus level, among the intestinal microbiota that correlated with both fecal Crp1 levels and NASH pathology (Fig. 4d), the proportions of Muribaculum, unassigned Muribaculaceae, Muribaculaceae uncultured bacterium, and unassigned Clostridiales were restored by R-Spo1 treatment, showing positive correlation with Crp1 levels. The outgrowth of Harryflintia was inhibited, showing negative correlation with Crp1 levels (Fig. 5g). Furthermore, the number of inflammatory foci was negatively correlated with Muribaculaceae uncultured bacterium, positively correlated with Harryflintia, and tended to be negatively correlated with Muribaculum, Muribaculaceae uncultured bacterium, and unassigned genus of the unassigned Clostridiales family. Both Muribaculum and unassigned Muribaculaceae showed negative correlation with the steatosis area, the number of TUNEL-positive cells, and the area of fibrosis. In addition, unassigned genus of the unassigned Clostridiales family was negatively correlated with the number of TUNEL-positive cells (Fig. 5h). Fecal Crp1 levels were negatively correlated with the number of inflammatory foci, the number of TUNEL-positive cells, and the fibrosis area, indicating restoration of Crp1 secretion ameliorates NASH pathology (Fig. 5i). Taken together, enhancing luminal α-defensin secretion from Paneth cells by R-Spo1 treatment reduced liver fibrosis via ameliorating inflammation and apoptosis in the liver along with preventing dysbiosis.
## Oral administration of Crp4 suppresses dysbiosis and ameliorates liver fibrosis
Finally, because granules of Paneth cells contain microbicidal components in addition to abundant α-defensins29, we investigated whether oral administration of an exogenous α-defensin prevents dysbiosis and subsequent NASH development. Oral administration of Crp4, which is known as the most potent in vitro bactericidal activities among Crps31 has been reported to improve homeostasis of the intestinal microbiota in mouse GVHD model and in chronic social defeat stress model24,27. Therefore, Crp4 was administered orally to CDAHFD-fed mice for 6 wk. Because C57BL/6 mouse strain lacks the *Crp4* gene32, dosing Crp4 introduces an exogenous α-defensin, and we confirmed that orally administered Crp4 reached the intestinal lumen by measuring levels of fecal Crp4 (Supplementary Fig. 8). Histological analysis of liver revealed that oral administration of Crp4 decreased the degree of lobular inflammation and hepatocellular ballooning and the area of fibrosis significantly, though, Crp4 administration did not affect the area of hepatic steatosis (Fig. 6a,b). Analysis of liver gene expression revealed that Crp4 administration increased Atg12 mRNA levels, decreased Cd11b and Tgfb1 mRNAs, and tended to decrease levels of Nox2 and Cd11c mRNAs. Genes that decreased in expression by R-Spo1 treatment including F$\frac{4}{80}$, Trailr2, and Bax were unaffected by oral Crp4 administration (Fig. 6c). TUNEL-positive cells in the liver decreased significantly in the Crp4-treated group compared to the untreated group (Fig. 6d). These findings collectively support the view that Crp4 prevented fibrosis by activating autophagy, ameliorating both inflammation and apoptosis in the liver, suggesting a different mechanism from R-Spo1 treatment. Figure 6Oral administration of Crp4 increases autophagy related gene expression, suppresses dysbiosis, and ameliorates liver fibrosis. Six-week-old C57BL/6 J mice were fed CDAHFD to induce NASH and orally administered with Crp4 at a dose of 110 µg or saline twice daily for 6 wk. ( a) Representative images of H&E- and Sirius red-stained liver sections. Scale bars: 50 µm. ( b) Quantification of steatosis area, lobular inflammation, hepatocellular ballooning, and Sirius red-positive area. The degree of lobular inflammation was evaluated by counting the number of inflammatory foci. ( c) Hepatic mRNA expression of ER stress, oxidative stress, autophagy, inflammation, and apoptosis marker genes. ( d) Representative images of TUNEL staining of liver sections and quantification of TUNEL-positive cells. Scale bars: 10 µm. ( e) Relative abundance of individual genera that were significantly recovered in CDAHFD + Crp4 group from CDAHFD group of 28 genera shown in Fig. 4d. ( f) *Correlation analysis* between relative abundance of individual genera and NASH pathology. Data are shown as mean ± SEM for $$n = 6$$–8 per group. * $P \leq 0.05$, **$P \leq 0.01$ and ***$P \leq 0.001$, by unpaired two-tailed Student’s t test (b,c,d,e) and Pearson correlation test (f).
To assess whether oral Crp4 ameliorated liver fibrosis by suppressing dysbiosis, the intestinal microbiota composition was analyzed. At the family level, among the intestinal bacteria that increased in the CDAHFD group during disease progression, the proportion of Eggerthellaceae, Streptococcaceae, and Ruminococcaceae decreased with Crp4 administration (Supplementary Fig. 9). At the genus level, the outgrowth of Harryflintia was suppressed by Crp4 administration comparably to R-Spo1 treatment. On the other hand, the increase of uncultured Lachnospiraceae and Ruminoclostridium 9, which were not reversed by R-Spo1 treatment reduced by Crp4 administration (Fig. 6e). Moreover, both Harryflintia and Ruminiclostridium 9 were positively correlated with the number of inflammatory foci and TUNEL-positive cells, showing that Crp4 administration ameliorated inflammation and apoptosis accompanied by fibrosis in the liver via preventing dysbiosis (Fig. 6f). Taken together, both R-Spo1 and Crp4 prevented liver fibrosis with amelioration of liver inflammation and apoptosis, suppressing dysbiosis, though, the gene expression in the liver and the intestinal microbiota composition, which improved by Crp4 administration, were different from those by R-Spo1 treatment.
## Discussion
Fibrosis is the major determinant of mortality in patients with NASH, though, development mechanisms of the fibrosis are not fully understood yet2. This study focused on the “gut-liver axis” that leads to liver fibrosis and analyzed the relationship between Paneth cell α-defensins and NASH pathology via dysbiosis. The liver inflammation and apoptosis are the main factors contributing to pathology of NASH patients4. In this study, we confirmed that inflammation and apoptosis are key drivers of liver fibrosis in the CDAHFD model consistent with previous reports33,34. In NASH patients, ER stress occurs in hepatocytes with steatosis, and mitochondrial dysfunction accompanied by oxidative stress and suppression of autophagy induce and aggravate liver inflammation4. Among ER stress markers, Pdia3, whose expression increases in NASH patients35, increased at 3 wk, followed by elevation of Perk and Chop. In addition, Nox2 increased, whereas Prdx6 decreased, and autophagy-related genes including Atg3, Atg12, and Lc3b decreased in the CDAHFD group, suggesting excessive diet-induced oxidative stress and impairment of autophagy. Also, Trailr2 and Bax, which are known to elevate due to toxic lipid-induced ER stress36, increased in the CDAHFD group. Our results suggested that inflammation and apoptosis in the liver accompanied by ER stress, oxidative stress, and impairment of autophagy contribute to liver pathology in the CDAHFD group, similar to NASH patients.
We showed that fecal Crp1, an abundant mouse α-defensin, significantly decreases before NASH onset along with disease progression in the CDAHFD group, and dysbiosis sharing some similar features with NAFLD/NASH patients occurs. At the family level, Prevotellaceae, previously reported to increase in NAFLD patients37, increased in the CDAHFD group. At the genus level, Bacteroides, known to increase and positively correlate with fibrosis score in NASH patients38, was elevated in the CDAHFD group. Although further studies are needed to address mechanisms on the intestinal hyperpermeability in detail, the CDAHFD group also showed elevated serum zonulin levels and bacterial translocation similar to NASH patients. The composition of the intestinal microbiota is affected by varied factors including diet, factors in intestinal environment, and host-derived factors39, and the cause of dysbiosis in NASH has not been well understood. α-Defensins secreted from Paneth cells have been known to regulate the symbiotic intestinal microbiota composition by the selective bactericidal activities18,21,22,27. We showed that Paneth cell α-defensin secretion decreased before onset of NASH in the CDAHFD group, and the amount of α-defensin secreted into the intestinal lumen was enhanced by administration of R-Spo1, ameliorating steatosis, inflammation, apoptosis, and fibrosis in the liver. Recent studies showed that Muribaculaceae, which increased by R-Spo1 treatment, is required for inner mucus layer formation of the intestine and SCFA production, known to have beneficial effects on the host40,41. Furthermore, human Paneth cell α-defensin HD5 does not elicit bactericidal activities against Muribaculaceae, suggesting a promotion of colonization42. On the other hand, Harryflintia, which decreased by R-Spo1 treatment, increased in the animal model of hyperlipidemia and atherosclerosis, and positively correlated with the disease phenotype43, suggesting that *Harryflintia is* deleterious microbiota. Enhancing luminal α-defensin secretion by R-Spo1 treatment prevented liver fibrosis possibly by the selective microbicidal activities, increasing Muribaculaceae and decreasing Harryflintia, which contribute to ameliorating the disease progression.
R-Spo1 enhanced luminal secretion of α-defensin by regeneration of Paneth cells, though, it has been known that Paneth cells secrete other antimicrobial proteins including lysozyme and secretory phospholipase A214. In addition, R-Spo1 is reported to increase not only numbers of Paneth cells but also goblet cells27. Goblet cells secrete mucus, which harbors the intestinal microbiota and influences its composition44. Therefore, we further tested whether increasing luminal α-defensin prevents NASH by oral administration of α-defensin. We used Crp4 but not Crp1 because Crp4 has the strongest microbicidal activities among Crps31 and also because we can distinguish administered Crp4 from endogenous α-defensins. Oral Crp4 administration prevented NASH pathology including fibrosis, and inhibited the proportion of Harryflintia as did R-Spo1 treatment. In contrast, Crp4 but not R-Spo1 administration decreased Ruminiclostridium 9 and uncultured Lachnospiraceae, suggesting possible different action for the microbiota. In addition, it has been reported that Ruminiclostridium 9, which was positively correlated with the number of inflammatory foci in this study, also increases in a high-fat/high-fructose diet-induced dyslipidemia model45. Although further studies are needed to confirm a direct association of Crp1, these findings suggest that orally administered Crp4 prevented progression of liver fibrosis by ameliorating inflammation along with inhibiting bacterial outgrowth of Harryflintia and Ruminiclostridium 9.
The composition and function of the intestinal microbiota differ between the mucus side and the luminal side in the intestine, forming a unique ecosystem46. Although both R-Spo1 and Crp4 corrected dysbiosis by increasing the amount of α-defensin in the intestinal lumen, they affected the intestinal microbiota differently except for Harryflintia. Muribaculaceae affected by R-Spo1 is considered as the mucus resident bacteria47, while Ruminiclostridium 9, which was affected by Crp4 is known as the luminal bacteria48. These findings provide an insight that R-Spo1 and Crp4 modified bacteria localized in mucus and lumen, respectively. It is possible that intravenously administered R-Spo1 enhanced α-defensin secretion from de novo Paneth cells into mucus layer and predominantly contributed to modifying mucosa-associated bacteria, whereas oral administration of Crp4 mainly affected luminal resident bacteria of the intestine. Furthermore, R-Spo1 decreased apoptosis associated genes, Trailr2 and Bax, whereas Crp4 increased expression of Atg12, an autophagy related protein. Free fatty acids in hepatocytes have been reported to induce TRAILR2 expression by CHOP induction and apoptosis36. R-Spo1 prevented lipid accumulation in liver, suggesting that R-Spo1 suppresses liver fibrosis by inhibiting apoptosis induced by lipotoxicity. On the other hand, oral Crp4 administration activated autophagy with less effects on liver steatosis and gene expression of Trailr2 and Bax. Atg12 whose gene expression increased by Crp4 has a crucial role for autophagy49, and loss of autophagy in hepatocytes causes apoptosis, inflammation, and fibrosis in the liver50. These results suggest that Crp4 prevented liver fibrosis by a different mechanism of action from R-Spo1, which may relate to the differences in the improving intestinal microbiota composition.
Paneth cells regulate the composition of the intestinal microbiota by secreting granules rich in α-defensins in response to cholinergic agents, bacteria, and dietary factors15,17,51. Secretory responses of Paneth cells need a biphasic increase in cytosolic Ca2+ concentration, and Ca2+-activated potassium channel KCNN4 is the essential modulator of Ca2+ concentration during Paneth cell secretion52. In addition, mice having defective cystic fibrosis transmembrane conductance regulator (CFTR) show unusual accumulation of Paneth cell granules in the intestinal crypt lumen, suggesting that CFTR is important for dissolution of secreted granules53. Our RNA-seq analysis of Paneth cells revealed that the expression of both KCNN4 and CFTR decreased in Paneth cells of the CDAHFD group. RAS-related GTP-binding protein (Rab) and soluble NSF attachment protein receptor (SNARE) family proteins, which allow vesicle transport and fusion, have key roles in granule secretion. The expression of Rab26, which has been reported to localize on secretory granules and required to amylase release in parotid acinar cells54 and vesicle-associated membrane protein 2 (VAMP2), essential for glucagon-like peptide 1 secretion in intestinal L cell55 decreased in the CDAHFD group. On the other hand, overexpression of syntaxin-binding protein 1 (Stxbp1) inhibits the SNARE complex assembly and decreases insulin secretion56. Paneth cell in the CDAHFD group showed decreased expression of Rab26 and Vamp2, which are essential for granule secretion, and increased expression of Stxbp1, which is a negative regulator of secretory response. Furthermore, decreased granule secretion from Paneth cells in the CDAHFD group was directly revealed by visualization and quantification of granule secretory response in enteroids. Because mutation of autophagy related genes such as Atg16L1 leads to the abnormal morphology of Paneth cell granules, autophagy elicits important roles in granule formation and secretion of Paneth cells29. Optn, which decreased in the CDAHFD group Paneth cells, has been reported to promote autophagosome formation via recruitment of Atg12-5-16L1 complex57. Our findings that Paneth cells in the CDAHFD group showed reduction of Optn expression suggests that abrogation of autophagy leads to abnormal granule formation and decreased granule secretion. Thus, it is possible that decreased granule secretion in the CDAHFD group occurs through impairment of influx of ions such as Ca2+, vesicle transport, and autophagy process in Paneth cells.
The relationship between diet and Paneth cell function has been reported. The excess nutrients decreased secretion of α-defensin from Paneth cells and led to dysbiosis58. Similarly, abnormal granule morphology and decreased expression of HD5 were reported in Paneth cells of obese individuals59. Furthermore, leucine and butyric acid induce secretion of α-defensin from Paneth cells51. These studies suggested that certain dietary factors and intestinal microbiota metabolites are directly involved in the induction of α-defensin secretion. Although CDAHFD model mice have limitations such as not showing obesity and insulin resistance usually observed in patients with NASH34 so that further studies are needed to clarify the roles of α-defensin in NASH, our findings are the initial discoveries providing novel insights into the process of NASH fibrosis in the “gut-liver axis” and may further suggest a novel therapeutic approach for NASH targeting Paneth cell α-defensin via regulation of the intestinal microbiota.
## Mice
Three-week-old male C57BL/6J mice were purchased from CLEA Japan Inc. and acclimated for 3 weeks prior to be using in experiments. After acclimation, six-week-old mice were divided into two groups: SD group fed with standard diet (SD; A06071314, Research Diets Inc.) and CDAHFD group fed with choline-deficient, L-amino acid-defined, high-fat diet with $0.1\%$ methionine (CDAHFD; A06071302, Research Diets Inc.). The mice were housed on 12-h light/dark cycle and had free access to food and tap water. All animal experiments in this study were conducted after obtaining approval from the Institutional Animal Care and Use Committee of the National University Corporation at Hokkaido University in accordance with Hokkaido University Regulations of Animal Experimentation. This study was also carried out in compliance with the ARRIVE guidelines.
## Histological analysis
The left lobe of liver was rapidly excised and fixed in $10\%$ neutral buffered formalin. 4 μm sections of paraffin-embedded tissue were stained with H&E or Sirius red. The average number of inflammatory foci and hepatocellular ballooning were obtained from 5 randomly selected fields (228 × 303 μm2) per slide by using image analysis software, NIS-Elements D ver. 4.13 (Nikon). Hepatocellular ballooning was scored as follow: 0 (ballooned cells are absent), 1 (ballooned cells are present). The average percentage of hepatic steatosis area and fibrosis area were assessed by quantification of lipid accumulation area and Sirius red-positive area, respectively, in at least 20 fields (272 × 361 μm2) per slide using the BZ-II analyzer (KEYENCE). TUNEL assay was performed using the apoptosis in situ detection kit (Wako) following manufacturer’s instruction. TUNEL-positive cells were quantified by counting positive nuclei in 5 randomly selected fields (118 × 156 μm2) per slide by using NIS-Elements D.
## Reverse transcription and quantitative PCR
DNA-free RNA was obtained from ileal or liver tissue using the RNeasy Mini Kit (QIAGEN) with Dnase treatment. 0.5 μg total RNA was reverse transcribed using SuperScript VILO MasterMix (Life Technologies) by thermal cycled at 25 °C for 10 min, 42 °C for 60 min, and 85 °C for 5 min using Mastercycler EP (Eppendorf). Quantitative PCR was performed using Roche LightCycler 96 (Roche) with fluorescence-labeled locked nucleic acid (LNA) hydrolysis probes (Roche) from the Universal Probe Library (UPL) following the manufacturer’s protocol. Gene expression was normalized to hypoxanthine guanine phosphoribosyl transferase-1 (Hprt1). The primer sequences are listed in Supplementary Table 2.
## Quantification of fecal Crps
Fecal samples were dried and powdered by using a Multi-beads shocker (Yasui Kikai). 30 mg of fecal samples was vortex mixed with 300 μL PBS for 12 h at 4 °C. Fecal suspension was centrifuged at 20,400 g for 10 min at 4 °C, and levels of Crp1 or Crp4 in supernatants were measured by sandwich ELISA as previously described23,60.
## Whole-mount immunofluorescent staining and image analysis
Whole-mount immunofluorescent staining was performed using a modification of a previously reported method61. The ileal tissue was fixed in $4\%$ paraformaldehyde (Sigma) for 2 h at room temperature. Fixed tissue was permeabilized with $0.5\%$ Triton X-100 (Sigma) overnight at room temperature, and then blocked with $10\%$ goat serum (Sigma) and $0.5\%$ Triton X-100 overnight at 4 °C. For SD and CDAHFD group, antibody reaction was performed with FITC labeled mouse anti-Crp1 antibody (50 μg/mL, clone 77-R63, produced in our laboratory) and Alexa Fluor 647-labeled anti-mouse/human CD324 (E-cadherin) antibody (1:100, clone DECMA-1, BioLegend). For CDAHFD + PBS and CDAHFD + R-Spo1 group, the primary antibody reaction was performed with rabbit anti-Olfactomedin 4 (Olfm4) antibody (1:80, clone D6Y5A, Cell Signaling) for 1 day at 4 °C, and then the secondary antibody reaction was performed with Alexa Fluor 555 conjugated F(ab′)2-goat anti-rabbit IgG (dilution 1:500, Thermo Fisher Scientific) and FITC labeled mouse anti-Crp1 antibody overnight at 4 °C. After washing, nuclei were stained with DAPI (Thermo Fisher Scientific). Samples were immersed in the optical-clearing solution (RapiClear 1.52, Sunjin Lab).
For quantification of Crp1 fluorescence intensity and counting numbers of Paneth cells and stem cells, Z-stack images were obtained using a confocal microscope (A1, Nikon) equipped with CFI Apo LWD 20X WI λS (Nikon). The number of Paneth cells was quantified by counting Crp1 immunostaining positive cells on 3 fields (150 × 150 μm2) per tissue. The number of stem cells was quantified counting Olfm4 positive cells on 3 fields (150 × 150 μm2) per tissue. Crp1 fluorescence intensity per Paneth cell was measured by creating a region of interest using image analysis software, NIS-Elements AR ver. 5.11 (Nikon), on 3 fields (150 × 150 μm2) per tissue, and the mean intensity per field was calculated. For quantification of the number and diameter of Paneth cell granules, Z-stack images were obtained using A1 with CFI Apo TIRF 60X Oil (Nikon). The number and diameter of Paneth cell granules were measured on 3 fields (33 × 33 μm2, 2 Paneth cells/field) per tissue.
## TEM
5-mm-long segments of terminal ileum were immediately fixed in $2\%$ paraformaldehyde and $2\%$ glutaraldehyde at 4 °C for overnight. Next, samples were post-fixed with $2\%$ osmium tetroxide at 4 °C for 2 h. Dehydration was carried out, followed by embedding in Quetol-812 epoxy resin (Nisshin EM). After staining with $2\%$ uranyl acetate and lead stain solution (Sigma), the ultrathin sections were examined with a JEM-1400Plus transmission electron microscope (JEOL Ltd.) at an acceleration voltage of 100 kV.
## Paneth cell RNA-seq analysis
Crypts were isolated from the small intestine of mice fed with SD or CDAHFD for 3 wk as previously described17. Briefly, the ileum segments were shaken in cold HBSS containing 30 mM EDTA. After vigorous shaking for ~ 300 times in HBSS, the isolated crypts were resuspended in HBSS containing 300 U/mL collagenase (Sigma), 10 µM Y-27632 (Sigma), and 1 mM N-acetylcysteine (Sigma), and shaken at 180 rpm for 5 min at 37 °C on a horizontal shaker (TAITEC). Then, 50 μg/μL Dnase I (Roche) was added, and the sample was mixed by pipetting. Cells were pelleted at 500 g for 5 min at 4 °C and resuspended in washing buffer (DMEM/F12 containing 10 μM Y-27632 and 1 mM N-acetylcysteine), then passed through 40-μm cell strainer (BD Falcon). Paneth cells were stained with Zinpyr-1 (Santa Cruz) and Allophycocyanin (APC) anti-CD24 (clone M$\frac{1}{69}$,)Abcam) in washing buffer for 10 min at 37 °C. After Paneth cell labeling, the cells were sorted by flow cytometry using a cell sorter (JSAN, Bay Bioscience). Single cells were gated by forward scatter and side scatter. Cells were sorted directly into lysis buffer for RNA isolation (PureLink RNA Mini Kit, Invitrogen). To make a pooled sample, at least 10,000 Paneth cells were sorted from 2 separate mice, and sorting experiment was repeated three times. 2 pools (each pool contains Paneth cells from 6 separate mice) were prepared for each group. Total RNA was isolated using Invitrogen® PureLink RNA Purification System according to manufacturer’s instructions, and cDNA was synthesized using a SMART-Seq. Sequencing libraries were built with the TruSeq RNA Library Prep Kit (Illumina) and then submitted to Illumina NovaSeq 6000 for 100-bp PE reads sequencing. Fragments per kilobase of transcript per million mapped reads (FPKM) values were used, genes with FPKM values below 1 were not included in the analysis, and fold change ≥ 1.5 or ≤ 0.67 was considered differentially expressed.
## Visualization and quantification of Paneth cell granule secretion
Paneth cell granule secretion was evaluated using a modification of a previously reported method17. Isolated crypts from the distal 20 cm of small intestine of mice fed with SD and CDAHFD for 3 wk were embedded in Matrigel (Corning) on a collagen-coated 8 well chamber cover (Matsunami). After Matrigel polymerization, the enteroid culture media were added and incubated for 1 h at 37 °C, $5\%$ CO2. The enteroids were stimulated by 1 μM of CCh (Sigma). The differential interference contrast images of Paneth cells before and 10 min after adding CCh were obtained from confocal microscopy (A1, Nikon). To quantify Paneth cell granule secretion, area of the granules at pre-and post-stimulation was measured by using the image analysis software, NIS-Elements AR, and calculated percent granule secretion. Paneth cells in 10 crypts of each mouse (4 mice of each group) were evaluated.
## Fecal DNA extraction
Fresh fecal samples were collected immediately after excretion, snap-frozen on dry ice, and stored at − 80 °C. For total DNA extraction, 200 mg fecal samples were processed by QIAamp Fast DNA Stool Mini Kit (QIAGEN) according to the manufacturer’s protocol. Final DNA concentrations were quantified from the absorption at 260 nm with a Nanodrop 2000 spectrometer (Thermo Fisher Scientific).
## 16S rDNA sequencing
The V3-V4 variable region of 16S ribosomal RNA genes were amplified from fecal DNA extracts using universal primer set of Bakt 341F (5-cctacgggnggcwgcag) and Bakt 805R (5-gactachvgggtatctaatcc)62. PCR amplification was performed using a 25 μL reaction volume mixtures comprising 0.5 ng/μL of DNA template, 200 nM of each primer and 1 × KAPA HiFi Hot Start Ready Mix (Kapa Biosystems) under the following conditions: initial denaturation at 95 °C for 3 min, 25 cycles of 95 °C for 30 s, 55 °C for 30 s and 72 °C for 30 s, followed by final extension at 72 °C for 5 min. Amplification products were subsequently purified using AMPure XP beads (Beckman Coulter). Then, index PCR was performed using a 50 μL reaction volume mixtures comprising 5 μL of purified PCR products, 5 μL of each index primer containing adapter sequence and sample specific 8 bp barcodes in the Nextera XT Index Kit v2 Set B (Illumina) and 1 × KAPA HiFi Hot Start Ready Mix under the following conditions: 95 °C for 3 min, 8 cycles of 95 °C for 30 s, 55 °C for 30 s and 72 °C for 30 s, followed by 72 °C for 5 min. Resulting amplification products were purified using AMPure XP beads, then quantified using the Qubit dsDNA HS Assay Kit (Invitrogen), and finally adjusted to 4 nM. Each amplicon was pooled and subjected to quantitative PCR using KAPA Library Quantification Kit Lightcycler 480 qPCR Mix (Kapa Biosystems), denatured following Illumina’s guideline, and then adjusted to 4 pM. Amplicon library was combined with $5\%$ of 4 pM PhiX Control v3 (Illumina) and subjected to pair-end sequencing using MiSeq instrument with a MiSeq 600-cycle v3 kit (Illumina). Resulting sequence reads were filtered for read quality, basecalled, and demultiplexed using bcl2fastq software (Illumina).
## 16S rDNA-based taxonomic analysis
Taxonomic analysis of FASTQ files generated from MiSeq was conducted by QIIME2 software (version 2019.7)63. Quality-filtering, denoising, and removal of chimeric sequences were carried out by DADA2 plugin64 using following parameters; –p-trim-left-f 17, –p-trim-left-r 21, –p-trunc-len-f 280, –p-trunc-len-r 200, –p-max-ee-f 2 –p-max-ee-r 2. Phylogenetic tree was constructed with FastTree65 after alignment with MAFFT66. Each feature was taxonomically assigned by a naïve-bayes classifier based on $99\%$ sequence identity to the SILVA database (v.132). α-Diversity (observed OTUs, PD whole tree, Shannon index and Simpson index) and β-diversity (weighted UniFrac distance) were calculated by Qiime2 pipeline. Statistical significance of β-diversity was determined by PERMANOVA test in Qiime2 pipeline.
## Measurement of serum zonulin
Zonulin concentration in serum was quantified by Mouse Haptoglobin ELISA Kit (Abcam, ab157714). The assay was performed according to the manufacture’s recommended methods.
## Bacterial translocation
To quantitate and identify bacteria, the spleen was removed and immediately placed into 500 μL of sterile Luria–Bertani (LB) medium. The spleen then was homogenized with a BioMasher (Nippi), and 200 μL was plated on LB agar plates and cultured either under aerobic conditions for 24 h or anaerobic conditions for 48 h at 37 °C. Colony-forming unit (CFU) were counted and calculated per organ.
## Administration of R-Spo1 and Crp4
Recombinant human R-Spo1 provided from Kyowa Kirin Co., Ltd was generated as previously reported25,26. R-Spo1 was intravenously administered at a dose of 600 μg three times a week for 3 wk. Recombinant Crp4, produced and purified as previously described67, was orally administered at a dose of 110 μg twice daily for 6 wk.
## Statistics
Data were analyzed using GraphPad Prism 8 software (GraphPad software), and results were expressed as individual points with mean values and error bars representing SEM. Statistical significance between 2 groups was determined by unpaired two-tailed Student’s t test. Pearson’s test was used to analyze the correlation. A P value of less than 0.05 was considered statistically significant.
## Supplementary Information
Supplementary Information. The online version contains supplementary material available at 10.1038/s41598-023-30997-y.
## References
1. Younossi Z. **Global perspectives on nonalcoholic fatty liver disease and nonalcoholic steatohepatitis**. *Hepatology* (2019) **69** 2672-2682. DOI: 10.1002/hep.30251
2. Dulai PS. **Increased risk of mortality by fibrosis stage in nonalcoholic fatty liver disease: systematic review and meta-analysis**. *Hepatology* (2017) **65** 1557-1565. DOI: 10.1002/hep.29085
3. Younossi ZM. **Global epidemiology of nonalcoholic fatty liver disease—Meta-analytic assessment of prevalence, incidence, and outcomes**. *Hepatology* (2016) **64** 73-84. DOI: 10.1002/hep.28431
4. Arab JP, Arrese M, Trauner M. **Recent insights into the pathogenesis of nonalcoholic fatty liver disease**. *Annu. Rev. Pathol. Mech. Dis.* (2018) **13** 321-350. DOI: 10.1146/annurev-pathol-020117-043617
5. Marchesini G, Petta S, Grave RD. **Diet, weight loss, and liver health in nonalcoholic fatty liver disease: pathophysiology, evidence, and practice**. *Hepatology* (2016) **63** 2032-2043. DOI: 10.1002/hep.28392
6. Tripathi A. **The gut-liver axis and the intersection with the microbiome**. *Nat. Rev. Gastroenterol. Hepatol.* (2018) **15** 397-411. DOI: 10.1038/s41575-018-0011-z
7. Cani PD. **Microbial regulation of organismal energy homeostasis**. *Nat. Metab.* (2019) **1** 34-46. DOI: 10.1038/s42255-018-0017-4
8. Bäckhed F. **The gut microbiota as an environmental factor that regulates fat storage**. *Proc. Natl. Acad. Sci. U.S.A.* (2004) **101** 15718-15723. DOI: 10.1073/pnas.0407076101
9. Ley RE, Turnbaugh PJ, Klein S, Gordon JI. **Microbial ecology: Human gut microbes associated with obesity**. *Nature* (2006) **444** 1022-1023. DOI: 10.1038/4441022a
10. Kolodziejczyk AA, Zheng D, Shibolet O, Elinav E. **The role of the microbiome in NAFLD and NASH**. *EMBO Mol. Med.* (2019) **11** e9302. DOI: 10.15252/emmm.201809302
11. Aron-Wisnewsky J. **Gut microbiota and human NAFLD: Disentangling microbial signatures from metabolic disorders**. *Nat. Rev. Gastroenterol. Hepatol.* (2020) **17** 279-297. DOI: 10.1038/s41575-020-0269-9
12. Allaire JM. **The intestinal epithelium: Central coordinator of mucosal immunity**. *Trends Immunol.* (2018) **39** 677-696. DOI: 10.1016/j.it.2018.04.002
13. Jones DE, Bevins CL. **Paneth cells of the human small intestine express an antimicrobial peptide gene**. *J. Biol. Chem.* (1992) **267** 23216-23225. DOI: 10.1016/S0021-9258(18)50079-X
14. Ouellette AJ. **Paneth cell α-defensins in enteric innate immunity**. *Cell. Mol. Life Sci.* (2011) **68** 2215-2229. DOI: 10.1007/s00018-011-0714-6
15. Ayabe T. **Secretion of microbicidal α-defensins by intestinal Paneth cells in response to bacteria**. *Nat. Immunol.* (2000) **1** 113-118. DOI: 10.1038/77783
16. Salzman NH, Ghosh D, Huttner KM, Paterson Y, Bevins CL. **Protection against enteric salmonellosis in transgenic mice expressing a human intestinal defensin**. *Nature* (2003) **422** 522-526. DOI: 10.1038/nature01520
17. Yokoi Y. **Paneth cell granule dynamics on secretory responses to bacterial stimuli in enteroids**. *Sci. Rep.* (2019) **9** 2710. DOI: 10.1038/s41598-019-39610-7
18. Masuda K, Sakai N, Nakamura K, Yoshioka S, Ayabe T. **Bactericidal activity of mouse α-defensin cryptdin-4 predominantly affects noncommensal bacteria**. *J. Innate Immun.* (2011) **3** 315-326. DOI: 10.1159/000322037
19. Salzman NH. **Enteric defensins are essential regulators of intestinal microbial ecology**. *Nat. Immunol.* (2010) **11** 76-83. DOI: 10.1038/ni.1825
20. Nakamura K, Sakuragi N, Takakuwa A, Ayabe T. **Paneth cell α-defensins and enteric microbiota in health and disease**. *Biosci. Microb. Food Health* (2016) **35** 57-67. DOI: 10.12938/bmfh.2015-019
21. Shimizu Y. **Paneth cell α-defensin misfolding correlates with dysbiosis and ileitis in Crohn’s disease model mice**. *Life Sci. Alliance* (2020) **3** e201900592. DOI: 10.26508/lsa.201900592
22. Eriguchi Y. **Graft-versus-host disease disrupts intestinal microbial ecology by inhibiting Paneth cell production of α-defensins**. *Blood* (2012) **120** 223-232. DOI: 10.1182/blood-2011-12-401166
23. Eriguchi Y. **Decreased secretion of Paneth cell α-defensins in graft-versus-host disease**. *Transpl. Infect. Dis.* (2015) **17** 702-706. DOI: 10.1111/tid.12423
24. Suzuki K. **Decrease of α-defensin impairs intestinal metabolite homeostasis via dysbiosis in mouse chronic social defeat stress model**. *Sci. Rep.* (2021) **11** 9915. DOI: 10.1038/s41598-021-89308-y
25. Kim K. **Mitogenic influence of human R-spondin1 on the intestinal epithelium**. *Science* (2005) **309** 1256-1260. DOI: 10.1126/science.1112521
26. Takashima S. **The Wnt agonist R-spondin1 regulates systemic graft-versus-host disease by protecting intestinal stem cells**. *J. Exp. Med.* (2011) **208** 285-294. DOI: 10.1084/jem.20101559
27. Hayase E. **R-Spondin1 expands Paneth cells and prevents dysbiosis induced by graft-versus-host disease**. *J. Exp. Med.* (2017) **214** 3507-3518. DOI: 10.1084/jem.20170418
28. Selsted ME, Miller SI, Henschen AH, Ouellette AJ. **Enteric defensins: Antibiotic peptide components of intestinal host defense**. *J. Cell Biol.* (1992) **118** 929-936. DOI: 10.1083/jcb.118.4.929
29. Wehkamp J, Stange EF. **An update review on the Paneth cell as key to ileal Crohn’s disease**. *Front. Immunol.* (2020) **11** 646. DOI: 10.3389/fimmu.2020.00646
30. Fasano A. **Zonulin, a newly discovered modulator of intestinal permeability, and its expression in coeliac disease**. *Lancet* (2000) **355** 1518-1519. DOI: 10.1016/S0140-6736(00)02169-3
31. Ouellette AJ. **Mouse Paneth cell defensins: Primary structures and antibacterial activities of numerous cryptdin isoforms**. *Infect. Immun.* (1994) **62** 5040-5047. DOI: 10.1128/iai.62.11.5040-5047.1994
32. Shanahan MT, Tanabe H, Ouellette AJ. **Strain-specific polymorphisms in paneth cell α-defensins of C57BL/6 mice and evidence of vestigial myeloid α-defensin pseudogenes**. *Infect. Immun.* (2011) **79** 459-573. DOI: 10.1128/IAI.00996-10
33. Takatani N. **Fucoxanthin inhibits hepatic oxidative stress, inflammation, and fibrosis in diet-induced nonalcoholic steatohepatitis model mice**. *Biochem. Biophys. Res. Commun.* (2020) **528** 305-310. DOI: 10.1016/j.bbrc.2020.05.050
34. Ulmasov B. **An inhibitor of arginine-glycine-aspartate-binding integrins reverses fibrosis in a mouse model of nonalcoholic steatohepatitis**. *Hepatol. Commun.* (2019) **3** 246-261. DOI: 10.1002/hep4.1298
35. Wang H. **ERp57 is up-regulated in free fatty acids-induced steatotic L-02 cells and human nonalcoholic fatty livers**. *J. Cell. Biochem.* (2010) **110** 1447-1456. DOI: 10.1002/jcb.22696
36. Cazanave SC. **Death receptor 5 signaling promotes hepatocyte lipoapoptosis**. *J. Biol. Chem.* (2011) **286** 39336-39348. DOI: 10.1074/jbc.M111.280420
37. Zhu L. **Characterization of gut microbiomes in nonalcoholic steatohepatitis (NASH) patients: A connection between endogenous alcohol and NASH**. *Hepatology* (2013) **57** 601-609. DOI: 10.1002/hep.26093
38. Boursier J. **The severity of nonalcoholic fatty liver disease is associated with gut dysbiosis and shift in the metabolic function of the gut microbiota**. *Hepatology* (2016) **63** 764-775. DOI: 10.1002/hep.28356
39. Schmidt TSB, Raes J, Bork P. **The human gut microbiome: From association to modulation**. *Cell* (2018) **172** 1198-1215. DOI: 10.1016/j.cell.2018.02.044
40. Smith BJ. **Changes in the gut microbiota and fermentation products associated with enhanced longevity in acarbose-treated mice**. *BMC Microbiol.* (2018) **19** 130. DOI: 10.1186/s12866-019-1494-7
41. Volk JK. **The Nlrp6 inflammasome is not required for baseline colonic inner mucus layer formation or function**. *J. Exp. Med.* (2019) **216** 2602-2618. DOI: 10.1084/jem.20190679
42. Ou J, Liang S, Guo XK, Hu X. **α-Defensins promote bacteroides colonization on mucosal reservoir to prevent antibiotic-induced dysbiosis**. *Front. Immunol.* (2020) **11** 2065. DOI: 10.3389/fimmu.2020.02065
43. Jiang T. **Lactobacillus mucosae strain promoted by a high-fiber diet in genetic obese child alleviates lipid metabolism and modifies gut microbiota in ApoE–/– mice on a western diet**. *Microorganisms* (2020) **8** 1225. DOI: 10.3390/microorganisms8081225
44. Sicard JF, Bihan GL, Vogeleer P, Jacques M, Harel J. **Interactions of intestinal bacteria with components of the intestinal mucus**. *Front. Cell. Infect. Microbiol.* (2017) **7** 387. DOI: 10.3389/fcimb.2017.00387
45. Horne RG. **High fat-high fructose diet-induced changes in the gut microbiota associated with dyslipidemia in Syrian hamsters**. *Nutrients* (2020) **12** 3557. DOI: 10.3390/nu12113557
46. Ringel Y. **High throughput sequencing reveals distinct microbial populations within the mucosal and luminal niches in healthy individuals**. *Gut Microbes* (2015) **6** 173-181. DOI: 10.1080/19490976.2015.1044711
47. Pereira FC. **Rational design of a microbial consortium of mucosal sugar utilizers reduces**. *Nat. Commun.* (2020) **11** 5104. DOI: 10.1038/s41467-020-18928-1
48. Wu M. **The differences between luminal microbiota and mucosal microbiota in mice**. *J. Microbiol. Biotechnol.* (2020) **30** 287-295. DOI: 10.4014/jmb.1908.08037
49. Hanada T. **The Atg12-Atg5 conjugate has a novel E3-like activity for protein lipidation in autophagy**. *J. Biol. Chem.* (2007) **282** 37298-37302. DOI: 10.1074/jbc.C700195200
50. Ni H. **Nrf2 promotes the development of fibrosis and tumorigenesis in mice with defective hepatic autophagy**. *J. Hepatol.* (2014) **61** 617-625. DOI: 10.1016/j.jhep.2014.04.043
51. Takakuwa A. **Butyric acid and leucine induce α-defensin secretion from small intestinal Paneth cells**. *Nutrients* (2019) **11** 2817. DOI: 10.3390/nu11112817
52. Ayabe T. **Modulation of mouse paneth cell α-defensin secretion by mIKCa1, a Ca2+-activated, intermediate conductance potassium channel**. *J. Biol. Chem.* (2002) **277** 3793-3800. DOI: 10.1074/jbc.M107507200
53. Clarke LL. **Abnormal Paneth cell granule dissolution and compromised resistance to bacterial colonization in the intestine of CF mice**. *Am. J. Physiol. Gastrointest. Liver Physiol.* (2004) **286** 1050-1058. DOI: 10.1152/ajpgi.00393.2003
54. Nashida T, Imai A, Shimomura H. **Relation of Rab26 to the amylase release from rat parotid acinar cells**. *Arch. Oral Biol.* (2006) **851** 89-95. DOI: 10.1016/j.archoralbio.2005.06.005
55. Li SK, Zhu D, Gaisano HY, Brubaker PL. **Role of vesicle-associated membrane protein 2 in exocytosis of glucagon-like peptide-1 from the murine intestinal L cell**. *Diabetologia* (2014) **57** 809-818. DOI: 10.1007/s00125-013-3143-2
56. Dong Y, Wan Q, Yang X, Bai L, Xu P. **Interaction of Munc18 and Syntaxin in the regulation of insulin secretion**. *Biochem. Biophys. Res. Commun.* (2007) **360** 609-614. DOI: 10.1016/j.bbrc.2007.06.107
57. Bansal M. **Optineurin promotes autophagosome formation by recruiting the autophagy-related Atg12-5-16L1 complex to phagophores containing the Wipi2 protein**. *J. Biol. Chem.* (2018) **293** 132-147. DOI: 10.1074/jbc.M117.801944
58. Zhou H. **Bile acid toxicity in Paneth cells contributes to gut dysbiosis induced by high-fat feeding**. *JCI Insight* (2020) **5** e138881. DOI: 10.1172/jci.insight.138881
59. Hodin CM. **Reduced Paneth cell antimicrobial protein levels correlate with activation of the unfolded protein response in the gut of obese individuals**. *J. Pathol.* (2011) **225** 276-284. DOI: 10.1002/path.2917
60. Nakamura K, Sakuragi N, Ayabe T. **A monoclonal antibody-based sandwich enzyme-linked immunosorbent assay for detection of secreted α-defensin**. *Anal. Biochem.* (2013) **443** 124-131. DOI: 10.1016/j.ab.2013.08.021
61. Yokoi Y, Adachi T, Sugimoto R, Kikuchi M, Ayabe T. **Simultaneous real-time analysis of Paneth cell and intestinal stem cell response to interferon-γ by a novel stem cell niche tracking method**. *Biochem. Biophys. Res. Commun.* (2021) **545** 14-19. DOI: 10.1016/j.bbrc.2021.01.050
62. Herlemann DPR. **Transitions in bacterial communities along the 2000 km salinity gradient of the Baltic Sea**. *ISME J.* (2011) **5** 1571-1579. DOI: 10.1038/ismej.2011.41
63. Bolyen E. **Reproducible, interactive, scalable and extensible microbiome data science using QIIME 2**. *Nat. Biotechnol.* (2019) **37** 852-857. DOI: 10.1038/s41587-019-0209-9
64. Callahan BJ. **DADA2: High-resolution sample inference from Illumina amplicon data**. *Nat. Methods* (2016) **13** 581-583. DOI: 10.1038/nmeth.3869
65. Price MN, Dehal PS, Arkin AP. **FastTree 2–approximately maximum-likelihood trees for large alignments**. *PLoS One* (2010) **5** e9490. DOI: 10.1371/journal.pone.0009490
66. Katoh K, Standley DM. **MAFFT multiple sequence alignment software version 7: Improvements in performance and usability**. *Mol. Biol. Evol.* (2013) **30** 772-780. DOI: 10.1093/molbev/mst010
67. Tomisawa S. **Efficient production of a correctly folded mouse α-defensin, cryptdin-4, by refolding during inclusion body solubilization**. *Protein Expr. Purif.* (2015) **112** 21-28. DOI: 10.1016/j.pep.2015.04.007
|
---
title: A comparative analysis of clinicopathological features and survival between
pre and postmenopausal breast cancer from an Indian cohort
authors:
- Vidya P. Nimbalkar
- Savitha Rajarajan
- Snijesh V P
- Annie Alexander
- Rohini Kaluve
- Sumithra Selvam
- Rakesh Ramesh
- Srinath B S
- Jyothi S. Prabhu
journal: Scientific Reports
year: 2023
pmcid: PMC9998443
doi: 10.1038/s41598-023-30912-5
license: CC BY 4.0
---
# A comparative analysis of clinicopathological features and survival between pre and postmenopausal breast cancer from an Indian cohort
## Abstract
Breast cancer (BC) among premenopausal women is an aggressive disease associated with poor outcome despite intensive treatment. Higher burden is observed in southeast Asian countries attributed to younger population structure. We compared the reproductive and clinicopathological characteristics, distribution of subtypes and survival between pre and postmenopausal women from a retrospective cohort of BC patients with median follow up over 6 years to examine the differences. In our cohort of 446 BC patients, $\frac{162}{446}$ ($36.3\%$) were premenopausal. Parity and age at last childbirth were significantly different between pre and postmenopausal women. Premenopausal BC had a higher proportion of HER2 amplified and triple negative breast cancer (TNBC) tumors ($$p \leq 0.012$$). Stratified analysis by molecular subtypes showed TNBC had significantly better disease free (DFS) and overall survival (OS) among premenopausal group (mean survival, pre vs. post, DFS = 79.2 vs. 54.0 months, OS = 72.5 vs. 49.5 months, $$p \leq 0.002$$ for both). Analysis on external datasets (SCAN-B, METABRIC) confirmed this finding for overall survival. Our data confirmed the previously observed association of clinical and pathological features between pre and postmenopausal BC. Exploration of better survival among premenopausal TNBC tumors is warranted in larger cohorts with long term follow up.
## Introduction
Breast cancer (BC) is the most common cancer worldwide. Though incidence is higher in developed nations, mortality due to BC is significantly higher in less developed countries1. As per the national cancer registry program [2020], the incidence of BC is rising and now is the most diagnosed cancer among women within India2. Women with BC in India are a decade younger in comparison to western counterparts, thus affecting younger women in premenopausal age group3,4. Previous studies have reported young women have higher mortality even if diagnosed early and receive an intense treatment compared to the women in the elderly age group5–7. BC in the young women tends to be more aggressive and are often diagnosed in the advanced stage of the disease and have unfavorable tumor characteristics like larger tumor size, lymph node positivity5,7.
Premenopausal breast cancer is a distinct disease as the hormonal milieu is different from the postmenopausal women. Association of common risk factors such as obesity and waist circumference also differ from postmenopausal BC, some studies even suggest inverse relationship8. Young women also tend to have a higher proportion of aggressive subtypes such as Human epidermal growth factor receptor 2 (HER2) amplified and triple negative breast cancer (TNBC) indicating difference in distribution of molecular subtypes9,10. Early detection of the BC is challenging in the premenopausal women due to high density of the breast tissue. Moreover, current therapeutic regimens are based on the menopausal status and molecular subtype of BC11 that contributes to differential prognosis and disease outcome between the pre and postmenopausal patients. Investigation into the trends of BC by menopausal status is important in countries with larger proportion of young premenopausal women from the public and patient health perspectives to inform prevention and early detection policies.
It has been an issue of debate if BC in Asia with occurrence at early age, is a different disease and examination of the population structure showed it could be due to strong cohort effect12. While the patterns of BC by menopausal status have been examined in broad geographical areas, information from regions such as individual countries and subregions is lacking. We evaluated the association of various reproductive and clinicopathological characteristics between pre and postmenopausal BC from a hospital-based cohort in India and examined the factors correlated with prognosis within them.
## Patient cohort
This is an observational, retrospective, hospital-based study conducted at St. John’s Research Institute, Bangalore. In this study, 446 BC patients were recruited from two tertiary cancer care hospitals in Bangalore, India between the year 2008 and 2013. Patients included in this study were women diagnosed with primary BC tumors which were confirmed histologically. These patients were followed-up to 9 years, with a total loss to follow-up of less than $4\%$ and a median follow-up duration of over 6 years. The study was approved by an individual institutional ethical review board. Informed consent was obtained from all the patients. All the clinical and histopathological information such as tumor size and location, grade of differentiation, lymphovascular invasion, perineural invasion, lymph node metastasis, treatment regimens, type of surgery was obtained from the patient’s clinical records. Information on disease progression such as local recurrence, distant metastases was collected during the regular follow up and duration to the event and survival period were recorded during the study.
Tumor cells showing ≥ $1\%$ immunopositivity were considered as positive for estrogen receptor (ER) and progesterone receptors (PR) expression. HER2 expression data was obtained from the hospital records that were assessed by IHC and FISH for equivocal cases. Cases with 3+ IHC staining, or FISH amplified were categorized as HER2 positive. Based on expression of ER, PR and HER2 tumors were classified into three subtypes: hormone receptor positive (HR +), HER2 amplified and TNBC. Women were categorized into pre and postmenopausal groups based on age 50 years as cut-off13.
To validate our findings, two publicly available external databases, Sweden Cancerome Analysis Network-Breast (SCAN-B) and The Molecular Taxonomy of Breast Cancer International Consortium (METABRIC) were used. SCAN-B is the large population-based cohort of primary breast tumors. Our analysis included a cohort of 3207 primary BC patients from the SCAN-B study (GSE202203)14 and data was downloaded from Gene expression Omnibus database (GEO)15. TNBC patient’s data ($$n = 320$$) (as per the clinical groups mentioned in the dataset) was used for analysis. METABRIC data (EGAS00000000098) was downloaded from cBioPortal16–18. TNBC patient’s data ($$n = 307$$) (as per the 3 gene classifier subtype mentioned in the dataset) from this database was considered for analysis.
## Statistical analysis
Statistical analysis was performed using SPSS statistical software Version 25.0 (SPSS, Chicago, IL, USA). Descriptive statistics was used to examine the distribution of the study variables. Association of various reproductive and clinicopathological factors with pre and postmenopausal groups were assessed using Chi square test. Disease free survival (DFS) was calculated as duration from date of surgery till the first evidence of metastasis/recurrence. Overall survival (OS) was calculated as the duration from surgery to the death of the patient. Survival probability was calculated by Kaplan Meier survival analysis and was compared between groups using log rank test, in an entire cohort and stratified by menopausal status, the risk factor associated towards the progression of disease was calculated using univariate and multivariable Cox proportional hazard model. Results are represented as hazard ratio (HR) with $95\%$ confidence interval (CI). In addition, stratified analysis was also carried out by subtype of tumor category to compare the disease-free survival and overall survival between pre and postmenopausal status. P value < 0.05 was considered as statistically significant.
## Ethical approval
The studies involving human participants were reviewed and approved by the Institutional Ethical Committee, St John’s Medical College and Hospital, Bangalore (No. IEC/$\frac{1}{655}$/2019). The patients/participants provided their written informed consent to participate in this study. This study was performed in line with the principles of the Declaration of Helsinki.
## Results
Median age of the cohort was 55 years, ranging between 24 and 88 years. There were 162 premenopausal ($36.3\%$) and 284 postmenopausal ($63.7\%$) BC patients. Large proportion of the patients were from urban background (80–$85\%$), $12\%$ were illiterate and $76\%$ belonged to home maker category. Around $13\%$ patients had family history of BC and $15\%$ patients had family history of any other cancer in the first-degree relatives. There was an overlap of $3\%$ cases having family history of BC and any other cancer in the first-degree relatives.
Duration of the lump felt was ranging from a week to few years. In majority of the cases duration of the lump felt was between > 1 month and 12 months ($55\%$). There were three cases that were detected at screening. Of the total patients, $52\%$ had tumor in the right breast, $45.5\%$ had in the left breast and $2.5\%$ had bilateral BC. Most common mode of detection was mammography ($54\%$) followed by fine needle aspiration cytology (FNAC) ($27\%$), ultrasound, biopsy/surgery, computed tomography (CT) scan and positron emission tomography (PET) CT. All patients received age and stage appropriate standard of care treatment. Majority of the patients (> $90\%$) received anthracycline or taxane based chemotherapy and all HR + patients received antihormonal therapy with tamoxifen or aromatase inhibitor. Most common site of metastasis was bone, liver, lung, followed by skeletal muscle, brain and others.
## Association of reproductive features between pre and postmenopausal breast cancer
The association of reproductive features like, age at menarche, age at first childbirth (FCB) and last childbirth (LCB) and parity with the risk of the BC development/ progression were assessed between pre and postmenopausal BC groups. Based on median cut-off of age at menarche, FCB and LCB and parity, BC patients were divided into two categories. Results are represented in Table1.Table 1Association of reproductive features between pre and postmenopausal breast cancer. VariableSubcategory ($$n = 446$$)Premenopausal group ($$n = 162$$)n (%)Postmenopausal group ($$n = 284$$)n (%)P valueAge at Menarche (years)< 14 ($$n = 193$$)66 (45.2)127 (48.7)0.503≥ 14 ($$n = 214$$)80 (54.8)134 (51.3)Age at FCB (years)< 23 ($$n = 169$$)55 (42.6)114 (48.1)0.316≥ 23 ($$n = 197$$)74 (57.4)123 (51.9)Age at LCB (years)< 30 ($$n = 138$$)53 (56.4)85 (42.7)0.029≥ 30 ($$n = 155$$)41 (43.6)114 (57.3)Parity< 2 ($$n = 54$$)29 (21.1)25 [10]0.003≥ 2 ($$n = 330$$)108 (78.9)222 [90]FCB First childbirth, LCB last childbirth.$p \leq 0.05$, Statistically significant (represented in bold).
We observed significantly higher proportion of postmenopausal BC women with older age at LCB (≥ 30 years) which was further supported by increased parity in this group. Age at menarche and FCB did not differ with menopausal status. We also did not find any association of family history of BC or any other cancer with pre and postmenopausal groups.
## Association of clinicopathological features between pre and postmenopausal breast cancer
Distribution of tumor characteristics such as tumor size (T size), lymph node (LN) status, grade, tumor infiltrating lymphocytes (TILs), did not differ between pre and postmenopausal groups. Premenopausal group had significantly higher proportion of ER negative ($$n = 162$$, premenopausal $$n = 69$$ ($43\%$), postmenopausal = 93 ($33\%$) ($$p \leq 0.038$$), HER2 amplified ($$n = 86$$, premenopausal $$n = 37$$ ($23\%$), postmenopausal = 49 ($17\%$) and TNBC tumors ($$n = 107$$, premenopausal $$n = 48$$ ($30\%$), postmenopausal $$n = 59$$ ($21\%$) ($$p \leq 0.012$$) compared to postmenopausal group. Association of various clinical characteristics of the tumor between pre and postmenopausal groups is represented in Table 2.Table 2Association of clinicopathological features between pre and postmenopausal breast cancer. VariableSubcategory ($$n = 446$$)Premenopausal ($$n = 162$$)n (%)Postmenopausal ($$n = 284$$)n (%)P valueT size (cm)T1 ($$n = 108$$)39 [27]69 [26]0.860T2 ($$n = 253$$)88 [60]165 [63]T3 ($$n = 49$$)19 [13]30 [11]LN statusPositive ($$n = 274$$)105 [65]169 [60]0.268Negative ($$n = 172$$)57 [35]115 [40]GradeI ($$n = 30$$)9 [7]21 [8]0.525II ($$n = 186$$)61[45]125 [49]III ($$n = 175$$)66 [48]109 [43]TILsPresent ($$n = 204$$)81 [57]123 [50]0.136Absent ($$n = 185$$)60 [43]125 [50]ERPositive ($$n = 284$$)93 [57]191 [67]0.038Negative [162]69 [43]93 [33]PGRPositive ($$n = 265$$)93 [57]172 [61]0.514Negative ($$n = 181$$)69 [43]112 [39]HER2Positive ($$n = 86$$)37 [23]49 [17]0.353Negative ($$n = 307$$)107 [66]200 [70]Equivocal ($$n = 53$$)18 [11]35 [13]SubtypeHR + ($$n = 253$$)77 [47]176 [62]0.012HER2 ($$n = 86$$)37 [23]49 [17]TNBC ($$n = 107$$)48 [30]59 [21]LN status lymph node status, TILs tumour infiltrating lymphocytes, ER estrogen receptor, PGR progesterone receptor, HER2 human epithelial growth factor receptor, HR + hormone receptor positive, TNBC triple negative breast cancer.$p \leq 0.05$, statistically significant (represented in bold).
## Association of clinicopathological features with disease progression within pre and postmenopausal tumors
The risk associated with each tumor characteristics towards the progression of the disease was evaluated by univariate Cox proportional hazard model in the entire cohort. Grade II and III tumors were categorised into high grade with grade I as the reference (low grade) for examination of association with survival. High grade (HR = 4.3 $95\%$ (1.06–17.5), larger tumor size (categorised as per AJCC criterion) (T3 vs. T1) (HR = 2.64, $95\%$ CI = 1.43–4.86), LN positivity (HR = 2.9, $95\%$ CI = 1.9–4.5), HER2 amplification (HR = 2.07 $95\%$ CI = 1.33–3.3) and TNBC subtype (HR = 1.62, $95\%$ CI = 1.03–2.56) were significantly associated with increased hazard ($p \leq 0.05$), whereas presence of the TILs infiltration (HR = 0.62, $95\%$ CI = 0.00–0.93), had significantly decreased hazard for the disease progression ($$p \leq 0.02$$). Pre and postmenopausal groups did not show any association with risk of the disease progression in our cohort. We further confirmed the association of these pivotal factors with progression of the disease by multivariate analysis. LN positivity, HER2 amplification and TNBC subtype were associated with significantly increased hazard whereas presence of TILs was associated with lower risk of disease progression in the entire cohort as mentioned in Table 3.Table 3Multivariate analysis of clinicopathological tumor characteristics in entire cohort, pre and postmenopausal groups for disease free survival. ReferenceVariableEntire cohort ($$n = 340$$)Premenopausal groups ($$n = 119$$)Postmenopausal groups ($$n = 221$$)HR ($95\%$CI)P valueHR ($95\%$CI)P valueHR ($95\%$CI)P valueT size (cm)T1T21.2 (0.66–1.95)0.6730.67 (0.3–1.5)0.331.7 (0.75–3.89)0.65T1T31.8 (0.86–3.47)0.1280.73 (0.2–2.7)0.633.34 (1.2–8.8)0.016LN statusNegativePositive3.75 (2.16–6.59)0.00013.0 (1.3–7.5)0.0134.65 (2.23–9.7)0.0001TILsAbsentPresent0.54 (0.00–0.83)0.0040.49 (0.24–1.05)0.0680.66 (0.38–1.2)0.140SubtypeHR + HER2 amplified2.17 (1.33–3.55)0.0021.94 (0.92–4.13)0.0862.05 (1.06–3.99)0.032HR + TNBC1.90 (1.1–3.27)0.0220.62 (0.22–1.73)0.363.55 (1.84–6.83)0.001HR hazard ratio, CI confidence interval, LN status lymph node status, TILs tumor infiltrating lymphocytes, HR + Hormone receptor positive, HER2 human epithelial growth factor receptor, TNBC triple negative breast cancer.$p \leq 0.05$, statistically significant (represented in bold).
Next, we evaluated the association of these factors independently in the pre and postmenopausal groups. Among the postmenopausal tumors, association of tumor size, LN positivity, HER2 amplification and TNBC subtype showed the similar trends of association as in the entire cohort. Presence of TILs was associated with decreased hazard and LN positivity was associated with increased hazard of disease progression in premenopausal BC.
Though TNBC subtype was associated with higher hazard in the entire cohort and within postmenopausal subgroup, TNBC tumors within premenopausal group did not show any association with the hazard in multivariate analysis (Table 3). To verify these findings further, we evaluated the hazard associated with menopausal status within TNBC tumors alone. Both univariate (HR = 0.28 (0.0–0.67), $$p \leq 0.004$$) and multivariate analysis (HR = 0.20 (0–0-0.56), $$p \leq 0.002$$) showed within TNBC subtype, premenopausal BC patients had lower hazard of disease progression.
Kaplan Meier survival analysis confirmed the results of multivariate analysis and showed no difference in the DFS between pre and postmenopausal BC (mean survival time 77.1 vs. 77.9 months) ($$p \leq 0.77$$). We next examined the influence of subtypes on DFS independently within pre and postmenopausal BC. In postmenopausal BC, patients with TNBC and HER2 amplified tumors were associated with poor outcome as expected and there was no difference in the survival between patients with HER2 amplified and TNBC tumors (mean survival, HR + vs. TNBC, 83.8 vs. 54 months, $p \leq 0.0001$, HR + vs. HER2 amplified tumor 83.8 vs. 63.3, $$p \leq 0.014$$, HER2 amplified vs. TNBC 63.3 vs. 54 months, $$p \leq 0.328$$).
However, comparison of DFS among different subtypes within premenopausal patients showed, no difference in the survival between HR + and TNBC breast cancer patients and both subtypes were associated with better outcome compared to patients with HER2 amplified tumors (Fig. 1A, mean survival, TNBC vs. HER2 amplified BC 79.2 vs. 61 months, $$p \leq 0.006.$$ mean survival HR + vs. HER2 amplified BC, 74.4 vs. 61 months, $$p \leq 0.058.$$ mean survival, TNBC vs. HR + tumors 79.2 vs. 74.4 months, $$p \leq 0.228$$).Figure 1Kaplan Meier survival plots. ( A) Disease free survival (DFS) across three subtypes in premenopausal breast cancer. ( B) DFS in TNBC tumors between pre and postmenopausal BC. ( C) Overall survival (OS) in TNBC tumors between pre and postmenopausal BC in SCAN-B database and (D) OS in TNBC tumors between pre and postmenopausal BC in METABRIC database.
Next, we examined the DFS and OS independently in each of the molecular subtypes that is HR +, HER2 amplified and TNBC between pre and postmenopausal tumors. There was no difference in the survival between the two groups in the HR + tumors (mean survival, pre vs post 74.4 vs. 83.8 months, $$p \leq 0.374$$) and in HER2 amplified tumors (mean survival, pre vs. post 61 vs. 63.3 months, $$p \leq 0.690$$). No difference was observed in OS either.
Interestingly in the TNBC subtype, premenopausal patients were associated with better outcome compared to postmenopausal BC patients (Fig. 1B, mean survival, pre vs. post 79.2 vs. 54 months, $$p \leq 0.002$$) which was contrary to the notion that TNBC tumors are always associated with aggressive behaviour. Similar trends were noted with overall survival (mean survival, pre vs. post, 72.5 vs. 49.5 months, $$p \leq 0.002$$). To rule out if these results could be due to the bias within our cohort, we validated our findings in two publicly available larger databases, SCAN-B and METABRIC with different population. OS analysis in these cohorts confirmed our findings on association of premenopausal BC patients with better prognosis within TNBC subtype (Fig. 1C,D, In SCAN-B database, mean survival, pre vs. post 94.2 vs. 84.4 months, $$p \leq 0.008$$ and in METABRIC database, mean survival, pre vs. post, 186.8 vs. 157.2 months, $$p \leq 0.033$$). We also performed the COX proportional hazard ratio model in these two cohorts to verify similar trends in pre and postmenopausal groups within TNBC subtype of tumors. In both SCAN-B and METABRIC datasets, TNBC tumors with premenopausal BC were associated with decreased hazard on univariate analysis [SCAN-B, HR = 0.42, $95\%$ CI = 0.00–0.82), $$p \leq 0.010$$, METABRIC, HR = 0.68, $95\%$ CI 0.00–0.98), $$p \leq 0.034$$] and multivariate analysis (SCAN-B, HR = 0.45, $95\%$ CI = 0.00–0.85), $$p \leq 0.014$$, METABRIC, HR = 0.56 (0.00–0.82), $$p \leq 0.003$$) compared to postmenopausal tumors.
## Discussion
Epidemiological studies have often distinguished between pre and postmenopausal BC mainly based on difference in risk factors like age and reproductive factors. Risk factors such as abdominal obesity, parity and age at childbirth are shown to be different between pre and postmenopausal BC19–21. Average age of women with BC in developing countries is about 10 years lower than developed nations, leading to higher burden of younger patients. As early onset of disease could reflect more fundamental etiologic differences, we investigated the different reproductive and clinical features and compared them between pre and postmenopausal BC.
Previous studies have evaluated the association and prognostic significance of reproductive features such as age at menarche, FCB, LCB and parity with breast cancer outcomes22,23. In our study cohort, women in postmenopausal age group had significantly higher age at LCB and higher parity. Early age at menarche, older age at menopause, older age at FCB and LCB are known to increase women’s risk of developing BC24. Zhang et al. evaluated the association of multiple reproductive factors with BC prognosis with respect to menopausal status and reported association of increased hazard with older age at FCB and LCB in premenopausal BC patients10. Significant inconsistencies are however found in association of the reproductive factors with prognosis in both pre and postmenopausal BC. A prospective cross-sectional study conducted in the Indian BC patient’s cohort reported that premenopausal BC patients had younger age at menarche, older age at FCB, lesser parity, lesser BMI and denser breast tissue compared to postmenopausal BC patients25. A study by Butt et al. reported that most of risk factors for pre and postmenopausal BC are same except less parity, which increased the risk for postmenopausal BC21.
In tune with the previous reports, our cohort of premenopausal BC had higher proportion of ER negative, HER2 amplified and TNBC tumors compared to postmenopausal tumors which were predominantly hormone receptor positive [9, 28, 2]. There was no difference in the relative distribution of other clinicopathological features between pre and postmenopausal groups. Other studies26,27 on premenopausal tumors have shown difference in tumor variables such as increased tumor size, lymph node metastasis, ER, PR and Ki67 expression and decreased HER2 and p53 expression compared to postmenopausal tumors. We did not find these differences between the pre or postmenopausal BC patients in our cohort. These differences might be due to different ethnicity as well as proportion of pre and postmenopausal patients in these cohorts compared to our study cohort. Premenopausal patients in these cohorts were proportionately much higher than in our study.
Among the molecular subtypes, previous studies such as the Carolina breast cancer study has reported premenopausal BC patients have higher prevalence of Basal-like breast tumors compared with postmenopausal patients28. The well-established clinical factors such as LN involvement, HER2 amplified status showed increased hazard of poor prognosis as expected. Better survival observed in the premenopausal TNBC tumors is unique finding of our study. To nullify the effect of inadequate chemotherapeutic treatment, which might influence adverse prognosis, we verified and confirmed that all the patients had received stage appropriate standard of care with anthracyclins and taxane based regimens. Multiple studies have reported higher proportion of aggressive subtype TNBC within Asian countries29,30. A very recent study by Nishimura et al. evaluated the clinical significance of menopausal status in TNBC. This study reported that higher proportion of TNBC tumors in postmenopausal patients and their association with favourable tumor characteristics and better DFS compared to premenopausal patients31 which is contrary to our observations. Disparities in TNBC outcomes have been reported based on race and ethnicity in multiple studies earlier32–34. Younger median age of our cohort and ethnicity-based differences between Indian and Japanese women might have contributed to the observed difference. Recent study by Yunzi et al.35 investigated tumor characteristics and mortality in south east Asian (SEA) women by regional distribution specific to countries in SEA and found the cancer specific survival was best for Japanese women among all SEA ethnic populations and suggests disaggregation by country or region of origin to identify subgroups that are at risk for worse outcomes.
TNBC tumors are heterogenous and consists of multiple molecular subtypes. PAM 50 molecular subtyping has shown inclusion of normal breast-like and claudin-low molecular subtypes other than basal-like tumors. Normal breast-like tumors have been clustered with basal-like tumors in the majority of the studies, but with better prognosis36. TNBC in the premenopausal age group may be enriched with normal breast like TNBC than basal like TNBC tumors, however there are very few studies to examine or compare prognostic significance of premenopausal TNBC tumors. Our findings need further validation in large cohorts with diverse ethnicity for confirmation.
Our study has limitations due to the retrospective cohort, with non-availability of information on weight, BMI and obesity amongst the all the women enrolled for the study. Previous studies have correlated the importance of waist circumference and central obesity as both risk factor for BC occurrence and prognosis37. Excess weight gain with onset of menopause among women with BC is also associated with other morbidities. We were unable to derive association of prognosis with density of the breast tissue due to lack of radiological information which is considered significant risk factor especially among young premenopausal women38. Lack of therapeutic interventions for ovarian suppression among the premenopausal women could have influenced the prognosis in this subgroup though majority of the women were treated with standard of care prevalent at the time of diagnosis. Findings from recent trails such as TEXT and SOFT39 showing improvements with use aromatase inhibitors along with ovarian suppression over the traditionally used tamoxifen is likely to influence the prognosis among young patients with BC in future.
In conclusion, findings of our study confirm the higher proportion of aggressive subtypes such as ER negative, HER2 amplified and TNBC within premenopausal BC suggesting the effect of hormonal milieu on the biology of tumor development. Parity and age at last childbirth were different between pre and postmenopausal women. We also noticed differential prognosis in TNBC tumors based on menopausal status in our cohort, which needs verification based on ethnicity and age distribution across other cohorts. Categorising tumor subtypes especially, TNBC tumors based on the menopausal status will help in developing better treatment strategies and in predicting outcome.
## Data availability
The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request. Publicly available datasets used in the current study: Sweden Cancerome Analysis Network-Breast (SCAN-B) (GSE202203; https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE202203). The Molecular Taxonomy of Breast Cancer International Consortium (METABRIC). ( EGAS00000000098; https://cbioportal-datahub.s3.amazonaws.com/brca_metabric.tar.gz).
## References
1. Sung H, Ferlay J, Siegel RL, Laversanne M, Soerjomataram I, Jemal A, Bray F. **GLobal cancer statistics 2020.pdf**. *CA Cancer J. Clin.* (2021) **71** 209-249. DOI: 10.3322/caac.21660
2. Mathur P. **Cancer statistics, 2020 report from national cancer registry programme**. *India. JCO Glob. Oncol.* (2020) **6** 1063-1075. DOI: 10.1200/GO.20.00122
3. Chopra B. **Age shift: Breast cancer is occurring in younger age groups—Is it true?**. *Clin. Cancer Investig. J.* (2014) **3** 526. DOI: 10.4103/2278-0513.142652
4. Malvia S, Bagadi SA, Dubey US, Saxena S. **Epidemiology of breast cancer in Indian women**. *Asia Pac. J. Clin. Oncol.* (2017) **13** 289-295. DOI: 10.1111/ajco.12661
5. Fredholm H. **Breast cancer in young women: Poor survival despite intensive treatment**. *PLoS ONE* (2009) **4** 1-9. DOI: 10.1371/journal.pone.0007695
6. Anders CK. **Young age at diagnosis correlates with worse prognosis and defines a subset of breast cancers with shared patterns of gene expression**. *J. Clin. Oncol.* (2008) **26** 3324-3330. DOI: 10.1200/JCO.2007.14.2471
7. Fredholm H. **Long-term outcome in young women with breast cancer: A population-based study**. *Breast Cancer Res. Treat.* (2016) **160** 131-143. DOI: 10.1007/s10549-016-3983-9
8. Schoemaker MJ. **Association of body mass index and age with subsequent breast cancer risk in premenopausal women**. *JAMA Oncol.* (2018) **4** 1-10. PMID: 28973558
9. Chollet-Hinton L. **Breast cancer biologic and etiologic heterogeneity by young age and menopausal status in the Carolina breast cancer study: A case-control study**. *Breast Cancer Res.* (2016) **18** 1-10. DOI: 10.1186/s13058-016-0736-y
10. Zhang JY. **Associations of reproductive factors with breast cancer prognosis and the modifying effects of menopausal status**. *Cancer Med.* (2020) **9** 385-393. DOI: 10.1002/cam4.2707
11. Heer E. **Global burden and trends in premenopausal and postmenopausal breast cancer: A population-based study**. *Lancet Glob. Heal.* (2020) **8** e1027-e1037. DOI: 10.1016/S2214-109X(20)30215-1
12. Lim YX, Lim ZL, Ho PJ, Li J. **Breast cancer in Asia: Incidence, mortality, early detection, mammography programs, and risk-based screening initiatives**. *Cancers Basel.* (2022) **14** 4218. DOI: 10.3390/cancers14174218
13. Phipps AI. **Defining menopausal status in epidemiologic studies: A comparison of multiple approaches and their effects on breast cancer rates**. *Maturitas* (2010) **67** 60-66. DOI: 10.1016/j.maturitas.2010.04.015
14. Dalal H. **Clinical associations of ESR2 (estrogen receptor beta) expression across thousands of primary breast tumors**. *Sci. Rep.* (2022) **12** 1-12. DOI: 10.1038/s41598-022-08210-3
15. Edgar R, Domrachev M, Lash AE. **Gene expression Omnibus: NCBI gene expression and hybridization array data repository**. *Nucleic Acids Res.* (2002) **30** 207-210. DOI: 10.1093/nar/30.1.207
16. Rueda OM. **Dynamics of breast-cancer relapse reveal late-recurring ER-positive genomic subgroups**. *Nature* (2019) **567** 399-404. DOI: 10.1038/s41586-019-1007-8
17. Pereira B. **The somatic mutation profiles of 2433 breast cancers refines their genomic and transcriptomic landscapes**. *Nat. Commun.* (2016) **7** 11098. PMID: 27020798
18. Curtis C, Shah SP, Chin S, Turashvili G. **Europe PMC funders group the genomic and transcriptomic architecture of 2000 breast tumours reveals novel subgroups**. *Nat. Commun.* (2012) **486** 346-352. DOI: 10.1038/nature10983
19. Amadou A, Hainaut P, Romieu I. **Role of obesity in the risk of breast cancer: Lessons from anthropometry**. *J. Oncol.* (2013) **2013** 19. DOI: 10.1155/2013/906495
20. Kruk J. **Association of lifestyle and other risk factors with breast cancer according to menopausal status: A case-control study in the region of western Pomerania (Poland)**. *Asian Pac. J. Cancer Prev.* (2007) **8** 513-524. PMID: 18260721
21. Butt Z. **A comparison between pre-menopausal and post-menopausal women**. *J. Res. Pharm. Pract.* (2014) **3** 12-18. PMID: 24991630
22. Butt S, Borgquist S, Anagnostaki L, Landberg G, Manjer J. **Parity and age at first childbirth in relation to the risk of different breast cancer subgroups**. *Int. J. Cancer* (2009) **125** 1926-1934. DOI: 10.1002/ijc.24494
23. Sighoko D. **Breast cancer in pre-menopausal women in West Africa: Analysis of temporal trends and evaluation of risk factors associated with reproductive life**. *Breast* (2013) **22** 828-835. DOI: 10.1016/j.breast.2013.02.011
24. Sun YS. **Risk factors and preventions of breast cancer**. *Int. J. Biol. Sci.* (2017) **13** 1387-1397. DOI: 10.7150/ijbs.21635
25. Surakasula A, Nagarjunapu G, Raghavaiah K. **A comparative study of pre- and post-menopausal breast cancer: Risk factors, presentation, characteristics and management**. *J. Res. Pharm. Pract.* (2014) **3** 12. DOI: 10.4103/2279-042X.132704
26. Kocaöz S, Korukluoǧlu B, Parlak Ö, Doǧan HT, Erdoǧan F. **Comparison of clinicopathological features and treatments between pre- And postmenopausal female breast cancer patients—A retrospective study**. *Prz. Menopauzalny* (2019) **18** 68-73. PMID: 31485202
27. Feng F. **Comparison of epidemiological features, clinicopathological features, and treatments between premenopausal and postmenopausal female breast cancer patients in western China: A retrospective multicenter study of 15,389 female patients**. *Cancer Med.* (2018) **7** 2753-2763. DOI: 10.1002/cam4.1503
28. Carey LA. **Race, breast cancer subtypes, and survival in the Carolina breast cancer study**. *J. Am. Med. Assoc.* (2006) **295** 2492-2502. DOI: 10.1001/jama.295.21.2492
29. Wang C. **Triple negative breast cancer in Asia: An insider’s view**. *Cancer Treat. Rev.* (2018) **62** 29-38. DOI: 10.1016/j.ctrv.2017.10.014
30. Thakur KK, Bordoloi D, Kunnumakkara AB. **Alarming burden of triple-negative breast cancer in India**. *Clin. Breast Cancer* (2018) **18** e393-e399. DOI: 10.1016/j.clbc.2017.07.013
31. Nishimura R. **Triple Negative Breast Cancer: An Analysis of the Subtypes and the Effects of Menopausal Status on Invasive Breast Cancer**. *J. Clin. Med.* (2022) **11** 2331. DOI: 10.3390/jcm11092331
32. Prakash O. **Racial disparities in triple negative breast cancer: A review of the role of biologic and non-biologic factors**. *Front. Public Health* (2020) **8** 1-14. DOI: 10.3389/fpubh.2020.576964
33. Makhani SS. **Racial and ethnic inequality in survival outcomes of women with triple negative breast cancer**. *Cureus* (2022) **14** 27120
34. Oppong BA. **Are there differences in outcomes by race among women with metastatic triple-negative breast cancer?**. *Breast Cancer Res. Treat.* (2022) **196** 399-408. DOI: 10.1007/s10549-022-06736-8
35. Yu AYL. **Disease characteristics and mortality among Asian women with breast cancer**. *Cancer* (2022) **128** 1024-1037. DOI: 10.1002/cncr.34015
36. Kumar P, Aggarwal R. **An overview of triple-negative breast cancer**. *Arch. Gynecol. Obstet.* (2016) **293** 247-269. DOI: 10.1007/s00404-015-3859-y
37. Nagrani R. **Central obesity increases risk of breast cancer irrespective of menopausal and hormonal receptor status in women of South Asian ethnicity**. *Eur. J. Cancer* (2016) **66** 153-161. DOI: 10.1016/j.ejca.2016.07.022
38. McCormack VA, Dos Santos Silva I. **Breast density and parenchymal patterns as markers of breast cancer risk: A meta-analysis**. *Cancer Epidemiol. Biomarkers Prev.* (2006) **15** 1159-1169. DOI: 10.1158/1055-9965.EPI-06-0034
39. Francis PA. **Tailoring adjuvant endocrine therapy for premenopausal breast cancer**. *N. Engl. J. Med.* (2018) **379** 122-137. DOI: 10.1056/NEJMoa1803164
|
---
title: 'The free bilamellar autograft (FBA) procedure: A comprehensive case series
of an alternative surgical approach to reconstruction of large eyelid defects'
authors:
- Heather M. McDonald
- Katherine A. McDonald
- Hector McDonald
journal: Frontiers in Surgery
year: 2023
pmcid: PMC9998481
doi: 10.3389/fsurg.2023.1038952
license: CC BY 4.0
---
# The free bilamellar autograft (FBA) procedure: A comprehensive case series of an alternative surgical approach to reconstruction of large eyelid defects
## Abstract
### Purpose
The free bilamellar autograft (FBA) procedure involves harvesting a free, full-thickness section of eyelid tissue from one of the patient’s healthy eyelids to reconstruct a large defect of the involved eyelid. No vascular augmentation is employed. The purpose of this study was to determine the structural and cosmetic results of this procedure.
### Methods
A case series was performed, looking at patients who underwent the FBA procedure for large, full-thickness eyelid defects (>$50\%$ eyelid length) between 2009 and 2020 at a single oculoplastic surgical centre. Basal cell carcinomas most frequently met criteria for the procedure. OHSN-REB waived ethics approval. All surgeries were performed by one surgeon. A single operation, with surgical steps reported in detail, was completed with follow-up documentation at ideally 1 day, 1 week, 1 month, 3 months, 6 months, and 1 year. The mean follow-up period was 28 months.
### Results
Thirty-one patients (17 males, 14 females, mean age 78-years-old) were included in the case series. Comorbidities included diabetes and smoking. Most patients had known basal cell carcinomas removed from the upper or lower eyelid. The mean widths of the recipient and donor sites were 18.8 and 11.5 mm, respectively. All 31 FBA surgeries resulted in structurally functional, cosmetically pleasing, and viable eyelids. Six patients developed minor graft dehiscence, 3 developed an ectropion, and 1 developed mild superficial graft necrosis secondary to frostbite, which fully recovered. Three healing phases were identified.
### Conclusion
This case series adds to the currently sparse data on the free bilamellar autograft procedure. The surgical technique is clearly articulated and illustrated. The FBA procedure is a simple and efficient alternative to current surgical techniques in the reconstruction of full-thickness upper and lower eyelid defects. The FBA provides functional and cosmetic success, despite the absence of an intact blood supply, with decreased operative time and faster recovery.
## Introduction
The eyelid plays an essential role in protecting the globe, maintaining a stable tear film, and nourishing the cornea. It is a complex tissue composed of an inner layer of mucosa, semi-rigid connective tissue of the tarsal plate, muscle, blood vessels, nerves, secretory glands, and an outer layer of skin. The eyelid is considered a bilamellar structure with the orbicularis oculi and skin comprising the anterior lamella and the tarsal plate and conjunctiva comprising the posterior lamella [1]. Large defects (>$50\%$ horizontal length) of the eyelid can arise from multiple conditions, most often post-excision of an eyelid skin cancer [2]. Basal cell carcinoma (BCC) is the most common type of skin cancer, with $20\%$ appearing in the periocular region [3]. Reconstruction is required to recreate the original anatomy of the lid. Current standards for reconstructing large, full-thickness eyelid defects include a wide variety of surgical options [4, 5]. Procedures include direct closure for smaller defects (<$33\%$) and lateral cantholysis with Tenzel semicircular flaps for moderate-sized defects ($33\%$–$66\%$). However, when the defect is larger than $66\%$, a combination of an anterior graft with posterior flap, anterior flap with posterior graft, or anterior and posterior flaps have historically been required.
Traditional teaching states that only one lamella can be formed using a free graft, and the other must be a vascularized flap to minimize the risk of central necrosis [6, 7]. Larger lid defects often require multi-step, time-consuming reconstructive procedures such as the Hughes tarsoconjunctival flap and the Cutler-Beard procedure. These surgeries require 2–8 weeks between steps, resulting in prolonged discomfort to the patient, temporary obstruction to their vision, a cosmetically displeasing appearance during this time, as well as increased operating room hours. These measures are a consequence of previous studies reporting high failure rates for full-thickness eyelid grafts without an augmented vascular supply [8]. However, such a procedure represents the best possible option for reconstructing anatomy and appearance.
Composite grafts have been used by some surgeons. These are full-thickness eyelid grafts which are often manipulated or combined with a vascular bed. The anterior and posterior lamellae may be offset [9], a skin flap may be utilized in combination with an eyelid margin/tarsus/conjunctiva graft [10], the orbicularis oculi muscle may be removed [11], or a “sandwich” technique may be employed (the orbicularis functions as an advancement flap and is sandwiched between a free posterior and anterior lamellar graft) [12]. The next step is exploring the use of a free, full-thickness eyelid graft. The periocular region is known to be well-vascularized, allowing for superior wound healing in the absence of supplemental vascular supply [13]. In 2021, Tenland et al. published a small case series of 10 patients who underwent successful full-thickness eyelid grafting, a procedure rarely described in textbooks and journals (14–17). Although done independently, we enhance the data provided in the Tenland et al. case series [16] by presenting a larger case series of a similar procedure, including a broader range of comorbidities, phases of healing, and an emphasis on the surgical technique. As per Tenland et al., we refer to this as a free bilamellar autograft (FBA) procedure.
## Materials and methods
The single-center case series data was collected from patients operated on between February 2009 to September 2020 in Ottawa, Canada. All surgeries took place at one oculoplastic surgical centre and were performed by one surgeon (HM). All patient data and outcomes were documented in the clinic’s electronic medical record (THINK EMR). The Ottawa Health Science Network Research Ethics Board (OHSN-REB) waived ethics approval for the case series. The case series adhered to the tenets of the Declaration of Helsinki. All referred patients with large, full-thickness eyelid defects (>$50\%$ eyelid length) considered too large for direct closure were offered the FBA procedure. Patients with clinically suggestive BCC limited to the eyelid were preferentially selected over squamous cell carcinoma (SCC) because SCC behaves more aggressively in the periocular region. Furthermore, the cases of SCC that presented to the clinic often extended beyond the eyelid skin and therefore these patients were better candidates for alternative reconstructive procedures. Informed verbal and written consent for the procedure included: [1] a description of the procedure, [2] risks and benefits of the procedure, [3] alternative surgical options, and [4] an explanation that a second procedure using a standard eyelid reconstructive technique might be required to vascularize the donor tissue if the graft became compromised. All patients who were offered the procedure consented to proceed. No cases were withheld from data presentation. Both upper and lower eyelids with malignant tumors qualified for the FBA procedure, and any uninvolved eyelid could be used for the full-thickness donor tissue. The graft was harvested from either the ipsilateral or contralateral side, but from the uninvolved upper or lower eyelid. Written consent was obtained from those patients whose photos appear in this study.
## Procedure set-up
Anesthesiologists used neuroleptic sedation for each patient with a combination of ketamine, midazolam, fentanyl and propofol. The surgeon used loupes with a 3.3X magnification and a headlight. The tumour was assessed, measured (Figure 1A), and marked with standard four millimetre surgical margins for BCC and seven millimeter surgical margins for melanoma in situ (MIS). The width of the excised area was documented. The donor tissue width was estimated and marked (Figure 1B). The donor lid was then stretched horizontally, ensuring that the secondary defect could undergo direct closure. One drop of topical anesthesia was placed in each eye and the operative site was prepared with controlled use of chlorhexidine to limit the risk of corneal toxicity. The surgeon performed subcutaneous infiltration of the tumor and donor sites using lidocaine $2\%$ with epinephrine 1:100,000; 2 ml or less per eyelid. All tissue was handled with 0.5 mm toothed forceps to preserve its architecture and integrity.
**Figure 1:** *A stepwise photo series of the FBA procedure. (A) Preliminary measurement of tumor, (B) donor lid harvested and tumor resected with 4 mm margins, (C) immediate post-procedure repair of graft and donor site, (D) 24 h post-procedure with marked ptosis, (E) 6 days after surgery, (F) 3 months post-procedure with resolved ptosis.*
## Procedure
Tumour excision and reconstruction occurred consecutively on the same day with post-operative pathology assessment. Each tumor was excised using Stevens scissors at the pre-delineated vertical and horizontal margins with clean single incisions through the full-thickness of the eyelid (Figure 1B). The autograft was always taken from the lateral aspect of the lid. Gentle monopolar cautery was carried out. Instruments were not replaced prior to reconstruction. Stevens scissors were used to create full-thickness single, deliberate, straight medial and lateral vertical incisions, joined by a clean horizontal cut to form a rectangular bilamellar composite graft (Figure 1B). The graft was immediately anchored in the recipient site with one 6–0 polyglactin suture (on a P1 needle) through both the donor and host eyelid skin and tarsal plates. 6–0 polyglactin sutures were placed at the recipient site near the eyelid margin of the two vertical interfaces and centrally at the horizontal interface, ensuring proper alignment of the donor and host lid margins under minimal tension. The two vertical lid margins were then approximated with 7–0 polyglactin (on a P1 needle) (Figure 2B), creating a raised interface. The procedure was repeated for the other vertical donor-host interface. The skin was gently approximated with simple interrupted 7–0 polyglactin sutures. The posterior lamella was not manipulated as proper alignment was achieved with the 6–0 polyglactin sutures. Repair of the donor site was performed with a 6–0 polyglactin suture attaching the tarsus to the periosteum of the lateral orbital rim. Skin was closed using 7–0 Polyglactin interrupted sutures (Figure 1C).
**Figure 2:** *Graphic illustration of the suturing technique used to secure the donor eyelid to the host eyelid. (A) The needle was passed: (1) at the surgical incision of the host just anterior to the tarsus but posterior to skin, through the host tarsus, and exiting just posterior to the tarsus but anterior to the palpebral conjunctiva to create a buried suture, then (2) at the surgical incision of the donor just anterior to the palpebral conjunctiva and posterior to the tarsus, through its tarsal plate and exiting just posterior to the eyelid skin. (B) The needle was passed: (1) Through the vertical edge of the host tarsus about 4 mm vertically from the eyelid margin, through the tarsus towards the palpebral fissure, aiming for the grey line of the host margin about 3 mm horizontally from the vertical incision; (2) Through the host grey line aiming back into the host tarsus; (3) Out of the host tarsus and into the donor tarsus about 2 mm vertically from the eyelid margin; (4) Back through the donor grey line of the eyelid margin; and (5) Passed back into the tarsus, coming out at the same level it went in to the host lid; (6) Creating a raised host-donor interface.*
## Postoperative care
Tobramycin/dexamethasone ointment was applied to all suture lines four times a day for the first week and then at bedtime for the second week after surgery. A patch was applied for the first 24 h and a shield was worn at bedtime for the first month following surgery. Patients were followed closely, ideally with appointments at 1 day (Figure 1D with acute ptosis), 1 week (Figure 1E), 1 month, 3 months (Figure 1F), 6 months, and 1 year after the operation. The mean follow-up period was 28 months for this case series (Table 1), and some patients were seen beyond the data collection period with no concerns. Non-identifying pictures were taken of the eyelid at each follow-up to assess for structural integrity, appearance, and viability of the autograft. Phases of healing were documented. Photos were subsequently assessed by all authors.
**Table 1**
| Patient # | Width of pre-excision recipient site (mm) | Width of pre-excision donor lid (mm) | Donor to recipient (±additional technique) | Length of follow up (months) | Complications |
| --- | --- | --- | --- | --- | --- |
| 24 | 15 | 10 | RUL to RLL, S | 11 | - |
| 2 | 16 | 12 | LUL to LLL | 3 | - |
| 6 | 16 | 12 | LUL to LLL | 103 | - |
| 7 | 16 | 12 | RUL to RLL | 17 | - |
| 10 | 16 | 12 | LUL to LLL | 3 | - |
| 17 | 16 | 12 | LLL to RLL | 59 | - |
| 18 | 16 | 12 | LUL to LLL | 67 | BCC recurrence at 2 and 6 years, re-excised. |
| 3 | 17 | 13 | LUL to LLL, C | 55 | - |
| 5 | 17 | 13 | RUL to RLL | 21 | - |
| 8 | 17 | 13 | LUL to LLL | 11 | - |
| 9 | 17 | 12 | LUL to LLL | 7* | - |
| 12 | 17 | 12 | LUL to LLL | 12 | Minor traumatic graft dehiscence, repaired |
| 16 | 17 | 10 | RUL to RLL | 3 | - |
| 20 | 17 | 10 | RUL to LLL | 5 | Minor traumatic graft dehiscence, repaired |
| 1 | 18 | 13 | RUL to RLL | 16 | - |
| 19 | 18 | 12 | RUL to LUL | 34 | - |
| 23 | 18 | 10 | RUL to RLL | 36 | Chalazion |
| 31 | 18 | 10 | LUL to LLL | 14 | Ectropion, repaired |
| 4 | 18.5 | 12 | LUL to LLL | 99 | Distichiasis A new primary BCC lateral to previous BCC at 6 years |
| 11 | 19 | 12 | LUL to LLL, C | 13 | - |
| 14 | 19 | 11 | RUL to RLL, C | 20 | - |
| 15 | 19 | 9 | LUL to LLL | 6 | Chalazion |
| 22 | 20 | 10 | RUL to RLL | 4 | Cyst inferior to autograft |
| 25 | 20 | 12 | RUL to RLL | 34 | Cyst in autograft |
| 27 | 20 | 10 | RUL to RLL, F | 6 | Traumatic donor site dehiscence, no repair required |
| 13 | 22 | 13 | RUL to RLL, S | 24 | - |
| 26 | 22 | 12 | LUL to LLL | 74 | Superficial necrosis secondary to frostbite Ectropion. Mild corneal exposure, required tarsorrhaphy |
| 30 | 24 | 12 | LUL to LLL | 13 | Minor graft dehiscence, repaired |
| 21 | 25 | 11 | LUL to LLL | 58 | Mild ectropion, no repair required |
| 29 | 25 | 11 | RUL to RLL, F | 29 | Minor graft dehiscence, repaired |
| 28 | 27 | 11 | RUL to RLL, F | 13 | Minor graft dehiscence, repaired |
| Mean (±SD) | 18.8 (±3.0) | 11.5 (±1.1) | | 28 (±28) | |
## Pathology
The excised tissue was sent to pathology for analysis after surgery with standard vertical paraffin sections.
## Results
Thirty-one patients were included in the case series: 14 females ($\frac{14}{31}$; $45\%$) and 17 males ($\frac{17}{31}$; $55\%$), with a mean age of 78-years-old (SD ± 9 years) (Table 2). Patients had a variety of comorbidities and 13 were on blood thinners, as indicated in Table 2. The average width of the tumor excision (recipient site pre-excision) was 18.8 mm (SD ± 3.0), ranging from 15 to 27 mm. The average donor site width pre-excision was 11.5 mm (SD ± 1.1 mm), ranging from 9 to 13 mm, and approximately 8–10 mm in height (Table 1). For very large recipient sites in 3 cases, a mini temporal rotation flap was also performed to help fill in the defect. A cantholysis was performed at the recipient site when the lid was under too much tension in an additional 3 cases. Two patients required a canalicular stent based on the location of their tumors. The upper eyelid was used for the ipsilateral lower lid most commonly, but also grafted to the contralateral lower or upper lid. The lower eyelid was grafted to the contralateral lower lid for one patient.
**Table 2**
| Patient # | Age at surgery | Sex | Diagnosis | Medical history | Blood thinners |
| --- | --- | --- | --- | --- | --- |
| 1 | 85 | F | BCC | CHF, arthritis | ASA |
| 2 | 87 | M | BCC | HTN | ASA |
| 3 | 81 | F | BCC | CAD, HTN, asthma, arthritis | ASA, Clopidogrel |
| 4 | 79 | F | BCC | HTN, previous eyelid surgery | |
| 5 | 85 | F | BCC | CAD, hypothyroid, HTN, laser eye surgery, arthritis, HSV, previous eyelid surgery | ASA |
| 6 | 67 | F | BCC | HTN, myotonic dystrophy | ASA |
| 7 | 79 | F | BCC | HTN, diverticulitis, arthritis, asthma/COPD, smoker | |
| 8 | 87 | M | BCC | GPA, COPD, previous eyelid surgery, smoker | |
| 9 | 58 | M | MIS | DM | |
| 10 | 83 | M | BCC | CAD, HTN, DM, asthma/COPD | |
| 11 | 82 | F | BCC | HTN, Parkinson’s | |
| 12 | 61 | M | BCC | Glaucoma | ASA |
| 13 | 79 | F | BCC | Glaucoma, previous eyelid surgery | |
| 14 | 88 | F | BCC | CVA, previous eyelid surgery | |
| 15 | 79 | M | BCC | DM, HTN, previous eyelid surgery, laser eye surgery, retinal detachment, smoker | ASA |
| 16 | 85 | F | BCC | Glaucoma | ASA |
| 17 | 78 | M | BCC | Previous eyelid surgery | |
| 18 | 58 | M | BCC | Charcot-Marie Tooth, cancer, arthritis, previous eyelid surgery | |
| 19 | 84 | F | BCC | HTN | |
| 20 | 87 | F | BCC | HTN, arthritis, DM, HSV, monocular secondary to VZV | ASA |
| 21 | 78 | M | BCC | CAD, HTN, arthritis, thyroid disease, asthma/COPD | |
| 22 | 92 | M | BCC | Arthritis | |
| 23 | 78 | M | BCC | Cancer, HSV, hepatitis | |
| 24 | 79 | F | BCC | CAD, thyroid disease, HSV | |
| 25 | 79 | F | BCC | HTN, kidney disease, arthritis, thyroid disease, previous retinal detachment, HSV | ASA |
| 26 | 66 | M | BCC | Skin cancer, COPD, previous eyelid surgery, smoker, alcohol-use disorder | Clopidogrel, ASA |
| 27 | 75 | M | BCC | CAD, stroke, GERD, kidney stones, smoker | ASA |
| 28 | 60 | M | BCC | CAD, smoker, marijuana use | |
| 29 | 87 | M | BCC | Asthma, DM, hepatitis, arthritis | |
| 30 | 66 | M | BCC | HTN, COPD, CAD, arthritis, smoker, marijuana use | |
| 31 | 73 | M | BCC | HTN, DM | Eliquis, ASA |
| Mean (±SD) | 78 (±9) | | | | |
In the surgical center’s region, it is not protocol for dermatopathologists to provide histopathological margin measurements for BCC. However, the reports stated that the peripheral and deep margins were negative for all patients. The case with MIS had 5 mm margins on histopathology. Radical excision was achieved in all cases.
## FBA eyelid function and complications
The FBA procedure results demonstrated success with respect to structural integrity, appearance, and viability of the full-thickness composite graft. No complications occurred during the procedure. The sutures did not have to be removed and never caused friction between the lid and cornea. All patients were satisfied with the functional and cosmetic result of their surgery. *In* general, complications were more common in patients who had a recipient site larger than 17 mm (Table 1).
A chalazion developed in the autograft of two eyelids post-surgery, demonstrating that its Meibomian glands likely remained functional. Three patients developed mild ectropion, two of which required repair. There was initial ptosis in many patients due to shortening of the donor eyelid’s horizontal length; however, this resolved over weeks. Lashes were generally lost or reduced in number on the grafted eyelid. Two patients developed cysts within or adjacent to the autograft that were inconsequential. Six patients developed minor dehiscence of the autograft, three of which were traumatic in nature due to rubbing the eye, wearing swimming goggles, and using a shield with a string post-operatively. Approximation sutures were required in five of these patients.
One patient had a recurrence of BCC at the recipient site many years after excision (Table 1). A different patient had a second primary BCC develop lateral to the excision site. One patient experienced superficial necrosis in the middle inferior aspect of the graft, after developing frostbite of the face. The graft ultimately healed but led to a mild ectropion. The patient required a temporary tarsorrhaphy due to exposure keratitis. No patients developed postoperative entropion, eyelid notching, distichiasis, trichiasis, lanugo hair irritation, donor lid contraction, dry eyes, or other complications (Table 1).
## Phases of autograft healing and cosmesis
The FBA procedure heals in three observable phases: the white phase, the blue phase, and the pink phase (Figure 3). Using the pictures taken at each follow up appointment and patient feedback, it was approximated that the white phase lasted from the time of operation to between 24 and 48 h post-operation, the blue phase lasted from 24 to 48 h and 7–12 days post-surgery, and the pink stage started around day 7–12 post-surgery with visible blood vessels in the autograft. The pink colour slowly faded over weeks to months.
**Figure 3:** *Healing phases of right-upper to right-lower FBA. (A) Pre-operative markings. (B) White phase immediately post-FBA. (C) Blue phase 4 days post-FBA. (D) Pink phase 1.5 months post-FBA. (E) Final outcome 6 months post-FBA.*
## Discussion
The FBA procedure offers an alternative method for surgeons to include in their armamentarium for treating large, full-thickness eyelid defects. Although previously called a free bilamellar autograft, it might be considered an autologous transplant. The term transplant could be favoured over graft due to the transfer of a full-thickness, unaltered piece of an accessory ocular structure, including skin, muscle, tarsal plate, secretory glands, and mucosal tissue, from one location to another with no enhanced vascular supply.
To the best of our knowledge, there is only one previous case report and one previous case series describing full-thickness free bilamellar autografts that do not require any augmented vascular supply or pedicle [16, 17]. Memarzadeh et al. described a single case in which a successful full-thickness free graft was transferred from the ipsilateral lower eyelid to the upper eyelid defect [17]. Tenland et al. reported a small series of 10 patients who underwent a similar procedure, but always used the contralateral eyelid as the donor [16]. Although surgical technique is not the emphasis in Tenland et al. ’s series, there are many similarities and results are comparable [16]. Tenland et al. reported success in all ten of their cases [16]. One of their cases ($10\%$) developed dehiscence due to excess tension. Three of our cases ($10\%$) experienced mild traumatic dehiscence, and three ($10\%$) developed minimal spontaneous dehiscence. Two of Tenland’s patients ($20\%$) had mild ectropion not requiring repair, and three patients ($10\%$) in our series developed the same, two of which required repair. Three of Tenland’s recipient sites ($30\%$) required a rotational flap to reduce tension; another 5 ($50\%$) donor sites required the same. Only 3 of our patients ($10\%$) required a rotational flap, but an additional 3 ($10\%$) required cantholysis to reduce tension. In all cases under tension in our series, the donor and recipient sites were ipsilateral. Tenland used a pentagonal excisional approach, whereas all our grafts were rectangular. The authors of both the previously published case report and series recommended that a larger series would be beneficial in supporting this procedure [16, 17]. This case series was started prior to the Tenland case series and conducted independently, resulting in the subtle differences in surgical technique [16].
Current standards for reconstruction are constrained by a necessitated vascular supply, often requiring multiple surgeries [6]. This is of particular concern in monocular patients, or pediatric patients who could develop amblyopia. The FBA is a single-step procedure that minimizes operating room time and post-procedural patient discomfort. Functional outcomes of currently used procedures can be hampered by complications including madarosis, ptosis, superior eyelid entropion, irregularity of the lid margin, inferior eyelid cicatrisation and retraction, as well as necrosis of the flap [6, 18]. Madarosis was common in the FBA procedure, ectropion infrequently occurred, and one patient developed superficial necrosis secondary to frostbite. Although shortening of the horizontal lid length caused ptosis of the donor lid was observed initially, this resolved over a few weeks in all cases [19].
The Meibomian gland function is often lost with classic eyelid reconstruction techniques. This is known to contribute to ocular surface disease [20]. Two patients in this case series developed a chalazion within their grafted tissue, suggesting that Meibomian gland function may be preserved following the FBA procedure, theoretically resulting in a more stable tear film [21]. This is further supported by no patients developing dry eyes. One patient had recurrence of their BCC years later, resulting in a recurrence rate of $3.2\%$. This is consistent with the known recurrence rate of $2\%$–$5\%$ in radical resections [22]. The periocular area is an area of higher risk for BCC recurrence [23]. Therefore, it may also be compared to the $2.9\%$ recurrence rate in a 2020 meta-analysis reviewing recurrences rates of periocular BCC following excision with Mohs micrographic surgery [24]. Although iatrogenic implantation has been raised as a potential concern if instruments are not replaced after the extirpative phase of a cancer surgery, there is no literature to suggest that instruments used during margin-negative resection on BCC can cause tumour seeding or recurrence [25]. Furthermore, the primary and secondary defects are often in proximity for this procedure, rendering complete separation impossible.
Patients who underwent the FBA procedure had a range of comorbidities including diabetes and tobacco use, which can be associated with poor wound healing and flap necrosis. [ 26]5. Smoking cessation prior to the procedure would be ideal. However, the presence of such comorbidities affecting microcirculation in patients with satisfactory FBA results supports that the procedure is robust.
The overall success of the FBA procedure suggests that an enhanced vascular supply is not always necessary when an intact, unaltered, bilamellar, full-thickness eyelid donor is auto-grafted. The eyelid’s naturally rich vascular supply provides the potential for plasmatic imbibition, inosculation, and revascularization of the arterial arcade. The procedure allows for the survival of all aspects of the organ except for the cilia.
## Phases of healing
The autograft tissue followed the same healing processes as previously described for skin transplants and composite grafts. The white phase demonstrates ischemia induced by surgery and injected epinephrine. Plasmatic imbibition ensues, wherein the FBA absorbs nutrients and oxygen from the host interface [27, 28]. The blue phase reveals the beginning of inosculation with an arterial-venous mismatch leading to mild venous congestion and a dark blue-purple discolouration. The pink phase demonstrates angiogenesis secondary to the eyelid’s rich vascular supply, with reliable arterial perfusion and venous drainage (Figure 2) [29]. In the final stages of healing there is epidermal proliferation.
These phases of healing may be further supported by the underlying tear film, which provides lubrication, antimicrobial support, promotes wound healing, suppresses inflammation, and scavenges free radicals [30]. The aqueous-mucin tear layer contains growth and supportive factors, some of which participate in epithelial or stromal wound healing as well as angiogenesis (e.g., vascular endothelial growth factor promoting blood vessel formation from pre-existing vessels) and neovascularization (e.g., epidermal growth factor; EGF; promoting de novo formation of blood vessels in addition to connection from pre-existing) [31]. Platelet-derived growth factors and transforming growth factor-beta potentiate tissue repair in vivo and EGF is a known promotor of adipose graft survival as a potent stimulator of neovascularization [32, 33].
Cold outdoor temperatures and physical trauma (e.g., swimming goggles) in the early phases of healing caused mild dehiscence and early signs of superficial necrosis, respectively (Table 1). There is likely an upper limit to the width of recipient site being repaired, which could not be determined in this study. While ipsilateral eyelids were used in this study as donor tissue, it may be wise to preserve this tissue and use contralateral lids in the event that a bridging flap is required later on. A potential limitation that was not encountered in this study is irradiated tissue. If radiation is being used in the eyelid region, the FBA procedure should not be considered, as it could be detrimental to flap survival [34].
The pathology was assessed with standard vertical paraffin sections, in which only $1\%$–$2\%$ of the margin can be evaluated by a pathologist [24]. Intraoperative horizontal (en face) frozen sections would allow for complete margin assessment [24], with the potential to further reduce recurrence rates.
This study is limited by the fact that it is a retrospective case series without a comparison group. There was no specific grading scale used to document the outcomes for each case; however, they were all assessed by one surgeon for consistency and photos were assessed by all authors to improve objectivity. Unfortunately, the height of the recipient and donor sites were not consistently measured and therefore cannot be specifically addressed, but were approximately all in the range of 8–10 mm. Finally, the number of patients in the case series is reasonable and a notable contribution to the current literature, but still too small to draw definitive conclusions regarding rates of complications.
## Conclusion
The FBA procedure was introduced in 2021 by Tenland et al. as a simple and efficient method of reconstructing large, full-thickness eyelid defects [16]. This case series adds to the literature and demonstrates that the procedure is robust and patients with a wide range of comorbidities are eligible. The ultimate structural and cosmetic results were almost universally successful across the 31 patients in this case series. Consequently, the FBA procedure may be considered as a reconstructive technique for large full-thickness eyelid defects in select patients.
## Data availability statement
The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding author.
## Ethics statement
Ethical review and approval was not required for the study on human participants in accordance with the local legislation and institutional requirements. Written informed consent for participation was not required for this study in accordance with the national legislation and the institutional requirements. Written informed consent was obtained from the individual(s) for the publication of any potentially identifiable images or data included in this article.
## Author contributions
HMM performed data collection, data analysis, and manuscript writing. KM created the study, performed data collection, data analysis, and manuscript writing. HM formulated the surgical technique, performed all surgeries, and contributed to data analysis and manuscript editing. All authors contributed to the article and approved the submitted version.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher's note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## References
1. Mustardé JC. **Major reconstruction of the eyelids: functional and aesthetic considerations**. *Clin Plast Surg* (1981) **8** 227-36. DOI: 10.1016/S0094-1298(20)30449-1
2. Deprez M, Uffer S. **Clinicopathological features of eyelid skin tumors. A retrospective study of 5504 cases and review of literature**. *Am J Dermatopathol* (2009) **31** 256-62. DOI: 10.1097/DAD.0b013e3181961861
3. Furdova A, Kapitanova K, Kollarova A, Sekac J. **Periocular basal cell carcinoma - clinical perspectives**. *Oncol Rev* (2020) **14** 420. DOI: 10.4081/oncol.2020.420
4. Porfiris E, Christopoulos A, Sandris P, Georgiou P, Ioannidis A, Popa CV. **Upper eyelid orbicularis oculi flap with tarsoconjunctival island for reconstruction of full-thickness lower lid defects**. *Plast Reconstr Surg* (1998) **103** 186-91. DOI: 10.1097/00006534-199901000-00030
5. Jackson IT. **Reconstruction of the lower eyelid defect in treacher collins syndrome**. *Plast Reconstr Surg* (1981) **67** 365-8. DOI: 10.1097/00006534-198103000-00017
6. Bernardini F, Skippen B. *Surgical ophthalmic oncology: A collaborate open access reference* (2019)
7. Lee WW, Erickson BP, Ko MJ, Liao SD, Neff A. **Advanced single-stage eyelid reconstruction: anatomy and techniques**. *Dermatol Surg* (2014) **40** S103-12. DOI: 10.1097/DSS.0000000000000117
8. Callahan A. *Reconstructive surgery of the eyelids and ocular Adnexa* (1966)
9. Callahan A. **The free composite lid graft**. *AMA Arch Ophthalmol* (1951) **45** 539-45. DOI: 10.1001/archopht.1951.01700010551006
10. Putterman AM. **Viable composite grafting in eyelid reconstruction**. *Am J Ophthalmol* (1978) **85** 237e41. DOI: 10.1016/s0002-9394(14)75955-0
11. Cannon PS, Madge SN, Kakizaki H, Selva D. **Composite grafts in eyelid reconstruction: the complications and outcomes**. *Br J Ophthalmol* (2011) **95** 1268-71. DOI: 10.1136/bjo.2009.170548
12. Paridaens D, van den Bosch WA. **Orbicularis muscle advancement flap combined with free posterior and anterior lamellar grafts: a 1-stage sandwich technique for eyelid reconstruction**. *Ophthalmology* (2008) **115** 189-94. DOI: 10.1016/j.ophtha.2007.03.022
13. Lowry JC, Bartley GB, Garrity JA. **The role of second-intention healing in periocular reconstruction**. *Ophthalmic Plast Reconstr Surg* (1997) **13** 174-88. DOI: 10.1097/00002341-199709000-00004
14. Smith B, Cherubini T. *Oculoplastic surgery* (1970) 30-2
15. Fox SA. **Autogenous free full-thickness eyelid grafts**. *Am J Ophthalmol* (1969) **67** 941-5. DOI: 10.1016/0002-9394(69)90090-7
16. Tenland K, Berggren J, Engelsberg K, Bohman E, Dahlstrand U, Castelo N. **Successful free bilamellar eyelid grafts for the repair of upper and lower eyelid defects in patients and laser speckle contrast imaging of revascularization**. *Ophthalmic Plast Reconstr Surg* (2021) **37** 168-72. DOI: 10.1097/IOP.0000000000001724
17. Memarzadeh K, Engelsberg K, Sheikh R, Malmsjö M. **Large eyelid defect repair using a free full-thickness eyelid graft**. *Plast Reconstr Surg Glob Open* (2017) **5** e1413. DOI: 10.1097/GOX.0000000000001413
18. Rahmi D, Mehmet B, Ceyda B, Sibel O. **Management of the large upper eyelid defects with cutler-beard flap**. *J Ophthalmol* (2014) **2014** 424567. DOI: 10.1155/2014/424567
19. Thaller VT, Then KY, Luhishi E. **Spontaneous eyelid expansion after full-thickness eyelid resection and direct closure**. *Br J Ophthalmol* (2001) **85** 1450-4. DOI: 10.1136/bjo.85.12.1450
20. Klein-Theyer A, Horwath-Winter J, Dieter FR, Haller-Schober EM, Riedl R, Boldin I. **Evaluation of ocular surface and tear film function following modified hughes tarsoconjunctival flap procedure**. *Acta Ophthalmol* (2014) **92** 286-90. DOI: 10.1111/aos.12034
21. Nichols KK, Foulks GN, Bron AJ, Glasgow BJ, Dogru M, Tsubota K. **The international workshop on meibomian gland dysfunction: executive summary**. *Invest Ophthalmol Vis Sci* (2011) **52** 1922-9. DOI: 10.1167/iovs.10-6997a
22. Iljin A, Zieliński T, Antoszewski B, Sporny S. **Clinicopathological analysis of recurrent basal cell carcinoma of the eyelid**. *Postepy Dermatol Alergol* (2016) **33** 42-6. DOI: 10.5114/pdia.2015.48039
23. Armstrong LTD, Magnusson MR, Guppy MPB. **Risk factors for recurrence of facial basal cell carcinoma after surgical excision: a follow-up analysis**. *J Plast Reconstr Aesthet Surg* (2017) **70** 1738-45. DOI: 10.1016/j.bjps.2017.04.006
24. Phan K, Oh LJ, Goyal S, Rutherford T, Yazdabadi A. **Recurrence rates following surgical excision of periocular basal cell carcinomas: systematic review and meta-analysis**. *J Dermatology Treat* (2020) **31** 597-601. DOI: 10.1080/09546634.2019.1698702
25. Berger-Richardson D, Xu RS, Gladdy RA, McCart JA, Govindarajan A, Swallow CJ. **Glove and instrument changing to prevent tumour seeding in cancer surgery: a survey of surgeons’ beliefs and practices**. *Curr Oncol* (2018) **25** e200-8. DOI: 10.3747/co.25.3924
26. Rees TD, Liverett DM, Guy CL. **The effect of cigarette smoking on skin-flap survival in the face lift patient**. *Plast Reconstr Surg* (1984) **73** 911-5. DOI: 10.1097/00006534-198406000-00009
27. Lindenblatt N, Calcagni M, Contaldo C, Menger MD, Giovanoli P, Vollmar B. **A new model for studying the revascularization of skin grafts in vivo: the role of angiogenesis**. *Plast Reconstr Surg* (2008) **122** 1669-80. DOI: 10.1097/PRS.0b013e31818cbeb1
28. Ray S, Rao K. *Full-thickness skin grafts* (2011)
29. Folkman J. **How is blood vessel growth regulated in normal and neoplastic tissue? G.H.A. clowes memorial award lecture**. *Cancer Res* (1986) **46** 467-73. DOI: 10.1158/0008-5472.CAN-16-1123
30. Pflugfelder SC, Stern ME. **Biological functions of tear film**. *Exp Eye Res* (2020) **197** 108115. DOI: 10.1016/j.exer.2020.108115
31. Klenkler B, Sheardown H, Jones L. **Growth factors in the tear film: role in tissue maintenance, wound healing, and ocular pathology**. *Ocul Surf* (2007) **5** 228-39. DOI: 10.1016/S1542-0124(12)70613-4
32. Park B, Kong JS, Kang S, Kim YW. **The effect of epidermal growth factor on autogenous fat graft**. *Aesthetic Plast Surg* (2011) **35** 738-44. DOI: 10.1007/s00266-011-9679-y
33. Pierce GF, Mustoe TA, Lingelbach J, Masakowski VR, Griffin GL, Senior RM. **Platelet-derived growth factor and transforming growth factor-beta enhance tissue repair activities by unique mechanisms**. *J Cell Biol* (1989) **109** 429-40. DOI: 10.1083/jcb.109.1.429
34. Patterson T, Berry R, Hopewell J, Wiernick G, WC G, MB M. **The effect of radiation in survival of experimental skin flaps**. *Skin flaps* (1975)
|
---
title: 'Global trends in the incidence rates of MDR and XDR tuberculosis: Findings
from the global burden of disease study 2019'
authors:
- Qingting Bu
- Rong Qiang
- Lingyan Fang
- Xiaokang Peng
- Hua Zhang
- Hua Cheng
journal: Frontiers in Pharmacology
year: 2023
pmcid: PMC9998482
doi: 10.3389/fphar.2023.1156249
license: CC BY 4.0
---
# Global trends in the incidence rates of MDR and XDR tuberculosis: Findings from the global burden of disease study 2019
## Abstract
Purpose: The study aimed to quantify the global trends of the incidence rates of multidrug-resistant (MDR) tuberculosis (MDR-TB) and extensively drug-resistant (XDR) tuberculosis (XDR-TB).
Methods: Cases, age-standardized rates (ASRs), and incidence rates of MDR-TB and XDR-TB during 2010–2019 were obtained from the Global Burden of Disease Study 2019. The incidence trends of MDR-TB and XDR-TB were evaluated using the estimated annual percentage changes (EAPCs) in ASRs. The relationships among the ASRs of MDR-TB and XDR-TB, the MDR rate, the XDR rate, and socio-demographic index (SDI) were assessed using locally weighted regression and Pearson’s correlation coefficient.
Results: The global ASR of MDR-TB on average decreased by $1.36\%$ (EAPC = −1.36, $95\%$ confidence interval [CI] = −2.19 to −0.52) per year whereas that of XDR-TB was stable (EAPC = 0.69, $95\%$ CI = −0.15–1.54) during 2010–2019. The incidence trends of MDR-TB in most regions and countries were decreasing, but those of XDR-TB were increasing. People aged 35–44 and 55–64 years had the highest incidence rates for MDR-TB and XDR-TB. The MDR and XDR rates both peaked in those aged 35–44 years. Areas with higher SDI tended to have lower ASRs of MDR-TB ($p \leq 0.001$, ρ = −0.43).
Conclusion: The current achievements for the incidence trends of MDR-TB and XDR-TB are insufficient. More strategies and tools need to be developed to further curb MDR-TB and XDR-TB, especially in high-risk areas and age groups, and in low SDI regions.
## Introduction
Tuberculosis is a chronic infectious disease seriously endangering human health that has become a major global public health and social problem (Kazempour Dizaji et al., 2018; World Health Organization, 2021a), with 1.3 million deaths due to TB in 2020 alone (World Health Organization, 2021a). One of the main reasons is that the drug resistance of TB continues to evolve. Standard treatment involving the two most-effective drugs (isoniazid and rifampicin) can achieve excellent cure rates for drug-sensitive patients with TB (Seung et al., 2015). In the treatment of drug-resistant TB, that of multidrug-resistant is very difficult since MDR-TB is resistant to the two most-effective first-line anti-TB drugs (isoniazid and rifampicin) (Trisakul et al., 2022). Nevertheless, extensively drug-resistant (XDR) TB, as a kind of MDR, is more concerning, and is resistant to isoniazid and rifampicin as well as all fluoroquinolone and second-line injectable drugs (World Health Organization, 2018; Lin et al., 2022; Trisakul et al., 2022). The overall cure rates of MDR-TB and XDR-TB were only $56\%$ and $39\%$, respectively (World Health Organization, 2018). For more than a decade, the proportion of MDR and rifampicin-resistant patients diagnosed with TB for the first time has remained around $3\%$–$4\%$, and that of patients previously treated for TB has remained at $18\%$–$21\%$. There are even countries with proportions of previously treated MDR-TB cases exceeding $50\%$ (World Health Organization, 2021a). The proportion of XDR-TB in TB is rarely reported.
According to the End TB Strategy of the World Health Organization (WHO) and the UN Sustainable Development Goals (United Nations, 2015; World Health Organization, 2015), global TB deaths must be reduced by $95\%$ in 2035 compared with 2015. With the current data, this goal is difficult to achieve (World Health Organization, 2021a), and so it is time for urgent action to end the global TB epidemic (Pan et al., 2020a; World Health Organization, 2021a). MDR-TB and XDR-TB increase the risk of death in patients with TB and hinder the achievement of the above goal. Studying the global incidence trends of MDR-TB and XDR-TB is helpful for preventing and treating TB, and thereby reducing deaths from TB. However, there has been no systematic summary addressing this issue. The purpose of this study was therefore to determine the global incidence trends of MDR-TB and XDR-TB using the Global Burden of Disease Study (GBD) 2019 data.
## Data source
Data sources for TB within the GBD 2019 data can be explored using the online GBD Results Tool (https://vizhub.healthdata.org/gbd-results/). The ICD-10 codes for TB are A10–A19.9, B90–B90.9, K67.3, K93.0, M49.0, and P37.0, while the ICD 9 codes are 010–019.9, 137–137.9, 138.0, 138.9, 139.9, 320.4, and 730.4–730.6. The GBD Results *Tool is* a data set developed and supported by the Institute for Health Metrics and Evaluation, which is an independent global health research center based at the University of Washington. This database provides epidemiological information on 369 diseases and injuries during 1990–2019 for 23 age groups; for males, females, and both sexes combined; and for 204 countries and territories that were grouped into 21 regions and 7 superregions. Previous studies have described the method of estimating TB incidence from the GBD database in detail (GBD 2019 Diseases and Injuries Collaborators, 2020; GBD 2019 Risk Factors Collaborators, 2020). Briefly, the TB data were derived from population-based surveys on tuberculin and cohort studies that examined the risk of developing active TB disease as a function of induration size. An updated systematic review was performed on the GBD 2019 which included routine surveillance and surveys reported to the WHO and the risk of MDR-TB (Mesfin et al., 2014; GBD 2019 Diseases and Injuries Collaborators, 2020). From the GBD 2019 database, we extracted the age-related number of cases and age-standardized rates (ASRs) or incidence rates during 2010–2019 globally among 5 socio-demographic index (SDI) regions, 21 geographical regions, and 204 countries and territories. The rates expressed as age-standardised are based on the GBD reference population (GBD 2017 Mortality Collaborators, 2018). In the GBD, the range of data point estimates is not expressed using $95\%$ confidence intervals (CIs), but instead using $95\%$ uncertainty intervals (UIs). Every estimate was calculated 1,000 times, and then the $95\%$ UI was determined by the 25th and 97fifth value of the 1,000 values after ordering them from smallest to largest (Bu et al., 2022). We also extracted the SDI of each country and region. SDI is a compound measure of income, average years of schooling, and the fertility in each location and year in the GBD database that is used to measure socio-demographic development (Pan et al., 2020b). It is the geometric mean of the 0 to 1 index of total fertility rate under 25 years of age, average education level of the population aged 15 and over, and lagging income per capita (GBD 2017 Disease and Injury Incidence and Prevalence Collaborators, 2018). The location with an SDI of 0 will have a theoretical minimum level of development related to health, while the location with an SDI of 1 will have a theoretical maximum level of development. For GBD 2019, the values of SDI were multiplied by 100 on a scale of 0–100 (GBD 2019 Diseases and Injuries Collaborators, 2020). It is divided into five levels: high, middle-high, middle, low-middle, and low.
## Statistical analysis
Estimated annual percentage changes (EAPCs) of incidence rates were used to evaluate the incidence trends during 2010–2019. EAPC is a summarizing and widely used measure that assesses ASR trends over a specified time period (Hankey et al., 2000; Sun et al., 2022). Natural logarithm of regression line fitting rates were used; that is, y = a + βx + e, where y = ln (ASR) and x is the calendar year. EAPC was calculated as 100×[exp(β)–1], and its $95\%$ CI was also obtained from the linear regression model. If EAPCs and the lower limit of the $95\%$ CI are both > 0, then ASR is considered to have an increasing trend. In contrast, if both EAPC estimation and the upper limit of the $95\%$ CI are < 0, ASR has a downward trend. For other values ASR is considered stable. We also assessed the relationships among ASR of MDR-TB and XDR-TB, MDR and XDR rates, and SDI using locally weighted regression and Pearson’s correlation coefficient. The MDR and XDR rates are the ratios of new MDR-TB and XDR-TB cases to new TB cases, respectively. The $p \leq 0.05$ was considered significant. R software (version 3.4.3) was used for the statistical analysis.
## Multidrug-resistant tuberculosis
Globally in 2019, the ASR of MDR-TB was 5.63 ($95\%$ UI = 3.12–9.73) per 100,000 among 450,600 cases ($95\%$ UI = 247,830–785,370), and the MDR incidence rate was $5.30\%$ (Table 1). The distribution of cases during 2010–2019 was almost U-shaped (Figure 1A). The ASR decreased on average by $1.36\%$ (EAPC = −1.36, $95\%$ CI = −2.19 to −0.52) per year during 2010–2019 (Table 1).
For SDI regions, the ASR exhibited a stable trend in the low and low-middle SDI regions and decreased in the other three SDI regions (Table 1). The high SDI regions had the fewest cases, and the lowest ASR and MDR rates (Table 1; Figure 1). The number of MDR-TB cases increased monotonically in the low- and low-middle SDI regions (Figure 1).
The ASR of MDR-TB in 12 of the 21 geographical regions exhibited decreasing trends during 2010–2019, with the largest decrease observed in East Asia (EAPC = −6.8, $95\%$ CI = −9.18 to −4.36), followed by Central Asia and Eastern Europe (Table 1). However, there were still two regions with stable ASR, and even seven with increased ASR (Table 1). Oceania had the largest increase (EAPC = 11.09, $95\%$ CI = 5.12–17.4). Central Asia and Eastern Europe had the highest MDR rates, at $20.51\%$ and $27.18\%$, respectively (Table 1).
The incidence trend of MDR-TB varied among the 204 countries and territories, decreasing in 116 of them, remaining stable in 35, and increasing in 53 (Supplementary Table S1). Countries with high MDR rates were mostly in Eastern Europe and Central Asia, which was consistent with the analysis at the regional level. For example, the ten regions with the highest MDR rates (in decreasing order) were the Republic of Moldova (MDR rate = $37.87\%$), Belarus ($36.91\%$), Ukraine ($29.52\%$), Russian Federation ($26.14\%$), Kyrgyzstan ($25.46\%$), Uzbekistan ($23.92\%$), Kazakhstan ($20.60\%$), Azerbaijan ($20.35\%$), Georgia ($18.12\%$), and Estonia ($17.58\%$) (Supplementary Table S1).
## Extensively drug-resistant tuberculosis
In 2019, there were 25,060 ($95\%$ UI = 17,090–36,470) new XDR-TB cases globally, which had increased by $22.5\%$ compared with 2010, and the ASR was 0.31 ($95\%$ UI = 0.21–0.45) per 100,000 (Table 1; Figure 2A). The XDR rate was $0.29\%$. The ASR was stable during 2010–2019 (EAPC = 0.69, $95\%$ CI = −0.15–1.54) (Table 1).
**FIGURE 2:** *The cases of XDR-TB from 2010 to 2019. (A) Global, (B) low SDI regions, (C) low-middle SDI regions, (D) middle SDI regions, (E) middle-high SDI regions and (F) high SDI regions. Abbreviations: XDR, extensively drug-resistant; TB, tuberculosis; SDI, socio-demographic index.*
For SDI regions, the ASR was only stable in the middle-high SDI regions (Table 1). The ASRs and numbers of cases increased in the other four SDI regions (Table 1; Figure 2). The middle-high SDI had the most XDR-TB cases, and the highest ASR and XDR rates (Table 1; Figure 2). The high SDI region had the fewest XDR-TB cases and lowest ASR rate (Table 1; Figure 2).
The ASRs increased in 16 of the 21 geographical regions, was stable in 4, and decreased only in Central Asia. The increase was largest in Oceania (EAPC = 16.14, $95\%$ CI = 9.49–23.19), followed by the Caribbean (EAPC = 9.92, $95\%$ CI = 7.86–12.03) and Australasia (EAPC = 9.44, $95\%$ CI = 7.53–11.39) (Table 1). Although the ASR decreased in Central Asia, its XDR rate was the second highest (XDR rate = $4.49\%$), and that of Eastern Europe was the highest (XDR rate = $5.96\%$).
The trend of ASR varied among the 204 countries and territories. The ASRs of most countries and territories (144 of 204) increased, while those of 24 were stable and it decreased in 36 (Supplementary Table S1). There was a close correspondence between countries with high MDR rates and high XDR rates; for example, the ten countries with the highest XDR rates also had the ten highest MDR rates (Supplementary Table S1).
## Age distributions of MDR-TB and XDR-TB incidence rates, and MDR and XDR rates
The age distributions of the MDR and XDR incidence rates were similar, with both having two peaks. The MDR-TB incidence rate peaked in those aged 35–44 and 55–64 years. The XDR-TB incidence rates were similar, also peaking in those aged 35–44 and 55–64 years (Figures 3A, B). The MDR and XDR rates both peaked in those aged 35–44 years (Figures 3C, D).
**FIGURE 3:** *Age distribution of MDR-TB incidence rate (A), XDR-TB incidence rate (B), MDR rate (C), and XDR rate (D). Abbreviations: MDR, multidrug-resistant; TB, tuberculosis; XDR, extensively drug-resistant.*
## Relationships among the ASRs of MDR-TB and XDR-TB, MDR and XDR rates, and SDI
We analyzed the relationships among the ASRs of MDR-TB and XDR-TB, MDR and XDR rates, and SDI based on national-level data. A significant negative correlation was found between the ASR of MDR-TB and SDI ($p \leq 0.001$, ρ = −0.43) (Figure 4A). No significant relationship was found between the ASR of XDR-TB ($$p \leq 0.54$$, ρ = 0.04) or the MDR incidence rate ($$p \leq 0.86$$, ρ = 0.01) and SDI (Figures 4B, C). A significant positive correlation was found overall between the XDR rate and SDI ($$p \leq 0.03$$, ρ = 0.15) (Figure 4D). However, as shown in Figure 4D, there was a negative relationship between them when the SDI exceeded about 0.75.
**FIGURE 4:** *The correlation between ASR of MDR-TB (A), ASR of XDR-TB (B), MDR rate (C), and XDR rate (D) and SDI. Each circle represents a country or territory. The size of the circle represents the number of cases. Abbreviations: ASR, age-standardized rate; MDR, multidrug-resistant; TB, tuberculosis; XDR, extensively drug-resistant; SDI, socio-demographic index.*
## Discussion
MDR-TB and XDR-TB are serious problems that represent great threats and challenges to human and public health (Akkerman et al., 2019; Borisov et al., 2019; Shang et al., 2022). According to the End TB Strategy, TB incidence and mortality should have declined by at least $20\%$ and $35\%$, respectively, between 2015 and 2020. However, the performance of the strategy has been suboptimal, with only $11\%$ and $9.2\%$ declines in TB incidence and mortality, respectively, by 2021 (Jeremiah et al., 2022). MDR-TB and XDR-TB played important roles in this poor performance (Seung et al., 2015; Jeremiah et al., 2022). In the present study, we analyzed the global incidence trends of MDR-TB and XDR-TB during 1990–2019 to help improve the current status of TB based on the GBD 2019 database.
The analyzed GBD database contains data from 1990 to 2019. We were more concerned about the current trend than the previous trend, and so this study focused on the data from the last 10-year period covered in the GBD. Previous data may affect the actual recent trends. For example, if the EAPC during 1990–2009 was significantly negative and that during 2010–2019 was significantly positive, it is possible that the EAPC during 1990–2019 would be significantly negative. Although this study found that the ASR of MDR-TB worldwide is declining, the current annual reduction in global TB incidence is $2\%$, which is too slow to achieve an end to the epidemic in the foreseeable future. According to the End TB Strategy (Uplekar et al., 2015; World Health Organization, 2019), the annual decline in global TB incidence rates must increase to $10\%$ annually by 2025. However, the EAPC of MDR-TB was −1.36, meaning that the ASR of MDR-TB decreased by $1.36\%$ per year, which is far less than $10\%$ or even $2\%$. This study found that the high ASR of MDR-TB was mostly attributable to Eastern Europe, South Asia, Southern Sub-Saharan Africa, Central Asia, Eastern Sub-Saharan Africa, and Central Sub-Saharan Africa. There are several possible reasons for the higher ASR of MDR-TB in these regions. The economic level is low and the accessibility to public health services is poor in these regions, and it includes many developing countries, which may have problems such as poverty, malnutrition, and poor living conditions (Lange et al., 2018). Our results also indicated that the ASR of MDR-TB had a significant negative correlation with SDI. SDI is a composite measure of income per capita, total fertility rate (age <25 years), and average education level (for those aged ≥15 years), and is used as a measure of sociodemographic development (GBD 2019 Risk Factors Collaborators, 2020; GBD 2019 Diabetes in the Americas Collaborators, 2022). As is well known, the incidence rate of TB is related to the socioeconomic and development levels (Nordholm et al., 2022; Soares et al., 2022). This affects the incidence rate of TB in various aspects. For example, public health systems are imperfect and the prevention and control of infectious diseases is weak in regions with low SDI. Both in terms of resource allocation and professional talent training, there are problems such as insufficient quantity, low quality, and unreasonable structure (Liu et al., 2019), which lead to an increase in the incidence rate of TB that can cause an increase in the incidence rate of MDR-TB. According to our analysis, the impact of social development level on the incidence rate of MDR-TB may be mostly attributed to its impact on the incidence rate of TB, rather than directly on that of MDR. We did not observe a significant correlation between MDR rate and SDI in this study. Eastern Europe and Central Asia had the highest MDR rates, which was consistent with previous reports (Dirlikov et al., 2015; Lange et al., 2018). It is promising that these regions had larger downward trends for the ASR of their MDR-TB compared with most regions. According to our results, XDR-TB should be considered because its ASR had no downward trend and actually increased in most regions. The trend was only declining in Central Asia. Regions with high MDR rate tend to have a high XDR rate, which was consistent with the principle of drug resistance in biology. There are two types of drug resistance in Mycobacterium tuberculosis: genetic and phenotypic resistance. Genetic drug resistance is caused by mutations in chromosomal genes in bacterial growth, while phenotypic resistance or drug tolerance is caused by epigenetic changes in gene expression and protein modification that induce drug tolerance in non-growing bacterial persisters (Zhang and Yew, 2015). These two types are mostly caused by drug use. A high MDR rate may result in a high XDR rate by increasing the use of second-line drugs. In the present study, XDR-TB and MDR-TB had different relationships with SDI; that is, the XDR rate had a significant positive correlation with SDI, but the ASR of XDR-TB was not significantly correlated with SDI. Although the relationship between XDR rate and SDI was significant, it was not strong, with a ρ value of only 0.15. The ρ value represents the strength of the correlation in Pearson’s coefficient (Pearson, 1920; Rodgers and Nicewander, 1988). This may mean that the XDR rate was more affected by other factors. We also analyzed the age distribution of MDR-TB and XDR-TB. The results indicated that there were two peaks for the incidence rates of MDR-TB and XDR-TB, in those aged 35–44 and 55–64 years. Nevertheless, there were also peaks for MDR and XDR rates, in those aged 35–44 years. The peaks for MDR and XDR rates were consistent with the first peaks of the incidence rates of MDR-TB and XDR-TB, which was logical since a high drug resistance leads to a high incidence rate in drug-resistant TB. A reasonable explanation for the absence of second peaks for MDR and XDR rates is that the mortality rate of TB is high among the elderly (Dhamnetiya et al., 2021), resulting in a small proportion of elderly patients having received previous treatment for TB. Drug-resistant TB mostly occurs in patients previously treated for TB (World Health Organization, 2021a). At the national level, India, China, and the Russian Federation are the countries with the three largest numbers of MDR-TB and XDR-TB cases, which account for most new cases in the world. This result for MDR-TB was consistent with a WHO report (World Health Organization, 2021b). WHO do not report the global incidence of XDR-TB, which is rarely reported. The incidence trends of MDR-TB and XDR-TB in China were declining. The incidence trend of MDR-TB in the Russian Federation was declining, while that of XDR was stable. India should receive more attention, because it has the most MDR-TB cases with a stable incidence trend and the second-highest rate of XDR-TB cases with an increasing incidence trend. Improving the incidence trends of MDR-TB and XDR-TB in *India is* important to improve control of the global incidence rates of MDR-TB and XDR-TB.
This study had several limitations, most notably being that the participants were from the GBD database and calculations were made using a model based on existing data in each country; that is, where data were not available, the results depended on predictive validity of the model for out-of-sample data. In addition, the MDR or XDR rate was the ratio of new MDR- or XDR-TB cases to new TB cases. Cases were point estimates, and their $95\%$ UIs were determined through 1,000 calculations. This approach made it impossible to estimate the UIs or CIs of MDR and XDR rates.
The present study has performed the most comprehensive analysis of the global trends of MDR-TB and XDR-TB during 2010–2019. Although the incidence of MDR-TB was declining, the rate of decline was too slow; moreover, the incidence trend of XDR-TB was not declining. The incidence trends of MDR-TB and XDR-TB varied markedly among different regions and countries. High-risk age groups, regions and countries with high burdens, and low-SDI regions require careful consideration, and effective tools need to be developed to curb MDR-TB and XDR-TB.
## Data availability statement
The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding author.
## Ethics statement
This study was performed in accordance with the Declaration of Helsinki and was approved by the institutional review board of Xi’an Children’s Hospital.
## Author contributions
Study design and data extraction: QB and HC; Statistical analysis: QB; Manuscript draft: QB, RQ, LF, XP, HZ, and HC; Charts and tables: QB and LF. All authors agreed to submit the final version of this 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/fphar.2023.1156249/full#supplementary-material
## References
1. Akkerman O., Aleksa A., Alffenaar J. W., Al-Marzouqi N. H., Arias-Guillén M., Belilovski E.. **Surveillance of adverse events in the treatment of drug-resistant tuberculosis: A global feasibility study**. *Int. J. Infect. Dis.* (2019) **83** 72-76. DOI: 10.1016/j.ijid.2019.03.036
2. Borisov S., Danila E., Maryandyshev A., Dalcolmo M., Miliauskas S., Kuksa L.. **Surveillance of adverse events in the treatment of drug-resistant tuberculosis: First global report**. *Eur. Respir. J.* (2019) **54** 1901522. DOI: 10.1183/13993003.01522-2019
3. Bu Q., Qiang R., Cheng H., Wang A., Chen H., Pan Z.. **Analysis of the global disease burden of down syndrome using YLDs, YLLs, and DALYs based on the global burden of disease 2019 data**. *Front. Pediatr.* (2022) **10** 882722. DOI: 10.3389/fped.2022.882722
4. Dhamnetiya D., Patel P., Jha R. P., Shri N., Singh M., Bhattacharyya K.. **Trends in incidence and mortality of tuberculosis in India over past three decades: A joinpoint and age-period-cohort analysis**. *BMC Pulm. Med.* (2021) **21** 375. DOI: 10.1186/s12890-021-01740-y
5. Dirlikov E., Raviglione M., Scano F.. **Global tuberculosis control: Toward the 2015 targets and beyond**. *Ann. Intern Med.* (2015) **163** 52-58. DOI: 10.7326/m14-2210
6. **Global, regional, and national incidence, prevalence, and years lived with disability for 354 diseases and injuries for 195 countries and territories, 1990-2017: A systematic analysis for the global burden of disease study 2017**. *Lancet* (2018) **392** 1789-1858. DOI: 10.1016/s0140-6736(18)32279-7
7. **Global, regional, and national age-sex-specific mortality and life expectancy, 1950-2017: A systematic analysis for the global burden of disease study 2017**. *Lancet* (2018) **392** 1684-1735. DOI: 10.1016/s0140-6736(18)31891-9
8. **Burden of diabetes and hyperglycaemia in adults in the Americas, 1990-2019: A systematic analysis for the global burden of disease study 2019**. *Lancet Diabetes Endocrinol.* (2022) **10** 655-667. DOI: 10.1016/s2213-8587(22)00186-3
9. **Global burden of 87 risk factors in 204 countries and territories, 1990-2019: A systematic analysis for the global burden of disease study 2019**. *Lancet* (2020) **396** 1223-1249. DOI: 10.1016/s0140-6736(20)30752-2
10. **Global burden of 369 diseases and injuries in 204 countries and territories, 1990-2019: A systematic analysis for the global burden of disease study 2019**. *Lancet* (2020) **396** 1204-1222. DOI: 10.1016/s0140-6736(20)30925-9
11. Hankey B. F., Ries L. A., Kosary C. L., Feuer E. J., Merrill R. M., Clegg L. X.. **Partitioning linear trends in age-adjusted rates**. *Cancer Causes Control* (2000) **11** 31-35. DOI: 10.1023/a:1008953201688
12. Jeremiah C., Petersen E., Nantanda R., Mungai B. N., Migliori G. B., Amanullah F.. **The WHO Global Tuberculosis 2021 Report - not so good news and turning the tide back to End TB**. *Int. J. Infect. Dis.* (2022) **124** S26-S29. DOI: 10.1016/j.ijid.2022.03.011
13. Kazempour Dizaji M., Kazemnejad A., Tabarsi P., Zayeri F.. **Risk factors associated with survival of pulmonary tuberculosis**. *Iran. J. Public Health* (2018) **47** 980-987. PMID: 30181996
14. Lange C., Chesov D., Heyckendorf J., Leung C. C., Udwadia Z., Dheda K.. **Drug-resistant tuberculosis: An update on disease burden, diagnosis and treatment**. *Respirology* (2018) **23** 656-673. DOI: 10.1111/resp.13304
15. Lin M., Chen Y. W., Li Y. R., Long L. J., Qi L. Y., Cui T. T.. **Systematic evaluation of line probe assays for the diagnosis of tuberculosis and drug-resistant tuberculosis**. *Clin. Chim. Acta* (2022) **533** 183-218. DOI: 10.1016/j.cca.2022.06.020
16. Liu H., Wang Y., Li C., Li Z., Cheng L., Sun A.. **Burden distribution of infectious diseases in areas and countries along the China-proposed belt and road initiative**. *Med. Soc.* (2019) **32** 7-11. DOI: 10.13723/j.yxysh.2019.01.002
17. Mesfin Y. M., Hailemariam D., Biadgilign S., Kibret K. T.. **Association between HIV/AIDS and multi-drug resistance tuberculosis: A systematic review and meta-analysis**. *PLoS One* (2014) **9** e82235. DOI: 10.1371/journal.pone.0082235
18. Nordholm A. C., Andersen A. B., Wejse C., Norman A., Ekstrøm C. T., Andersen P. H.. **Social determinants of tuberculosis: A nationwide case-control study, Denmark, 1990-2018**. *Int. J. Epidemiol.* (2022) **51** 1446-1456. DOI: 10.1093/ije/dyac109
19. Pan Z., Zhang J., Bu Q., He H., Bai L., Yang J.. **The gap between global tuberculosis incidence and the first milestone of the WHO end tuberculosis strategy: An analysis based on the global burden of disease 2017 database**. *Infect. Drug Resist* (2020a) **13** 1281-1286. DOI: 10.2147/idr.S248875
20. Pan Z., Zhang J., Cheng H., Bu Q., Li N., Deng Y.. **Trends of the incidence of drug use disorders from 1990 to 2017: An analysis based on the global burden of disease 2017 data**. *Epidemiol. Psychiatr. Sci.* (2020b) **29** e148. DOI: 10.1017/s2045796020000657
21. Pearson K.. **Notes on the history of correlation**. *Biometrika* (1920) **13** 25-45. DOI: 10.1093/biomet/13.1.25
22. Rodgers J. L., Nicewander W. A.. **Thirteen ways to look at the correlation coefficient**. *Am. Statistician* (1988) **42** 59-66. DOI: 10.1080/00031305.1988.10475524
23. Seung K. J., Keshavjee S., Rich M. L.. **Multidrug-resistant tuberculosis and extensively drug-resistant tuberculosis**. *Cold Spring Harb. Perspect. Med.* (2015) **5** a017863. DOI: 10.1101/cshperspect.a017863
24. Shang W. J., Jing W. Z., Wang Y. P., Kang L. Y., Du M., Liu J.. **Epidemic situation of multidrug-resistant tuberculosis and change trend in Belt and Road countries**. *Zhonghua Liu Xing Bing Xue Za Zhi* (2022) **43** 1060-1065. DOI: 10.3760/cma.j.cn112338-20220126-00079
25. Soares P., Aguiar A., Leite A., Duarte R., Nunes C.. **Ecological factors associated with areas of high tuberculosis diagnosis delay**. *Public Health* (2022) **208** 32-39. DOI: 10.1016/j.puhe.2022.04.010
26. Sun H., Xue L., Guo Y., Du J., Nan K., Li M.. **Global, regional and national burden of non-hodgkin lymphoma from 1990 to 2017: Estimates from global burden of disease study in 2017**. *Ann. Med.* (2022) **54** 633-645. DOI: 10.1080/07853890.2022.2039957
27. Trisakul K., Nonghanphithak D., Chaiyachat P., Kaewprasert O., Sakmongkoljit K., Reechaipichitkul W.. **High clustering rate and genotypic drug-susceptibility screening for the newly recommended anti-tuberculosis drugs among global extensively drug-resistant**. *Emerg. Microbes Infect.* (2022) **11** 1857-1866. DOI: 10.1080/22221751.2022.2099304
28. **Sustainable development goals [Online]**. (2015)
29. Uplekar M., Weil D., Lonnroth K., Jaramillo E., Lienhardt C., Dias H. M.. **WHO's new end TB strategy**. *Lancet* (2015) **385** 1799-1801. DOI: 10.1016/s0140-6736(15)60570-0
30. **Global tuberculosis report 2019**. (2019)
31. **Global tuberculosis report 2021**. (2021a)
32. **Rapid сommunication: Key changes to treatment of multidrug- and rifampicin-resistant tuberculosis (MDR/RR-TB)**. (2018)
33. **The end TB strategy [Online]**. (2015)
34. **WHO global lists of high burden countries for TB, multidrug/rifampicin-resistant TB (MDR/RR-TB) and TB/HIV, 2021–2025**. (2021b)
35. Zhang Y., Yew W. W.. **Mechanisms of drug resistance in**. *Int. J. Tuberc. Lung Dis.* (2015) **19** 1276-1289. DOI: 10.5588/ijtld.15.0389
|
---
title: The role of indoleamine 2,3-dioxygenase 1 in early-onset post-stroke depression
authors:
- Hengshu Chen
- Xia Huang
- Chang Zeng
- Dongren Sun
- Fan Liu
- Jingyuan Zhang
- Qiao Liao
- Shihang Luo
- Weiye Xu
- Yeqing Xiao
- Danfeng Zeng
- Mingyu Song
- Fafa Tian
journal: Frontiers in Immunology
year: 2023
pmcid: PMC9998486
doi: 10.3389/fimmu.2023.1125634
license: CC BY 4.0
---
# The role of indoleamine 2,3-dioxygenase 1 in early-onset post-stroke depression
## Abstract
### Background
The immune-inflammatory response has been widely considered to be involved in the pathogenesis of post-stroke depression (PSD), but there is ambiguity about the mechanism underlying such association.
### Methods
According to Diagnostic and Statistical Manual of Mental Disorders (5th edition), depressive symptoms were assessed at 2 weeks after stroke onset. 15 single nucleotide polymorphisms (SNPs) in genes of indoleamine 2,3-dioxygenase (IDO, including IDO1 and IDO2) and its inducers (including pro-inflammatory cytokines interferon [IFN]-γ, tumor necrosis factor [TNF]-α, interleukin [IL]-1β, IL-2 and IL-6) were genotyped using SNPscan™ technology, and serum IDO1 levels were detected by double-antibody sandwich enzyme-linked immune-sorbent assay.
### Results
Fifty-nine patients ($31.72\%$) were diagnosed with depression at 2 weeks after stroke onset (early-onset PSD). The IDO1 rs9657182 T/T genotype was independently associated with early-onset PSD (adjusted odds ratio [OR] = 3.008, $95\%$ confidence interval [CI] 1.157-7.822, $$p \leq 0.024$$) and the frequency of rs9657182 T allele was significantly higher in patients with PSD than that in patients with non-PSD (χ2 = 4.355, $$p \leq 0.037$$), but these results did not reach the Bonferroni significance threshold ($p \leq 0.003$). Serum IDO1 levels were also independently linked to early-onset PSD (adjusted OR = 1.071, $95\%$ CI 1.002-1.145, $$p \leq 0.044$$) and patients with PSD had higher serum IDO1 levels than patients with non-PSD in the presence of the rs9657182 T allele but not homozygous C allele (t = -2.046, $$p \leq 0.043$$). Stroke patients with the TNF-α rs361525 G/G genotype had higher serum IDO1 levels compared to those with the G/A genotype (Z = -2.451, $$p \leq 0.014$$).
### Conclusions
Our findings provided evidence that IDO1 gene polymorphisms and protein levels were involved in the development of early-onset PSD and TNF-α polymorphism was associated with IDO1 levels, supporting that IDO1 which underlie strongly regulation by cytokines may be a specific pathway for the involvement of immune-inflammatory mechanism in the pathophysiology of PSD.
## Introduction
There is increasingly robust evidence that the activation of immune-inflammatory pathways plays an etiological role in the development and progression of post-stroke depression (PSD) [1], a complex and common post-stroke complication associated with increased morbidity and mortality [2], although its exact pathogenesis remains undetermined. The involvement of inflammation in the pathophysiology of PSD was initially based on an inflammatory hypothesis [3, 4] in which acute stroke induces a wide spectrum of central and peripheral immune-inflammatory responses, accompanied by upregulation of various pro-inflammatory cytokines (interleukin [IL]-1β, tumor necrosis factor [TNF]-α, interferon [IFN]-γ and IL-6, for instance) [5], subsequently resulting in increased expression of the gene encoding enzyme indoleamine 2,3-dioxygenase (IDO) 1 that triggers the depletion of serotonin, a unanimously identified feature of depression [6]. Increasing data have since proven that pro-inflammatory cytokines can act as biomarkers of PSD. Su et al. found that TNF-α, IFN-γ and IL-6 levels were elevated in patients suffering from PSD within 1 year after stroke [7]. In addition, Kim et al. observed that high serum levels of TNF-α and IL-1β were associated with an increased risk of PSD, especially in the acute stage of stroke and in patients carrying susceptible genes [8]. Additionally, results from Kang et al. demonstrated that higher serum IL-18 levels were independently related to PSD in the early and chronic phase after stroke [9], in line with the investigation by Yang et al. showing the predictive role of IL-18 levels in the risk of PSD [10]. Despite the robust association of pro-inflammatory cytokines with the occurrence of PSD, it is unclear whether the effect of increased immune activation resulting from stroke on the risk of PSD is associated with IDO1 expression.
IDO1 is an enzyme strictly regulated by cytokines and can be expressed in a variety of cells throughout the body in response to immunological signals, including IFN-γ, TNF-α, IL-1β, IL-2, and IL-6 stimulation [6, 11]. It has been well-established that IDO1 plays an important role in the etiology of depression through two mechanisms [6, 12], one is that the overactivated IDO1, an initial and key rate-limiting enzyme of the tryptophan catabolite pathway, tends to direct tryptophan down the kynurenine pathway that releases quinolinic acid, a powerful N-methyl-D-aspartate (NMDA) receptor agonist with definite neurotoxic effects which is involved in the onset of depression [13, 14], and the other is that IDO1 shunts tryptophan from the serotonin synthesis route, thereby favoring depression [15]. These observations indicate that IDO1 activation is relatively unique to inflammation-induced depression [16]. The available results, furthermore, reveal that upregulation of IDO1 activation is a characteristic of the post-stroke inflammatory response [17, 18]. Based on these findings, IDO1 may be a pivotal mediator of the contribution of stroke-associated inflammatory processes to the development of PSD. As such, the present study was designed to investigate the association between gene polymorphisms of IDO (IDO1 and IDO2, a recently recognized enzyme structurally and functionally similar to IDO1 [19]) and its inducers (including IFN-γ, TNF-α, IL-1β, IL-2 and IL-6) and PSD at 2 weeks after stroke (early-onset PSD), taking into account the role of serum IDO1 levels in the pathophysiology of early-onset PSD. In parallel, we analyzed the association between cytokine SNPs and serum IDO1 levels.
## Study population and clinical assessment
186 acute stroke patients hospitalized at the Department of Neurology, Xiangya Hospital of Central South University were recruited from July 2019 to February 2021. Inclusion criteria were: [1] age from 18-75 years, [2] diagnosed with acute stroke by brain magnetic resonance imaging or computerized tomography imaging within 2 weeks since onset, [3] ability to complete all necessary evaluations. The exclusion criteria were performed as described previously [20]. Written informed consents were signed from all patients, as approved by Medical Ethics Committee of the Xiangya Hospital of Central South University. We collected the information on demographic data (age, gender and years of education), vascular risk factors (hypertension, diabetes, heart disease, hyperlipidemia, current smoking and drinking), history of stroke, transient ischemic attack (TIA), intravenous thrombolysis and/or endovascular treatment, type of stroke (ischemic, hemorrhagic or subtypes according to the Trial of Org 10,172 in Acute Stroke Treatment [TOAST] classification) [21], stroke hemisphere (left, right or bilateral) and location (anterior, posterior or both), National Institute of Health Stroke Scale (NIHSS) score and Mini-Mental State Examination (MMSE) score, time from stroke onset to the blood sample collection, the complete blood counts (leukocyte, neutrophil, monocyte, lymphocyte and platelet counts) from the first blood routine results, pulmonary and/or urinary tract infection, and antibiotic. And the assessment of depressive symptoms and grouping of patients have been described in our previous investigation [20].
## Gene polymorphism selection and genotyping
We identified 15 SNPs in genes of IDO (IDO1 and IDO2) and its inducers (including IFN-γ, TNF-α, IL-1β, IL-2 and IL-6) selected from previous literature associated with PSD or depression or stroke, with a minor allele frequency > 0.05 indexed in the Chinese Han population dataset of a genetic database (http://www.ensembl.org). Each SNP was genotyped using SNPscan™ multiple SNP genotyping technology [22] unaware of the sample status.
## Measurement of serum IDO1 levels
Fasting venous blood samples were obtained within 2 weeks after stroke onset. Quantitative IDO1 assay was performed for the determination of IDO1 concentrations in serum using double-antibody sandwich enzyme-linked immune-sorbent assay kit provided by Shanghai Tianhao Biotechnology Co., Ltd., China. The minimum detectable dose was less than 0.1 ng/mL. The inter- and intra-assay coefficients of variation were less than $10\%$ and $15\%$, respectively.
## Statistical analysis
The data was analyzed using the IBM SPSS Statistics for Windows, version 26 (IBM Corp, Armonk, NY, USA). Shapiro-Wilk test was used to determine the normality of continuous variables. Continuous variables with normal distribution were summarized as means ± standard deviations assessed by Student’s t-test, and continuous variables with non-normal distribution were presented as median (interquartile range [IQR]) analyzed by Mann-Whitney U test. Categorical variables were reported as absolute number (percentage value) compared by Chi-squared test or Fisher’s exact test. Hardy-*Weinberg equilibrium* (HWE) was examined using the Chi-squared test based on the genotype distribution in non-PSD group. The binary logistic regression model allowing adjustment for statistically significant confounding factors, in addition, was performed to identify independent risk factors for PSD. A two-side p value of less than 0.05 was considered statistically significant and the multiple comparisons were adjusted by the Bonferroni correction with a corrected p-value threshold ($$p \leq 0.05$$/15 = 0.003), given that fifteen tests were performed for the association of each SNP with PSD. Furthermore, depending on data distribution, the correlation between cytokine SNPs and serum IDO1 levels was evaluated using one-way analysis of variance with Bonferroni post-hoc test or non-parametric test (Kruskal-Wallis or Mann-Whitney U test) ($p \leq 0.05$).
## Demographic and clinical characteristics
General demographic and clinical characteristics of the cohort were shown in Table 1. In the participants as a whole, a median (IQR) age of them was 57 (51–65) years and 57 ($30.65\%$) were female, 15 ($8.06\%$) of whom were diagnosed with hemorrhagic stroke. PSD was found in 59 ($31.72\%$) patients who experienced an assessment of depressive symptoms along with non-PSD patients on Day 14 after stroke onset. Compared to the non-PSD group, the PSD group exhibited a higher distribution of female ($40.68\%$ versus $25.98\%$, $$p \leq 0.043$$), lower median (IQR) of years of education (9 [6-12] versus 12 [9-12], $$p \leq 0.018$$), and lower MMSE score reflecting cognitive function (24 [21-28] versus 26 [23-29], $$p \leq 0.033$$), but none of them were independently associated with PSD status (Table 2). No statistically significant differences, however, were observed between the PSD and non-PSD groups with respect to age, vascular risk factors, history of stroke, TIA, intravenous thrombolysis and/or endovascular treatment, type of stroke, stroke hemisphere and location, NIHSS score, time from stroke onset to the blood sample collection, the complete blood counts from the first blood routine results, pulmonary and/or urinary tract infection, and antibiotic.
## Genotype and allelic frequencies by PSD status
The genotype and allele frequencies of IDO (IDO1 and IDO2) and its inducers (including IFN-γ, TNF-α, IL-1β, IL-2 and IL-6) SNPs in PSD and non-PSD patients were summarized in Table I in the Supplementary Material. There was no deviation from HWE for any genotype in non-PSD group (all $p \leq 0.05$). As shown in Table I, the genotype and allele distributions of IDO1 polymorphisms in PSD group were distinct from those in patients with non-PSD, showing that the frequency of the IDO1 rs9657182 T/T genotype was significantly higher in PSD patients than non-PSD patients before (odds ratio [OR] = 2.615, $95\%$ confidence interval [CI] 1.048-6.529, $$p \leq 0.039$$) and after adjusting potential confounders including gender, years of education, MMSE score and serum IDO1 levels (adjusted OR = 3.008, $95\%$ CI 1.157-7.822, $$p \leq 0.024$$) (Table 2), and that the frequency of the rs9657182 T allele was also significantly higher in patients with PSD compared to those with non-PSD (χ2 = 4.355, $$p \leq 0.037$$), but these differences disappeared after Bonferroni correction ($p \leq 0.003$). We did not observe any significant differences in either the genotype or allelic frequencies of IDO1 rs7820268 and IDO2 rs2929115 between PSD and non-PSD groups, as did gene polymorphisms of IDO inflammatory stimuli (including IFN-γ, TNF-α, IL-1β, IL-2 and IL-6). According to the foregoing, the IDO1 rs9657182 T/T genotype was independent risk factor of early-onset PSD and the rs9657182 T allele conferred an elevated risk for the development of early-onset PSD, providing suggestive association of the IDO1 rs9657182 polymorphism with the risk of early-onset PSD (0.003 < $P \leq 0.05$).
## Association of serum IDO1 levels with PSD status
In the study population, in comparison to patients with non-PSD, serum IDO1 levels were higher in patients with PSD (t = -2.040, $$p \leq 0.043$$) (Table 1), suggesting that the increased serum IDO1 levels were related to an elevated risk of early-onset PSD. Moreover, the association with PSD status remained stable even after controlling for possible covariates shown in Table 2 (adjusted OR = 1.071, $95\%$ CI 1.002-1.145, $$p \leq 0.044$$). Additionally, serum IDO1 levels were further analyzed depending on the IDO1 rs9657182 allele distribution and PSD status, and we found that patients with PSD showed greater serum IDO1 levels than patients with non-PSD in the presence of the rs9657182 T allele but not homozygous C allele (t = -2.046, $$p \leq 0.043$$) (Figure 1). These results indicated that there was an independent association between serum IDO1 levels and early-onset PSD, and that the IDO1 rs9657182 T allele increased the risk of early-onset PSD by enhancing serum IDO1 levels.
**Figure 1:** *Comparison of serum indoleamine 2,3-dioxygenase (IDO) 1 concentrations according to the presence or absence of the IDO1 rs9657182 T allele between PSD and non-PSD patients. The result was presented as median (interquartile range) on the left side of
Figure 1
compared by the Mann-Whitney U-test, and the data was shown as means ± standard deviations on the right side of
Figure 1
compared by the Student’s t-test. *p < 0.05.*
## Association of serum IDO1 levels with cytokine SNPs
Serum IDO1 levels of the study population by gene polymorphisms of pro-inflammatory cytokines including IFN-γ, TNF-α, IL-1β, IL-2 and IL-6 were listed in Table II in the Supplementary Material. Serum IDO1 levels were significantly higher in stroke patients carrying the TNF-α rs361525 G/G genotype than in those carrying the G/A genotype (Z = -2.451, $$p \leq 0.014$$), indicating that TNF-α rs361525 polymorphism was associated with serum IDO1 levels. Besides, there was a correlation between TNF-α rs1799964 polymorphism and serum IDO1 levels ($F = 3.564$, $$p \leq 0.030$$) but in adjusted Bonferroni post-hoc comparisons this finding did not reach the threshold of statistical significance (T/T versus T/C genotype, $$p \leq 0.076$$; T/T versus C/C genotype, $$p \leq 0.297$$; T/C versus C/C genotype, $$p \leq 1.000$$). No significant relationships of serum IDO1 levels with SNPs of IFN-γ, IL-1β, IL-2 and IL-6 were found ($p \leq 0.05$).
## Discussion
The present study examined for the first time the role of gene polymorphisms of IDO (IDO1 and IDO2) and its inducer and serum IDO1 levels in the risk of PSD at 2 weeks after stroke onset and also discussed the relationship between cytokine SNPs and serum IDO1 levels. The results preliminarily indicated here that IDO1 rs9657182 T allele was a suggestive predisposing factor for early-onset PSD, which was associated with increased serum IDO1 levels possibly due to the T allele affecting the transcriptional activity of the promoter region of the IDO1 gene, and that patients with acute stroke who carry the rs9657182 T/T genotype had an increased susceptibility to early-onset PSD, and that serum IDO1 levels were an independent risk factor for early-onset PSD, and that serum IDO1 levels in stroke settings were relevant to the TNF-α rs361525 polymorphism. These findings provided further support for the cytokine hypothesis of PSD during the acute phase of stroke and also suggested that the rs9657182 polymorphism and serum IDO1 levels might be novel diagnostic biomarkers and/or intervention targets for early-onset PSD.
Considering the overwhelming evidence on a responsible role of pro-inflammatory cytokines in the etiology of PSD [1] and cytokine-inducible IDO1 as a key factor in inflammation-induced depression [23], gene polymorphisms involved in determining the functional activity of IDO1 and its inducer cytokines are promising candidate contributors to PSD. In this regarding, gene polymorphisms of pro-inflammatory cytokines that stimulate IDO1 expression, including IFN-γ, TNF-α, IL-1β, IL-2 and IL-6, were analyzed between PSD and non-PSD patients. The result here that SNPs in genes encoding the above IDO-associated inflammatory stimulants were not correlated with early-onset PSD was in accordance with our previous study showing that there was no link between gene polymorphisms of pro-inflammatory cytokines (IFN-γ, TNF-α, IL-1β and IL-6) and early-onset PSD [20]. Furthermore, Kim et al. also found that pro-inflammatory cytokines, such as TNF-α, IL-1β and IL-6, were independent of depression during the acute phase of stroke [24]. A previous investigation, however, highlighted the potential synergistically effects of TNF-α and IL-1β, considering their corresponding alleles together, on the risk of PSD at 2 weeks post-stroke [8]. These inconsistent findings may be related to the following events: the observability of functional polymorphisms of pro-inflammatory cytokines is influenced by population distribution [24] and there is heterogeneity in respect to diagnostic criteria of depression, sample size and subject selection.
In parallel, the association of the rs9657182 polymorphism, localized in the promoter region of the IDO1 gene [25], with PSD status was also explored in present study. Although polymorphisms in these pro-inflammatory cytokines were not risk factors for early-onset PSD status, we observed that the IDO1 rs9657182 polymorphism was correlated with early-onset PSD, providing suggestive evidence for a potential causal link between the IDO1 polymorphism and the risk of early-onset PSD. Specifically, the T/T genotype of IDO1 rs9657182 investigated in this study was an independent precipitating factor for early-onset PSD and the rs9657182 T allele was also a risk factor for it, similar to the finding of Smith et al. suggesting that the rs9657182 polymorphism was a predictor of the development of cytokine-induced depressive symptoms during treatment with IFN-α in Caucasian patients with chronic hepatitis C [25]. The data derived from animal models, besides, displayed that inflammatory stimuli did not induce depression-like behavior when IDO1 gene was genetically deficient [26]. These investigations highlighted the causative role of IDO1 polymorphism in the pathophysiology of depression mediated by immune stimulation. However, there was no association between early-onset PSD and the rs2929115 polymorphism of IDO2, adjacent to IDO1 gene and with a similarity of IDO2 to IDO1 in protein structure [27], indicating either that IDO2 has limited correlation with early-onset PSD or that the primary acts of IDO2 lie outside of its enzymatic function [28]. And we further assessed the relevance of serum IDO1 levels to early-onset PSD status. Our study demonstrated that the correlation of serum IDO1 levels with early-onset PSD varied by the rs9657182 allele distribution. Increased serum IDO1 levels in stroke patients carrying the T allele but not homozygous C allele of the rs9657182 conveyed a liability to early-onset PSD, and in combination with the evidence on the contribution of the T allele to early-onset PSD formation, it could be assumed that the polymorphism in the IDO1 gene promoter region at position rs9657182 affected the expression levels of IDO1 in response to cytokines stimulation following stroke, giving rise to the development of early-onset PSD. It was interesting to note that serum IDO1 levels were associated with an increased risk of early-onset PSD, independent of the presence of the IDO1 rs9657182 polymorphism, possibly because IDO1 production was subject to the complex interplay and/or synergy of numerous parameters in immune-inflammatory settings, such as other IDO1 gene polymorphisms and multiple pro-inflammatory cytokine concentrations. Our finding was further supported by a preclinical study that denoted increased IDO expression in the nucleus accumbens, hippocampus, and hypothalamus of PSD-like phenotype mice [29]. There are several underlying mechanisms that may account for the association of IDO1 with PSD: the elevated levels of pro-inflammatory cytokines after stroke upregulate IDO1 expression that causes the depletion of serotonin precursor tryptophan and increased neurotoxic kynurenine metabolite (quinolinic acid), an agonist of NMDA receptors, leading to the occurrence of depression (14, 30–32). IDO1 has been seen as a central hub linking immune-inflammatory processes to the monoaminergic [33] and glutamatergic systems implicated in depression [14]. Our results, taken together, supported that the interactions of cytokine- serotonin and -glutamate via IDO1 could play a key role in PSD [3, 4].
In the state of stroke-induced immune activation, upon analysis on the relationship between individual genetic variation in cytokines and circulating IDO1 levels, we observed that stroke patients harboring the G/G genotype at the rs361525 locus of the TNF-α gene had higher serum IDO1 levels compared to the G/A genotype. We hypothesized that the association of serum IDO1 levels with the rs361525 polymorphism was related to the expression of TNF-α. The possible explanation is that the SNP rs361525 in the promoter region of TNF-α gene enhance the production of TNF-α [34, 35], which synergistically promotes IDO1 upregulation with other cytokines [36, 37]. It should be noted that in our cases there was no mutant homozygous A allele of the rs361525. More credible large-scale studies are warranted to clarify the role of the TNF-α levels in combination with its gene polymorphism in IDO1 expression, especially in the context of immune activation. Interestingly for our purposes, the TNF-α rs361525 polymorphism may be indirectly involved in the development of early-onset PSD, given the findings in the present study that the rs361525 polymorphism was associated with serum IDO1 levels, which were an independent risk factor for early-onset PSD. Our data, collectively, provided reasonable grounds to assume that IDO1 may be a crucial mediator linking inflammation and early-onset PSD.
The limitations of the current study need to be considered. Firstly, this cross-sectional study contributed to establishing a preliminary association between IDO1 and early-onset PSD, but it would be informative to design longitudinal studies to assess the value of IDO1 in late-onset PSD. Secondly, although there is a degree of overlap between brain and peripheral IDO1 activity [23], results based on serum IDO1 levels could not be readily applied to the central nervous system and their relationship needs to be specifically investigated. Thirdly, due to the limitation of sample size, we did not further evaluate the severity of PSD and it could be interesting to understand if there is an association between gene polymorphisms, serum IDO1 levels and severity of depression in a larger study population. Fourthly, serum levels of IDO inflammatory stimulants, including IFN-γ, TNF-α, IL-1β, IL-2 and IL-6, were not measured and the combination of cytokine levels with their corresponding gene polymorphisms helps to identify the optimal risk factors for PSD.
In conclusion, our study had shed light on that gene polymorphisms as suggestive risk predictors and expression levels of IDO1 were involved in depression occurring during the acute stage of stroke and the TNF-α rs361525 polymorphism was associated with serum IDO1 levels. These findings may be a meaningful addition to the neuroimmune pathways in the pathophysiology of early-onset PSD and serum IDO1 levels, alone or in combination with its corresponding polymorphism, may allow for a different intervention focusing on more specific etiologically-based management for early-onset PSD. Future adequately powerful trials are necessary to elucidate the role of IDO1 associated with the post-stroke immune-inflammatory responses in the pathogenesis of PSD, especially depression during the acute phase of stroke, from different perspectives.
## Data availability statement
The original contributions presented in the study are included in the article/ Supplementary Materials. Further inquiries can be directed to the corresponding authors.
## Ethics statement
The studies involving human participants were reviewed and approved by Medical Ethics Committee of the Xiangya Hospital of Central Department of Clinical Pharmacology, Xiangya Hospital, Central South University, Changsha, China South University.
## Author contributions
HC, XH, DS, FL, MS and FT contributed to conception and design of the study. HC, XH, JZ, QL and MS organized the database. HC wrote the first draft of the manuscript. HC, CZ, SL, WX, YX, DZ wrote sections of the manuscript. All authors contributed to the article and approved the submitted version.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu.2023.1125634/full#supplementary-material
## References
1. Wijeratne T, Sales C. **Understanding why post-stroke depression may be the norm rather than the exception: The anatomical and neuroinflammatory correlates of post-stroke depression**. *J Clin Med* (2021) **10** 1674. DOI: 10.3390/jcm10081674
2. Loubinoux I, Kronenberg G, Endres M, Schumann-Bard P, Freret T, Filipkowski RK. **Post-stroke depression: mechanisms, translation and therapy**. *J Cell Mol Med* (2012) **16**. DOI: 10.1111/j.1582-4934.2012.01555.x
3. Spalletta G, Bossù P, Ciaramella A, Bria P, Caltagirone C, Robinson RG. **The etiology of poststroke depression: a review of the literature and a new hypothesis involving inflammatory cytokines**. *Mol Psychiatry* (2006) **11**. DOI: 10.1038/sj.mp.4001879
4. Li W, Ling S, Yang Y, Hu Z, Davies H, Fang M. **Systematic hypothesis for post-stroke depression caused inflammation and neurotransmission and resultant on possible treatments**. *Neuro Endocrinol Lett* (2014) **35**
5. Iadecola C, Anrather J. **The immunology of stroke: from mechanisms to translation**. *Nat Med* (2011) **17** 796-808. DOI: 10.1038/nm.2399
6. Maes M, Leonard BE, Myint AM, Kubera M, Verkerk R. **The new '5-HT' hypothesis of depression: cell-mediated immune activation induces indoleamine 2,3-dioxygenase, which leads to lower plasma tryptophan and an increased synthesis of detrimental tryptophan catabolites (TRYCATs), both of which contribute to the onset of depression**. *Prog Neuropsychopharmacol Biol Psychiatry* (2011) **35**. DOI: 10.1016/j.pnpbp.2010.12.017
7. Su JA, Chou SY, Tsai CS, Hung TH. **Cytokine changes in the pathophysiology of poststroke depression**. *Gen Hosp Psychiatry* (2012) **34**. DOI: 10.1016/j.genhosppsych.2011.09.020
8. Kim JM, Kang HJ, Kim JW, Bae KY, Kim SW, Kim JT. **Associations of tumor necrosis factor-α and interleukin-1β levels and polymorphisms with post-stroke depression**. *Am J Geriatr Psychiatry* (2017) **25**. DOI: 10.1016/j.jagp.2017.07.012
9. Kang HJ, Bae KY, Kim SW, Kim JT, Park MS, Cho KH. **Effects of interleukin-6, interleukin-18, and statin use, evaluated at acute stroke, on post-stroke depression during 1-year follow-up**. *Psychoneuroendocrinology* (2016) **72**. DOI: 10.1016/j.psyneuen.2016.07.001
10. Yang L, Zhang Z, Sun D, Xu Z, Zhang X, Li L. **The serum interleukin-18 is a potential marker for development of post-stroke depression**. *Neurol Res* (2010) **32**. DOI: 10.1179/016164110x12656393665080
11. Myint AM, Kim YK. **Cytokine-serotonin interaction through IDO: a neurodegeneration hypothesis of depression**. *Med Hypotheses* (2003) **61**. DOI: 10.1016/s0306-9877(03)00207-x
12. Müller N, Schwarz MJ. **The immune-mediated alteration of serotonin and glutamate: towards an integrated view of depression**. *Mol Psychiatry* (2007) **12** 988-1000. DOI: 10.1038/sj.mp.4002006
13. Maes M, Mihaylova I, Ruyter MD, Kubera M, Bosmans E. **The immune effects of TRYCATs (tryptophan catabolites along the IDO pathway): relevance for depression - and other conditions characterized by tryptophan depletion induced by inflammation**. *Neuro Endocrinol Lett* (2007) **28**
14. Wu X, Chen B, Di Z, Jiang S, Xu H, Shi M. **Involvement of kynurenine pathway between inflammation and glutamate in the underlying etiopathology of CUMS-induced depression mouse model**. *BMC Neurosci* (2022) **23** 62. DOI: 10.1186/s12868-022-00746-4
15. Chaves Filho AJM, Lima CNC, Vasconcelos SMM, de Lucena DF, Maes M, Macedo D. **IDO chronic immune activation and tryptophan metabolic pathway: A potential pathophysiological link between depression and obesity**. *Prog Neuropsychopharmacol Biol Psychiatry* (2018) **80**. DOI: 10.1016/j.pnpbp.2017.04.035
16. Dantzer R, O'Connor JC, Freund GG, Johnson RW, Kelley KW. **From inflammation to sickness and depression: when the immune system subjugates the brain**. *Nat Rev Neurosci* (2008) **9** 46-56. DOI: 10.1038/nrn2297
17. Mo X, Pi L, Yang J, Xiang Z, Tang A. **Serum indoleamine 2,3-dioxygenase and kynurenine aminotransferase enzyme activity in patients with ischemic stroke**. *J Clin Neurosci* (2014) **21**. DOI: 10.1016/j.jocn.2013.08.020
18. Jackman KA, Brait VH, Wang Y, Maghzal GJ, Ball HJ, McKenzie G. **Vascular expression, activity and function of indoleamine 2,3-dioxygenase-1 following cerebral ischaemia-reperfusion in mice**. *Naunyn Schmiedebergs Arch Pharmacol* (2011) **383**. DOI: 10.1007/s00210-011-0611-4
19. Murray MF. **The human indoleamine 2,3-dioxygenase gene and related human genes**. *Curr Drug Metab* (2007) **8** 197-200. DOI: 10.2174/138920007780362509
20. Chen H, Liu F, Sun D, Zhang J, Luo S, Liao Q. **The potential risk factors of early-onset post-stroke depression from immuno-inflammatory perspective**. *Front Immunol* (2022) **13**. DOI: 10.3389/fimmu.2022.1000631
21. Adams HP, Bendixen BH, Kappelle LJ, Biller J, Love BB, Gordon DL. **Classification of subtype of acute ischemic stroke. definitions for use in a multicenter clinical trial. TOAST. trial of org 10172 in acute stroke treatment**. *Stroke* (1993) **24** 35-41. DOI: 10.1161/01.str.24.1.35
22. Cheng J, Li Y, He Q, Luo L, Zhang Y, Gao Y. **Essential hypertension in patients exposed to high-arsenic exposed areas in western China: Genetic susceptibility and urinary arsenic metabolism characteristics**. *J Trace Elem Med Biol* (2021) **67**. DOI: 10.1016/j.jtemb.2021.126778
23. Walker AK, Kavelaars A, Heijnen CJ, Dantzer R. **Neuroinflammation and comorbidity of pain and depression**. *Pharmacol Rev* (2014) **66** 80-101. DOI: 10.1124/pr.113.008144
24. Kim JM, Stewart R, Kim SW, Shin IS, Kim JT, Park MS. **Associations of cytokine gene polymorphisms with post-stroke depression**. *World J Biol Psychiatry* (2012) **13**. DOI: 10.3109/15622975.2011.588247
25. Smith AK, Simon JS, Gustafson EL, Noviello S, Cubells JF, Epstein MP. **Association of a polymorphism in the indoleamine- 2,3-dioxygenase gene and interferon-α-induced depression in patients with chronic hepatitis c**. *Mol Psychiatry* (2012) **17**. DOI: 10.1038/mp.2011.67
26. O'Connor JC, Lawson MA, André C, Briley EM, Szegedi SS, Lestage J. **Induction of IDO by bacille calmette-guérin is responsible for development of murine depressive-like behavior**. *J Immunol* (2009) **182**. DOI: 10.4049/jimmunol.0802722
27. Fatokun AA, Hunt NH, Ball HJ. **Indoleamine 2,3-dioxygenase 2 (IDO2) and the kynurenine pathway: characteristics and potential roles in health and disease**. *Amino Acids* (2013) **45**. DOI: 10.1007/s00726-013-1602-1
28. Merlo LMF, Peng W, DuHadaway JB, Montgomery JD, Prendergast GC, Muller AJ. **The immunomodulatory enzyme IDO2 mediates autoimmune arthritis through a nonenzymatic mechanism**. *J Immunol* (2022) **208**. DOI: 10.4049/jimmunol.2100705
29. Koo YS, Kim H, Park JH, Kim MJ, Shin YI, Choi BT. **Indoleamine 2,3-Dioxygenase-Dependent neurotoxic kynurenine metabolism contributes to poststroke depression induced in mice by ischemic stroke along with spatial restraint stress**. *Oxid Med Cell Longev* (2018) **2018**. DOI: 10.1155/2018/2413841
30. Raison CL, Dantzer R, Kelley KW, Lawson MA, Woolwine BJ, Vogt G. **CSF concentrations of brain tryptophan and kynurenines during immune stimulation with IFN-alpha: relationship to CNS immune responses and depression**. *Mol Psychiatry* (2010) **15** 393-403. DOI: 10.1038/mp.2009.116
31. Savitz J. **The kynurenine pathway: a finger in every pie**. *Mol Psychiatry* (2020) **25**. DOI: 10.1038/s41380-019-0414-4
32. Gao L, Gao T, Zeng T, Huang P, Wong NK, Dong Z. **Blockade of indoleamine 2, 3-dioxygenase 1 ameliorates hippocampal neurogenesis and BOLD-fMRI signals in chronic stress precipitated depression**. *Aging (Albany NY)* (2021) **13**. DOI: 10.18632/aging.202511
33. Werner ER, Bitterlich G, Fuchs D, Hausen A, Reibnegger G, Szabo G. **Human macrophages degrade tryptophan upon induction by interferon-gamma**. *Life Sci* (1987) **41**. DOI: 10.1016/0024-3205(87)90149-4
34. Laddha NC, Dwivedi M, Begum R. **Increased tumor necrosis factor (TNF)-α and its promoter polymorphisms correlate with disease progression and higher susceptibility towards vitiligo**. *PloS One* (2012) **7**. DOI: 10.1371/journal.pone.0052298
35. Sharma S, Ghosh B, Sharma SK. **Association of TNF polymorphisms with sarcoidosis, its prognosis and tumour necrosis factor (TNF)-alpha levels in Asian indians**. *Clin Exp Immunol* (2008) **151**. DOI: 10.1111/j.1365-2249.2007.03564.x
36. Robinson CM, Hale PT, Carlin JM. **The role of IFN-gamma and TNF-alpha-responsive regulatory elements in the synergistic induction of indoleamine dioxygenase**. *J Interferon Cytokine Res* (2005) **25** 20-30. DOI: 10.1089/jir.2005.25.20
37. Werner-Felmayer G, Werner ER, Fuchs D, Hausen A, Reibnegger G, Wachter H. **Tumour necrosis factor-alpha and lipopolysaccharide enhance interferon-induced tryptophan degradation and pteridine synthesis in human cells**. *Biol Chem Hoppe Seyler* (1989) **370**. DOI: 10.1515/bchm3.1989.370.2.1063
|
---
title: Characteristic gene expression in the liver monocyte-macrophage-DC system is
associated with the progression of fibrosis in NASH
authors:
- Xiaoxiao Wang
- Zilong Wang
- Baiyi Liu
- Rui Jin
- Yuyun Song
- Ran Fei
- Xu Cong
- Rui Huang
- Xiaohe Li
- Jia Yang
- Lai Wei
- Huiying Rao
- Feng Liu
journal: Frontiers in Immunology
year: 2023
pmcid: PMC9998489
doi: 10.3389/fimmu.2023.1098056
license: CC BY 4.0
---
# Characteristic gene expression in the liver monocyte-macrophage-DC system is associated with the progression of fibrosis in NASH
## Abstract
### Background
The monocyte-macrophage-dendritic cell (DC) (MMD) system exerts crucial functions that may modulate fibrogenesis in nonalcoholic steatohepatitis (NASH). In this study, we explored the cell characteristics, distribution and developmental trajectory of the liver MMD system in NASH mice with fibrosis and clarified characteristic genes of the MMD system involved in liver fibrosis progression in NASH mice and patients.
### Methods
Single cells in liver tissue samples from NASH and normal mice were quantified using single-cell RNA sequencing (scRNA-seq) analysis. Differentially expressed genes (DEGs) in the MMD system by pseudotime analysis were validated by tyramide signal amplification (TSA)-immunohistochemical staining (IHC) and analyzed by second harmonic generation (SHG)/two-photon excitation fluorescence (TPEF).
### Results
Compared with control mice, there were increased numbers of monocytes, Kupffer cells, and DCs in two NASH mouse models. From the transcriptional profiles of these single cells, we identified 8 monocyte subsets (Mono1-Mono8) with different molecular and functional properties. Furthermore, the pseudotime analysis showed that Mono5 and Mono6 were at the beginning of the trajectory path, whereas Mono2, Mono4, Kupffer cells and DCs were at a terminal state. Genes related to liver collagen production were at the late stage of this trajectory path. DEGs analysis revealed that the genes Fmnl1 and Myh9 in the MMD system were gradually upregulated during the trajectory. By TSA-IHC, the Fmnl1 and Myh9 expression levels were increased and associated with collagen production and fibrosis stage in NASH mice and patients.
### Conclusions
Our transcriptome data provide a novel landscape of the MMD system that is involved in advanced NASH disease status. Fmnl1 and Myh9 expression in the MMD system was associated with the progression of NASH fibrosis.
## Introduction
Nonalcoholic fatty liver disease (NAFLD) is defined as excessive fat deposition in the liver and in absence of heavy drinking and other chronic liver diseases [1, 2]. The natural history of NAFLD consists of steatosis, nonalcoholic steatohepatitis (NASH), fibrosis, cirrhosis and hepatocellular carcinoma (HCC) (1–3). Approximately one-quarter of the general population has NAFLD, 10-$30\%$ of NAFLD patients are at risk of progressing to NASH and liver fibrosis, and 0.3-$3\%$ of NASH patients with fibrosis progress to cirrhosis and HCC annually [4]. In particular, fibrosis is considered as a crucial adverse result in the natural process of NASH progression.
Fibrosis is independently associated with liver transplantation and liver-related risk events in patients with NASH (5–7). The relative risks of hazardous events increased from fibrosis stage 2, especially in NASH patients with cirrhosis, the occurrence of liver decompensation and the mortality of liver-related events and all-cause mortality increased substantially [5, 8, 9]. Therefore, growing emphasis has been placed on improving or treating fibrosis in NASH [10]. However, there are currently no approved drugs for NAFLD and NASH by the Food and Drug Administration (FDA), especially for NASH fibrosis.
At present, the mechanism of liver fibrogenesis in NASH is unclear. Previous studies have shown that monocytes, macrophages and dendritic cells (DCs) exert diversity in modulating fibrogenesis [11, 12]. In the healthy state, the majority of macrophages are liver-resident yolk sac-derived Kupffer cells [13], while in the setting of liver injury, including NASH, there is marked infiltration of monocyte-derived macrophages (MoMFs) [14]. Furthermore, under inflammatory settings, monocytes can also differentiate into a special subset of DCs, called moDCs (monocyte-derived DCs) [15]. However, until recently, the characteristic genes expressed in the monocyte-macrophage-DC (MMD) system that are involved in the progression of NASH fibrosis were not fully understood. In recent years, single-cell transcriptomics provides high-dimensional information about cellular composition of various tissues, which reshapes and updates the understanding of complex biological systems for researchers [16]. Although single-cell RNA sequencing (scRNA-seq) has been conducted and applied in the livers of mice and humans, few studies have investigated the characteristics of the MMD system in the fibrogenesis of NASH. In this study, we explored the cell composition and developmental trajectory of cells in the MMD system and clarified the relationship between hub genes from the MMD system and collagen deposition in NASH fibrosis.
## Experimental mouse models
C57BL/6 mice (RRID: MGI:2159769) were fed a high-fat diet (HFD, $23.6\%$ fat, $41.5\%$ carbohydrate and $0.02\%$ cholesterol; MD12032, Medicience Ltd., China) or western diet (WD, $21.2\%$ fat, $48.5\%$ carbohydrate, and $1.25\%$ cholesterol; TD120528, Medicience Ltd., China) and fructose solution (23.1 g fructose and 18.9 g glucose in a liter of tap water) plus intraperitoneal injection of CCl4 (carbon tetrachloride, $10\%$, 2.5ul/g body weight, once a week) (HFD+F+CCl4 and WD+F+CCl4) for 16 weeks. In addition, the control mice were given normal diet ($18\%$ kcal fat, $58\%$ kcal carbohydrate, SPF Biotechnology Co., Ltd.) and drinking water. Before model construction, mice were assigned into control and two experimental groups using randomizations. After 16 weeks, no mice were excluded from our study and all eight mice are available for final analysis. This study was approved by the Ethics Committee of Peking University People’s Hospital (2021PHE111) and conformed to the ethical guidelines of the 1975 Declaration of Helsinki.
## Tissue dissociation, preparation of single-cell suspensions, Chromium 10x Genomics library and sequencing
Liver tissues of mice were obtained from wild-type mice ($$n = 2$$) and NASH mice with fibrosis (HFD+F+CCl4, $$n = 3$$; WD+F+CCl4, $$n = 3$$) after heart perfusion. Subsequently, liver tissue of each mice was transferred into a sterile RNase-free culture dish containing 1x PBS (calcium and magnesium-free) on ice and cut it into 0.5 mm2 pieces. Liver tissues were dissociated into single cells in dissociation solution ($0.35\%$ collagenase IV5, 2 mg/ml papain, 120 Units/ml DNase I) in 37 °C water bath with shaking for 20 min at 100 rpm. The cell suspension was filtered by passing through 70-30 um stacked cell strainer and centrifuged at 300g for 5 min at 4°C. The cell pellet was resuspended in 100 ul 1x PBS and added with 1 ml 1x red blood cell lysis buffer (MACS 130-094-183) and incubated at room temperature for 2-10 min to lyse remaining red blood cells. After incubation, the suspension was centrifuged at 300g for 5 min. The suspension was resuspended in 100μl Dead Cell Removal MicroBeads (MACS 130-090-101) and remove dead cells using Miltenyi® Dead Cell Removal Kit (MACS 130-090-101). Then the suspension was resuspended in 1 xPBS and centrifuged at 300 g for 3 min. The cell pellet was resuspended in 50 μl of 1 x PBS. The overall cell viability was confirmed by trypan blue exclusion, which needed to be above $85\%$.
Depending on the manufacturer’s instructions of the 10X Genomics Chromium Single-Cell 3’ Kit (V3), single-cell suspensions were loaded into 10x Chromium. Then, according to the standard operation protocols, we performed cDNA amplification and library construction. Libraries were sequenced on an Illumina NovaSeq 6000 sequencing system (paired-end multiplexing run, 150 bp) at a minimum depth of 20,000 reads per cell by LC-BioTechnology Co., Ltd. (HangZhou, China).
## Single-cell RNA sequencing data analysis
Using Illumina bcl2fastq software (version 2.20), sequencing results were demultiplexed and converted to FASTQ format. Sample demultiplexing, barcode processing and single-cell 3’ gene counting were performed depending on the Cell Ranger pipeline (version 5.0.1, https://support.10xgenomics.com/single-cell-geneexpression/software/pipelines/latest/what-is-cell-ranger), and scRNA-seq data were aligned to the Ensembl genome GRCh38/GRCm38 reference genome. About 88,000 single cells captured from 8 mice were processed using 10X Genomics Chromium Single Cell 3’ Solution. The Cell Ranger output was loaded into Seurat (version 3.1.1) and used for dimensional reduction, clustering, and scRNA-seq data analysis.
## Cell annotation and comparison of differentially expressed genes
We completed our manual clusters annotations depending on marker genes of specific cells from published papers and CellMarker (CellMarker/index.jsp). The key differentially expressed transcripts that define each cell cluster are exhibited in Table 1. We performed bioinformatic analysis using the OmicStudio tools at https://www.omicstudio.cn/tool.
**Table 1**
| Cell | Marker |
| --- | --- |
| Hepatocyte | Afp, Alb |
| Hepatic stellate cell | Acta 1, Des |
| Endothelial cell | Bmp2, Lyvel |
| Dendritic cell | Itgax, CD83, CD80 |
| monocyte | Ly6c1, Ly6c2 and Itgam |
| Kupffer cell | Clec4f, Adgre1 |
| T cell | CD3d, CD3e, CD3g, CD4, CD8a |
| B cell | CD22 |
| granulocyte | Itgam and Ly6g |
## Cell developmental trajectory analysis
Using Monocle2, we inferred the cell lineage trajectory of the MMD system. We first used the transcript count data (e.g. UMI) and created an object with the parameter ‘‘expression Family = negbinomial. Size ()’’ following the Monocle2 tutorial. The ‘‘differentialGeneTest’’ function was used to derive differentially expressed genes (DEGs) from each cluster, and genes with a q-value < 0.05 were used to order the cells in pseudotime analysis. Furthermore, DEGs along the pseudotime were detected and analyzed using the ‘‘differentialGeneTest’’ function after the cell trajectories were constructed.
## Preparation and pathological evaluation of human and mouse liver slices
Human paraffin-embedded liver tissues were obtained from the Department of Pathology at Peking University People’s Hospital. In this study, 6 liver tissue samples from biopsy-proven NASH patients with fibrosis and 3 samples from healthy donors (age ≥18 years) were eligible and included. Patients with viral hepatitis, alcoholic liver disease, drug-induced liver disease, autoimmune liver disease, cholestatic liver disease or hereditary metabolic liver disease were not ineligible. We did not assess whether subjects were male or female because sex was not the analysis object in this study. We did not check for sample sizes using a power analysis because our study does not report statistics on between groups or within group variables. Informed consent forms were obtained from all participants. This study was approved by the Ethics Committee of Peking University People’s Hospital (2022PHB088-001) and conformed to the ethical guidelines of the 1975 Declaration of Helsinki.
Each sample (human and mouse) was sectioned at a thickness of 4 μm for histological assessment and immunohistochemistry (IHC). Using hematoxylin and eosin (H&E) and Sirius Red (SR) staining, liver histology for all participants was evaluated by two specialized pathologists who were blinded to the patient and mouse details according to the NASH Clinical Research Network (NASH CRN) System (steatosis was scored from 0-3, ballooning: 0-2, lobular inflammation: 0-3, portal inflammation: 0-2, and liver fibrosis: 0 to 4).
## Second harmonic generation/two-photon excitation fluorescence
Images of unstained sections of liver tissues were acquired using a Genesis system (HistoIndex Pte. Ltd., Singapore). Collagen in liver tissues was visualized by SHG microscopy, and TPEF microscopy was used to visualize the other cell structures.
## Multiplexed immunohistochemical staining
Multiplex staining of paraffin-embedded liver tissues was performed using a PANO 7-plex IHC kit (Cat# 0004100100, Panovue). Fmnl1 (1:200, Novus Cat# NBP1-88460, RRID: AB_11040849), Myh9 (1;3000, Abcam Cat# ab238131, RRID: AB_2924880), CD11c (1:150, CST Cat# 97585, RRID: AB_2800282), Ly6C (1:50, Abcam Cat# ab54223, RRID: AB_881384), CLEC4F (1:50, R and D System Cat# MAB2784, RRID: AB_2081338), CD11c (1:800, CST #45581, RRID: AB_2799286), CD68 (1:800, Abcam Cat# ab955, RRID: AB_307338) and CCR2 (1:500, Origene Cat# TA337218, RRID: AB_2924881) antibodies were sequentially applied, followed by horseradish peroxidase-conjugated secondary antibody incubation and tyramide signal amplification (TSA). After all antigens above have been labeled, nuclei were stained with 4’-6’-diamidino-2-phenylindole (DAPI, Sigma-Aldrich, Missouri, USA, Cat# D9542). The stained slides were scanned to obtain multispectral images using the Mantra System (PerkinElmer) and analyzed using InForm image analysis software (version 2.4, PerkinElmer).
## Cell culture, detection, and analysis
On the one hand, THP1 cells were seeded 1x106 cells per well in six-well plate and treated with palmitate (PA,1 mM, Sigma-Aldrich, USA). Meanwhile, different siRNA-lipofectamine™ 3000 (Cat# L3000015, Invitogen) mixture was added into cell culture medium. After 48h, total RNA of THP1 cells were extracted to detect the levels of Jun, spp1, Socs3 and Rac1. On the other hand, THP1 cells were seeded 1x106 cells per well on six-well upper trans-well insert (0.4µM), LX2 cells were seeded 1x105 cells in the lower wells of trans-well plates and attached overnight. After 24h, THP1 cells were washed with phosphate-buffered saline solution (PBS), and treated with PA (1 mM) and different siRNA-lipofectamine™ 3000 mixture, each upper insert was transferred to a lower plate containing the LX2 cells. Thereafter, THP1 cells and LX2 cells were co-cultured in serum-free medium for 48 hours. Total RNA of LX2 cells were extracted to detect the levels of α-smooth muscle actin (SMA), collagen type I alpha 1 (Col1a1) and Fibronectin (Fn). The sequences of different siRNA were listed in Supplementary Table 1.
## Statistical analysis
Statistical analysis was performed using R 4.1.2 (Vienna, Austria), SPSS 20.0 (SPSS, Chicago, IL, United States) and Graphpad Prism 5.0 (Graphpad Software Inc., San Diego, CA, United States). Comparisons of cell distribution and gene expression among different groups were performed using one-way ANOVA. Comparisons of mRNA levels between two coculture groups were analyzed using the Mann−Whitney U test. Two-sided p-values less than 0.05 were considered significant.
## NASH mice showed obvious liver steatosis, inflammation, and fibrosis
Compared with mice fed a normal diet, two NASH mouse models with fibrosis developed pronounced obesity. In response to HFD/WD feeding, fructose water and CCl4, NASH mice showed significantly increased body weight and serum alanine aminotransferase (ALT), aspartate aminotransferase (AST) and triglyceride (TG) levels, especially in WD+F+CCl4 mice (Figure 1A). In addition, by HE and SR staining evaluation, there was obvious macrovesicular and microvesicular steatosis (score 2-3), lobular inflammation (score 2-3), and lobular and portal fibrosis (score 3-4) in NASH mice (Figure 1B) after constructed for 16 weeks. Although there was no statistical difference, we observed obvious collagen deposition and steatosis signals in images from SHG/TPEF scanning of NASH mice (Figure 1B and Supplementary Figure 1).
**Figure 1:** *Metabolic and pathological characteristics of NASH mice with fibrosis. (A) Body weight, ALT, AST and TG levels and (B) Hematoxylin and eosin (H&E), Sirius Red staining and SHG/TPEF scanning images in control, HFD+F+CCl4 and WD+F+CCl4 mice. ps: ALT, alanine aminotransferase; AST, aspartate aminotransferase; and TG, triglyceride.*
## ScRNA-seq profiling of liver cells in NASH mice with fibrosis
To construct a liver cell atlas in NASH mouse models, we performed cell classification and marker gene identification. After the filtration of low-quality cells, there were 14956 cells in the control group, 31956 cells in the HFD+F+CCl4 group, and 33562 cells in the WD+F+CCl4 group. The requirements for high-quality cells in this study are as follows: 1. The number of genes identified in a single cell (500 Inf); 2. The total number of UMIs in a single cell was less than Inf; 3. The percentage of mitochondrial gene expression in a single cell was less than $25\%$. In addition, genes were filtered to retain genes expressed in at least one cell. A total of 27 cell clusters were identified and visualized using the T-distributed stochastic neighbor embedding (t-SNE) method in control mice and NASH mice with fibrosis (Supplementary Figures 2A, B). According to the representative markers of live cells in Table 1, we identified 10 different cells (Figure 2A). The parenchymal cells of the liver are mainly hepatocytes, and nonparenchymal cells consist of cholangiocytes, hepatic stellate cells (HSCs), endothelial cells, DCs, monocytes, Kupffer cells, T cells, B cells, and granulocytes (Figures 2A, B and Supplementary Figure 3). All of these cell subtypes were shared among control mice and NASH mice with fibrosis, however, at different proportions. Compared with those of control mice, higher proportions of monocytes, Kupffer cells and DCs were observed in HFD+F+CCl4 and WD+F+CCl4 mice (Figures 2C, D). In addition, the two NASH fibrosis models shared similar cell compositions and proportions (Figure 2C).
**Figure 2:** *ScRNA-seq profiling of liver cells from normal and NASH mice with fibrosis. (A) The annotation and color codes for the 10 identified liver cells. (B) The t-SNE plot shows the cell origins in the control, HFD+F+CCl4 mice and WFD+F+CCl4 mice. (C) Heatmap showing the expression of marker genes in the indicated cell types. (D) Histogram indicating the proportion of cells in the liver tissue of each mouse. ps: C01-02: control mice 01-02; H01-03: HFD+F+CCl4 mice 01-03; W01-03: WFD+F+CCl4 mice 01-03.*
## The components and function of the monocyte-macrophage-DC system in the livers of NASH mice with fibrosis
We next conducted unsupervised clustering of monocytes, Kupffer cells and DCs. A total of 10 clusters emerged within the monocyte-macrophage-DC system, including eight clusters of monocytes (Mono 1-Mono 8), one cluster of Kupffer cells and one cluster of DCs (Figures 3A–C), namely, the MMD system. Compared with those of the control mice, there were higher proportions of Mono 1-4, Mono 7, Mono 8, Kupffer cells and DCs and lower proportions of Mono 5 and Mono 6 in the two NASH mouse models with fibrosis (Figure 3D). Cells in the MMD system in the two NASH fibrosis models displayed high expression levels of C-Jun, spp1, Rac1 and Socs3 (Figures 3E, F). IHC staining verified the increased abundance of these markers in NASH mice with fibrosis (all $p \leq 0.05$, Figure 3G and Supplementary Figure 4). KEGG enrichment analysis showed increased phagosome and lysosome signaling pathways in Mono 2 and increased PPAR signaling pathways and fatty acid degradation in Mono 5 (Figures 3H, I). In vitro, Jun, Spp1, Rac1 or Socs3 in THP1 cells was knock-down (KD), and palmitic acid (PA) was used to induce lipogenesis in THP1 cells. After 48h, the mRNA levels of molecules in phagocytosis and lysosomal signal pathway (Ctsz, Psap, LAMP1) in THP1 cells decreased (Supplementary Figure 6).
**Figure 3:** *The components of the monocyte-macrophage cell-DC (MMD) system in normal and NASH mice with fibrosis. (A, B) t-SNE plot showing the subtypes of cells in the MMD system from normal and NASH mice with fibrosis. Each cluster is color-coded according to cell type and mouse group. (C) Heatmap showing the expression of the marker genes in each cell type of the MMD system. (D) Bubble map showing the proportion of each cell type in the MMD system from normal and NASH mice with fibrosis. (E, F) Volcano plot showing the differentially expressed genes in the MMD system between control and HFD+F+CCl4 mice/control and WD+F+CCl4 mice. (G) Representative images of IHC staining (Jun, SPP1, Rac1 and Socs3) in formalin-fixed paraffin-embedded (FFPE) tissue in control, HFD+F+CCl4 and WD+F+CCl4 mice. (H, I) Bubble plot showing the KEGG enrichment of specific pathways in monocyte 2 (Mono2) and Mono5. ps: C01-02: control mice 01-02; H01-03: HFD+F+CCl4 mice 01-03; W01-03: WFD+F+CCl4 mice 01-03.*
## The transition trajectory of cells in the MMD system of NASH mice with fibrosis
Having defined different subsets in the MMD system, we aimed to analyze the dynamic immune states, cell transitions and pathways that further guide the different fates of cells. First, a complete dataset of monocytes, Kupffer cells and DCs was extracted and analyzed with Monocle2, which compare all single-cell transcriptomes in multidimensional space using machine-learning algorithms and orders cells along a path representing a developmental trajectory in theoretical time, known as pseudotime. Outcomes of pseudotime analysis showed that Mono5, Mono6 and Mono7 were at the beginning of the trajectory path, whereas Mono1, Mono2, Mono4, Kupffer cells and DCs were at the terminal state of the trajectory path. Surprisingly, Mono5 and Mono7 were primarily distributed in control mice, with few cells identified at the start of transition trajectory, whereas Kupffer cells and Mono1-4 were primarily distributed in the two NASH mouse models with fibrosis and mainly at the end of the transition path (Figures 4A–C).
**Figure 4:** *The transition trajectory and regulatory networks of cells in the MMD system. (A, B) Pseudotime-ordered analysis of cells in the MMD system from control and NASH mice. (C) Pseudotime plot showing the transition trajectory of Kupffer cells, DCs and 8 monocytes. (D) Heatmap showing the dynamic changes in gene expression along the pseudotime axis. (E-J) Characteristic genes in the select signaling pathway that are differentially expressed across the pseudotime of cells in the MMD system. ps: C, control mice; H, HFD+F+CCl4 mice; W, WFD+F+CCl4 mice.*
According to the ordered transition trajectory, we further performed DEG analysis to explore changes in the function of pseudotime and identified four broad patterns of gene expression dynamics (Figure 4D). Group 1 genes were upregulated along the trajectory early and maintained continuously, and a number of factors within carbohydrate metabolic processes were enriched in this group (Figure 4E). Group 2 genes represented a series of genes associated with lipid metabolic processes. *These* genes were at higher levels at the beginning of the trajectory and then upregulated significantly and downregulated at the end of the trajectory, which was consistent with the change in liver steatosis in the progression of NASH (Figure 4F). Group 3 encompassed genes associated with oxidative phosphorylation, which was downregulated consistently along the trajectory (Figure 4G). Groups 4 and 5 were enriched for genes related to lysosome and phagosome and autophagy, which increased slightly during the trajectory (Figures 4H, I). Group 6 included genes associated with collagen production, which increased in expression later along the trajectory path in accordance with the progression of fibrosis in NASH (Figure 4J).
## The upregulated expression of Fmnl1 and Myh9 in the MMD system is associated with collagen production in NASH mice with fibrosis
By TSA-IHC, the expression levels of Fmnl1 and Myh9 in liver sections were increased in the two NASH mouse models with fibrosis compared with control mice. The number of MMD system cells (Ly6c/CLEC4F/CD11c+ cells) in the two NASH models was also increased (all $p \leq 0.05$, Figure 5 and Supplementary Figure 5). Furthermore, Fmnl1 and Myh9 were mainly expressed in the MMD system. In the livers of NASH mice with fibrosis, the expression levels of Fmnl1 and Myh9 in the MMD system were associated with the distribution of collagen detected by SHG (Figure 5). In the human liver, the expression levels of Fmnl1 and Myh9 in the MMD system (CCR2/CD68/CD11c+ cells in humans) were also increased with collagen deposition, and the expression levels of Fmnl1 and Myh9 increased with aggravation of the liver fibrosis stage (all $p \leq 0.05$, Figure 6 and Supplementary Figure 5). In order to validate the function of Fmnl1 and Myh9 in MMD system, in vitro, Fmnl1 or Myh9 in THP1 cells was knock-down (KD), and palmitic acid (PA) was used to induce lipogenesis of THP1 cells. Then THP1 cells were co-cultured with hepatic stellate cell line (LX2 cells). After 48h co-culture, compared with controls, in the Fmnl1 or Myh9 KD group, the mRNA levels of collagen related markers (α-SMA, Col1a1, Fibronectin) in LX2 cells were downregulated (Supplementary Figure 7).
**Figure 5:** *Fmnl1 and Myh9 expression in MMD system cells and collagen deposition in livers from control mice and NASH mice with fibrosis. ps: MMD: monocyte-macrophage-DC (MMD); DAPI: cell nucleus, Fmnl1: Fmnl1+ cells (white arrow), Myh9: Myh9+cells (white arrow); Ly6C/CLEC4F/CD11c/Fmnl1: Fmnl1expression in MMD system cells (white arrow); Ly6C/CLEC4F/CD11c/Myh9: Myh9 expression in MMD system cells (white arrow); SHG, collagen detected by second harmonic generation.* **Figure 6:** *Fmnl1 and Myh9 expression in MMD system cells and collagen deposition in livers from healthy controls and NASH patients with fibrosis. ps: MMD: monocyte-macrophage-DC (MMD); F1-2: Fibrosis stage 1-2, F3-4: Fibrosis stage 3-4. DAPI: cell nucleus, Fmnl1: Fmnl1+ cells (white arrow), Myh9: Myh9+cells (white arrow); CCR2/CD68/CD11c/Fmnl1: Fmnl1 expression in MMD system cells (white arrow); CCR2/CD68/CD11c/Myh9: Myh9 expression in MMD system cells (white arrow); SHG, collagen detected by second harmonic generation.*
## Discussion
Monocytes, macrophages and DCs are crucial nonparenchymal cells in the liver, and these three immune cells have the same origin during NASH [17]. In our study, they were collectively named the MMD system. During the pathogenesis of NASH, monocytes migrate from peripheral blood and transform into macrophages and/or DCs in the liver, which participate in the progression of liver fibrosis [14, 15, 17, 18]. In this study, using scRNA-seq, we found that the number and distribution of monocytes, Kupffer cells and DCs changed significantly during NASH fibrosis. In the process of NASH fibrosis, some types of monocytes in the MMD system transformed into new monocytes, macrophages and DC cells, and Fmnl1 and Myh9 levels in the MMD system were also significantly increased at the end of the trajectory path, which was associated with the deposition of liver collagen in NASH mice and patients with fibrosis.
Compared with control mice, the numbers and types of monocytes, Kupffer cells and DCs in NASH mice with fibrosis were increased, especially monocytes. Similarly, by CITE-seq, researchers found that the NAFLD mouse model showed diverse macrophage subtypes [19]. Seidman et al. found that there were five major macrophage subsets in the liver of NASH mice, including normal KCs, NASH KCs, recruited macrophages (RM), Ly6Chi-RM and Ly6Clo-RM [20]. During liver injury, whether acute or chronic injury, a large number of monocytes are recruited to sites of hepatic injury, at this time, monocytes can differentiate into macrophages and become the main macrophage population, which promotes the activation of hepatic stellate cells (HSCs) to become myofibroblasts and contributes to NASH fibrogenesis (21–24). Furthermore, DCs in the liver also appear to be involved in liver fibrosis in the progression of NAFLD [25]. During fibrosis induced by TAA and recombinant leptin, the number of DCs were elevated up to 7-fold, producing redundant IL-6 and TNFα and activating HSCs though TNFα and/or direct cell contact [26]. Rahman AH et al. proposed that DCs regulate the number and activity of cells (e.g. natural killer (NK) cells and CD8+ cells) which involved in fibrosis progression and tissue remodeling [27].
We found that NASH-induced genes in cells from the MMD system were highly enriched for the pathways responsible for fatty acid catabolism (adipor1), extracellular matrix (ECM) remodeling (Rac1, TGFβ1) and immunoregulation (Jun, Socs3, SPP1). By IHC, the expression levels of SPP1, Jun, Socs3 and Rac1 in NASH mice with fibrosis were increased significantly compared with those in control mice. Previous studies have shown that the expression of SPP1 was positively correlated with liver steatosis, inflammation, fibrosis and insulin resistance in obese individuals or mice [28, 29]. Multiple signaling pathways, such as the extrinsic apoptosis and fibroblast proliferation, which medicated by SPP1 could regulate the progression of NASH [30]. As we all known, c-*Jun is* an immediate early gene that is regulated by the N-terminal kinase of c-Jun (JNK) [31]. It was reported that the expression level of c-Jun elevated with the progression from liver steatosis to NASH [32]. In human adipocytes, JNK2 was proven to be involved in fatty acid synthesis through regulating SREBP-1c [33]. Rac1, the small GTP-binding protein, is required for saturated fatty acids (SFA)-stimulated MLK3-dependent JNK activation in hepatocytes [34]. SOCS3, as a suppressor of cytokine signaling, is considered to promote insulin resistance by inhibiting insulin and leptin signaling during the inflammatory response [35]. However, the relationship between the expression of these genes and NASH fibrosis was not clear. In this study, we proposed for the first time that they are related to fibrosis in NASH (Supplementary Figure 8).
During our research, Fmnl1 was upregulated in the terminal path of the transition trajectory, and its expression was positively associated with collagen deposition. Fmnl1 is an essential component of macrophage podosomes [36]. In the liver, moderate to strong staining of Fmnl1 was shown in macrophages, whereas hepatocytes and biliary epithelium remained negative [37, 38]. The expression level of Fmnl1 was associated with cell phagocytosis, adhesion and podosome dynamics, migration and survival of macrophages (37, 39–42). Targeted suppression of Fmnl1 resulted in decreases in macrophage adhesion and migration [36, 42]. Additionally, the expression of Fmnl1 was specifically upregulated during monocyte differentiation to macrophages [36]. Therefore, combined the results of this study, we speculated that the mechanism of Fmnl1 participating in NASH fibrosis might be that the increased expression of Fmnl1 facilitated the differentiation of monocytes to macrophages and the dynamic changes of podosomes, promoted the release of pro-fibrogenic cytokines and activated HSCs, thereby promoted the production of liver collagen in NASH.
Myh9 was also upregulated in the terminal path of the transition trajectory, and its expression was positively associated with NASH fibrosis. Myh9, the gene which encodes the heavy chain (MHCII) of non-muscle myosin II A (NMII-A), is involved in cell migration, adhesion, division, polarity and morphogenesis and signal transduction [43, 44]. Previous studies have suggested that Myh9 was highly upregulated in the process of neutrophil differentiation and played a necessary and fundamental role in neutrophil trafficking (45–47). Furthermore, in hepatocellular carcinoma (HCC), targeting Myh9 improved the survival of HCC-bearing mice markedly and promoted sorafenib sensitivity of HCC cells [48]. In our study, we proposed for the first time that Myh9 was associated with liver fibrosis in NASH; however, the mechanism was unclear. It is possible that Myh9 participates in the differentiation and migration of monocytes and then communicates with HSCs to promote collagen formation.
Although our study used scRNA-seq data and our own samples to verify the correlation between characteristic genes in MMD system and the progression of NASH fibrosis, there were some limitations to be considered. First, the mechanism of Fmnl1 and *Myh9* genes which involved in NASH fibrosis was still unclear. In the future, we will try our best to explore the signaling pathway of these two genes in collagen production. Second, due to the limitation of our own liver samples, the mechanism of the interaction between Fmnl1 and Myh9 in the NASH progression has not been fully clarified. Third, in this research, only cross-sectional samples were used for the validation of Fmnl1 and Myh9 expression. It was necessary for us to include cohort samples of NASH fibrosis to deeply analyze the clinical value of the characteristic genes in MMD systems.
## Conclusions
During NASH, monocytes, macrophages and DCs in the MMD system are diverse and involved in fibrogenesis. The expression of Fmnl1 and Myh9 in the transition trajectory of the MMD system was related to liver collagen deposition and fibrosis progression in NASH.
## Data availability statement
The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/ Supplementary Material.
## Ethics statement
This study was approved by the Ethics Committee of Peking University People’s Hospital (2022PHB088-001). The patients/participants provided their written informed consent to participate in this study. This study was approved by the Ethics Committee of Peking University People’s Hospital (2021PHE111).
## Author contributions
XW, HR, and FL contributed to conception and design of the study. ZW, BL, RJ, and YS constructed mouses models. RF, XC, RH, XL, and JY performed the experiments. XW, LW, HR, and FL analyzed and interpreted of data. XW and FL wrote the first draft of the manuscript. FL and HR made critical revision of the manuscript for important intellectual content. All authors contributed to manuscript revision, read, and approved the submitted version.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu.2023.1098056/full#supplementary-material
## References
1. Petroni ML, Brodosi L, Bugianesi E, Marchesini G. **Management of non-alcoholic fatty liver disease**. *BMJ* (2021) **372**. DOI: 10.1136/bmj.m4747
2. Ludwig J, Viggiano TR, McGill DB, Oh BJ. **Nonalcoholic steatohepatitis: Mayo clinic experiences with a hitherto unnamed disease**. *Mayo Clin Proc* (1980) **55**
3. Hui JM, Kench JG, Chitturi S, Sud A, Farrell GC, Byth K. **Long-term outcomes of cirrhosis in nonalcoholic steatohepatitis compared with hepatitis c**. *Hepatology* (2003) **38**. DOI: 10.1053/jhep.2003.50320
4. Tacke F, Weiskirchen R. **Non-alcoholic fatty liver disease (Nafld)/Non-alcoholic steatohepatitis (Nash)-related liver fibrosis: Mechanisms, treatment and prevention**. *Ann Transl Med* (2021) **9** 729. DOI: 10.21037/atm-20-4354
5. Vilar-Gomez E, Calzadilla-Bertot L, Wai-Sun Wong V, Castellanos M, Aller-de la Fuente R, Metwally M. **Fibrosis severity as a determinant of cause-specific mortality in patients with advanced nonalcoholic fatty liver disease: A multi-national cohort study**. *Gastroenterology* (2018) **155** 443-57 e17. DOI: 10.1053/j.gastro.2018.04.034
6. Diehl AM, Day C. **Cause, pathogenesis, and treatment of nonalcoholic steatohepatitis**. *N Engl J Med* (2017) **377**. DOI: 10.1056/NEJMra1503519
7. Angulo P, Kleiner DE, Dam-Larsen S, Adams LA, Bjornsson ES, Charatcharoenwitthaya P. **Liver fibrosis, but no other histologic features, is associated with long-term outcomes of patients with nonalcoholic fatty liver disease**. *Gastroenterology* (2015) **149** 389-97 e10. DOI: 10.1053/j.gastro.2015.04.043
8. Sanyal AJ, Harrison SA, Ratziu V, Abdelmalek MF, Diehl AM, Caldwell S. **The natural history of advanced fibrosis due to nonalcoholic steatohepatitis: Data from the simtuzumab trials**. *Hepatology* (2019) **70**. DOI: 10.1002/hep.30664
9. Taylor RS, Taylor RJ, Bayliss S, Hagstrom H, Nasr P, Schattenberg JM. **Association between fibrosis stage and outcomes of patients with nonalcoholic fatty liver disease: A systematic review and meta-analysis**. *Gastroenterology* (2020) **158** 1611-25 e12. DOI: 10.1053/j.gastro.2020.01.043
10. Friedman SL, Neuschwander-Tetri BA, Rinella M, Sanyal AJ. **Mechanisms of nafld development and therapeutic strategies**. *Nat Med* (2018) **24**. DOI: 10.1038/s41591-018-0104-9
11. Almeda-Valdes P, Aguilar Olivos NE, Barranco-Fragoso B, Uribe M, Mendez-Sanchez N. **The role of dendritic cells in fibrosis progression in nonalcoholic fatty liver disease**. *BioMed Res Int* (2015) **2015**. DOI: 10.1155/2015/768071
12. Gao CC, Bai J, Han H, Qin HY. **The versatility of macrophage heterogeneity in liver fibrosis**. *Front Immunol* (2022) **13**. DOI: 10.3389/fimmu.2022.968879
13. Gomez Perdiguero E, Klapproth K, Schulz C, Busch K, Azzoni E, Crozet L. **Tissue-resident macrophages originate from yolk-Sac-Derived erythro-myeloid progenitors**. *Nature* (2015) **518**. DOI: 10.1038/nature13989
14. Kazankov K, Jorgensen SMD, Thomsen KL, Moller HJ, Vilstrup H, George J. **The role of macrophages in nonalcoholic fatty liver disease and nonalcoholic steatohepatitis**. *Nat Rev Gastroenterol Hepatol* (2019) **16**. DOI: 10.1038/s41575-018-0082-x
15. Dominguez PM, Ardavin C. **Differentiation and function of mouse monocyte-derived dendritic cells in steady state and inflammation**. *Immunol Rev* (2010) **234** 90-104. DOI: 10.1111/j.0105-2896.2009.00876.x
16. Terkelsen MK, Bendixen SM, Hansen D, Scott EAH, Moeller AF, Nielsen R. **Transcriptional dynamics of hepatic sinusoid-associated cells after liver injury**. *Hepatology* (2020) **72**. DOI: 10.1002/hep.31215
17. Sutti S, Locatelli I, Bruzzi S, Jindal A, Vacchiano M, Bozzola C. **Cx3cr1-expressing inflammatory dendritic cells contribute to the progression of steatohepatitis**. *Clin Sci (Lond)* (2015) **129** 797-808. DOI: 10.1042/CS20150053
18. Yoshio S, Kanto T. **Macrophages as a source of fibrosis biomarkers for non-alcoholic fatty liver disease**. *Immunol Med* (2021) **44**. DOI: 10.1080/25785826.2020.1868664
19. Remmerie A, Martens L, Thone T, Castoldi A, Seurinck R, Pavie B. **Osteopontin expression identifies a subset of recruited macrophages distinct from kupffer cells in the fatty liver**. *Immunity* (2020) **53** 641-57 e14. DOI: 10.1016/j.immuni.2020.08.004
20. Seidman JS, Troutman TD, Sakai M, Gola A, Spann NJ, Bennett H. **Niche-specific reprogramming of epigenetic landscapes drives myeloid cell diversity in nonalcoholic steatohepatitis**. *Immunity* (2020) **52** 1057-74 e7. DOI: 10.1016/j.immuni.2020.04.001
21. Krenkel O, Puengel T, Govaere O, Abdallah AT, Mossanen JC, Kohlhepp M. **Therapeutic inhibition of inflammatory monocyte recruitment reduces steatohepatitis and liver fibrosis**. *Hepatology* (2018) **67**. DOI: 10.1002/hep.29544
22. Ju C, Tacke F. **Hepatic macrophages in homeostasis and liver diseases: From pathogenesis to novel therapeutic strategies**. *Cell Mol Immunol* (2016) **13**. DOI: 10.1038/cmi.2015.104
23. Weiskirchen R, Tacke F. **Liver fibrosis: From pathogenesis to novel therapies**. *Dig Dis* (2016) **34**. DOI: 10.1159/000444556
24. Wang T, Ma C. **The hepatic macrophage pool in Nash**. *Cell Mol Immunol* (2021) **18**. DOI: 10.1038/s41423-021-00690-z
25. Handa P, Kowdley KV. **Dendritic cells in Nash: Friend or foe**. *Ann Hepatol* (2013) **12**
26. Connolly MK, Bedrosian AS, Mallen-St Clair J, Mitchell AP, Ibrahim J, Stroud A. **In liver fibrosis, dendritic cells govern hepatic inflammation in mice**. *J Clin Invest* (2009) **119**. DOI: 10.1172/JCI37581
27. Rahman AH, Aloman C. **Dendritic cells and liver fibrosis**. *Biochim Biophys Acta* (2013) **1832** 998-1004. DOI: 10.1016/j.bbadis.2013.01.005
28. Kiefer FW, Neschen S, Pfau B, Legerer B, Neuhofer A, Kahle M. **Osteopontin deficiency protects against obesity-induced hepatic steatosis and attenuates glucose production in mice**. *Diabetologia* (2011) **54**. DOI: 10.1007/s00125-011-2170-0
29. Bertola A, Deveaux V, Bonnafous S, Rousseau D, Anty R, Wakkach A. **Elevated expression of osteopontin may be related to adipose tissue macrophage accumulation and liver steatosis in morbid obesity**. *Diabetes* (2009) **58**. DOI: 10.2337/db08-0400
30. He W, Huang C, Zhang X, Wang D, Chen Y, Zhao Y. **Identification of transcriptomic signatures and crucial pathways involved in non-alcoholic steatohepatitis**. *Endocrine* (2021) **73** 52-64. DOI: 10.1007/s12020-021-02716-y
31. Wang J, Ma J, Nie H, Zhang XJ, Zhang P, She ZG. **Hepatic regulator of G protein signaling 5 ameliorates nonalcoholic fatty liver disease by suppressing transforming growth factor beta-activated kinase 1-C-Jun-N-Terminal Kinase/P38 signaling**. *Hepatology* (2021) **73**. DOI: 10.1002/hep.31242
32. Schulien I, Hockenjos B, Schmitt-Graeff A, Perdekamp MG, Follo M, Thimme R. **The transcription factor c-Jun/Ap-1 promotes liver fibrosis during non-alcoholic steatohepatitis by regulating osteopontin expression**. *Cell Death Differ* (2019) **26**. DOI: 10.1038/s41418-018-0239-8
33. Ito M, Nagasawa M, Omae N, Tsunoda M, Ishiyama J, Ide T. **A novel Jnk2/Srebp-1c pathway involved in insulin-induced fatty acid synthesis in human adipocytes**. *J Lipid Res* (2013) **54**. DOI: 10.1194/jlr.M031591
34. Sharma M, Urano F, Jaeschke A. **Cdc42 and Rac1 are major contributors to the saturated fatty acid-stimulated jnk pathway in hepatocytes**. *J Hepatol* (2012) **56**. DOI: 10.1016/j.jhep.2011.03.019
35. Jiang ZY, Zhou Y, Zhou L, Li SW, Wang BM. **Identification of key genes and immune infiltrate in nonalcoholic steatohepatitis: A bioinformatic analysis**. *BioMed Res Int* (2021) **2021**. DOI: 10.1155/2021/7561645
36. Mersich AT, Miller MR, Chkourko H, Blystone SD. **The formin Frl1 (Fmnl1) is an essential component of macrophage podosomes**. *Cytoskeleton (Hoboken)* (2010) **67**. DOI: 10.1002/cm.20468
37. Gardberg M, Heuser VD, Iljin K, Kampf C, Uhlen M, Carpen O. **Characterization of leukocyte formin Fmnl1 expression in human tissues**. *J Histochem Cytochem* (2014) **62**. DOI: 10.1369/0022155414532293
38. Krainer EC, Ouderkirk JL, Miller EW, Miller MR, Mersich AT, Blystone SD. **The multiplicity of human formins: Expression patterns in cells and tissues**. *Cytoskeleton (Hoboken)* (2013) **70**. DOI: 10.1002/cm.21113
39. Schuster IG, Busch DH, Eppinger E, Kremmer E, Milosevic S, Hennard C. **Allorestricted T cells with specificity for the Fmnl1-derived peptide Pp2 have potent antitumor activity against hematologic and other malignancies**. *Blood* (2007) **110**. DOI: 10.1182/blood-2006-11-058750
40. Yayoshi-Yamamoto S, Taniuchi I, Watanabe T. **Frl, a novel formin-related protein, binds to rac and regulates cell motility and survival of macrophages**. *Mol Cell Biol* (2000) **20**. DOI: 10.1128/MCB.20.18.6872-6881.2000
41. Colon-Franco JM, Gomez TS, Billadeau DD. **Dynamic remodeling of the actin cytoskeleton by Fmnl1gamma is required for structural maintenance of the golgi complex**. *J Cell Sci* (2011) **124**. DOI: 10.1242/jcs.083725
42. Miller MR, Blystone SD. **Human macrophages utilize the podosome formin Fmnl1 for adhesion and migration**. *Cellbio (Irvine Calif)* (2015) **4** 1-11. DOI: 10.4236/cellbio.2015.41001
43. Vicente-Manzanares M, Ma X, Adelstein RS, Horwitz AR. **Non-muscle myosin ii takes centre stage in cell adhesion and migration**. *Nat Rev Mol Cell Biol* (2009) **10**. DOI: 10.1038/nrm2786
44. Asensio-Juarez G, Llorente-Gonzalez C, Vicente-Manzanares M. **Linking the landscape of Myh9-related diseases to the molecular mechanisms that control non-muscle myosin ii-a function in cells**. *Cells* (2020) **9**. DOI: 10.3390/cells9061458
45. Marigo V, Nigro A, Pecci A, Montanaro D, Di Stazio M, Balduini CL. **Correlation between the clinical phenotype of Myh9-related disease and tissue distribution of class ii nonmuscle myosin heavy chains**. *Genomics* (2004) **83**. DOI: 10.1016/j.ygeno.2003.12.012
46. Shin JW, Buxboim A, Spinler KR, Swift J, Christian DA, Hunter CA. **Contractile forces sustain and polarize hematopoiesis from stem and progenitor cells**. *Cell Stem Cell* (2014) **14** 81-93. DOI: 10.1016/j.stem.2013.10.009
47. Zehrer A, Pick R, Salvermoser M, Boda A, Miller M, Stark K. **A fundamental role of Myh9 for neutrophil migration in innate immunity**. *J Immunol* (2018) **201**. DOI: 10.4049/jimmunol.1701400
48. Lin X, Li AM, Li YH, Luo RC, Zou YJ, Liu YY. **Silencing Myh9 blocks hbx-induced Gsk3beta ubiquitination and degradation to inhibit tumor stemness in hepatocellular carcinoma**. *Signal Transduct Target Ther* (2020) **5**. DOI: 10.1038/s41392-020-0111-4
|
---
title: A novel prognostic scoring model based on copper homeostasis and cuproptosis
which indicates changes in tumor microenvironment and affects treatment response
authors:
- Yun-Long Ma
- Ya-Fei Yang
- Han-Chao Wang
- Chun-Cheng Yang
- Lun-Jie Yan
- Zi-Niu Ding
- Bao-Wen Tian
- Hui Liu
- Jun-Shuai Xue
- Cheng-Long Han
- Si-Yu Tan
- Jian-Guo Hong
- Yu-Chuan Yan
- Xin-Cheng Mao
- Dong-Xu Wang
- Tao Li
journal: Frontiers in Pharmacology
year: 2023
pmcid: PMC9998499
doi: 10.3389/fphar.2023.1101749
license: CC BY 4.0
---
# A novel prognostic scoring model based on copper homeostasis and cuproptosis which indicates changes in tumor microenvironment and affects treatment response
## Abstract
Background: Intracellular copper homeostasis requires a complex system. It has shown considerable prospects for intervening in the tumor microenvironment (TME) by regulating copper homeostasis and provoking cuproptosis. Their relationship with hepatocellular carcinoma (HCC) remains elusive.
Methods: In TCGA and ICGC datasets, LASSO and multivariate Cox regression were applied to obtain the signature on the basis of genes associated with copper homeostasis and cuproptosis. Bioinformatic tools were utilized to reveal if the signature was correlated with HCC characteristics. Single-cell RNA sequencing data analysis identified differences in tumor and T cells’ pathway activity and intercellular communication of immune-related cells. Real-time qPCR analysis was conducted to measure the genes’ expression in HCC and adjacent normal tissue from 21 patients. CCK8 assay, scratch assay, transwell, and colony formation were conducted to reveal the effect of genes on in vitro cell proliferation, invasion, migration, and colony formation.
Results: We constructed a five-gene scoring system in relation to copper homeostasis and cuproptosis. The high-risk score indicated poor clinical prognosis, enhanced tumor malignancy, and immune-suppressive tumor microenvironment. The T cell activity was markedly reduced in high-risk single-cell samples. The high-risk HCC patients had a better expectation of ICB response and reactivity to anti-PD-1 therapy. A total of 156 drugs were identified as potential signature-related drugs for HCC treatment, and most were sensitive to high-risk patients. Novel ligand-receptor pairs such as FASLG, CCL, CD40, IL2, and IFN-Ⅱ signaling pathways were revealed as cellular communication bridges, which may cause differences in TME and immune function. All crucial genes were differentially expressed between HCC and paired adjacent normal tissue. Model-constructed genes affected the phosphorylation of mTOR and AKT in both Huh7 and Hep3B cells. Knockdown of ZCRB1 impaired the proliferation, invasion, migration, and colony formation in HCC cell lines.
Conclusion: We obtained a prognostic scoring system to forecast the TME changes and assist in choosing therapy strategies for HCC patients. In this study, we combined copper homeostasis and cuproptosis to show the overall potential risk of copper-related biological processes in HCC for the first time.
## 1 Introduction
As an indispensable human body element, copper participates in various physiological and metabolic functions, including coagulation, oxidative metabolism, and hormone production (Bhattacharjee et al., 2017). There are inherent complex mechanisms in cells to maintain copper homeostasis. Copper homeostasis disorders involve a wide range of diseases, including degenerative neurological diseases (Bisaglia and Bubacco, 2020), metabolic diseases (Lowe et al., 2017), cardio-cerebrovascular diseases (Fukai et al., 2018), and tumors (Oliveri, 2022). The elevated copper level has been found in various solid tumors, promoting proliferation, invasion, migration, and angiogenesis (Oliveri, 2022). Excess copper caused by copper homeostasis disorder such as transporter mutation leads to programmed cell death, which was recently identified as cuproptosis (Tsvetkov et al., 2022). Cuproptosis is induced via copper-dependent protein fatty acylation, accompanied by tricarboxylic acid cycle changes, and influenced by mitochondrial function (Tsvetkov et al., 2022). Copper homeostasis is not only related to the drug resistance of traditional chemotherapeutic drugs but also can affect specific immune checkpoints and change the anti-tumor immune response (da Silva et al., 2022; Voli et al., 2020). Given the vital role of copper in cancer, copper ion carriers and copper complexes have been developed as anticancer drugs (Chen et al., 2006; Cen et al., 2004; O'Day et al., 2009; O'Day et al., 2013; Tsang et al., 2020). Still, the metabolic heterogeneity of different cancers is the main obstacle to their application. To achieve a more stable and reliable anticancer effect by affecting the copper homeostasis of tumor cells, the corresponding receptors of specific types of tumor cells should be targeted (da Silva et al., 2022). As key players in this novel cell death form, the genes related to copper homeostasis and cuproptosis possibly be promising cancer therapy targets. The specific mechanism of cuproptosis was covered; nevertheless, for further targeted drug development and clinical application, understanding different targets of copper homeostasis and cuproptosis in various tumors is still far from sufficient.
Although there are a variety of measures for diagnosis and treatment, mortality and prognosis are still poor for HCC because of a wide range of predisposing factors and unobvious early clinical manifestations (Hartke et al., 2017). Compared with mature traditional therapy, non-invasive diagnosis and targeted therapy are still challenging. Recently, patients’ prognosis and life quality have been improved by systemic therapies (Llovet et al., 2021). The disorder of copper homeostasis can cause cuproptosis, which has great potential in developing new therapies for HCC. Copper content is closely linked to liver cirrhosis and HCC (Zhang et al., 1994). Ionizing radiation can increase the radiation resistance caused by intracellular copper and inhibit ferroptosis and the degradation of HIF1α (Yang et al., 2022). Copper-binding enzyme LOXL4 causes the immunosuppressive phenotype of macrophages and promotes the progression of HCC (Tan et al., 2021). Given the critical role of copper in HCC, new copper complexes for specific targets have been developed. A new copper complex can induce cell senescence by inhibiting methionine cycle metabolism, which depends on mitochondrial carrier protein (Jin et al., 2020). Another targeted nanoparticle containing copper complex effectively reduces the growth of mice’s HCC (Xu et al., 2020). The evidence above suggests that genes related to copper homeostasis and cuproptosis have remarkable research prospects in expanding systemic therapy and improving patient prognosis in clinical application.
This study developed a novel prognostic scoring system that incorporates genes related to copper homeostasis and cuproptosis to predict the clinical outcome of HCC patients. To demonstrate the predictive value of the signature, we explored the underlying mechanisms based on bulk and single-cell RNA sequencing data. Novel receptor-ligand pairs were proposed to help understand tumor-immune cell interactions and explain the differences in TME related to the signature. Finally, potential targeted and chemotherapeutic drugs were predicted for different scoring samples. Our predictive model showed great potential in identifying the risk of copper-related physiological processes and assisting in the therapy of HCC patients.
## 2.1 Acquisition of multiomics data
The following bulk RNA-sequencing expression profiles and corresponding clinical data were downloaded from the TCGA database (https://portal.gdc.com $$n = 377$$). Raw sequencing reads were aligned using the STAR aligner and expressed as fragments per million mapped reads (FPKM). Gene expression profiles were standardized using R (https://www.r-project.org/). Only patients with complete clinical information related to the analysis were retained. Training and testing groups were randomly assigned in a ratio of 1:1 among the patients. To establish an independent validation cohort, Clinical pathology and RNA-Seq mRNA expression data were obtained for 232 samples from the ICGC portal (https://dcc.icgc.org/projects/LIRI-JP). The UCSC Xena server was used to retrieve somatic mutations and methylation data for HCC (https://xenabrowser.net/). The GEO database was used to download data for single-cell RNA sequencing of primary HCC tissues (GSE149614, $$n = 10$$). “ Seurat” and “NormalizeData” R packages were used for the standardization of the single-cell RNA-Seq data. “ FingVariableGenes” R package was used for the identification of the top 3,000 highly variable genes. The determination of cell types was as shown in Supplementary Figure S1A (Malignant cell markers-GPC3, CD24, MDK, KRT18; Meyloid cell markers-CD68, AIF1, C1QA, TPSAB1; T cell markers-CD3D, CD3E, CD2; B cell markers-MZB1, MS4A1, CD79A; Fibroblast cell markers-COL1A2, COL3A1, ACTA2; Endothelial cell markers-FLT1, RAMP2, PLVAP).
## 2.2 Identification of genes related to copper homeostasis and cuproptosis
25 genes (SLC31A1, SLC31A2, ATOX1, PDHB, COX11, COX17, PDHA1, NLRP3, NFE2L2, CCS, MTF1, LIPT2, LIPT1, LIAS, GLS, GCSH, FDX1, DLST, DLD, DLAT, DBT, CDKN2A, ATP7B, ATP7A, SCO1) directly involved in copper death and copper homeostasis processes were obtained from previous studies (Bian et al., 2022; da Silva et al., 2022; Inesi, 2017; Tsvetkov et al., 2022). An analysis of the differential expression of these genes was conducted in HCC. To screen related genes, Pearson correlation analysis was conducted (correlation coefficient>0.4, $p \leq 0.001$). *Qualified* genes were associated with cuproptosis or copper homeostasis.
## 2.3 Development of the signature related to copper homeostasis and cuproptosis
With R package “glmnet,” genes associated with copper homeostasis and cuproptosis were screened using univariate cox regression. Then the least absolute shrinkage and selection operator (LASSO) Cox regression and multivariate Cox regression models were used to creating the copper metabolism and cuproptosis gene signature in the training cohort. Gene expression values and coefficients of crucial genes were multiplied to determine the score of each sample. The median value of the score determined high-risk and low-risk groups. ROC curves analysis and Kaplan-Meier survival analysis were conducted to evaluate the signature. Independent prognostic analysis was conducted to determine if the risk score affected survival in patients with HCC. A stratified clinical examination was performed according to the patient’s clinical pathological characteristics (age, gender, grading, staging). The “ggDCA” R package was used to analyze different diagnostic models.
## 2.4 Nomogram development and validation
Multivariate Cox regression results were used to develop the Nomogram model. The final model was chosen using the Akaike information criterion (AIC) as a backward selection criterion (Harrell et al., 1996). Nomogram validation was conducted using a calibration curve generated via regression analysis. The nomogram was developed following the nomogram guide (Iasonos et al., 2008).
## 2.5 Functional enrichment and genetic alterations analysis
KEGG and GO analyses were performed using the R package “clusterProfiler” (Yu et al., 2012). *The* genetic variation between groups of Risk Scores was analyzed using R package “Maftools.” The ssGSEA score was calculated with R package “GSVA,” which was also used for functional enrichment analysis in malignant cells and T cells of single-cell RNA-Seq data. MATH score was used to evaluate tumor heterogeneity (Mroz and Rocco, 2013).
## 2.6 Drug sensitivity prediction
Drug sensitivity in cancer was predicted using the Genomics of Drug Sensitivity in Cancer database (GDSC: https://www.cancerxgene.org). “ pRRophetic” R package was used to calculate half maximal inhibitory concentration (IC50) (Geeleher et al., 2014).
## 2.7 Immune profile analysis and cell communication
“Immunedeconv” R package was applied to evaluate the immune score (Sturm et al., 2020). VEGFB, TNFSF4, TNFRSF4, TNFRSF18, TIGIT, TGFB1, SELP, PDCD1, LAG3, IL1A, IL12A, IDO1, HMGB1, HAVCR2, EDNRB, CTLA4, CD276, CD274, CTLA4, BTLA, and ARG1 were chosen as immune checkpoints. The TIDE procedure was combined with subclass mapping and immunophenoscore (IPS) to calculate potential ICB responses (Kim et al., 2008). The IPS of HCC patients included in the analysis came from the TCIA database (https://www.tcia.at/home).
‘‘Celltalker’’ R package was applied to analyze crosstalk between malignant cells and immunocytes based on the single-cell RNA-Seq data.
## 2.8 Human tissues
Surgically resected HCC and normal adjacent tissue samples were obtained from twenty-one HCC patients at the Qilu Hospital of Shandong University (Jinan, China) and stored in liquid nitrogen. All HCC samples were confirmed through clinicopathological features. The hospital’s ethical committee approved the study, and each patient signed a written informed consent form.
## 2.9 qRT-PCR, Western blot, and immunohistochemistry
The cells were washed with PBS and lysed in RIPA buffer (Beyotime, CN) containing phosphatase inhibitors and protease inhibitors (Beyotime, CN) at the indicated time points. The BCA Protein Assay kit (Beyotime, CN) was used to determine protein lysate concentration. After centrifuging at 12,000×g for 15 min, the supernatant was mixed with the 5×SDS-PAGE loading buffer (Beyotime, CN), and boiled at 95°C for 5 min. A standard Western blot procedure was then followed. 0.2 um PVDF membrane was obtained from Thermo Fisher Scientific (Thermo Fisher Scientific, United States). Enhanced chemiluminescence was obtained from Thermo Fisher Scientific (Thermo Fisher Scientific, United States). Antibodies against AKT, p-AKT, mTOR, p-mTOR and GAPDH were obtained from Cell Signaling Technology (Cell Signaling Technology, CN).
Trizol reagent (Thermo Fisher Scientific, United States) was used to prepare total RNA from tissues or cells. PrimeScript™ RT Master Mix (Takara Bio, JP) was used for reverse transcription. qRT-PCR analysis was performed with the CFX Connect system (Bio-Rad, United States) and CharmQ SYBR qPCR Master Mix (Takara, Japan). Supplementary Table S3 lists the primers used in this study.
In accordance with standard protocols, immunohistochemistry was performed on HCC and adjacent normal tissue. Antibody against ZCRB1 was obtained from Thermo Fisher Scientific (Thermo Fisher Scientific, United States).
## 2.10 Cell lines and cell culture
Hep3B and Huh-7 cells were obtained from the Shanghai Cell Collection. DMEM with $10\%$ FBS and $1\%$ Penicillin-Streptomycin was used to culture the cells at 37°C with $5\%$ CO2.
## 2.11 Cell transfection and Cell Counting Kit-8 assay
siCDKN2A, siDLAT, siGEMIN2, siZCRB1, and siKLF9 were obtained from Ribobio (CN). As directed by the manufacturer, JetPRIME® transfection kit (BIOFIL, CN) was used for transfection. After 24 h, total RNA was extracted for qRT-PCR.
Huh7 or Hep3B cells were inoculated into 96-well plates 24 h after transfection at a density of 1,000 cells per well. Each group was replicated five times. Cell Counting Kit-8 (Dojindo, JP) was used for the measurement of cell proliferation. As the culture progressed, absorbance values were measured after 0, 24, 48, and 72 h.
## 2.12 Migration, invasion, and colony formation assay
The scratch assay was applied to evaluate cell migration and repair. After reaching $90\%$–$100\%$ confluency in wells of culture plates, cells were exposed to serum-free medium for 6 h, and each cultured well was scraped with a pipette tip in the same specification. Cells were washed with PBS to remove fragments. Microscope images of the same positions were acquired in after 0 and 30 h. Based on the percentage of wound closure area, cell migration was determined.
Transwell migration and invasion assays were conducted to evaluate the ability of cell migration and invasion. 24-well transwell chambers (Corning, United States) were used in the assay. For the invasion assay, matrigel (Corning, United States) was applied to the upper ventricle surface of the basement membrane of the transwell chamber. The insert was filled with 30,000 cells suspended in 150 ul serum-free serum before the assay. In the lower chamber, 700 ul medium containing $12\%$ fetal serum was added for chemotactic stimulation. Cells were cultured for 24 h for migration assays and 40 h for invasion assays. Then cotton swabs were used to remove cells from the surface of the membrane. Cultured cells were fixed with $100\%$ methanol and stained with $0.1\%$ crystal violet. Random visual fields of 3 different inserts were captured, and the number of cells was counted.
After inoculating 3000 cells per well, Huh7 and Hep3B cells were grown for 8 and 10 days respectively in 6 well plates in complete medium. Cultured cells were fixed with $100\%$ methanol and stained with $0.1\%$ crystal violet. Each well was counted for the number of colonies.
## 2.13 Statistical analysis
R packages and analysis methods were executed with R (version 4.0.3). Quantitative variables were evaluated using independent samples t-tests. The unpaired Wilcoxon rank sum test was applied for the gene difference significance. For categorical data, Chi-square tests were applied. The ROC curve and Kaplan-Meier model judged efficacy in predicting survival outcomes. The relationships between prognostic classification, survival outcomes, and other clinical parameters were revealed with the Cox proportional model. A p-value less than 0.05 indicates statistical significance. * means a p-value less than 0.05; ** means a p-value less than 0.01; *** means a p-value less than 0.001; **** means a p-value less than 0.0001. For multiple corrections, the Benjamini–Hochberg method was applied.
## 3.1 Identification of genes related to copper homeostasis and cuproptosis and development of prognostic signature
We sorted out 25 genes from previous studies that have been proven to participate in cuproptosis and the maintenance of copper homeostasis directly. *Most* genes ($\frac{22}{25}$) were differently expressed in HCC and normal tissues (Figure 1A). Given the crucial role of copper in cancer, the signature related to cuproptosis and copper homeostasis could assist in evaluating tumor microenvironment changes and other pathological processes in HCC induced by copper. A correlation analysis was carried out according to the coefficient, and 95 genes were screened.
**FIGURE 1:** *Identifying genes related to copper homeostasis and cuproptosis and development of the signature. (A) Expression of 25 genes in HCC and normal tissues. Unpaired Wilcoxon Rank Sum and Signed Rank Test was applied for difference significance analysis. (B) Ten-fold cross-validation of the LASSO Cox regression model’s tuning parameter (λ) selection. Based on the minimum criteria and the 1-SE criteria, vertical lines were drawn at the optimal values. (C) LASSO coefficient profiles of the prognostic genes. (D) The correlation between the five genes in the signature and the genes directly participating in copper homeostasis and cuproptosis. (E) The expression of the five hub genes in the TCGA cohort. (F) Differential expression of the hub genes between HCC and normal tissues. (G) Kaplan−Meier plots of CDKN2A, KLF9, DLAT, GEMIN2, and ZCRB1.*
Including the original set of genes and their related genes, one hundred twenty candidate genes were confirmed as genes related to copper homeostasis and cuproptosis [Supplementary Table S1 (S1)], which were input into a univariate COX analysis. A LASSO regression was conducted on the genes with prognostic significance. Five hub genes were obtained (CDKN2A, DLAT, KLF9, GEMIN2, ZCRB1) (Figures 1B, C). There is a strong correlation between hub genes and genes directly participating in copper homeostasis and cuproptosis (Figure 1D). The signature was developed with a multivariate Cox proportional model. Risk Score = −0.2629*KLF9+0.6633*ZCRB1 +0.3994*DLAT+ 0.2121* CDKN2A+0.7650*GEMIN2. Gene expression in the cohort was visualized using a heat map (Figure 1E). HCC and normal tissues expressed all five genes differently (Figure 1F). Kaplan-Meier survival analysis was used to verify their relationship with HCC prognosis (Figure 1G). Five hub genes were significantly different in expression between HCC and normal tissues, and their expression was correlated with prognosis.
## 3.2 Clinical prognostic validation of the signature
The overall survival (OS) of high-risk patients was briefer in all cohorts (Figure 2A). In the TCGA cohort, 1-year, 3-year, and 5-year AUC values were 0.746, 0.703, and 0.718. Compared with clinical features, the risk score has higher prediction accuracy (Figure 2B). In addition, the Progression Free Survival (PFS) of high-risk patients in the TCGA cohort was also shorter (Figure 2C). Clinical characteristics and risk scores of HCC patients were analyzed by univariate and multivariate Cox regression, demonstrating that prognosis was independently predicted by the risk score (Figure 2D). The correlation between the risk score and pathological characteristics was examined. T stage, TNM stage, and histological grade were significantly correlated with risk score (Table 1). In stratified clinical analysis, there were significant differences in OS between high-risk and low-risk patients in all subgroups (Figure 2E). As an independent external validation set, the ICGC dataset was processed using the same methodology as the TCGA dataset. AUC, pathological characteristics analysis, and Kaplan-*Meier analysis* of the ICGC dataset once again demonstrated the prognostic value of the signature (Figures 2A, B) (Table 1). According to the above results, the signature was associated with HCC progression.
**FIGURE 2:** *Clinical validation of the prognostic signature. (A) The Kaplan-Meier overall survival (OS) curves of the risk score in the training, testing, TCGA, and ICGC cohorts. (B) Receiver operating characteristic (ROC) curve based on the risk score and other clinicopathological features for predicting OS in HCC patients. (C) Kaplan-Meier Progression–free Survival (PFS) curve. (D) Univariate (left) and multivariate (right) Cox regression analyses. (E) Kaplan-Meier survival subgroup analysis stratified by clinical characteristics. Grouping criteria: age>60/≤60, gender, histological grade, TNM stage. (F) OS nomogram (G) Nomogram calibration for the OS nomogram.* TABLE_PLACEHOLDER:TABLE 1 *As a* means of facilitating the clinical application of prognostic signatures, a nomogram was constructed through the combination of traditional clinical information (age, tumor stage, tumor grade) and risk scores (Figure 2F). Based on the second-generation sequencing result of the patients, the overall survival can be estimated by combining pathological characteristics with the risk scores. The nomogram performed well at predicting according to the calibration curve (Figure 2G).
## 3.3 Distribution of model-constructing genes and risk score in UMAP
The risk score of ten single-cell sequencing samples was calculated according to the Cox proportional model above for further analysis [Supplementary Table S1 (S2)]. We divided the single-cell sequencing samples into high- and low-risk scoring groups, and there was a significant statistical difference between the two groups (Figure 3A). Given the excellent predictability of the risk score for clinical prognosis, we explored the distribution of genes participating in the risk model and the risk score distribution in the uniform manifold approximation and projection (UMAP) based on the single-cell RNA sequencing data. The cells with high-risk scores were mainly malignant (Figures 3B, C). All the genes involved in the model construction were expressed to a certain extent in malignant cells (Figure 3D). Different from other genes, increased expression of KLF9 is associated with a lower risk score and better prognosis for patients, while it’s mainly expressed in fibroblasts and endothelial cells (Figure 3D).
**FIGURE 3:** *Distribution of model genes and risk score in UMAP based on single-cell sequencing data (GSE149614). (A) Comparison of risk scores of different samples. (B) UMAP of 23,590 cells from primary HCC tumors of ten HCC patients. (C) Distribution of risk score in UMAP. (D) Distribution of model construction genes in UMAP.*
## 3.4 Functional enrichment analysis revealed risk score correlated with HCC malignant degree
To reveal the potential mechanism causing the clinical characteristic in HCC patients with different risk scores, a cut-off of a p-value of 0.05 and a |FC| > 2 was used for screening differentially expressed genes (DEGs) between high- or low-risk groups [Supplementary Table S1 (S3)]. An analysis of GO and KEGG was then conducted. Results showed cell proliferation-related biological processes enriched mostly (Figures 4A, B). Then we collected a set of genes in tumor-related pathways and calculated the enrichment scores for every patient using the ssGSEA method. The high-risk group showed significant upregulation of proliferation and cell cycle pathways, including G2M checkpoint, DNA replication, DNA repair, MYC targets, and PI3K/AKT/mTOR pathway, which was in agreement with the results of the KEGG and GO (Figure 4C). We also found the upregulation of cell response to hypoxia, which can lead to an increase in tumor invasiveness. The OCLR algorithm was subsequently applied to calculate mRNAsi (Malta et al., 2018). A higher mRNAsi score was found in the high-risk group, reflecting the loss of cell differentiation phenotype and acquisition of stem cell-like characteristics (Figure 4D). Based on single-cell sequencing data, we performed GSVA to analyze the pathway enrichment in HCC malignant cells of high- and low-risk samples (Figure 4E). A series of cancer-promoting pathways in the high-risk samples were upregulated, such as oxidative phosphorylation, MYC targets, and DNA repair. In contrast, low-risk samples showed increased activity of more cancer-inhibiting pathways, such as the P53 pathway, apoptosis process, and IL2-STAT5 signal pathway. The analysis results of bulk RNA-Seq and single-cell RNA-Seq revealed that a higher risk score predicted stronger proliferative ability and malignancy in HCC.
**FIGURE 4:** *Functional enrichment analysis of high- and low-risk groups. GO analysis (A) and KEGG analysis (B) of DEGs between the high-risk and low-risk groups. (C) Heatmap of ssGSEA scores in proliferation-related pathways. (D) Correlation between mRNAsi score and risk score of the signature. (E) GSVA for malignant cells from single-cell RNA-Seq.*
## 3.5 Genomic changes of cuproptosis and copper homeostasis related signature
An investigation of the relationship between somatic mutations and the signature was conducted in high-risk and low-risk patients. The fifteen genes with the highest mutation rate were identified (Figure 5A). In spite of the fact that there was no significant difference in tumor mutation burden between groups with high- and low-risk scores (Supplementary Figure S1C), there were differences in tumor heterogeneity (Figure 5B) and mutation rates of several high-frequency mutant genes. High-risk individuals exhibited higher mutation rates of TP53 ($$p \leq 0$$), LRP1B ($$p \leq 0.008$$), and OBSCN ($$p \leq 0.008$$), all of which were identified as crucial tumor suppressors.
**FIGURE 5:** *Features of mutation and methylation of high- and low-risk groups in the TCGA-cohort. (A) Diagrams of the 15 most substantially changed genes in the high-risk and low-risk subgroups. (B) The MATH scores of HCC patients from high-risk and low-risk subgroups. (C) The top 10 genes with the most significant positive or negative β value difference in high- and low-risk groups.*
In addition, the methylation of genes was compared between the high and low-risk groups [Supplementary Table S1 (S3)]. β value was used to measure the methylation level of genes. The top 10 genes with the most significant positive or negative β value differences were displayed respectively (Figure 5C). A higher level of methylation was found in the high-risk group for the following genes: SH3BP4, ADI1, AEN, ELK4, C9orf5, BAIAP2, HFE2, RAD54L2, and SLC23A2. Methylation levels of the following genes were more significant in the low-risk group: UCK2, DHX9, FLVCR1, LQK1, CDKN2BAS, PACS1, CASP2, SPP1, SLCA5, HIF1a. The complete data was shown in [Supplementary Table S1 (S2)]. Finally, we compared copy number variations (CNV) in two groups, but no significant difference was found in the results (Supplementary Figure S1D).
## 3.6 Immune landscape analysis revealed immunosuppressive tendency of high-risk score sample
The TCGA cohort’s immune-related processes’ scores were calculated using ssGSEA (Figure 6A). The results showed decreased response to IFN-1 and IFN-2, decreased CCR activity, decreased cytolytic activity, and increased expression of MHC-1 in high-risk HCC patients. Using the quantiseq algorithm, the immune score of tumor tissue was quantified to further reveal the effect of different risk scores on the immune-related TME (Figure 6B). The high-risk group showed significant increases in B cells, M2 macrophages, monocytes, and T cells, but a decrease in NK cells. The immune infiltration in the external validation cohort (ICGC) was analyzed using the same method. The high-risk group showed significant increases in B cell and M2 macrophage, but a decrease in NK cells (Figure 6C), which was roughly in line with the TCGA cohort. Also, immune checkpoint molecules were examined that inhibit immune cells and allow tumors to escape immune recognition. A significant increase in the expression of most chosen immune checkpoint molecules ($\frac{17}{20}$) was observed in high-risk individuals (Figure 6D). The ICGC cohort also revealed significant differences in immune checkpoint expression in different risk groups ($\frac{12}{20}$) (Figure 6E), which confirms the TCGA cohort’s results. Analysis of the relationship between cancer immune cycle and risk score was carried out using TIP (http://biocc.hrbmu.edu.cn/TIP) (Figure 6F) (Xu et al., 2018). Risk scores and step 1 (antigen release from cancer cells) of the immune process were positively correlated, but step 5 (immune cell infiltration into tumors) was negatively correlated.
**FIGURE 6:** *The immune landscape of high- and low-risk groups. (A) Heatmap of ssGSEA scores in the activity of immune-related processes in TCGA cohort. The quantiseq method calculates the proportion of 10 types of immune cells in low- and high-risk score groups of the TCGA cohort (B) and ICGC cohort (C). The expression of immune checkpoints in the high-risk and low-risk groups of the TCGA cohort (D) and ICGC cohort (E). (F) Correlation between cancer-immunity cycle scores and risk scores in the model in the TCGA cohort. (G) The proportion of malignant cells and immune-related cells in high-risk and low-risk samples of single-cell sequencing data. Detailed data was shown in supplementary Table S1 (S4). (H) GSVA analysis of T cell function in the single-cell cohort.*
In the analysis of the sc-RNA data, We compared the contents of tumor cells and different types of immune cells in different groups. It was found that samples at high risk contained a higher proportion of malignant cells and a lower proportion of immune cells and other cells. ( Figure 6G). T cells in TME are essential participants in tumor-related immune processes but are usually inhibited by various signals. T cells from high-risk and low-risk groups were compared using GSVA to investigate whether risk score impacts T cell function in TME (Figure 6H). The low-risk group showed significantly higher activity in T cell activation pathways than the high-risk group, such as cytotoxicity, chemicals, T cell functions, negative regulation of T cell apoptosis, IMmotion150 teff, cytokines, IMmotion 150 myoid inflammation, leucocyte function.
## 3.7 Ligand–receptors pairs analysis between immunocytes and HCC cells
The analysis above revealed that the infiltration rate of immune cells was different between high- and low-risk scores. As a result, we conducted a communication analysis between malignant cells and other immune-related cells based on the single-cell sequencing data to find the pathways and corresponding targets (Figure 7A). In the high-risk samples, fibroblasts dominated signal input and output. In comparison, the signal input of the low-risk samples was dominated by endothelial cells. In the communication between tumor cells and other cells, Signal intensity and communication process were significantly different between high- and low-risk groups for the following pathways: FASLG (FASL-FAS) signal pathway (Figure 7B), CCL (CCL5-CCR5) signal pathway (Figure 7C), CD40 (CD40L-(ITFA5, IGTB1)) signal pathway (Figure 7D), IL2 (IL7R-IL7RG) signal pathway (Figure 7E), and IFN-II (IFNG-(IFNGR1-2)) signal pathway (Figure 7F). The high-risk group has different degrees of signal intensity reduction in these pathways, the activation of which could assist in the anti-tumor process. The total information flow between high-risk and low-risk groups also differed significantly across other signaling pathways (Figure 8A). To a certain extent, this explains the decrease in immune cell infiltration and the tendency of immunosuppression in the HCC TME of high-risk samples.
**FIGURE 7:** *Differences of Ligand-Receptors in cell communication between high- and low-risk samples of single-cell RNA-Seq. (A) Dot graphs show how intensively each cell type communicates in high-risk and low-risk samples. (B) FASLG signaling network of high- and low-risk samples. (C) CCL signaling network of high- and low-risk samples. (D) CD40 signaling network of high- and low-risk samples. (E) IL2 signaling network of high- and low-risk samples. (F) IFN-II signaling network of high- and low-risk samples.* **FIGURE 8:** *Pathways with different overall information between high- and low-risk single-cell sequencing samples. Predicting treatment response. (A) Signaling pathways with significant differences in the overall intercellular information between single-cell RNA-Seq samples at high- and low-risk. (B) The IC50s of chemotherapeutic agents and targeted drugs related to the 5-gene signature. (C) The prediction results show the distribution of immune response scores in the high- or low-risk groups of the TCGA cohort. (D) Differences in sensitivity of subgroups based on risk score to immunotherapy. Submap classing analysis manifested that the high-risk score group could be more sensitive to the programmed cell death protein 1 (PD-1) inhibitor (Bonferroni-corrected p = 0.008).*
## 3.8 Responses prediction of chemotherapeutic and immune therapy
Non-operative treatment of HCC faces the challenge of drug resistance, and copper has been proven to alter tumor cell drug resistance. To evaluate risk characteristics’ role in clinical treatment, we compared high- and low-risk patients’ sensitivity to chemotherapeutics and target therapy. In total, 156 differential drugs and molecular compounds were screened out, with 127 drugs being more sensitive to the high-risk group and 29 drugs being more sensitive to the low-risk group [Supplementary Table S1 (S5)]. The IC50 value estimated for low-risk cancer patients is higher than that found in low-risk cancer patients for the following clinically common targeted therapy drugs: Tipifarnib, Tivozanib, Masitinib, Dasatinib, Sunitinib, and chemotherapy drugs: Gemcitabine, Vinorelbine, Rapamycin, Paclitaxel, Pyrimethamine (Figure 8B).
Based on TIDE algorithm, immune checkpoint inhibitor responses in different patient groups were predicted. The TIDE score was higher in low-risk patients than in high-risk patients, which means the efficacy of immune checkpoint blocking therapy (ICB) was worse, and the survival time was shorter after ICB treatment (Figure 8C). Based on the IPS of HCC patients, the response to immunotherapy targeted specifically at CTLA-4 and PD-1 in high-risk and low-risk HCC patients was examined using a subclass mapping approach. The high-risk patients responded well to anti-PD-1 therapy, while the low-risk patients had no reaction to either anti-PD-1 or anti-CTLA4 therapy. ( Figure 8D). We found that the treatment options mentioned above were more likely to benefit high-risk patients, regardless of whether they were traditional chemotherapy or targeted therapy.
## 3.9 Verification of biological function and expression level of model-constructed genes
Twenty-one HCC patients were tested using qRT-PCR on paired tumors and normal adjacent tissues. CDKN2A, GEMIN2, DLAT, and ZCRB1 were expressed at higher levels in tumors than in normal tissues, while the expression of KLF9 in tumor tissue was lower (Figure 9A). For the subsequent study, we selected Hep3B and Huh7 cell lines transfected with sh-RNAs for knockdown experiments of five model-constructed genes. The plasmid transfection efficiency of all five genes for both cell lines was greater than $50\%$ (Figure 9B). The proliferation and viability of cells were assessed by the CCK-8 assay. Knockdown of CDKN2A, GEMIN2, and ZCRB1 prominently impaired cell growth of both Huh7 and Hep3B, while the knockdown of KLF9 improved the cell growth of both cell lines (Figure 9C).
**FIGURE 9:** *Model-constructed genes expression in HCC tissues and effects on HCC cell biological signature. (A) qRT-PCR analysis of model-constructed gene expression in paired tumors and adjacent normal tissues from 21 HCC patients. (B) The efficiency of transfection was determined by qRT-PCR. (C) CCK-8 was used to determine growth curves for transfected Huh7 and Hep3B cellsB cells. (D) AKT, p-AKT, mTOR, and p-mTOR expression in different groups were detected with western blot.*
According to previous findings (Figure 4C), different risk groups differed significantly in proliferative capacity and activity of the PI3K/AKT/mTOR signal pathway, which is widely implicated in mitochondrial metabolism and tumor drug resistance. There is evidence that phosphorylation of AKT and mTOR affects copper-induced disease progression in a variety of diseases, including cancer. The results showed that the knockdown of CDKN2A, GEMIN2, DLAT, and ZCRB1 prominently impaired the phosphorylation of both AKT and mTOR in Huh7 and Hep3B, while the knockdown of KLF9 improved the phosphorylation (Figure 9D).
## 3.10 In Vitro effects of ZCRB1 on biological behavior of liver cancer cells
Considering the knockout of ZCRB1 had the strongest tumor-inhibiting effect in the proliferation experiment, specifically, we selected ZCRB1 as the target to determine its effect on proliferation, invasion, migration, and colony formation. In the scratch assay experiment, scratches in the knock-down group healed slower than in the control group (Figure 10A). In addition, transwell invasion and migration experiments confirmed that ZCRB1 downregulation significantly reduced tumor cell invasion and migration (Figures 10B, C). ZCRB1 also inhibited the growth of Huh7 and Hep3B colonies after knockdown, suggesting that ZCRB1 boosts colony formation (Figure 10D). Besides, we performed immunohistochemical analyses of human HCC tissues and adjacent normal tissue using ZCRB1 antibody. The expression of ZCRB1 in HCC tissue was significantly stronger than in adjacent normal tissue. The expression of ZCRB1 in HCC tissue was significantly stronger than in adjacent normal tissue Figure 10E. Through the Human Protein Atlas (HPA) database, we also supplied the immunohistochemical images of the remaining model-constructed molecules in the HPA database, which indicated that there was a higher expression of GEMIN2, CDKN2A, and DLAT in HCC than in normal tissues (Supplementary Figure S1E), but the immunohistochemical data of KLF9 were not obtained. As a result, in vitro experiments suggest that ZCRB1 expression is closely related to malignant behavior in tumor cells, and it could become a new therapeutic target for copper homeostasis and cuproptosis.
**FIGURE 10:** *(A) Scratch experiments were used to determine migration and wound healing. (B) Transwell migration assays. (C) Transwell invasion assays. (D) Clonogenic assays. (E) Phenotypic experiments on ZCRB1 in vitro and immunohistochemistry results in clinical samples.*
## 4 Discussion
Compared with other trace elements in the human body, copper has unique redox activity, making it an essential catalytic cofactor (Kim et al., 2008). Copper homeostasis disorder can lead to intracellular copper overload, leading to cellular protein toxic stress, which is the reason for acute cell death in cuproptosis (Tsvetkov et al., 2022). During cell proliferation, copper participates in the signal cascade (Tsang et al., 2020), promotes proliferation and diffusion, and participates in tumor microenvironment changes (Soncin et al., 1997). The critical role of copper uptake, distribution, and effluent ligand/pump expression in cancer has been confirmed (Itoh et al., 2008; Blockhuys et al., 2017; Blockhuys et al., 2020). Because of copper’s role in cancer development and the crucial position of the liver in the process of copper storage and metabolism, we searched for the related genes of copper homeostasis and cuproptosis through correlation analysis and then constructed a copper homeostasis and cuproptosis associated gene signature by Lasso regression and multivariate COX analysis. HCC patients’ prognoses could be well predicted with the model, and clinical characteristics were combined with risk scores to construct a nomogram model for facilitating clinical research and application. Several studies have examined the relationship between cuproptosis and patient prognosis in HCC (Ding et al., 2022; Peng et al., 2022; Xie et al., 2022; Zhang et al., 2022). For example, Peng et al. developed a prognostic model based on cuproptosis-related genes (Peng et al., 2022). Xie et al. built a cuproptosis-related immune checkpoint gene signature to identify the prognosis of HCC patients (Xie et al., 2022). Ding et al. also built a cuproptosis-related prognosis model and discussed it in different cuproptosis subtypes (Ding et al., 2022). Before the concept of cuproptosis was proposed, copper and copper homeostasis had been revealed to be related to many diseases, including cancers. Cuproptosis is mainly involved in participants of the TCA cycle within mitochondrial metabolism while maintaining intracellular copper homeostasis requires an intracellular multi-structure system. No research has been conducted based on copper homeostasis in hepatocellular carcinoma or other diseases. In fact, the integrity role of copper homeostasis and cuproptosis in disease has been recognized. A recent study published on signal transport and target therapy comprehensively elaborated on broad application prospects of copper homeostasis and cuproptosis and proposed that reliable biomarkers are scarce as of now (Chen et al., 2022). Besides, the clinical utility of specific models is crucial. Decision Curve Analysis (DCA) is a method to evaluate and compare multiple clinical prediction models in clinical utility, which was proposed by Dr. Andrew Vickers. This method allows us to compare our study with other cuproptosis-related models. At the time of 1, 3, and 5 years, our model exhibits better application value than the models based solely on cuproptosis (Supplementary Figure S1F). Additionally, our study was the first to explore in detail the function of tumor cells and T cells as well as the intercellular communication among different risk groups based on single-cell sequencing data, which clarifies the impact of copper-related physiological processes on different components in tumor microenvironments and provides a new perspective for follow-up readers’ studies.
Previous studies have shown that copper in the TME can directly or indirectly activate metalloenzyme function and oxidative stress (Ma et al., 1999). Without regular intracellular disposal, excessive oxidative stress will induce tumor cell transformation and uncontrolled proliferation (Hsu et al., 1994). In the functional analysis of the TCGA cohort and single-cell RNA-Seq samples, high-risk patients and malignant cell clustering showed higher proliferative capacity and viability under hypoxia. The risk score in malignant cell clustering is significantly higher than in others. Furthermore, patients with high-risk scores had higher mRNAsi scores, reflecting an acquired stem cell-like phenotype and loss of cell differentiation. The high-risk group had an increased mutation rate for TP53, LRP1B, and OBSCN. A previous study showed that OBSCN is an effective tumor suppressor in various cancers (Guardia et al., 2021). It is also known that TP53 mutation frequency is higher in cancers with increased malignancy. LRP1B is a tumor suppressor, but LRP1B-mutated cancers have improved outcomes with ICIs, the underlying mechanism of which has not yet been clarified (Brown et al., 2021). In addition, many genes with different methylation levels are associated with proliferation. This explains in one way why patients with high-risk scores had poorer clinical outcomes, indicating that the genes related to copper homeostasis and cuproptosis affect the tumor proliferation and malignancy in HCC and may even be involved in forming cancer stem cells.
Cuproptosis is mainly involved in participants of the TCA cycle within mitochondrial metabolism, while mitochondrial metabolism and glycolysis are highly related to the phosphorylation of AKT (Stiles, 2009). Early studies have confirmed that tumorigenesis can be reduced by inhibiting the copper transporter 1-copper axis via AKT signaling (Chen et al., 2021; Guo et al., 2021). And before cuproptosis was revealed, some studies had previously attempted to change cancer cells’ tolerance to specific drugs by blocking the activity of AKT (Banerjee et al., 2016; Wu et al., 2018). By knocking out the model-constructed genes, we revealed that the model-constructed genes were strongly correlated with AKT and mTOR phosphorylation levels. Despite the function of some crucial molecules has been proved, there are still unsolved mysteries. In a recent study, bioinformatics and experimental verification were combined to prove the effect of DLAT on AKT phosphorylation in HCC (Zhou et al., 2022). According to another study, CDKN2A-mediated AKT phosphorylation influences cervical cancer malignancy (Luan et al., 2021), but there are no relevant studies in HCC. KLF9 is downregulated in HCC, which could stabilize p53 and induce apoptosis (Sun et al., 2014), while it remains unknown whether it affects the activity of AKT-related pathways in HCC. A particular interest of ours is ZCRB1, which is an RNA-binding protein. As a tumor suppressor gene, ZCRB1 phosphorylates JMJD5 to regulate aerobic glycolysis in GBM through the cyclic RNA HEATR5B (Song et al., 2022). There are few studies on the role of ZCRB1 in cancer. According to our results, however, knocking out ZCRB1 significantly inhibits the malignant phenotype of HCC, as well as inhibiting the phosphorylation of AKT and mTOR. Considering the heterogeneity of copper-related metabolic processes in different tissues, ZCRB1 may combine different circRNAs and complete the phosphorylation of AKT/mTOR through different signal axes. And this process may be caused by an imbalance in copper homeostasis or cuproptosis.
Copper participates in human immunity, which promotes leukocyte differentiation, maturation, and proliferation and maintains the phagocytosis of neutrophils (Djoko et al., 2015). The role of copper in antitumor immunity has been demonstrated in recent studies. In the immune regulation of cancer, disulfiram as a copper carrier can make cancer cells carry excess copper and maintain the stability of PD-L1 in HCC (Zhou et al., 2019). In the immune-activation mouse model of neuroblastoma (GBM), copper chelation therapy with TEPA can reduce the PD-L1 expression of GBM, improve the anti-GBM immune response mediated via NK cells, and inhibit the immune checkpoint (Voli et al., 2020). These studies advise that reducing the concentration of copper in the tumor can stimulate the anti-cancer immune response and promote new immune cell clones in tumors. We found that the immune cells infiltrating the tumor tissues of high-risk patients were significantly reduced. The proportion of NK cells decreased, while immunosuppressive cells (Macrophage M2, Tregs) increased significantly. Changes in pathways related to immune cells and chemokines indicate that it is believed that copper accumulation in TME reduces the ability of immune cells to infiltrate and weakens the body’s immune response to malignant cells as a result.
In the TME, T cells play a significant role in anti-tumor immunity. Besides being a target for immune checkpoint therapy, it can also promote tumor immune escape, so understanding its characteristics is crucial (Oh et al., 2021). Based on single-cell sequencing data, GSVA was performed on T cells of samples from different risk groups, and many pathways related to T cell activity were suppressed, which suggests T cell activity may be regulated by genes involved in copper homeostasis and cuproptosis. Furthermore, positive correlations were found between several immune checkpoints and risk scores. As shown above, immunosuppression tends to be more common in high-risk patients, and the score of the signature reflects that tendency.
Our analysis of single-cell sequencing data revealed that the difference in cell communication between different cells could be the mechanism behind TME changes associated with the signature. An array of signaling pathways and corresponding receptor-ligand pairs were identified. Fas is found in virtually all cells, while the FasL gene is predominantly expressed in activated T cells. Inducing apoptosis and cell death is the primary function of Fas/FasL. T cells and NK cells trigger tumor cell apoptosis through FasL, a tumor suppressor gene (Villa-Morales and Fernández-Piqueras, 2012). CCL5’s role in tumors has been controversial. Some studies suggest that its production induces immunosuppression (Chang et al., 2012), while others suggest it promotes tumor immunity (Harlin et al., 2009; Liu et al., 2015). Tumor necrosis factor (TNF) receptors include the CD40 receptor. CD40 activates dendritic cells, which then activate CD8 + T cells Vonderheide, 2020. Monoclonal CD40 has shown efficacy in tumor therapy (Cancer Discov, 2017). In addition to proliferating effector T cells, IL-2 regulates the growth of Treg cells. IL-2-based anticancer treatments are becoming increasingly popular (Mullard, 2021). Anti-CTLA-4 resistance is affected by the expression of IFNG1 (Cancer Discov, 2017), whereas anti-PD-1 resistance is affected by the expression of IFNG2 (Williams et al., 2020). The signal intensity of the above pathways and their receptor-ligand pairs decreased to varying degrees in high-risk samples. Taking into account the change in T cell activity, the above ligand-receptor pairs may be required for genes related to copper homeostasis and cuproptosis to participate in communicating intercellularly, which may begin with the activation of immune-helper cells, such as dendritic cells, followed by the activation of effector T cells such as CD8+ and NK cells. Consequently, this will lead to a change in immune-related TME and ICBs sensitivity.
As a traditional therapy for HCC, chemotherapy can’t wholly remove tumor cells because of inherent or acquired drug resistance (Siddik, 2003). Applying copper-based complexes and copper-chelating agents is sufficient to bypass cisplatin resistance in different types of cancer (Mo et al., 2018; Rochford et al., 2020; Vančo et al., 2021). Similar methods were used in clinical trials of breast and prostate cancers (Henry et al., 2006; Pass et al., 2008; Chan et al., 2020). In several studies, chemotherapy resistance was associated with the downregulation of copper transporters and the upregulation of pumps and chaperones for copper efflux. ( Katano et al., 2002; Safaei and Howell, 2005; Yu et al., 2020). The relationship between copper metabolism and chemotherapy resistance is disease-specific. For example, the clinical correlation between copper transporter 1(CTR1) expression and the efficacy of platinum chemotherapeutic drugs were contradictory in different studies (Ishida et al., 2010; Lee et al., 2011; Akerfeldt et al., 2017). Therefore, applying the risk model requires a prediction of the chemotherapeutic drug’s sensitivity. In our research, 127 drugs and compounds were expected to be sensitive to high-risk HCC patients. Additionally, patients at high risk responded better to immunotherapy targeting PD-1, which provides a reference for further research and clinical application. Considering the copper dependence on cancer progression and the low cytotoxicity of copper-chelating drugs (Hsu et al., 1994), it has the potential to use genes related to copper homeostasis and cuproptosis as immunotherapy targets. Due to the limitation of understanding the metabolic process of copper and the related mechanisms of copper homeostasis in different tumor drug resistance, no copper complexes have been used in anti-tumor therapy. More copper-related targets and pathways in cells must be found to develop more stable drug ligands. Our research provides a new application direction for traditional chemotherapeutic and targeted drugs.
## 5 Conclusion
A novel scoring model related to copper homeostasis and cuproptosis was developed in this study. High-risk scores predicted poor prognosis, high tumor malignancy, and tumor immunosuppression in HCC patients. Novel receptor-ligand pairs were proposed as targets for the changes in immune function and TME based on the intercellular communication status. Targeted and chemotherapeutic drugs with potential effects were predicted. Meanwhile, model-constructed genes were validated in terms of their clinical and functional significance, but further study is needed to understand the mechanism in more detail. As a result of our research, we are able to evaluate the malignant degree, TME changes, and cross-talk between malignant cells and immunocytes in patients with HCC, which can provide suggestions for treatment.
## Data availability statement
The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding authors.
## Ethics statement
The studies involving human participants were reviewed and approved by the Regional Ethics Committee at Shandong University. Written informed consent for participation was not required for this study in accordance with the national legislation and the institutional requirements.
## Author contributions
TL and Y-LM conceived the study. Y-LM and D-XW made a major contribution to the design and drafting of the article. H-CW contributed to the design and application of tools of statistical analysis. Y-FY, C-CY, L-JY, Z-ND, B-WT, HL, J-SX, C-LH, S-YT, J-GH, D-XW, and X-CM contributed equally to the statistical analysis. The final manuscript was read and approved by all authors.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fphar.2023.1101749/full#supplementary-material
## References
1. Akerfeldt M. C., Tran C. M., Shen C., Hambley T. W., New E. J.. **Interactions of cisplatin and the copper transporter CTR1 in human colon cancer cells**. *J. Biol. Inorg. Chem.* (2017) **22** 765-774. DOI: 10.1007/s00775-017-1467-y
2. Banerjee K., Das S., Sarkar A., Chatterjee M., Biswas J., Choudhuri S. K.. **A copper chelate induces apoptosis and overcomes multidrug resistance in T-cell acute lymphoblastic leukemia through redox imbalance and inhibition of EGFR/PI3K/Akt expression**. *Biomed. Pharmacother. = Biomedecine Pharmacother.* (2016) **84** 71-92. DOI: 10.1016/j.biopha.2016.08.056
3. Bhattacharjee A., Chakraborty K., Shukla A.. **Cellular copper homeostasis: Current concepts on its interplay with glutathione homeostasis and its implication in physiology and human diseases**. *Metallomics* (2017) **9** 1376-1388. DOI: 10.1039/c7mt00066a
4. Bian Z., Fan R., Xie L.. **A novel cuproptosis-related prognostic gene signature and validation of differential expression in clear cell renal cell carcinoma**. *Genes (Basel)* (2022) **13** 851. DOI: 10.3390/genes13050851
5. Bisaglia M., Bubacco L.. **Copper ions and Parkinson's disease: Why is homeostasis so relevant?**. *Biomolecules* (2020) **10** 195. DOI: 10.3390/biom10020195
6. Blockhuys S., Brady D. C., Wittung-Stafshede P.. **Evaluation of copper chaperone ATOX1 as prognostic biomarker in breast cancer**. *Breast Cancer* (2020) **27** 505-509. DOI: 10.1007/s12282-019-01044-4
7. Blockhuys S., Celauro E., Hildesjö C., Feizi A., Stål O., Fierro-González J. C.. **Defining the human copper proteome and analysis of its expression variation in cancers**. *Metallomics* (2017) **9** 112-123. DOI: 10.1039/c6mt00202a
8. Brown L. C., Tucker M. D., Sedhom R., Schwartz E. B., Zhu J., Kao C.. **LRP1B mutations are associated with favorable outcomes to immune checkpoint inhibitors across multiple cancer types**. *J. Immunother. Cancer* (2021) **9** e001792. DOI: 10.1136/jitc-2020-001792
9. **IFNγ mutations prompt CTLA-4 inhibitor resistance**. *Cancer Discov.* (2017) **7** OF3. DOI: 10.1158/2159-8290
10. Cen D., Brayton D., Shahandeh B., Meyskens F. L., Farmer P. J.. **Disulfiram facilitates intracellular Cu uptake and induces apoptosis in human melanoma cells**. *J. Med. Chem.* (2004) **47** 6914-6920. DOI: 10.1021/jm049568z
11. Chan N., Willis A., Kornhauser N., Ward M. M., Lee S. B., Nackos E.. **Correction: Influencing the tumor microenvironment: A phase II study of copper depletion using tetrathiomolybdate in patients with breast cancer at high risk for recurrence and in preclinical models of lung metastases**. *Clin. Cancer Res.* (2020) **26** 5051. DOI: 10.1158/1078-0432.CCR-20-3177
12. Chang L. Y., Lin Y. C., Mahalingam J., Huang C. T., Chen T. W., Kang C. W.. **Tumor-derived chemokine CCL5 enhances TGF-β-mediated killing of CD8(+) T cells in colon cancer by T-regulatory cells**. *Cancer Res.* (2012) **72** 1092-1102. DOI: 10.1158/0008-5472.CAN-11-2493
13. Chen D., Cui Q. C., Yang H., Dou Q. P.. **Disulfiram, a clinically used anti-alcoholism drug and copper-binding agent, induces apoptotic cell death in breast cancer cultures and xenografts via inhibition of the proteasome activity**. *Cancer Res.* (2006) **66** 10425-10433. DOI: 10.1158/0008-5472.CAN-06-2126
14. Chen H., Feng X., Gao L., Mickymaray S., Paramasivam A., Abdulaziz Alfaiz F.. **Inhibiting the PI3K/AKT/mTOR signalling pathway with copper oxide nanoparticles from houttuynia cordata plant: Attenuating the proliferation of cervical cancer cells**. *Artif. cells, nanomedicine, Biotechnol.* (2021) **49** 240-249. DOI: 10.1080/21691401.2021.1890101
15. Chen L., Min J., Wang F.. **Copper homeostasis and cuproptosis in health and disease**. *Signal Transduct. Target. Ther.* (2022) **7** 378. DOI: 10.1038/s41392-022-01229-y
16. da Silva D. A., De Luca A., Squitti R., Rongioletti M., Rossi L., Machado C. M. L.. **Copper in tumors and the use of copper-based compounds in cancer treatment**. *J. Inorg. Biochem.* (2022) **226** 111634. DOI: 10.1016/j.jinorgbio.2021.111634
17. Ding L., Li W., Tu J., Cao Z., Li J., Cao H.. **Identification of cuproptosis-related subtypes, cuproptosis-related gene prognostic index in hepatocellular carcinoma**. *Front. Immunol.* (2022) **13** 989156. DOI: 10.3389/fimmu.2022.989156
18. Djoko K. Y., Ong C. L., Walker M. J., McEwan A. G.. **The role of copper and zinc toxicity in innate immune defense against bacterial pathogens**. *J. Biol. Chem.* (2015) **290** 18954-18961. DOI: 10.1074/jbc.R115.647099
19. Fukai T., Ushio-Fukai M., Kaplan J. H.. **Copper transporters and copper chaperones: Roles in cardiovascular physiology and disease**. *Am. J. Physiol. Cell Physiol.* (2018) **315** C186-C201. DOI: 10.1152/ajpcell.00132.2018
20. Geeleher P., Cox N., Huang R. S.. **pRRophetic: an R package for prediction of clinical chemotherapeutic response from tumor gene expression levels**. *PLoS One* (2014) **9** e107468. DOI: 10.1371/journal.pone.0107468
21. Guardia T., Eason M., Kontrogianni-Konstantopoulos A.. **Obscurin: A multitasking giant in the fight against cancer**. *Biochim. Biophys. Acta Rev. Cancer* (2021) **1876** 188567. DOI: 10.1016/j.bbcan.2021.188567
22. Guo J., Cheng J., Zheng N., Zhang X., Dai X., Zhang L.. **Copper promotes tumorigenesis by activating the PDK1-AKT oncogenic pathway in a copper transporter 1 dependent manner**. *Adv. Sci. (Weinheim, Baden-Wurttemberg, Ger.* (2021) **8** e2004303. DOI: 10.1002/advs.202004303
23. Harlin H., Meng Y., Peterson A. C., Zha Y., Tretiakova M., Slingluff C.. **Chemokine expression in melanoma metastases associated with CD8+ T-cell recruitment**. *Cancer Res.* (2009) **69** 3077-3085. DOI: 10.1158/0008-5472.CAN-08-2281
24. Harrell F. E., Lee K. L., Mark D. B.. **Multivariable prognostic models: Issues in developing models, evaluating assumptions and adequacy, and measuring and reducing errors**. *Stat. Med.* (1996) **15** 361-387. DOI: 10.1002/(SICI)1097-0258(19960229)15:4<361:AID-SIM168>3.0.CO;2-4
25. Hartke J., Johnson M., Ghabril M.. **The diagnosis and treatment of hepatocellular carcinoma**. *Semin. Diagn Pathol.* (2017) **34** 153-159. DOI: 10.1053/j.semdp.2016.12.011
26. Henry N. L., Dunn R., Merjaver S., Pan Q., Pienta K. J., Brewer G.. **Phase II trial of copper depletion with tetrathiomolybdate as an antiangiogenesis strategy in patients with hormone-refractory prostate cancer**. *Oncology* (2006) **71** 168-175. DOI: 10.1159/000106066
27. Hsu H. Y., Lin S. Y., Huang C. J., Lian S. L., Ho Y. H.. **Changes of serum copper and zinc levels in patients with nasopharyngeal carcinoma by radiotherapy**. *Biol. Trace Elem. Res.* (1994) **46** 1-13. DOI: 10.1007/BF02790064
28. Iasonos A., Schrag D., Raj G. V., Panageas K. S.. **How to build and interpret a nomogram for cancer prognosis**. *J. Clin. Oncol.* (2008) **26** 1364-1370. DOI: 10.1200/JCO.2007.12.9791
29. Inesi G.. **Molecular features of copper binding proteins involved in copper homeostasis**. *IUBMB Life* (2017) **69** 211-217. DOI: 10.1002/iub.1590
30. Ishida S., McCormick F., Smith-McCune K., Hanahan D.. **Enhancing tumor-specific uptake of the anticancer drug cisplatin with a copper chelator**. *Cancer Cell* (2010) **17** 574-583. DOI: 10.1016/j.ccr.2010.04.011
31. Itoh S., Kim H. W., Nakagawa O., Ozumi K., Lessner S. M., Aoki H.. **Novel role of antioxidant-1 (Atox1) as a copper-dependent transcription factor involved in cell proliferation**. *J. Biol. Chem.* (2008) **283** 9157-9167. DOI: 10.1074/jbc.M709463200
32. Jin C., Li Y., Su Y., Guo Z., Wang X., Wang S.. **Novel copper complex CTB regulates methionine cycle induced TERT hypomethylation to promote HCC cells senescence via mitochondrial SLC25A26**. *Cell Death Dis.* (2020) **11** 844. DOI: 10.1038/s41419-020-03048-x
33. Katano K., Kondo A., Safaei R., Holzer A., Samimi G., Mishima M.. **Acquisition of resistance to cisplatin is accompanied by changes in the cellular pharmacology of copper**. *Cancer Res.* (2002) **62** 6559-6565. PMID: 12438251
34. Kim B. E., Nevitt T., Thiele D. J.. **Mechanisms for copper acquisition, distribution and regulation**. *Nat. Chem. Biol.* (2008) **4** 176-185. DOI: 10.1038/nchembio.72
35. Lee Y. Y., Choi C. H., Do I. G., Song S. Y., Lee W., Park H. S.. **Prognostic value of the copper transporters, CTR1 and CTR2, in patients with ovarian carcinoma receiving platinum-based chemotherapy**. *Gynecol. Oncol.* (2011) **122** 361-365. DOI: 10.1016/j.ygyno.2011.04.025
36. Liu J., Li F., Ping Y., Wang L., Chen X., Wang D.. **Local production of the chemokines CCL5 and CXCL10 attracts CD8+ T lymphocytes into esophageal squamous cell carcinoma**. *Oncotarget* (2015) **6** 24978-24989. DOI: 10.18632/oncotarget.4617
37. Llovet J. M., De Baere T., Kulik L., Haber P. K., Greten T. F., Meyer T.. **Locoregional therapies in the era of molecular and immune treatments for hepatocellular carcinoma**. *Nat. Rev. Gastroenterol. Hepatol.* (2021) **18** 293-313. DOI: 10.1038/s41575-020-00395-0
38. Lowe J., Taveira-da-Silva R., Hilário-Souza E.. **Dissecting copper homeostasis in diabetes mellitus**. *IUBMB Life* (2017) **69** 255-262. DOI: 10.1002/iub.1614
39. Luan Y., Zhang W., Xie J., Mao J.. **CDKN2A inhibits cell proliferation and invasion in cervical cancer through LDHA-mediated AKT/mTOR pathway**. *Clin. Transl. Oncol.* (2021) **23** 222-228. DOI: 10.1007/s12094-020-02409-4
40. Ma Y., Ogino T., Kawabata T., Li J., Eguchi K., Okada S.. **Cupric nitrilotriacetate-induced apoptosis in HL-60 cells association with lipid peroxidation, release of cytochrome C from mitochondria, and activation of caspase-3**. *Free Radic. Biol. Med.* (1999) **27** 227-233. DOI: 10.1016/s0891-5849(99)00083-0
41. Malta T. M., Sokolov A., Gentles A. J., Burzykowski T., Poisson L., Weinstein J. N.. **Machine learning identifies stemness features associated with oncogenic dedifferentiation**. *Cell* (2018) **173** 338-354.e15. DOI: 10.1016/j.cell.2018.03.034
42. Mo Q., Deng J., Liu Y., Huang G., Li Z., Yu P.. **Mixed-ligand Cu(II) hydrazone complexes designed to enhance anticancer activity**. *Eur. J. Med. Chem.* (2018) **156** 368-380. DOI: 10.1016/j.ejmech.2018.07.022
43. Mroz E. A., Rocco J. W.. **MATH, a novel measure of intratumor genetic heterogeneity, is high in poor-outcome classes of head and neck squamous cell carcinoma**. *Oral Oncol.* (2013) **49** 211-215. DOI: 10.1016/j.oraloncology.2012.09.007
44. Mullard A.. **Restoring IL-2 to its cancer immunotherapy glory**. *Nat. Rev. Drug Discov.* (2021) **20** 163-165. DOI: 10.1038/d41573-021-00034-6
45. O'Day S., Gonzalez R., Lawson D., Weber R., Hutchins L., Anderson C.. **Phase II, randomized, controlled, double-blinded trial of weekly elesclomol plus paclitaxel versus paclitaxel alone for stage IV metastatic melanoma**. *J. Clin. Oncol.* (2009) **27** 5452-5458. DOI: 10.1200/JCO.2008.17.1579
46. O'Day S. J., Eggermont A. M., Chiarion-Sileni V., Kefford R., Grob J. J., Mortier L.. **Final results of phase III SYMMETRY study: Randomized, double-blind trial of elesclomol plus paclitaxel versus paclitaxel alone as treatment for chemotherapy-naive patients with advanced melanoma**. *J. Clin. Oncol.* (2013) **31** 1211-1218. DOI: 10.1200/JCO.2012.44.5585
47. Oh D. Y., Fong L., Newell E. W., Turk M. J., Chi H., Chang H. Y.. **Toward a better understanding of T cells in cancer**. *Cancer Cell* (2021) **39** 1549-1552. DOI: 10.1016/j.ccell.2021.11.010
48. Oliveri V.. **Selective targeting of cancer cells by copper ionophores: An overview**. *Front. Mol. Biosci.* (2022) **9** 841814. DOI: 10.3389/fmolb.2022.841814
49. Pass H. I., Brewer G. J., Dick R., Carbone M., Merajver S.. **A phase II trial of tetrathiomolybdate after surgery for malignant mesothelioma: Final results**. *Ann. Thorac. Surg.* (2008) **86** 383-389. DOI: 10.1016/j.athoracsur.2008.03.016
50. Peng X., Zhu J., Liu S., Luo C., Wu X., Liu Z.. **Signature construction and molecular subtype identification based on cuproptosis-related genes to predict the prognosis and immune activity of patients with hepatocellular carcinoma**. *Front. Immunol.* (2022) **13** 990790. DOI: 10.3389/fimmu.2022.990790
51. Rochford G., Molphy Z., Kavanagh K., McCann M., Devereux M., Kellett A.. **Cu(ii) phenanthroline-phenazine complexes dysregulate mitochondrial function and stimulate apoptosis**. *Metallomics* (2020) **12** 65-78. DOI: 10.1039/c9mt00187e
52. Safaei R., Howell S. B.. **Copper transporters regulate the cellular pharmacology and sensitivity to Pt drugs**. *Crit. Rev. Oncol. Hematol.* (2005) **53** 13-23. DOI: 10.1016/j.critrevonc.2004.09.007
53. Siddik Z. H.. **Cisplatin: Mode of cytotoxic action and molecular basis of resistance**. *Oncogene* (2003) **22** 7265-7279. DOI: 10.1038/sj.onc.1206933
54. Soncin F., Guitton J. D., Cartwright T., Badet J.. **Interaction of human angiogenin with copper modulates angiogenin binding to endothelial cells**. *Biochem. Biophys. Res. Commun.* (1997) **236** 604-610. DOI: 10.1006/bbrc.1997.7018
55. Song J., Zheng J., Liu X., Dong W., Yang C., Wang D.. **A novel protein encoded by ZCRB1-induced circHEATR5B suppresses aerobic glycolysis of GBM through phosphorylation of JMJD5**. *J. Exp. Clin. cancer Res. CR* (2022) **41** 171. DOI: 10.1186/s13046-022-02374-6
56. Stiles B. L.. **PI-3-K and AKT: Onto the mitochondria**. *Adv. drug Deliv. Rev.* (2009) **61** 1276-1282. DOI: 10.1016/j.addr.2009.07.017
57. Sturm G., Finotello F., List M.. **Immunedeconv: An R package for unified access to computational methods for estimating immune cell fractions from bulk RNA-sequencing data**. *Methods Mol. Biol.* (2020) **2120** 223-232. DOI: 10.1007/978-1-0716-0327-7_16
58. Sun J., Wang B., Liu Y., Zhang L., Ma A., Yang Z.. **Transcription factor KLF9 suppresses the growth of hepatocellular carcinoma cells**. *Cancer Lett.* (2014) **355** 25-33. DOI: 10.1016/j.canlet.2014.09.022
59. Tan H. Y., Wang N., Zhang C., Chan Y. T., Yuen M. F., Feng Y.. **Lysyl oxidase-like 4 fosters an immunosuppressive microenvironment during hepatocarcinogenesis**. *Hepatology* (2021) **73** 2326-2341. DOI: 10.1002/hep.31600
60. Tsang T., Posimo J. M., Gudiel A. A., Cicchini M., Feldser D. M., Brady D. C.. **Copper is an essential regulator of the autophagic kinases ULK1/2 to drive lung adenocarcinoma**. *Nat. Cell Biol.* (2020) **22** 412-424. DOI: 10.1038/s41556-020-0481-4
61. Tsvetkov P., Coy S., Petrova B., Dreishpoon M., Verma A., Abdusamad M.. **Copper induces cell death by targeting lipoylated TCA cycle proteins**. *Science* (2022) **375** 1254-1261. DOI: 10.1126/science.abf0529
62. Vančo J., Trávníček Z., Hošek J., Malina T., Dvořák Z.. **Copper(II) complexes containing natural flavonoid pomiferin show considerable**. *Int. J. Mol. Sci.* (2021) **22** 7626. DOI: 10.3390/ijms22147626
63. Villa-Morales M., Fernández-Piqueras J.. **Targeting the Fas/FasL signaling pathway in cancer therapy**. *Expert Opin. Ther. Targets* (2012) **16** 85-101. DOI: 10.1517/14728222.2011.628937
64. Voli F., Valli E., Lerra L., Kimpton K., Saletta F., Giorgi F. M.. **Intratumoral copper modulates PD-L1 expression and influences tumor immune evasion**. *Cancer Res.* (2020) **80** 4129-4144. DOI: 10.1158/0008-5472.CAN-20-0471
65. Vonderheide R. H.. **CD40 agonist antibodies in cancer immunotherapy**. *Annu. Rev. Med.* (2020) **71** 47-58. DOI: 10.1146/annurev-med-062518-045435
66. Williams J. B., Li S., Higgs E. F., Cabanov A., Wang X., Huang H.. **Tumor heterogeneity and clonal cooperation influence the immune selection of IFN-γ-signaling mutant cancer cells**. *Nat. Commun.* (2020) **11** 602. DOI: 10.1038/s41467-020-14290-4
67. Wu N., Zhang C., Wang C., Song L., Yao W., Gedanken A.. **Zinc-doped copper oxide nanocomposites reverse temozolomide resistance in glioblastoma by inhibiting AKT and ERK1/2**. *Nanomedicine Lond. Engl.* (2018) **13** 1303-1318. DOI: 10.2217/nnm-2017-0359
68. Xie Y., Zhang W., Sun J., Sun L., Meng F., Yu H.. **A novel cuproptosis-related immune checkpoint gene signature identification and experimental validation in hepatocellular carcinoma**. *Sci. Rep.* (2022) **12** 18514. DOI: 10.1038/s41598-022-22962-y
69. Xu L., Deng C., Pang B., Zhang X., Liu W., Liao G.. **Tip: A web server for resolving tumor immunophenotype profiling**. *Cancer Res.* (2018) **78** 6575-6580. DOI: 10.1158/0008-5472.can-18-0689
70. Xu Y., Kong Y., Xu J., Li X., Gou J., Yin T.. **Doxorubicin intercalated copper diethyldithiocarbamate functionalized layered double hydroxide hybrid nanoparticles for targeted therapy of hepatocellular carcinoma**. *Biomater. Sci.* (2020) **8** 897-911. DOI: 10.1039/c9bm01394f
71. Yang M., Wu X., Hu J., Wang Y., Wang Y., Zhang L.. **COMMD10 inhibits HIF1α/CP loop to enhance ferroptosis and radiosensitivity by disrupting Cu-Fe balance in hepatocellular carcinoma**. *J. Hepatol.* (2022) **76** 1138-1150. DOI: 10.1016/j.jhep.2022.01.009
72. Yu G., Wang L. G., Han Y., He Q. Y.. **clusterProfiler: an R package for comparing biological themes among gene clusters**. *OMICS* (2012) **16** 284-287. DOI: 10.1089/omi.2011.0118
73. Yu Z., Cao W., Ren Y., Zhang Q., Liu J.. **ATPase copper transporter A, negatively regulated by miR-148a-3p, contributes to cisplatin resistance in breast cancer cells**. *Clin. Transl. Med.* (2020) **10** 57-73. DOI: 10.1002/ctm2.19
74. Zhang Y. J., Zhao D. H., Huang C. X.. *Zhonghua Nei Ke Za Zhi* (1994) **33** 113-116. PMID: 8070291
75. Zhang Z., Zeng X., Wu Y., Liu Y., Zhang X., Song Z.. **Cuproptosis-related risk score predicts prognosis and characterizes the tumor microenvironment in hepatocellular carcinoma**. *Front. Immunol.* (2022) **13** 925618. DOI: 10.3389/fimmu.2022.925618
76. Zhou B., Guo L., Zhang B., Liu S., Zhang K., Yan J.. **Disulfiram combined with copper induces immunosuppression via PD-L1 stabilization in hepatocellular carcinoma**. *Am. J. Cancer Res.* (2019) **9** 2442-2455. PMID: 31815045
77. Zhou Y., Gu H., Shao B., Zhang S., Pall H., Peixoto R. D.. **Glycolysis-related gene dihydrolipoamide acetyltransferase promotes poor prognosis in hepatocellular carcinoma through the Wnt/β-catenin and PI3K/Akt signaling pathways**. *Ann. Transl. Med.* (2022) **10** 1240. DOI: 10.21037/atm-22-5272
|
---
title: Use of zinc deposited in deciduous teeth as a retrospective measurement of
dietary zinc exposure during early development
authors:
- N. A. Wahono
- L. A. Wakeling
- W. Dirks
- D. A. Banks
- T. J. Shepherd
- D. Ford
- R. A. Valentine
journal: Frontiers in Oral Health
year: 2023
pmcid: PMC9998501
doi: 10.3389/froh.2023.1119086
license: CC BY 4.0
---
# Use of zinc deposited in deciduous teeth as a retrospective measurement of dietary zinc exposure during early development
## Abstract
### Purpose
We proposed that zinc (Zn) deposition in deciduous teeth would be a timed record of exposure to this essential micronutrient over very early life. We tested this hypothesis by gathering information on the maternal and child's diet during pregnancy and early infancy and measuring mineral deposition in the dentine at points during deciduous tooth development.
### Methods
We developed a short food frequency questionnaire (S-FFQ) to record consumption of food containing Zn during pregnancy and over the first year of life of the child in an Indonesian population. Zn, Sr and Ca were measured by laser ablation ICP-MS in a series of points across the developmental timeline in deciduous teeth extracted from 18 children undergoing the process as part of dental treatment whose mothers completed the SFFQ. Mothers and children were classified into either high Zn or low Zn groups according to calculated daily Zn intake.
### Results
The Zn/Sr ratio in dentine deposited over late pregnancy and 0–3 months post-partum was higher ($p \leq 0.001$, 2-way ANOVA; $p \leq 0.05$ by Holm-Sidak post hoc test) in the teeth of children of mothers classified as high Zn consumers ($$n = 10$$) than in children of mothers classified as low Zn consumers ($$n = 8$$).
### Conclusion
The S-FFQ was validated internally as adequately accurate to measure zinc intake retrospectively during pregnancy and post-partum (∼7 years prior) by virtue of the correlation with measurements of zinc in deciduous teeth. The ratio of Zn/Sr in deciduous teeth appears to be a biomarker of exposure to zinc nutrition during early development and offers promise for use as a record of prior exposure along a timeline for research studies and, potentially, to identify individuals at heightened risk of detrimental impacts of poor early life zinc nutrition on health in later life and to implement preventative interventions.
## Introduction
The nutritional environment encountered during development in utero and early childhood has been shown in numerous studies to have potential lifelong consequences for health through a range of mechanisms including physiological impacts and epigenetic recording [1, 2]. Although numerous studies have uncovered specific relationships, the need for longitudinal measurement in the context of the long human lifespan is a constraint, and much of the work has been done in animal models. Retrospective measurement of exposure in humans would be a highly valuable tool in such research. Also, some negative impacts may remain dormant and be manifest only in combination with further environmental exposures and/or as a result of ageing. Thus, it is important to uncover biomarkers of early life nutritional exposure, ideally in childhood while plasticity is still high, to intervene and re-set a better health trajectory.
The essentiality of zinc in myriad cellular and physiological functions accounts for the remarkably large component of the proteome that comprises zinc metalloproteins, which has been estimated to be $10\%$ [3]. It is beyond the scope of this paper to provide a full exposition of the myriad roles of Zn in normal physiological function and to describe in depth how zinc dyshomeostasis can affect health and increase susceptibility to or cause disease. However, there are particularly strong and well-studied associations, and recent rigorous reviews, concerning the role of zinc in immune function [4], skin health [5] and glycaemic control [6].
This high prevalence of zinc in fundamental biological function leads to the prediction that the zinc supply in utero will have profound effects on lifelong health. However, while there is a sizeable body of work on the value of zinc supplementation on outcomes for pregnancy and early infancy [7], little attention has been given to the potential role of maternal zinc status during pregnancy on health of the progeny in later life. Evidence of likely important impact includes a report that zinc concentration in maternal plasma during the first trimester was associated negatively with motor score and language ability at 1 year of age [8]. The lack of a robust biomarker of current zinc status, despite studies over several decades that have proposed measures in hair and urine and measurement of the expression of zinc-responsive proteins such as metallothioneins, as well as measurement in plasma, introduces a level of uncertainty into studies on the effects of zinc status on health [9, 10]. We propose that, because zinc is incorporated into the dental hard tissues, seemingly via zinc-regulated zinc transporters [11], and has been shown to influence the physical properties and thus probably resilience of enamel [12], the dietary zinc supply during development may have longer term implications for oral health. There is a need to gather robust evidence on the effects of early life nutritional exposure to zinc. To do this, the ability to determine retrospectively the level of zinc exposure of individuals during the period in utero would be a valuable tool. This would also facilitate the identification of individuals at risk of any detrimental impacts so that dietary remediation or other interventions to protect against specific effects of poor early zinc nutrition uncovered through future research can be recommended.
Measures of mineral deposition in teeth in relation to morphological features that identify periods of growth and weaning have a long history of use by anthropologists in the study of human life history [13]. Many of these studies focused on the detection of early dietary transitions, such as weaning, using strontium (Sr) and barium (Ba) [reviewed in [14]] and exposure to the neurotoxicant lead [e.g., [15, 16]]. However, zinc deposition in enamel and dentine has also been investigated, revealing that the enamel surface is highly enriched in zinc [17, 18], which was attributed to preferential binding of zinc by matrix metalloproteinase-20 and kallikrein-4, which are active during the two-stage enamel mineralisation process of secretion and maturation. These studies dismissed the use of dentine for recording trace element incorporation because of its porous nature. However, other studies have revealed that dentine does, in fact, incorporate trace elements in a predictable way, with clear zonation and time resolution from incremental markings [15, 16, 19]. This was further demonstrated using synchrotron x-ray fluorescence to map calcium, strontium and zinc at the neonatal line (NNL), which revealed increased levels of zinc in prenatal enamel and in dentine leading to the suggestion that zinc can be used to help identify the NNL [20]. A more recent study that used LA-ICP-MS single line rastering of the entire crown [21] found that zinc was elevated just after birth in $65\%$ of the sample, but variable in the rest, highlighting the need for more studies and a need to contextualise the results and take account of influencing factors. Evidence of environmental influences on metal deposition in the dentine of deciduous teeth was gathered in a pilot study on a community-based population, which used a retrospective method that accounted for water sources for both mother and infant, breastfeeding duration, formula feeding and demographic information and found correlations between early diet and trace element concentrations and timing of incorporation [19].
Given these promising preliminary results, we propose that zinc deposited in the dentine of deciduous teeth can be used as a biomarker of zinc status during early development. To address this hypothesis, we developed a food frequency questionnaire (FFQ) to measure retrospectively zinc consumption by mothers of children who presented for the extraction of deciduous teeth at the Child Integration Clinic, Dental Hospital of The Dentistry Faculty, Universitas Indonesia and we measured zinc in the dentine of these extracted teeth by laser ablation inductively-coupled plasma mass spectrometry (LA-ICP-MS) to investigate if a relationship existed.
## Materials and methods
Ethical permission was obtained from the Committee of The Medical Research Ethics of the Faculty of Medicine University of Indonesia (526/UN2.F1/ETIK/2014). All participants in this study were treated based on the guidelines assigned in the Declaration of Helsinki and gave informed consent.
## Development and validation of food frequency questionnaires (FFQ) for dietary Zn intake
A semi-quantitative FFQ to estimate dietary Zn intake in the Indonesian population was developed, focusing particularly on pregnancy and the period of infancy. In an initial phase, participants filled out an online questionnaire to report their recollection of all foods consumed in the previous 24 h (Q-24 h) to gather information on foods commonly consumed in Indonesia. The food items gathered from the Q-24 h were used in the development of a FFQ to be used to estimate habitual Zn intake over a longer period (LFFQ). As this study focused on Zn intake, food items not captured through the Q-24 h but known to be good sources of Zn were added. The LFFQ comprised 82 food items.
A series of food photographs were produced to enable participants to estimate their usual food portion and were based on the recommendations of a previous study [22]. Each food was presented as four portion sizes comprising $25\%$, $50\%$, $100\%$ and $125\%$ of a portion commonly consumed or portion on the package label of commercial products. Portions were measured out using an electrical scale (TANITA digital food scale). The amount of Zn in each food was obtained from USDA Nutrient Database for Standard Reference (United States Department of Agriculture 2013; https://data.nal.usda.gov/dataset/usda-national-nutrient-database-standard-reference-legacy-release), the Indonesian food database Nutrisurvey 2007 (http://www.nutrisurvey.de), or from previous studies [23, 24]. A plate or bowl containing the food was arranged together with a spoon and fork on each side. Food was photographed on a white background using a digital camera with a macro lens (Nikon 3100D) and photographs were printed at a size of 4 cm × 8 cm. In parallel, a shorter version of the FFQ (S-FFQ), which comprised fewer food items (28 items), was developed with the aim of reducing the required time for completion and thus pressure on the interviewer and participant during the clinic visit. To develop the S-FFQ, the number of food items was reduced by focusing on Zn-rich foods, such as red meat, offal, avocado, broccoli, spinach, grouping vegetables with lower Zn content, such as cabbage, carrot and lettuce, into a category of “other vegetables” and excluding items that were found to be rarely or never consumed by this population, such as brown rice, veal and pork. The L-FFQ and S-FFQ were compared with one and other and with a 3-day food record (Q3-d).
Both the L-FFQ and the S-FFQ consisted of five sections, which were: [1] personal information about child and parents, which included name, date of birth, birth weight, parents' educational background and occupation; [2] prenatal and birth history; [3] post-natal history, including feeding in the first six months, weaning age and foods, and consumption of food supplements; [4] retrospective record of foods consumed during pregnancy; [5] retrospective record of foods consumed by the child during infancy (from weaning up to age one year old); and [6] record of foods consumed by the child at the point of sampling. The S-FFQ is included as supplementary information.
## Calculation of current dietary Zn intake and intake during pregnancy and infancy
Children who attended the Child Integration Clinic, Dental Hospital of The Dentistry Faculty, Universitas Indonesia, from July-August 2014, for the removal of deciduous teeth to address dental health issues were eligible for this study. The inclusion criteria were: [1] the child and mother declared their willingness to participate in the study by signing a consent form explained to them previously; [2] the crown structure of the extracted tooth/teeth was still intact. Dental health personnel carried out the clinical assessment and tooth extraction following standard dental procedures.
Mothers were interviewed at the single visit when extraction was carried out to complete the overarching study questionnaire and the S-FFQ for themselves (retrospective recall for the period during pregnancy) and their children (over the period from weaning to one year old and current). Before completing the FFQ for foods consumed during pregnancy, the mothers were asked about their health during pregnancy, morning sickness and any food cravings. They were also asked to describe all foods they consumed on a daily (or frequent) basis, from breakfast to dinner. Portion size was estimated by the participants from the food photographs, and frequency of consumption was recorded. Daily Zn intake from the food source was calculated from these parameters and all intakes were summed to calculate total average daily Zn intake. Any information given by the participants was treated as confidential. They repeated the FFQ, but this time describing the food consumed by their child during infancy and again a FFQ for their child at their current age.
## Zn and Sr distribution in dentine of human primary teeth
Extracted teeth were rinsed in PBS (137 mM NaCl, 2.7 mM KCl, 4.3 mM Na2HPO4.7H2O, 1.4 mM KH2PO4, pH 7.3) to remove blood and debris then placed immediately in RNAlater® Stabilization Solution (Invitrogen) at 4°C. Teeth were stored at 4°C overnight then transferred to −20°C storage in the Oral Biology Laboratory, Faculty of Dentistry, Universitas Indonesia. Transfer to the Oral Biology Laboratory, School of Dental Sciences, Newcastle University, was in RNALater.
Teeth were sectioned to 200 µm thickness using a using a low-speed saw. Sections were prepared from the middle one third in the mesial-distal direction for incisors and at least two cusp tips were included in the buccal-lingual direction for molars. Trace elements including Ca, the major cation of dentine, were measured by Laser Ablation Inductively Coupled Mass Spectrometry (LA-ICP-MS) at the Faculty of Environment, School of Earth and Environment, University of Leeds based on a published protocol [15]. Briefly, a Geolas 193 nm ArF excimer laser coupled to an Agilent 750°C ICP-mass spectrometer was used to vaporise dentine tissue from a series of 100 µm diameter pits (constant energy density of 10 J/cm2; pulse rate of 5 Hz). Ablation pit transects, which crossed the neonatal line, were generated along the dentinal tubules from the enamel dentine junction (EDJ) towards the dentine-pulp chamber (DPC) (Figure 1). The laser ablation process was captured using a video camera with visible light sources integrated into the optical array to be displayed and monitored continuously on a computer screen. The specification and operating conditions of the laser and mass spectrometer are summarised in Table 1.
**Figure 1:** *A representative image of a longitudinal ground tooth section of an incisor. Labelling shows the enamel, enamel dentine junction (EDJ, purple arrows), dentine, neonatal line (NL, green arrows), ablation pits in the transect (orange arrows), dental pulp chamber (DPC), and tertiary dentine (black arrow).* TABLE_PLACEHOLDER:Table 1 Reference material NIST SRM Glass 610 was analysed before and after each analytical session and used for instrument calibration. This also permitted cross referencing to replicate analyses of NIST SRM Glasses 612 and 614 to determine the instrument performance and within-run standard errors. The isotopes chosen for elemental measurement (44Ca, 66Zn, 88Sr) were checked and proven free from isobaric interferences. Measured ion intensities were processed offline using SILLS, a software programme specifically written for signal integration of laboratory laser systems by Murray Allan (University of Leeds) and later modified by Dimitri Meier and Marcel Guillong (Die Eidgenossische Technische, Zurich). The data were then presented as elemental ion intensities (cps) relative to Ca ion intensities (cps) as Zn/Ca and Sr/Ca cps ratios. In the absence of international reference standards for dentine (bio-apatite), cps ratios permitted relative variation in Zn and Sr in tooth dentine to be studied without the need for absolute concentrations.
## Determination of tooth age from pregnancy up to twelve months of age
After the LA-ICP-MS procedure, all ground sections were observed under a polarized light microscope (Olympus BX51). The images were captured (x4 objective) and processed using a Q-Imaging Micropublisher 3.3 RTV camera and Improvision Openlab 5.0.2 image software. The tooth sections were then further prepared by polishing down the section into 80–100 µm thickness, and individually mounted on a microscope slide. Each ablation pit was identified, and the diameter was measured in micrometer (µm) units. The distance between ablation pits were also determined. The daily secretion rate of dentine (DSR) in each area was calculated by measuring the distance between Von Ebner's lines (x40 objective). All measurements were carried out using ImageJ software. The period of time covered by each pit and gap was determined by dividing the width of each by the DSR. Using the neonatal line (NNL; indicated on Figure 1 NNL is more commonly used in histology literature.) in dentine as the zero point, the age range (days) for each ablation pit was determined, taking the median value as the age point (days) of the ablation pit. Age points for ablation pits laid in dentine from the NNL to the enamel dentine junction (EDJ) are stated as negative values.
## Development and validation of a food frequency questionnaire (FFQ) for dietary Zn intake
The S-FFQ was validated as a tool to measure Zn intake in adult women by comparing the daily Zn intake recorded using the S-FFQs with a 3-day diet record (3DR) in six volunteers. A Bland-Altman plot was used to evaluate the limits of agreement between the S-FFQ and 3DR. The differences between the two assessments were within the standard deviation of the mean difference (0.19 for 3DR minus S-FFQ), which was 3.97 mg/d for the upper limit and −3.59 mg/d for the lower limit. Thus, the S-FFQ was considered adequately accurate and used in the study to estimate the daily Zn intake of the participants.
The L-FFQ and S-FFQ as tools to measure Zn intake in children were compared through completion by thirteen mothers of children who attended Al-Hidayah kindergarten, Bekasi, West Java, Indonesia. On the first day, participants were interviewed to complete the S-FFQ to calculate the daily Zn intake of their children in the previous month. A follow-up interview was carried out three days later to complete the L-FFQ for the same month. Statistical analysis confirmed a significant and positive strong correlation in dietary Zn intake measured using the two FFQs (Spearman analysis, $r = 0.997$, $$p \leq 0.01$$). Bland-*Altman analysis* showed that the daily Zn intake measured using the S-FFQ was in good agreement with the intake calculated using the L-FFQ. The differences in daily Zn intake measured using the L-FFQ and S-FFQ were within the standard deviation of the mean difference (−0.33 for L-FFQ minus S-FFQ), which was 2.6 mg/d for the upper limit and −3.26 mg/d for the lower limit. It was thus deemed that the S-FFQ was an acceptable replacement for the L-FFQ and that its use in the study would not compromise the accuracy of data obtained on dietary Zn intake in the child participants.
## Estimation of dietary Zn intake during pregnancy and infancy
Eighteen children paired with their mothers were eligible and recruited in this study. Characteristics of participants, including age, medical history, socioeconomic level, prenatal and postnatal history, are shown in Table 2. The dental hospital is located at central Jakarta, and the majority of the participants originated from the local area.
**Table 2**
| General Information | General Information.1 | n (%) |
| --- | --- | --- |
| Child's Gender | Male | 8 (40.00) |
| Child's Gender | Female | 10 (50.00) |
| Child's age (years) (mean ± SD) | 7.94 ± 1.25 | |
| Special needs | 0 (0.00) | |
| Medical history | Illness | 0 (0.00) |
| Medical history | Allergies | 5 (26.32) |
| Socioeconomic level | | |
| Father's education | below high school | 1 (5.00) |
| Father's education | high school | 12 (60.00) |
| Father's education | higher education | 5 (25.00) |
| Mother's education | below high school | 2 (10.00) |
| Mother's education | high school | 11 (55.00) |
| Mother's education | higher education | 5 (25.00) |
| Family income (Indonesian Rupiah) | <1,000,000 | 2 (10.00) |
| Family income (Indonesian Rupiah) | 1,000,000–<3,000,000 | 8 (40.00) |
| Family income (Indonesian Rupiah) | 3,000,000 < 5,000,000 | 5 (25.00) |
| Family income (Indonesian Rupiah) | ≤5,000,000 | 3 (15.00) |
| Prenatal and birth history | | |
| Mother's age during pregnancy [years (mean ± SD)] | 27.94 ± 5.07 | |
| Hyperemesis | 7 (35.00) | |
| Hospitalized | 1 (5.00) | |
| Drugs/Food Supplements/Vitamins | | 18 (90.00) |
| Delivery time | pre-term | 3 (15.00) |
| Delivery time | at-term | 16 (80.00) |
| Delivery method | vaginal birth | 14 (70.00) |
| Delivery method | caesarean section | 4 (20.00) |
| Birthweight (kg) (mean ± SD) | 3.17 ± 0.57 | |
| First six-months of feeding | | |
| Exclusively breastfed | 7 (50.00) | |
| Breastfed + early weaned | 4 (22.4) | |
| Breastfed + formula-fed | 3 (16.68) | |
| Breastfed + formula-fed + early weaned | 3 (16.68) | |
| Formula-fed | 1 (5.56) | |
| Weaning history | | |
| Weaning age (months) (mean ± SD) | 4.86 ± 2.29 | |
| Daily Zn intake (mg) (mean ± SD) | | |
| Pregnancy | 11.79 (7.31) | |
| Infancy | 5.96 (2.89) | |
| At point of sampling | 7.54 (3.70) | |
The daily Zn intake of the mother during pregnancy and the child during infancy were calculated using the S-FFQ. The estimated average requirement (EAR) for Zn in the mixed or refined vegetarian group defined by the IZiNCG [International Zinc Nutrition Consultative Group (IZiNCG) et al., 2004] was used to classify participants into low and high Zn intake groups, taking the EAR as the boundary between groups. The EAR for pregnant women is 8 mg/d and for children aged 1–3 years old the EAR is 3 mg/d. Eight mothers were in the low Zn intake (LZM) group (intake <8 mg/d); ten mothers were in the high Zn intake (HZM) group (intake >8 mg/d). Two children were in the LZC group (intake <3 mg/d); sixteen were in the HZC group (intake >3 mg/d).
The foods consumed most frequently during pregnancy, infancy, and at the point of sampling were rice, cereals and noodles (grouped). The food consumed least frequently was offal. Plant-based protein made the highest total contribution to daily Zn intake in mothers, followed by (in rank order) dairy and its products, rice, cereals and noodles (grouped), red meat and its products, eggs, beans, seeds and nuts (grouped), Zn-rich vegetables and fruits (grouped), fish and seafood, white meat and its products, and offal.
During infancy, the greatest contribution to daily Zn intake was from dairy products, followed by (in rank order) eggs, plant-based protein, rice, cereals and noodles (grouped), white meat and its products, Zn-rich vegetables and fruits (grouped), red meat and its products, offal, fish and seafood, beans, seeds and nuts (grouped).
At the time of sampling, the greatest contribution to daily Zn intake in children was from rice, cereals and noodles (grouped), followed by (in rank order) dairy products, red meat and its products, plant-based proteins, white meats and its products, eggs, Zn-rich vegetables and fruits (grouped), beans, seeds and nuts (grouped), fish and seafood, and offal.
Figure 2 shows the relative contribution of these foods for each group/time point.
**Figure 2:** *The contribution of different food groups to daily Zn intake of the study sample, measured using the S-FFQ, during pregnancy and – for children – infancy and the point of sampling.*
## Distribution of Zn and Sr in the dentine of human primary teeth
66Zn, 40Ca, and 88Sr were measured in one primary tooth from each of the study participants, which consisted of 14 incisors and 4 molars. Sr distribution and content was measured for comparison with Zn distribution and content, and we posited that it would be unrelated to developmental exposure to dietary Zn. Results presented relative to Ca ion intensity use a multiplication factor of 10,000 for Zn and 1,000 for Sr.
A previous study [15] found a sharp increase in Zn in ablation pits within approximately 200 µm of the DPC, and hypothesized this was associated with the formation of secondary dentine. In this area, the dentinal tubule density and diameter increase and accommodate more odontoblast cells involved in secondary dentine formation. Consistent with the findings of this earlier study, we found that the Zn/Ca ratio was significantly higher in the region within 500 μm of the DPC compared with other points (Figure 3A). The Sr/Ca ratio was greater closer to the pulp and the EDJ compared with the area 501–750 μm from the DPC (Figure 3B). These results suggest differences in the mechanisms and their regulation for incorporation of Zn and Sr in dental hard tissue.
**Figure 3:** *Distribution of Zn and Sr between the pulp and EDJ in deciduous teeth (median and 25th–75th percentile with whiskers extending from smallest to largest value; n = 18). (A) The Zn/Ca ratio differed in ablation pits according to the distance from the pulp to the EDJ (Friedman test, p < 0.0001). The ratio was significantly higher in ablation pits closest to the pulp [a,b p < 0.05, c,d p < 0.01 (Dunn's post-hoc test)]. (B) The Sr/Ca ratio in ablation pits also differed according to distance from the pulp to the EDJ (Friedman test, p < 0.0001) and was significantly lower in ablation pits in the region 501–750μm from the DPC compared with the region closer to and more distant from the DPC (*p < 0.05, Dunn's post-hoc test).*
To align ablation pits, and thus the corresponding measurements of Zn and Sr, with time points during development, we excluded measurements up to 500 μm from the DPC, on the basis that it was likely to constitute a region of secondary dentine formation. Thus, for incisors the median (minimum-maximum) time points covered were 111 (47–220) days before birth to 321 (277–517) days after birth, and for molars the median (minimum-maximum) time points covered 87 (40–127) days before birth to 492 (427–559) days after birth.
## Zn and Sr distribution in dentine from developmental time points of pregnancy up to twelve months of age
The Zn/Ca ratio remained fairly constant between late pregnancy and 0–3 months but then increased up to the latest sampling point of 9–12 months, differing significantly between 0 and 3 months and 9–12 months and between 3 and 6 months and 9–12 months (Figure 4A). In contrast there was a significant decrease in the Sr/Ca ratio across all time intervals measured (Figure 4B).
**Figure 4:** *Zn and Sr distribution across different points in time during the development of deciduous teeth (median and 25th–75th percentile with whiskers extending from smallest to largest value; n = 18). (A) The Zn/Ca ratio differed significantly across time points (Friedman test, p < 0.001); ***p < 0.001, ****p < 0.0001 (Dunn's post hoc test). (B) The Sr/Ca ratio also differed significantly across time points (Friedman Test, p < 0.001); *p < 0.05, **p < 0.01, ***p < 0.001 (Dunn's post hoc test).*
## The impact of Zn nutrition on Zn and Sr distribution in dentine
Separation of the data on Zn/Ca and Sr/Ca ratios in dentine according to maternal classification into high or low zinc intake groups (HZM ($$n = 10$$) and LZM ($$n = 8$$), respectively) revealed a different trend for Zn compared with Sr that was consistent with our hypothesis that higher Zn intake during development in utero and early infancy would be recorded as higher Zn levels in dentine but that Sr would not show this relationship (Figure 5). The median Zn/Ca ratio for the HZM group from ablation pits corresponding with all time points sampled was numerically higher than for the LZM group, although there were no statistically-significant differences (Figure 7A). In contrast, Sr/Ca ratios did not show this trend but showed an opposite trend, which we did not predict, with ratios from ablation pits corresponding with all time points sampled being numerically higher in LZM than in HZM groups, though again not differing significantly (Figure 5B). However, when data were expressed as Zn/Sr ratios there was a significant difference according to dietary zinc group ($p \leq 0.001$, 2-way ANOVA; Figure 5C). The Zn/Sr ratio in ablation pits corresponding to late pregnancy and 0–3 months post-partum was higher in the teeth of children of HZ mothers compared with LZ mothers ($$p \leq 0.024$$, Holm-Sidak post hoc test).
**Figure 5:** *Ratios of Zn, Sr and Ca across different points in time during the development of deciduous teeth shown as scatter plots with data points coded for mothers classified as consumers of a high zinc (HZM; n = 10) or low zinc (LZM; n = 8) diet during pregnancy. Medians are shown as horizontal bars. (A) Data expressed as Zn/Ca ratios. (B) Data expressed as Sr/Ca ratios. (C) Data expressed as Zn/Sr ratios; p < 0.001, 2-way ANOVA; *p < 0.05 for HZM vs. LZM (Holm-Sidak post hoc test).*
## Discussion
The most significant finding of this study was that dietary intake of Zn during pregnancy measured using a food frequency questionnaire developed for the purpose of measuring this retrospectively in the mothers of children of ages at which the primary dentition was still present (median 7.94 years) was correlated with the quantity of Zn as a ratio of Sr deposited during tooth development. The internal consistency of this finding indicates that retrospective recall of the diet eaten by the mothers in the sample during pregnancy was an adequate basis on which to determine Zn intake over this period and that the S-FFQ developed was a sufficiently accurate tool to capture this information. External validation of the S-FFQ per se was demonstrated by its use to measure current dietary Zn intake in adult women living in Indonesia through comparison with 3-day food records in a sample of 6 women.
Given the much broader importance to lifelong health of the child of the maternal diet more generally during pregnancy, knowledge that a short food frequency questionnaire administered retrospectively can be informative with regard to measuring intake of a specific nutrient is useful information to other researchers wishing to survey diet during pregnancy in a similar retrospective manner. However, an important caveat is that extension of this principle to other cultures cannot be assumed. Cultural dietary norms may introduce more uncertainly and inaccuracy in other populations, thus similar tools would require suitable validation in other contexts.
Studies on zinc intake and zinc status of mothers, infants and children in Indonesia indicate that inadequate zinc intake and zinc deficiency are prevalent. For example, in West Java, $25\%$ of mothers and $17\%$ of their infants (2.4–10.5 months) were found to be zinc-deficient [25], and a recent study found that $43.2\%$ of a sample of children in a rural village in Indonesia had inadequate zinc intake and the majority of these had low serum zinc concentration [26]. A sample of women living in the coastal area of Makassar all had low serum zinc concentration (<66 microg/100 ml) 4–6 weeks after giving birth and $21\%$ had had zinc intake of less that $80\%$ RDI (average 15.9 mg/d) [27]. The calculated average daily zinc intake in mothers in our sample (11.8 mg) was lower than that of this sample from Makassar (15.9 mg), which indicates that children in our sample from urban areas around Jakarta will have been exposed to sub-optimal zinc nutrition during early life. The apparent prevalence of inadequate zinc intake in Indonesian mothers and children mean that future studies on the impacts of inadequate early zinc nutrition on the lifelong health of the child are important. A retrospective measurement of this, as we show to be afforded by measurement in the deciduous dentition, will – compared with longitudinal studies - expediate outcomes of research on this topic.
Our discovery that the Zn/Sr ratio in deciduous teeth correlated with maternal Zn intake during late pregnancy (median 111 days or 87 days before giving birth for incisors and molars, respectively) and 0–3 months post-partum, when most infants were breastfed, should be agnostic of population and dietary culture. Thus, we have uncovered a biomarker of early developmental exposure to Zn nutrition that could be applied independently of dietary records in other studies and in other populations to uncover new information about the effect of early life Zn nutrition on later health outcomes.
Seven of the 18 children (3 of high Zn mothers and 4 of low Zn mothers) were breastfed exclusively up to the age of 6 months. The significant effect of the maternal diet on the Zn/Sr ratio was still evident in the areas mineralised from late pregnancy to 3–6 months in this group. This is consistent with `the mineral record being reflective of indirect dietary exposure through breast milk rather than only a consequence of receiving foods more similar to the maternal diet in the full group of 18 children. However, this significant effect was lost when all time points (also 6–9 and 9–12 months) were included in the model. A likely explanation is a combination of a reduction in the power of the statistical analysis combined with a reduced influence of the maternal diet as breastfeeding was replaced with other foods.
Evidence that Zn exposure in utero can affect susceptibility to features of the metabolic syndrome, such as cardiovascular disease [28], and the association of *Zn status* with type 2 diabetes mellitus [6], points to a hypothesis that *Zn status* in very early life may influence susceptibility to type 2 diabetes in later life. We posit an interrogation of this hypothesis as one of many potential future studies using retrospective measures of zinc exposure during very early life made using laser ablation ICP-MS in deciduous teeth.
We measured Sr primarily because some data suggest that the Sr/Ca ratio can be used as an indicator of the period of breastfeeding. We observed that the median Sr/Ca ratio decreased across the time intervals sampled. It was argued, and demonstrated in a small sample of human deciduous teeth, that the ratio should decline after birth if the infant is breastfed because of a greater activating effect of the mammary gland than the placenta on Ca transfer but increase if the infant is bottle-fed [29]. Extending this argument predicts that weaning should coincide with an increase in the ratio. In our study, the ratio decreased after birth in 14 of the 18 teeth, commensurate with this prediction and these previous observations. However, the predicted relationship with bottle-feeding was not observed. Two of the infants were bottle-fed immediately or within 2 weeks of birth and both showed a decline in Sr/Ca after birth. We observed no increase in Sr/Ca in any of the teeth after weaning (reported by mothers as ranging from 2 to 8 months; mode 6 months) nor on the reported point at which table foods were introduced, which for 5 of the infants was within the 12 month sampling period. Thus, although the pattern of change in the Sr/Ca ratio was consistent with the observations reported previously our observations do not support the hypothesis that this is due to an activating effect of breast milk on Ca transfer. The measurement of Sr in deciduous teeth along with other elements, as in this study, may be useful in future research in the context of its purported but debated cariostatic properties [30]. While some in vitro studies have demonstrated that Sr – particularly in combination with Fl - can promote caries rehardening - e.g., [31] - epidemiological data are confounded by the co-presence in the water supply of other trace elements with possible similar properties. Use of the trace element profile of the deciduous dentition alongside records of caries may deconvolute some of these interactions and shed more light on the role Sr can play in protection against dental caries.
## Conclusion
Maternal dietary Zn intake correlated with the ratio of Zn/Sr deposited in the developing tooth over the period of late pregnancy and early infancy. This measure is a promising tool to record exposure to Zn during this period of development for use in research and also, potentially, to identify individuals at heightened risk of detrimental impacts of poor early life Zn nutrition on health in later life and to implement preventative interventions.
## 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 Committee of The Medical Research Ethics of the Faculty of Medicine University of Indonesia (526/UN2.F1/ETIK/2014). Written informed consent to participate in this study was provided by the participants’ legal guardian/next of kin.
## Author contributions
NAW: collected the tooth samples, administered the FFQs and conducted the statistical analysis of data. LAW: provided supervision to NAW. WD: provided training to NAW on tooth preparation and advised on aspects of the manuscript. DAB and TJS: conducted the LA-ICP-MS measurements. DF: interpreted the study findings and provided substantial input into writing of the manuscript. RAV: conceived the study, provided supervision to NAW and provided input into the manuscript. All authors contributed to the article and approved the submitted version.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher's note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/froh.2023.1119086/full#supplementary-material.
## References
1. Burton MA, Lillycrop KA. **Nutritional modulation of the epigenome and its implication for future health**. *Proc Nutr Soc* (2019) **78** 305-12. DOI: 10.1017/S0029665119000016
2. Tarry-Adkins JL, Ozanne SE. **Poor early growth and age-associated disease**. *Subcell Biochem* (2019) **91** 1-19. DOI: 10.1007/978-981-13-3681-2_1
3. Andreini C, Banci L, Bertini I, Rosato A. **Counting the zinc-proteins encoded in the human genome**. *J Proteome Res* (2006) **5** 196-201. DOI: 10.1021/pr050361j
4. Wessels I, Fischer HJ, Rink L. **Dietary and physiological effects of zinc on the immune system**. *Annu Rev Nutr* (2021) **41** 133-75. DOI: 10.1146/annurev-nutr-122019-120635
5. Al-Khafaji Z, Brito S, Bin BH. **Zinc and zinc transporters in dermatology**. *Int J Mol Sci* (2022) **23** 16165. DOI: 10.3390/ijms232416165
6. Fernandez-Cao JC, Warthon-Medina M, Moran VH, Arija V, Doepking C, Serra-Majem L. **Zinc intake and status and risk of type 2 diabetes mellitus: a systematic review and meta-analysis**. *Nutrients* (2019) **11**. DOI: 10.3390/nu11051027
7. Carducci B, Keats EC, Bhutta ZA. **Zinc supplementation for improving pregnancy and infant outcome**. *Cochrane Database Syst Rev* (2021) **3** CD000230. DOI: 10.1002/14651858
8. Polanska K, Hanke W, Krol A, Gromadzinska J, Kuras R, Janasik B. **Micronutrients during pregnancy and child psychomotor development: opposite effects of zinc and selenium**. *Environ Res* (2017) **158** 583-9. DOI: 10.1016/j.envres.2017.06.037
9. Lowe NM, Fekete K, Decsi T. **Methods of assessment of zinc status in humans: a systematic review**. *Am J Clin Nutr* (2009) **89** 2040S-51S. DOI: 10.3945/ajcn.2009.27230G
10. Hennigar SR, Kelley AM, McClung JP. **Metallothionein and zinc transporter expression in circulating human blood cells as biomarkers of zinc Status: a systematic review**. *Adv Nutr* (2016) **7** 735-46. DOI: 10.3945/an.116.012518
11. Wahono NA, Ford D, Wakeling LA, Valentine RA. **The presence and response to Zn of ZnT family mRNAs in human dental pulp**. *Metallomics* (2019) **11** 613-20. DOI: 10.1039/c8mt00343b
12. Klimuszko E, Orywal K, Sierpinska T, Sidun J, Golebiewska M. **The evaluation of zinc and copper content in tooth enamel without any pathological changes - an in vitro study**. *Int J Nanomedicine* (2018) **13** 1257-64. DOI: 10.2147/IJN.S155228
13. Smith TM. **Teeth and human life-history evolution**. *Annu Rev Anthropol* (2013) **42** 191-208. DOI: 10.1146/annurev-anthro-092412-155550
14. Smith TM, Cook L, Dirks W, Green DR, Austin C. **Teeth reveal juvenile diet, health and neurotoxicant exposure retrospectively: what biological rhythms and chemical records tell us**. *Bioessays* (2021) **43** e2000298. DOI: 10.1002/bies.202000298
15. Shepherd TJ, Dirks W, Manmee C, Hodgson S, Banks DA, Averley P. **Reconstructing the life-time lead exposure in children using dentine in deciduous teeth**. *Sci Total Environ* (2012) **425** 214-22. DOI: 10.1016/j.scitotenv.2012.03.022
16. Shepherd TJ, Dirks W, Roberts NM, Patel JG, Hodgson S, Pless-Mulloli T. **Tracing fetal and childhood exposure to lead using isotope analysis of deciduous teeth**. *Environ Res* (2016) **146** 145-53. DOI: 10.1016/j.envres.2015.12.017
17. Humphrey LT, Jeffries TE, Dean MC, Irish JD, Nelson GC. **Micro spatial distributions of lead and zinc in human deciduous tooth enamel**. *Technique and application in dental anthropology* (2008) 87-110
18. Müller W, Nava A, Evans D, Rossi PF, Alt KW, Bondioli L. **Enamel mineralization and compositional time-resolution in human teeth evaluated via histologically-defined LA-ICPMS profiles**. *Geochim Cosmochim Acta* (2019) **255** 105-26. DOI: 10.1016/j.gca.2019.03.005
19. Friedman A, Bauer JA, Austin C, Downs TJ, Tripodis Y, Heiger-Bernays W. **Multiple metals in children's deciduous teeth: results from a community-initiated pilot study**. *J Expo Sci Environ Epidemiol* (2022) **32** 408-17. DOI: 10.1038/s41370-021-00400-x
20. Dean MC, Spiers KM, Garrevoet J, Le Cabec A. **Synchrotron x-ray fluorescence mapping of Ca, Sr and Zn at the neonatal line in human deciduous teeth reflects changing perinatal physiology**. *Arch Oral Biol* (2019) **104** 90-102. DOI: 10.1016/j.archoralbio.2019.05.024
21. Smith TM, Austin C, Ávila JN, Dirks W, Green DR, Williams IS. **Permanent signatures of birth and nursing initiation are chemically recorded in teeth**. *J Archaeol Sci* (2022) **140** 105564. DOI: 10.1016/j.jas.2022.105564
22. Nelson M, Haraldsdottir J. **Food photographs: practical guidelines II. Development and use of photographic atlases for assessing food portion size**. *Public Health Nutr* (1998) **1** 231-7. DOI: 10.1079/phn19980039
23. Chan S, Ferguson E, Bailey K, Fahmida U, Harper T, Gibson R. **The concentrations of iron, calcium, zinc and phytate in cereals and legumes habitually consumed by infants living in East Lombok, Indonesia**. *J Food Compos Anal* (2007) **27** 609-17. DOI: 10.1016/j.jfca.2007.03.003
24. Norhaizan ME, Nor Faizadatul Ain AW. **Determination of phytate, iron, zinc, calcium contents and their molar ratios in commonly consumed raw and prepared food in Malaysia**. *Malays J Nutr* (2009) **15** 213-22. PMID: 22691819
25. Dijkhuizen MA, Wieringa FT, West CE. **Concurrent micronutrient deficiencies in lactating mothers and their infants in Indonesia**. *Am J Clin Nutr* (2001) **73** 786-91. DOI: 10.1093/ajcn/73.4.786
26. Flora R, Fajar N, Febry F, Yuliana I, Yuliarti Y, Nurlaili N. **Zinc intake, zinc Serum levels, and intelligence in school children in rural Macedonian**. *J Med Sci* (2021) **9** 391-7. DOI: 10.1038/ejcn.2015.60
27. Indriasari R, Syam A, Kurniati Y, Mansur M. **Zinc deficiency and inadequate zinc intake among postpartum women in coastal area of Makassar, Indonesia**. *J Nutr Sci Vitaminol* (2020) **66** S99-102. DOI: 10.3177/jnsv.66.S99
28. Tomat AL, Costa Mde L, Arranz CT. **Zinc restriction during different periods of life: influence in renal and cardiovascular diseases**. *Nutrition* (2011) **27** 392-8. DOI: 10.1016/j.nut.2010.09.010
29. Humphry LT, Dean MC, Jeffries TE, Bailey SE, Hublin J-J. **An evaluation of changes in strontium/calcium ratios across the neonatal line in human deciduous teeth**. *Dental perspectives on human evolution* (2007) 303-19
30. Lippert F, Hara AT. **Strontium and caries: a long and complicated relationship**. *Caries Res* (2013) **47** 34-49. DOI: 10.1159/000343008
31. Lippert F. **The effects of fluoride, strontium, theobromine and their combinations on caries lesion rehardening and fluoridation**. *Arch Oral Biol* (2017) **80** 217-21. DOI: 10.1016/j.archoralbio.2017.04.022
|
---
title: 'Clinical features and independent predictors of postoperative refractory trauma
to anal fistula combined with T2DM: A propensity score-matched analysis-retrospective
cohort study'
authors:
- Xiao Tang
- Taohong He
- Xinyi Li
- Ya Liu
- Yuqi Wu
- Gehang You
- Jie Li
- Yu Yun
- Lei Wu
- Li Li
- Jian Kang
journal: Frontiers in Surgery
year: 2023
pmcid: PMC9998506
doi: 10.3389/fsurg.2023.1119113
license: CC BY 4.0
---
# Clinical features and independent predictors of postoperative refractory trauma to anal fistula combined with T2DM: A propensity score-matched analysis-retrospective cohort study
## Abstract
### Background
Refractory wound is a common postoperative complication in anal fistula surgery, when combined with type 2 diabetes mellitus (T2DM) it presents a slower recovery time and more complex wound physiology. The study aims to investigate factors associated with wound healing in patients with T2DM.
### Materials and methods
365 T2DM patients who underwent anal fistula surgery at our institution were recruited from June 2017 to May 2022. Through propensity score-matched (PSM) analysis, multivariate logistic regression analysis was applied to determine independent risk factors affecting wound healing.
### Results
122 pairs of patients with no significant differences were successfully established in matched variables. Multivariate logistic regression analysis revealed that uric acid (OR: 1.008, $95\%$ CI: 1.002–1.015, $$p \leq 0.012$$), maximal fasting blood glucose (FBG) (OR: 1.489, $95\%$ CI: 1.028–2.157, $$p \leq 0.035$$) and random intravenous blood glucose (OR: 1.130, $95\%$ CI: 1.008–1.267, $$p \leq 0.037$$) elevation and the incision at 5 o’clock under the lithotomy position (OR: 3.510, $95\%$ CI: 1.214–10.146, $$p \leq 0.020$$) were independent risk factors for impeding wound healing. However, neutrophil percentage fluctuating within the normal range can be considered as an independent protective factor (OR: 0.906, $95\%$ CI: 0.856–0.958, $$p \leq 0.001$$). After executing the receiver operating characteristic (ROC) curve analysis, it was found that the maximum FBG expressed the largest under curve area (AUC), glycosylated hemoglobin (HbA1c) showed the strongest sensitivity at the critical value and maximum postprandial blood glucose (PBG) had the highest specificity at the critical value. To promote high-quality healing of anal wounds in diabetic patients, clinicians should not only pay attention to surgical procedures but also take above-mentioned indicators into consideration.
## Introduction
Anal fistula, mostly formed by inflammatory cells, collagen and epithelial tissue is a pathological channel between the anal or rectum and the skin [1]. Its incidence fluctuates between $\frac{10.4}{100}$,000 to $\frac{23.2}{100}$,000 [2, 3], however, due to the privacy of the lesion site and the low consultation rate, the true incidence might be higher. The causes of anal fistulas are complex, but most of them are formed after the rupture of the perianal abscess which is caused by anal gland infection [4]. Sometimes they are considered to be different stages of the same disease. The abundance of perianal connective tissue and tissue spaces can contribute to complex and variable anal fistula alignment. About $59.0\%$–$71.0\%$ of them are low anal fistulas, and $62.3\%$–$67.0\%$ of them are intersphincteric fistulas [5, 6]. The nature of the anal fistula and the high cure rate have dictated that anal fistulectomy and cutting seton surgery are still the most commonly performed clinical procedures (7–9). However, after these procedures, surgeons must face the cruel fact that these wounds rely on a large amount of granulation tissue to fill the defects. Previous studies have found even in procedures with smaller trauma areas, wound recovery time remained long (10–12). It is crucial to reduce the effect of confounding factors by adjusting the baseline data.
When combined with other underlying diseases, the trauma-healing process is even lumpier. Patients with T2DM in developing countries, represented by China and India, are expected to increase by $69\%$ between 2010 and 2030, reaching a staggering 693 million by 2045 (13–16). Diabetic patients are often associated with slow wound healing, and when it comes to wounds occurring in the dense anal nerve, it greatly increases patients’ pain and decreases their quality of life. As a complex recovery process, the circulatory metabolic state of the body could continuously influence wound healing by reshaping the internal environment. Elevated blood glucose affects wound healing process by upgrading inflammation levels, dysfunction oxidative stress, and slowing down angiogenesis (17–20). Previous reports of other surgeries reveal that abnormal blood glucose metabolism increases the risk of infection and impedes wound healing after other procedures [21, 22]. However, there is a lack of studies on postoperative wound recovery in anal fistula patients with T2DM. This study included statistics of laboratory tests, surgical modalities, and postoperative treatment. The information gap was filled on anal fistula wound healing in T2DM patients. As a continuous and changing process, wound recovery is more complex and variable in diabetic patients. PSM is necessary to exclude some unpredictable confounding factors and logistic regression could be applied to target risk factors for wound healing. Meanwhile, the effect of blood glucose has been elaborated by using the receiver operating characteristic (ROC) curve. The study provides a therapeutic direction to promote rapid healing after anal fistula surgery.
## Diseases definition and data collection
The diagnosis of anal fistula is based on the German S3 guidelines: anal abscess and fistula [23]. All patients were diagnosed with anal fistula by anal finger examination, anoscope examination, radiographic examination (including rectal endoluminal ultrasound, pelvic CT, or MRI), or intraoperative probe/methylene blue staining, and the number of internal orifices was counted by these techniques. The diagnostic criteria for T2DM were based on the latest Chinese guidelines for the prevention and treatment of T2DM set by the Chinese Diabetes Society [24, 25]. And the diagnosis was assigned by an endocrinologist. Relevant data were collected on the cases, including demographic characteristics, clinical features, laboratory and ancillary tests at admission, anal fistula-related information (e.g., previous surgical history, anal fistula types, number of internal orifices, etc.), pre- and post-surgical treatments, and surgical modalities. Non-healing (refractory) group refers to trauma that cannot be repaired in time with conventional therapy or wounds that can not achieve functional recovery and anatomical integrity [26]. The last routine dressing change time in the outpatient clinic was collected as the outcome indicator. Judged by the specialist anorectologist and the definition of the relevant literature, patients were divided into the non-healing (refractory) group or healing group according to whether its recovery period is longer than 35 days (27–29).
Among the underlying diseases, hypertensive disease and non-alcoholic fatty liver diseases are listed independently. Chronic cardiovascular diseases included coronary atherosclerotic heart disease and lacunar cerebral infarction. Chronic lung diseases included tuberculosis, chronic obstructive pulmonary disease, and chronic pulmonary heart disease. Chronic liver diseases included chronic viral hepatitis B, cirrhosis of the liver, hepatic hemangioma, etc.
## Inclusion exclusion criteria and follow up
This clinical retrospective study collected 408 cases of anal fistula combined with T2DM attending the Hospital of Chengdu University of Chinese Medicine from June 2017 to May 2022. The study excluded anal fistula caused by Crohn's disease or exotic injuries (from indigestible diets or external devices). Extremely complicated fistulas that were not suitable for fistulotomy, cutting seton surgery or a combination procedure were screened out. And fistulas in the active phase of anal fistula were excluded. Research recruited 408 patients, but 43 patients were eliminated because of recurrence (within 6 months) ($$n = 35$$), severe clinical data deficits, failure to perform surgery, or missed visits. In the end, a total of 365 patients were included. Up on wound recovery time, patients were divided into the non-healing (refractory) group ($$n = 170$$) and the healing group ($$n = 195$$), respectively. After matching, there were 122 patients in each group. The detailed operation procedure was shown in Figure 1. This study was approved by the ethics committee of the Hospital of Chengdu University of Traditional Chinese Medicine (ethics number: 2022KL-018). The research was a retrospective clinical study and only the patients’ previous treatment data were extracted through the medical system. Therefore, no informed consent was required from the patients.
**Figure 1:** *Research flow chart.*
## Statistical analysis
All statistical analysis were performed with SPSS version 22.0 (IBM, Armonk, NY, USA). Continuous variables were expressed as mean ± standard (SD) deviation or M (QL, QU). Before conducting propensity score, the student's t-test and Mann–Whitney U test were applied for continuous variables, whereas the χ2 test was used for dichotomous variables. To eliminate relevant confounders and increase comparability between different group, PSM analysis was performed using nearest-neighbor matching (1:1) adjusted for baseline data including: gender, age, body mass index (BMI), history of anal fistula surgery, the number of internal orifices (≥2 or not) and the number of wounds (≥3 or not), with the caliper value set at 0.02. After PSM, variables were tested by paired t-test and Wilcoxon rank sum test for continuous variables. McNemar test and Fisher's exact test for categorical variables. Risk factors for non-healing wound were identified by univariate and multivariate logistic regression models, and the degree of association was shown by calculating odds ratio (OR). Variance inflation factor (VIF) and tolerance were applied to determine the covariance between internal orifice and incision, as well as multicollinearity among variables before multivariate regression. To prevent overfitting, multiple variables with statistically different ($p \leq 0.05$) were selected for multivariate logistic regression analysis to identify independent predictors of healing trauma. Also, we used ROC curve to assess the sensitivity and specificity of glucose indicators in wound healing. All statistical tests were two-sided, and $p \leq 0.05$ was considered statistically significant. The experimental procedure of the article was carried out with reference to the relevant authoritative literature [27].
## Description of baseline data
The majority patients were male, reaching astonishing 347 cases ($95.07\%$). The healing group had more female patients than the refractory group. However, there was no discernible difference between each other ($$p \leq 0.248$$). BMI (26.83 ± 0.33 vs. 26.97 ± 0.28 Kg/m2) and previous surgical history (19 vs. 21) in the healing group were slightly higher than those in the non-healing group. According to more than two types of ancillary findings, the number of patients with multiple internal orifices was larger in the healing group ($$n = 173$$, $88.72\%$), while cases with ≥3 wounds were more in the refractory group (42 vs. 89, $$p \leq 0.003$$). No significant difference was discovered between two groups in terms of underlying diseases ($p \leq 0.05$).
122 pairs of patients were successfully matched (Table 1) with no significant difference between gender ($$p \leq 1.000$$), age ($$p \leq 0.923$$), BMI ($$p \leq 0.547$$), previous anal fistula surgery history ($$p \leq 1.000$$), number of internal orifices ($$p \leq 1.000$$), and number of wounds ($$p \leq 1.000$$).
**Table 1**
| Variables | Before PSM | Before PSM.1 | Before PSM.2 | Before PSM.3 | After PSMa | After PSMa.1 | After PSMa.2 | After PSMa.3 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Variables | Total (n = 365) | Refractory group (n = 175) | Healing group (n = 190) | p valueb | Total (n = 244) | Refractory group (n = 122) | Healing group (n = 122) | p valuec |
| Gender, n (%) | | | | 0.248 | | | | 1.000 |
| Male | 347 (95.07%) | 164 (96.47%) | 183 (93.85%) | | 240 (98.26%) | 120 (98.36%) | 120 (98.36%) | |
| Female | 18 (4.93%) | 6 (3.53%) | 12 (6.15%) | | 4 (1.64%) | 2 (1.67%) | 2 (1.67%) | |
| Age (year) | 50 (40, 57) | 50 (41, 59) | 50 (44, 55) | 0.779 | 48.54 ± 0.79 | 48.62 ± 1.17 | 48.45 ± 1.07 | 0.923 |
| BMI (Kg/m2) | 26.91 ± 0.21 | 26.83 ± 0.33 | 26.97 ± 0.28 | 0.744 | 27.05 ± 0.25 | 27.11 ± 0.41 | 26.98 ± 0.33 | 0.547 |
| Previous anal fistula surgery, n (%) | | | | 0.901 | | | | 1.000 |
| Yes | 40 (10.96%) | 19 (11.18%) | 21 (10.77%) | | 26 (10.66%) | 14 (11.48%) | 12 (9.84%) | |
| No | 325 (89.04%) | 151 (88.82%) | 174 (89.23%) | | 218 (89.34%) | 108 (88.52%) | 110 (90.16%) | |
| Number of internal orifices, n (%) | | | | 0.003* | | | | 1.000 |
| =1 | 304 (83.29%) | 131 (77.06%) | 173 (88.72%) | | 212 (86.89%) | 106 (86.89%) | 106 (86.89%) | |
| ≥2 | 61 (16.71%) | 39 (22.94%) | 22 (11.28%) | | 32 (13.11%) | 16 (13.11%) | 16 (13.11%) | |
| Number of wounds, n (%) | | | | <0.001* | | | | 1.000 |
| <3 | 234 (64.11%) | 81 (47.64%) | 153 (78.76%) | | 160 (65.57%) | 80 (65.57%) | 80 (65.57%) | |
| ≥3 | 131 (35.89%) | 89 (52.35%) | 42 (21.54%) | | 84 (34.43%) | 42 (34.43%) | 42 (34.43%) | |
| Other comorbidities, n (%) | | | | | | | | |
| Hypertensive disease | 105 (28.77%) | 54 (31.76%) | 51 (26.15%) | 0.238 | 76 (34.84%) | 42 (34.43%) | 34 (27.87%) | 0.788 |
| Non-alcoholic fatty liver diseases | 235 (64.38%) | 114 (67.06%) | 121 (62.05%) | 0.319 | 161 (65.98%) | 79 (64.75%) | 82 (67.21%) | 0.807 |
| Prostatic hyperplasia | 61 (16.71%) | 26 (15.29%) | 35 (17.95%) | 0.498 | 43 (17.62%) | 19 (15.57%) | 24 (19.67%) | 0.188 |
| Chronic pulmonary disease | 10 (2.74%) | 6 (3.53%) | 4 (2.05%) | 0.574 | 7 (2.87%) | 4 (3.28%) | 3 (2.46%) | 0.727 |
| Chronic liver disease | 16 (4.38%) | 6 (3.53%) | 10 (5.13%) | 0.610 | 10 (4.10%) | 5 (4.10%) | 5 (4.10%) | 0.727 |
| Cardiovascular diseases | 15 (4.11%) | 10 (5.88%) | 5 (2.56%) | 0.111 | 10 (4.10%) | 5 (4.10%) | 5 (4.10%) | 1.000 |
## Laboratory and ancillary tests
The results of the detailed laboratory and ancillary tests are presented in Table 2. The concentration of plasma sodium (137.45 mmol/L [135.78, 140.10] vs. 139.45 mmol/L [137.58, 140.83], $p \leq 0.001$) and plasma chloride (102.94 ± 0.34 vs. 103.94 ± 0.29 mmol/L, $$p \leq 0.039$$) showed a higher level in non-healing group. No significant abnormalities or intergroup differences were discovered in the rest of electrolytes such as potassium, calcium, magnesium and phosphorus in plasma. Uric acid, one of the key metabolic byproducts in the body, demonstrated an aberrant range in the non-healing group, where the level reached 359.50 (292.50, 433.50) mmol/L, compared to the healing group's 314.00 mmol/L (255.50.372.00). In terms of lipid metabolism, both groups displayed abnormal ranges in cholesterol (5.46 mmol/L [4.43, 6.25] vs. 5.16 mmol/L [4.37, 6.20]) and triglyceride (1.99 mmol/L [1.34, 2.78] vs. 2.30 mmol/L [1.67, 3.41]). However, no statistical significance ($p \leq 0.05$) was detected in cholesterol, triglyceride, high-density lipoprotein (LDL), low-density lipoprotein (HDL). Meanwhile, except for the A/G ratio (1.60 [1.40, 1.80] vs. 1.70 [1.50, 1.95], $$p \leq 0.041$$), the remaining indicators of albumin, globulin, alanine transaminase (ALT) and aspartate transaminase (AST) were not found to be different from the normal range or demonstrate differences between groups.
**Table 2**
| Variables | Normal range | Total (n = 244) | Refractory group (n = 122) | Healing group (n = 122) | p value |
| --- | --- | --- | --- | --- | --- |
| Axillary temperature (°C) | 36.0–37.1 | 36.4 (36.2, 36.7) | 36.5 (36.2, 36.7) | 36.4 (36.2, 36.6) | 0.113 |
| K (mmol/L) | 3.5–5.3 | 4.11 (3.80, 4.45) | 4.21 (3.83, 4.51) | 4.02 (3.79, 4.36) | 0.064 |
| Na (mmol/L) | 137–147 | 138.60 (136.90, 140.48) | 137.45 (135.78, 14 0.10) | 139.45 (137.58, 140.83) | <0.001* |
| Cl (mmol/L) | 99–110 | 103.44 ± 0.23 | 102.94 ± 0.34 | 103.94 ± 0.29 | 0.039* |
| Ca (mmol/L) | 2.11–2.52 | 2.32 (2.22, 2.40) | 2.30 (2.21, 2.40) | 2.33 (2.23, 2.42) | 0.064 |
| Mg (mmol/L) | 0.75–1.02 | 0.82 ± 0.01 | 0.81 ± 0.01 | 0.82 ± 0.01 | 0.422 |
| P (mmol/L) | 0.85–1.51 | 1.10 ± 0.01 | 1.12 ± 0.02 | 1.09 ± 0.02 | 0.353 |
| Blood urea nitrogen (mmol/L) | 2.6–7.5 | 5.90 (4.66, 7.00) | 5.87 (4.13, 6.91) | 5.97 (5.00, 7.09) | 0.079 |
| SCR (mmol/L) | 41–73 | 68.8 (61.00, 81.38) | 69.70 (60.60, 84.93) | 67.90 (61.98, 78.33) | 0.467 |
| Uric acid (mmol/L) | 155–357 | 330.00 (274.00, 399.00) | 359.50 (292.50, 433.50) | 314.00 (255.50.372.00) | <0.001* |
| Random intravenous blood glucose (mmol/L) | 3.93–6.11 | 10.50 (7.45, 15.33) | 12.39 (8.27, 16.64) | 8.88 (6.61, 12.41) | <0.001* |
| Albumin (g/L) | 40–55 | 46.10 (43.23, 49.18) | 46.05 (43.28, 49.00) | 46.15 (43.20, 49.23) | 0.751 |
| Globulin (g/L) | 20–40 | 28.20 (24.53, 32.20) | 28.75 (25.00, 33.10) | 27.45 (24.38, 31.35) | 0.150 |
| A/G | 1.2–2.4 | 1.61 (1.47, 1.90) | 1.60 (1.40, 1.80) | 1.70 (1.50, 1.95) | 0.041* |
| ALT (U/L) | 7–40 | 27.00 (17.00, 44.75) | 26.00 (17.00, 45.00) | 28.50 (18.00, 44.25) | 0.335 |
| AST (U/L) | 13–35 | 23.50 (18.00, 34.00) | 23.00 (17.00, 34.00) | 24.00 (18.00, 34.00) | 0.960 |
| GGT (U/L) | 35–100 | 37.50 (24.00, 59.00) | 38.50 (25.75, 59.00) | 37.00 (22.75, 60.50) | 0.490 |
| Total bile acid (μmol/L) | 0–10 | 4.80 (2.63, 7.50) | 4.90 (2.48, 7.80) | 4.75 (2.85, 7.50) | 0.536 |
| Total bilirubin (μmol/L) | 0–21 | 14.50 (10.33, 20.35) | 14.90 (10.25, 20.50) | 13.85 (10.48, 19.30) | 0.510 |
| Cholesterol (mmol/L) | <5.18 | 5.20 (4.39, 6.21) | 5.46 (4.43, 6.25) | 5.16 (4.37, 6.20) | 0.839 |
| TG (mmol/L) | <1.7 | 2.09 (1.53, 3.10) | 1.99 (1.34, 2.78) | 2.3 (1.67, 3.41) | 0.091 |
| HDL (mmol/L) | >1 | 1.10 (0.93, 1.41) | 1.09 (0.91, 1.57) | 1.10 (0.93, 1.29) | 0.168 |
| LDL (mmol/L) | <3.3 | 2.70 ± 0.06 | 2.59 ± 0.08 | 2.78 ± 0.10 | 0.093 |
| White blood cell count (×109/L) | 3.5–9.5 | 7.28 (5.84, 9.07) | 7.39 (5.88, 9.15) | 7.04 (5.77, 8.78) | 0.493 |
| Neutrophil count (×109/L) | 1.8–6.3 | 4.90 (3.66, 6.28) | 4.86 (3.70, 6.30) | 4.93 (3.64, 6.27) | 0.803 |
| Lymphocyte count (×109/L) | 1.1–3.2 | 1.67 (1.29, 2.16) | 1.66 (1.35, 2.23) | 1.70 (1.24, 2.14) | 0.501 |
| Monocyte count (×109/L) | 0.1–0.6 | 0.42 (0.3, 0.56) | 0.45 (0.32, 0.56) | 0.40 (0.29, 0.56) | 0.397 |
| NEUT% | 40–75 | 67.30 (58.00, 74.20) | 66.15 (54.78, 74.13) | 67.80 (61.00, 74.88) | 0.014* |
| LY% | 20–50 | 26.20 (18.15, 47.33) | 28.25 (18.00, 53.23) | 26.20 (18.15, 47.33) | 0.039* |
| MO% | 3–10 | 7.35 (5.30, 29.53) | 8.75 (5.60, 34.50) | 7.35 (5.30, 29.53) | 0.008* |
| Red blood cell count (×109/L) | 3.8–5.1 | 5.06 ± 0.03 | 5.07 ± 0.04 | 5.05 ± 0.05 | 0.838 |
| Hemoglobin (g/L) | 115–160 | 152.24 ± 0.98 | 152.00 ± 1.33 | 152.48 ± 1.44 | 0.754 |
| RDW-SD (fl) | 36.4–46.3 | 41.30 (38.90, 43.38) | 40.75 (37.88, 42.80) | 42.15 (39.58, 43.75) | 0.001* |
| PLT (×109/L) | 100–300 | 211.14 ± 3.78 | 215.22 ± 5.39 | 207.25 ± 5.33 | 0.322 |
| C-reactive protein (mg/L) | 0–5 | 11.66 (3.91, 24.03) | 13.15 (4.89, 30.45) | 9.86 (2.51, 21.63) | 0.064 |
| Glycosylated hemoglobin (%) | 4.1–6.1 | 8.10 (6.93, 9.70) | 8.80 (7.68, 10.03) | 7.40 (6.50, 8.73) | 0.001* |
| Maximum FBG (mmol/L) | 3.9–6.0 | 8.00 (6.90, 9.80) | 9.00 (7.50, 10.40) | 7.35 (6.70, 8.25) | <0.001* |
| Maximum PBG (mmol/L) | 7.8–11.1 | 14.50 (12.80, 16.60) | 15.15 (13.20, 17.53) | 14.05 (12.28, 15.53) | <0.001* |
| Urine glucose | Negative | | | | 0.203 |
| Normal | | 60 (24.59%) | 25 (20.49%) | 35 (28.69%) | |
| Abnormal | | 184 (75.41%) | 97 (79.51%) | 87 (71.31%) | |
| Urine ketone | Negative | | | | 0.008* |
| Normal | | 190 (77.87%) | 85 (69.68%) | 105 (86.07%) | |
| Abnormal | | 54 (22.13%) | 37 (30.33%) | 17 (13.93%) | |
| Endoanal ultrasound, n (%) | | 143 (58.61%) | 71 (58.20%) | 72 (59.02%) | 1.000 |
| Detection of ultrasound | Positive | 136 (55.74%) | 68 (55.74%) | 68 (55.74%) | 1.000 |
Anal fistula lesions are confined by collagen and epithelial tissue and the infection is mostly under control. The results showed that white blood cell count (WBC) and neutrophil percentage (NEUT%) fluctuated in the normal range in both groups, while C-reactive protein (CRP) was higher than normal range. The NEUT% in the healing group reached $67.80\%$ (61.00, 74.88) was greater than in non-healing group [$66.15\%$ (54.78, 74.13), $$p \leq 0.014$$]. At the same time, lymphocyte percentage ($28.25\%$ [18.00, 52.23] vs. $26.20\%$ [18.15, 47.33], $$p \leq 0.039$$) and monocyte percentage ($8.75\%$ [5.60, 34.50] vs. $7.35\%$ [5.30, 29.53], $$p \leq 0.008$$) in refractory group also exhibited a superior level. 143 patients carried out anorectal endoscopic ultrasonography, and the positive detection of fistulas internal orifices in the implemented patients reached $94.44\%$.
It exhibits a clear distinction between glucose indicators. It is worth mentioning that random intravenous blood glucose (tested with other biochemical indicators) and HbA1c were extracted preoperatively. The finger-prick glucose test was used to monitor FBG and PBG. Patients in the healing group had lower random intravenous blood glucose (8.88 mmol/L [6.61–12.41]) compared to 12.39 mmol/L (8.27–16.64) in refractory group, with a p value less than 0.001. In response to medium and long-term blood glucose levels, HbA1c levels at admission and maximum PBG, and FBG levels throughout hospitalization has been collected. We noticed that HbA1c indicated a higher standard in non-healing group (8.80 mmol/L [7.68, 10.03] vs. 7.40 mmol/L [6.50, 8.73], $$p \leq 0.001$$). During the hospitalization, patients’ maximum FBG and the maximum PBG respectively reached 9.00 mmol/L (7.50, 10.40) and 15.15 mmol/L (13.20, 17.53) in the refractory group. Striking statistical differences between the two groups were shown ($p \leq 0.001$). Among the glucose related indicators, the area under the ROC curve of maximum FBG reached 0.724, followed by HbA1c 0.667, random blood glucose 0.665 and maximum PBG 0.640 (Figure 2). The critical values of HbA1c, random blood glucose, maximum FBG and maximum PBG were $7.69\%$, 12.50, 8.35 and 16.05 mmol/L, respectively. The sensitivity of each index at the critical values was $75.4\%$, $50.0\%$, $64.8\%$, and $41.8\%$, and the specificity was $58.2\%$, $76.2\%$, $74.6\%$, and $82.8\%$, respectively.
**Figure 2:** *Receiver operating characteristic (ROC) curves of different glycemic indicators in patients with anal fistula and T2DM. (A) ROC curve of random intravenous blood glucose. (B) ROC curve of preoperative HbA1c. (C) ROC curve of maximum FBG. (D) ROC curve of maximum PBG. *p < 0.05 with Wilcoxon rank sum test.*
## Surgery-related treatment results
Of these 244 patients, $40.98\%$ ($$n = 100$$) were diagnosed with high anal fistula which patients in the refractory group was slightly higher than in the healing group ($$n = 56$$ vs. $$n = 44$$). Different surgical techniques were carried out by the surgeons depending on the type of fistula. 30 ($12.30\%$) patients underwent cutting seton surgery, 144 ($59.02\%$) patients performed anal fistulectomy and 70 ($28.69\%$) patients underwent both procedure (hybrid surgery). Fisher's exact test did not reveal any statistical difference either overall or between the two different procedures ($p \leq 0.05$). In order to explore the effects of different incisions in wound healing, we summarized the characteristics of the distribution of wounds under lithotomy position. 6 o’clock incision was chosen most ($$n = 123$$, $50.41\%$), followed by the 3 o’clock incision ($$n = 64$$, $26.23\%$), 5 o’clock incision ($$n = 60$$, $24.59\%$), and 1 o’clock incision ($$n = 53$$, $21.27\%$). Both 1 o’clock and 5 o’clock wounds exhibited a higher incidence in the non-healing group and there were statistical differences in both groups ($$p \leq 0.005$$, $$p \leq 0.029$$). Milligan-Morgan surgery were conducted in 70 patients ($28.69\%$) at the same time, while in the healing group this proportion only reached 25 cases ($20.49\%$). The full surgery-related data are shown in Table 3.
**Table 3**
| Variables | Total (n = 244) | Refractory group (n = 122) | Healing group (n = 122) | p value |
| --- | --- | --- | --- | --- |
| Hospital day (day) | 8 (7, 10) | 8 (7, 10) | 8 (7, 9) | 0.104 |
| Incision before surgery, n (%) | | | | 0.344 |
| Yes | 12 (4.92%) | 4 (3.28%) | 8 (6.56%) | |
| No | 232 (95.08%) | 118 (96.72%) | 114 (93.44%) | |
| Anal fistula type, n (%) | | | | 0.104 |
| High anal fistula | 100 (40.98%) | 56 (45.90%) | 44 (36.07%) | |
| Low anal fistula | 144 (59.02%) | 66 (54.10%) | 78 (63.93%) | |
| Surgery method, n (%) a | | | | 0.416 |
| Anal fistulectomy | 144 (59.02%) | 67 (54.92%) | 77 (63.11%) | Anal fistulectomy vs. Cutting seton surgery: 0.324a |
| Cutting seton surgery | 30 (12.30%) | 13 (10.66%) | 17 (13.93%) | Cutting seton surgery vs. Hybrid surgery: 0.501a |
| Hybrid surgery | 70 (28.69%) | 32 (26.23%) | 38 (31.15%) | Anal fistulectomy vs. Hybrid surgery: 0.179a |
| Position of incision, n (%) | | | | |
| 1 o'clock incision | 53 (21.72%) | 36 (29.51%) | 17 (13.93%) | 0.005* |
| 2 o'clock incision | 21 (8.61%) | 11 (9.02%) | 10 (4.92%) | 1.000 |
| 3 o'clock incision | 64 (26.23%) | 32 (26.23%) | 32 (26.23%) | 1.000 |
| 4 o'clock incision | 16 (6.56%) | 8 (6.56%) | 8 (6.56%) | 1.000 |
| 5 o'clock incision | 60 (24.59%) | 38 (31.15%) | 22 (18.03%) | 0.029* |
| 6 o'clock incision | 123 (50.41%) | 60 (49.18%) | 63 (51.64%) | 0.791 |
| 7 o'clock incision | 43 (17.62%) | 19 (15.57%) | 24 (19.67%) | 0.500 |
| 8 o'clock incision | 14 (5.74%) | 5 (4.10%) | 9 (7.38%) | 0.424 |
| 9 o'clock incision | 44 (18.03%) | 25 (20.49%) | 19 (15.57%) | 0.430 |
| 10 o'clock incision | 18 (7.38%) | 7 (5.74%) | 11 (9.02%) | 0.424 |
| 11 o'clock incision | 48 (19.67%) | 28 (22.95%) | 20 (16.39%) | 0.268 |
| 12 o'clock incision | 14 (5.74%) | 7 (5.74%) | 7 (5.74%) | 1.000 |
| Milligan-Morgan surgery, n (%) | | | | 0.006* |
| Yes | 70 (28.69%) | 45 (36.89%) | 25 (20.49%) | |
| No | 174 (71.31%) | 77 (63.11%) | 97 (79.51%) | |
## Pre- and post-operative treatment
Treatment for anal fistula and T2DM must be administrated simultaneously when they are both present. To highlight the relationship between various therapies and wound healing, we gathered several interventions and results presented in Table 4. Among these patients, 92 ($37.70\%$) patients had their first T2DM diagnosis with no statistically significant difference ($$p \leq 0.665$$). Insulin pumps were used for glycemic control in 61 ($25.00\%$) patients during hospitalization, including 41 ($33.61\%$) in the refractory group and 20 ($16.39\%$) in the healing group. Among the patients who used them, the duration did not show a difference (6.49 ± 0.39 d vs. 5.00 ± 0.64 d, $$p \leq 0.053$$). There was no significant difference between the two groups regarding the adoption of subcutaneous insulin, but it did show a higher utilization rate on oral hypoglycemics in the healing group (71 vs. 90, $$p \leq 0.015$$).
**Table 4**
| Variables | Total (n = 244) | Refractory group (n = 122) | Healing group (n = 122) | p value |
| --- | --- | --- | --- | --- |
| Previous diagnosis of T2DM, n (%) | | | | 0.665 |
| Yes | 92 (37.70%) | 57 (46.72%) | 35 (28.69%) | |
| No | 152 (62.30%) | 65 (53.28%) | 87 (71.31%) | |
| Usage of the insulin pump, n (%) | | | | 0.002* |
| Yes | 61 (25.00%) | 41 (33.61%) | 20 (16.39%) | |
| No | 183 (75.00%) | 81 (66.39%) | 102 (83.61%) | |
| Duration of insulin pump use (day) | 6.02 ± 0.34 | 6.49 ± 0.39 | 5.00 ± 0.64 | 0.053 |
| Oral hypoglycemic, n (%) | | | | 0.015 |
| Yes | 161 (65.98%) | 71 (58.20%) | 90 (73.77%) | |
| No | 83 (34.02%) | 51 (41.80%) | 32 (26.23%) | |
| Subcutaneous Insulin Injections, n (%) | | | | 0.519 |
| Yes | 48 (19.67%) | 26 (21.31%) | 22 (18.03%) | |
| No | 196 (80.33%) | 96 (78.69%) | 100 (81.97%) | |
| Preoperative antibiotic therapy, n (%) | | | | <0.001* |
| Yes | 159 (65.16%) | 65 (53.28%) | 94 (77.05%) | |
| No | 85 (34.84%) | 57 (46.72%) | 28 (22.95%) | |
| Amount of antibiotics, n (%) | | | | 0.839 |
| 1 | 219 (89.75%) | 109 (89.34%) | 110 (90.16%) | |
| 2 | 25 (10.25%) | 13 (10.66%) | 12 (9.84%) | |
| Duration of antibiotic use after surgery (day) | 5 (5, 7) | 5 (5, 7) | 5 (5, 7) | 0.140 |
| Usage of polymyxin B, n (%) | | | | 0.134 |
| Yes | 100 (40.98%) | 44 (36.07%) | 56 (45.90%) | |
| No | 144 (59.02%) | 78 (63.93%) | 66 (54.10%) | |
In 244 cases, the rate of preoperative antibiotics reached $65.16\%$, with a higher proportion in the refractory group. The duration of postoperative intravenous antibiotics usage (5d [5, 7] vs. 5d [5, 7], $$p \leq 0.14$$) did not demonstrate a significant difference. $89.75\%$ ($$n = 219$$) of the patients used only one antibiotic and $10.25\%$ ($$n = 25$$) used two or more antibiotics after operation, no significant difference was seen between the groups ($p \leq 0.05$).
## Results of univariate and multivariate logistic regression analysis
After the univariate logistic regression analysis of statistically significant indicators, all outcomes revealed various degrees of associations with wound healing with the exception of LY% (Table 5). Among the laboratory indices, elevated plasma sodium, chloride, A/G ratio, and NEUT% showed varying degrees of protection. Random intravenous blood glucose (OR: 1.137, $95\%$ CI: 1.072–1.206, $p \leq 0.001$), HbA1c (OR: 1.284, $95\%$ CI: 1.104–1.494, $$p \leq 0.001$$), maximal FBG (OR: 1.581, $95\%$ CI: 1.1.313–1.904, $p \leq 0.001$) and excessive levels of maximal PBG (OR:1.182, $95\%$ CI: 1.079–1.296, $p \leq 0.001$) were shown to be impairing factors for wound healing. Intriguingly, the usage of insulin pump had an OR > 1 in the univariate logistic regression analysis, whereas oral hypoglycemic drugs unsurprisingly played a protective role (OR = 0.50). In surgical treatment, 1 and 5 o’clock incision and concurrent M-M surgery both served to retard wound healing.
**Table 5**
| Variables | Univariate OR | 95% CI | p value |
| --- | --- | --- | --- |
| Na | 0.824 | 0.734–0.913 | <0.001* |
| Cl | 0.931 | 0.869–0.998 | 0.043* |
| Uric acid | 1.005 | 1.004–1.006 | <0.001* |
| Random intravenous blood glucose | 1.137 | 1.072–1.206 | <0.001* |
| A/G | 0.412 | 0.197–0.880 | 0.022* |
| NEUT% | 0.974 | 0.952–0.998 | 0.030* |
| LY% | 1.01 | 0.998–1.022 | 0.090 |
| MO% | 1.021 | 1.005–1.038 | 0.011* |
| RDW-SD | 0.869 | 0.799–0.945 | 0.001* |
| Glycosylated hemoglobin | 1.284 | 1.104–1.494 | 0.001* |
| Maximum FBG | 1.581 | 1.313–1.904 | <0.001* |
| Maximum PBG | 1.182 | 1.079–1.296 | <0.001* |
| Urine ketone | 2.5 | 1.280–4.883 | 0.007* |
| Usage of insulin pump | 3.0 | 1.467–6.137 | 0.003* |
| Oral hypoglycemic | 0.5 | 0.288–0.867 | 0.014* |
| Preoperative antibiotic therapy | 0.318 | 0.174–0.518 | <0.001* |
| Placement of drainage tube | 8.0 | 1.001–63.962 | 0.050* |
| 1 o'clock incision | 2.583 | 1.327–5.030 | 0.005* |
| 5 o'clock incision | 2.0 | 1.097–3.645 | 0.024* |
| Milligan-Morgan surgery | 2.429 | 1.303–4.525 | 0.005* |
To prevent overfitting, the research included plasma sodium, chloride, uric acid, random intravenous blood glucose, HbA1c, maximum FBG, maximum PBG, insulin pump usage and oral hypoglycemic drug use, 1 and 5 o’clock incision into the multivariate analysis by referring relevant literature and clinical experience. The results of multivariate logistic regression analysis are presented in Table 6. It revealed that the decrease of NEUT% (OR: 0.906, $95\%$ CI: 0.856–0.958, $$p \leq 0.001$$), elevation of uric acid (OR: 1.008, $95\%$ CI: 1.002–1.015, $$p \leq 0.012$$), maximum FBG (OR:1.489, $95\%$ CI:1.028–2.157, $$p \leq 0.035$$), random intravenous blood glucose (OR: 1.130, $95\%$ CI: 1.008–1.267, $$p \leq 0.037$$) and incision at 5 o’clock (OR: 3.510, $95\%$ CI: 1.214–10.146, $$p \leq 0.020$$) were independent risk factors for refractory wounds.
**Table 6**
| Variables | Multivariate OR | 95% CI | p value |
| --- | --- | --- | --- |
| NEUT% | 0.906 | 0.856–0.958 | 0.001* |
| Uric acid | 1.008 | 1.002–1.015 | 0.012* |
| Random intravenous blood glucose | 1.13 | 1.008–1.267 | 0.037* |
| Maximum FBG | 1.489 | 1.028–2.157 | 0.035* |
| 5 o'clock incision | 3.51 | 1.214–10.146 | 0.020* |
## Discussion
It is widely accepted that surgical treatment is the only way to cure anal fistula. Surgical methods such as ligation of intersphincteric fistula tract (LIFT), video-assisted anal fistula treatment (VAAFT), mucosal advancement flap surgery, and biomaterial occlusion have developed rapidly [5, 6, 9]. Fistulas fistulotomy and cutting seton surgery can fulfill general anal fistula treatment needs and are still the most commonly used procedure in clinical practice (7–9). However, due to the special physiological functions of the anus, the failure rate of anal fistula surgery reached $3.90\%$–$31\%$ [30, 31]. Excessive healing time after anal fistula surgery is an important cause of surgical failure. It has been proved that the choice of surgical approach [32], increase of systemic inflammation [33] are risk factors for refractory wounds. In this research, the fistulas fistulotomy and cutting seton surgery do not show a difference in wound healing, this may indicate that infection played a more important role in wound healing than the trauma area, which fits with Ho KS's report [7]. The key point of anal fistula surgery is the proper drainage of secretions. We found trauma at 5 o’clock is an independent risk factor for healing. Because the trauma located on the posterior side of the anal may constantly face large impact from defecation and leftover stool tended to penetrate from the 6 o'clock incision resulting in inefficient diversion effects.
Most purulent material has been drained by external orifices, infection has been limited in anal fistula patients [33]. It is reasonable that WBC and NEUT% fluctuated within the normal range. Previous reports indicated that persistently elevated levels of inflammation at the trauma surface could delay trauma recovery, especially in diabetic patients [34]. By multifactorial logistic regression, however, we found that acceptable evaluation of NEUT% favors wound healing. Unfortunately, this only represents the preoperative result, inflammation indicators after the surgery was unable to be collected, so the correlation between it and wound recovery can not be elucidated yet.
Successful surgery is the first step in securing wound healing, but the dynamics of the trauma environment will constantly affect wound healing. Mina Sarofim [30] reported that diabetes mellitus was an important risk factor for the recurrence of anal fistula. In this study, we found that elevated maximal FBG and RBG were independent risk factors for impeding wound healing. Escalated blood glucose delays wound healing in many ways. First and foremost, it increases the level of inflammation and prolongs the duration of infection. The persistent infiltration of inflammatory cells in the hyperglycemic state secretes large amounts of pro-inflammatory factors, aggravating the traumatic inflammation [35]. Long glycolysis time increases neutrophil [36] and monocytes accumulation [37] and also slows down macrophage cell transformation, thus increasing secretion of cytotoxic substances. In addition, it's found that patients with refractory trauma exhibited higher levels of serum uric acid. Abnormalities in uric acid metabolism laterally contribute to the rise in inflammation. Previous studies have already suggested that inflammatory macrophage phenotype persistence in T2DM wounds leads to derangement catabolic process of uric acid, the retention of uric acid may lead to its crystallization in wounds, thus increasing the inflammation level [38, 39].
Secondly, chronic hyperglycemia status in T2DM impaired wound redox response, and excessive production of reactive oxygen (ROS) decreases the quality of wound healing. Toll-like receptor expression is upregulated, and hypoxia-inducible factors (HIFs) destabilization cause the imbalance of redox responses [24]. In contrast, an oxygen-rich environment accelerates the survival and migration of keratin-forming cells and fibroblasts, thus promoting trabecular vascular growth [40]. Third, the high glycemic state prevents the growth of blood vessels and the migration of newborn cells. Persistent hyperglycemia impaired epithelial and macrophage function reduced expression of growth factors and weakened pro-angiogenic signaling. The decline in growth factors directly impairs the proliferation, migration and differentiation processes of keratin-forming cells and fibroblasts, damaging body's ability to repair traumas [41]. After ROC curve analysis, maximum FBG shows the largest AUC (0.724), which has the strongest predictive effect for postoperative wound healing. Although HbA1c has the strongest sensitivity at the critical value, it mainly reflects the blood glucose status before surgery. Patients intervened by professional endocrinologists after admission which made the blood glucose fluctuation within the hospital more in line with the long-term situation.
Hyperinflammatory, hypoxic, and disorders of angiogenesis formed an interactive network that feeds back to each other and restrains trauma healing. Well-managed blood sugar promotes high-quality wound healing. Patients with an insulin pump have an increased risk of non-healing trauma, which is not consistent with previously reported results [42]. The main reason is that in clinical practice, insulin pumps were mostly used in patients with high blood glucose levels. From long-term perspective, these patients had worse glucose metabolism status. The critical solution is to optimize glucose metabolism to avoid delaying wound healing.
*In* general, wound healing is a multifactorial repair process. This study reports independent factors affecting wound healing in patients with anal fistula combined with T2DM. Most of the risk factors had a strong connection with inflammation and glycolysis. Uric acid, blood glucose and maximum fasting blood glucose elevation, and incision at 5 o’clock are independent risk factors for impeding wound healing. Elevated NEUT% within normal levels is a protective factor for trauma healing. We recommend that physicians actively monitor these indicators to ensure rapid wound healing. However, it has to be mentioned that there is still lack of RCT trials to support the certainty of its risk factors. Also, as a retrospective study, selection bias is inevitable, although we used PSM to minimize bias as much as possible.
## Data availability statement
To protect patients' privacy, raw study data are available from the authors upon request.
## Ethics statement
The studies involving human participants were reviewed and approved by Ethics Committee of the Hospital of Chengdu University of Traditional Chinese Medicine ethics number: 2022KL-018). Written informed consent for participation was not required for this study in accordance with the national legislation and the institutional requirements.
## Author contributions
XT and TH is the principal investigator of this study and was involved in the critical revision of important intellectual content. JK was involved in the conception and design of this manuscript. XL was involved in the conception and design of the manuscript. YL and YW were involved in the conception and design and statistical advice. GY, JL and YY were involved in the conception and design and coordination development of the trial. LW and LL were involved in the conception and design. All authors critically revised the protocol for important intellectual content and approved the final manuscript.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher's note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## References
1. Sugrue J, Nordenstam J, Abcarian H, Bartholomew A, Schwartz JL, Mellgren A. **Pathogenesis and persistence of cryptoglandular anal fistula: a systematic review**. *Tech Coloproctol* (2017) **21** 425-32. DOI: 10.1007/s10151-017-1645-5
2. Hokkanen SR, Boxall N, Khalid JM, Bennett D, Patel H. **Prevalence of anal fistula in the United Kingdom**. *World J Clin Cases* (2019) **7** 1795-804. DOI: 10.12998/wjcc.v7.i14.1795
3. Amato A, Bottini C, De Nardi P, Giamundo P, Lauretta A, Realis Luc A. **Evaluation and management of perianal abscess and anal fistula: SICCR position statement**. *Tech Coloproctol* (2020) **24** 127-43. DOI: 10.1007/s10151-019-02144-1
4. Sözener U, Gedik E, Kessaf Aslar A, Ergun H, Halil Elhan A, Memikoğlu O. **Does adjuvant antibiotic treatment after drainage of anorectal abscess prevent development of anal fistulas? A randomized, placebo-controlled, double-blind, multicenter study**. *Dis Colon Rectum* (2011) **54** 923-9. DOI: 10.1097/DCR.0b013e31821cc1f9
5. Liu D, Li W, Wang X, Qiu J, Wang L, Xiong F. **Classification of anal fistulas based on magnetic resonance imaging**. *Zhonghua Wei Chang Wai Ke Za Zhi* (2018) **21** 1391-5. PMID: 30588591
6. Kołodziejczak M, Santoro GA, Obcowska A, Lorenc Z, Mańczak M, Sudoł-Szopińska I. **Three-dimensional endoanal ultrasound is accurate and reproducible in determining type and height of anal fistulas**. *Colorectal Dis* (2017) **19** 378-84. DOI: 10.1111/codi.13580
7. Ho KS, Tsang C, Seow-Choen F, Ho YH, Tang CL, Heah SM. **Prospective randomised trial comparing ayurvedic cutting seton and fistulotomy for low fistula-in-ano**. *Tech Coloproctol* (2001) **5** 137-41. DOI: 10.1007/s101510100015
8. Emile SH, Elfeki H, El-Said M, Khafagy W, Shalaby M. **Modification of parks classification of cryptoglandular anal fistula**. *Dis Colon Rectum* (2021) **64** 446-58. DOI: 10.1097/dcr.0000000000001797
9. Pescatori M. **Surgery for anal fistulae: state of the art**. *Int J Colorectal Dis* (2021) **36** 2071-9. DOI: 10.1007/s00384-021-03917-7
10. La Torre M, Lisi G, D'Agostino E, Boccuzzi M, Campanelli M, Varriale M. **Lift and vaaft for high trans-sphincteric anal fistula: a single center retrospective analysis**. *Int J Colorectal Dis* (2020) **35** 1149-53. DOI: 10.1007/s00384-020-03584-0
11. Zhao B, Wang Z, Han J, Zheng Y, Cui J, Yu S. **Long-term outcomes of ligation of the inter-sphincteric fistula tract plus bioprosthetic anal fistula plug (lift-plug) in the treatment of trans-sphincteric perianal fistula**. *Med Sci Monit* (2019) **25** 1350-4. DOI: 10.12659/msm.914925
12. Al Sebai OI, Ammar MS, Mohamed SH, El Balshy MA. **Comparative study between intersphinecteric ligation of perianal fistula versus conventional fistulotomy with or without seton in the treatment of perianal fistula: a prospective randomized controlled trial**. *Ann Med Surg* (2021) **61** 180-4. DOI: 10.1016/j.amsu.2020.12.014
13. Cho NH, Shaw JE, Karuranga S, Huang Y, da Rocha Fernandes JD, Ohlrogge AW. **IDF diabetes atlas: global estimates of diabetes prevalence for 2017 and projections for 2045**. *Diabetes Res Clin Pract* (2018) **138** 271-81. DOI: 10.1016/j.diabres.2018.02.023
14. Shaw JE, Sicree RA, Zimmet PZ. **Global estimates of the prevalence of diabetes for 2010 and 2030**. *Diabetes Res Clin Pract* (2010) **87** 4-14. DOI: 10.1016/j.diabres.2009.10.007
15. Zheng Y, Ley SH, Hu FB. **Global aetiology and epidemiology of type 2 diabetes mellitus and its complications**. *Nat Rev Endocrinol* (2018) **14** 88-98. DOI: 10.1038/nrendo.2017.151
16. Saeedi P, Petersohn I, Salpea P, Malanda B, Karuranga S, Unwin N. **Global and regional diabetes prevalence estimates for 2019 and projections for 2030 and 2045: results from the international diabetes federation diabetes atlas, 9(th) edition**. *Diabetes Res Clin Pract* (2019) **157** 107843. DOI: 10.1016/j.diabres.2019.107843
17. Roy R, Zayas J, Singh SK, Delgado K, Wood SJ, Mohamed MF. **Overriding impaired FPR chemotaxis signaling in diabetic neutrophil stimulates infection control in murine diabetic wound**. *Elife* (2022) **11** e72071. DOI: 10.7554/eLife.72071
18. Feng J, Dong C, Long Y, Mai L, Ren M, Li L. **Elevated Kallikrein-binding protein in diabetes impairs wound healing through inducing macrophage M1 polarization**. *Cell Commun Signal* (2019) **17** 60. DOI: 10.1186/s12964-019-0376-9
19. Catrina SB, Zheng X. **Hypoxia and hypoxia-inducible factors in diabetes and its complications**. *Diabetologia* (2021) **64** 709-16. DOI: 10.1007/s00125-021-05380-z
20. Das S, Majid M, Baker AB. **Syndecan-4 enhances PDGF-BB activity in diabetic wound healing**. *Acta Biomater* (2016) **42** 56-65. DOI: 10.1016/j.actbio.2016.07.001
21. Ko KI, Sculean A, Graves DT. **Diabetic wound healing in soft and hard oral tissues**. *Transl Res* (2021) **236** 72-86. DOI: 10.1016/j.trsl.2021.05.001
22. Gortler H, Rusyn J, Godbout C, Chahal J, Schemitsch EH, Nauth A. **Diabetes and healing outcomes in lower extremity fractures: a systematic review**. *Injury* (2018) **49** 177-83. DOI: 10.1016/j.injury.2017.11.006
23. Ommer A, Herold A, Berg E, Fürst A, Post S, Ruppert R. **German S3 guidelines: anal abscess and fistula (second revised version)**. *Langenbecks Arch Surg* (2017) **402** 191-201. DOI: 10.1007/s00423-017-1563-z
24. **Chinese Guidelines for the prevention and treatment of type 2 diabetes (2017 edition)**. *Chin J Pract Intern Med* (2018) **38** 292-344. DOI: 10.19538/j.nk2018040108
25. **Chinese Guidelines for the prevention and treatment of type 2 diabetes (2020 edition) (top)**. *Chin J Pract Intern Med* (2021) **41** 668-95. DOI: 10.19538/j.nk2021080106
26. Lazarus GS, Cooper DM, Knighton DR, Percoraro RE, Rodeheaver G, Robson MC. **Definitions and guidelines for assessment of wounds and evaluation of healing**. *Wound Repair Regen* (1994) **2** 165-70. DOI: 10.1046/j.1524-475X.1994.20305.x
27. Shafik AA, El Sibai O, Shafik IA. **Combined partial fistulectomy and electro-cauterization of the intersphincteric tract as a sphincter-sparing treatment of complex anal fistula: clinical and functional outcome**. *Tech Coloproctol* (2014) **18** 1105-11. DOI: 10.1007/s10151-014-1208-y
28. Madbouly KM, El Shazly W, Abbas KS, Hussein AM. **Ligation of intersphincteric fistula tract versus mucosal advancement flap in patients with high transsphincteric fistula-in-ano: a prospective randomized trial**. *Dis Colon Rectum* (2014) **57** 1202-8. DOI: 10.1097/dcr.0000000000000194
29. Mushaya C, Bartlett L, Schulze B, Ho YH. **Ligation of intersphincteric fistula tract compared with advancement flap for complex anorectal fistulas requiring initial seton drainage**. *Am J Surg* (2012) **204** 283-9. DOI: 10.1016/j.amjsurg.2011.10.025
30. Sarofim M, Ooi K. **Reviewing perianal abscess management and recurrence: lessons from a trainee perspective**. *ANZ J Surg* (2022) **92** 1781-3. DOI: 10.1111/ans.17750
31. Anaraki F, Nikshoar MR, Ketabforoush A, Chegini R, Borumandnia N, Tavallaei M. **Fistulectomy and primary sphincteroplasty in complex anal fistula treatment: a hospital-based long-term follow-up study**. *Tech Coloproctol* (2023) **27** 145-52. DOI: 10.1007/s10151-022-02722-w
32. Göttgens KW, Smeets RR, Stassen LP, Beets G, Breukink SO. **Systematic review and meta-analysis of surgical interventions for high cryptoglandular perianal fistula**. *Int J Colorectal Dis* (2015) **30** 583-93. DOI: 10.1007/s00384-014-2091-8
33. Wang JP, Cai C, Du JL, Shi HQ, Liu QW, Dai ZH. **Role of interleukin-17 in the pathogenesis of perianal abscess and anal fistula: a clinical study on 50 patients with perianal abscess**. *ANZ J Surg* (2019) **89** 244-7. DOI: 10.1111/ans.14874
34. Mocanu V, Dang JT, Ladak F, Tian C, Wang H, Birch DW. **Antibiotic use in prevention of anal fistulas following incision and drainage of anorectal abscesses: a systematic review and meta-analysis**. *Am J Surg* (2019) **217** 910-7. DOI: 10.1016/j.amjsurg.2019.01.015
35. Louiselle AE, Niemiec SM, Zgheib C, Liechty KW. **Macrophage polarization and diabetic wound healing**. *Transl Res* (2021) **236** 109-16. DOI: 10.1016/j.trsl.2021.05.006
36. **Consensus of Chinese experts on the diagnosis and treatment of anal fistula (2020)**. *Zhonghua Wei Chang Wai Ke Za Zhi* (2020) **23** 1123-30. DOI: 10.3760/cma.j.cn.441530-20200925-00537
37. Adamo K, Gunnarsson U, Eeg-Olofsson K, Strigård K, Brännström F. **Risk for developing perianal abscess in type 1 and type 2 diabetes and the impact of poor glycemic control**. *Int J Colorectal Dis* (2021) **36** 999-1005. DOI: 10.1007/s00384-020-03818-1
38. Fernandez ML, Upton Z, Shooter GK. **Uric acid and xanthine oxidoreductase in wound healing**. *Curr Rheumatol Rep* (2014) **16** 396. DOI: 10.1007/s11926-013-0396-1
39. Kimball AS, Davis FM, denDekker A, Joshi AD, Schaller MA, Bermick J. **The histone methyltransferase Setdb2 modulates macrophage phenotype and uric acid production in diabetic wound repair**. *Immunity* (2019) **51** 258-71.e5. DOI: 10.1016/j.immuni.2019.06.015
40. Asano K, Ueki S, Tamari M, Imoto Y, Fujieda S, Taniguchi M. **Adult-onset eosinophilic airway diseases**. *Allergy* (2020) **75** 3087-99. DOI: 10.1111/all.14620
41. Pastar I, Stojadinovic O, Yin NC, Ramirez H, Nusbaum AG, Sawaya A. **Epithelialization in wound healing: a comprehensive review**. *Adv Wound Care* (2014) **3** 445-64. DOI: 10.1089/wound.2013.0473
42. Fan HJ, Yu JH, Cui GM, Zhang WY, Yang X, Dong QJ. **Insulin pump for the treatment of diabetes in combination with ulcerative foot infections**. *J Biol Regul Homeost Agents* (2016) **30** 465-70. PMID: 27358133
|
---
title: 'A randomized controlled trial to test the effectiveness of two technology-enhanced
diabetes prevention programs in primary care: The DiaBEAT-it study'
authors:
- Fabio A. Almeida
- Wen You
- Fabiana A. Brito
- Thais F. Alves
- Cody Goessl
- Sarah S. Wall
- Richard W. Seidel
- Brenda M. Davy
- Mark H. Greenawald
- Jennie L. Hill
- Paul A. Estabrooks
journal: Frontiers in Public Health
year: 2023
pmcid: PMC9998510
doi: 10.3389/fpubh.2023.1000162
license: CC BY 4.0
---
# A randomized controlled trial to test the effectiveness of two technology-enhanced diabetes prevention programs in primary care: The DiaBEAT-it study
## Abstract
### Objective
To evaluate the effectiveness of two technology-enhanced interventions for diabetes prevention among adults at risk for developing diabetes in a primary care setting.
### Methods
The DiaBEAT-it study employed a hybrid 2-group preference (Choice) and 3-group randomized controlled (RCT) design. This paper presents weight related primary outcomes of the RCT arm. Patients from Southwest Virginia were identified through the Carilion Clinic electronic health records. Eligible participants (18 and older, BMI ≥ 25, no Type 2 Diabetes) were randomized to either Choice ($$n = 264$$) or RCT ($$n = 334$$). RCT individuals were further randomized to one of three groups: [1] a 2-h small group class to help patients develop a personal action plan to prevent diabetes (SC, $$n = 117$$); [2] a 2-h small group class plus automated telephone calls using an interactive voice response system (IVR) to help participants initiate weight loss through a healthful diet and regular physical activity (Class/IVR, $$n = 110$$); or [3] a DVD with same content as the class plus the same IVR calls over a period of 12 months (DVD/IVR, $$n = 107$$).
### Results
Of the 334 participants that were randomized, 232 ($69\%$) had study measured weights at 6 months, 221 ($66\%$) at 12 months, and 208 ($62\%$) at 18 months. Class/IVR participants were less likely to complete weight measures than SC or DVD/IVR. Intention to treat analyses, controlling for gender, race, age and baseline BMI, showed that DVD/IVR and Class/IVR led to reductions in BMI at 6 (DVD/IVR −0.94, $p \leq 0.001$; Class/IVR −0.70, $p \leq 0.01$), 12 (DVD/IVR −0.88, $p \leq 0.001$; Class/IVR-0.82, $p \leq 0.001$) and 18 (DVD/IVR −0.78, $p \leq 0.001$; Class/IVR −0.58, $p \leq 0.01$) months. All three groups showed a significant number of participants losing at least $5\%$ of their body weight at 12 months (DVD/IVR $26.87\%$; Class/IVR $21.62\%$; SC $16.85\%$). When comparing groups, DVD/IVR were significantly more likely to decrease BMI at 6 months ($p \leq 0.05$) and maintain the reduction at 18 months ($p \leq 0.05$) when compared to SC. There were no differences between the other groups.
### Conclusions
The DiaBEAT-it interventions show promise in responding to the need for scalable, effective methods to manage obesity and prevent diabetes in primary care settings that do not over burden primary care clinics and providers.
### Registration
https://clinicaltrials.gov/ct2/show/NCT02162901, identifier: NCT02162901.
## 1. Introduction
The Centers for Disease Control and Prevention (CDC) estimates that there are 34.2 million ($10.5\%$) Americans with diabetes, in addition to the 88 million ($34.5\%$ of the population) with prediabetes in the United States, and strongly recommends healthcare approaches to prevent diabetes [1]. Approximately 5–$10\%$ of individuals with prediabetes develop type 2 diabetes (T2D) yearly with an American Diabetes Association (ADA) [2] expert panel estimating that up to $70\%$ of individuals with prediabetes will eventually progress to diabetes, further highlighting the importance of intervening [3]. Finally, due to the continued growth of the obesity epidemic, the burden of prediabetes and diabetes are expected to continue to rise [4]. As there is no known treatment available to cure diabetes and self-management for those with diabetes remains a challenge, the importance of prevention is paramount [5].
The Diabetes Prevention Program (DPP) was seminal in demonstrating that a modest weight loss achieved through diet and exercise was superior to medication in delaying the onset of T2D [6]. The DPP program found that 30 min of physical activity per day 5 times a week coupled with a 5–$10\%$ weight loss resulted in a $58\%$ reduction in the incidence of diabetes [6]. Following on the success of the DPP, researchers have sought to determine the effectiveness of the DPP in more typical community and clinical settings (7–9). However, barriers to large-scale implementation of these adaptations still exist, where information on primary prevention and management of T2D is still limited [10]. Recent studies [11, 12] have shown the lack of availability of these programs in underserved areas, with lifestyle coaches reporting lack of space, administrative support, sufficient allocation of their own time for the program, overall costs, and difficulty scheduling as barriers to broad dissemination of these programs [13]. On the other hand, participants have reported distance, work schedules, lack of transportation and childcare needs as remaining issues that prevent them from fully engaging in these in-person group adaptations of the DPP (14–16).
To address these barriers, several interventions have used technology to successfully adapt and deliver the DPP. Several systematic reviews have shown that technology-based resources can optimize diabetes prevention intervention to achieve clinically significant weight loss [17]. A review by Levine et al. [ 18] found that technology-assisted weight loss interventions that included some form of human coaching were successful in helping individuals lose weight in primary care settings. Another review [19] of technology-mediated diabetes prevention interventions found that these types of programs can result in a clinically significant amount of weight loss in patients with prediabetes. This review included studies that used a variety of technologies including DVDs, e-videos, web-based resources, videoconferencing, telephone calls, interactive voice response, text messages, e-counseling, email, and online group forums with a variety of Supplementary material (e.g., Physical Activity and Nutrition workbooks, log books, and in person group DPP). Joiner et al. [ 20] found similar results further supporting the effectiveness of technology-based interventions in helping individuals at risk for developing diabetes to lose clinically significant amounts of weight. However, questions remain regarding the effectiveness of eHealth interventions within primary care settings and in promoting weight loss maintenance or weight gain prevention (18, 21–23).
The original diaBEAT-it study [24] was a pragmatic clinical trial employing a hybrid preference/randomized control trial (RCT). The study focused primarily on the individual-level factors of reach, effectiveness, maintenance, and cost [24] but each active intervention was designed for broad dissemination and scalability within and across healthcare and, potentially, public health systems. The overarching goal of the study was to determine the effectiveness and maintenance of effects of two technology-enhanced interventions relative to standard care (SC) in reducing body weight within the context of a traditional RCT, while concurrently determining the relative reach of these two interventions within the context of a two-group preference design where participants had the option to choose which intervention they would like to participate in (Choice). The diaBEAT-it study has been fully described elsewhere [24]. The purpose of the present paper is to evaluate the effectiveness of the two technology-enhanced interventions in supporting patients to reduce their body mass index (BMI) over an 18-month period relative to a minimal standard care intervention. We hypothesized that compared to minimal standard care, each intervention would result in greater mean reduction in BMI over 18 months.
## 2. Design and methods
Patients at risk for developing diabetes were randomly assigned [2-1] by the project manager to either an RCT or Choice study arm using a blocked (groups of 4) randomization table stratified by sex created by the study statistician. Patients in the RCT arm were further randomized [1-1-1] to one of three groups: [1] a 2-h small group class designed to help patients develop a personal action plan to prevent diabetes (SC) [25]; [2] a 2-h small group class plus automated telephone calls using an interactive voice response system (IVR) to help participants initiate weight loss via the promotion of a healthful diet and regular physical activity, and maintain their behavior changes over a period of 12 months (Class/IVR); or [3] a DVD with same content as the class plus the same IVR calls over a period of 12 months (DVD/IVR).
This paper presents weight related outcomes associated with the randomized control trial arm of the DiaBEAT-it study [24]. We powered our study to detect statistically significant body weight changes at 6 and 12 months in Class/IVR and DVD/IVR when compared with the SC group within the RCT design. Sample size was determined by using the average weight loss and standard deviations found in our previous studies (25–27) for the 6-month effect and averages from the literature [28, 29] for the 12-month effect. As such, assuming a correlation of 0.5 between repeated measures, we estimated that a sample size of 78 participants per group would give us a $90\%$ power to detect a minimum detectable difference in change in weight of 2.3 lbs. at 6 months and 2.7 lbs. at 12 months. The goal for enrollment was 120 participants per group to achieve a sample size of 78 after an estimated $35\%$ attrition at 18 months. The trial design and methods have been described in detail elsewhere [24]. Supplementary Figure 1 provides the CONSORT information for the RCT study arm. This study and protocol were approved by the Carilion Clinic Institutional Review Board and was registered at clinicaltrials.gov (NCT02162901).
## 2.1. Participant eligibility and recruitment
Potential participants were initially identified through a Carilion Clinic electronic health records (EHR) query of primary care patients between January 2014 and August 2015 [24]. EHR eligibility included patients that over the previous 12 months were 18 years of age and older, BMI ≥ 25, had ICD-9 codes for prediabetes, glucose intolerance, metabolic syndrome, and/or obesity while excluding those with ICD-9 codes indicating diagnosed diabetes, congestive heart failure, and coronary artery disease. A list of patients meeting initial eligibility criteria were sent to their physicians for final approval. All approved patients were recruited via a physician letter providing general information about the study and went through telephone screening for final eligibility determination. During the phone screening, a research assistant reiterated key points of the letter, answered any questions, and determined diabetes risk and study eligibility. Diabetes risk was determined using the Diabetes Risk Calculator (DRC) [30]. Individuals with a score of 5 or higher are considered to be in particularly greater risk and were the target for recruitment.
Individuals were eligible if they were 18 years of age or older with a BMI of at least 25 kg/m2 (BMI > 22 for Asian), spoke English, were not pregnant or planning to become pregnant in the following 18 months, were not diagnosed with T2D, congestive heart failure, or coronary artery disease, had no contraindication for physical activity (PA) or weight loss, had access to a phone, and had a DRC test score indicative of high risk for developing T2D (Score of 5 or higher). Eligibility was broadly defined to allow for most typical primary care patients to be eligible to participate in the study.
Prior to the baseline visit, the project manager created sealed opaque envelopes with group assignment information according to the blocked (groups of 4) randomization table stratified by gender created by the study statistician to blind research staff to intervention assignment. Informed consent procedures were initiated during the screening telephone call with participants receiving the informed consent via mail prior to their initial visit so they could prepare for the first study visit. These procedures were completed at the time of the first study visit with participants [1] receiving information on the risks and benefits of participating on the trial, [2] being given the opportunity to ask questions, clarifications, and raise any concerns, and [3] being informed of the interventions of interest and their rights as a research subject. All assessments took place after full consent was given by study participants. Once all assessments were completed, a research assistant randomized participants in the RCT study arm to one of the three study groups using the previously created envelopes. Participants randomized to SC received information about the class and a workbook. Those randomized to the Class/IVR group received a workbook and were assisted by a research assistant in signing into an IVR account to select days and times best for their calls and setup a security PIN. Research assistants also helped participants in completing an initial test of the system to familiarize themselves with the IVR calls. Participants randomized to the DVD/IVR group received a workbook, a DVD and a brief instruction on how to use the TV to navigate the DVD in addition to the IVR system setup. Finally, all randomized participants received $25.00 as a thank you for their time in completing the baseline assessments.
## 2.2.1. Standard care
Participants in the SC comparison group took part in a 2-h small group session class [25] taught by two trained Carilion Clinic employees (Certified Diabetes Educators and Registered Dietitians). This class has been offered for the past 6 years and although they are available to all Carilion Clinic patients, for the purpose of the project, separate classes to each intervention group were offered. As such, both groups attended project specific classes for their given study group (SC or Class/IVR). The content, format, and individuals teaching these classes did not differ from the currently taught classes. Participants in the SC group received no additional intervention contact after the initial class. They were contacted 6, 12, and 18 months following their class date for follow-up assessments. During the class participants were encouraged to develop their own personal action plan to preventing T2D by setting a goal of losing $10\%$ of their current weight over 12 months and to be physically active for 60 min, 5 days per week. The personal action plan also included a listing of motivational reasons to prevent diabetes, personal goals for weight management, physical activity, and healthful eating, identifying barriers, strategies to overcome barriers, and upholding accountability for these goals through a commitment to enlist friends and/or family members in the change process [25]. Class instructors provided detailed information on current recommendations for physical activity and healthy eating (MyPlate guidelines) and gave a workbook covering all 22 session topics following a similar curriculum as developed by the original DPP. The class is fully described elsewhere [24].
## 2.2.2. Class/IVR group
Participants in this group attended the 2-h class described above, received a workbook, completed a “Live” counseling call [31], and received 22 tailored IVR calls over a period of 12 months with the final 6 months focusing on maintenance and relapse prevention based on DPP's after Core program. This intervention was designed to help participants initiate moderate weight loss through physical activity and healthful eating and maintain these behavior changes. All participants developed a personal action plan with the goal of losing $10\%$ of their current weight in 12 months and being physically active for 60 min a day, 5 days per week. Workbook content topics focused on achieving a balanced diet through the reduction of fat and caloric intake plus adding regular physical activity to enhance initial weight loss and prevent weight regain. Additionally, we used the 5 A's model to assist participants in setting physical activity and healthful eating goals necessary for weight loss and maintenance [32]. One week after class completion, participants received a telephone call lasting 45–60 min to reinforce learning objectives and provide further clarifications [31]. Research assistants delivered this call using teach-to-goal and teach-back strategies to allow participants to describe key intervention concepts (i.e., MyPlate guidelines, types, and length of physical activity) using their own words and provide additional rounds of education until the participant demonstrated a firm understanding of the information. For those participants that did not attend the initial 2-h class, the research assistants provided the full content of the class and assisted them in creating their personal action plan. One week after the live telephone call the participants began receiving IVR support calls. There were 22 IVR calls lasting between 15 and 30 min with 8 weekly calls, followed by 8 biweekly calls and 6 monthly calls focusing on maintenance and relapse prevention. Participants were required to complete one call before moving on to the next call, as such, it was not possible to skip IVR calls and content. For those participants that did not complete an IVR call, reminder contacts using telephone, text, and email were used for up to 2 weeks to try and get participants back on track. Each IVR call included an assessment of current weight, PA, and dietary behaviors, feedback on goal progression, content related to the session topic (i.e., Move Those Muscles, Being Active: A Way of Life, Healthy Eating With MyPlate, Be A Fat Detective), teach to goal reinforcement of key messages, and a homework assignment. New action plans were created every month (Calls 4, 8, 12, and 16) through Call 16 and then on every call during the maintenance and relapse prevention phase. This included updating goals, identifying new barriers, selecting strategies to resolve barriers, and goal setting-feedback loops.
## 2.2.3. DVD/IVR group
This group was identical to the Class/IVR group but was initiated with a DVD that replicated the class content. The DVD included the following segments: [1] *What is* pre-diabetes? [ 2] What are the risk factors for diabetes? [ 3] Developing your DiaBEAT-it action plan, [4] Goal setting for physical activity and healthy eating, [5] putting together a toolbox of resources, and [6] making a commitment to change. The DVD was about 60 min in duration with several planned pauses to allow for completion of activities. This replicated the 5 A's approach that guided the class and included the completion of an action plan page in the accompanying workbook. Finally, the DVD included an appendix with additional free-of-charge, online nutrition and physical activity informational videos. Participants received their live counseling call within 7 days of being given the DVD. Similar to the Class/IVR group, those participants that reported not watching the DVD the research assistants provided full information and guided them through developing their personal action plan during the “Live” call. The IVR structure and content was the same as described above.
## 2.3. Outcome measures
Trained research assistants unaware of group assignment collected data at baseline, 6, 12, and 18 months. The primary outcome was change in BMI from baseline to 18 months. Secondary outcomes included percentage of participants achieving weight loss goals of $5\%$ or more, changes in percent weight reduction as well as maintenance of those changes at 12 and 18 months. Height was assessed in stocking feet with a calibrated stadiometer with a fixed vertical backboard and adjustable headboard. Weight was assessed with the calibrated Health-O-Meter 2101KL digital stand-on scale (www.homscales.com). Body Mass Index was calculated in kg/m2. Demographic characteristics were collected using a computer-based questionnaire (https://surveymethods.com/). Research assistants were available on site to answer any questions and help participants with potential computer/survey issues. All assessments took place at a research facility.
## 2.4. Statistical analysis
Statistical analysis included descriptive statistics for age, sex, race, ethnicity, education, income, health literacy, employment, health insurance, Diabetes Risk Calculator (DRC), and weight status. Chi-square and independent t-tests were conducted to determine if any of the groups differed on baseline characteristics (Supplementary Table 1). Data were examined for the presence of outliers, violations of normality (for those continuous variables) and missing data. No violations of normality were detected. Between group differences in changes in BMI and other weight outcomes were prespecified using intention-to-treat (ITT) analysis. To simultaneously account for individual effects regardless of the condition, we employed a linear mixed effect model to a multi-treatment framework [33] for the treatment effect analysis [34]. To be specific, two group dummies are in the model along with assessment time dummies and their interactions. This model allows us to control error non-independence of over time assessment within the same individual and heteroskedasticity caused by between individual heterogeneity, and a-priori-determined covariates that are influencing factors of outcome-specific production. The goal was to make more robust inferences about the treatment effect of main outcomes of interest: for example, the effect of Class/IVR and DVD/IVR in reducing BMI over 18 months when compared to SC group. For those participants with missing outcome measurements, we replaced the missing data with their baseline value following the Baseline Carried Forward approach.
Additionally, we conducted analysis based on participants completing at least 4 sessions (i.e., meeting NDPP threshold for recognition standards) [35], at least 6 months (i.e., core intervention effects), and the full 12 months (i.e., post-core effects). For the purposes of these analyses, class and “Live-Call” completion were calculated based on attendance, DVD was based on participant self-report, and IVR call completion was based on the voice files for the lesson of the week being played [24]. Further, for the dichotomous outcome measures (i.e., achieve $5\%$ weight loss goal), we treated those models as linear probability models in order to retain the straight-forward treatment effect interpretation of the results by applying generalized linear models in the analysis. Means and standard deviations for all primary and secondary outcomes at baseline, 6, 12, and 18 months are also presented. All statistical analyses were conducted in Stata v16 and the $5\%$ significance level was used.
## 3.1. Participant enrollment and characteristics
Supplementary Figure 1 presents participant enrollment and retention at 6, 12, and 18 months. A total of 3,115 were identified as potentially eligible to join the study. Of those, 1,712 ($55\%$) were reached by phone with 689 completing screening questions and 427 scheduling an initial study visit to determine full eligibility. A total of 358 patients were eligible to participate in the study with 334 ($93\%$) completing full baseline assessments and being randomized (SC = 117, Class/IVR = 110, DVD/IVR = 107). The mean age of participants was 52.3 (±12.1) years with a mean BMI of 37.2 (±7.3) kg/m2 (Supplementary Table 1). At baseline, $68.1\%$ of participants were female, $76.8\%$ were non-Hispanic white, $20.0\%$ were Non-Hispanic black, $25.8\%$ had high school or lower education, and $55.8\%$ were employed full time (Supplementary Table 1). Intervention groups (Class/IVR and DVD/IVR) participants were less likely to be retired ($$P \leq 0.036$$), had higher average diabetes risk scores ($$P \leq 0.019$$) and higher average BMI ($$P \leq 0.004$$) when compared with SC participants (Supplementary Table 1). Of the 334 participants that were randomized, 232 ($69\%$) had study measured weights at 6 months, 221 ($66\%$) at 12 months, and 208 ($62\%$) at 18 months. Class/IVR participants were less likely to complete weight measures than SC or DVD/IVR (Supplementary Figure 1).
## 3.2. Weight loss
Supplementary Table 2 reports estimated mean changes in BMI and weight over an 18-month period. A total of eight participants (SC = 1, Class/IVR = 6, DVD/IVR = 1) were eliminated from full analysis due to becoming pregnant during trial (Supplementary Figure 1). ITT results show that at month 6, the mean ± SE change in BMI from baseline in DVD/IVR was significant with −0.94 ± 0.21 ($$P \leq 0.022$$ vs. SC; $$P \leq 0.450$$ vs. Class/IVR), significant in Class/IVR with −0.70 ± 0.24 ($$P \leq 0.206$$ vs. SC), and non-significant in SC with −0.33 ± 0.17. At month 12, the mean ± SE change in BMI from baseline in DVD/IVR was significant with −0.88 ± 0.20 ($$P \leq 0.058$$ vs. SC; $$P \leq 0.853$$ vs. Class/IVR), significant in Class/IVR with −0.82 ± 0.25 ($$P \leq 0.141$$ vs. SC), and non-significant in SC with −0.36 ± 0.19. At month 18, the mean ± SE change in BMI from baseline in DVD/IVR was significant with −0.78 ± 0.22 ($$P \leq 0.030$$ vs. SC; $$P \leq 0.550$$ vs. Class/IVR), significant in Class/IVR with −0.58 ± 0.23 ($$P \leq 0.160$$ vs. SC), and non-significant in SC with −0.18 ± 0.17.
At month 6, mean percent weight loss ± SE change from baseline was significant in all three conditions (DVD/IVR: −2.77 ± 0.48; Class/IVR: −1.42 ± 0.46; SC: −1.40 ± 0.42) with DVD/IVR significantly losing more weight than Class/IVR ($$P \leq 0.046$$) and SC ($$P \leq 0.031$$) (Supplementary Table 2). At month 12, the mean percent weight loss ± SE change remained significant in all three conditions (DVD/IVR: −2.56 ± 0.50; Class/IVR: −1.80 ± 0.50; SC: −1.47 ± 0.44) with no between group differences (Supplementary Table 2). At month 18, the mean percent weight loss ± SE change remained significant in all three conditions (DVD/IVR: −2.18 ± 0.54; Class/IVR: −1.27 ± 0.48; SC: −1.11 ± 0.44) with no between group differences (Supplementary Table 2). Finally, results show positive time effects for DVD/IVR (6M: $25.84\%$, 12M: $26.87\%$, 18M: $20.69\%$), Class/IVR (6M: $18.59\%$, 12M: $21.62\%$, 18M: $18.59\%$), and SC (6M: $15.94\%$, 12M: $16.85\%$, 18M: $16.85\%$) participants achieving $5\%$ weight loss across all three timepoints with no treatment effect found across groups (Supplementary Table 2).
## 3.3. Intervention participation rates: CDC recognition standards
On average participants in the DVD/IVR group completed 15.5 (±8.6) sessions compared to 14.1 (±8.3) for Class/IVR. Approximately, $86.3\%$ of participants in the intervention groups (DVD/IVR: $86.6\%$; Class/IVR: $86\%$) met the CDC threshold of completing at least 4 sessions with $48.4\%$ (DVD/IVR: $52.6\%$; Class/IVR: $44\%$) staying in the program for at least 6 months, and $29.5\%$ (DVD/IVR: $37.1\%$; Class/IVR: $21.5\%$) completing every session during the 12-month program. Average percent weight loss at 12 months for those meeting the CDC threshold were $3.24\%$ (DVD/IVR) and $2.74\%$ (Class/IVR) with $35.74\%$ (DVD/IVR) and $33.44\%$ (Class/IVR) achieving a $5\%$ weight loss.
## 3.3.1. Adverse events
During the trial, 40 adverse events (AE) were reported; 6 were classified as serious adverse events (SAE). The majority were associated with immune system disorders (allergic reactions–21). Additional categories included cardiac disorders [1], musculoskeletal disorders [3], general disorders [1], infections [1], injury or procedure complications [3], neoplasms benign, malignant and unspecified [1], nervous systems disorders [1], respiratory disorders [1], and vascular disorders [2]. Twenty-one AEs were determined to be related to the study and 3 had insufficient information to make a determination. The 21 related AEs were all associated with a skin irritation as result of the application of the accelerometer used in the study. One SAE also associated with the application of the accelerometer led to a severe reaction and hospitalization. Overall events were equally balanced between groups, with 13 in SC, 11 in Class/IVR, and 14 in DVD/IVR.
## 4. Discussion and conclusion
The randomized control trial arm of our study demonstrated that two technology-enhanced diabetes prevention programs both led to modest reductions in body weight over an 18-month period. Most importantly, the DVD/IVR group showed significant reductions in BMI when compared to the SC group confirming our original hypothesis. However, there were no significant differences between Class/IVR and SC groups. Participants in the DVD/IVR group lost a mean 2.79 kg over 12 months with $26.9\%$ of participants losing $5\%$ or more of initial body weight in ITT analyses. These numbers improve for both technology-enhanced groups as the number of sessions attended increased.
Our results support the findings of several recent reviews on technology mediated DPPs [19, 20], eHealth obesity interventions [23], weight loss interventions in primary care [18, 36], and self-help weight loss interventions [22]. Joiner et al. [ 20] found an estimated mean percent weight loss from baseline to 15 months to be −$3.98\%$ across the 22 studies included in the review. This magnitude of effect varied from −$3.32\%$ for stand-alone eHealth interventions to −$4.49\%$ for interventions with behavioral support given by a counselor remotely to −$4.65\%$ for interventions with behavioral support given by a counselor in-person.
When investigating the effects of eHealth obesity interventions, Hutchesson et al. [ 23] found that eHealth interventions demonstrated significantly greater weight loss (kg) than control groups (−2.70), or minimal intervention comparisons (−1.40). This review of 84 studies also showed significant weight loss for web-based interventions that incorporated non-eHealth components (−3.70); mobile interventions (−2.40) and web-based interventions delivered only using eHealth technologies (−2.21). Levine et al. [ 18] found 12 interventions that achieved weight loss (range: 0.08 kg −5.4 kg) compared to controls, 5–$45\%$ of patients losing at least $5\%$ of baseline weight with trial duration and attrition ranging from 3 to 36 months and 6–$80\%$, respectively. On another review [36] of 15 RCTs focusing on weight loss in primary care settings, the authors showed pooled results from meta-analysis indicating a mean weight loss of −1.36 kg at 12 months, and −1.23 kg at 24 months. Hartmann-Boyce et al. [ 22] found similar results in their meta-analysis of 23 studies of self-help interventions for weight loss in overweight and obese adults. They found that intervention participants lost significantly more weight than controls at 6 months (−1.85 kg) with no significant effect at 12 months (−0.76). They also showed that programs using some form of interactivity appeared to be more effective than controls at 6 months (−0.94 kg).
Taken altogether, our results support existing evidence on the effectiveness of technology mediated DPPs [19, 20], eHealth interventions for weight loss [18, 23, 36], and weight loss interventions in primary care settings [18, 36]. Our intent-to-treat weight loss magnitude across all three conditions at 6 (SC: −1.52, Class/IVR: −1.70, DVD/IVR: −3.04), 12 (SC: −1.56, Class/IVR: −2.04, DVD/IVR: −2.79), and 18 months (SC: −1.15, Class/IVR: −1.46, DVD/IVR: −2.55) were well within the range found in these reviews (−0.08 to −3.76). Most importantly, our study presents results at 18 months with significant reductions in BMI, which represents a significant addition to current literature on the weight loss maintenance effect of technology enhanced interventions delivered within a primary care setting [18, 23, 36]. Additionally, our attrition rates (31–$38\%$) and percent of individuals achieving at least $5\%$ weight loss in all study groups across all timepoints (ITT: 15.94–$26.87\%$) are well within the range found by Levine et al. [ 18] when investigating technology-assisted weight loss interventions in primary care.
Additionally, while not the original purpose of this study, results from this trial seem to indicate that both the DVD/IVR and the Class/IVR groups could meet CDC recognition standards [35] minus the “Live” health coaching requirement. Most importantly, our CDC threshold results indicate similar results to the latest NDPP evaluation [37]. We found that on average participants completed 15.5 sessions in DVD/IVR and 14.1 in Class/IVR (NDPP: 14) with $86.3\%$ of participants in the intervention groups (NDPP: $86.6\%$) meeting the CDC threshold of completing at least 4 sessions, $48.4\%$ (NDPP: $48.3\%$) staying in the program for at least 6 months, and $29.5\%$ (NDPP: $10.4\%$) attending at least the full 12-month program. Average weight lost at 12 months for DVD/IVR participants meeting the CDC threshold was $3.24\%$ compared to $2.74\%$ for Class/IVR (NDPP: $3.6\%$) with $35.74\%$ of DVD/IVR participants and $33.44\%$ of Class/IVR (NDPP: $35.5\%$) achieving $5\%$ weight loss goal.
These are important findings when considering that several barriers to large scale implementation remain for technology enhanced DPPs [19, 20] and weight loss interventions in primary care settings [18, 23, 36]. In fact, recent studies [11, 12] have shown that the NDPP as currently delivered and its technology-based options are not able to reach a large proportion of the American population. These studies have shown a lack of availability of these programs in underserved areas [11, 12] where primary care is overburdened and under resourced, and when these programs are offered, they fail to attract a large and representative sample of the target population [12]. Further, the spotty access to internet services and reliance on data plans presents a barrier to engagement in traditional eHealth programs requiring Internet connection [38]. The use of smart automated telephone calls to deliver DPP content shows promise in addressing these issues. The IVR system addresses barriers at multiple implementation levels. At the organizational, setting level the IVR system reduces the need for space, staff time, overall costs, and scheduling barriers. At the organizational, staff-level the IVR system reduces staff burden of delivering NDPP, and difficulty on scheduling participants, and allows staff to spend more time building relationships with participants to improve overall engagement. At the individual level, the IVR system addresses issues of distance, lack of transportation, work schedules, unreliable access to the Internet, and childcare needs. This is particularly important for primary care settings in underserved communities where there is a lack of resources, (i.e., medically underserved areas, space, competing demands, and expertise) and geographic segregation makes it difficult to deliver the NDPP or weight loss programs (13–16).
The present trial is not without limitations. First, study participants in the DVD/IVR group presented significantly higher BMI with a higher proportion being at Class III *Obesity status* at baseline. While we used randomization procedures, we did not stratify by BMI status. Nevertheless, we accounted for these initial differences by using baseline BMI values as control variable in our models. Second, we had an overall high attrition rate. As such, our results must be seen with caution as up to $38\%$ of our participants did not complete follow-up assessments. These numbers were particularly higher among Class/IVR participants reaching $50\%$ at 18 months. Nonetheless, we used Intent-to-treat analysis to include all participants with a baseline value in our models. When comparing with other studies, we also see similar attrition rates for weight loss programs in primary care [18]. Future studies should continue to investigate factors influencing participant engagement and retention in weight loss interventions delivered in primary care settings. Finally, our trial lasted only 18 months and did not include glycemic control (e.g., HbA1) or event based (e.g., T2D incidence) outcomes. Thus, long-term effects of the three groups await further investigation.
In closing, our findings show that a technology-enhanced diabetes prevention program was effective in reducing BMI at 6 months and maintaining these results at 12 and 18 months in a group of primary care patients at risk for developing T2D. The DiaBEAT-it interventions respond to the need for scalable, effective methods to manage obesity and prevent diabetes in primary care settings that do not over burden primary care clinics and providers. Further, the CDC requires the inclusion of a lifestyle coach in any in-person or technology-based program as one of the standards for recognition in its National Diabetes Prevention Program [35]. Consideration of expanded program criteria to include fully-automated systems and or the possibility of engaging clinical staff as engagement agents instead of lifestyle coaches, may reduce the burden placed on many resource strapped primary care clinics and improve the potential for adoption and sustainability of DPP adaptations. Effective automated technologies such as DiaBEAT-it represent one of these strategies with the potential to serving a large, representative and geographically distant population, while decreasing the need for organizational resources and reliance on Internet availability.
## 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 Carilion Clinic Institutional Review Board. The patients/participants provided their written informed consent to participate in this study.
## Author contributions
FA, WY, RS, BD, MG, JH, and PE contributed to the conception of the protocol and study design. FA, WY, FB, TA, CG, and SW were involved with the data collection and analysis. All authors were involved in writing the paper and had final approval of the submitted 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.2023.1000162/full#supplementary-material
## References
1. 1.Centers for Disease Control and Prevention. Estimates of Diabetes and Its Burden in the United States. National Diabetes Statistics Report. Atlanta, GA: US Department of Health and Human Services (2020).. *Estimates of Diabetes and Its Burden in the United States. National Diabetes Statistics Report* (2020)
2. Tabák AG, Herder C, Rathmann W, Brunner EJ, Kivimäki M. **Prediabetes: a high-risk state for developing diabetes**. *Lancet.* (2012) **379** 11. DOI: 10.1016/S0140-6736(12)60283-9
3. Nathan DM, Davidson MB, DeFronzo RA, Heine RJ, Henry RR, Pratley R. **Impaired fasting glucose and impaired glucose tolerance: implications for care**. *Diabetes Care.* (2007) **30** 753-9. DOI: 10.2337/dc07-9920
4. 4.Centers for Disease Control and Prevention. Adult Obesity Facts. Atlanta, GA: US Department of Health and Human Services (2020).. *Adult Obesity Facts* (2020)
5. 5.Centers for Disease Control and Prevention. Prediabetes - Your Chance to Prevent Type 2 Diabetes. Atlanta, GA: US Department of Health and Human Services (2020).. *Prediabetes - Your Chance to Prevent Type 2 Diabetes* (2020)
6. Knowler WC, Barrett-Connor E, Fowler SE. **Reduction in the incidence of type 2 diabetes with lifestyle intervention or metformin**. *N Engl J Med.* (2002) **346** 393-403. DOI: 10.1056/NEJMoa012512
7. Ali MK, Echouffo-Tcheugui J, Williamson DF. **How effective were lifestyle interventions in real-world settings that were modeled on the Diabetes Prevention Program?**. *Health Aff.* (2012) **31** 67-75. DOI: 10.1377/hlthaff.2011.1009
8. Whittemore R. **A systematic review of the translational research on the Diabetes Prevention Program**. *Transl Behav Med.* (2011) **1** 480-91. DOI: 10.1007/s13142-011-0062-y
9. Dunkley AJ, Bodicoat DH, Greaves CJ, Russell C, Yates T, Davies MJ. **Diabetes prevention in the real world: effectiveness of pragmatic lifestyle interventions for the prevention of type 2 diabetes and of the impact of adherence to guideline recommendations: a systematic review and meta-analysis**. *Diabetes Care.* (2014) **37** 922-33. DOI: 10.2337/dc13-2195
10. Dugani SB, Mielke MM, Vella A. **Burden and management of type 2 diabetes in rural United States**. *Diabetes Metab Res Rev* (2020) **37** e3410. DOI: 10.1002/dmrr.3410
11. Ariel-Donges AH, Gordon EL, Dixon BN, Eastman AJ, Bauman V, Ross KM. **Rural/urban disparities in access to the National Diabetes Prevention Program**. *Transl Behav Med.* (2020) **10** 1554-8. DOI: 10.1093/tbm/ibz098
12. Ali MK, Bullard KM, Imperatore G, Benoit SR, Rolka DB, Albright AL. **Reach and use of diabetes prevention services in the United States, 2016–2017**. *JAMA Netw Open* (2019) **2** e193160. DOI: 10.1001/jamanetworkopen.2019.3160
13. Halley MC, Petersen J, Nasrallah C, Szwerinski N, Romanelli R, Azar KM. **Barriers and facilitators to real-world implementation of the diabetes prevention program in large healthcare systems: lifestyle coach perspectives**. *J Gen Intern Med.* (2020) **35** 8. DOI: 10.1007/s11606-020-05744-y
14. Venditti EM, Wylie-Rosett J, Delahanty LM, Mele L, Hoskin MA, Edelstein SL. **Short and long-term lifestyle coaching approaches used to address diverse participant barriers to weight loss and physical activity adherence**. *Int J Behav Nutr Phys Act.* (2014) **11** 16. DOI: 10.1186/1479-5868-11-16
15. Laws RA, Fanaian M, Jayasinghe UW, McKenzie S, Passey M, Davies GP. **Factors influencing participation in a vascular disease prevention lifestyle program among participants in a cluster randomized trial**. *BMC Health Serv Res.* (2013) **13** 12. DOI: 10.1186/1472-6963-13-201
16. Ackermann RT. **Working with the YMCA to implement the Diabetes Prevention Program**. *Am J Prev Med* (2013) **44** S352-6. DOI: 10.1016/j.amepre.2012.12.010
17. Rhoon LV, Byrne M, Morrissey E, Murphy J, McSharry J. **A systematic review of the behaviour change techniques and digital features in technologydriven type 2 diabetes prevention interventions**. *Digital Health.* (2020) **6** 26. DOI: 10.1177/2055207620914427
18. Levine DM, Savarimuthu S, Squires A, Nicholson J, Jay M. **Technology-assisted weight loss interventions in primary care: a systematic review**. *J Gen Intern Med.* (2015) **30** 107-17. DOI: 10.1007/s11606-014-2987-6
19. Bian RR, Piatt GA, Sen A, Plegue MA, De Michele ML, Hafez D. **The effect of technology-mediated diabetes prevention interventions on weight: a meta-analysis**. *J Med Internet Res.* (2017) **19** e76. DOI: 10.2196/jmir.4709
20. Joiner KL, Nam S, Whittemore R. **Lifestyle interventions based on the diabetes prevention program delivered**. *Prev Med.* (2017) **100** 13. DOI: 10.1016/j.ypmed.2017.04.033
21. Block G, Azar KM, Romanelli RJ, Block TJ, Hopkins D, Carpenter HA. **Diabetes prevention and weight loss with a fully automated behavioral intervention by Email, Web, and Mobile Phone: a randomized controlled trial among persons with prediabetes**. *J Med Internet Res.* (2015) **17** e240. DOI: 10.2196/jmir.4897
22. Hartmann-Boyce J, Jebb SA, Fletcher BR, Aveyard P. **Self-help for weight loss in overweight and obese adults: systematic review and meta-analysis**. *Am J Public Health.* (2015) **105** e43-57. DOI: 10.2105/AJPH.2014.302389
23. Hutchesson MJ, Rollo ME, Krukowski R, Ells L, Harvey J, Morgan PJ. **eHealth interventions for the prevention and treatment of overweight and obesity in adults: a systematic review with meta-analysis**. *Obes Rev.* (2015) **16** 376-92. DOI: 10.1111/obr.12268
24. Almeida FA, Pardo KA, Seidel RW, Davy BM, You W, Wall SS. **Design and methods of “fthodsA, Seid a hybrid preference/randomized control trial design using the RE-AIM framework**. *Contemp Clin Trials.* (2014) **38** 383-96. DOI: 10.1016/j.cct.2014.06.010
25. Almeida FA, Shetterly S, Smith-Ray RL, Estabrooks PA. **Reach and effectiveness of a weight loss intervention in patients with prediabetes in Colorado**. *Prev Chronic Dis.* (2010) **7** A103. PMID: 20712930
26. Smith-Ray RL, Almeida FA, Bajaj J, Foland S, Gilson M, Heikkinen S. **Translating efficacious behavioral principles for diabetes prevention into practice**. *Health Promot Pract.* (2009) **10** 58-66. DOI: 10.1177/1524839906293397
27. Estabrooks PA, Smith-Ray RL. **Piloting a behavioral intervention delivered through interactive voice response telephone messages to promote weight loss in a pre-diabetic population**. *Patient Educ Couns.* (2008) **72** 34-41. DOI: 10.1016/j.pec.2008.01.007
28. Harvey-Berino J, Pintauro S, Buzzell P, Gold EC. **Effect of internet support on the long-term maintenance of weight loss**. *Obes Res.* (2004) **12** 320-9. DOI: 10.1038/oby.2004.40
29. Perri MG, Limacher MC, Durning PE, Janicke DM, Lutes LD, Bobroff LB. **Treatment of obesity in underserved rural settings (TOURS): a randomized trial of extended-care programs for weight management**. *Arch Intern Med.* (2008) **168** 7. DOI: 10.1001/archinte.168.21.2347
30. Heikes KE, Eddy DM, Arondekar B, Schlessinger L. **Diabetes risk calculator: a simple tool for detecting undiagnosed diabetes and pre-diabetes Diabetes**. *Care.* (2008) **31** 1040-5. DOI: 10.2337/dc07-1150
31. Goessl C, Estabrooks P, You W, Britigan D, DeAlba A, Almeida F. **Effectiveness of DVD vs. group-initiated diabetes prevention on information uptake for high & low health literacy participants**. *Patient Educ Couns.* (2019) **102** 968-75. DOI: 10.1016/j.pec.2018.12.026
32. Estabrooks PA, Glasgow RE, Dzewaltowski DA. **Physical activity promotion through primary care**. *JAMA.* (2003) **289** 2913-6. DOI: 10.1001/jama.289.22.2913
33. Laird NM, Ware JH. **Random-effects models for longitudinal data**. *Biometrics.* (1982) **38** 963-74. DOI: 10.2307/2529876
34. Verbeke G, Molenberghs G. *Linear Mixed Models for Longitudinal Data* (2000)
35. 35.Centers for Disease Control and Prevention. Diabetes Prevention Recognition Program: Standards and Operating Procedures. Atlanta, GA: US Department of Health and Human Services (2021).. *Diabetes Prevention Recognition Program: Standards and Operating Procedures* (2021)
36. Booth HP, Prevost TA, Wright AJ, Gulliford MC. **Effectiveness of behavioural weight loss interventions delivered in a primary care setting: a systematic review and meta-analysis**. *Family Practice.* (2014) **31** 10. DOI: 10.1093/fampra/cmu064
37. Ely EK, Gruss SM, Luman ET, Gregg EW, Ali MK, Nhim K. **A national effort to prevent type 2 diabetes: participant-level evaluation of CDC's national diabetes prevention program**. *Diabetes Care.* (2017) **40** 10. DOI: 10.2337/dc16-2099
38. Perrin A. *Digital Gap Between Rural and Nonrural America Persists* (2019)
|
---
title: 'Adherence to 24-h Movement Guidelines and Depressive Status During the Coronavirus
Disease Outbreak: A Cross-Sectional Japanese Survey'
authors:
- Takahisa Ohta
- Madoka Ogawa
- Naoki Kikuchi
- Hiroyuki Sasai
- Takanobu Okamoto
journal: International Journal of Public Health
year: 2023
pmcid: PMC9998524
doi: 10.3389/ijph.2023.1604647
license: CC BY 4.0
---
# Adherence to 24-h Movement Guidelines and Depressive Status During the Coronavirus Disease Outbreak: A Cross-Sectional Japanese Survey
## Abstract
Objectives: *The coronavirus* disease 2019 (COVID-19) pandemic has affected people’s physical activity, sedentary behavior, and sleep. This study aimed to clarify the association between combining these factors, integrated as adherence to 24-h movement guidelines, and depressive status during the COVID-19 pandemic.
Methods: At the end of October 2020, we sent self-administered questionnaires to 1,711 adults aged ≥18. We assessed physical activity, sedentary behavior, sleep duration, adherence to 24-h movement guidelines, depressive status, and confounding factors.
Results: Of the 640 valid responses, 90 ($14.1\%$) reported a depressive status. Multivariable odds ratios ($95\%$ confidence interval) of depressive status were 0.22 (0.07, 0.71) for all three recommendations of the 24-h movement guidelines and those who met none of the recommendations as reference. The number of guidelines met was associated with depressive status in a dose-response fashion.
Conclusion: Meeting the 24-h movement guidelines was associated with a lower prevalence of depressive status during the COVID-19 pandemic. Adults should adhere to these guidelines to maintain their mental health during future quarantine life.
## Introduction
The coronavirus disease 2019 (COVID-19) outbreak has drastically affected the economy and people’s lifestyles worldwide [1, 2] Physical activity (PA) decreased during the pandemic due to implementing a quarantined lifestyle [3] Sadly, behavioral changes such as decreased PA and prolonged sedentary behavior (SB) may increase the risk of adverse health outcomes, including cancer, cardiovascular disease, diabetes, and depression [4, 5]. The prevalence of depression during the COVID-19 pandemic was reportedly $27\%$–$45\%$ worldwide and has become a serious public health concern [6–11].
Depression is a common illness in Japan and a proven risk factor for suicide; it also brings a substantial economic burden [12, 13] Therefore, preventive measures are necessary to avoid the onset of depression and promote mental health. Previous studies before the COVID-19 pandemic suggested that moderate PA slows down the onset of depression. A systematic review and meta-analysis showed a non-linear dose-response association between PA and health outcomes, with higher PA preventing the onset of depression compared with lower PA [14]. A longer time spent in SB also carried a higher risk of depression, with a relative risk of 1.25 ($95\%$ confidence interval [CI]: 1.16–1.35) for a longer versus a shorter time spent in SB [15].
A rapid systematic review during the pandemic demonstrated that a higher volume and frequency of moderate-to-vigorous physical activity (MVPA) was associated with a $12\%$–$32\%$ and $15\%$–$34\%$ lower prevalence of depression and anxiety, respectively [16]. Other studies showed that low PA was also associated with mental illness during the COVID-19 pandemic [17, 18]. However, these studies had notable limitations and challenges.
First, the above rapid systematic review focused only on individual components such as PA, which may have unknown confounding factors. An individual’s 24-h daily activity comprises three components, PA, SB, and sleep duration. These three components are codependent rather than independent, and seeing them as a single frame is beginning to take hold as the latest public health concept. The World Health Organization (WHO) and academic societies from several countries strongly support this concept [17–19] In particular, a 24-h movement guideline of the Canadian Society of Exercise Physiology recommends limiting SB to <8 h per day, MVPA to at least 150 min per week, and 7–9 h of sleep (7–8 h for adults ≥65 years) [17]. Furthermore, few studies have examined the association between mental health and 24-h movement guidelines, including depression, although associations with adiposity markers and cardiovascular markers have been reported [20].
Second, there is a lack of information on the associations between mental health and 24-h movement for the Japanese population during the COVID-19 pandemic. The policies (e.g., lockdown) to prevent the spread of COVID-19 differ among countries. Since Japan declared a state of emergency with fewer legal restrictions than other countries, it is impossible to compare Japan with other countries. For example, the studies conducted during the lockdown showed a decrease in PA and an increase in SB among U.S. college students and adults compared to pre-lockdown conditions, and adverse mental health consequences were also observed [21, 22]. In Japan, an increase in suicide rates during the pandemic has been reported, and suggestions for preventive strategies to maintain mental wellbeing must be an urgent issue [23]. Thus, clarifying the importance of 24-h movement guidelines for each country’s response during a pandemic may be a piece of important fundamental knowledge for preparing for the next unknown infectious disease.
This study aimed to clarify the association between adherence to the 24-h movement guidelines (in terms of PA, SB, and sleep duration) and the prevalence of depressive status among middle-aged and older adults during the COVID-19 outbreak in Japan. The findings of this study will allow us to add to the knowledge of preventive medicine targeted toward avoiding depressive symptoms in current and future similar crises that may require implementing a quarantine lifestyle.
## Design and Participants
This cross-sectional survey, a part of the Setagaya- Aoba study, was conducted at the end of October 2020 [24]. On either side of the survey period, states of emergency were declared in Japan due to the COVID-19 pandemic. The first state of emergency lasted from 7 April to 25 May 2020, and the second from 8 January to 21 March 2021. The Nippon Sport Science *University is* located in Setagaya, Tokyo, and Aoba, Yokohama-city has direct access to the metropolitan area and is called a commuter town. The Nippon Sport Science University holds physical fitness test events for citizens annually. These events involve approximately 1,000 persons living around the university each year and contribute their health assessments for free. Event guidance is provided on the bulletin board and the university website from September to October every year. After an online survey announcement, individuals who had participated in our fitness test events since 2017 received the mail-based questionnaire. Participants were excluded for the following reasons: age <18 years old, non-response, or response without physical activity or a depressive questionnaire.
## Physical Activity, SB, and Sleep Duration
Physical activity was assessed using the Global Physical Activity Questionnaire (GPAQ, Japanese version), which was developed initially by the WHO [25]. The weekly cut-off for physical activity among adults was based on the 24-h movement guidelines and defined as performing moderate-to-vigorous-intensity aerobic physical activity for >150 min per week. SB was also assessed using the GPAQ using the question, “How much time do you usually spend sitting or reclining on a typical day?” SB was evaluated based on the 24-h movement guidelines and categorized as < 8 h per day according to the duration. Sleep duration was assessed using a single self-administered question, “How long do you usually sleep a day?” Participants could respond by reporting the time in hours and minutes. Based on 24-h movement guidelines, the limits were set as 7–9 h per day for adults 18–64 years old and 7–8 h for adults ≥65 years old [17].
## Adherence to 24-h Movement Guidelines
Participants were categorized into eight groups to assess adherence to components of the Canadian 24-h movement guidelines as follows: 1) not meeting any recommendations, 2) meeting only SB recommendations, 3) meeting only MVPA recommendations, 4) meeting only sleep duration recommendations, 5) meeting SB and MVPA recommendations, 6) meeting SB and sleep duration recommendations, 7) meeting MVPA and sleep duration recommendations, and 8) meeting all recommendations of the 24-h movement guidelines. In addition, participants were categorized into the following four groups to assess their adherence to the number of guideline recommendations: 1) not meeting any recommendations, 2) meeting only one recommendation, 3) meeting two recommendations, and 4) meeting all recommendations.
## Depressive Status
The Center for Epidemiologic Studies Depression Scale (CES-D) was used to assess depressive status [26]. The CES-D is a valid and reliable tool comprising 20 items. The CES-D instrument evaluates whether the respondent has experienced depression symptoms, such as restless sleep, poor appetite, and feeling of loneliness, within the preceding week. Responses consisted of symptom frequency and ranged from 0 to 3 for each item. For example, the responses “rarely/never,” “sometimes,” “moderately/most of the time,” and “mostly/almost all the time” were scored as 0, 1, 2, and 3 points, respectively. We also summed the score of the 20 items (range: 0–60) and assigned a score of ≥16 as the cut-off for depressive states.
## Confounding Factors
Body mass index was calculated using self-reported height and weight. We used a self-administered questionnaire to obtain data regarding smoking status (yes/no), alcohol consumption (yes/no), changes in income in 2019 (compared with before the COVID-19 outbreak), living alone (yes/no), daily use of social networking service (SNS; yes/no), history of depression (yes/no), the number of chronic diseases requiring medicine, and menopause (yes/no).
## Statistical Analysis
To reduce the bias of missing values, we conducted multiple imputations using every explanatory variable and obtained 20 datasets. First, an analysis of each dataset was conducted each dataset, and then the results of 20 multiple imputation datasets were integrated into one result. The continuous variables are presented as means (standard deviations) or medians (interquartile ranges) based on the normality test results. The categorical variables were shown as numbers (percentages).
A logistic regression model was used to perform a univariate analysis between each confounding factor and depressive status. The logistic regression analyses examined the association between meeting the 24-h movement guideline and the prevalence of depressive status. The models were adjusted for age, body mass index, smoking habits [27], alcohol habits [28], changes in income since 2019, presence of a housemate, daily use of SNS [29] a history of depression, number of chronic diseases requiring medicine [30], and menopause [31]. The multivariable odds ratios (ORs) and $95\%$ CIs for the associations between meeting the 24-h movement guidelines and depressive status were calculated using the “not meeting the guidelines” group as a reference. A continuous variable (the number of guidelines met) was included in a separate model to evaluate the linear associations. All the statistical analyses were performed using the SPSS version 25 (IBM, Inc., Chicago, IL, United States). A p-value of <0.05 was considered to be statistically significant.
## Results
During 2017–2019, 1,711 adults participated in physical fitness tests at the Tokyo Setagaya and Yokohama Kenshidai (Aoba-ward) Nippon Sport Science University campuses. All these individuals were eligible to participate in the study. Of the 640 valid responders, 90 ($14.1\%$) reported a depressive status (Table 1).
**TABLE 1**
| Unnamed: 0 | Overall | Number of guideline meeting | Number of guideline meeting.1 | Number of guideline meeting.2 | Number of guideline meeting.3 |
| --- | --- | --- | --- | --- | --- |
| | Overall | 0 | 1 | 2 | 3 |
| | n = 640 | n = 38 | n = 166 | n = 293 | n = 143 |
| No. of cases, n (%) | 90 (14.1) | 12 | 27 | 38 | 13 |
| CES-D score | 9.4 (6.7) | 13.2 (8.0) | 9.7 (7.4) | 9.1 (6.1) | 8.6 (6.3) |
| Sex | Sex | Sex | Sex | Sex | Sex |
| Men, n (%) | 268 (41.7) | 15 | 73 | 119 | 61 |
| Women, n (%) | 372 (58.3) | 23 | 93 | 174 | 82 |
| Age, year | 64.1 (14.0) | 60.9 (13.1) | 60.7 (14.0) | 65.2 (13.6) | 66.5 (14.3) |
| Body mass index, kg/m2 | 21.8 (3.4) | 22.3 (3.9) | 21.7 (3.7) | 22.0 (3.3) | 21.6 (3.0) |
| Missing, n (%) | 7 (1.1) | — | 3 (1.8) | 4 (1.4) | - |
| Smoking, n (%) | 30 (4.7) | 5 (13.2) | 7 (4.2) | 11 (3.8) | 7 (4.9) |
| Missing, n (%) | 9 (1.4) | — | 6 (3.6) | 3 (1.0) | - |
| Alcohol, n (%) | 304 (47.3) | 17 (44.7) | 76 (45.8) | 140 (47.8) | 71 (49.7) |
| Missing, n (%) | 9 (1.4) | — | 5 (3.0) | 3 (1.0) | 1 (0.7) |
| Sleep duration, hr | Sleep duration, hr | Sleep duration, hr | Sleep duration, hr | Sleep duration, hr | Sleep duration, hr |
| <6 h, n (%) | 124 (19.4) | 16 (42.1) | 44 (26.5) | 64 (21.8) | — |
| ≤6, <8, n (%) | 431 (67.3) | 21 (55.3) | 114 (68.7) | 197 (67.2) | 99 (69.2) |
| 8 ≤, n (%) | 85 (13.3) | 1 (2.6) | 8 (4.8) | 32 (10.9) | 44 (30.8) |
| Changes in income, n (%) | Changes in income, n (%) | Changes in income, n (%) | Changes in income, n (%) | Changes in income, n (%) | Changes in income, n (%) |
| Decreased | 79 (14.8) | 3 (7.9) | 18 (10.8) | 40 (13.7) | 18 (12.6) |
| Unchanged | 440 (68.8) | 29 (76.3) | 115 (69.3) | 119 (40.6) | 97 (67.8) |
| Increased | 14 (2.2) | 3 (7.9) | 4 (2.4) | 5 (1.7) | 2 (1.4) |
| Missing, n (%) | 107 (16.9) | 3 (7.9) | 29 (17.5) | 49 (16.7) | 26 (18.2) |
| Using SNS, n (%) | 410 (36.3) | 23 (60.5) | 109 (65.7) | 185 (63.1) | 93 (65.0) |
| Living alone, n (%) | 75 (12.1) | 4 (10.5) | 28 (16.9) | 30 (10.2) | 13 (9.1) |
| Missing, n (%) | 10 (1.6) | — | 4 (2.4) | 5 (1.7) | 1 (0.7) |
| History of depression, (%) | 13 (2.0) | 1 (2.6) | 4 (2.4) | 6 (2.0) | 2 (1.4) |
| Number of chronic diseases, n (%) | — | | | | |
| 0 | 387 (60.5) | 25 (65.8) | 112 (67.5) | 160 (54.6) | 90 (62.9) |
| 1 | 175 (27.3) | 9 (23.7) | 42 (25.3) | 84 (28.7) | 40 (28.0) |
| 2 | 55 (8.6) | 2 (5.3) | 8 (4.8) | 36 (12.3) | 9 (6.3) |
| ≥3 | 23 (3.6) | 2 (5.3) | 4 (2.4) | 13 (4.4) | 4 (2.8) |
| Menopause (women only) | 290 (77.9) | 17 (73.9) | 70 (75.3) | 142 (81.6) | 61 (74.4) |
Table 2 shows the association between the prevalence of depressive status and each variable. Age and the existence of housemates were associated with the prevalence of depressive status.
**TABLE 2**
| Variables | No. | No. of cases | Prevalence a | OR (95% CIs) |
| --- | --- | --- | --- | --- |
| Sex | Sex | Sex | Sex | Sex |
| Women | 372 | 62 | 166.7 | 1.00 (References) |
| Men | 268 | 28 | 104.5 | 0.58 (0.36, 0.94) |
| Age b | 640 | 90 | 140.6 | 0.82 (0.70, 0.96) |
| Body mass index | 633 | 90 | 142.2 | 1.05 (0.98, 1.13) |
| Smoking | Smoking | Smoking | Smoking | Smoking |
| No | 601 | 82 | 136.4 | 1.00 (References) |
| Yes | 30 | 6 | 200.0 | 1.73 (0.71, 4.22) |
| Alcohol | Alcohol | Alcohol | Alcohol | Alcohol |
| No | 327 | 46 | 140.7 | 1.00 (References) |
| Yes | 304 | 41 | 134.9 | 0.94 (0.60, 1.48) |
| Sleep duration | Sleep duration | Sleep duration | Sleep duration | Sleep duration |
| <6 h | 124 | 25 | 201.6 | 1.69 (1.00, 2.85) |
| 6 ≤, <8 | 431 | 56 | 129.9 | 1.00 (References) |
| 8 ≤ | 85 | 9 | 105.9 | 0.79 (0.38, 1.67) |
| Changes in income | Changes in income | Changes in income | Changes in income | Changes in income |
| Decreased | 79 | 20 | 253.2 | 1.00 (References) |
| Unchanged | 440 | 52 | 118.2 | 0.48 (0.27, 0.84) |
| Increased | 14 | 3 | 214.3 | 0.58 (0.19, 1.82) |
| Using SNS | Using SNS | Using SNS | Using SNS | Using SNS |
| No | 230 | 33 | 143.5 | 1.00 (References) |
| Yes | 410 | 57 | 139.0 | 0.96 (0.61, 1.53) |
| Living alone | Living alone | Living alone | Living alone | Living alone |
| No | 555 | 72 | 129.7 | 1.00 (References) |
| Yes | 75 | 16 | 213.3 | 1.87 (1.03, 3.41) |
| History of depression | History of depression | History of depression | History of depression | History of depression |
| No | 627 | 88 | 140.4 | 1.00 (References) |
| Yes | 13 | 2 | 153.8 | 1.11 (0.24, 5.11) |
| Medication of for chronic disease | Medication of for chronic disease | Medication of for chronic disease | Medication of for chronic disease | Medication of for chronic disease |
| 0 | 387 | 45 | 116.3 | 1.00 (References) |
| 1 | 175 | 34 | 194.3 | 1.83 (1.12, 2.98) |
| 2 | 55 | 9 | 163.6 | 1.49 (0.68, 3.24) |
| ≥3 | 23 | 2 | 87.0 | 0.72 (0.16, 3.19) |
| Menopause (women only) | Menopause (women only) | Menopause (women only) | Menopause (women only) | Menopause (women only) |
| No | 73 | 17 | 232.9 | 1.00 (References) |
| Yes | 290 | 43 | 148.3 | 0.57 (0.31, 1.08) |
The association between meeting the 24-h movement guidelines and depressive status is shown in Table 3. After adjusting for various confounding factors and using “not meeting the guidelines” as a reference, participants who complained about MVPA, sleep duration independently or in combination with sleep duration, and MVPA or SB were not significantly different. In addition, an inverse association was observed between the number of components met and the depressive status (P for the trend = 0.006). The multivariable ORs were 0.40 ($95\%$ CI: 0.17, 0.92) for one component met, 0.34 ($95\%$ CI: 0.15, 0.75) for two components met, and 0.24 ($95\%$ CI: 0.10, 0.61) for all components met, compared with the “none of the components met” as a reference.
**TABLE 3**
| Unnamed: 0 | Number of participants | Number of cases | Crude OR | Multivariable OR a |
| --- | --- | --- | --- | --- |
| Meeting recommendations | Meeting recommendations | Meeting recommendations | Meeting recommendations | Meeting recommendations |
| | 38 | 12 | 1.00 (References) | 1.00 (References) |
| SB only | 58 | 6 | 0.25 (0.08, 0.74) | 0.33 (0.09, 1.24) |
| MVPA only | 79 | 16 | 0.55 (0.23, 1.32) | 0.52 (0.15, 1.80) |
| Sleep duration only | 29 | 5 | 0.45 (0.14, 1.47) | 0.93 (0.22, 3.88) |
| SB and MVPA | 190 | 23 | 0.30 (0.13, 0.67) | 0.19 (0.06, 0.58) |
| SB and sleep duration | 37 | 6 | 0.42 (0.14, 1.27) | 0.44 (0.11, 1.74) |
| MVPA and sleep duration | 66 | 9 | 0.34 (0.13, 0.91) | 0.42 (0.11, 1.60) |
| SB, MVPA, and sleep duration | 143 | 13 | 0.22 (0.09, 0.52) | 0.22 (0.07, 0.70) |
| Number of meeting recommendations | Number of meeting recommendations | Number of meeting recommendations | Number of meeting recommendations | Number of meeting recommendations |
| | 38 | 12 | 1.00 (References) | 1.00 (References) |
| 1 | 166 | 27 | 0.42 (0.19, 0.94) | 0.50 (0.18, 1.42) |
| 2 | 293 | 38 | 0.32 (0.15, 0.69) | 0.26 (0.09, 0.72) |
| 3 b | 143 | 13 | 0.22 (0.09, 0.53) | 0.22 (0.07, 0.71) |
| p for trend | | | 0.001 | 0.004 |
## Discussion
This study investigated the association between adherence to 24-h movement guidelines and the prevalence of depressive status using a survey conducted between Japan’s first and second states of emergency. Our primary findings were as follows. First, those who met the 24-h movement guidelines had a lesser depressive status than those who did not. Second, there was a strong association between SB and MVPA with a less depressive status. These findings emphasized the importance of meeting the guidelines, especially for adults. Thus, SB and PA were more important preventive components of a depressive status during the COVID-19 outbreak. However, the causal relationship remains unclear. Therefore, further longitudinal studies are required.
The prevalence of depressive symptoms during the lockdown was $27.8\%$ in the United States and $24\%$–$31\%$ in China; WHO has reported an increase in the global prevalence of depressive symptoms to $25\%$ due to the COVID-19 pandemic [32–34]. This study’s depressive symptoms prevalence was $14.1\%$, which was lower than the values mentioned above. It is unclear whether this is because the participants were not subjected to firm behavioral restrictions such as a lockdown. Yamamoto et al. conducted a nationwide online survey in Japan and found depressive symptoms prevalence of $17.9\%$, which is consistent with the present study [35].
Complete adherence to the recommended 24-h movement guidelines may improve health outcomes rather than independently meeting each component’s guidelines. Numerous previous investigations have shown an association between each component and depressive status. In addition, MVPA may improve health outcomes such as cardiovascular disease, biomarkers, and all-cause mortality [20]. However, to our knowledge, only a few studies have evaluated the association between adherence to the 24-h movement guidelines and mental health, including depression, for adults. Therefore, we believe this study is the first evidence of the benefit of 24-h movement guidelines for mental health. For those who stay at home due to the COVID-19 pandemic, reducing sedentary time to MVPA may improve or maintain mental health.
Those who decrease SB alone or increase MVPA, which meets the recommendation independently, may not be protected sufficiently from the onset of depression. Our study showed that those who met the recommendation in combination with SB and MVPA had less depressive status than those who met the recommendation for sleep duration alone or combined with sleep duration and MVPA or SB. The detailed mechanisms of this association are still unknown. However, the association of each component with depression is well established. Previous studies have reported dose-response relationships between MVPA, a lower risk of depression or SB, and a high risk of depression, regardless of the COVID-19 pandemic [5, 15, 16]. Additionally, a substantial health benefit may be obtained through SB reallocated to MVPA, not sleep duration [20]. Furthermore, there was a linear association between a less depressive status and meeting the recommendations of the Canadian 24-h movement guidelines. Accordingly, our study showed that there may have been an additive association between meeting the SB recommendations and the PA recommendations for reducing depression.
Plausible mechanisms of these findings may exist in multiple cascades. One possible mechanism is the biological aspect. Those with depression are often observed as having decreased hippocampal volume and increased oxidative marker levels [36]. PA can regulate the oxidant marker and hippocampal volume through activated regulation of brain-derived neurotrophic factors [37, 38]. A meta-analysis reported that a PA intervention might improve self-worth [39]. Additionally, increasing PA and subsequent cardiorespiratory fitness benefits cognitive function [40]. Finally, PA may improve mental health through an intricate interaction with each mechanism, while increasing SB may displace PA. Heavy SB decreases social interaction, which may increase the risk of depression. During the COVID-19 pandemic, increasing SB due to the increased time spent staying at home or in quarantine may have resulted in less time for face-to-face interactions when working and more time spent communicating remotely. Researching how to improve and prevent depression is necessary.
This study had several strengths. First, the survey was conducted during the COVID-19 outbreak, which caused a crucial health burden for many people. Although studies worldwide have reported on the association between PA and mental health during the COVID-19 pandemic [16], corresponding evidence among the Japanese population is limited. Moreover, few previous studies have adjusted for social factors to investigate the association between PA and mental health. Therefore, this study contributed substantially to the current literature. Second, PA was evaluated using a well-validated self-reporting questionnaire (GPAQ) [25]. A systematic review demonstrated the association of PA with depression during COVID-19 but did not include an appropriate PA assessment. Finally, to our knowledge, this study may have been one of few studies in adults and older adults to assess compliance with 24-h movement guidelines. This observation will provide essential knowledge to prepare for the next unknown infectious disease or disaster.
## Limitations
However, there were also several limitations to our study. First, the design of this study was cross-sectional; therefore, we could not determine a causal relationship between adherence to the 24-h movement guidelines and depressive status. Second, PA, SB, and sleep duration were self-reported. Although the GPAQ is a well-established and valid version of questionnaires, such as the International Physical Activity Questionnaire, the responses are subjective. There may have been recall bias. Objective measurements of PA, SB, and sleep duration are needed to exclude this bias. Third, the small sample size and restricted population were considered possible limitations. Fourth, habitual dietary information was not collected. Since it may play a role in the residual bias for the prevalence of depressive status, nutritional intake data should also be collected in future longitudinal studies [41]. Fifth, reverse causality may have existed when considering the associations between depressive status and PA or sleep duration [42]. A longitudinal design to investigate the onset of depression is needed in the future. Finally, the 24-h movement guidelines consensus panel removed the ≥10-min bouts of PA recommendation because accumulated <10-min bouts can promote health benefits [17]. However, the GPAQ asked participants to consider ≥10-min bouts of every PA domain; there was an inconsistency between the questionnaire and the guidelines. Thus, this inconsistency may have underestimated the PA. This discrepancy should be removed using an objective tool such as a 3-axis accelerometer.
## Conclusion
Meeting all recommendations of the 24-h movement guidelines, or at least meeting SB and MVPA recommendations, was associated with a lower prevalence of depressive status. This finding suggested that adults meet the 24-h movement guidelines to maintain a healthy mental status. Evaluating human life across 24 h is closer to the reality of life than evaluating each element in isolation. In addition, PA, SB, and sleep duration may be critical factors to consider in preventing depressive status during the COVID-19 outbreak, which could be useful knowledge for the next unknown pandemic crises. Nevertheless, further longitudinal and interventional studies are needed to confirm the effectiveness of the 24-h movement guidelines.
## Ethics Statement
The studies involving human participants were reviewed and approved by the Ethics Committee of the Nippon Sport Science University. The patients/participants provided their written informed consent to participate in this study.
## Author Contributions
TOh designed the study; TOh, MO, NK, and TOk collected the data; TOh analyzed the data; TOh wrote the first draft; HS advised crucial to the draft and provided funding; HS, NK, and TOk oversaw the entire project. All authors have read and approved the final version of the manuscript and agree with the order of presentation of the authors.
## Conflict of Interest
The authors declare that they do not have any conflicts of interest.
## References
1. Kabir M, Afzal MS, Khan A, Ahmed H. **COVID-19 Pandemic and Economic Cost**. *impact on forcibly displaced people* (2020) **35** 101661
2. Mattioli Av, Sciomer S, Cocchi C, Maffei S. **Quarantine during COVID-19 Outbreak: Changes in Diet and Physical Activity Increase the Risk of Cardiovascular Disease. Nutrition**. *Metab Cardiovasc Dis* (2020) **30** 1409-0417
3. Goethals L, Barth N, Guyot J, Hupin D, Celarier T, Bongue B. **Impact of home Quarantine on Physical Activity Among Older Adults Living at home during the COVID-19 Pandemic: Qualitative Interview Study**. *J Med Internet Res* (2020) **22** 1-5
4. Kyu HH, Bachman VF, Alexander LT, Mumford JE, Afshin A, Estep K. **Physical Activity and Risk of Breast Cancer, colon Cancer, Diabetes, Ischemic Heart Disease, and Ischemic Stroke Events: Systematic Review and Dose-Response Meta-Analysis for the Global Burden of Disease Study 2013**. *BMJ (Online)* (2016) **354** i3857-10. DOI: 10.1136/bmj.i3857
5. Schuch F, Vancampfort D, Firth J, Rosenbaum S, Ward P, Reichert T. **Physical Activity and Sedentary Behavior in People with Major Depressive Disorder: A Systematic Review and Meta-Analysis**. *J Affect Disord [Internet]* (2017) **210** 139-50. DOI: 10.1016/j.jad.2016.10.050
6. Chen J, Farah N, Dong RK, Chen RZ, Xu W, Yin J. **Mental Health during the Covid-19 Crisis in Africa: A Systematic Review and Meta-Analysis**. *Int J Environ Res Public Health* (2021) **18** 10604. DOI: 10.3390/ijerph182010604
7. Zhang SX, Miller SO, Xu W, Yin A, Chen BZ, Delios A. **Meta-analytic Evidence of Depression and Anxiety in Eastern Europe during the COVID-19 Pandemic**. *Eur J Psychotraumatol* (2022) **13** 2000132. DOI: 10.1080/20008198.2021.2000132
8. Chen X, Chen J, Zhang M, Dong RK, Li J, Dong Z. **Meta-Regression on the Heterogenous Factors Contributing to the Prevalence of Mental Health Symptoms during the COVID-19 Crisis Among Healthcare Workers**. *Front Psychiatry* (2022) **13** 833865. DOI: 10.3389/fpsyt.2022.833865
9. Zhang SX, Batra K, Xu W, Liu T, Dong RK, Yin A. **Mental Disorder Symptoms during the COVID-19 Pandemic in Latin America – a Systematic Review and Meta-Analysis**. *Epidemiol Psychiatr Sci* (2022) **31** e23. DOI: 10.1017/S2045796021000767
10. Zhang SX, Chen RZ, Xu W, Yin A, Dong RK, Chen BZ. **A Systematic Review and Meta-Analysis of Symptoms of Anxiety, Depression, and Insomnia in Spain in the COVID-19 Crisis**. *Int J Environ Res Public Health* (2022) **19** 1018. DOI: 10.3390/ijerph19021018
11. Brailovskaia J, Cosci F, Mansueto G, Miragall M, Herrero R, Baños RM. **The Association between Depression Symptoms, Psychological burden Caused by Covid-19 and Physical Activity: An Investigation in Germany, Italy, Russia, and Spain**. *Psychiatry Res* (2021) **295** 113596. DOI: 10.1016/j.psychres.2020.113596
12. Gilman SE, Sucha E, Kingsbury M, Horton NJ, Murphy JM, Colman I. **Depression and Mortality in a Longitudinal Study: 1952-2011**. *Cmaj* (2017) **189** E1304-E1310-10. DOI: 10.1503/cmaj.170125
13. 13.
Ministry of Health Lobour and Welfare. Vital Statistics of Japan. Ministry of Health, Labour and Welfare (2019). p. 21. [Internet]Available from: https://www.mhlw.go.jp/toukei/saikin/hw/jinkou/kakutei19/dl/15_all.pdf
(Accessed June 16, 2021). *Vital Statistics of Japan* (2019) 21
14. Schuch FB, Vancampfort D, Firth J, Rosenbaum S, Ward PB, Silva ES. **Physical Activity and Incident Depression: A Meta-Analysis of Prospective Cohort Studies**. *Am J Psychiatry* (2018) **175** 631-48. DOI: 10.1176/appi.ajp.2018.17111194
15. Zhai L, Zhang Y, Zhang D. **Sedentary Behaviour and the Risk of Depression: A Meta-Analysis**. *Br J Sports Med* (2015) **49** 705-9. DOI: 10.1136/bjsports-2014-093613
16. Wolf S, Seiffer B, Zeibig JM, Welkerling J, Brokmeier L, Atrott B. **Is Physical Activity Associated with Less Depression and Anxiety during the COVID-19 Pandemic? A Rapid Systematic Review**. *Sports Med.* (2021) **51** 1771-1783. DOI: 10.1007/s40279-021-01468-z
17. Ross R, Chaput JP, Giangregorio LM, Janssen I, Saunders TJ, Kho ME. **Canadian 24-Hour Movement Guidelines for Adults Aged 18-64 Years and Adults Aged 65 Years or Older: an Integration of Physical Activity, Sedentary Behaviour, and Sleep**. *Appl Physiol Nutr Metab* (2020) **45** S57-S102-102. DOI: 10.1139/apnm-2020-0467
18. Tremblay MS, Chaput JP, Adamo KB, Aubert S, Barnes JD, Choquette L. **Canadian 24-Hour Movement Guidelines for the Early Years (0-4 Years): An Integration of Physical Activity, Sedentary Behaviour, and Sleep**. *BMC Public Health* (2017) **17** 874. DOI: 10.1186/s12889-017-4859-6
19. 19.
World Health Organaization. Guidelines on Physical Activity, Sedentary Behaviour and Sleep. World Health Organization (2019). p. 4. [Internet]p. Available from: https://apps.who.int/iris/bitstream/handle/10665/325147/WHO-NMH-PND-2019.4-eng.pdf?sequence=1&isAllowed=y%0Ahttp://www.who.int/iris/handle/10665/311664%0Ahttps://apps.who.int/iris/handle/10665/325147
(Accessed June 16, 2021). *Guidelines on Physical Activity, Sedentary Behaviour and Sleep* (2019) 4
20. Janssen I, Clarke AE, Carson V, Chaput JP, Giangregorio LM, Kho ME. **A Systematic Review of Compositional Data Analysis Studies Examining Associations between Sleep, Sedentary Behaviour, and Physical Activity with Health Outcomes in Adults**. *Appl Physiol Nutr Metab* (2020) **45** S248-S257-57. DOI: 10.1139/apnm-2020-0160
21. Romero-Blanco C, Rodríguez-Almagro J, Onieva-Zafra MD, Parra-Fernández ML, Prado-Laguna MDC, Hernández-Martínez A. **Physical Activity and Sedentary Lifestyle in university Students: Changes during Confinement Due to the Covid-19 Pandemic**. *Int J Environ Res Public Health* (2020) **17** 6567-13. DOI: 10.3390/ijerph17186567
22. Meyer J, McDowell C, Lansing J, Brower C, Smith L, Tully M. **Changes in Physical Activity and Sedentary Behavior in Response to Covid-19 and Their Associations with Mental Health in 3052 Us Adults**. *Int J Environ Res Public Health* (2020) **17** 6469-13. DOI: 10.3390/ijerph17186469
23. Tanaka T, Okamoto S. **Increase in Suicide Following an Initial Decline during the COVID-19 Pandemic in Japan**. *Nat Hum Behav* (2021) **5** 229-38. DOI: 10.1038/s41562-020-01042-z
24. Ohta T, Nagashima J, Sasai H, Kikuchi N, Nakazato K, Okamoto T. **Sport Program Service Study and Setagaya- Aoba Study**. *J Phys Fit Sports Med* (2021) 1-10
25. Bull FC, Maslin TS, Armstrong T. **Global Physical Activity Questionnaire (GPAQ): Nine Country Reliability and Validity Study**. *J Phys Act Health* (2009) **6** 790-804. DOI: 10.1123/jpah.6.6.790
26. Radloff LS. **The CES-D Scale: A Self-Report Depression Scale for Research in the General Population**. *Appl Psychol Meas* (1977) **1** 385-401. DOI: 10.1177/014662167700100306
27. Tomita A, Manuel JI. **Evidence on the Association between Cigarette Smoking and Incident Depression from the South African National Income Dynamics Study 2008-2015: Mental Health Implications for a Resource-Limited Setting**. *Nicotine Tob Res* (2020) **22** 118-23. DOI: 10.1093/ntr/nty163
28. Boden JM, Fergusson DM. **Alcohol and Depression**. *Addiction* (2011) **106** 906-14. DOI: 10.1111/j.1360-0443.2010.03351.x
29. Yoon S, Kleinman M, Mertz J, Brannick M. **Is Social Network Site Usage Related to Depression? A Meta-Analysis of Facebook–Depression Relations**. *J Affect Disord [Internet]* (2019) **248** 65-72. DOI: 10.1016/j.jad.2019.01.026
30. Read JR, Sharpe L, Modini M, Dear BF. **Multimorbidity and Depression: A Systematic Review and Meta-Analysis**. *J Affect Disord* (2017) **221** 36-46. DOI: 10.1016/j.jad.2017.06.009
31. Cole MG, Dendukuri N. **Risk Factors for Depression Among Elderly Community Subjects: A Systematic Review and Meta-Analysis**. *Am J Psychiatry* (2003) **160** 1147-56. DOI: 10.1176/appi.ajp.160.6.1147
32. Ettman CK, Abdalla SM, Cohen GH, Sampson L, Vivier PM, Galea S. **Prevalence of Depression Symptoms in US Adults before and during the COVID-19 Pandemic**. *JAMA Netw Open* (2020) **3** e2019686. DOI: 10.1001/jamanetworkopen.2020.19686
33. Vasilj I, Herceg K, Covic I, Santi M, Curlin M, Ljevak I. **Determinants of the COVID-19 Pandemic in the West Herzegovina Canton**. *Psychiatr Danub* (2020) **32** 221-5
34. 34.
World Health Organization. COVID-19 Pandemic Triggers 25% Increase in Prevalence of Anxiety and Depression Worldwide (2022). [Internet][cited 2022 Oct 18]. Available from: https://www.who.int/news/item/02-03-2022-covid-19-pandemic-triggers-25-increase-in-prevalence-of-anxiety-and-depression-worldwide
(Accessed October 18, 2022). *COVID-19 Pandemic Triggers 25% Increase in Prevalence of Anxiety and Depression Worldwide* (2022)
35. Yamamoto T, Uchiumi C, Suzuki N, Yoshimoto J, Murillo-Rodriguez E. **The Psychological Impact of ‘mild Lockdown’ in Japan during the COVID-19 Pandemic: A Nationwide Survey under a Declared State of Emergency**. *Int J Environ Res Public Health* (2020) **17** 9382-19. DOI: 10.3390/ijerph17249382
36. Lindqvist D, Dhabhar FS, James SJ, Hough CM, Jain FA, Bersani FS. **Oxidative Stress, Inflammation and Treatment Response in Major Depression**. *Psychoneuroendocrinology* (2017) **76** 197-205. DOI: 10.1016/j.psyneuen.2016.11.031
37. Erickson KI, Leckie RL, Weinstein AM. **Physical Activity, Fitness, and gray Matter Volume**. *Neurobiol Aging* (2014) **35** S20S20-88. DOI: 10.1016/j.neurobiolaging.2014.03.034
38. Szuhany KL, Bugatti M, Otto MW. **A Meta-Analytic Review of the Effects of Exercise on Brain-Derived Neurotrophic Factor**. *J Psychiatr Res [Internet]* (2015) **60** 56-64. DOI: 10.1016/j.jpsychires.2014.10.003
39. Liu M, Wu L, Ming Q. **How Does Physical Activity Intervention Improve Self-Esteem and Self-Concept in Children and Adolescents? Evidence from a Meta-Analysis**. *PLoS One* (2015) **10** e0134804-17. DOI: 10.1371/journal.pone.0134804
40. Angevaren M, Aufdemkampe G, Verhaar H, Aleman A, Vanhees L. **Physical Activity and Enhanced Fitness to Improve Cognitive Function in Older People without Known Cognitive Impairment**. *Cochrane Database Syst Rev* (2008). DOI: 10.1002/14651858.cd005381.pub2
41. Quirk SE, Williams LJ, O’Neil A, Pasco JA, Jacka FN, Housden S. **The Association between Diet Quality, Dietary Patterns and Depression in Adults: A Systematic Review**. *BMC Psychiatry* (2013) 13. PMID: 23297686
42. Stubbs B, Koyanagi A, Schuch F, Firth J, Rosenbaum S, Gaughran F. **Physical Activity Levels and Psychosis: A Mediation Analysis of Factors Influencing Physical Activity Target Achievement Among 204 186 People across 46 Low- and Middle-Income Countries**. *Schizophr Bull* (2017) **43** 536-45. DOI: 10.1093/schbul/sbw111
|
---
title: Hypoglycemic and H2O2-induced oxidative injury protective effects and the phytochemical
profiles of the ethyl acetate fraction from Radix Paeoniae Alba
authors:
- Lu Zhang
- Chun-yan Peng
- Pei-xin Wang
- Linju Xu
- Jia-hui Liu
- Xing Xie
- Ling Lu
- Zong-cai Tu
journal: Frontiers in Nutrition
year: 2023
pmcid: PMC9998525
doi: 10.3389/fnut.2023.1126359
license: CC BY 4.0
---
# Hypoglycemic and H2O2-induced oxidative injury protective effects and the phytochemical profiles of the ethyl acetate fraction from Radix Paeoniae Alba
## Abstract
Radix Paeonia Alba (RPA) is often used as food and medicine. This study aimed to enrich and identify the antioxidant and hypoglycemic bioactive compounds from RPA. The results indicated that the ethyl acetate fraction (EAF) showed the highest total phenolic content, DPPH, ABTS+ scavenging ability, and α-glucosidase inhibition ability (IC50 = 7.27 μg/ml). The EAF could alleviate H2O2-induced oxidative stress in HepG2 cells by decreasing the MDA and ROS levels, improving cell apoptosis, increasing the enzyme activity of GPX-Px, CAT, SOD, Na+/K+-ATP, and Ca2+/Mg2+-ATP, and stimulating T-AOC expression, which also enhanced the glucose uptake of insulin-resistant HepG2 cells. In addition, the EAF significantly reduced the fasting blood glucose level and improved glucose tolerance in diabetic mice. An HPLC-QTOF-MS/MS analysis displayed that procyanidin, digallic acid isomer, methyl gallate, tetragalloylglucose isomer, dimethyl gallic acid, and paeoniflorin were the major compounds in the EAF. These findings are meaningful for the application of the EAF in the medicinal or food industry to prevent and treat oxidative stress and diabetes mellitus.
## 1. Introduction
Diabetes mellitus (DM) is a worldwide-prevalent chronic disease inherited from natural or acquired insufficiencies and ineffectiveness of insulin secretion. It can be further divided into type I diabetes, type II diabetes (T2D), gestational diabetes mellitus, and others. Over $90\%$ of the cases of type I diabetes are T2D [1]. The International Diabetes Federation estimated that 537 million adults are living with diabetes in 2021 and that the amount will increase by 74 million compared with 2019, indicating an increment of $16\%$, and it is expected to reach 783 million by 2045 (https://diabetesatlas.org/data/en/country/42/cn.html). T2D is a complex metabolic abnormality characterized by chronic hyperglycemia and impaired pancreatic β-cell function and may negatively influence the structure and function of many organ systems by increasing the risk of cardiovascular disease, heart failure, diabetic kidney disease, diabetic retinopathy, etc. ( 2–4). Currently, mitigating glucose absorption, promoting insulin secretion, alleviating insulin resistance, inhibiting glucagon-like peptide-1 (GLP-1) receptor, sodium-glucose cotransporter-2, and diet control are the predominant modalities of T2D management [5]. Meanwhile, oxidative stress is an important piece for understanding the complex mechanism involved in the development of diabetes and its complications [6, 7].
Reactive oxygen species (ROS), including singlet oxygen (O2), superoxide anion radicals (O2), hydrogen peroxide (H2O2), and hydroxyl radicals (OH), are formed during the respiration process and play an important role in biological functions, such as cell proliferation, apoptosis, and signal transduction in organisms [8]. While excessive ROS will cause impaired glycometabolism in the liver, enhance insulin resistance in the liver and skeletal muscle cells, and reduce the function of pancreatic β-cells, it also promotes the development of diabetes [7]. Antioxidants can alleviate oxidative stress, prevent or delay ROS-triggered apoptosis, and might be a reasonable way to treat diabetes and other metabolic syndromes. Antioxidant enzymes, antioxidants, and proteins that separate transition metals are the main ROS defense systems in an organism. In the antioxidant enzyme protection system, superoxide dismutase (SOD), glutathione peroxidase (GSH-Px), and catalase (CAT) can help to scour free radicals and reduce or eliminate oxidative damage [9, 10]. Antioxidants, natural substances obtained from natural plants and synthetic chemicals that show strong radical scavenging ability, can prevent oxidative injury by removing excessive ROS, decreasing malondialdehyde (MDA), and enhancing the activity of antioxidant enzymes [11]. For example, N-acetylcysteine could improve insulin secretion and insulin signaling by mitigating oxidative stress [6]. However, the application of synthetic antioxidants was limited due to their potential teratogenicity, carcinogenicity, and mutagenicity [12]. Thus, compounds derived from foods or herbs that exhibit low toxicity and high antioxidant ability may reduce oxidative damage by balancing the ROS level in the body.
Radix Paeonia Alba (RPA) is the dry root of herbaceous peony and is widely used in Chinese food due to its rich nutrition and health care function, especially in various stews and soups, such as stewed RPA with pig's feet, stewed RPA with pigeon, paeonia–glycyrrhiza soup, oyster–RPA soup, and papaya–RPA soup. Modern pharmacological studies have shown that RPA can regulate the immune, digestive, and cardiovascular systems [13]. Di et al. [ 14] found that paeoniflorin pretreatment drastically attenuates the ROS level in H2O2-induced Schwann cell injury. The total glucosides of paeonia clearly improved the kidney-related symptoms in diabetic rats [15]. Meanwhile, several bodies of literature have reported that the crude extract or pure compounds of RPA scavenged radicals (DPPH, ABTS+, etc.), inhibited nitric oxide (NO) production, and prevented diabetes-associated renal damage [16]. However, the in vitro and in vivo antioxidant and hypoglycemic abilities of an RPA extract and its main active compounds still need to be researched further.
In this study, RPA was extracted and fractionated with different solvents, and the total phenolic content (TPC), radical scavenging, and α-glucosidase inhibition activities were evaluated to screen the fraction with the strongest hypoglycemic and antioxidant activities. Then, HepG2 cell models with oxidative injury induced by H2O2 and insulin resistance were applied to investigate the effect of RPA and its fractions on oxidative stress and glucose absorption, respectively. The in vivo hypoglycemic and oral glucose tolerance test (OGTT) abilities were evaluated with db/db mice. Finally, the main chemical composition of the RPA fraction with the strongest activity was identified by high-performance liquid chromatography-tandem quadrupole time-of-flight mass spectrometry (HPLC-ESI-QTOF-MS/MS).
## 2.1. Materials and chemicals
Dried Radix Alba Paeoniae was purchased from Anqing Chunyuan pharmacy in Anqing city, Anhui province, China on June 2021. Analytic-grade chloroform, ethyl acetate, and n-butanol and chromatographic-grade acetonitrile and formic acid were from Aladdin Reagent Int. ( Shanghai, China). Metformin, acarbose, 1,1-diphenyl-2-picrylhydrazyl (DPPH), 2,2-azinobis-(3-ethylbenzothiazoline-6-sulfonic acid) (ABTS), α-glucosidase, and p-nitrophenyl-α-D-galactopyranoside (pNPG) were purchased from Sigma-Aldrich (St. Louis, MO, USA). Human hepatocellular carcinoma cells (HepG2) and culture media were purchased from the BeNa Culture Collection (Beijing, China). All other chemicals were of analytical grade and from Sinopharm Chemical Reagent Co., Ltd. (Shanghai, China).
## 2.2. Preparation of samples
Dried RPA was pulverized into powder by a disintegrator and soaked for 24 h in $70\%$ ethanol solution (m/v, 1:20) at room temperature. After 7 days, the mixtures were filtered and the residues were extracted for two times under the same extraction conditions. The supernatants obtained through three times of extraction were combined and concentrated to yield the crude extract. Finally, the extracts were dissolved in distilled water and fractionated by chloroform, ethyl acetate, and n-butanol sequentially to yield the chloroform fraction (DCF), the ethyl acetate fraction (EAF), and the n-butanol fraction (nBuF) for further analyses.
## 2.3. Determination of the total phenolic content
The TPC was quantified by the Folin-Ciocalteau method [17] with some modifications. Briefly, 200 μl of properly diluted samples or standard were mixed with 100 μl of Folin-Ciocalteau reagent. After 5 min, 300 μl of Na2CO3 ($7.5\%$, w/v) and 1.0 ml of water were added. After 30 min, the absorbance at 765 nm was measured using a microplate reader (Biotek, Vermont, USA). The TPC was calculated based on the calibration curve plotted using gallic acid (0–200 μg/ml) and the results were expressed as μg of gallic acid equivalents per gram of extract [(μg GAE)/g E].
## 2.4. Determination of DPPH·and ABTS+ scavenging ability
The DPPH·and ABTS+ scavenging activity were determined according to the methods reported in our previous study [2]. Freshly prepared DPPH·or ABTS+ working solution (150 μl) was mixed with 50 μl of the sample at various concentrations in 96-well microplates. After 30 or 6 min of incubation at room temperature, the absorbance was measured at 517 or 734 nm. Ethanol and quercetin were used as the negative control and positive control, respectively. The DPPH·and ABTS+ scavenging activity were expressed as the IC50 value, which was calculated by a nonlinear curve fitting of percentage inhibition ratio vs. sample concentration (μg/ml) using Origin 2019 (OriginLab Co., US).
## 2.5.1. Cell cultivation and cell viability assay
HepG2 cells were cultured in Dulbecco's Modified Eagle's Medium (DMEM) supplemented with $10\%$ fetal bovine serum (FBS), 100 U/ml penicillin, and 100 μg/ml streptomycin. The cells were grown at 37°C under a humidified $5\%$ CO2 atmosphere. The CCK-8 method was used to determine cell viability [18]. The HepG2 cells were seeded in 96-well plates at a density of 1.0 × 104 cells/well and incubated for 24 h. Then, the cells were exposed to 10–300 μg/ml of the EAF or PBS for 24 h, and the cell viability was measured by a CCK-8 kit.
For oxidative damage protection analysis, the blank group, the control group, and the EAF group were set after 24 h of adherent incubation. The EAF and control groups were treated with 10–300 μg/ml of EAF and serum-free DMEM for 24 h, followed by the exposure of H2O2 for another 4 h. The blank group was treated with serum-free DMEM for 28 h. Finally, the cell viability was measured by a CCK-8 kit.
## 2.5.2. Intracellular ROS level assay
The intracellular ROS level was determined using 2′7′-dichlorofluorescein diacetate (DCFH-DA) staining and flow cytometry [14]. The HepG2 cells were seeded in 6-well plates (1 × 105 cells/well), cultured in a humidified incubator with $5\%$ CO2 at 37°C for 24 h, and treated with different concentrations of EAF and 300 μM of H2O2 in sequence. The cells were then stained with 10 μM of DCFH-DA in serum-free DMEM at 37°C for 30 min in darkness and washed two times with PBS and followed by digestion with trypsin and centrifugation at 1,000 rpm for 5 min. The fluorescent intensity was measured by an Accuri C6 flow cytometry (Becton, Dickinson and Company, USA) at an excitation wavelength of 488 nm, and the intracellular redox status was calculated based on the ratio of green vs. yellow.
## 2.5.3. Measurement of MDA level and enzyme activity
HepG2 cells at the log phase were prepared as single-cell suspensions and seeded into 6-well plates (1 × 106 cells/well) at 37°C for 24 h. After treatment with the EAF for 24 h, the cells were incubated with H2O2 for 4 h, washed two times with a PBS solution and lysed in lysis buffer (Biyuntian Biotechnology Co., Ltd, Shanghai, China). A BCA protein assay kit (Jiancheng Bioengineering Institute, Nanjing, China) was used to measure the intracellular protein content. The malondialdehyde (MDA) content, total antioxidant ability (T-AOC), catalase (CAT), glutathione peroxidase (GSH-Px), superoxide dismutase (SOD), Na+/K+-adenosine triphosphate (Na+/K+-ATP) and Ca2+/Mg2+-adenosine triphosphate (Ca2+/Mg2+-ATP) enzyme activity were determined with the corresponding assay kits according to the manufacturer's instructions (Jiancheng Bioengineering Institute, Nanjing, China).
## 2.5.4. Detection of HepG2 apoptosis
Cell apoptosis was measured by AnnexinV FITC/PI staining and flow cytometry [14]. HepG2 cells were cultured as described in Section 2.5.1. After digestion with trypsin containing EDTA (Solarbio, China), the cells were centrifuged at 1,000 g for 5 min at 4°C and washed two times with precooled PBS solution. Then, 1.0 × 106 cells were collected and centrifuged at 1,000 g for 5 min and allowed to react with 5.0 μl of Annexin V-FITC at 4°C for 15 min in darkness, followed by incubation with 10 μl of PI staining solution. After incubation at 4°C for 5 min, the early, viable, late, and apoptotic cells were collected, and the percentage ratio was analyzed with an Accuri C6 flow cytometry. The x and y coordinates refer to the fluorescence intensities of annexin V and PI, respectively.
## 2.6. α-glucosidase inhibition assay
The inhibition activity of α-glucosidase was assessed by the previous method [18]. Properly diluted samples (50 μl) were incubated with 50 μl of the α-glucosidase solution (0.2 U/ml) at 37°C for 10 min prior to the addition of 50 μl pNPG. After 50 min, 100 μl of the 0.2 M Na2CO3 solution was added to stop the reaction, absorbance was measured at 405 nm against a blank without α-glucosidase, and the system without sample was run in parallel as the control group. Acarbose was used as the positive control, and the IC50 value was used to evaluate the inhibition ability.
## 2.7. Determination of glucose consumption by IR-HepG2 cells
An insulin resistance (IR) model was induced as indicated in a previous study [3]. HepG2 cells were seeded in 96-well plates at 2 × 105 cells/well (100 μl/well) and cultured for 24 h. The cells were washed with PBS three times, treated with serum-free high-glucose DMEM (4.5 g/L) containing 1.0 mM insulin, and incubated at 37°C under $5\%$ CO2 atmosphere for 36 h to induce the IR group. The cells cultured in high-glucose DMEM supplemented with $10\%$ FBS and $1\%$ antibiotic antimycotic solution were taken as the negative control group (Con). Then, the EAF solutions were added to the IR model cells for 24 h. The glucose content in the culture was detected using a glucose assay kit according to the manufacturer's instructions (Jiancheng Bioengineering Institute, Nanjing, China). The glucose consumption was calculated as follows: Glucose consumption = Cm–Cs, where Cm and Cs (mmol/L) refer to the glucose content in cells before and after 24 h of incubation, respectively.
## 2.8.1. Animal experiments
A total of 508-week-old male C57BL6/J mice, including 40 db/db and 10 db/m mice, were purchased from Changzhou Cavens Laboratory Animal Co., Ltd [license key: SCXK (Su) 2016-0010, Jiangsu, China]. All experimental protocols in this research were approved by the Committee on the Ethics of Animal Experiments of the Jiangxi Normal University (Permission Number: JNU20210311-001). The breeding environment of mice was strictly controlled (25 ± 2°C, relative humidity 50 ± $5\%$, and $\frac{12}{12}$ h diurnal cycle). After 1 week of adaptive feed, the 40 db/db mice were randomly divided into 5 groups: the diabetes group (Mod), the metformin group (250 mg/kg body weight per day, Met250), the low-dose group (100 mg/kg body weight per day, EAF100), the middle-dose group (250 mg/kg body weight per day, EAF250), and the high-dose group (400 mg/kg body weight per day, EAF400) and fed with D12451 during the period of drug administration. The non-diabetic mice (db/m) fed with D12450B only were used as the control group (Con). The drugs were dissolved in $5\%$ Macrogol 400 solution; the Con and Mod groups were gavaged with a corresponding dose of $5\%$ Macrogol 400 solution per day. The body weight and fast blood glucose (FBG) of all mice were measured weekly during the continuous administration for 8 weeks.
## 2.8.2. Oral glucose tolerance test
The OGTT was conducted as described in a previous study [19]. Briefly, the mice fasted for 8 h after 8 weeks of intervention, and all mice were orally administered with 2 g/kg glucose solution. The FBG of mice was measured at 0, 30, 60, 90, 120, and 180 min by using a blood glucose monitor (ACCU-CHEK Performa, India). The OGTT values were calculated based on the curve areas.
## 2.9. HPLC-QTOF-MS/MS analysis
The separation and identification of compounds were performed using an HPLC-QTOF-MS/MS system [20]. Briefly, the samples were separated by a YMC-Triart C18 column (4.6 × 250 mm, 5 μm, Japan) at a flow rate of 0.8 ml/min. Formic acid water (A) and acetonitrile (B) were used as the mobile phase. The elution program was set as follows: 0 min, $5\%$ B; 6 min, $9\%$ B; 7 min, $18\%$ B; 38 min, and $40\%$ B. The elutes were interfered with a Hybrid Quadrupole-TOF Mass Spectrometer Triple TOF 5600+ system (SCIEX, USA) directly for mass identification. The MS and MS/MS data were acquired under the negative ion mode with an electrospray ionization resource. The full-scan mass spectrum was detected at a mass range of m/z 100–1,500 and 50–1,500 in the MS and MS/MS models.
## 2.10. Statistical analysis
Statistical analyses were carried out using the SPSS 22.0 software (IBM, Armonk, NY, USA). The plots were drawn with Origin 2019 (OriginLab, Northampton, MA, USA), and all data were expressed as mean ± SD (standard deviation). Significant difference among data was calculated by Tukey's-b and one-way analysis of variance (ANOVA). A P-value of < 0.05 was considered significant.
## 3.1. Total phenolic content
The TPC of the crude extract and its fractions from RPA are listed in Table 1. The EAF showed the highest TPC, followed by the nBuF and DCF; the contents were 844.83, 210.11, and 106.22 mg GAE/g E, respectively, and the lowest TPC was found in the crude extract (26.70 mg GAE/g E). The results indicated that ethyl acetate had the best-enriching efficiency for the phenolics in RPA extract. Our previous research also found that ethyl acetate possessed better enriching efficiency for the polyphenols in *Acer palmatum* “Atropurpureum,” *Acer palmatum* Thunb [2], and Ipomoea batatas leaves [17] than n-butanol and chloroform. In addition, the TPC was much higher than that reported by Wang et al. [ 13] with a value of 58.60 μg/mg E, which could be explained by the difference in extraction solvents. Water is not a proper solvent to recover polyphenols due to their high polarity.
**Table 1**
| Samples | Total phenolics (mg GAE/g Fr.) | DPPH·(IC50, μg/ml) | ABTS+ (IC50, μg/ml) | α-glucosidase (μg/ml) |
| --- | --- | --- | --- | --- |
| Crude extracts | 26.70 ± 0.78d | 184.5 ± 0.71c | 132.12 ± 10.35b | 69.09 ± 2.04b |
| DCMF | 106.22 ± 11.86c | 420.80 ± 4.18a | 173.17 ± 4.83a | 21.35 ± 0.61d |
| EAF | 844.83 ± 29.75a | 19.57 ± 1.95e | 7.82 ± 1.03d | 7.27 ± 1.82e |
| nBuF | 210.11 ± 6.45b | 206.64 ± 3.34b | 52.66 ± 1.11c. | 29.63 ± 1.58c |
| Standards | – | 27.35 ± 0.33d | 10.04 ± 1.51d | 197.01 ± 6.42a |
## 3.2. In vitro free radical scavenging abilities
In in vitro antioxidant models, DPPH and ABTS+ scavenging abilities were used to assess the antioxidant activity of RPA fractions. As shown in Table 1, all fractions were able to scavenge DPPH and ABTS+. However, the EAF exhibited the strongest DPPH and ABTS+ scavenging abilities with IC50 values of 19.57 and 7.82 μg/ml, respectively. The DPPH scavenging ability was higher than that of the positive control quercetin, and the ABTS+ scavenging ability was comparable with quercetin, suggesting the excellent antioxidant potential of the EAF. The DPPH and ABTS+ scavenging abilities of nBuF and DCF were much lower than those of quercetin, especially for the DCF fraction, as the IC50 values were 420.80 and 173.17 μg/ml, respectively. Meanwhile, the DPPH scavenging abilities of DCF and nBuF were lower than that of the crude extract, while their ABTS+ scavenging ability was better than crude extracts. Considerable radical scavenging ability was also found by You et al. [ 21] by the $50\%$ aqueous methanol extracts of raw and processed RPA. Therefore, the EAF was selected for further cell assays to evaluate oxidative injury protection.
## 3.3.1. Cell viability
H2O2 is one of the commonly used substances for establishing oxidative damage in cells as it can penetrate cell membranes and cause cell damage and has been used in the osteoblasts, the nerve cells, the vascular endothelial cells, and the hepatocytes [9]. Therefore, the H2O2-induced HepG2 cells oxidative damage model was used to further determine the oxidative protection effect of the EAF. The cytotoxicity of EAF toward natural HepG2 cells was presented in Figure 1A; with the increase in EAF incubation concentrations, the cell survival rate decreased gradually. Only $74.83\%$ cell viability was achieved after treatment with 300 μg/ml of the EAF, and the viability at 200 μg/ml was $87.51\%$. Thus, 25–250 μg/ml was used for further experiments. The H2O2 damage degree described as cell viability is listed in Figure 1B, and dose-dependent cytotoxicity was observed. The cell viability was reduced from 97.99 to $59.71\%$ after incubating with 50–300 μM of H2O2 for 4 h. Therefore, 300 μM of H2O2 was selected to damage HepG2 cells. To investigate the oxidative injury protection of the EAF, HepG2 cells were exposed to 25–250 μg/ml of the EAF before being stimulated with 300 μM of H2O2. It was clear that the EAF could significantly increase the cell viability ($P \leq 0.05$) and exhibit a dose-dependent relationship (Figure 1B). The cell viability reached up to $85.63\%$ when 250 μg/ml of the EAF was added. The above results suggested that the EAF can protect HepG2 cells from the damage induced by H2O2.
**Figure 1:** *The cell viability of HepG2 cells in the presence or absence of the EAF (A). The cell viabilzity of H2O2-induced HepG2 cells with or without the EAF (B). The annotation ** and *** indicate the P-value of 0.05 and 0.01 compared to the Con group or the Mod group, respectively.*
## 3.3.2. Effects on the levels of MDA, ROS, and antioxidant enzymes
The MDA and ROS levels in liver cells are generally considered to be essential indicators of peroxidation and antioxidative defenses [14]. Therefore, the effect of the EAF on the production of MDA and ROS in HepG2 cells was measured. As shown in Figures 2A, B, compared with the Con group, the H2O2 treatment significantly increased the ROS and MDA levels in HepG2 cells ($P \leq 0.05$), and they were clearly decreased when the cells were pretreated with the EAF. The ROS level was reduced from 11.37 to $2.72\%$ when 100 μg/ml of the EAF was added. Meanwhile, the MDA level decreased from 53.03 to 29.44 nmol/ml when the concentration of the EAF was increased from 0 to 250 μg/ml. However, no obvious difference was observed between the lower dose (25 μg/ml) and the Mod dose EAF groups ($P \leq 0.05$). Therefore, the EAF could reduce the MDA and ROS levels and possess benefits for HepG2 cells against oxidative stress.
**Figure 2:** *The MDA, ROS, CAT, SOD, GSH-Px, T-AOC, Na+/K+-ATP, and Ca2+/Mg2+-ATP levels in H2O2-induced HepG2 cells treated with different concentrations of the EAF (A–H). Con: free of H2O2 treatment; Mod: induced by 300 μM H2O2; and EAF: induced by 300 μM H2O2 and different doses of the EAF. The annotation *** indicates a P-value of < 0.01 vs. the Mod group.*
The GPX-Px, CAT, and SOD enzymes play a pivotal role in protecting cells from free radical damage, and the activation of the enzymes may help to scavenge active radicals and protect cells from high oxidative stress. The T-AOC reflects the total antioxidant levels of the enzymatic and non-enzymatic systems in the body, which are responsible for maintaining health [12]. The effects of the EAF on the activities of GPX-Px, CAT, and SOD on H2O2-induced HepG2 cells are given in Figures 2C–E, and the activities of the antioxidant enzymes were remarkably reduced in the Mod group compared with the Con group. As expected, the EAF significantly increased the activities of GSH-Px, CAT, and SOD ($P \leq 0.05$). In particular, when the concentration of the EAF was 250 μg/ml, the expression levels of GPX-Px, CAT, and SOD were enhanced 2.07, 4.01, and 1.21 times, respectively ($P \leq 0.05$). The T-AOC was enhanced from 1.80 to 5.50 U/ml (illustrated in Figure 2F). Therefore, the EAF could regulate H2O2-induced oxidative stress by stimulating the activities of CAT, SOD, and GSH-Px and upregulating the T-AOC level. Ming and Dong [22] also found that intragastric administration of the RPA powder at doses of ~2–8 g/kg body weight/day could significantly reduce the serum MDA content and enhance SOD activity in Wistar rats.
## 3.3.3. Effect on the activities of the energy metabolism enzymes
The ATPases of Na+/K+-ATP and Ca2+/Mg2+-ATP participate in the growth and reproduction of organisms, play an important role in signal transmission and energy metabolism, and even provide the necessary nutrients for cells [23]. To further investigate the effect of the EAF on cell energy metabolism, the protective effects of the EAF on the Na+/K+-ATP and Ca2+/Mg2+-ATP enzymes were analyzed, and the results are shown in Figures 2G, H. The highest Na+/K+-ATP and Ca2+/Mg2+-ATP enzyme activities were found in the Con group with the values of 1.26 and 2.38 U/mg/protein, respectively. Induction with 300 μM of H2O2 could significantly reduce the activity of these two ATPases. Howver, the enzyme activity was individually increased by 1.44 and 1.94 times in the HepG2 cells pretreated with 250 μg/ml of the EAF prior to H2O2 damage. The enhanced Na+/K+-ATP and Ca2+/Mg2+-ATP enzyme activities were observed in oxidation-damaged HepG2 cells treated with *Capparis spinosa* L. [23]. The results suggested that improved Na+/K+-ATP and Ca2+/Mg2+-ATP enzyme activity might contribute to the protection of EAF against H2O2 induced HepG2 cell oxidative stress.
## 3.3.4. Inhibition rate of H2O2-induced HepG2 cell apoptosis
Oxidative stress and cell apoptosis were known to have a close relationship, and the apoptosis of cells was seen as the end result of oxidative damage [14]. From Figure 3, the apoptosis percentage detected in HepG2 cells cultured with H2O2 was $29.7\%$, which was much higher than that of natural HepG2 cells (Con group) ($P \leq 0.05$). Pretreatment with 50, 100, and 250 μg/ml of the EAF significantly attenuated the apoptosis percentage, and the values reduced to 24.8, 19.5, and $9.1\%$, respectively. The data indicated that the EAF could improve the apoptosis of H2O2-induced oxidative damage HepG2 cells and alleviate the oxidative stress of HepG2 cells.
**Figure 3:** *Annexin V-FITC/PI staining by flow cytometry. Representative image of fluorescence-activated cell sorting analysis (A). Quantitative analysis of apoptotic HepG2 cells (B). The annotation ** and *** indicate the P-value of 0.05 and 0.01 compared with the Con group or the Mod group, respectively.*
## 3.4. α-glucosidase inhibition ability
α-glucosidase is an important carbohydrate hydrolase, and its inhibition activity can effectively retard the hydrolysis of carbohydrates and reduce the intestinal absorption of glucose. The α-glucosidase inhibitors are usually considered a promising approach to decrease fasting and postprandial blood glucose, and this approach was considered to prevent and treat diabetes [17]. To investigate the potential capability of RPA extracts for diabetes treatment, the α-glucosidase inhibitory activity of RPA extracts was evaluated. As shown in Table 1, the RPA extract and its fractions (DCMF, EAF, and nBuF) were effective, and the detected IC50 values were 69.09, 21.35, 7.27, and 29.63 μg/ml, respectively. In addition, the EAF exhibited the highest α-glucosidase inhibitory when compared with other fractions and acarbose (IC50 = 197.01 μg/ml). The α-glucosidase inhibitory capabilities of DCMF, EAF, and nBuF were enhanced by 3.24, 9.50, and 2.33 times than that of crude extract, respectively. The α-glucosidase inhibition activity of RPA extracts and its fractions was highly correlated with the TPC and antioxidant activity (R2 = 0.76, 0.84), suggesting that phenolics and antioxidants in RPA contributed much to its α-glucosidase inhibition. Thus, based on these results, the EAF fraction was selected to further evaluate the effect on glucose uptake in IR-HepG2 cell models.
## 3.5. Promote glucose uptake in IR-HepG2 cells
HepG2 cells are widely used in biochemical and nutritional studies as they can retain the morphology and functions in culture and still are suitable cell models to investigate IR [24]. In this study, an IR-HepG2 model was used to estimate the ability of the EAF to modulate glucose uptake in vitro, and the results are shown in Figure 4A. After pretreatment with the EAF, the glucose uptake by IR-HepG2 cells was remarkably increased in comparison with the model group, which enhanced 0.78 mM when 100 μg/ml of the EAF was added ($P \leq 0.05$). However, the glucose uptake values of all other IR-HepG2 cell groups were lower than the Con group. Therefore, the EAF may treat T2D by increasing the glucose uptake in liver cells and hence the hypoglycemic effect in vivo is worth further evaluation.
**Figure 4:** *The effect of the EAF fraction on the glucose uptake of IR-HepG2 cells (A), FBG (B), OGTT (C), and the area of AUC (D) in T2D mice. The different letters (a–d) in each column represent the significant difference (P < 0.05). **P < 0.05.*
## 3.6. Effects of the EAF on the FBG level and OGTT in T2D mice
As shown in Figure 4B, there were no significant differences in the initial FBG levels among the T2D mice in all groups. After oral administration for 8 weeks, the FBG level of the Mod group was significantly increased, while those of the Met and EAF groups were decreased and showed a dosage-dependent effect. In the last week, the FBG levels were reduced from 29.28 to 13.86, 19.98, 18.35, and 14.76 mmol/L in the Met 250, EAF100, 250, and 400 groups, respectively. The results suggested that the EAF pretreatment could improve the FBG levels of T2D mice.
An OGTT was often used to evaluate the abilities of the samples to regulate glucose metabolism in vivo [25]. As given in Figures 4C, D, the blood glucose (BG) levels of all mice reached the peak at 30 min and then declined until 180 min, while the Mod group possessed clearly abnormal glucose tolerance. Compared with the Mod group, the AUC values of the Met 250, EAF100, EAF250, and EAF400 groups were reduced by ~27.46, 5.98, 9.18, and $44.46\%$, respectively. There was no significant difference between the EAF100 and EAF250 groups ($P \leq 0.05$). The results showed that supplementation of the EAF could enhance the glucose tolerance of the T2D mice, and the effect of the high dose of the EAF fraction (400 mg/kg body weight per day) was comparable with that of Met250.
## 3.7. Identification of phytochemical profiling
In this study, the HPLC-QTOF-MS/MS technology was performed to investigate the active compounds in the EAF. The compounds were identified by comparing the retention time, molecular formula, and MS/MS information with the references and database. The TIC spectrum of the EAF is displayed in Figure 5, and the MS/MS information is given in Table 2. A total of 23 compounds were identified in the EAF, including 5 flavonoids, 2 phenolic acids, 9 tannins, 3 terpenoids, and 4 other compounds.
**Figure 5:** *The total ion chromatogram (TIC) for the compounds in the EAF.* TABLE_PLACEHOLDER:Table 2
## 3.7.1. Flavonoids
Peak 5 was assigned as the procyanidin, and the MS/MS ion at 289 implied the existence of the epicatechin residue [20]. Peak 7 with MS/MS ions at 271, 245, 205, 179, and 125 accounted for the characterization of catechin [26]. Peak 13 with MS/MS ions at 169, 241, and 317 displayed the loss of a galloyl moiety and myricetin, which was proposed as galloylmyricetin [26]. Peak 21 was suggested as glochiflavanoside B due to the same fragment ions in the literature [27]. Peak 30 was identified as naringenin based on the fragment ions at 151 ([M–C6H5O–CO]−) [28].
## 3.7.2. Phenolic acids
Peaks 1 and 4 were identified as dihydrocaffeic acid and gallic acid by comparing them with the standards.
## 3.7.3. Tannins
Gallic acid derivatives and gallotannins contain one or more galloyl moieties and showed the characteristic fragment ions at 169.0133 ([gallic acid-H]−) and 125.0232 ([gallic acid-CO2-H]−) [20]. Peak 6 with the MS/MS ion at 169 was suggested as digallic acid, which was generated by the dehydration condensation of two gallic acids. Peak 8 produced the fragment ion at 168 ([M–OH]−) and 124 ([M–CO2-OH]−) and was assigned as methyl gallate. Peak 12 was suggested as dimethyl gallic acid resulting from the MS/MS ions at 169 and 172 [26]. Peak 11 was identified as tetragalloylglucose, with the MS/MS ions at 617 and 169 indicating the presence of 1 and 3 galloyl moieties [29]. Peaks 14, 16, 17, and 19 showed the same MS ion and molecular formula and were tentatively identified as galloylpaeoniflorin and its isomers. The MS/MS ion at 313 implied the existence of galloyl glucose, which produced an ion at 169 by losing a glucosyl residue [29]. The molecular weight of peak 18 was 152 Da ([gallic acid-OH–H]−), which is higher than peak 8 and was identified as methyl digallate. Similarly, peaks 22 and 25 were considered as isomers of dimethyl digallate, which were methyl-substituted compounds of peak 18 [2]. Peaks 28 and 29 with MS/MS ions at 535, 431, and 121 were proposed as mudanpioside B isomers and have been found in the genus Paeonia [30].
## 3.7.4. Terpenoids
Peaks 9, 10, 15, and 20 showed common MS ions at 479, which were identified as isomers of paeoniflorin [29]. The MS/MS ions at 283 and 121 correspond to the loss of benzoic acid and glucose residues. Peak 27 was identified as mascaroside based on the MS/MS ions at 431, 375, 195, and 165 according to the literature [31].
## 3.7.5. Others
Peak 2 was identified as sucrose using the standard. Peak 3 was proposed as citric acid, and the MS/MS ion at 129 indicated the loss of the carboxyl and hydroxyl groups [31]. Peak 23 with MS/MS ion at 229 indicated the presence of one phenol and three hydroxyl groups and was suggested as 6-(3,4-dihydroxybenzyl)-5,7-dihydroxy-2-(4-hydroxyphenyl)-4H-1-benzopyran-4-one [32]. Peak 26 was assigned as pinen-10-Yl vicianoside by matching MS/MS ions with the report by Nöst et al. [ 33].
## 4. Discussion
Many studies found that long-term drug therapies for patients with T2D would result in various side effects and generate drug resistance, while natural plant extracts have been demonstrated as alternative therapeutic agents or supplements to treat T2D by alleviating oxidative damage and hyperglycemia [34, 35]. Currently, the exploration of novel antidiabetic drugs or dietary supplements from natural products is a hotspot in the research field of T2D. RPA, as a medicinal plant and has been reported to show various biological activities like antioxidant and hypoglycemic effects, but the main active compounds are still not clear. Therefore, the aim of this study was to investigate the phytochemical composition and the antioxidant and hypoglycemic activities of RPA in vitro and in vivo.
In this study, compared with other fractions, the EAF of RPA showed the highest total phenolic content, DPPH and ABTS+ scavenging capacities, and α-glucosidase inhibition ability, which were 31.6, 9.5, 16.9, and 9.5 times higher than the crude extracts, respectively. The Pearson correlation coefficients of the TPC with α-glucosidase inhibitory and DPPH and ABTS+ scavenging activities were −0.759, −0.753, and −0.816, which indicated that phenolics may be the main antioxidants and α-glucosidase inhibitors in RPA. A previous study also found that ethyl acetate displayed a stronger enrichment effect on phenolics in two *Acer palmatum* cultivars than n-butanol and water, which could be explained by the similar polarities [2]. You et al. [ 21] reported that the DPPH scavenging capacity of $50\%$ aqueous methanol extract from RPA was lower than that of our study with an IC50 value of 310 μg/ml. Simultaneously, HPLC-QTOF-MS/MS analysis suggested that flavonoids, tannins, and terpenoids were the main compounds in RPA, which was in accordance with the report of Xiong et al. [ 29]. Galloylmyricetin, catechin, paeoniflorin, and gallic acid and its derivatives have been demonstrated to exhibit excellent inhibition abilities on free radicals and α-glucosidase and may have contributed to the biological activities of the EAF in vitro [3, 19, 24].
Furthermore, the effect of the EAF on oxidative damage was determined by an H2O2-induced HepG2 cell oxidative damage model. After pretreatment with the EAF, the levels of MDA and ROS in HepG2 cells remarkably decreased, and the activities of SOD, CAT, GPX-Px, T-AOC, Na+/K+-ATP, and Ca2+/Mg2+-ATP obviously increased. In addition, the addition of the EAF inhibited the apoptosis of HepG2 cells induced by H2O2. The results confirmed that the EAF could alleviate oxidative stress damage in vitro. Phenolics like tannins and terpenoids have been proven to prevent oxidative damage by improving antioxidant enzymes [36], which were abundant in the EAF. Moreover, flavonoids like procyanidin could improve energy metabolism by increasing ATP synthesis [37]. As studied by Yuan et al. [ 9], paeoniflorin suppressed oxidative stress by enhancing the SOD and CAT levels in H2O2-induced HepG2 cells, and its derivatives were found in high contents in the EAF. Hydrolyzable tannins, as dominant ingredients in the EAF, have been reported to show a modulation effect on antioxidant enzyme levels and the activities of Na+/K+- and Mg2+-ATP in erythrocyte membranes [38]. It has been found that catechin could improve the antioxidant enzyme (CAT and GSH-Px) contents in vitro and in vivo and protect HepG2 cells from apoptosis caused by H2O2 [39]. Crispo et al. [ 10] also proved that methyl gallate reduced H2O2-induced apoptosis percentage in PC12 cells. In addition, oxidative stress was highly associated with insulin resistance (IR), which may influence glucose metabolism in cells. The result indicated that the EAF significantly enhanced the glucose uptake of IR-HepG2 cells. Similarly, pretreatment with procyanidin from grape seeds significantly increased glucose consumption in HepG2 cells from 36.78 to 55.92 μmol/mg cell protein [3]. Many studies have also demonstrated that catechin and gallic acid and its derivatives treatment promoted glucose consumption in IR-HepG2 cells, and paeoniflorin treatment could alleviate IR in HepG2 cells [18]. Therefore, the identified bioactive compounds in the EAF greatly contributed to improving H2O2-induced HepG2 cell oxidative injury and increasing glucose uptake in IR-HepG2 cells.
Db/db mice were used as a model to further investigate the hypoglycemic effect of the EAF in vivo. After consumption of the EAF for 8 weeks, the FBG level of db/db mice declined sharply by comparison with the Mod group and showed a dosage effect. Liu et al. [ 40] confirmed that the ethanolic extract from peony seeds led to a decrease in the glucose level in high-fat diet-induced T2D mice and suggested the same phenomenon in our research. The OGTT results exhibited that the AUC value of the EAF400 group dropped and was $44.46\%$ lower than that of the Mod group, which revealed that the oral administration of the EAF could regulate insulin sensitivity and glucose metabolism of T2D mice, corresponding to our result in vitro. Flavonoids, phenolic acids, tannins, and terpenoids have been proven to show excellent effects in the management of T2D in various ways [5, 36]. The phenolics and terpenoids from the EAF displayed good antioxidant and α-glucosidase inhibition activities, which might effectively reduce the absorption of glucose and improve IR and lower the FBG level in diabetic mice. As reported, the catechin-enriched extract treatment reversed the FBG level in db/db mice [19]. Compared with the Mod group, the FBG level and the AUC value of 30 mg/kg paeoniflorin group decreased to 3.71 mmol/L and 10.19 after treatment for 4 weeks, and 4 paeoniflorin isomers were identified in the EAF. Based on the above results, it can be concluded that the EAF had a strong hypoglycemic effect in vitro and in vivo.
## 5. Conclusion
The EAF, as the best bioactive fraction from RPA extracts, could effectively scavenge free radicals, protect oxidative stress injury, enhance glucose uptake, and decrease hyperglycemia, which were attributed to the higher TPC (844.83 mgGAE/g) and components like flavonoids, phenolic acids, tannins, and terpenoids. The ability of the EAF to scavenge DPPH and ABTS+ and suppress α-glucosidase was higher than that of the corresponding standard (Vc and acarbose). The EAF treatment alleviated oxidative damage in HepG2 cells caused by H2O2 by increasing the SOD and CAT enzyme levels and reducing the production of MDA and ROS. In addition, the Na+/K+-ATP and Ca2+/Mg2+-ATP enzyme levels and cell apoptosis were also improved by the EAF. When the concentration reached 50 and 100 μg/ml, the EAF effectively increased the glucose uptake of IR-HepG2 cells. Here, a strong ability to decelerate FBG levels (from 29.98 to 14.76 mmol/L) in db/db mice was found at the gavage dose of the EAF 400 mg/kg/day. Therefore, the present study reveals that EAF has a great potential for development as a natural drug or a dietary supplement for the treatment of T2D.
## Data availability statement
The original contributions presented in the study are included in the article/supplementary material, further inquiries can be directed to the corresponding authors.
## Ethics statement
All animal procedures were approved by the Institutional Animal Care and Use Committee at Hunter Biotechnology, Inc. [Approval number: IACUC-2020-2574-01, Use license number: SYXK (zhe) 2022-0004]. The feeding and management were accredited by the Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC) International (No. 001458).
## Author contributions
LZ: methodology, supervision, writing—review and editing, funding acquisition, and project administration. C-yP: investigation, validation, formal analysis, and writing. P-xW: methodology, investigation, and original draft. LX: software, investigation, and statistical analysis. J-hL: investigation and validation. XX: investigation and writing—review and editing. LL: investigation and methodology. Z-cT: conceptualization, supervision, and funding acquisition. All authors contributed to the article and approved the submitted version.
## Conflict of interest
LZ and LL were employed by Jiangxi Deshang Pharmaceutical Co., Ltd., Yichun, Jiangxi, China. The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher's note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## References
1. Alam S, Hasan MK, Neaz S, Hussain N, Hossain MF, Rahman T. **Diabetes mellitus: insights from epidemiology, biochemistry, risk factors, diagnosis, complications and comprehensive management**. *Diabetology.* (2021) **2** 36-50. DOI: 10.3390/diabetology2020004
2. Zhang L, Tu ZC, Xie X, Lu Y, Wang ZX, Wang H. **Antihyperglycemic, antioxidant activities of two**. *Ind Crop Prod* (2016) **89** 522-32. DOI: 10.1016/j.indcrop.2016.06.005
3. Zhang HJ, Ji BP, Chen G, Zhou F, Luo YC, Yu HQ. **A combination of grape seed-derived procyanidins and gypenosides alleviates insulin resistance in mice and HepG2 cells**. *J Food Sci.* (2009) **74** 1-7. DOI: 10.1111/j.1750-3841.2008.00976.x
4. Huang ZR, Huang QZ, Chen KW, Huang ZF, Liu Y, Jia RB. *Front Nutr.* (2022) **9** 1013466. DOI: 10.3389/fnut.2022.1013466
5. Egbuna C, Awuchi CG, Kushwaha G, Rudrapal M, Patrick KC, Singh O. **Bioactive compounds effective against type 2 diabetes mellitus: a systematic review**. *Curr Top Med Chem.* (2021) **21** 1067-95. DOI: 10.2174/18734294MTE1ENjAgx
6. Argaev FL, Rosenzweig T. **Complexity of NAC action as an antidiabetic agent: opposing effects of oxidative and reductive stress on insulin secretion and insulin signaling**. *Int J Mol Sci.* (2022) **23** 2965-87. DOI: 10.3390/ijms23062965
7. Carlos PA, Javier RG, Pablo P, Marcelo SM, Bastián VG, Cristián CB. **Exercise and oxidative stress in type 2 diabetes mellitus**. *Int J Mol Sci* (2018) **146** 362-72. DOI: 10.4067/s0034-98872018000300362
8. Habtemariam S. **Modulation of reactive oxygen species in health and disease**. *Antioxidants.* (2019) **8** 513-7. DOI: 10.3390/antiox8110513
9. Yuan J, Lu Y, Wang H, Feng Y, Jiang S, Gao XH. **Paeoniflorin resists H**. *Front Pharmacol* (2020) **11** 536-47. DOI: 10.3389/fphar.2020.00536
10. Crispo JA, Piche M, Ansell DR, Eibl JK, Tai IT, Kumar A. **Protective effects of methyl gallate on H**. *Biochem Bioph Res Co.* (2010) **393** 773-8. DOI: 10.1016/j.bbrc.2010.02.079
11. Pisoschi AM, Pop A, Iordache F, Stanca L, Predoi G, Serban AI. **Oxidative stress mitigation by antioxidants an overview on their chemistry and influences on health status**. *Eur J Med Chem.* (2021) **209** 112891-943. DOI: 10.1016/j.ejmech.2020.112891
12. Wang L, Xie Y, Yang W, Yang Z, Jiang S, Zhang C. **Alfalfa polysaccharide prevents H**. *Sci Rep* (2019) **9** 1782-93. DOI: 10.1038/s41598-018-38466-7
13. Wang S, Xu J, Wang C, Li J, Wang Q, Kuang H. *In J Biol Macromol.* (2020) **164** 1554-64. DOI: 10.1016/j.ijbiomac.2020.07.214
14. Zhang D, Yang B, Chang SQ, Ma SS, Sun JX, Yi L. **Protective effect of paeoniflorin on H**. *Chin J Nat Med.* (2021) **19** 90-9. DOI: 10.1016/S1875-5364(21)60010-9
15. Xu XX, Qi XM, Zhang W, Zhang CQ, Wu XX, Wu YG. **Effects of total glucosides of**. *Phytomedicine.* (2014) **21** 815-23. DOI: 10.1016/j.phymed.2013.12.003
16. Parker S, May B, Zhang C, Zhang AL, Lu C, Xue CC. **pharmacological review of bioactive constituents of**. *Phytother Res* (2016) **30** 1445-73. DOI: 10.1002/ptr.5653
17. Zhang L, Tu ZC, Yuan T, Wang H, Xie X, Fu ZF. **Antioxidants and**. *Food Chem.* (2016) **208** 61-88. DOI: 10.1016/j.foodchem.2016.03.079
18. Ma Y, Meng A, Dai Y, Yang H, Huang Y, Liu X. **Hypoglycemic activity of phenols from**. *J Funct Food.* (2022) **97** 105261-70. DOI: 10.1016/j.jff.2022.105261
19. Wein S, Schrader E, Rimbach G, Wolffram S. **Oral green tea catechins transiently lower plasma glucose concentrations in female db/db mice**. *J Med Food* (2013) **16** 312-7. DOI: 10.1089/jmf.2012.0205
20. Li Z, Tu ZC, Wang H, Zhang L. **Ultrasound-assisted extraction optimization of alpha-glucosidase inhibitors from**. *Molecules.* (2020) **25** 4507-28. DOI: 10.3390/molecules25194507
21. You G, Li C, Mao Y, Ren X, Liu Y, Wang M. **comparative analysis of raw and processed**. *Biomed Chromatogr.* (2020) **34** e4901-10. DOI: 10.1002/bmc.4901
22. Ming Y, Dong H. **The molecular mechanism of compatibility of achyranthes bidentata and**. *Indian J Pharm Scll* (2022) **84** 20-25. DOI: 10.36468/pharmaceutical-sciences.spl.407
23. Yu L, Yang J, Wang X, Jiang B, Sun Y, Ji Y. **Antioxidant and antitumor activities of**. *Oncol Rep.* (2017) **37** 357-67. DOI: 10.3892/or.2016.5249
24. Nabila B, Katarzyna KG, Krzysztof M, Lynda B, Anna K. *Mitochondrion* (2021) **59** 190-213. DOI: 10.1016/j.mito.2021.06.003
25. Zhao W, Chen L, Zhou H, Deng C, Han Q, Chen Y. **Protective effect of**. *Mol Med Rep.* (2021) **24** 741-52. DOI: 10.3892/mmr.2021.12381
26. Li C, Seeram NP. **Ultra-fast liquid chromatography coupled with electrospray ionization time-of-flight mass spectrometry for the rapid phenolic profiling of red maple (**. *J Sep Sci.* (2018) **41** 2331-46. DOI: 10.1002/jssc.201800037
27. Otsuka H, Hirata E, Shinzato T, Takeda Y. **Glochiflavanosides A-D: flavanol glucosides from the leaves of glochidion zeylanicum (Gaertn) A juss**. *Chem Pharm Bull.* (2001) **49** 921-3. DOI: 10.1248/cpb.49.921
28. Zengin G, Mahomoodally MF, Paksoy MY, Picot-Allain C, Glamocilja J, Sokovic M. **Phytochemical characterization and bioactivities of five**. *Ind Crop Prod* (2019) **135** 107-21. DOI: 10.1016/j.indcrop.2019.04.033
29. Xiong P, Qin S, Li K, Liu M, Zhu L, Peng J. **Identification of the tannins in traditional chinese medicine**. *Ara J Chem* (2021) **14** 103398-412. DOI: 10.1016/j.arabjc.2021.103398
30. He C, Peng B, Dan Y, Peng Y, Xiao P. **Chemical taxonomy of tree peony species from China based on root cortex metabolic fingerprinting**. *Phytochemistry* (2014) **107** 69-79. DOI: 10.1016/j.phytochem.2014.08.021
31. Wang X, Zhong XJ, Zhou N, Cai N, Xu JH, Wang QB. **Rapid characterizaiton of chemical constituents of the tubers of**. *Molecules.* (2020) **25** 898-917. DOI: 10.3390/molecules25040898
32. Li MC, Yao Z, Takaishi Y, Tang SA, Duan HQ. **Isolation of novel phenolic compounds with multidrug resistance (MDR) reversal properties from**. *Chem Bio* (2011) **8** 1112-20. DOI: 10.1002/cbdv.201000224
33. Nöst X, Pferschy-Wenzig EM, Nikles S, He X, Fan D, Lu A. **Identification of constituents affecting the secretion of pro-inflammatory cytokines in LPS-Induced U937 cells by UHPLC-HRMS-based metabolic profiling of the traditional Chinese medicine formulation Huangqi Jianzhong Tang**. *Molecules.* (2019) **24** 3116-43. DOI: 10.3390/molecules24173116
34. Jugran AK, Rawat S, Devkota HP, Bhatt ID, Rawal RS. **Diabetes and plant-derived natural products: from ethnopharmacological approaches to their potential for modern drug discovery and development**. *Phytother Res.* (2021) **35** 223-45. DOI: 10.1002/ptr.6821
35. Jia RB, Wu J, Luo D, Lin L, Chen C, Xiao C. **The beneficial effects of two polysaccharide fractions from**. *Foods.* (2022) **11** 1416. DOI: 10.3390/foods11101416
36. Zhang H, Tsao R. **Dietary polyphenols, oxidative stress and antioxidant and anti-inflammatory effects**. *Curr Opin Food Sci* (2016) **8** 33-42. DOI: 10.1016/j.cofs.2016.02.002
37. Skemiene K, Liobikas J, Borutaite V. **Anthocyanins as substrates for mitochondrial complex I: protective effect against heart ischemic injury**. *FEBS J* (2015) **282** 963-71. DOI: 10.1111/febs.13195
38. Yang B, Liu P. **Composition and biological activities of hydrolyzable tannins of fruits of**. *J Agr Food Chem* (2014) **62** 529-41. DOI: 10.1021/jf404703k
39. Singh BN, Shankar S, Srivastava RK. **Green tea catechin, epigallocatechin-3-gallate (EGCG): mechanisms, perspectives and clinical applications**. *Biochem Pharmacol* (2011) **82** 1807-21. DOI: 10.1016/j.bcp.2011.07.093
40. Liu ZG Li MZ, Qian DW, Liu ZA, Shu QG. **Phytochemical profiles and the hypoglycemic effects of tree**. *Food Funct.* (2021) **12** 11777-89. DOI: 10.1039/D1FO02341A
|
---
title: Vitamin D-VDR (vitamin D receptor) alleviates glucose metabolism reprogramming
in lipopolysaccharide-induced acute kidney injury
authors:
- Qing Dai
- Hao Zhang
- Shiqi Tang
- Xueqin Wu
- Jianwen Wang
- Bin Yi
- Jishi Liu
- Zhi Li
- Qin Liao
- Aimei Li
- Yan Liu
- Wei Zhang
journal: Frontiers in Physiology
year: 2023
pmcid: PMC9998528
doi: 10.3389/fphys.2023.1083643
license: CC BY 4.0
---
# Vitamin D-VDR (vitamin D receptor) alleviates glucose metabolism reprogramming in lipopolysaccharide-induced acute kidney injury
## Abstract
Background: *Our previous* study showed that vitamin D (VD)-vitamin D receptor (VDR) plays a nephroprotective role in lipopolysaccharide (LPS)-induced acute kidney injury (AKI). Recently, glucose metabolism reprogramming was reported to be involved in the pathogenesis of AKI.
Objective: To investigate the role of VD-VDR in glucose metabolism reprogramming in LPS-induced AKI.
Methods: We established a model of LPS-induced AKI in VDR knockout (VDR-KO) mice, renal proximal tubular-specific VDR-overexpressing (VDR-OE) mice and wild-type C57BL/6 mice. In vitro, human proximal tubular epithelial cells (HK-2 cells), VDR knockout and VDR overexpression HK-2 cell lines were used.
Results: Paricalcitol (an active vitamin D analog) or VDR-OE reduced lactate concentration, hexokinase activity and PDHA1 phosphorylation (a key step in inhibiting aerobic oxidation) and simultaneously ameliorated renal inflammation, apoptosis and kidney injury in LPS-induced AKI mice, which were more severe in VDR-KO mice. In in vitro experiments, glucose metabolism reprogramming, inflammation and apoptosis induced by LPS were alleviated by treatment with paricalcitol or dichloroacetate (DCA, an inhibitor of p-PDHA1). Moreover, paricalcitol activated the phosphorylation of AMP-activated protein kinase (AMPK), and an AMPK inhibitor partially abolished the protective effect of paricalcitol in LPS-treated HK-2 cells.
Conclusion: VD-VDR alleviated LPS-induced metabolic reprogramming in the kidneys of AKI mice, which may be attributed to the inactivation of PDHA1 phosphorylation via the AMPK pathway.
## 1 Introduction
Acute kidney injury (AKI) is a critical clinical syndrome with high incidence and mortality in hospitalized patients, especially in intensive care unit (ICU) patients, and there are very limited treatment options at hand (Ronco et al., 2019). Sepsis is the most common cause of severe AKI in critically ill patients (Hoste et al., 2015), and its mortality rate is approximately $30\%$ (Bouchard et al., 2015). Lipopolysaccharide (LPS, an endotoxin from the gram-negative bacterial wall), a well-known component that induces sepsis, is widely used in research on sepsis-associated AKI (SA-AKI) (Stasi et al., 2017).
SA-AKI is associated with glomerular and tubular cell damage, mainly due to systemic inflammation, changes in renal hemodynamics, and several other mechanisms (Dellepiane et al., 2016). Recently, the role of metabolic reprogramming in SA-AKI progression has been recognized, and targeting metabolic reprogramming represents a potentially effective therapeutic strategy for the progression of SA-AKI (Toro et al., 2021). Glucose metabolism reprogramming refers to the process of switching the glucose metabolism pathway from oxidative phosphorylation to glycolysis in the presence of sufficient oxygen, during which the activity of hexokinase increases, lactic acid accumulates, and the activity of the pyruvate dehydrogenase complex (PDHc) decreases (Vander Heiden et al., 2009; Biswas, 2015; Zhu et al., 2022). It has been found that the metabolism in septic mice induced by cecal ligation and puncture (CLP) and LPS-treated proximal tubule cells changes from oxidative phosphorylation to glycolysis in the presence of sufficient oxygen (Li et al., 2020; Tan et al., 2021). Moreover, the glycolysis induced by LPS-injected mice was associated with decreased renal function (Smith et al., 2014). These findings suggest the involvement of reprogramming glucose metabolism in SA-AKI.
Pyruvate dehydrogenase E1 subunit alpha 1 (PDHA1), the key regulatory site of PDHc, catalyzes the conversion of pyruvate into acetyl-CoA after it enters mitochondria (Zhou et al., 2001). It regulates the activity of PDHc through phosphorylation and dephosphorylation to affect the metabolic flux of glycolysis and the tricarboxylic acid cycle in mitochondria (Holness and Sugden, 2003). Kolobova et al. [ 2001] confirmed that specific site phosphorylation of PDHA1 can inhibit PDHc activity (Kolobova et al., 2001). It has been reported that the phosphorylation level of PDHA1 increased in CLP-AKI model mice (Li et al., 2020). Moreover, Mao et al. found that reducing the phosphorylation level of PDHA1 mitigated LPS-induced endothelial barrier dysfunction (Mao et al., 2022). Therefore, we speculate that inhibition of PDHA1 phosphorylation may be the target of SA-AKI treatment, but there are no relevant research reports at present.
VDR (vitamin D receptor) is highly expressed in the kidney and exerts nephroprotective effects through multiple mechanisms. In our previous studies, we demonstrated that 1, 2, 5(OH)2D3 (active vitamin D) or its active analogs can exert renal protection in lipopolysaccharide (LPS)-induced acute kidney injury by activating VDR (Du et al., 2019). Current studies have shown that 1,25(OH)2D3 treatment can alleviate the abnormal reprogramming of glucose metabolism in breast cancer (Santos et al., 2018), and it can also promote the transformation of dendritic cells and HEK293T cells to aerobic oxidation (Ferreira et al., 2015; Santos et al., 2017). However, whether VD-VDR can alleviate SA-AKI by regulating glucose metabolism reprogramming is unclear. In this work, we aimed to explore the role of VD-VDR in glucose metabolism reprogramming in an AKI model induced by LPS and elucidate its potential regulatory mechanism.
## 2.1 Animal experiment
Wild type male C57BL/6 mice were purchased from *Slyke jingda* Biotechnology Company (CertificateSCXK 2016-0002; Hunan, China). The VDR knockout (VDR-KO), renal proximal tubular specific VDR overexpressing (VDR-OE) mice and littermates were constructed in cooperation with the Model Animal Research Center of Nanjing University. All experimental mice were fed under SPF conditions, and the experimental protocols were approved by the Laboratory Animal Ethics Committee of Central South University.
To induce AKI, 8-week-old male mice received intraperitoneal injections of either PBS (WT group) or 20 mg/kg LPS (O111:B4, L2630, Sigma Aldrich, LPS group) for one dose. To investigate the effect of VDR activation on LPS-induced AKI, mice were injected intraperitoneally with paricalcitol (an activated vitamin D analog, a present from professor Yan Chun Li, Chicago university, 0.2 μg/kg/day for wild-type mice (Du et al., 2019) and 0.1 μg/kg/day for VDR-OE mice (Jiang et al., 2021), respectively, 1 week before LPS injection) or the same volume of solvent. Finally, wild type male C57BL/6 mice were randomly divided into WT, WT + P, LPS group and LPS + P group; VDR-KO mice and their littermates were randomly divided into WT, KO, LPS, KO + LPS groups; and OE mice and their littermates were randomly divided into WT, OE, LPS, and OE + LPS groups. All mice were sacrificed 24 h after LPS administration. Their blood and kidneys were collected for subsequent experimental analysis.
## 2.2 Cell culture and treatment
Human proximal tubular epithelial cells (HK-2 cells) and their VDR knockout (VDR-KO) cell lines were used in this study, provided by the Institute of Nephrology, Central South University. HK-2 cells were transfected with VDR plasmid and blank plasmid using Lipofectamine 2000, and cultured in F12 (1:1) DMEM supplemented with $10\%$ fetal bovine serum. Cells were seeded in six-well plates at a rate of 5 × 104 cells/well. After 24 h of incubation at 37°C, $5\%$ CO2, the cell cultures were supplemented with paricalcitol (200 nM for HK-2cells, 100 nM for VDR-OE cells) for 24 h, and then treated with LPS (1 μg/mL) for a further 16–24 h to harvest cells for follow-up experiments. In some experiments, DCA (5 mM, 2 h pretreatment, 347795, Sigma Aldrich), Compound C (10 μM, 1h pretreatment, HY-13418, MedChemExpress) were used.
## 2.3 Measurement of BUN, Cr, lactate, and hexokinase activity
Blood and renal tissues were collected for biochemical analysis. BUN and creatinine levels in serum, and lactate levels, hexokinase activity in renal tissues were measured using the corresponding detection kits in accordance with the manufacturer’s instructions. All these ELISA kits purchased from Nanjing JianCheng Bioengineering Institute (Nanjing, China). In addition, lysate of HK-2 cells was also collected for detection of lactate levels (KTB1100, Abbkine).
## 2.4 Mitochondrial morphology observation by electron microscopy
Tissues were embedded and cut into 50–100 nm ultrathin sections by an ultramicrotome and a diamond knife. Then, they were double stained with $3\%$ uranyl acetate and lead nitrate and examined with a Hitachi HT- 7700 electron microscope.
## 2.5 Renal tissue histopathological and immunofluorescence staining
Briefly, isolated mouse kidney tissue was fixed immediately with formalin and embedded in paraffin, then, the tissue was cut into 5 μm thick sections for hematoxylin-eosin (H&E) staining, TUNEL fluorescent staining, F$\frac{4}{80}$ fluorescent staining, p-PDHA1 fluorescent staining and PDHA1 fluorescent staining. The intensities of p-PDHA1 and PDHA1 in the photos were detected by Image J software.
## 2.6 Western blot analysis
After extraction of tissue and cellular protein, protein was separated by SDS/PAGE and electro-transferred to PVDF membranes. The resulting membranes were blocked with $0.1\%$ (w/v) BSA solution on a shaker for 1 h. Then, they were incubated with primary antibodies at 4°C overnight. The antibodies including: VDR (ab109234, Abcam), cleaved caspase-3 (ab49822, ab214430, Abcam), bcl2 (226593-1-AP, ProteinTech), PDHA1 (sc-377092, Santa Curz Biotechnology), p-PDHA1 (ab177461, Abcam), p-AMPK (2535, Cell Signaling Technology), AMPK (2532, Cell Signaling Technology), β-actin (20536-1-AP, ProteinTech), α-tublin (AF7010, Affinity). After that, the membrane was incubated with the fluorescent secondary antibody for 1 h before three times with TBST. Finally, the protein expression levels were visualized by the Image Studio software and band intensities were quantified with Image J gel analysis software.
## 2.7 Real-time quantitative PCR
Total RNA was extracted from the renal cortex and cell using the corresponding detection kits in accordance with the manufacturer’s instructions. cDNA was synthesized using a reverse transcription kit. Real-time quantitative PCR was performed using SYBR Green PCR Master Mix on a Roche Light Cycler 480 system. PCR primer sequences are shown in Table 1 (Chen et al., 2013; Li et al., 2022). The relative mRNA levels were calculated using the 2−ΔΔCT formula.
**TABLE 1**
| Unnamed: 0 | Mouse | Human |
| --- | --- | --- |
| IL6 | F: ATAGTCCTTCCTACCCCAATTTCC | F: GACAGCCACTCACCTCTTCA |
| IL6 | R: CTGACCACAGTGAGGAATGTCCAC | R: GCCTCTTTGCTGCTTTCACA |
| MCP1 | F: CACTCACCTGCTGCTACTCA | F: AGCAGCAAGTGTCCCAAAGA |
| MCP1 | R: CTTCTTGGGGTCAGCACAGA | R: CGGAGTTTGGGTTTGCTTGT |
| ACTIN | F: CATTGCTGACAGGATGCAGAAGG | F: CATGTACGTTGCTATCCAGGC |
| ACTIN | R: TGCTGGAAGGTGGACAGTGAGG | R: CTCCTTAATGTCACGCACGAT |
## 2.8 Oxygen consumption rate (OCR)
Oxygen consumption rate (OCR) was measured using the SeahorseXF96 Extracellular Flux Analyzer (Seahorse Bioscience, North Billerica, MA, United States). HK-2 cells were seeded into 96-well cell culture plate at a density of 1.5 × 104 cells. When the cell confluence was about $90\%$, cells were washed twice with assay medium (49.5 mL basal medium, 500 μL sodium pyruvate and basal medium) and incubated in a non-CO2 incubator for 40–60 min, OCR was measured. The working fluid concentration was as follows: oligomycin (1 μM), FCCP (2.5 μM), rotenone and antimycin A (1 μM).
## 2.9 Statistical analysis
All data were presented as means ± SD. Statistical comparisons were carried out using unpaired two-tailed Student’s t-test or one-way analysis of variance (ANOVA) as appropriate. Statistical significance was defined as $p \leq 0.05.$
## 3.1 VDR deficiency aggravated LPS-induced renal injury and glucose metabolism reprogramming
As shown in Figure 1A, LPS-induced loss of renal function featured with elevated BUN and creatinine levels, and the levels were further increased in VDR-KO mice. HE staining revealed severe tubular dilation, cell shedding and brush-border disruption in the kidney cortex after LPS injection, and these pathological lesions were more serious in VDR-KO mice than in their WT littermates (Figure 1B). Additionally, immunofluorescence analysis of TUNEL and F$\frac{4}{80}$ (a marker of macrophage infiltration) in kidney sections showed that elevated tubular cell apoptosis and interstitial inflammatory infiltrate in LPS-treated mice were aggravated in the KO + LPS group (Figure 1C). In addition, compared with the LPS group, the KO + LPS group showed more robust caspase-3 (cleaved) activation, weaker bcl2 in western blot analysis and higher mRNA expression of IL-6 and MCP-1 by real-time RT‒PCR (Figures 1D, E). These results are consistent with those of our previous study (Du et al., 2019).
**FIGURE 1:** *Effects of VDR deletion on LPS-induced AKI mice. (A) Serum concentrations of BUN and Cr at 24 h after LPS administration. (B) H&E staining of kidney sections. (C) Immunofluorescence analysis and its quantitative analysis of TUNEL (top) and F4/80 (bottom) of kidney sections. (D) Western blot analysis (left) and densitometric quantitation (right) of VDR, bcl2 and cleaved caspase3 was performed in the four groups of mice. (E) Real-time RT-PCR quantification of IL-6 and MCP1 in the renal cortex of the four groups of mice. *p < 0.05; **p < 0.01; ***p < 0.001. VDR, vitamin D receptor.*
More importantly, we found that the lactate concentration and hexokinase activity in renal homogenate were increased after LPS injection at 24 h and further increased in VDR-KO mice (Figure 2A). Western blot and immunofluorescence analyses showed that the protein expression level of p-PDHA1/PDHA1 (a key catalytic enzyme that adjusts the tricarboxylic acid cycle and oxidative phosphorylation during glycolysis through phosphorylation and dephosphorylation) was increased in LPS-injected mice, and VDR-KO mice showed a further increase in the p-PDHA1/PDHA1 ratio in renal tissue (Figures 2B, C). Since mitochondria are the main site of aerobic oxidation, we observed morphological changes in mitochondria through transmission electron microscopy, and the results showed that LPS-treated kidney tissue had more swollen mitochondria and a reduced number of cristae with a more lamellar phenotype, and these characteristics were significantly exacerbated in VDR-KO mice (Figure 2D). These results confirm that VDR deficiency aggravated LPS-induced renal injury and glucose metabolism reprogramming.
**FIGURE 2:** *VDR deletion aggravated the abnormal glycolysis of LPS-induced AKI mice. (A) Renal lactate content and hexokinase activity of the four groups. (B) Western blot analysis (left) and densitometric quantitation (right) of PDHA1 and p-PDHA1 and was performed in the four groups of mice. (C) Immunofluorescence analysis and its quantitative analysis of p-PDHA1 (green) and PDHA1 (red) of kidney sections. White arrow: glomerulus; yellow arrow: renal tubules. (D) Images of mitochondrial injury of proximal tubule epithelial cells of mice. *p < 0.05; **p < 0.01; ***p < 0.001.*
## 3.2 VDR overexpression alleviates renal injury and glucose metabolism reprogramming in LPS-induced AKI
To further confirm the role of VDR in glucose metabolism reprogramming, we constructed an LPS-induced AKI model in transgenic mice with renal proximal tubular-specific VDR-overexpressing (VDR-OE). As expected, compared with WT littermates treated with the same dose of LPS, renal function, kidney cortex pathological lesions, tubular cell apoptosis and interstitial inflammation were partially ameliorated in VDR-OE mice (Figure 3). Consistently, the increased lactate concentration and hexokinase activity in renal homogenate were lowered by overexpression of VDR (Figure 4A), and the ratio of renal p-PDHA1/PDHA1 expression was also significantly decreased in VDR-OE mice (Figures 4B, C). Electron microscopy showed that the swelling of mitochondria in VDR-OE mice treated with LPS was improved compared to their WT littermates treated with LPS (Figure 4D). These results suggest that VDR overexpression alleviated renal injury and glucose metabolism reprogramming in LPS-induced AKI.
**FIGURE 3:** *Effects of VDR overexpression on LPS-induced AKI mice. (A) Serum concentrations of BUN and Cr at 24 h after LPS administration. (B) H&E staining of kidney sections. (C) Immunofluorescence analysis and its quantitative analysis of TUNEL (top) and F4/80 (bottom) of kidney sections. (D) Western blot analysis (left) and densitometric quantitation (right) of VDR, bcl2 and cleaved caspase3 was performed in the four groups of mice. (E) Real-time RT-PCR quantification of IL-6 and MCP1 in the renal cortex of the four groups of mice. *p < 0.05; **p < 0.01; ***p < 0.001.* **FIGURE 4:** *VDR overexpression lightened the abnormal glycolysis of LPS-induced AKI mice. (A) Renal lactate content and hexokinase activity of the four groups. (B) Western blot analysis (left) and densitometric quantitation (right) of PDHA1 and p-PDHA1 and was performed in the four groups of mice. (C) Immunofluorescence analysis and its quantitative analysis of p-PDHA1 (green) and PDHA1 (red) of kidney sections. White arrow: glomerulus; yellow arrow: renal tubules. (D) Images of mitochondrial injury of proximal tubule epithelial cells from the four groups of mice. *p < 0.05; **p < 0.01; ***p < 0.001.*
## 3.3 The VDR agonist paricalcitol protected against LPS-induced AKI and alleviated glucose metabolism reprogramming
Paricalcitol (pari), an active vitamin D analog, was used in our study to illustrate the role of vitamin D in LPS-induced AKI. Our results show that pari treatment ameliorated renal insufficiency and pathological damage induced by LPS in C57 mice (Figures 5A, B). Additionally, elevated tubular cell apoptosis and interstitial inflammatory infiltrate in LPS-treated mice were inhibited in the LPS + P group (Figures 5C–E), which is consistent with our previous study (Du et al., 2019). Moreover, the increased lactate accumulation, hexokinase activity and p-PDHA1/PDHA1 ratio in LPS-treated mice were relieved by pari treatment (Figures 6A–C). The mitochondrial damage induced by LPS injection was significantly attenuated by pari treatment (Figure 6D). These results confirm that paricalcitol alleviated glucose metabolism reprogramming in LPS-induced acute kidney injury.
**FIGURE 5:** *Paricalcitol alleviated renal injury on LPS-induced AKI mice. (A) Serum concentrations of BUN and Cr at 24 h after LPS administration. (B) H&E staining of kidney sections. (C) Immunofluorescence analysis and its quantitative analysis of TUNEL (top) and F4/80 (bottom) of kidney sections. (D) Western blot analysis (left) and densitometric quantitation (right) of VDR, cleaved caspase3 and bcl2 was performed in the four groups of mice. (E) Real-time RT-PCR quantification of IL-6 and MCP1 in the renal cortex of the four groups of mice. *p < 0.05; **p < 0.01; ***p < 0.001. P, Paricalcitol; BUN, blood urea nitrogen; Cr, creatinine; H&E, Hematoxylin and eosin; MCP1, monocyte chemoattractant protein-1.* **FIGURE 6:** *Paricalcitol alleviated glucose metabolism reprogramming of LPS-induced AKI mice. (A) Renal lactate content and hexokinase activity of the four groups. (B) Western blot analysis (left) and densitometric quantitation (right) of PDHA1 and p-PDHA1 and was performed in the four groups of mice. (C) Immunofluorescence analysis and its quantitative analysis of p-PDHA1 (green) and PDHA1 (red) of kidney sections. White arrow: glomerulus; yellow arrow: renal tubules. (D) Images of mitochondrial injury of proximal tubule epithelial cells from the four groups of mice by TEM. *p < 0.05; **p < 0.01; ***p < 0.001.*
## 3.4 VDR alleviates LPS-induced glucose metabolism reprogramming and cell injury in HK2 cells
We constructed VDR knockout and VDR overexpression HK-2 cell lines to verify the effect of VDR on LPS-induced glucose metabolism reprogramming in vitro. Our results show that LPS induced glucose metabolism reprogramming, including a decreased oxygen consumption rate (OCR) and increased lactate levels, which was more serious in VDR-KO cells (Figures 7A, B). On the contrary, overexpression of VDR significantly attenuated LPS induced reprogramming of glucose metabolism featured with restored OCR and decreased lactate levels (Figures 8A, B). Furthermore, compared with the LPS group, the KO + LPS group had a higher p-PDHA1/PDHA1 ratio (Figure 7C), while the OE + LPS group had a lower p-PDHA1/PDHA1 ratio (Figure 8C). The expression of caspase-3 (cleaved) and bcl2 by western blot (Figures 7C, 8C) and the mRNA expression of IL-6 and MCP-1 (by PCR) (Figures 7D, 8D) showed that VDR knockout could promote cell apoptosis and inflammation in HK-2 cells, while VDR overexpression could improve these alterations induced by LPS. These results suggests that VDR alleviates LPS-induced glucose metabolism reprogramming and cell injury in renal tubular cells.
**FIGURE 7:** *VDR deletion aggravated abnormal glycolysis and injury in LPS-induced HK-2 cell. (A) Oxygen consumption rate (OCR) (top) measured by Seahorse metabolic analyzer and quantitative analysis (bottom) of mitochondrial function parameters (basal respiration, maximal respiration). (B) lactate content in HK-2 cells and VDR-KO cells treated with LPS for 24 h. (C) Western blot analysis (left) and densitometric quantitation (right) of VDR, PDHA1, p-PDHA1, cleaved caspase3 and bcl2 was performed in the four groups of HK-2 cells. (D) Real-time RT-PCR quantification of IL-6 and MCP1. *p < 0.05; **p < 0.01; ***p < 0.001. Oligo, oligomycin; Rote, rotenone; Anti, antimycin A.* **FIGURE 8:** *VDR overexpression lightened LPS-induced abnormal glycolysis and injury in HK-2 cell. (A) Oxygen consumption rate (OCR) (top) measured by Seahorse metabolic analyzer and quantitative analysis (bottom) of mitochondrial function parameters (basal respiration, maximal respiration). (B) lactate content in HK-2 cells and VDR-OE cells treated with LPS for 24 h. (C) Western blot analysis (left) and densitometric quantitation (right) of VDR, PDHA1, p-PDHA1, cleaved caspase3 and bcl2 was performed in the four groups of HK-2 cells. (D) Real-time RT-PCR quantification of IL-6 and MCP1. *p < 0.05; **p < 0.01; ***p < 0.001. Oligo, oligomycin; Rote, rotenone; Anti, antimycin A.*
## 3.5 VD-VDR alleviates LPS-induced glucose metabolism reprogramming by inhibiting the phosphorylation of PDHA1
The above results indicate that VD-VDR can reduce the level of phosphorylated PDHA1 (p-PDHA1) but has no evident effect on the total protein level of PDHA1 in LPS-induced glucose metabolism reprogramming. Thus, to confirm that VDR can play a protective role in glucose metabolism reprogramming and renal injury in LPS-induced AKI by regulating PDHA1, DCA (a p-PDHA1 inhibitor) was used to evaluate LPS-induced tubular cell injury in vitro. In HK2 cells, the decreased OCR levels and increased cellular lactate accumulation induced by LPS were protected by paricalcitol or the p-PDHA1 inhibitor DCA, respectively. Importantly, when treated with both DCA and pari, the protective effect on the glucose metabolism reprogramming of HK-2 cells was no better than that of DCA alone (Figures 9A, B), and the same phenomenon also appeared in the impact on the ratio of p-PDHA1/PDHA1 expression (Figure 9C). Interestingly, the anti-apoptotic and anti-inflammatory effects of the DCA and pari combination in LPS-induced HK2 cells were comparable to those of pari alone (Figures 9C, D). These data confirm that VDR activation can alleviate LPS-induced glucose metabolism reprogramming by inhibiting the phosphorylation of PDHA1.
**FIGURE 9:** *Paricalcitol alleviated LPS-induced injury through phosphorylation of PDHA1. (A) Oxygen consumption rate (OCR) (top) measured by Seahorse metabolic analyzer and quantitative analysis (bottom) of mitochondrial function parameters (basal respiration, maximal respiration). (B) lactate content in HK-2 cells treated with LPS, LPS + P, LPS + D (D: DCA, 5mM, pretreated 2 h), LPS + P + D for 24 h. (C) Western blot analysis (top) and densitometric quantitation (bottom) of PDHA1, p-PDHA1, cleaved caspase3 and bcl2 was performed in the five groups of HK-2 cells. (D) Real-time RT-PCR quantification of IL-6 and MCP1. *p < 0.05; **p < 0.01; ***p < 0.001. P, paricalcitol; D: DCA, dichloroacetic acid solution. Oligo, oligomycin; Rote, rotenone; Anti, antimycin A.*
## 3.6 VD-VDR alleviates glucose metabolism reprogramming via the activation of AMPK in LPS-induced renal cell injury
We further investigated how VD-VDR inhibits the phosphorylation of PDHA1. Since our previous research confirmed that VD-VDR can activate AMPK in diabetic nephropathy (Li et al., 2022), we detected the level of protein expression of p-AMPK in LPS-induced AKI mice and LPS-treated HK-2 cells. As Figures 10, 11 show, p-AMPK levels were increased in LPS-induced AKI mice and LPS-treated HK-2 cells, and paricalcitol or VDR overexpression further promoted the expression of p-AMPK (Figures 10B, C, 11B). In addition, the increased level of p-AMPK was weakened in VDR-KO mice and HK-2 cells (Figures 10A, 11A).
**FIGURE 10:** *VD-VDR active AMPK in LPS-induced AKI mice. (A) Western blot analysis (left) and densitometric quantitation (right) of AMPK and p-AMPK was performed in group of WT, KO, LPS and KO + LPS. (B) Western blot analysis (left) and densitometric quantitation (right) of AMPK and p-AMPK was performed in group of WT, OE, LPS and OE + LPS. (C) Western blot analysis (left) and densitometric quantitation (right) of AMPK and p-AMPK was performed in group of WT, WT + P, LPS and LPS + P. *p < 0.05; **p < 0.01; ***p < 0.001.* **FIGURE 11:** *VD alleviate glucose metabolism reprogramming via the activation of AMPK pathway. (A) Western blot analysis (left) and densitometric quantitation (right) of AMPK and p-AMPK in VDR-KO cells treated with LPS for 24 h. (B) Western blot analysis (left) and densitometric quantitation (right) of AMPK and p-AMPK in VDR-OE cells treated with LPS for 24 h. (C) Oxygen consumption rate (OCR) (top) measured by Seahorse metabolic analyzer and quantitative (bottom) analysis of mitochondrial function parameters (basal respiration, maximal respiration). (D) lactate content in HK-2 cells treated with LPS, LPS + P, LPS + C (C: compound C, 10 μM, pretreated 1 h), LPS + P + C for 12–16 h. (E) Western blot analysis (left) and densitometric quantitation (right) of AMPK, p-AMPK, PDHA1, p-PDHA1, cleaved caspase3 and bcl2 was performed in the five groups of HK-2 cells. (F) Real-time RT-PCR quantification of IL-6 and MCP1. *p < 0.05; **p < 0.01; ***p < 0.001. Oligo, oligomycin; Rote, rotenone; Anti, antimycin A.*
It has been reported that the glycolysis shift induced by LPS is related to AMP-activated protein kinase (AMPK) (Tan et al., 2021). Therefore, we examined the protein expression of p-AMPK/AMPK in LPS-treated HK2 cells and observed glycolytic metabolism in HK2 cells treated with Compound C, an AMPK inhibitor. As our data show, Compound C and LPS induced the metabolic state switch from oxidative phosphorylation (lower OCR levels) to glycolysis (higher lactic acid levels), while the effects on alleviating glucose metabolism reprogramming by pari were greatly weakened after Compound C treatment (Figures 11C, D). Similar results were observed in the effect of Compound C on the expression of p-PDHA1/PDHA1 and caspase-3 (cleaved) (Figure 11E). Real-time PCR analysis indicated that Compound C could induce increased expression of IL-6 mRNA, and the reduction in IL-6 by paricalcitol was also largely abolished in the presence of Compound C, but it had no influence on MCP-1 mRNA (Figure 11F).
## 4 Discussion
In this study, we report the role of VD-VDR in renal glucose metabolism reprogramming induced by LPS for the first time. Our data show that paricalcitol treatment or VDR-specific overexpression restored glucose metabolism reprogramming and renal injury in LPS-induced AKI, whereas VDR-KO resulted in a more severe glycolytic shift and renal injury. In addition, we also initially found that paricalcitol attenuated LPS-induced reprogramming of glucose metabolism in HK-2 cells partially via the AMPK pathway.
Glucose metabolism reprogramming is increasingly recognized as a potentially effective therapeutic strategy for the progression of AKI (Li et al., 2021). Researchers have proposed that metabolic reprogramming exerts renal protection by making up for the damaged energy supply in a short time at the early stage of SA-AKI (Gómez et al., 2017). However, most evidence shows that the aerobic glycolysis transition of renal tubular epithelial cells is harmful and aggravates the damage to renal function. This is because during sepsis, continuous and different injuries may significantly magnify tubular injuries in cells (Toro et al., 2021). In addition, accumulation of lactate and the end product of glycolysis can activate innate immune and inflammatory responses through TLR-mediated NF-κB signaling and inflammasomes (Samuvel et al., 2009), while the content of pyruvate, which has anti-inflammatory and antioxidant effects, is reduced (Zager et al., 2014), while inflammatory factors are continuously secreted, resulting in a persistent inflammatory state and mitochondrial damage, leading to renal tubular epithelial cell damage. Moreover, the inhibition of aerobic glycolysis alleviates SA-AKI (Tan et al., 2021). Therefore, it is important to reduce glucose metabolic reprogramming to restore renal function in SA-AKI.
Glucose metabolism reprogramming is regulated by many metabolic enzymes. Hexokinase (HK2), the first rate-limiting enzyme of glycolytic metabolism, and lactic acid, the final metabolite of glycolysis under anaerobic conditions, both reflect the glycolysis activity. The OCR level and dephosphorylation level of PDHA1 reflect the aerobic oxidation activity. During SA-AKI, lactate is elevated in septic pigs (Chvojka et al., 2008), and sepsis induces a metabolic shift to aerobic glycolysis in CLP mice (Waltz et al., 2016) and LPS-treated HK-2 cells (Ji et al., 2021). Our data show that LPS-injected mice had a higher lactate concentration and hexokinase activity in renal homogenate than wild-type mice, and LPS-treated HK-2 cells had decreased OCR levels and higher lactate levels than the control group, which is consistent with previously mentioned reports. This runaway reprogramming of glucose metabolism can be restored by paricalcitol, a VDR agonist. VDR can inhibit glycolysis in colorectal cancer (Zuo et al., 2020), and VD supplementation can improve mitochondrial respiration in primary trophoblasts isolated from obese women (Phillips et al., 2022). Ryan et al. [ 2016] found that phosphorylated pyruvate dehydrogenase (PDH) (Ser-293) decreased in 1α,25(OH)2D3-treated human skeletal muscle cells. However, the effect of VD-VDR on the reprogramming of glucose metabolism has not been studied in AKI. Our results show that paricalcitol reduced the elevated lactate concentration and hexokinase activities in LPS-induced AKI mice; more importantly, paricalcitol inhibited the phosphorylation of PDHA1. Additionally, we constructed an LPS-AKI model in VDR-KO and VDR-OE mice to explore the effect of VDR on the reprogramming of glucose metabolism in LPS-AKI. As expected, knockout of VDR aggravated glucose metabolism reprogramming in LPS-AKI mice, including lactate accumulation and hexokinase activity, whereas overexpression of VDR attenuated glucose metabolism reprogramming. Moreover, our in vitro experimental data are consistent with those found in animal experiments. These results confirm that VD-VDR could alleviate glucose metabolism reprogramming in LPS-induced acute kidney injury.
In our animal experiments, we confirmed the regulatory effect of VDR on p-PDHA1; however, the specific regulatory mechanism is still unclear. PDHA1 is a key site regulating PDHc, and its phosphorylation is involved in the pathological mechanism of many diseases. Oh et al. found that phosphorylation of PDHA1 mediates cisplatin-induced acute kidney injury and may be a therapeutic target for cisplatin-induced acute kidney injury (Oh et al., 2017). Pan et al. found that PDHA1 dephosphorylation reduces pyroptosis-induced inflammation (Pan et al., 2022). In the present work, we used DCA, an inhibitor of p-PDHA1, to explore the regulation of p-PDHA1 by VDR. Our results show that in LPS-treated HK2 cells, inhibition of p-PDHA1 reduces cell glucose metabolic reprogramming, cell inflammation and apoptosis. It seems that the protective effects of the DCA and paricalcitol combination in LPS-induced HK2 cells were comparable to those of paricalcitol alone. This finding supports that paricalcitol can alleviate glucose metabolism reprogramming by inhibiting p-PDHA1.
AMP-activated protein kinase (AMPK) is a classic energy receptor. Under conditions of metabolic stress, such as hypoxia and ischemia, AMPK is activated to increase ATP production and reduce ATP consumption to maintain cellular energy homeostasis (Burkewitz et al., 2014; Hardie, 2015). Zhang et al. [ 2022] showed that AMPK antagonizes nickel-refining fume-induced aerobic glycolysis. Similarly, Tang et al. [ 2021] found that enhanced glycolysis in lung fibroblasts induced by LPS can be prevented by regulating the AMPK pathway. Similarly, Tang et al. [ 2021] found that enhanced glycolysis in lung fibroblasts induced by LPS can be prevented by regulating the AMPK pathway (Jin et al., 2020). AMPK plays an important role in glucose metabolism reprogramming and is also a fundamental regulator of many pathways involved in energy metabolism (Hardie et al., 2012). Herein, we detected the expression of p-AMPK/AMPK experimentally and confirmed that the AMPK pathway was activated in LPS-induced AKI mice, and the activation of VDR could further increase the expression of p-AMPK. Notably, the activation of AMPK during sepsis is an early adaptive response to injury, while the pharmacological activation of AMPK can protect against AKI and improve the survival rate of SA-AKI mice (Jin et al., 2020). In our experiment, VD-VDR mediates the activation of the AMPK pathway and plays a protective role against AKI, which is consistent with our previous report (Li et al., 2022).
It has been reported that PDHc activity can be restored by treatment with an AMPK activator (Dugan et al., 2013). Cai et al. [ 2020] also found that S293 phosphorylation of PDHA1 was increased in AMPK knockout cells, while in AMPK-activated cells, S293 phosphorylation of PDHA1 was decreased and PDHc was activated. As expected, our results show that the protective effect of paricalcitol on metabolic reprogramming was weakened by the inhibition of the AMPK pathway. This result indicates that the AMPK pathway is involved in the regulation of the ratio of p-PDHA1/PDHA1 by VDR and exerts a reno-protective effect. The results from the above cellular experiments demonstrate that VDR regulates PDHA1 phosphorylation by activating AMPK.
Considering that VDR is a nuclear transcription factor, we wondered whether there is transcriptional regulation of PDHA1 by VDR. The results show that the expression of PDHA1 was only slightly downregulated in the kidney tissue of LPS-injected VDR-KO mice, which is inconsistent with the downregulation of VDR. However, the overall effect of VDR on the inhibition of metabolic reprogramming and renal protection is clearly presented, and we speculate that VD-VDR may affect metabolic reprogramming by regulating other molecules, which requires further investigation in the future.
Overall, our data demonstrate that VD-VDR could alleviate glucose metabolism reprogramming in lipopolysaccharide-induced acute kidney injury, mediated by activation of the AMPK pathway. Our work provides new insight into the renoprotective effect of vitamin D-VDR in SA-AKI and provides a promising target for AKI prevention and treatment.
## Data availability statement
The raw data supporting the conclusion of this article will be made available by the authors, without undue reservation.
## Ethics statement
The animal study was reviewed and approved by the Laboratory Animal Ethics Committee of Central South University.
## Author contributions
WZ, AL, and YL conceived the idea and designed the experiments. JW and QL contributed to the literature research. QD, ST, and XW performed the experiments. ZL and JL contributed to the data analysis and statistical analysis. QD, WZ, and AL prepared the figures, drafted the introduction and results. YL and BY contributed to the discussion. WZ, AL, and HZ revised the manuscript. All authors have approved the final version of the manuscript.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fphys.2023.1083643/full#supplementary-material
## References
1. Biswas S. K.. **Metabolic reprogramming of immune cells in cancer progression**. *Immunity* (2015) **43** 435-449. DOI: 10.1016/j.immuni.2015.09.001
2. Bouchard J., Acharya A., Cerda J., Maccariello E. R., Madarasu R. C., Tolwani A. J.. **A prospective international multicenter study of AKI in the intensive care unit**. *Clin. J. Am. Soc. Nephrol.* (2015) **10** 1324-1331. DOI: 10.2215/cjn.04360514
3. Burkewitz K., Zhang Y., Mair W. B.. **AMPK at the nexus of energetics and aging**. *Cell Metab.* (2014) **20** 10-25. DOI: 10.1016/j.cmet.2014.03.002
4. Cai Z., Li C. F., Han F., Liu C., Zhang A., Hsu C. C.. **Phosphorylation of**. *Mol. Cell* (2020) **80** 263-278.e7. DOI: 10.1016/j.molcel.2020.09.018
5. Chen Y., Liu W., Sun T., Huang Y., Wang Y., Deb D. K.. **1,25-Dihydroxyvitamin D promotes negative feedback regulation of TLR signaling via targeting microRNA-155-SOCS1 in macrophages**. *J. Immunol.* (2013) **190** 3687-3695. DOI: 10.4049/jimmunol.1203273
6. Chvojka J., Sykora R., Krouzecky A., Radej J., Varnerova V., Karvunidis T.. **Renal haemodynamic, microcirculatory, metabolic and histopathological responses to peritonitis-induced septic shock in pigs**. *Crit. Care* (2008) **12** R164. DOI: 10.1186/cc7164
7. Dellepiane S., Marengo M., Cantaluppi V.. **Detrimental cross-talk between sepsis and acute kidney injury: New pathogenic mechanisms, early biomarkers and targeted therapies**. *Crit. Care* (2016) **20** 61. DOI: 10.1186/s13054-016-1219-3
8. Du J., Jiang S., Hu Z., Tang S., Sun Y., He J.. **Vitamin D receptor activation protects against lipopolysaccharide-induced acute kidney injury through suppression of tubular cell apoptosis**. *Am. J. Physiol. Ren. Physiol.* (2019) **316** F1068-f1077. DOI: 10.1152/ajprenal.00332.2018
9. Dugan L. L., You Y. H., Ali S. S., Diamond-Stanic M., Miyamoto S., DeCleves A. E.. **AMPK dysregulation promotes diabetes-related reduction of superoxide and mitochondrial function**. *J. Clin. Invest.* (2013) **123** 4888-4899. DOI: 10.1172/jci66218
10. Ferreira G. B., Vanherwegen A. S., Eelen G., Gutiérrez A. C. F., Van Lommel L., Marchal K.. **Vitamin D3 induces tolerance in human dendritic cells by activation of intracellular metabolic pathways**. *Cell Rep.* (2015) **10** 711-725. DOI: 10.1016/j.celrep.2015.01.013
11. Gómez H., Kellum J. A., Ronco C.. **Metabolic reprogramming and tolerance during sepsis-induced AKI**. *Nat. Rev. Nephrol.* (2017) **13** 143-151. DOI: 10.1038/nrneph.2016.186
12. Hardie D. G.. **Ampk: Positive and negative regulation, and its role in whole-body energy homeostasis**. *Curr. Opin. Cell Biol.* (2015) **33** 1-7. DOI: 10.1016/j.ceb.2014.09.004
13. Hardie D. G., Ross F. A., Hawley S. A.. **Ampk: A nutrient and energy sensor that maintains energy homeostasis**. *Nat. Rev. Mol. Cell Biol.* (2012) **13** 251-262. DOI: 10.1038/nrm3311
14. Holness M. J., Sugden M. C.. **Regulation of pyruvate dehydrogenase complex activity by reversible phosphorylation**. *Biochem. Soc. Trans.* (2003) **31** 1143-1151. DOI: 10.1042/bst0311143
15. Hoste E. A., Bagshaw S. M., Bellomo R., Cely C. M., Colman R., Cruz D. N.. **Epidemiology of acute kidney injury in critically ill patients: The multinational AKI-EPI study**. *Intensive Care Med.* (2015) **41** 1411-1423. DOI: 10.1007/s00134-015-3934-7
16. Ji R., Chen W., Wang Y., Gong F., Huang S., Zhong M.. **The warburg effect promotes mitochondrial injury regulated by uncoupling protein-2 in septic acute kidney injury**. *Shock* (2021) **55** 640-648. DOI: 10.1097/shk.0000000000001576
17. Jiang S., Zhang H., Li X., Yi B., Huang L., Hu Z.. **Vitamin D/**. *J. Steroid Biochem. Mol. Biol.* (2021) **206** 105789. DOI: 10.1016/j.jsbmb.2020.105789
18. Jin K., Ma Y., Manrique-Caballero C. L., Li H., Emlet D. R., Li S.. **Activation of AMP-activated protein kinase during sepsis/inflammation improves survival by preserving cellular metabolic fitness**. *Faseb J.* (2020) **34** 7036-7057. DOI: 10.1096/fj.201901900R
19. Kolobova E., Tuganova A., Boulatnikov I., Popov K. M.. **Regulation of pyruvate dehydrogenase activity through phosphorylation at multiple sites**. *Biochem. J.* (2001) **358** 69-77. DOI: 10.1042/0264-6021:3580069
20. Li A., Yi B., Han H., Yang S., Hu Z., Zheng L.. **Vitamin D-**. *Autophagy* (2022) **18** 877-890. DOI: 10.1080/15548627.2021.1962681
21. Li Y., Nourbakhsh N., Pham H., Tham R., Zuckerman J. E., Singh P.. **Evolution of altered tubular metabolism and mitochondrial function in sepsis-associated acute kidney injury**. *Am. J. Physiol. Ren. Physiol.* (2020) **319** F229-f244. DOI: 10.1152/ajprenal.00390.2019
22. Li Z., Lu S., Li X.. **The role of metabolic reprogramming in tubular epithelial cells during the progression of acute kidney injury**. *Cell Mol. Life Sci.* (2021) **78** 5731-5741. DOI: 10.1007/s00018-021-03892-w
23. Mao L., Sun M., Chen Z., Zeng Z., Wu J., Chen Z.. **The pyruvate dehydrogenase complex mitigates LPS-induced endothelial barrier dysfunction by metabolic regulation**. *Shock* (2022) **57** 308-317. DOI: 10.1097/shk.0000000000001931
24. Oh C. J., Ha C. M., Choi Y. K., Park S., Choe M. S., Jeoung N. H.. **Pyruvate dehydrogenase kinase 4 deficiency attenuates cisplatin-induced acute kidney injury**. *Kidney Int.* (2017) **91** 880-895. DOI: 10.1016/j.kint.2016.10.011
25. Pan S. S., Wang F., Hui Y. P., Chen K. Y., Zhou L., Gao W. L.. **Insulin reduces pyroptosis-induced inflammation by PDHA1 dephosphorylation-mediated NLRP3 activation during myocardial ischemia-reperfusion injury**. *Perfusion* (2022) **4** 026765912210998. DOI: 10.1177/02676591221099807
26. Phillips E. A., Hendricks N., Bucher M., Maloyan A.. **Vitamin D supplementation improves mitochondrial function and reduces inflammation in placentae of obese women**. *Front. Endocrinol. (Lausanne)* (2022) **13** 893848. DOI: 10.3389/fendo.2022.893848
27. Ronco C., Bellomo R., Kellum J. A.. **Acute kidney injury**. *Lancet* (2019) **394** 1949-1964. DOI: 10.1016/s0140-6736(19)32563-2
28. Ryan Z. C., Craig T. A., Folmes C. D., Wang X., Lanza I. R., Schaible N. S.. **1α,25-Dihydroxyvitamin D3 regulates mitochondrial oxygen consumption and dynamics in human skeletal muscle cells**. *J. Biol. Chem.* (2016) **291** 1514-1528. DOI: 10.1074/jbc.M115.684399
29. Samuvel D. J., Sundararaj K. P., Nareika A., Lopes-Virella M. F., Huang Y.. **Lactate boosts TLR4 signaling and NF-kappaB pathway-mediated gene transcription in macrophages via monocarboxylate transporters and MD-2 up-regulation**. *J. Immunol.* (2009) **182** 2476-2484. DOI: 10.4049/jimmunol.0802059
30. Santos G. C., Zeidler J. D., Pérez-Valencia J. A., Sant'Anna-Silva A. C. B., Da Poian A. T., El-Bacha T.. **Metabolomic analysis reveals vitamin D-induced decrease in polyol pathway and subtle modulation of glycolysis in HEK293T cells**. *Sci. Rep.* (2017) **7** 9510. DOI: 10.1038/s41598-017-10006-9
31. Santos J. M., Khan Z. S., Munir M. T., Tarafdar K., Rahman S. M., Hussain F.. **Vitamin D(3) decreases glycolysis and invasiveness, and increases cellular stiffness in breast cancer cells**. *J. Nutr. Biochem.* (2018) **53** 111-120. DOI: 10.1016/j.jnutbio.2017.10.013
32. Smith J. A., Stallons L. J., Schnellmann R. G.. **Renal cortical hexokinase and pentose phosphate pathway activation through the EGFR/Akt signaling pathway in endotoxin-induced acute kidney injury**. *Am. J. Physiol. Ren. Physiol.* (2014) **307** F435-F444. DOI: 10.1152/ajprenal.00271.2014
33. Stasi A., Intini A., Divella C., Franzin R., Montemurno E., Grandaliano G.. **Emerging role of Lipopolysaccharide binding protein in sepsis-induced acute kidney injury**. *Nephrol. Dial. Transpl.* (2017) **32** 24-31. DOI: 10.1093/ndt/gfw250
34. Tan C., Gu J., Li T., Chen H., Liu K., Liu M.. **Inhibition of aerobic glycolysis alleviates sepsis-induced acute kidney injury by promoting lactate/Sirtuin 3/AMPK-regulated autophagy**. *Int. J. Mol. Med.* (2021) **47** 19. DOI: 10.3892/ijmm.2021.4852
35. Tang C. J., Xu J., Ye H. Y., Wang X. B.. **Metformin prevents PFKFB3-related aerobic glycolysis from enhancing collagen synthesis in lung fibroblasts by regulating AMPK/mTOR pathway**. *Exp. Ther. Med.* (2021) **21** 581. DOI: 10.3892/etm.2021.10013
36. Toro J., Manrique-Caballero C. L., Gómez H.. **Metabolic reprogramming and host tolerance: A novel concept to understand sepsis-associated AKI**. *J. Clin. Med.* (2021) **10** 4184. DOI: 10.3390/jcm10184184
37. Vander Heiden M. G., Cantley L. C., Thompson C. B.. **Understanding the warburg effect: The metabolic requirements of cell proliferation**. *Science* (2009) **324** 1029-1033. DOI: 10.1126/science.1160809
38. Waltz P., Carchman E., Gomez H., Zuckerbraun B.. **Sepsis results in an altered renal metabolic and osmolyte profile**. *J. Surg. Res.* (2016) **202** 8-12. DOI: 10.1016/j.jss.2015.12.011
39. Zager R. A., Johnson A. C., Becker K.. **Renal cortical pyruvate depletion during AKI**. *J. Am. Soc. Nephrol.* (2014) **25** 998-1012. DOI: 10.1681/asn.2013070791
40. Zhang T., Wang Y., Chen Y., Gao Y., Zhang D., Jin S.. **Metformin alleviates nickel-refining fumes-induced aerobic glycolysis via AMPK/GOLPH3 pathway**. *Ecotoxicol. Environ. Saf.* (2022) **236** 113461. DOI: 10.1016/j.ecoenv.2022.113461
41. Zhou Z. H., McCarthy D. B., O'Connor C. M., Reed L. J., Stoops J. K.. **The remarkable structural and functional organization of the eukaryotic pyruvate dehydrogenase complexes**. *Proc. Natl. Acad. Sci. U. S. A.* (2001) **98** 14802-14807. DOI: 10.1073/pnas.011597698
42. Zhu S., Gu H., Peng C., Xia F., Cao H., Cui H.. **Regulation of glucose, fatty acid and amino acid metabolism by ubiquitination and SUMOylation for cancer progression**. *Front. Cell Dev. Biol.* (2022) **10** 849625. DOI: 10.3389/fcell.2022.849625
43. Zuo S., Wu L., Wang Y., Yuan X.. **Long non-coding RNA MEG3 activated by vitamin D suppresses glycolysis in colorectal cancer via promoting c-myc degradation**. *Front. Oncol.* (2020) **10** 274. DOI: 10.3389/fonc.2020.00274
|
---
title: Cell surface GRP78 regulates TGFβ1-mediated profibrotic responses via TSP1
in diabetic kidney disease
authors:
- Jackie Trink
- Usman Ahmed
- Kian O’Neil
- Renzhong Li
- Bo Gao
- Joan C. Krepinsky
journal: Frontiers in Pharmacology
year: 2023
pmcid: PMC9998550
doi: 10.3389/fphar.2023.1098321
license: CC BY 4.0
---
# Cell surface GRP78 regulates TGFβ1-mediated profibrotic responses via TSP1 in diabetic kidney disease
## Abstract
Introduction: Diabetic kidney disease (DKD) is the leading cause of kidney failure in North America, characterized by glomerular accumulation of extracellular matrix (ECM) proteins. High glucose (HG) induction of glomerular mesangial cell (MC) profibrotic responses plays a central role in its pathogenesis. We previously showed that the endoplasmic reticulum resident GRP78 translocates to the cell surface in response to HG, where it mediates Akt activation and downstream profibrotic responses in MC. Transforming growth factor β1 (TGFβ1) is recognized as a central mediator of HG-induced profibrotic responses, but whether its activation is regulated by cell surface GRP78 (csGRP78) is unknown. TGFβ1 is stored in the ECM in a latent form, requiring release for biological activity. The matrix glycoprotein thrombospondin 1 (TSP1), known to be increased in DKD and by HG in MC, is an important factor in TGFβ1 activation. Here we determined whether csGRP78 regulates TSP1 expression and thereby TGFβ1 activation by HG.
Methods: Primary mouse MC were used. TSP1 and TGFβ1 were assessed using standard molecular biology techniques. Inhibitors of csGRP78 were: 1) vaspin, 2) the C-terminal targeting antibody C38, 3) siRNA downregulation of its transport co-chaperone MTJ-1 to prevent GRP78 translocation to the cell surface, and 4) prevention of csGRP78 activation by its ligand, active α2-macroglobulin (α2M*), with the neutralizing antibody Fα2M or an inhibitory peptide.
Results: TSP1 transcript and promoter activity were increased by HG, as were cellular and ECM TSP1, and these required PI3K/Akt activity. Inhibition of csGRP78 prevented HG-induced TSP1 upregulation and deposition into the ECM. The HG-induced increase in active TGFβ1 in the medium was also inhibited, which was associated with reduced intracellular Smad3 activation and signaling. Overexpression of csGRP78 increased TSP-1, and this was further augmented in HG.
Discussion: *These data* support an important role for csGRP78 in regulating HG-induced TSP1 transcriptional induction via PI3K/Akt signaling. Functionally, this enables TGFβ1 activation in response to HG, with consequent increase in ECM proteins. Means of inhibiting csGRP78 signaling represent a novel approach to preventing fibrosis in DKD.
## 1 Introduction
As the largest cause of kidney failure worldwide, diabetic kidney disease (DKD) places a significant burden on our healthcare system. Current therapies include stabilizing blood glucose and blood pressure, the use of inhibitors of the renin-angiotensin system and SGLT2 inhibitors in type 2 diabetics. Further studies are currently assessing efficacy of newer therapies such as inhibitors of DPP-4 and Rho-kinase, as well as the immunomodulatory peptide AcSDKP and activation of the sirtuin SIRT3 (Omata et al., 2006; Kolavennu et al., 2008; Castoldi et al., 2013; Zuo et al., 2013; Peng et al., 2016; Lin et al., 2018; Xu et al., 2018; Locatelli et al., 2020). However, the current standard of care for DKD can only slow disease progression, with many patients developing end-stage kidney disease and requiring costly therapies including dialysis or kidney transplant (Johnson and Spurney, 2015; Tuttle et al., 2020). Thus, identifying novel therapeutic interventions that can prevent disease progression is crucial. The primary hallmark of DKD manifestation begins with structural changes to the glomerulus, the filtering unit of the kidney. These pathological changes include glomerular basement membrane thickening and the overproduction of extracellular matrix proteins (ECM) by glomerular mesangial cells (MC) which leads to glomerulosclerosis and ultimately loss of filtration ability (Reidy et al., 2014; Alicic et al., 2017; Sagoo and Gnudi, 2020). Thus, MC play a critical role in the pathogenesis of DKD and are an important therapeutic target for attenuating fibrosis.
The profibrotic cytokine transforming growth factor β1 (TGFβ1) is well characterized for its role in promoting ECM production by MC as well as other kidney cell types including endothelial cells, fibroblasts and podocytes in response to high glucose (HG) through Smad-dependent and -independent signaling pathways (Verrecchia et al., 2001; Li et al., 2003; Edeling et al., 2016). Some of these latter pathways include Wnt, hedge-hog, SIRT3, FGFR1, glucocorticoid receptor-mediated signaling and PI3k/Akt. Together, these and others contribute to cellular transformation to a profibrotic phenotype (such as macrophage-to-myofibroblast transition (MMT), epithelial mesenchymal transition (EMT) and endothelial-to-mesenchymal transition (EndMT) (Kato et al., 2006; Lebleu et al., 2013; Nikolic-Paterson et al., 2014; Cho et al., 2018). TGFβ1 is secreted in an inactive form, with the mature cytokine non-covalently attached to its N-terminal propeptide, the latency-associated peptide (LAP). As LAP binding blocks TGFβ1 receptor binding, extracellular dissociation from LAP is required to reveal the receptor recognition site and thus cytokine activation (Munger et al., 1997; Koli et al., 2001). Integrin activation may dissociate LAP either through traction and mechanical release or through protease activation and LAP cleavage (Worthington et al., 2011). However, activation of TGFβ1 by HG in MC was shown to be dependent on the extracellular glycoprotein thrombospondin 1 (TSP1), which enables the non-proteolytic release of active TGFβ1 (Crawford et al., 1998; Yevdokimova et al., 2001). This interaction with LAP is specific to TSP1, as TSP2 lacks the peptide sequence required to non-proteolytically release and activate TGFβ1 (Hugo, 2003).
TSP1 interacts with both the mature portion of TGFβ1 and with LAP through distinct sites. Binding to the mature domain orients TSP1 to enable additional binding to LAP. This binding disrupts the intramolecular LAP-mature domain interaction to expose the receptor binding sequence of TGFβ1, enabling receptor binding and signaling (Murphy-Ullrich and Suto, 2018). Upregulation and increased deposition of TSP1 into the ECM as well as its role in TGFβ1 activation has been reported in HG in MC as well as in vivo in DKD mouse models and in human DKD (Poczatek et al., 2000; Wahab et al., 2005; Hohenstein et al., 2008; Lu et al., 2011). Furthermore, TSP1 knockout mice were protected from the development of DKD (Daniel et al., 2007), and the use of the peptide LSKL which blocks the interaction of TSP1 with LAP, protected type 1 diabetic Akita mice from DKD (Lu et al., 2011). Reduced active TGFβ1 and downstream signaling were seen in both studies. Thus, TSP1 is critical for the activation and profibrotic signaling of TGFβ1, but how TSP1 is upregulated in DKD has not yet been elucidated.
Previously, our lab has shown that the endogenous endoplasmic reticulum resident GRP78 translocates to the cell surface of MC in response to HG. Here, it acts as a receptor for the activated form of the protease inhibitor alpha 2 macroglobulin, the binding of which enables HG-induced downstream profibrotic signaling (Van Krieken et al., 2019; Trink et al., 2021; Trink et al., 2022). PI3K/*Akt is* a key mediator of cell surface (cs)GRP78 signaling, a pathway known to facilitate TGFβ1 synthesis as well as ECM production in response to HG (Wu et al., 2009; Van Krieken et al., 2019). We have also shown that csGRP78 inhibition attenuates the HG-induced synthesis of TGFβ1 (Trink et al., 2022). However, whether csGRP78 also contributes to TGFβ1 activation in HG through regulation of its non-proteolytic activator TSP1 has not yet been determined and is addressed in these studies.
## 2.1 Cell culture
Primary MC from C57Bl/6 mice were isolated for culture using Dynabeads. They were cultured in DMEM (1,000 mg/L or 5.6 mM glucose) supplemented with $20\%$ FBS, 100 µg/mL streptomycin, and 100 µg/mL penicillin at 37°C in $95\%$ O2, $5\%$ CO2. Cells were serum-deprived in $0.5\%$ FBS for 24 h prior to treatment with HG (30 mM) with or without inhibitors of csGRP78 signaling: the C-terminus targeting GRP78 antibody C38 (2 µg/mL) (Munro and Pelham, 1987; De Ridder et al., 2012; Trink et al., 2022) or vaspin (100 ng/mL) (Nakatsuka et al., 2012; Nakatsuka et al., 2013; Abdolahi et al., 2022), or the following PI3K/Akt inhibitors: wortmannin (100 ng/mL), LY294002 (20 µM), and Akt Inhibitor VIII (10 µM).
## 2.2 Protein extraction and immunoblotting
MC protein extraction was previously described (Krepinsky et al., 2003). Protein expression was assessed using SDS-PAGE and immunoblotting. Antibodies used for Western blotting were: TSP1 (1:1,000, R&D Systems), pAkt Ser473 (1:1,000, Cell Signaling), total Akt (1:1,000, Cell Signaling), pSmad3 Ser$\frac{423}{425}$ (1:4,000, Novus), total Smad3 (1:1,000, Abcam), MTJ1 (1:1,000, Cedarlane), LAP-TGFβ1 (1:1,000, R&D Systems), GRP78 (1:1,000, BD Biosciences), α-tubulin (1:40,000, Sigma).
## 2.3 Luciferase and transfection
For transfection experiments, MC were plated at $50\%$ confluency and transfected with either 1 µg of the mouse TSP1 luciferase reporter construct [mTSP1-luciferase, a gift from P. Bornstein, Plasmid #12409, Addgene (Michaud-Levesque and Richard, 2009)] with 0.05 µg pCMV β-galactosidase (β-Gal, Clonetech) using Effectene (Qiagen) or 100 nM of MTJ1 or control siRNA (Silencer Select, ThermoFisher) using Lipofectamine (Invitrogen). After 18 h, cells were serum-deprived and treated as above for protein collection for siRNA experiments. For luciferase harvest, 1× Reporter Lysis Buffer (Promega) was added to the plate which was then stored at −80°C overnight prior to cell lysis. Luciferase activity was measured on clarified lysate using the Luciferase Assay System (Promega) with a luminometer (Junior LB 9509, Berthold). Β-Gal activity was used to normalize transfection efficiency, measured using the β-Galactosidase Enzyme Assay System (Promega) with a SpectraMax Plus 384 Microplate Reader (Molecular Devices) set to read absorbance at 420 nm.
The pcDNA3.1 plasmid which contains GRP78 lacking its ER retention sequence KDEL was transfected by electroporation as previously described (Trink et al., 2022). This was used to overexpress GRP78 at the cell surface. The empty vector pcDNA 3.1 was used as a control.
## 2.4 RNA extraction and qtPCR
RNA was extracted from MC using Trizol (Invitrogen), and 0.5 µg of RNA was reverse transcribed using qScript Supermix Reagent (Quanta Biosciences). Expression of TSP1 mRNA relative to 18S was determined using the ΔΔCt method. Quantitative PCR was performed using Power SYBR Green PCR Master Mix on the Applied Biosystems Vii 7 Real-Time PCR System. The following primers were used: TSP1 forward 5′-TGGCCAGCGTTGCCA -3′ and reverse 5′- TCTGCAGCACCCCCTGAA-3′ and 18S forward 5′- GCCGCTAGAGGTGAAATTCTTG-3′ and reverse 5′- CATTCTTGGCAAATGCTTTCG-3’.
## 2.5 ELISA for active TGFβ1
To measure biologically active TGFβ1 in conditioned MC media, the TGFβ1 Quantikine ELISA kit (R&D Systems) was used, with omission of the activation step described in the protocol.
## 2.6 TGFβ1 bioassay with mink lung epithelial cells (MLECs)
MLEC stably transfected with the PAI-1 luciferase promoter construct, generously provided by Dr. T. Tsuda, were used. MC and MLEC were cocultured in MEM with $10\%$ FBS, plated on a 12-well plate at 5,000 and 25,000 cells/well, respectively (1:5 ratio MLEC: MC). The following day, cells were serum deprived for 18 h followed by treatment with HG and various inhibitors. At collection, cells were lysed in 1× Reporter Lysis Buffer (Promega) and stored at −80°C overnight before analysis of PAI-1 luciferase activity as described above.
## 2.7 Extracellular matrix extraction
MC were lysed with $0.5\%$ sodium deoxycholate [DOC; $0.5\%$ DOC, 50 mM Tris pH 8.0, 150 mM NaCl, $1\%$ Nonidet P-40; as used in (Klingberg et al., 2014)] three times to allow complete removal of cells while maintaining ECM adhesion to plates. Plates were then washed twice with cold DOC lysis buffer and three times with cold 1xPBS. Lysis buffer (PBS pH 7.4, 5 mM EDTA, 5 mM EGTA, 10 mM sodium pyrophosphate, 50 mM NaF, 1 mM NaVO3, $1\%$ Triton) containing 1 mM DTT, 60 mM n-octyl glucopyranoside, and protease inhibitors was heated at 100°C for 2 min before being added to the plate. After thorough scraping, ECM lysate was transferred into an Eppendorf tube and boiled for an additional 10 min. Samples were assessed using SDS-PAGE and immunoblotting. An aliquot from the first DOC extraction of each sample was taken and run alongside the final ECM extraction to confirm the complete removal of cellular debris by probing for tubulin.
## 2.8 Surface protein co-immunoprecipitation from live cells
MC were washed three times with 1xPBS and then incubated in $1\%$ BSA with 5 µg anti-TSP1 antibody at 4°C for 2 h on a shaker set to low speed. Cells were then washed and lysed by passing through a 25-gauge needle (Precision Glide Needle, B) 5 times. Lysates were clarified and normalized with an equal amount of Protein G beads (rProtein G agarose, Invitrogen) added to each sample with gentle rocking overnight at 4°C. Samples were then washed in lysis buffer and eluted from the beads by boiling for 5 min in PSB. Samples were assessed by SDS-PAGE and immunoblotting.
## 2.9 Statistical analysis
All points presented in graphs represent individual data points. A two-tailed t-test or one-way ANOVA was used to analyze differences between two or more groups, respectively. Tukey’s post hoc analysis was used to compare differences between two or more groups. ImageJ was used for the quantification of experiments and GraphPad Prism 6.0 was used for the analysis of data. Statistical significance was set to $p \leq 0.05$ and data are presented as mean ± SEM.
## 3.1 PI3K/Akt signaling is required for HG-induced TSP1 regulation
Previously we showed that HG-induced PI3K/Akt activation requires csGRP78 in MC (Van Krieken et al., 2019; Trink et al., 2021). Since PI3K/Akt signaling was shown to mediate TSP1 expression by complement in MC (Wang et al., 2006), we hypothesized that csGRP78 would promote TSP1 expression in HG through this pathway. To first test whether PI3K/Akt are required for HG-induced TSP1 upregulation, we used the following inhibitors: LY294002 and wortmannin to inhibit PI3K activity and Akt inhibitor VIII to inhibit Akt activity. We previously showed that the osmotic control mannitol has no effect on the cell surface translocation of GRP78 or the upregulation and activation of α2M in MC (Van Krieken et al., 2019; Trink et al., 2021). Further, mannitol does not induce TSP1 expression (Yevdokimova et al., 2001). Thus, mannitol was omitted from these experiments. We observed significantly increased TSP1 expression in response to HG, which was attenuated by all three inhibitors (Figures 1A–C). We next assessed TSP1 transcript regulation. Similar to its protein expression, HG-induced TSP1 transcript upregulation was attenuated by the Akt inhibitor (Figure 1D). Next, we confirmed that TSP1 promoter activity was also regulated by PI3K/Akt signaling. Both PI3K inhibitors and the Akt inhibitor suppressed TSP1 promoter activation by HG. Some suppression of basal activity was also seen with LY294002 and Akt VIII (Figures 1E–G). Taken together, PI3K/*Akt is* required for HG-induced TSP1 regulation.
**FIGURE 1:** *HG-induced TSP1 upregulation requires PI3K/Akt. HG (48 h)-induced TSP1 upregulation in MC was attenuated by the PI3K inhibitors (A) LY294002 and (B) wortmannin as well as (C) Akt inhibitor VIII (n = 10, **p < 0.01, ***p < 0.005). (D) TSP1 transcript upregulation by HG (24 h) was inhibited by Akt inhibitor VIII (n = 4, *p < 0.05). HG (48 h)-induced TSP1 promoter activity, assessed using a luciferase reporter construct, was also inhibited by (E) LY294002 (n = 3, *p < 0.05, **p < 0.01, ***p < 0.005, ****p < 0.0001), (F) wortmannin (n = 9, *p < 0.05, **p < 0.01, ****p < 0.001), or (G) Akt inhibitor VIII (n = 9, ****p < 0.001) in MC.*
## 3.2 PI3K/Akt inhibition attenuates HG-induced TGFβ1 activation and signaling
Given that TSP1 is an important activator of TGFβ1 in response to HG (Yevdokimova et al., 2001), we next assessed whether PI3K/Akt inhibition would also abolish TGFβ1 activation and downstream signaling. In Figures 2A–C, the increase in active TGFβ1 in the medium seen with HG, as assessed by ELISA, was prevented by both PI3K and Akt inhibitors. Activation of the major mediator of TGFβ1 signaling Smad3 was then assessed by immunoblotting for its activated form, phosphorylated at its C-terminus Ser$\frac{473}{475.}$ As anticipated, PI3K and Akt inhibition blocked HG-induced Smad3 activation (Figures 2D–F). Thus, PI3K/Akt are required for TSP1-mediated activation of TGFβ1 and its downstream signaling.
**FIGURE 2:** *HG-induced TGFβ1 activation and signaling are blocked by PI3K/Akt inhibition. HG (48 h)-induced TGFβ1 activation, assessed by ELISA of medium without acid activation, was prevented by PI3K inhibition using either (A) LY294002 (n = 4, **p < 0.01, ***p < 0.005) or (B) wortmannin (n = 4, **p < 0.01, ****p < 0.0001), as well as Akt inhibition using (C) Akt inhibitor VIII (n = 4, **p < 0.01). Smad3 signaling downstream of TGFβ1 was assessed by its C-terminal phosphorylation at Ser473/475 (pSmad3). HG (48 h)-induced Smad3 phosphorylation was prevented by (D) LY294002, (E) wortmannin, and (F) Akt inhibitor VIII (n = 4, *p < 0.05).*
## 3.3 csGRP78 mediates HG-induced TSP1 expression through Akt activation
We next wished to determine whether csGRP78 was an upstream mediator of TSP1 regulation by HG through its activation of Akt. Cell surface GRP78 signaling was shown to be blocked by the C-terminus targeting GRP78 antibody C38 (Munro and Pelham, 1987; De Ridder et al., 2012) and the adipokine vaspin (visceral adipose tissue-derived serine proteinase inhibitor) in other settings (Nakatsuka et al., 2012; Nakatsuka et al., 2013). We thus used these to assess whether csGRP78 mediates TSP1 upregulation by HG. Figure 3A shows that C38, but not a control IgG, inhibits HG-induced TSP1 transcript upregulation. Similar inhibitory effects were seen on TSP1 promoter activation and increased protein expression by HG with both C38 and vaspin (Figures 3B–E). In Figures 3D, E, we confirmed that both csGRP78 inhibitors prevented HG-induced Akt activation, as assessed by its phosphorylation at Ser473. We further tested the effects of downregulating MTJ1, a co-chaperone required for GRP78 translocation to the cell surface in response to HG (Van Krieken et al., 2019). Short interfering (si)RNA knockdown of MTJ1 inhibited both HG-induced Akt activation and TSP1 upregulation (Figure 3F).
**FIGURE 3:** *Cell surface GRP78 mediates HG-induced TSP1 expression through Akt activation. (A) csGRP78 inhibition using the C38 antibody, but not a control IgG (2 µg for each antibody) prevented HG (24 h)-induced TSP1 transcript upregulation (n = 6–8, *p < 0.05). HG (48 h)-induced TSP1 promoter activity, assessed using a luciferase reporter construct, was also inhibited by csGRP78 inhibition using either (B) the C38 antibody (n = 3, **p < 0.01, ***p < 0.005, ****p < 0.0001) or (C) vaspin (n = 3, **p < 0.01, ***p < 0.005, ****p < 0.001). Further, csGRP78 inhibition also prevented HG (48 h)-induced Akt activation assessed by its phosphorylation at Ser473 as well as TSP1 upregulation using (D) C38, but not control IgG, antibody (n = 5, *p < 0.05, **p < 0.01), (E) vaspin (n = 10, *p < 0.05, **p < 0.01, ***p < 0.005) and (F) siRNA knockdown of MTJ1, the chaperone required for cell surface translocation of GRP78 in HG (n = 3, *p < 0.05, **p < 0.01). (G) Antibody neutralization of the known ligand and activator for csGRP78, α2M*, prevented HG (48 h)-induced TSP1 upregulation. Control IgG had no effect (10 µg for each antibody, n = 3, *p < 0.05). (H) An inhibitory peptide (Pep) that prevents the interaction between α2M* and csGRP78 also inhibited TSP1 upregulation by HG (48 h), with scrambled peptide (Scr) having no effect (100 nM, n = 5, *p < 0.05, **p < 0.01, ***p < 0.005).*
Lastly, we previously showed that the high-affinity ligand for csGRP78, activated alpha 2-macroglobulin (denoted α2M*), mediates PI3K/Akt activation in response to HG through csGRP78 (Trink et al., 2021). Using a neutralizing antibody specific for α2M*, as well as an inhibitory peptide that prevents the interaction between csGRP78 and α2M*, we show that HG-induced TSP1 upregulation is attenuated by α2M* inhibition (Figures 3G, H). Altogether, these data show the importance of α2M*-csGRP78 in the upregulation of TSP1 through Akt signaling in response to HG.
## 3.4 Deposition of TSP1 in the ECM is mediated by csGRP78
Increased deposition of TSP1 into the ECM under HG conditions has previously been shown (Yevdokimova et al., 2001; Xu et al., 2020). Here we investigated whether its deposition is mediated by csGRP78. In Figures 4A, B, inhibition of csGRP78 by C38 or vaspin prevented ECM deposition of TSP1 stimulated by HG. Similar effects were observed with downregulation of MTJ1 to inhibit GRP78 cell surface translocation (Figure 4C). Lastly, the localization of TSP1 in the ECM was visualized using immunofluorescence. As detailed in Methods, cells were removed after treatment, with the remaining DOC-insoluble matrix assessed for TSP1 presence. Inhibition of csGRP78 by the C38 antibody, but not the isotype control IgG, decreased HG-induced TSP1 deposition in the ECM (Figure 4D). Thus, csGRP78 is crucial for TSP1 upregulation as well as ECM deposition in response to HG in MC.
**FIGURE 4:** *Deposition of TSP1 in the ECM is mediated by csGRP78. ECM was extracted as DOC-insoluble material after HG for 48 h. The increased deposition of TSP1 into the ECM by MC was prevented by csGRP78 inhibition using (A) C38 antibody, but not control IgG antibody, (B) vaspin and (C) MTJ1 knockdown with siRNA (n = 3 for each experiment). (D) Immunofluorescent staining of remaining ECM. Prior to decellularization, cells were treated with HG for 48 h, causing increased TSP1 deposition into the matrix. This was prevented by C38, but not control IgG (n = 6, **p < 0.01, ***p < 0.005, ****p < 0.0001).*
## 3.5 HG-induced TGFβ1 activation requires csGRP78
Since TSP1 is an important regulator of HG-induced TGFβ1 activation in MC and diabetic kidneys (Poczatek et al., 2000; Wahab et al., 2005; Hohenstein et al., 2008; Lu et al., 2011), and csGRP78 is required for its upregulation, we next investigated whether csGRP78 would be required for TGFβ1 activation by HG. In Figures 5A–C we observed a marked increase in the activation of TGFβ1 in the medium in response to HG. This was prevented by csGRP78 inhibitors C38 antibody and vaspin, as well as the knockdown of MTJ1. We further assessed TGFβ1 activation functionally by using MLECs that are stably transfected with the Smad3-responsive PAI-1 luciferase construct. We first established that PAI-1 luciferase activity is not increased in MLEC in response to HG (not shown). We then co-incubated MLEC with MC. After treatment with HG, increased luciferase activity was observed (Figures 5D–F), reflecting increased bioactive TGFβ1 in the medium. Inhibition of csGRP78 with the C38 antibody (Figure 5D) or vaspin (Figure 5E), as well as inhibition of α2M* using a neutralizing antibody (Figure 5F) or an inhibitory peptide (Figure 5G) all prevented HG-induced TGFβ1 activation. We further confirmed that inhibition of csGRP78 also abrogated TGFβ1 downstream signaling in MC exposed to HG. Figure 5H shows that csGRP78 inhibition with vaspin abrogated HG-induced activation of Smad3 as assessed by its phosphorylation at Ser$\frac{473}{475.}$ Knockdown of MTJ1 similarly prevented Smad3 activation in HG (Figure 5I).
**FIGURE 5:** *HG-induced TGFβ1 activation requires csGRP78. Assessment of active TGFβ1 in the medium by ELISA showed that csGRP78 inhibition by (A) C38, but not a control IgG antibody (n = 8, ****p < 0.0001, (B) vaspin (n = 4, **p < 0.01, ***p < 0.005) or (C) MTJ1 siRNA knockdown (n = 4, **p < 0.01, ***p < 0.005) prevented HG (48 h)-induced TGFβ1 activation. (D–F) MC were co-cultured with mink lung epithelial cells (MLEC) stably transfected with the Smad3-regulated PAI-1 promoter luciferase. The HG (48 h)-induced increase in PAI-1 luciferase activation, reflecting bioactive TGFβ1, was prevented by csGRP78 inhibition with (D) C38, but not control IgG antibody or (E) vaspin, as well as by α2M* inhibition with (F) a neutralizing antibody, but not control IgG (10 µg) or (G) a peptide that prevents csGRP78/α2M* interaction (100 nM, n = 12, *p < 0.05, **p < 0.01, ***p < 0.005, ****p < 0.0001). In MC, signaling downstream of TGFβ1 was assessed by immunoblotting for Smad3 phosphorylated at Ser473/475. Inhibition of csGRP78 using both (H) vaspin and (I) MTJ1 siRNA knockdown prevented Smad3 activation in HG (n = 3, *p < 0.05, **p < 0.01). (J) TSP1 was immunoprecipitated from live cells after HG (48 h) using C38, with IgG used as a control. Immunoblotting shows association with both LAP and csGRP78 in response to HG (n = 3, ****p < 0.0001).*
In the ECM, TSP1 dual interaction with LAP and the mature TGFβ1 in its latent complex enables TGFβ1 activation (Crawford et al., 1998; Yevdokimova et al., 2001). We previously showed that HG induces interaction between csGRP78 and LAP (Trink et al., 2022). We thus sought to determine whether csGRP78 could also interact with TSP1. We immunoprecipitated TSP1 from live cell cultures to isolate the cell surface/extracellular TSP1 as outlined in Methods. Figure 5J shows that HG induces an interaction between TSP1, LAP, and csGRP78 which is prevented by csGRP78 inhibition with the C38 antibody. Overall, these data support an important role for csGRP78 in regulating the activation of TGFβ1 in response to HG and suggest that physical interaction at the cell surface is likely important for this regulation.
## 3.6 Overexpression of csGRP78 augments TSP1 production and TGFβ1 activation in HG
We have previously shown that overexpressing GRP78 lacking the ER retention sequence KDEL (GRP78 ΔKDEL) increases GRP78 at the cell surface, profibrotic signaling, and matrix production and augments HG responses (Trink et al., 2021). Here we wanted to assess whether csGRP78 overexpression could also augment TSP1 production and TGFβ1 activation. In Figure 6A, we observed increased basal activity of the TSP1 promoter luciferase in cells expressing GRP78 ΔKDEL, at a level similar to that of HG-induced TSP1 luciferase activity. HG further augmented TSP1 promoter activity, although this did not reach statistical significance. We next assessed the effects on the TSP1 protein. Similarly to TSP1 luciferase, protein was increased by GRP78 ΔKDEL alone and further increased with HG (Figure 6B). In parallel, we observed an increase in active TGFβ1 in the medium (Figure 6C) and an increase in TGFβ1 biologic activity assessed using the MLEC system described above (Figure 6D) with GRP78 ΔKDEL alone, also augmented by HG.
**FIGURE 6:** *csGRP78 overexpression augments TSP1 production and TGFβ1 activation. Overexpression of GRP78 ΔKDEL, which increases csGRP78, increased basal and HG (48 h)-induced TSP1 (A) promoter activity and (B) protein expression (n = 6, *p < 0.05, **p < 0.01, ***p < 0.005, ****p < 0.0001). (C) Activation of TGFβ1 measured by ELISA on medium, as well as its downstream signaling assessed using (D) MC co-cultured with MLEC stably expressing PAI-1 luciferase, showed that GRP78 ΔKDEL transfection increased basal and HG (48 h)-induced TGFβ1 activity (n = 9, *p < 0.05, ****p < 0.0001). The induction of TSP1 expression by GRP78 ΔKDEL was attenuated by α2M* inhibition using either (E) a neutralizing antibody (10 µg) or (F) a peptide preventing csGRP78/α2M* interaction (100 nM) (n = 4, **p < 0.01).*
Our previous data show that α2M* is required not only for HG responses, but also for csGRP78 effects seen with overexpression of GRP78 ΔKDEL (Trink et al., 2021). To determine whether this extends to TSP1 effects, we treated cells in which csGRP78 was overexpressed using GRP78 ΔKDEL with α2M* inhibitors. Both its neutralizing antibody (Figure 6E) and the inhibitory peptide (Figure 6F) blocked GRP78 ΔKDEL-induced expression of TSP1, showing the importance of this ligand-receptor pair in regulating TSP1 production.
## 4 Discussion
We have previously shown the de novo expression of csGRP78 in DKD and the importance of HG-induced GRP78 translocation to the cell surface to mediate profibrotic signaling in MC (Van Krieken et al., 2019; Trink et al., 2021). While these studies showed that csGRP78 was important for HG-induced TGFβ1 upregulation (Trink et al., 2021; Trink et al., 2022), whether it also regulated activation of TGFβ1 from its latent state was not clear. In these follow-up studies, we chose to focus on TSP1 given its central role in the non-proteolytic activation of TGFβ1 by HG and in DKD, as well as the limited information on how TSP1 production is regulated in this setting. We now present evidence, summarized in Figure 7, that csGRP78 facilitates the production and consequent extracellular deposition of TSP1 through PI3K/Akt signaling in response to HG, with inhibition of this pathway preventing the bioactivation of TGFβ1. These novel data provid further support for the potential therapeutic value of inhibiting csGRP78 to prevent the fibrotic phenotype seen in DKD. Indeed, direct TGFβ1 inhibition is not feasible due to its pleiotropic homeostatic effects. This was shown in a clinical trial evaluating the efficacy of a neutralizing TGFβ1 antibody in patients with DKD, in which dosing that limited adverse effects was ineffective in slowing DKD (Voelker et al., 2017). Indirect methods of TGFβ1 attenuation targeting disease-specific abnormally heightened signaling, while leaving homeostatic signaling undisturbed, are thus of high therapeutic interest. Our study supports csGRP78 as a potential anti-fibrotic therapeutic target that could provide a disease-specific mechanism by which to inhibit TGFβ1 profibrotic signaling in DKD. Current studies are underway to test the efficacy of csGRP78 inhibition in an in vivo DKD model.
**FIGURE 7:** *Proposed role for csGRP78 in mediated TGFβ1 activation through regulation of TSP1 in MC. HG promotes GRP78 localization on the cell surface. Increased activation of the protease inhibitor α2M (denoted α2M*) enables its high-affinity interaction with csGRP78. This induces downstream PI3K/Akt activation and consequent TSP1 gene upregulation. The increased TSP1 localizes to the ECM, where others have shown it to interact with latent TGFβ1 and serve as an important regulator of TGFβ1 activation. The nature and role of csGRP78 interaction with this complex requires further study. These data support a role for csGRP78/α2M* in mediating TGFβ1 profibrotic activation in MC by HG, thus highlighting a potential therapeutic target for indirect TGFβ1 inhibition. Created with BioRender.*
Previous studies have shown that TSP1 is increased in MC and some non-kidney cells in response to HG as well as other stimuli relevant to DKD such as angiotensin II (Poczatek et al., 2000; Naito et al., 2004; Zhou et al., 2006). Importantly, TSP1 upregulation was also seen in both murine models of DKD and human type 1 and 2 diabetic kidneys (Yevdokimova et al., 2001; Wahab et al., 2005; Hohenstein et al., 2008; Lu et al., 2011; Murphy-Ullrich, 2019). Several studies have shown the pathologic importance of TSP1 as an activator of latent TGFβ1, contributing to the pathogenesis of fibrosis that is characteristic of DKD. Notably, TSP1 was shown to be expressed in a mesangial pattern in DKD (Daniel et al., 2007). In cultured MC, HG-induced TGFβ1 activation was blocked by a peptide that inhibited interaction between TSP1 and latent TGFβ1 (Yevdokimova et al., 2001; Lu et al., 2011). TSP1 knockout mice with type 1 diabetes induced by streptozotocin showed reduced DKD compared to their wild-type counterparts. Less glomerular matrix accumulation, inflammation, and proteinuria were associated with reduced active glomerular TGFβ1 and downstream signaling, with no difference in total TGFβ1 levels (Daniel et al., 2007). In another study, inhibition of latent TGFβ1-TSP1 interaction with a specific peptide attenuated DKD in type 1 diabetic Akita mice, with a reduction in proteinuria, urinary active TGFβ1 and Smad$\frac{2}{3}$ activation. Importantly, peptide treatment also reduced tubulointerstitial fibrosis, a pathologic finding seen in the later stages of DKD (Lu et al., 2011). TSP1 thus contributes to both early and later-stage DKD development through its regulation of TGFβ1 activation.
Several studies suggest that the important role of TSP1 in kidney fibrosis extends beyond DKD. For example, increased TSP1 was found in the kidneys of remnant rats, a model of reduced kidney mass and fibrosis as is seen in chronic kidney disease of any etiology. This increase preceded and was predictive of the development of tubular interstitial fibrosis (Hugo et al., 2002). Additionally, TSP1 deficiency was found to be protective against the development of proteinuria, fibrosis, and inflammation induced by the chemotherapeutic agent adriamycin in mice (Maimaitiyiming et al., 2016). Interstitial fibrosis that developed following unilateral ureteral obstruction was significantly attenuated by a peptide inhibiting TSP1/latent TGFβ1 interaction, and this was associated with a reduction in active TGFβ1 and Smad activation (Xie et al., 2010). Finally, TSP1 was also found to contribute to the longer-term development of fibrosis after acute injury in an ischemia/reperfusion model, again associated with TGFβ1 activation (Julovi et al., 2020). While requiring assessment, it is probable that csGRP78 is increased in response to various pathogenic stimuli in kidney cells in addition to HG. Indeed, our unpublished data show that angiotensin II, a common pathogenic mediator of fibrosis in diabetic and non-diabetic chronic kidney disease, also induced GRP78 localization to the cell surface in MC. The effect of additional stimuli in MC and other kidney cell types, as well as the localization of GRP78 to the cell surface in various non-DKD models, will be determined in future studies.
While the upregulation of TSP1 by HG in MC and DKD is well documented, the mechanism by which this occurs is less well-defined. An important role for cGMP-dependent protein kinase G (PKG) has been identified, with PKG normally repressing TSP1 transcription. PKG is activated by nitric oxide signaling to activate soluble guanylate cyclase, thereby increasing cGMP levels. In MC exposed to HG, a reduction in nitric oxide in the medium and intracellular cGMP was seen, and a nitric oxide donor or overexpression of a constitutively active PKG prevented HG-induced TSP1 upregulation and TGFβ1 activation (Wang et al., 2002; Wang et al., 2003). TSP1 upregulation in HG was mediated by the transcription factor upstream stimulatory factor 2 (USF2), which under normal glucose conditions is repressed by PKG (Wang et al., 2004). In our studies, we showed an important role for PI3K/Akt activation in TSP1 upregulation. This pathway was also shown to mediate complement-induced TSP1 upregulation and consequent TGFβ1 activation in MC (Wang et al., 2006). Interestingly, increased USF2 production by HG in MC was shown to be dependent on the transcription factor CREB, well known to be activated by Akt phosphorylation (Du and Montminy, 1998; Shi et al., 2008). Future studies will explore whether csGRP78/Akt regulation of TSP1 intersects with PKG/USF2 signaling.
Our previous studies have shown that the high-affinity ligand for csGRP78, α2M*, is locally produced by MC in HG and in DKD to enable PI3K/Akt signaling and downstream ECM production (Trink et al., 2021). We showed not only its activation in the diabetic kidney, but also its upregulation at the protein and transcript levels. Additional data available in the RNA-seq database NephroSeq also show a disease-specific increase in α2M in DKD patients in both glomeruli and the tubulointerstitium compared to healthy control patients (Supplementary Figure S1) (Ju et al., 2013; Ju et al., 2015). Further, α2M was shown to be increased in human endothelial cells and MC in patients with focal segmental glomerular sclerosis (FSGS), another fibrotic disease of the kidney. This study showed that α2M was specifically upregulated in disease and its higher expression was associated with poorer kidney outcomes (Menon et al., 2020), suggesting the importance of α2M* signaling may also be relevant to non-diabetic kidney disease. Our current data extend these observations by showing that α2M* is also required for csGRP78-mediated TSP1 upregulation and TGFβ1 activation in HG. Furthermore, as csGRP78 is positioned entirely extracellularly, we identified integrin β1 as a transmembrane signaling partner for csGRP78 that enables intracellular signal transmission and profibrotic responses (Van Krieken et al., 2019; Trink et al., 2022). Interestingly, TSP1 was also shown to bind integrin β1 (Calzada et al., 2004). Whether activation of TGFβ1 by TSP1 in HG requires binding to integrin β1 and/or csGRP78 needs further study, as do the details of their interaction at a molecular level.
In conclusion, we have shown that HG-induced csGRP78 and α2M* mediate TGFβ1 activation and downstream signaling through the regulation of TSP1. This pathway provides an indirect method of TGFβ1 inhibition that is entirely extracellular, and thus an attractive potential therapeutic target. Further studies will investigate whether inhibition of csGRP78, α2M*, or their interaction is an effective approach to DKD treatment and prevention.
## Data availability statement
The raw data supporting the conclusion of this article will be made available by the authors, without undue reservation.
## Author contributions
JT, UA, RL, KO’N, and BG performed experiments; JK conceived the experimental design; JT analysed the data; JT and JK wrote the manuscript. All authors have read and agreed to the published version of the manuscript.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fphar.2023.1098321/full#supplementary-material
## References
1. Abdolahi A., Vahabzadeh Z., Izadpanah E., Moloudi M. R.. **‘Vaspin attenuates steatosis-induced fibrosis via GRP78 receptor by targeting AMPK signaling pathway’**. *J. Physiology Biochem.* (2022) **7878** 185-197. DOI: 10.1007/S13105-021-00852-7
2. Alicic R. Z., Rooney M. T., Tuttle K. R.. **‘Diabetic kidney disease: Challenges, progress, and possibilities’**. *Clin. J. Am. Soc. Nephrol.* (2017) **12** 2032-2045. DOI: 10.2215/CJN.11491116
3. Calzada M. J., Annis D. S., Zeng B., Marcinkiewicz C., Banas B., Lawler J.. **‘Identification of novel beta1 integrin binding sites in the type 1 and type 2 repeats of thrombospondin-1’**. *J. Biol. Chem.* (2004) **279** 41734-41743. DOI: 10.1074/JBC.M406267200
4. Castoldi G., di Gioia C. R. T., Bombardi C., Preziuso C., Leopizzi M., Maestroni S.. **‘Renal antifibrotic effect of N-Acetyl-Seryl-Aspartyl-Lysyl-Proline in diabetic rats’**. *Am. J. Nephrol.* (2013) **37** 65-73. DOI: 10.1159/000346116
5. Cho J. G., Lee A., Chang W., Lee M. S., Kim J.. **‘Endothelial to mesenchymal transition represents a key link in the interaction between inflammation and endothelial dysfunction’**. *Front. Immunol.* (2018) **9** 294. DOI: 10.3389/FIMMU.2018.00294
6. Crawford S. E., Stellmach V., Murphy-Ullrich J. E., Ribeiro S. M., Lawler J., Hynes R. O.. **‘Thrombospondin-1 is a major activator of TGF-beta1**. *Cell* (1998) **93** 1159-1170. DOI: 10.1016/S0092-8674(00)81460-9
7. Daniel C., Schaub K., Amann K., Lawler J., Hugo C.. **‘Thrombospondin-1 is an endogenous activator of TGF-beta in experimental diabetic nephropathy**. *Diabetes* (2007) **56** 2982-2989. DOI: 10.2337/DB07-0551
8. De Ridder G. G., Ray R., Pizzo S. V.. **‘A murine monoclonal antibody directed against the carboxyl-terminal domain of GRP78 suppresses melanoma growth in mice’**. *Melanoma Res.* (2012) **22** 225-235. DOI: 10.1097/CMR.0b013e32835312fd
9. Du K., Montminy M.. **‘CREB is a regulatory target for the protein kinase Akt/PKB’**. *J. Biol. Chem.* (1998) **273** 32377-32379. DOI: 10.1074/JBC.273.49.32377
10. Edeling M., Ragi G., Huang S., Pavenstadt H., Susztak K.. **‘Developmental signalling pathways in renal fibrosis: The roles of notch, Wnt and hedgehog’**. *Nat. Rev. Nephrol.* (2016) **712** 426-439. DOI: 10.1038/nrneph.2016.54
11. Hohenstein B., Daniel C., Hausknecht B., Boehmer K., Riess R., Amann K. U.. **‘Correlation of enhanced thrombospondin-1 expression, TGF-beta signalling and proteinuria in human type-2 diabetic nephropathy’**. *Nephrol. Dial. Transplant.* (2008) **23** 3880-3887. DOI: 10.1093/NDT/GFN399
12. Hugo C., Kang D. H., Johnson R. J.. **‘Sustained expression of thrombospondin-1 is associated with the development of glomerular and tubulointerstitial fibrosis in the remnant kidney model’**. *Nephron* (2002) **90** 460-470. DOI: 10.1159/000054735
13. Hugo C.. **‘The thrombospondin 1–TGF‐β axis in fibrotic renal disease’**. *Nephrol. Dial. Transplant.* (2003) **18** 1241-1245. DOI: 10.1093/NDT/GFG159
14. Johnson S. A., Spurney R. F.. **‘Twenty years after ACEIs and ARBs: Emerging treatment strategies for diabetic nephropathy ’**. *Am. J. physiology. Ren. physiology* (2015) **309** F807-F820. DOI: 10.1152/ajprenal.00266.2015
15. Ju W., Greene C. S., Eichinger F., Nair V., Hodgin J. B., Bitzer M.. **‘Defining cell-type specificity at the transcriptional level in human disease’**. *Genome Res.* (2013) **23** 1862-1873. DOI: 10.1101/GR.155697.113
16. Ju W., Nair V., Smith S., Zhu L., Shedden K., Song P. X. K.. **‘Tissue transcriptome-driven identification of epidermal growth factor as a chronic kidney disease biomarker’**. *Sci. Transl. Med.* (2015) **7** 316ra193. DOI: 10.1126/SCITRANSLMED.AAC7071
17. Julovi S. M., Sanganeria B., Minhas N., Ghimire K., Nankivell B., Rogers N. M.. **‘Blocking thrombospondin-1 signaling via CD47 mitigates renal interstitial fibrosis’**. *Lab. Investig.* (2020) **100** 1184-1196. DOI: 10.1038/s41374-020-0434-3
18. Kato M., Yuan H., Xu Z. G., Lanting L., Li S. L., Wang M.. **‘Role of the akt/FoxO3a pathway in TGF-beta1-mediated mesangial cell dysfunction: A novel mechanism related to diabetic kidney disease’**. *J. Am. Soc. Nephrol.* (2006) **17** 3325-3335. DOI: 10.1681/ASN.2006070754
19. Klingberg F., Chow M. L., Koehler A., Boo S., Buscemi L., Quinn T. M.. **‘Prestress in the extracellular matrix sensitizes latent TGF-β1 for activation’**. *J. Cell Biol.* (2014) **207** 283-297. DOI: 10.1083/jcb.201402006
20. Kolavennu V., Zeng L., Peng H., Wang Y., Danesh F. R.. **‘Targeting of RhoA/ROCK signaling ameliorates progression of diabetic nephropathy independent of glucose control’**. *Diabetes* (2008) **57** 714-723. DOI: 10.2337/DB07-1241
21. Koli K., Saharinen J., Hyytiäinen M., Penttinen C., Keski-Oja J.. **Latency, activation, and binding proteins of TGF-beta**. *Microsc. Res. Tech.* (2001) **52** 354-362. DOI: 10.1002/1097-0029(20010215)52:4<354::AID-JEMT1020>3.0.CO;2-G
22. Krepinsky J. C., Ingram A. J., Tang D., Wu D., Liu L., Scholey J. W.. **‘Nitric oxide inhibits stretch-induced MAPK activation in mesangial cells through RhoA inactivation’**. *J. Am. Soc. Nephrol.* (2003) **14** 2790-2800. DOI: 10.1097/01.ASN.0000094085.04161.A7
23. Lebleu V. S., Taduri G., O'Connell J., Teng Y., Cooke V. G., Woda C.. **‘Origin and function of myofibroblasts in kidney fibrosis’**. *Nat. Med.* (2013) **19** 1047-1053. DOI: 10.1038/nm.3218
24. Li J. H., Huang X. R., Zhu H. J., Johnson R., Lan H. Y.. **‘Role of TGF-β signaling in extracellular matrix production under high glucose conditions’**. *Kidney Int.* (2003) **63** 2010-2019. DOI: 10.1046/j.1523-1755.2003.00016.x
25. Lin J. R., Zheng Y. J., Zhang Z. B., Shen W. L., Li X. D., Wei T.. **‘Suppression of endothelial-to-mesenchymal transition by SIRT (sirtuin) 3 alleviated the development of hypertensive renal injury’**. *Hypertens. (Dallas, Tex* (2018) **72** 350-360. DOI: 10.1161/HYPERTENSIONAHA.118.10482
26. Locatelli M., Zoja C., Zanchi C., Corna D., Villa S., Bolognini S.. **‘Manipulating Sirtuin 3 pathway ameliorates renal damage in experimental diabetes’**. *Sci. Rep.* (2020) **110** 8418-8430. DOI: 10.1038/s41598-020-65423-0
27. Lu A., Miao M., Schoeb T. R., Agarwal A., Murphy-Ullrich J. E.. **‘Blockade of TSP1-dependent TGF-β activity reduces renal injury and proteinuria in a murine model of diabetic nephropathy’**. *Am. J. Pathology* (2011) **178** 2573-2586. DOI: 10.1016/J.AJPATH.2011.02.039
28. Maimaitiyiming H., Zhou Q., Wang S.. **‘Thrombospondin 1 deficiency ameliorates the development of adriamycin-induced proteinuric kidney disease’**. *PLOS ONE* (2016) **11** e0156144. DOI: 10.1371/JOURNAL.PONE.0156144
29. Menon R., Otto E. A., Hoover P., Eddy S., Mariani L., Godfrey B.. **‘Single cell transcriptomics identifies focal segmental glomerulosclerosis remission endothelial biomarker ’**. *JCI insight* (2020) **5** e133267. DOI: 10.1172/jci.insight.133267
30. Michaud-Levesque J., Richard S.. **‘Thrombospondin-1 is a transcriptional repression target of PRMT6’**. *J. Biol. Chem.* (2009) **284** 21338-21346. DOI: 10.1074/JBC.M109.005322
31. Munger J. S., Harpel J. G., Gleizes P. E., Mazzieri R., Nunes I., Rifkin D. B.. **‘Latent transforming growth factor-beta: Structural features and mechanisms of activation’**. *Kidney Int.* (1997) **51** 1376-1382. DOI: 10.1038/KI.1997.188
32. Munro S., Pelham H. R. B.. **‘A C-terminal signal prevents secretion of luminal ER proteins’**. *Cell* (1987) **48** 899-907. DOI: 10.1016/0092-8674(87)90086-9
33. Murphy-Ullrich J. E., Suto M. J.. **‘Thrombospondin-1 regulation of latent TGF-β activation: A therapeutic target for fibrotic disease’**. *Matrix Biol.* (2018) **68** 28-43. DOI: 10.1016/j.matbio.2017.12.009
34. Murphy-Ullrich J. E.. **‘Thrombospondin 1 and its diverse roles as a regulator of extracellular matrix in fibrotic disease’**. *J. Histochem. Cytochem.* (2019) **67** 683-699. DOI: 10.1369/0022155419851103
35. Naito T., Masaki T., Nikolic-Paterson D. J., Tanji C., Yorioka N., Kohno N.. **‘Angiotensin II induces thrombospondin-1 production in human mesangial cells via p38 mapk and jnk: A mechanism for activation of latent TGF-β 1’**. *Am. J. Physiology - Ren. Physiology* (2004) **286** 278-287. DOI: 10.1152/ajprenal.00139.2003
36. Nakatsuka A., Wada J., Iseda I., Teshigawara S., Higashio K., Murakami K.. **‘Vaspin is an adipokine ameliorating ER stress in obesity as a ligand for cell-surface GRP78/MTJ-1 complex’**. *Diabetes* (2012) **61** 2823-2832. DOI: 10.2337/db12-0232
37. Nakatsuka A., Wada J., Iseda I., Teshigawara S., Higashio K., Murakami K.. **‘Visceral adipose tissue-derived serine proteinase inhibitor inhibits apoptosis of endothelial cells as a ligand for the cell-surface GRP78/voltage-dependent anion channel complex’**. *Circulation Res.* (2013) **112** 771-780. DOI: 10.1161/CIRCRESAHA.111.300049
38. Nikolic-Paterson D. J., Wang S., Lan H. Y.. **‘Macrophages promote renal fibrosis through direct and indirect mechanisms’**. *Kidney Int. Suppl.* (2014) **4** 34-38. DOI: 10.1038/KISUP.2014.7
39. Omata M., Taniguchi H., Koya D., Kanasaki K., Sho R., Kato Y.. **‘N-acetyl-seryl-aspartyl-lysyl-proline ameliorates the progression of renal dysfunction and fibrosis in WKY rats with established anti-glomerular basement membrane nephritis’**. *J. Am. Soc. Nephrol. JASN* (2006) **17** 674-685. DOI: 10.1681/ASN.2005040385
40. Peng H., Li Y., Wang C., Zhang J., Chen Y., Chen W.. **‘ROCK1 induces endothelial-to-mesenchymal transition in glomeruli to aggravate albuminuria in diabetic nephropathy’**. *Sci. Rep.* (2016) **6** 20304. DOI: 10.1038/SREP20304
41. Poczatek M. H., Hugo C., Darley-Usmar V., Murphy-Ullrich J. E.. **‘Glucose stimulation of transforming growth factor-beta bioactivity in mesangial cells is mediated by thrombospondin-1’**. *Am. J. pathology* (2000) **157** 1353-1363. DOI: 10.1016/S0002-9440(10)64649-4
42. Reidy K., Kang H. M., Hostetter T., Susztak K.. **‘Molecular mechanisms of Diabetic kidney disease’**. *J. Clin. Invest* (2014) **124** 2333-2340. DOI: 10.1172/JCI72271
43. Sagoo M. K., Gnudi L.. **‘Diabetic nephropathy: An overview’**. *Methods Mol. Biol.* (2020) **2067** 3-7. DOI: 10.1007/978-1-4939-9841-8_1
44. Shi L., Liu S., Nikolic D., Wang S.. **‘High glucose levels upregulate upstream stimulatory factor 2 gene transcription in mesangial cells’**. *J. Cell. Biochem.* (2008) **103** 1952-1961. DOI: 10.1002/JCB.21585
45. Trink J., Li R., Squire E., O'Neil K., Zheng P., Gao B.. **‘Integrin β1/cell surface GRP78 complex regulates TGFβ1 and its profibrotic effects in response to high glucose’**. *Biomedicines* (2022) **10** 2247. DOI: 10.3390/biomedicines10092247
46. Trink J., Li R., Palarasah Y., Troyanov S., Andersen T. E., Sidelmann J. J.. **‘Activated alpha 2-macroglobulin is a novel mediator of mesangial cell profibrotic signaling in diabetic kidney disease’**. *Biomedicines* (2021) **9** 1112. DOI: 10.3390/biomedicines9091112
47. Tuttle K. R., Brosius F. C., Cavender M. A., Fioretto P., Fowler K. J., Heerspink H. J. L.. **‘SGLT2 inhibition for CKD and cardiovascular disease in type 2 diabetes: Report of a scientific workshop sponsored by the national kidney foundation’**. *Am. J. Kidney Dis.* (2020) **77** 94-109. DOI: 10.1053/j.ajkd.2020.08.003
48. Van Krieken R., Mehta N., Wang T., Zheng M., Li R., Gao B.. **‘Cell surface expression of 78-kDa glucose-regulated protein (GRP78) mediates diabetic nephropathy’**. *J. Biol. Chem.* (2019) **294** 7755-7776. DOI: 10.1074/jbc.RA118.006939
49. Verrecchia F., Chu M. L., Mauviel A.. **‘Identification of novel TGF-beta/Smad gene targets in dermal fibroblasts using a combined cDNA microarray/promoter transactivation approach’**. *J. Biol. Chem.* (2001) **276** 17058-17062. DOI: 10.1074/JBC.M100754200
50. Voelker J., Berg P. H., Sheetz M., Duffin K., Shen T., Moser B.. **‘Anti-TGF-**. *J. Am. Soc. Nephrol.* (2017) **28** 953-962. DOI: 10.1681/ASN.2015111230
51. Wahab N. A., Schaefer L., Weston B. S., Yiannikouris O., Wright A., Babelova A.. **‘Glomerular expression of thrombospondin-1, transforming growth factor beta and connective tissue growth factor at different stages of diabetic nephropathy and their interdependent roles in mesangial response to diabetic stimuli’**. *Diabetologia* (2005) **48** 2650-2660. DOI: 10.1007/S00125-005-0006-5
52. Wang S., Shiva S., Poczatek M. H., Darley-Usmar V., Murphy-Ullrich J. E.. **‘Nitric oxide and cGMP-dependent protein kinase regulation of glucose-mediated thrombospondin 1-dependent transforming growth factor-beta activation in mesangial cells’**. *J. Biol. Chem.* (2002) **277** 9880-9888. DOI: 10.1074/JBC.M108360200
53. Wang S., Skorczewski J., Feng X., Mei L., Murphy-Ullrich J. E.. **‘Glucose up-regulates thrombospondin 1 gene transcription and transforming growth factor-beta activity through antagonism of cGMP-dependent protein kinase repression via upstream stimulatory factor 2’**. *J. Biol. Chem.* (2004) **279** 34311-34322. DOI: 10.1074/JBC.M401629200
54. Wang S., Wu X., Lincoln T. M., Murphy-Ullrich J. E.. **‘Expression of constitutively active cGMP-dependent protein kinase prevents glucose stimulation of thrombospondin 1 expression and TGF-activity’**. *Diabetes* (2003) **52** 2144-2150. DOI: 10.2337/diabetes.52.8.2144
55. Wang Y., Qiu W., Xu W., Xu J., Tong J.. **‘Sublytic complement C5b-9 complexes induce thrombospondin-1 production in rat glomerular mesangial cells via PI3-k/Akt: Association with activation of latent transforming growth factor-beta1’**. *Clin. Exp. Immunol.* (2006) **144** 326-334. DOI: 10.1111/j.1365-2249.2006.03069.x
56. Worthington J. J., Klementowicz J. E., Travis M. A.. **‘TGFβ: A sleeping giant awoken by integrins’**. *Trends Biochem. Sci.* (2011) **36** 47-54. DOI: 10.1016/J.TIBS.2010.08.002
57. Wu D., Peng F., Zhang B., Ingram A. J., Kelly D. J., Gilbert R. E.. **‘PKC-beta1 mediates glucose-induced Akt activation and TGF-beta1 upregulation in mesangial cells’**. *J. Am. Soc. Nephrol.* (2009) **20** 554-566. DOI: 10.1681/ASN.2008040445
58. Xie X. S., Li F. y., Liu H. c., Deng Y., Li Z., Fan J. m.. **‘LSKL, a peptide antagonist of thrombospondin-1, attenuates renal interstitial fibrosis in rats with unilateral ureteral obstruction’**. *Archives Pharmacal Res.* (2010) **3333** 275-284. DOI: 10.1007/S12272-010-0213-6
59. Xu J., Wang J., Cheng Y., Li X., He M., Zhu J.. **‘Glucagon-like peptide-1 mediates the protective effect of the dipeptidyl peptidase IV inhibitor on renal fibrosis via reducing the phenotypic conversion of renal microvascular cells in monocrotaline-treated rats’**. *BioMed Res. Int.* (2018) **2018** 1864107. DOI: 10.1155/2018/1864107
60. Xu L., Zhang Y., Chen J., Xu Y.. **‘Thrombospondin-1: A key protein that induces fibrosis in diabetic complications’**. *J. Diabetes Res.* (2020) **2020** 8043135. DOI: 10.1155/2020/8043135
61. Yevdokimova N., Abdel Wahab N., Mason R. M.. **‘Thrombospondin-1 is the key activator of TGF-beta1 in human mesangial cells exposed to high glucose’**. *J. Am. Soc. Nephrol.* (2001) **12** 703-712. DOI: 10.1681/ASN.V124703
62. Zhou Y., Poczatek M. H., Berecek K. H., Murphy-Ullrich J. E.. **‘Thrombospondin 1 mediates angiotensin II induction of TGF-β activation by cardiac and renal cells under both high and low glucose conditions’**. *Biochem. Biophysical Res. Commun.* (2006) **339** 633-641. DOI: 10.1016/J.BBRC.2005.11.060
63. Zuo Y., Chun B., Potthoff S. A., Kazi N., Brolin T. J., Orhan D.. **‘Thymosin β4 and its degradation product, Ac-SDKP, are novel reparative factors in renal fibrosis’**. *Kidney Int.* (2013) **84** 1166-1175. DOI: 10.1038/KI.2013.209
|
---
title: High prevalence of reduced fertility and use of assisted reproductive technology
in a German cohort of patients with peripartum cardiomyopathy
authors:
- Tobias J. Pfeffer
- Manuel List
- Cordula Schippert
- Bernd Auber
- Melanie Ricke-Hoch
- Valeska Abou-Moulig
- Dominik Berliner
- Johann Bauersachs
- Denise Hilfiker-Kleiner
journal: Clinical Research in Cardiology
year: 2022
pmcid: PMC9998571
doi: 10.1007/s00392-022-02034-x
license: CC BY 4.0
---
# High prevalence of reduced fertility and use of assisted reproductive technology in a German cohort of patients with peripartum cardiomyopathy
## Abstract
### Background
Over the past decades the use of assisted reproduction technology (ART) increased worldwide. ARTs are associated with an elevated risk for cardiovascular complications. However, a potential relation between subfertility/ARTs and the heart disease peripartum cardiomyopathy (PPCM) has not been systematically analyzed yet.
### Methods
A retrospective cohort study was carried out, including $$n = 111$$ PPCM patients from the German PPCM registry. Data from PPCM patients were compared to those from postpartum women in the *German* general population.
### Results
The prevalence of reported subfertility was high among PPCM patients ($30\%$; $\frac{33}{111}$). Most of the subfertile PPCM patients ($55\%$; $\frac{18}{33}$) obtained vitro fertilizations (IVF) or intracytoplasmic sperm injections (ICSI). PPCM patients were older ($p \leq 0.0001$), the percentage of born infants conceived by IVF/ICSI was higher ($p \leq 0.0001$) with a higher multiple birth ($p \leq 0.0001$), C-section ($p \leq 0.0001$) and preeclampsia rate ($p \leq 0.0001$), compared to postpartum women. The cardiac outcome was comparable between subfertile and fertile PPCM patients. Whole exome sequencing in a subset of $$n = 15$$ subfertile PPCM patients revealed that $33\%$ ($\frac{5}{15}$) carried pathogenic or likely pathogenic gene variants associated with cardiomyopathies and/or cancer predisposition syndrome.
### Conclusions
Subfertility occurred frequently among PPCM patients and was associated with increased age, hormonal disorders, higher twin pregnancy rate and high prevalence of pathogenic gene variants suggesting a causal relationship between subfertility and PPCM. Although this study found no evidence that the ART treatment per se increases the risk for PPCM or the risk for an adverse outcome, women with subfertility should be closely monitored for signs of peripartum heart failure.
### Supplementary Information
The online version contains supplementary material available at 10.1007/s00392-022-02034-x.
## Introduction
In industrialized countries the pregnancy rate among young women is falling due to a wider usage of long-acting reversible contraceptives (LARCs), emergency contraceptives and changes in the social structure, leading to a rising maternal age at first birth [1]. Fertility of women decreases with increasing age requiring more frequently assisted reproductive technologies (ART) such as in vitro fertilizations (IVF) and, since male fertility decreases as well, intracytoplasmic sperm injections (ICSI) [2, 3]. Simultaneously, insurance coverage of ARTs expanded in various countries, contributing to an increasing use of ARTs in the past decades [4–6]. In Germany, the rate of born infants conceived by IVF/ICSI is $2.2\%$ whereas in other European countries this rate ranges from $0.9\%$ in Malta to $6.4\%$ in Denmark (in 2014) [4]. In the United States the rate of infants conceived by IVF/ICSI has been calculated to be $1.6\%$ in 2014 [7] and $1.8\%$ in 2016 [5].
Pregnancies conceived or supported by ART procedures are associated with higher rates of hypertensive disorders such as preeclampsia [8]. Hypertensive disorders in turn are well-known risk factors for the development of peripartum cardiomyopathy (PPCM) [9]. PPCM is a life-threatening heart disease with onset in the last months of pregnancy, during delivery or in the first months after delivery in previously heart-healthy women [10]. PPCM is defined by heart failure due to left ventricular (LV) systolic dysfunction (LV-ejection fraction < $45\%$) [11, 12].
Hereby, we examine the prevalence of subfertility and ART procedures in a cohort of patients from the German PPCM registry in comparison to postpartum women in the *German* general population. In addition, we compared, outcome and potential risk factors including genetic risk factors and comorbidities in PPCM patients with and without subfertility and ART procedures.
## Data collection
This study was approved by the local ethics committee of Hannover Medical School, Hanover, Germany (7970_BO_K_2018). The study complies with the Declaration of Helsinki and all patients and postpartum healthy controls gave written informed consent.
All PPCM patients who were treated at Hannover Medical School between January 2005 and March 2019 were screened for subfertility. Data regarding fertility status and applied subfertility treatment were collected via phone survey or in our PPCM outpatient clinic (Suppl. Figure 1). In this study, diagnosis of subfertility was based on either reported use of ART procedures or on reported failure to achieve a pregnancy after more than six months of regular unprotected sexual intercourse. The results were discussed and reviewed by gynecologists for reproductive medicine, who were blinded for the analysis of the medical findings.
Medical data at diagnosis, such as clinical signs of heart failure, echocardiographic parameters, laboratory assessments, drug treatment, adverse events and medical history, were collected. Follow-up (FU) data were obtained after 6 months (range 16–32 weeks). Very few data sets are incomplete in this registry as index PPCM was partly diagnosed in remote community hospitals or in an outpatient clinic from which patients were referred.
For a better evaluation of the outcome, we divided our collective in three subgroups, mainly depending on the LVEF at six months FU, as it has already been established in former studies [13]. In brief, full recovery was defined by an LVEF of ≥ $50\%$, partial recovery by an LVEF of 35–$49\%$ and no recovery was defined by an LVEF < $35\%$ or the occurrence of an adverse event (implantation of a left ventricular assist device (LVAD), heart transplantation (HTX) or death).
To compare the prevalence of IVF/ICSI in our collective with the prevalence in the general population, the percentage of born infants conceived by IVF/ICSI among the total number of born infants of all interviewed patients was calculated for both collectives, using data from the Statistical Office of Germany and the German IVF-registry [14, 15]. Numbers were pooled for 2002 up to 2017 as the first live birth in our PPCM collective was in 2002 and the latest available data for the general population reaches up to 2017. These analyses were performed in cooperation with our local institute for biometrics.
## Exome sequencing and variant classification
Exome sequencing was performed on all PPCM patients with conducted IVF or ICSI whose blood samples were available. Furthermore, all PPCM patients with reported subfertility and history of cancer were analyzed. DNA was extracted from whole blood samples. DNA enrichment and library preparation were performed using the xGen® Exome Research Panel (Integrated DNA Technologies, Inc., Coralville, USA). Sequencing was performed on an Illumina NextSeq 500 using the NextSeq $\frac{500}{550}$ High Output v2 kit (Illumina, San Diego, CA). Alignment to the reference genome build (GRCh37) was performed using megSAP, version 0.1-710-g52d2b0c (https://github.com/imgag/megSAP). Variant prioritization and visualization was performed with GSvar, version 2018_04 (https://github.com/imgag/ngs-bits), IGV [16], version 2.4.14 and with Alamut® visual, version 2.11 (Interactive Biosoftware, Rouen, France). Variants were classified according to the criteria proposed by the American College of Medical Genetics and Genomics (ACMG) [17], and 248 genes were analyzed per patient. Genes associated with dilated cardiomyopathy (DCM) were selected using the human phenotype ontology database [18] term “dilated cardiomyopathy” (HP:0001644; 115 genes); genes associated with DNA damage response (DDR) and general cancer predisposition syndromes (CPS) were composed of all genes listed in a benchmark study regarding cancer predisposition gene testing in adult patients [19] (Suppl. Table 1 and 2 for detailed information). *Only* genetic variants which were classified as ACMG class $\frac{4}{5}$ (likely pathogenic/pathogenic) were considered. Part of the genetic data published in this study where already published and discussed in a former study, investigating the connection between PPCM and cancer [20].
## Statistical analysis
Data were analyzed using GraphPad Prism version 5.0f for Mac (GraphPad Software, La Jolla California, USA). Continuous data are expressed as mean ± SD or median (IQR) and categorical data as frequencies (%). Normal distribution was assessed using D'Agostino & *Pearson omnibus* normality test. Fisher’s exact test or Chi-square test was used for discrete variables depending on sample size, unpaired t test or Mann–Whitney U test were used for continuous variables. For multiple group comparison we used Chi-square test for discrete variables and one-way non-parametric ANOVA (Kruskal–Wallis test) for non-normally distributed continuous variables. A p-value of less than 0.05 was considered statistically significant.
## Prevalence of ART and infants conceived by IVF/ICSI in PPCM patients compared to the general population in Germany
In the present study, $$n = 111$$ patients with confirmed diagnosis of PPCM and known fertility status were included. Out of these, $30\%$ ($\frac{33}{111}$) reported subfertility and/or received fertility treatment (SF-PPCM). Any kind of fertility treatment (ST-PPCM) was received by $26\%$ ($\frac{29}{111}$) of all PPCM patients and $16\%$ ($\frac{18}{111}$) were treated with IVF or ICSI (IVF-PPCM) (Fig. 1a). Among the 18 patients of IVF-PPCM, 16 patients were treated with IVF or ICSI prior to the index pregnancy, leading to the diagnosis of PPCM, and two patients underwent IVF/ICSI 45 or 17 months before delivery of index pregnancy. There were four patients with subfertility who did not receive any fertility-related treatment (NT-PPCM). In the present PPCM cohort $12\%$ ($\frac{26}{209}$) of all born infants were conceived by IVF/ICSI. Currently, the percentage of infants conceived by IVF/ICSI in the general population in *Germany is* about $2.2\%$ [14, 15] within the same time period, which is significantly ($p \leq 0.0001$) lower than in the PPCM cohort (Fig. 1b, c).Fig. 1Distribution of PPCM patients in the different subgroups depending on fertility status and applied treatment protocol (a) and proportion of infants conceived by IVF/ICSI (ART infants) among PPCM patients (b) and in the *German* general population for the time period 2002–2017 (c)
## Age, gravidity, parity and cardiac function in PPCM patients with and without subfertility
The $$n = 111$$ PPCM patients presented with a mean LVEF of 28 ± $10\%$ at diagnosis. The median parity in all 111 patients was 1 (range: 1–6, Table 1). In one patient, index pregnancy resulted in a stillbirth, whereas 110 patients had pregnancies with live births. Parity was significantly lower in SF-PPCM compared to PPCM patients with unimpaired fertility (F-PPCM) (1.3 ± 0.7 vs. 1.8 ± 1.1; $$p \leq 0.0240$$) and the twin pregnancy rate was significantly higher ($36\%$ vs. $15\%$; $$p \leq 0.0221$$). All other baseline parameters were comparable between both groups including all parameters associated with severity of the disease such as NYHA class, LVEF, and NT-proBNP levels (Suppl. Figure 2). Furthermore, we compared the baseline parameters of our cohort with the baseline parameters of peripartum women in the normal population in Germany. PPCM patients were significantly older (34 ± 4 years vs. 31 ± 5 years; $p \leq 0.0001$), and had a significantly higher rate of multiple births ($22\%$ vs. $1.7\%$; $p \leq 0.0001$), C-sections ($68\%$ vs. $29\%$; $p \leq 0.0001$) and preeclampsia ($31\%$ vs. $2.6\%$; $p \leq 0.0001$). In contrast, smoking status ($33\%$ vs. $37\%$; ns) and parity (1.6 vs. 1.4; ns) were comparable between both groups and the prevalence of gestational diabetes was significantly lower in PPCM patients ($6\%$ vs. $13\%$; $p \leq 0.0.05$).Table 1Baseline characteristics of PPCM patients with and without subfertilityBaseline characteristicsAll PPCM ($$n = 111$$)Subfertile PPCM (SF-PPCM; $$n = 33$$)Subgroup with IVF/ICSI (IVF-PPCM; $$n = 18$$)Fertile PPCM (F-PPCM; $$n = 78$$)Age in years34 ± 435 ± 536 ± 4*34 ± 4Gravidity1 (1–7)1 (1–4)1 [1–4]*2 (1–7)Parity1 (1–6)1 (1–4)*1 (1–4)*1 (1–6)Twin pregnancy$22\%$ ($\frac{24}{111}$)$36\%$ ($\frac{12}{33}$)*$50\%$ ($\frac{9}{18}$)*$15\%$ ($\frac{12}{78}$)C-Section$68\%$ ($\frac{71}{104}$)$82\%$ ($\frac{27}{33}$)$94\%$ ($\frac{17}{18}$)*$62\%$ ($\frac{44}{71}$)Time from birth to first echo in days33 ± 5423 ± 4817 ± 3038 ± 57Medical history Hypertension$13\%$ ($\frac{14}{109}$)$9\%$ ($\frac{3}{33}$)$0\%$ ($\frac{0}{18}$)$14\%$ ($\frac{11}{76}$) Diabetes$4\%$ ($\frac{4}{109}$)$3\%$ ($\frac{1}{33}$)$6\%$ ($\frac{1}{18}$)$4\%$ ($\frac{3}{76}$) (Former) nicotine abuse$33\%$ ($\frac{31}{94}$)$27\%$ ($\frac{8}{30}$)$13\%$ ($\frac{2}{15}$)$36\%$ ($\frac{23}{64}$) History of cancer in total$11\%$ ($\frac{12}{111}$)$24\%$ ($\frac{8}{33}$)$17\%$ ($\frac{3}{18}$)$5\%$ ($\frac{4}{78}$) Cancer before PPCM$7\%$ ($\frac{8}{111}$)$18\%$ ($\frac{6}{33}$)$17\%$ ($\frac{3}{18}$)$3\%$ ($\frac{2}{78}$) Cancer after PPCM$5\%$ ($\frac{5}{111}$)$9\%$ ($\frac{3}{33}$)$0\%$ ($\frac{0}{18}$)$3\%$ ($\frac{2}{78}$)Pregnancy-related conditions Preeclampsia$31\%$ ($\frac{34}{111}$)$30\%$ ($\frac{10}{33}$)$44\%$ ($\frac{8}{18}$)$31\%$ ($\frac{24}{78}$) Gestational diabetes$6\%$ ($\frac{7}{110}$)$6\%$ ($\frac{2}{33}$)$6\%$ ($\frac{1}{18}$)$6\%$ ($\frac{5}{77}$)Heart failure medication D2-agonist$87\%$ ($\frac{92}{106}$)$94\%$ ($\frac{31}{33}$)$94\%$ ($\frac{17}{18}$)$84\%$ ($\frac{61}{73}$) ACE inhibitor$77\%$ ($\frac{82}{106}$)$79\%$ ($\frac{26}{33}$)$78\%$ ($\frac{14}{18}$)$77\%$ ($\frac{56}{73}$) AT1-antagonist$18\%$ ($\frac{19}{105}$)$15\%$ ($\frac{5}{33}$)$17\%$ ($\frac{3}{18}$)$19\%$ ($\frac{14}{72}$) MR-antagonist$74\%$ ($\frac{78}{105}$)$61\%$ ($\frac{20}{33}$)*$56\%$ ($\frac{10}{18}$)$81\%$ ($\frac{58}{72}$) Beta-blocker$92\%$ ($\frac{99}{108}$)$88\%$ ($\frac{29}{33}$)$89\%$ ($\frac{16}{18}$)$93\%$ ($\frac{70}{75}$) Diuretic$81\%$ ($\frac{86}{106}$)$76\%$ ($\frac{25}{33}$)$72\%$ ($\frac{13}{18}$)$84\%$ ($\frac{61}{73}$)*Continuous data* are expressed as mean ± SD or median (range) according to normality of distribution. Dichotomous data are represented as percentage (number). Comparison between two groups (fertile group values were tested for statistical difference against subfertile group values or IVF/ICSI subgroup values) using unpaired t-test for metrical values with normal distribution, Mann–Whitney U test for metrical values with non-normal distribution and Fisher’s exact test for dichotomous valuesIVF/ICSI, in vitro fertilization/intracytoplasmic sperm injection; LVEF, left ventricular ejection fraction; NT-proBNP, N-terminal pro-B-natriuretic peptide; NYHA, New York Heart Association; BL, baseline; D2-agonist, D2 dopamine receptor agonist; ACE inhibitor, angiotensin converting enzyme inhibitor; AT1-antagonist, AT1 angiotensin receptor antagonist; MR-antagonist, mineralocorticoid receptor antagonist*$p \leq 0.05.$ Numbers may not add up to $100\%$ because of roundingTable 2Baseline characteristics of PPCM patients in comparison to peripartum women in the general populationBaseline characteristicsAll PPCM ($$n = 111$$)General populationAge in years34 ± 431 ± 5**** [15]Parity1.631.41 [34]Multiple births$22\%$ ($\frac{24}{111}$)$1.7\%$ (232,$\frac{068}{13}$,299,837)**** [35]C-sections$68\%$ ($\frac{71}{104}$)$29\%$ (3,780,$\frac{978}{13}$,124,061)**** [36](Former) nicotine abuse$33\%$ ($\frac{31}{94}$)$37\%$ [37]Preeclampsia$31\%$ ($\frac{34}{111}$)$2.6\%$ ($\frac{3654}{138}$,571)**** [38]Gestational diabetes$6\%$ ($\frac{7}{110}$)$13\%$ (75,$\frac{034}{567}$,191)* [39]*Continuous data* are expressed as mean ± SD or median (range) according to normality of distribution. Dichotomous data are represented as percentage (number)As source data for parity and smoking status was not available, we could not perform statistical analysis comparing both groups for these properties. Both groups were compared using unpaired t-test for metrical values with normal distribution and Fisher’s exact test or Chi-square test for dichotomous values. * $p \leq 0.05$; ****p << 0.0001Age in years for peripartum women in the *German* general population was calculated as pooled mean of ages for every woman’s birth in Germany from 2009 to 2018 whereas for PPCM patients only years of births of the PPCM pregnancies were included. All other data from the Statistical Office of Germany (parity, multiple births, C-sections) were pooled for the time period 2000–2018 analogous to the range of date of diagnosis in our questioned collective. The rate of (former) nicotine abuse for *German* general population factors in data from the German micro-census 2017 including women of age 20–45 (corresponding to the age range of 23–43 in our questioned collective)
## Causes of subfertility
In $45\%$ ($\frac{15}{33}$) of SF-PPCM predominantly female causes of subfertility could be identified. Of those, ovulatory dysregulations presenting with menstrual dysfunction were reported in $44\%$ ($\frac{14}{32}$), polycystic ovary syndrome (PCOS) in $12\%$ ($\frac{4}{33}$), history of hyperprolactinemia in $19\%$ ($\frac{6}{32}$), hypothyroidism in $45\%$ ($\frac{15}{33}$) and diminished ovarian reserve defined by an anti-Müllerian hormone level below respective reference value in $18\%$ ($\frac{2}{11}$). Organic causes such as endometriosis ($12\%$, $\frac{4}{33}$), tubal/ovarian abnormalities ($22\%$, $\frac{7}{32}$), uterus abnormalities ($6\%$, $\frac{2}{32}$) or past surgical procedures on the uterus/fallopian tubes/ovaries ($34\%$, $\frac{11}{32}$) were also present in SF-PPCM. Further risk factors that were present in our collective and that have shown to have an impact on subfertility, were advanced age (≥ 40 years) at start of successful fertility treatment ($15\%$, $\frac{5}{33}$), overweight defined by body mass index ≥ 25 kg/m2 ($31\%$, $\frac{10}{32}$), (former) nicotine abuse ($27\%$ $\frac{8}{30}$), history of radio-/chemotherapy ($15\%$ $\frac{5}{33}$) and heterozygous Fragile X syndrome ($3\%$, $\frac{1}{33}$) [21, 22].
In $18\%$ ($\frac{6}{33}$) a predominantly male causation, indicated by changes in the concerned men’s spermiograms, and in another $18\%$ ($\frac{6}{33}$) a combined contribution of both partners to the subfertility was assumed (Suppl. Table 3).
In $15\%$ ($\frac{5}{33}$) no apparent cause neither in the female patients nor in their partners were found. One patient ($3\%$, $\frac{1}{33}$) conducted a prophylactic cryopreservation prior to chemotherapy and was included in our analysis analogous to the definition of our inclusion criteria given above (Table 2).
## Fertility treatment protocols applied to PPCM patients
For further analyses, only fertility treatment regimens applied less than three months before conception of index PPCM were included ($$n = 24$$). Sixteen patients underwent an ART procedure directly prior to the index pregnancy. Out of these, nine patients obtained ICSI and six patients underwent IVF without intracytoplasmic injection. For one patient no data from the treating fertility center with regard to the fertility procedure was available.
Nine patients underwent controlled ovarian hyperstimulation (COH) directly prior to the embryo transfer with seven patients obtaining a long protocol with administration of gonadotropin-releasing hormone (GnRH)-agonists and two patients obtaining a short protocol with administration of GnRH-antagonists. Five patients underwent an embryo transfer in a cryo-thaw cycle without any additional medication for that cycle and the remaining two patients got pregnant after ovum donation.
For luteal support, progesterone was administered in $91\%$ ($\frac{20}{22}$) and estrogen was given in $45\%$ ($\frac{10}{22}$) of the ST-PPCM patients. Human chorionic gonadotropin (hCG) was utilized for ovulation induction in $74\%$ ($\frac{17}{23}$) of these PPCM patients. For ovarian stimulation, follicle-stimulating hormone (FSH) was administered in $52\%$ ($\frac{12}{23}$), human chorionic gonadotrophin (hMG) in $13\%$ ($\frac{3}{23}$) and clomiphene in $9\%$ ($\frac{2}{23}$) for the same objective. These numbers vary if applied medications of previous unsuccessful treatment cycles are included (Suppl. Figure 3).
## Genetic analysis
*Since* genetics contribute to both, PPCM and subfertility, exome sequencing was performed in a subset of $$n = 15$$ SF-PPCM patients with reported subfertility. Out of these, 12 patients received fertility treatment. In five out of these 15 patients ($33\%$), either pathogenic (P, class 5) or likely pathogenic (LP, class 4) variants associated with dilatative (DCM) or hypertrophic (HCM) cardiomyopathies ($$n = 5$$) and/or cancer predisposition syndrome (CPS, $$n = 3$$), with two patients carrying mutations associated with both pathologies, were found (Suppl. Table 1 and 2). Of note, eight patients of SF-PPCM had a history of cancer prior to ($$n = 6$$) or after ($$n = 3$$) PPCM (one patient suffered from both, cancer prior to and after PPCM) and $88\%$ ($\frac{7}{8}$) of the pathogenic or likely pathogenic mutations were found in patients with a history of cancer.
## Impact of subfertility and fertility treatments on the cardiac outcome in PPCM patients
To evaluate the impact of subfertility on the clinical course of PPCM, we measured LVEF and NT-proBNP-levels 3, 6 and 12 months after diagnosis. At all time points, LVEF as well as NT-proBNP levels did not differ significantly between F-PPCM and SF-PPCM (Fig. 2a and Suppl. Figure 2a, b). Furthermore, mortality ($0\%$ vs. $0\%$) and the rate of implanted LVAD ($3\%$ vs. $1\%$, $\frac{1}{33}$ vs. $\frac{1}{78}$) or HTX ($3\%$ vs. $0\%$, $\frac{1}{33}$ vs. $\frac{0}{78}$) were similar in both groups. Fig. 2Course of LVEF in SF-PPCM and F-PPCM at baseline, 6 and 12 months follow-up (a) and recovery status at 6 months follow-up of SF-PPCM and F-PPCM (b) and of IVF-PPCM and F-PPCM (c). BL, baseline; LVAD, left ventricular assist device; LVEF, left ventricular ejection fraction; HTX, heart transplantation; IVF/ICSI, in vitro fertilization/intracytoplasmic sperm injection; M, months. Red column: non-recovery (LVEF ≤ $35\%$, HTX, LVAD-implantation or death); yellow column: partial recovery (LVEF > 35-$49\%$); green column: full recovery (LVEF ≥ $50\%$). Chi-square tests were performed on the distribution of the three recovery groups and did not retrieve any statistically significant results Moreover, the recovery status at 6 and 12 months FU did neither differ between SF-PPCM and F-PPCM nor between IVF-PPCM and F-PPCM (Fig. 2b, c).
## Discussion
The present study is the first to address subfertility and the high rate of conducted ART procedures among PPCM patients in depth. A risk factor analysis of an Israeli cohort of 36 PPCM patients revealed a high rate of patients conceiving by IVF [23], which we also observed in our larger cohort of 111 German PPCM patients. The use of ARTs, especially IVF/ICSI procedures, was significantly higher in PPCM patients compared to the general population in Germany.
In our study cohort, well-described risk factors for both PPCM and subfertility such as high maternal age and hormonal disorders, were elevated compared to pregnant women in the general population. These risk factors were also more prevalent than in PPCM collectives with different ethnical backgrounds. In the largest African PPCM registry, the PEACE Registry [24], PPCM patients were younger, parity was higher and hormonal disorders were less frequent compared to our German PPCM collective. These distinctions go in line with the findings from the EORP registry, the largest international registry including 739 women with PPCM from 49 different countries [10]. Therein, European PPCM patients were older and had a lower number of previous pregnancies compared to PPCM patients from Africa, Asia–Pacific or Middle East, suggesting a distinct set of risk factors for individual ethnic groups. Interestingly, hormonal disorders were more frequent in the PPCM cohorts from Asia and Middle East. In comparison to the largest Asian PPCM study from Taiwan with 925 PPCM patients [25], our German PPCM collective displayed a higher maternal age, a higher multiparity rate and a similar rate of previous deliveries ($33\%$ vs. $36\%$). To evaluate the role of ethnic backgrounds as well as the accessibility and usage of assisted reproductive technology in other countries with regard to an increased risk for PPCM, our findings should be validated in a larger, ethnically diverse cohort of PPCM patients.
In addition, twin pregnancies, which occur more commonly after IVF/ICSI due to general practice of multiple embryo transfers, are a known risk factor for PPCM [26]. Indeed, in the present cohort, twin pregnancy rate was higher in PPCM patients with ART compared to PPCM patients with naturally conceived pregnancies. With regard to the causes for the observed higher incidence of subfertility in PPCM patients, subfertile PPCM patients frequently carried pathogenic or likely pathogenic gene variants associated with either DCM/HCM and/or CPS, which may impact fertility due to lower rate of functional eggs and increased rate of early abortion due to genetic impairment in embryos. In addition, some of the subfertile PPCM patients underwent anticancer treatment prior to the index pregnancy, which can also impair fertility [27] and can lead to a higher risk for cardiomyopathies [20] at the same time. Thus, the anticancer treatment may have also contributed to the connection between subfertility and PPCM.
Furthermore, shared pathophysiological pathways could also play a role for the observed high prevalence of subfertility and ART in PPCM patients. For example, excessive oxidative stress is known to play a key role in the pathophysiology of PPCM [28, 29] and has also been described as a key factor in subfertility-associated diseases such as endometriosis [30] and polycystic ovary syndrome (PCOS) [31, 32].
Interestingly, both subfertility and the use of ART procedures did not affect the clinical course of PPCM, as the cardiac outcome was comparable between subfertile and fertile PPCM patients, suggesting that neither the condition nor the ART treatment are influencing the course of the disease.
Concluding from these data, women with subfertility problems, especially when receiving fertility treatment, should be closely monitored for cardiovascular complications in the peripartum phase. Suspicious signs of heart failure like dyspnea should be taken seriously and lead to early cardiac examination to rule out PPCM. Peripartum determination of NT-proBNP could serve as a biomarker for cardiac health in women who conceived by IVF/ICSI and thus help to diagnose PPCM early.
Patients with diagnosed PPCM and a history of subfertility who are planning a subsequent pregnancy and seek medical assistance (especially ART) should consult their treating physician and should be carefully informed about the potential risks. In these patients, ART procedures should only be conducted as result of a risk–benefit analysis after informed consent between cardiologist, gynecologist and the patient. Once pregnant, these patients should remain under close clinical monitoring whereby the use of biomarkers such as NT-proBNP and echocardiographic examinations may help to identify worsening of cardiac function during and after pregnancy. In this context it is important to note that PPCM can also occur after early abortions.
Further studies are needed to examine the connection of PPCM and subfertility more closely and to elucidate the pathophysiological mechanism behind the accumulation of subfertility among PPCM patients.
## Limitations
Not all patients included in this study were diagnosed in our hospital and thus baseline data are partly incomplete. For the same reason we were not able to retrieve blood samples at baseline from all PPCM patients.
The definition of subfertility varies in the literature as much as the concept itself is not widely recognized as a narrowly defined term. The time to pregnancy is a subjective measure to estimate the severity of subfertility. We were guided by a German study for subfertility that determined the lightest form of impaired fertility at 6 months of unsuccessful conception [33].
## Conclusions
Subfertility and fertility treatments are more frequent in PPCM patients compared to the general population. It remains unclear whether fertility treatments are an underlying risk factor for the development of PPCM or whether these patients carry risk factors and pathophysiological conditions that cause both, subfertility and a predisposition for PPCM. Whole exome sequencing revealed an increased number of mutations in PPCM patients with subfertility, but these mutations occurred mainly in PPCM patients with history of cancer, making genetic alterations an improbable connection between these two entities. The cardiac outcome did not differ significantly between PPCM patients with and without subfertility.
Women receiving fertility treatment should be closely monitored for symptoms of heart failure and should be immediately referred to a cardiologist in case of suspected PPCM. PPCM patients who are planning a subsequent pregnancy should only consider fertility treatment under close clinical monitoring and after informed consent between cardiologist, gynecologist and patient.
## Supplementary Information
Below is the link to the electronic supplementary material. Supplemental Fig. 1 Study flow diagram of our analysis of subfertility and fertility treatments among PPCM patients. ART, assisted reproductive technology. Supplemental Fig. 2 LVEF (a) and NT-proBNP (b) in the respective groups of questioned PPCM patients at baseline, 3, 6 and 12 months follow-up. NYHA class at 6 months follow-up for the same groups (c): BL, baseline; FU, follow-up; IVF/ICSI, in vitro fertilization/intracytoplasmic sperm injection; LVEF, left ventricular ejection fraction; M, months; NT-proBNP, N-terminal pro-B-natriuretic peptide; NYHA, New York Heart Association. LVEF and NT-proBNP values at the different time points of each group were compared using one-way non-parametric ANOVA (Kruskal-Wallis test) and for the comparison of the distribution of FU NYHA class in the different groups a Chi-square test was carried out. Neither of the tests revealed any statistically significant results. Supplemental Fig. 3 Applied fertility treatment medication within 3 months prior to index pregnancy and in total. GnRH (gonadotropin-releasing hormone); FSH (follicle-stimulating hormone); LVEF (left ventricular ejection fraction); LH (luteinizing hormone); hMG (human menopausal gonadotropin); hCG (human chorionic gonadotropin) (PPTX 671 kb)Supplementary file2 (PDF 281 kb)
## References
1. 1.OECD (2019) SF2.3: Age of mothers at childbirth and age-specific fertility. https://www.oecd.org/els/soc/SF_2_3_Age_mothers_childbirth.pdf. Accessed 23 Nov 2021
2. Menken J, Trussell J, Larsen U. **Age and infertility**. *Science* (1986.0) **233** 1389-1394. DOI: 10.1126/science.3755843
3. Dunson DB, Baird DD, Colombo B. **Increased infertility with age in men and women**. *Obstet Gynecol* (2004.0) **103** 51-56. DOI: 10.1097/01.AOG.0000100153.24061.45
4. De Geyter C, Calhaz-Jorge C, Kupka MS, Wyns C, Mocanu E, Motrenko T. **ART in Europe, 2014: results generated from European registries by ESHRE: the European IVF-monitoring Consortium (EIM) for the European Society of Human Reproduction and Embryology (ESHRE)**. *Hum Reprod* (2018.0) **33** 1586-1601. DOI: 10.1093/humrep/dey242
5. Sunderam S, Kissin DM, Zhang Y, Folger SG, Boulet SL, Warner L. **Assisted reproductive technology surveillance—United States, 2016**. *MMWR Surveill Summ* (2019.0) **68** 1-23. DOI: 10.15585/mmwr.ss6804a1
6. Teoh PJ, Maheshwari A. **Low-cost in vitro fertilization: current insights**. *Int J Womens Health* (2014.0) **6** 817-827. DOI: 10.2147/IJWH.S51288
7. Sunderam S, Kissin DM, Crawford SB, Folger SG, Jamieson DJ, Warner L. **Assisted reproductive technology surveillance—United States, 2014**. *MMWR Surveill Summ* (2017.0) **66** 1-24. DOI: 10.15585/mmwr.ss6606a1
8. Okby R, Harlev A, Sacks KN, Sergienko R, Sheiner E. **Preeclampsia acts differently in in vitro fertilization versus spontaneous twins**. *Arch Gynecol Obstet* (2018.0) **297** 653-658. DOI: 10.1007/s00404-017-4635-y
9. Behrens I, Basit S, Lykke JA, Ranthe MF, Wohlfahrt J, Bundgaard H. **Hypertensive disorders of pregnancy and peripartum cardiomyopathy: a nationwide cohort study**. *PLoS ONE* (2019.0) **14** e0211857. DOI: 10.1371/journal.pone.0211857
10. Sliwa K, Petrie MC, van der Meer P, Mebazaa A, Hilfiker-Kleiner D, Jackson AM. **Clinical presentation, management, and 6-month outcomes in women with peripartum cardiomyopathy: an ESC EORP registry**. *Eur Heart J* (2020.0) **41** 3787-3797. DOI: 10.1093/eurheartj/ehaa455
11. Sliwa K, Petrie MC, Hilfiker-Kleiner D, Mebazaa A, Jackson A, Johnson MR. **Long-term prognosis, subsequent pregnancy, contraception and overall management of peripartum cardiomyopathy: practical guidance paper from the Heart Failure Association of the European Society of Cardiology Study Group on Peripartum Cardiomyopathy**. *Eur J Heart Fail* (2018.0) **20** 951-962. DOI: 10.1002/ejhf.1178
12. Bauersachs J, Konig T, van der Meer P, Petrie MC, Hilfiker-Kleiner D, Mbakwem A. **Pathophysiology, diagnosis and management of peripartum cardiomyopathy: a position statement from the Heart Failure Association of the European Society of Cardiology Study Group on peripartum cardiomyopathy**. *Eur J Heart Fail* (2019.0) **21** 827-843. DOI: 10.1002/ejhf.1493
13. Hilfiker-Kleiner D, Haghikia A, Berliner D, Vogel-Claussen J, Schwab J, Franke A. **Bromocriptine for the treatment of peripartum cardiomyopathy: a multicentre randomized study**. *Eur Heart J* (2017.0) **38** 2671-2679. DOI: 10.1093/eurheartj/ehx355
14. 14.Blumenauer V, Czeromin U, Fehr D, Fiedler K, Gnoth C, Krüssel JS et al (2018) D.I.R annual 2017—the German IVF-registry. J Reprod Med Endocrinoll 15(5–6):217–49. https://www.deutsches-ivf-register.de/perch/resources/dir-annual-2017-english-final-190320.pdf. Accessed 23 Nov 2021
15. 15.Statistical Office of Germany (2009–2018) (2021) Live births: Germany, years, age of the mother, live birth order (Code: 12612–0005). https://www-genesis.destatis.de/genesis/online/?operation=sprachwechsel&language=en. Accessed 23 Nov 2021
16. Thorvaldsdottir H, Robinson JT, Mesirov JP. **Integrative Genomics Viewer (IGV): high-performance genomics data visualization and exploration**. *Brief Bioinform* (2013.0) **14** 178-192. DOI: 10.1093/bib/bbs017
17. Richards S, Aziz N, Bale S, Bick D, Das S, Gastier-Foster J. **Standards and guidelines for the interpretation of sequence variants: a joint consensus recommendation of the American College of Medical Genetics and Genomics and the Association for Molecular Pathology**. *Genet Med* (2015.0) **17** 405-424. DOI: 10.1038/gim.2015.30
18. Kohler S, Carmody L, Vasilevsky N, Jacobsen JOB, Danis D, Gourdine JP. **Expansion of the Human Phenotype Ontology (HPO) knowledge base and resources**. *Nucleic Acids Res* (2019.0) **47** D1018-D1027. DOI: 10.1093/nar/gky1105
19. Whitworth J, Smith PS, Martin JE, West H, Luchetti A, Rodger F. **Comprehensive cancer-predisposition gene testing in an adult multiple primary tumor series shows a broad range of deleterious variants and atypical tumor phenotypes**. *Am J Hum Genet* (2018.0) **103** 3-18. DOI: 10.1016/j.ajhg.2018.04.013
20. Pfeffer TJ, Schlothauer S, Pietzsch S, Schaufelberger M, Auber B, Ricke-Hoch M. **Increased cancer prevalence in peripartum cardiomyopathy**. *JACC CardioOncol* (2019.0) **1** 196-205. DOI: 10.1016/j.jaccao.2019.09.008
21. Homan GF, Davies M, Norman R. **The impact of lifestyle factors on reproductive performance in the general population and those undergoing infertility treatment: a review**. *Hum Reprod Update* (2007.0) **13** 209-223. DOI: 10.1093/humupd/dml056
22. Kelly-Weeder S, Cox CL. **The impact of lifestyle risk factors on female infertility**. *Women Health* (2006.0) **44** 1-23. DOI: 10.1300/j013v44n04_01
23. Shani H, Kuperstein R, Berlin A, Arad M, Goldenberg I, Simchen MJ. **Peripartum cardiomyopathy—risk factors, characteristics and long-term follow-up**. *J Perinat Med* (2015.0) **43** 95-101. DOI: 10.1515/jpm-2014-0086
24. Karaye KM, Ishaq NA, Sa'idu H, Balarabe SA, Talle MA, Isa MS. **Incidence, clinical characteristics, and risk factors of peripartum cardiomyopathy in Nigeria: results from the PEACE Registry**. *ESC Heart Fail* (2020.0) **7** 236-244. DOI: 10.1002/ehf2.12562
25. 25.Wu VC-C, Chen T-H, Yeh J-K, Wu M, Lu C-H, Chen S-W et al (2017) Clinical outcomes of peripartum cardiomyopathy: a 15-year nationwide population-based study in Asia. Med (Baltim) 96(43):e8374-e. 10.1097/MD.0000000000008374
26. Haghikia A, Podewski E, Libhaber E, Labidi S, Fischer D, Roentgen P. **Phenotyping and outcome on contemporary management in a German cohort of patients with peripartum cardiomyopathy**. *Basic Res Cardiol* (2013.0) **108** 366. DOI: 10.1007/s00395-013-0366-9
27. Waring AB, Wallace WH. **Subfertility following treatment for childhood cancer**. *Hosp Med* (2000.0) **61** 550-557. DOI: 10.12968/hosp.2000.61.8.1398
28. Hilfiker-Kleiner D, Kaminski K, Podewski E, Bonda T, Schaefer A, Sliwa K. **A cathepsin D-cleaved 16 kDa form of prolactin mediates postpartum cardiomyopathy**. *Cell* (2007.0) **128** 589-600. DOI: 10.1016/j.cell.2006.12.036
29. Ricke-Hoch M, Bultmann I, Stapel B, Condorelli G, Rinas U, Sliwa K. **Opposing roles of Akt and STAT3 in the protection of the maternal heart from peripartum stress**. *Cardiovasc Res* (2014.0) **101** 587-596. DOI: 10.1093/cvr/cvu010
30. Sharma RK, Agarwal A. **Role of reactive oxygen species in gynecologic diseases**. *Reprod Med Biol* (2004.0) **3** 177-199. DOI: 10.1111/j.1447-0578.2004.00068.x
31. Zhao Y, Zhang C, Huang Y, Yu Y, Li R, Li M. **Up-regulated expression of WNT5a increases inflammation and oxidative stress via PI3K/AKT/NF-kappaB signaling in the granulosa cells of PCOS patients**. *J Clin Endocrinol Metab* (2015.0) **100** 201-211. DOI: 10.1210/jc.2014-2419
32. Bedaiwy MA, Falcone T, Sharma RK, Goldberg JM, Attaran M, Nelson DR. **Prediction of endometriosis with serum and peritoneal fluid markers: a prospective controlled trial**. *Hum Reprod* (2002.0) **17** 426-431. DOI: 10.1093/humrep/17.2.426
33. Gnoth C, Godehardt E, Frank-Herrmann P, Friol K, Tigges J, Freundl G. **Definition and prevalence of subfertility and infertility**. *Hum Reprod* (2005.0) **20** 1144-1147. DOI: 10.1093/humrep/deh870
34. 34.Statistical Office of Germany (2000–2018) (2021) Total fertility rate (per woman): Germany, years, age groups of the women (Code: 12612-0009). https://www-genesis.destatis.de/genesis/online/?operation=sprachwechsel&language=en. Accessed 23 Nov 2021
35. 35.Statistical Office of Germany (2000–2018) (2021) Multiple births: Germany, years, type of multiple birth (Code: 12612-0016). https://www-genesis.destatis.de/genesis/online/?operation=sprachwechsel&language=en. Accessed 23 Nov 2021
36. 36.Statistical Office of Germany (2000–2018) (2021) Child deliveries in hospitals (number and percentage). Classification: years, region, type of child delivery. https://www-genesis.destatis.de/genesis/online/?operation=sprachwechsel&language=en. Accessed 23 Nov 2021
37. 37.Statistical Office of Germany (2017) Distribution of the population to groups in terms of smoking habits (in percent). Classification: years, Germany, age, sex, smoking habits (filtered for female and age 15–50). https://www-genesis.destatis.de/genesis/online/?operation=sprachwechsel&language=en. Accessed 23 Nov 2021
38. Kuc S, Wortelboer EJ, van Rijn BB, Franx A, Visser GH, Schielen PC. **Evaluation of 7 serum biomarkers and uterine artery Doppler ultrasound for first-trimester prediction of preeclampsia: a systematic review**. *Obstet Gynecol Surv* (2011.0) **66** 225-239. DOI: 10.1097/OGX.0b013e3182227027
39. Melchior H, Kurch-Bek D, Mund M. **The prevalence of gestational diabetes**. *Dtsch Arztebl Int* (2017.0) **114** 412-418. DOI: 10.3238/arztebl.2017.0412
|
---
title: Gut microbiota alters host bile acid metabolism to contribute to intrahepatic
cholestasis of pregnancy
authors:
- Bo Tang
- Li Tang
- Shengpeng Li
- Shuang Liu
- Jialin He
- Pan Li
- Sumin Wang
- Min Yang
- Longhui Zhang
- Yuanyuan Lei
- Dianji Tu
- Xuefeng Tang
- Hua Hu
- Qin Ouyang
- Xia Chen
- Shiming Yang
journal: Nature Communications
year: 2023
pmcid: PMC9998625
doi: 10.1038/s41467-023-36981-4
license: CC BY 4.0
---
# Gut microbiota alters host bile acid metabolism to contribute to intrahepatic cholestasis of pregnancy
## Abstract
Intrahepatic cholestasis of pregnancy (ICP) is a female pregnancy-specific disorder that is characterized by increased serum bile acid and adverse fetal outcomes. The aetiology and mechanism of ICP are poorly understood; thus, existing therapies have been largely empiric. Here we show that the gut microbiome differed significantly between individuals with ICP and healthy pregnant women, and that colonization with gut microbiome from ICP patients was sufficient to induce cholestasis in mice. The gut microbiomes of ICP patients were primarily characterized by *Bacteroides fragilis* (B. fragilis), and B. fragilis was able to promote ICP by inhibiting FXR signaling via its BSH activity to modulate bile acid metabolism. B. fragilis-mediated FXR signaling inhibition was responsible for excessive bile acid synthesis and interrupted hepatic bile excretion to ultimately promote the initiation of ICP. We propose that modulation of the gut microbiota-bile acid-FXR axis may be of value for ICP treatment.
Intrahepatic cholestasis of pregnancy (ICP) is a liver disease that sometimes develops during pregnancy and is characterized by increased serum bile acid levels. Here the authors report that the gut microbiome species B. fragilis is enriched in patients with ICP and promotes ICP development in mice via inhibition of signalling though the bile acid receptor FXR.
## Introduction
Intrahepatic cholestasis of pregnancy (ICP) is the most common pregnancy-related liver disease, which predominantly occurs in the second or third trimester and is mainly characterized by maternal pruritus and increased levels of serum bile acid and liver transaminases1. ICP is mainly associated with increased adverse perinatal outcomes, such as spontaneous preterm delivery; fetal distress, intrauterine death and growth restriction; low Apgar score and meconium-stained fluid with a small degree of maternal risk2,3. The aetiology and pathophysiology of ICP remain poorly understood, and therapies have been largely empiric. Ursodeoxycholic acid (UDCA) is commonly used to treat ICP and is effective in ameliorating pruritus by reducing alanine aminotransferase in ICP4–6. However, previous studies have shown that there is insufficient evidence to recommend UDCA to improve fetal outcomes and bile acid levels6,7, leading to controversy about its use in ICP.
The gut microbiota in the gastrointestinal tract are able to impact the functions related to immune, metabolic and inflammatory diseases8,9 and their composition can be shaped by various environmental factors, such as diet, immune system, host genetics, and hormones10–14. It is well known that the human body undergoes substantial immunological, hormonal and metabolic changes during a normal pregnancy15,16. Previous studies have shown that the gut microbiota in healthy pregnant women undergoes profound alterations from the first to the third trimester17–19. Moreover, gut microbiota dysbiosis may be sufficient to induce disease even in otherwise nonpredisposed individuals20. Our group recently reported that the gut microbiota of individuals with preeclampsia (PE) were dysbiotic and that microbiota transplantation from patients with PE could promote a PE-like phenotype in pregnant mice21. Additionally, previous studies have revealed a close link between microbial dysbiosis and many pregnancy-related diseases22–24. This highlights the importance of understanding microbiome changes during pregnancy, which might offer a promising approach of preemptively modulating the microbiota before conception or during early pregnancy to reduce the risk of adverse outcomes. A previous study revealed that there was a notable metabolic change during the gestation period that induced elevated serum bile acids in pregnancy25. Primary bile acids synthesized from cholesterol in the liver are conjugated with glycine and taurine and secreted into the intestine. Then, the primary bile acids are transformed into secondary bile acids, and the deconjugation of the bile acids is mediated by intestinal bacteria26, with FXR being the key receptor that regulates hepatic bile acid biosynthesis, transport and secretion26.
Recent studies using 16S rRNA sequencing have revealed that compared to healthy controls, patients with ICP have different gut microbiota profiles27,28. However, there is insufficient evidence to resolve the cause-and-effect relationships and to determine whether gut microbiota changes are a consequence of disease or contribute to the development of ICP. In particular, the particular species of gut microbiota involved and the underlying mechanism by which the gut microbiota impact ICP pathogenesis are still unknown. Herein, we aimed to determine whether ICP is associated with a specific gut microbiome profile and to clarify the role of the gut microbiota in ICP pathogenesis.
Our present study showed that a distinct microbial profile was present in ICP patients, and that the abundance of *Bacteroides fragilis* was substantially increased in individuals with ICP. Transplantation of the faecal microbiota from patients with ICP was sufficient to promote an ICP-like phenotype. We further suggest a mechanism involving microbe-mediated bile acid metabolism and discuss how it might regulate FXR signaling and the incidence of ICP. Our data highlight the importance of the gut microbiota in ICP development and provide a potential target for the clinical management of ICP.
## Gut microbiota from patients with ICP is altered significantly and sufficiently to promote intrahepatic cholestasis of pregnancy in mice
BMI- and age-matched pregnant women with ICP ($$n = 50$$) and control pregnant women ($$n = 41$$) were recruited. We then carried out whole-genome shotgun sequencing on faecal samples from these healthy pregnant women and patients with ICP (for demographic and clinical characteristics see Supplementary Table 1). A metagenomic dataset with an average of 40,420,272 ± 3,544,978 paired-end reads per sample was obtained. Raw reads were preprocessed using KneadData to eliminate human DNA sequences and to filter sequences with poor quality, which removed $5.26\%$ of the reads. Ultimately, 40,420,272 ± 3,544,978 reads per sample were obtained. The sequences were analysed with MetaPhLan3 implemented within the HUMAnN3 pipeline. We selected a total of 904 clades including 12 phyla (L2), 23 classes (L3), 37 orders (L4), 73 families (L5), 192 genera (L6) and 567 species (L7). Alpha diversity was calculated to evaluate the differences between healthy controls and ICP patients. No significant difference in alpha diversity was observed among the taxon levels (Wilcoxon rank-sum test; Supplementary Fig. 1). For beta diversity, weighted Unifrac distances were analysed and plotted by principal coordinate analysis (PCoA) (ANOSIM; $$P \leq 0.006$$; Fig. 1a). The weighted Unifrac rank between the two groups was much higher than that within each group (Fig. 1b).Fig. 1Gut microbiota from ICP patients is altered significantly and sufficient to promote ICP in mice.a Weighted Unifrac PCoA (principal coordinate analysis) plot of individuals with ICP patients and healthy controls. The ANOSIM test was used to calculate the significance of dissimilarity (ANOSIM, $$P \leq 0.001$$). b A truncated violin plot shows the comparison of Weighted Unifrac range of samples between ICP patients and healthy controls. The notch in the middle of the box represents the median. The top and bottom of box represent the 75th and 25th quartiles, respectively. The upper and lower whiskers extended 1.5× the interquartile range from the upper edge and lower edge of the box represent maximum and minimum, respectively. With a truncated violin plot, the curve of the violin extends only to the minimum and maximum values in the data set. Outlier values were not shown in this plot. $$n = 50$$ individuals with ICP, $$n = 41$$ individuals in healthy controls. The ANOSIM test was used to calculate the significance of dissimilarity. The exact P values were shown. c Experimental schematic: following antibiotics treatment, the recipient mice were transplanted with fecal samples from ICP patients or healthy controls. Tissues and samples were collected at E18d. d Analysis of serum levels of total bile acids, ALT, AST, ALP, and GGT ($$n = 8$$ mice per group). Data are presented as mean ± SEM. P values were determined by two-tailed Student’s t-test. e Hematoxylin and eosin staining of representative livers. Scale bar: 100 μm. f, g Number of pups per litter ($$n = 8$$ mice per group). The representative fetuses are from the same litter with the same gestational age. h–j Fetal weight (h), placenta weight (i) and ratio of dead fetus (j) in the two groups ($$n = 8$$ mice per group). P values were determined by two-tailed Student’s t-test for g–j. k Placenta tissues were stained with H&E. Representative images were shown. Scale bar: 50 μm. Data are presented as mean ± SEM for g–j. FMT fecal microbiota transplantation, ABX antibiotics cocktail, PC principal coordinate analysis, ANOSIM analysis of similarities. Source data are provided as a Source Data file.
To investigate the role of the gut microbiota in ICP, stool samples from healthy controls or patients with ICP was transplanted into recipient mice (Fig. 1c). We further collected faeces from mice receiving ICP stool transplants and from mice receiving healthy donor stool transplants and performed 16S rRNA sequencing to validate the fidelity of the microbiota transplantation. The gut microbiota of mice receiving ICP stool transplants was altered significantly compared to that of control mice (Supplementary Fig. 2). Compared with mice receiving stool transplants from healthy pregnant women, mice receiving ICP stool transplants displayed increased levels of total bile acid (TBA), aspartate transaminase (AST), alanine aminotransferase (ALT), alkaline phosphatase (ALP) and gamma-glutamyl transpeptidase (GGT) (Fig. 1d). The hepatic tissue in mice receiving stool transplants from healthy controls exhibited a normal morphology, whereas cytoplasm rarefaction, nuclear condensation, vacuolar degeneration, portal oedema and a loss of hepatic structure in periportal areas were observed in hepatic tissues of mice receiving stool transplants with microbiota from patients with ICP (Fig. 1e). The number of live pups (Fig. 1f, g), fetal weight (Fig. 1h) and placental weight (Fig. 1i) were significantly decreased in the mice receiving stool transplants from women with ICP, and the dead fetus rate (Fig. 1j) in these mice was higher than the rate in mice receiving transplants of stool from healthy controls. Histomorphological analysis revealed that fecal transplantation from individuals with ICP resulted in intracellular oedema and severe atrophy of trophoblasts in the placenta (Fig. 1k). Thus, our data demonstrated that microbiota transplantation from patients with ICP into mice could transfer ICP-related phenotypes.
## B. fragilis induced intrahepatic cholestasis in pregnant mice
We further sought to identify the differentiating patterns in bacterial taxa between the two groups and to explore the microbial species associated with ICP parameters. The microbiomes of individuals with ICP were primarily characterized by Bacteriodes fragilis, Klebsiella pneumoniae, Klebsiella variicola, Klebsiella quasipneumoniae, Weissella confusa, Citrobacter youngae, and *Enterobacter cloacae* (Wilcoxon rank-sum test; Fig. 2a). In addition, some of these microbial species were associated with ICP clinical parameters, such as levels of TBA, ALT, AST, ALP, GGT, birth weight and Apgar score (Fig. 2b). To further evaluate the relationship between microbiota and ICP severity, we divided the patients with ICP into two groups on the basis of TBA levels (mild group, serum TBA range 10–39.9 μmol/L; severe group serum TBA ≥ 40 μmol/L). We compared the demographic characteristics between the mild and severe group and found that there were significant differences in TBA levels, ALT, ALP, GGT, neonate weight and Apgar score (Supplementary Table 2). Although there was no notable difference in the alpha and beta diversity between the mild and severe groups (Supplementary Fig. 3; Fig. 2c), we observed that the microbiomes of individuals with severe ICP were primarily characterized by *Bacteriodes fragilis* (B. fragilis) (Fig. 2d), and that the abundance of B. fragilis was markedly increased in the severe group compared to the mild group (Fig. 2e). B. fragilis was positively correlated with the levels of TBA, ALT and AST but negatively related to gestational age, birth weight and Apgar score (Fig. 2b).Fig. 2B. fragilis induced intrahepatic cholestasis of pregnancy in mice.a Differentially enriched bacteria between ICP patients and healthy controls. Pink dots represent bacteria with higher abundance in samples of ICP patients. Blue dots represent bacteria with higher abundance in samples of healthy controls. Grey dots represent bacteria with no significant difference between the two groups. P values were adjusted by Benjamini & Hochberg (BH) method to control FDR. FDR-adjusted $P \leq 0.05$ was shown. b Correlation heatmap of differentially enriched bacteria with the clinical characteristics. Pearson correlation analysis with FDR-adjusted P value. * $P \leq 0.05$; **$P \leq 0.01$; ***$P \leq 0.001.$ c Weighted Unifrac PCoA plot of samples between the severe and mild groups in ICP patients. The ANOSIM test was used to calculate the significance of dissimilarity (ANOSIM, $$P \leq 0.869$$). d Differentially enriched bacteria between the severe and mild groups in ICP patients. Pink dots represent bacteria with relatively higher abundance in severe ICP patients. Grey dots represent bacteria with no significant difference between the two groups. P values were adjusted by Benjamini & Hochberg (BH) method to control FDR. FDR-adjusted $P \leq 0.05$ was shown. e The comparison of B. fragilis abundance between the two groups. The horizontal bar within box represents median. The top and bottom of box represent 75th and 25th quartiles, respectively. The upper and lower whiskers extended 1.5× the interquartile range from the upper edge and lower edge of the box represent maximum and minimum, respectively. P value was determined by two-tailed Mann–Whitney test. f Experimental schematic for gavaging B. fragilis. g Analysis of serum levels of total bile acids, ALT and AST in each group. Data are presented as mean ± SEM. P values were determined by ordinary one-way ANOVA with Tukey’s correction or Welch ANOVA with Games-Howell’s multiple comparisons test. $$n = 6$$ mice in control group, $$n = 8$$ mice in B. fragilis and EE2 group. h Representative images of H&E staining of livers in each group. Scale bar: 100 μm. i–l Number of pups per litter (i) in each group. $$n = 6$$ in control group, $$n = 8$$ in B. fragilis and EE2 group. Number of fetal weight (j) in each group. $$n = 6$$ in control group, $$n = 7$$ in B. fragilis and EE2 group. Number of placenta weight (k) and ratio of dead fetus (l) in each group. $$n = 6$$ in control group, $$n = 8$$ in B. fragilis and EE2 group. Data are presented as mean ± SEM for i–l. P values were determined by ordinary one-way ANOVA with Tukey’s correction for i–l. m Representative images of H&E staining of placenta in each group. Scale bar: 50 μm. ANOSIM analysis of similarities, EE2 17α-ethynylestradiol; *Source data* are provided as a Source Data file.
We further tested whether B. fragilis might play an important role in the development of ICP. Hence, we transplanted B. fragilis into C57BL/6 mice using oral gavage after antibiotic-based microbiota depletion (Fig. 2f). The EE2-induced ICP mouse model was used as the positive control (Fig. 2f). Notably, we observed that colonization of B. fragilis caused increased levels of TBA (Fig. 2g) and cholestatic liver injury (Fig. 2h; Supplementary Fig. 4a–c), disrupted normal fetal growth (Fig. 2i–l; Supplementary Fig. 4d), and induced histopathological abnormalities of the placenta in recipient mice (Fig. 2m). Our data suggest that B. fragilis could promote intrahepatic cholestasis in a pregnant mouse model.
## B. fragilis promotes ICP through its BSH activity mediating bile acid metabolism
Since there was a difference in microbiome composition between healthy controls and ICP patients, we further investigated the functional differences by analysing the functional pathways. We performed Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis and observed that several pathways were enriched in the ICP patients. Among these pathways, bile acid metabolism was the top metabolic pathway affected by the ICP microbiome (Fig. 3a). We found that the serum concentrations of most conjugated bile acids were much higher in patients with ICP (Supplementary Fig. 5a), and that several conjugated bile acids were decreased in fecal samples from patients with ICP (Supplementary Fig. 5b, c). To identify whether fecal transplantation from ICP patients or B. fragilis colonization was responsible for the bile acid metabolism patterns, we performed targeted metabolomics analyses. We observed that fecal transplantation from patients with ICP could decrease the ratio of conjugated bile acids (Supplementary Fig. 6). We also found that the levels of conjugated bile acids in the small intestine were dramatically decreased in B. fragilis-colonized mice (Fig. 3b, c), with a significant reduction in tauro-α-muricholic acid (TαMCA), tauro-β-muricholic acid (TβMCA), and taurodeoxycholic acid (TDCA).Fig. 3B. fragilis promotes the ICP through its BSH activity mediating bile acid metabolism.a Kyoto Encyclopedia of Genes and Genomes (KEGG) annotation of key altered metabolic pathways in individuals with ICP. P values were adjusted by Benjamini & Hochberg (BH) method to control FDR. FDR-adjusted $P \leq 0.05$ was shown. b Levels of small intestinal bile acids in B. fragilis transplanted mice or control group ($$n = 6$$ mice per group). P values were determined by two-tailed Student’s t-test. * $P \leq 0.05.$ c Conjugated and unconjugated bile acids in B. fragilis transplanted mice or control group. $$n = 6$$ per group. P values were determined by two-sided Mann–Whitney test. d Hydrolysis efficiency of GDCA by B. fragilis with or without CAPE in vitro. P values were determined by Welch ANOVA with Games-Howell’s multiple comparisons test. $$n = 6$$ per group. e–i Mice were divided into three groups (control, B. fragilis and B. fragilis + CAPE). Analysis of serum levels of total bile acids (e), ALT (f), AST (g), ALP (h) and GGT (i) in each group. $$n = 6$$ per group. P values were determined by ordinary one-way ANOVA with Tukey’s correction. Fetal weight (j) and placenta weight (k) in each group ($$n = 6$$ in control group; $$n = 8$$ in B. fragilis group and B. fragilis + CAPE group). P values were determined by ordinary one-way ANOVA with Tukey’s correction. l Representative images of H&E staining of livers and placenta in each group. Scale bar = 100 μm for liver; scale bar = 50 μm for placenta. m Levels of small intestinal bile acids in indicated groups ($$n = 6$$ per group). P value was determined by one-way ANOVA with Tukey’s correction. * $P \leq 0.05.$ Data are presented as mean ± SEM for b–k, m. CAPE caffeic acid phenethyl ester. Source data are provided as a Source Data file.
It is well known that BSH enzymes in microbial genera can deconjugate glycine or taurine-conjugated bile acids26. Furthermore, we investigated whether B. fragilis with high BSH activity could directly mediate the deconjugation of bile acids in vitro. We observed that B. fragilis could mediate the deconjugation of conjugated bile acids which was attenuated by the BSH inhibitor caffeic acid phenethyl ester (CAPE) (Fig. 3d). B. fragilis had lower activity in conjugated CA or CDCA compared to conjugated DCA (Supplementary Fig. 7), demonstrating the substrate specificity of the B. fragilis BSH enzyme. To explore the function of BSH activity enriched in B. fragilis in the development of the ICP phenotype, CAPE was added during the B. fragilis gavage (Supplementary Fig. 8a). CAPE significantly reversed B. fragilis-induced cholestatic liver injury (Fig. 3e–i; Supplementary Fig. 8b), fetal growth inhibition (Fig. 3j, k; Supplementary Fig. 8c), and placental abnormalities (Fig. 3l). Similarly, CAPE supplementation attenuated B. fragilis-induced deconjugation of conjugated bile acids in the mouse small intestine in vivo (Fig. 3m). Thus, our results suggested that B. fragilis could promote the ICP phenotype through its deconjugation of BSH activity.
## B. fragilis suppressed FXR signaling by mediating bile acid metabolism to increase hepatic bile acid accumulation
Since massive bile acid accumulation is closely associated with the incidence and development of cholestatic liver diseases, we previously investigated and found that B. fragilis could increase the total serum bile acid level. We further observed that hepatic bile acid levels were significantly increased by B. fragilis colonization in mice (Fig. 4a). We found that hepatic levels of T-αMCA, T-βMCA and TCA were significantly elevated in B. fragilis-colonized mice (Fig. 4b). Notably, B. fragilis caused a significant increase in the serum 7α-hydroxy-4-cholesten-3-one (C4) level, which is a surrogate of bile acid synthesis (Fig. 4c). We also observed that the increased conjugated bile acids induced by B. fragilis colonization could activate sphingosine-1-phosphate receptor 2 (S1PR2) and promote inflammation in the liver (Supplementary Fig. 9). Since bile acids mainly exert their functions by interacting with receptors such as FXR, we further tested intestinal and liver FXR signaling. In the intestine, FXR activation induced FGF15 expression in mice, and FGF15 bound to the FGF receptor/β-Klotho complex on hepatocytes and repressed bile acid synthesis. Our results showed that FGF15 levels in the serum of mice receiving transplants with ICP microbiota were dramatically decreased (Supplementary Fig. 10a), and the expression levels of downstream genes of FXR were markedly suppressed in the ileum and liver of ICP microbiota-transplanted mice (Supplementary Fig. 10b, c), suggesting that intestinal and hepatic FXR signaling was inhibited by ICP gut microbiota transplantation. Additionally, the expression of FGF15 and Shp in ileal tissues was significantly reduced by B. fragilis colonization (Fig. 4d, e). The expression of FXR target genes in hepatic tissues was similarly reduced by B. fragilis colonization (Fig. 4f). These data suggest that B. fragilis administration markedly suppressed intestinal and hepatic FXR signaling. Fig. 4B. fragilis suppressed FXR signaling through mediating bile acid metabolism to increase hepatic bile acids accumulation.a Liver total bile acid levels in B. fragilis transplanted mice or control group ($$n = 6$$ per group). b Hepatic bile acids profiles in B. fragilis transplanted mice or control group ($$n = 6$$ per group). P values were determined by two-tailed Student’s t-test for a, b. *$P \leq 0.05.$ c, d Serum C4 levels (c) and FGF15 levels (d) in B. fragilis transplanted mice or control group ($$n = 6$$ per group). P values were determined by two-tailed Student’s t-test. e, f Relative expression of intestinal (e) and hepatic (f) FXR mRNA and its target genes in mice colonized with B. fragilis or control ($$n = 6$$ per group). P values were determined by two-tailed Student’s t-test. g TR-FRET FXR coactivator recruitment assay to assess the action of GDCA on FXR; CDCA and GW4064 was used as positive control. GUDCA was used as negative control ($$n = 3$$ per group). h Serum C4 levels in each group ($$n = 6$$ per group). i Liver total bile acid levels in each group ($$n = 6$$ per group). P values were determined by one-way ANOVA with Tukey’s correction for h–j Representative images of H&E staining of livers and placenta in each group. Scale bar = 100 μm for liver; scale bar = 50 μm for placenta. k, l Relative expression of intestinal FXR target genes (k) and hepatic FXR target genes (l) in indicated groups. $$n = 6$$ per group. P values were determined by one-way ANOVA with Tukey’s correction. Data are presented as mean ± SEM for a–i, k, l. Source data are provided as a Source Data file.
Since we previously found that B. fragilis could directly mediate the deconjugation of GDCA through its BSH activity, we further investigated whether the FXR signaling change resulted from B. fragilis-related bile acid metabolism. We found that GDCA could substantially induce FXR transactivation, which was inhibited by GUDCA (Supplementary Fig. 11a). Furthermore, we observed that GDCA induced the expression levels of the FXR target genes FGF19 and Shp in a dose-dependent manner in the human intestinal Caco-2 cell line (Supplementary Fig. 11b). To further clarify the interaction between GDCA and FXR, in silico molecular docking studies were carried out. The docking results showed that GDCA could bind well to FXR, with a docking score of 13.43. The hydrogen bonding between the residues Met265, Met328, Arg331 and GDCA played an important role in the observed interaction (Supplementary Fig. 12a). In addition, the molecular backbone of GDCA is known to have a wide range of hydrophobic interactions with the receptor, facilitating the stabilization of the hydrophobic core of the receptor, which might have acted as a possible direct agonist of FXR in this study (Supplementary Fig. 12b). We further performed a TR-FRET FXR coactivator assay to validate whether GDCA was a direct FXR agonist. Similar to CDCA and GW4064, which were the positive controls, GDCA showed agonistic action (Fig. 4g). We further observed that supplementation of GDCA to B. fragilis-treated mice decreased the serum C4 level (Fig. 4h), consistent with previous studies that have shown GDCA can inhibit bile acid synthesis29,30. Supplementation with GDCA decreased serum and hepatic TBA levels (Supplementary Fig. 13a; Fig. 4i) and liver injury (Fig. 4j; Supplementary Fig. 13b–e) and increased pup numbers (Supplementary Fig. 13f, g), fetal weight (Supplementary Fig. 13h), placental weight (Supplementary Fig. 13i), and morphological changes in the placenta (Fig. 4j), while deoxycholic acid (DCA) and glycochenodeoxycholic acid (GCDCA) were unable to mitigate and even aggravated B. fragilis-induced cholestasis (Supplementary Fig. 14). Furthermore, the expression levels of FXR target gene mRNAs were significantly reversed by GDCA (Fig. 4k, l). Collectively, our data indicate that B. fragilis suppressed FXR signaling by modulating bile acid metabolism and induced hepatic bile acid accumulation to promote ICP development.
## FXR signaling is required for B. fragilis-induced ICP in mice
Since our previous data revealed that B. fragilis-mediated FXR signaling suppression, we further sought to investigate the role of FXR in B. fragilis-induced ICP phenotype. We hypothesized that FXR activation may restore the effects of B. fragilis. To test this hypothesis, we used the GW4064, a general FXR agonist, and found that GW4064 could substantially reverse B. fragilis-induced cholestasis (Fig. 5a), liver injury (Fig. 5b–d; Supplementary Fig. 15a–d), fetal growth inhibition (Fig. 5e, f; Supplementary Fig. 15e), and morphological changes in the placenta (Fig. 5g).Fig. 5FXR signaling is required for B. fragilis-induced ICP in mice.a–c Analysis of serum levels of total bile acids (a), ALT (b) and AST (c) in each group ($$n = 6$$ per group). P values were determined by one-way ANOVA with Tukey’s correction or Welch ANOVA with Games-Howell’s multiple comparisons test. d Representative images of H&E staining of liver in each group. Scale bar: 50 μm. e, f Fetal weight (e) and placenta weight (f) in each group ($$n = 6$$ per group). P values were determined by one-way ANOVA with Tukey’s correction. g Representative images of H&E staining of placenta in each group. Scale bar: 50 μm. h Liver total bile acid levels in each group ($$n = 6$$ per group). i, j Relative expression of intestinal FXR target genes (i) and hepatic FXR target genes (j) in indicated groups. $$n = 6$$ per group. P values were determined by one-way ANOVA with Tukey’s correction. k, l Serum levels of total bile acids, ALT, AST, ALP, GGT in each group ($$n = 6$$ per group). m–o Number of pups per litter (m), fetal weight (n), and placenta weight (o) in each group ($$n = 6$$ per group). P values were determined by one-way ANOVA with Tukey’s correction or Welch ANOVA with Games-Howell’s multiple comparisons test or Kruskal-Wallis with Dunn’s multiple comparisons test depending on the sample distribution type for h-o. Data are presented as mean ± SEM for a–c, e, f, h–o. Source data are provided as a Source Data file.
Moreover, we observed that GW4064 significantly reversed B. fragilis-induced hepatic bile acid levels (Fig. 5h). As expected, the reduced expression of FXR target genes in intestine and hepatic tissues were almost completely restored by GW4064 (Fig. 5i, j). Furthermore, B. fragilis was gavaged into FXR knockout (FXR−/−) mice and wild-type (WT) mice. Consistent with previous results, B. fragilis greatly suppressed intestinal and hepatic FXR signaling in WT mice, while it had no effects in FXR−/− mice (Supplementary Fig. 16a, b). B. fragilis substantially promoted cholestatic liver injury (Fig. 5k, l; Supplementary Fig. 16c, d), disrupted fetal growth (Fig. 5m, n), and induced placental dysfunction in WT mice (Fig. 5o; Supplementary Fig. 16e), whereas it had no effect in FXR−/− mice because of FXR deletion in these mice. Thus, our data revealed that FXR signaling was required for B. fragilis-mediated ICP-like phenotype.
## B. fragilis induces excessive bile acid synthesis and inhibits hepatic bile acid excretion through suppression of FXR signaling to promote ICP
The possible mechanisms by which B. fragilis/FXR signaling regulates bile acid metabolism and cholestasis in ICP were further investigated. Since cholestasis is mainly characterized by bile flow impairment that mainly involves excessive bile acid synthesis and disrupted hepatic bile excretion, hepatic bile acid synthesis and excretory genes were then measured in different mouse models. Mice transplanted with ICP stool exhibited much higher levels of Cyp7a1, Cyp8b1 and Cyp27a1 mRNAs and lower expression levels of BSEP and MRP2 than control livers (Fig. 6a). Similarly, we observed that B. fragilis induced bile acid synthesis markers and decreased the bile excretory genes BSEP and MRP2 in the liver (Fig. 6b). Notably, the relative mRNA expression levels of Cyp7a1, Cyp8b1, Cyp27a1, BSEP and MRP2 in the liver were dramatically reversed by GDCA administration in B. fragilis-treated mice (Fig. 6c). Immunohistochemical staining confirmed that GDCA administration significantly reversed the protein expression levels of the hepatic bile acid synthetic enzymes Cyp7a1, Cyp8b1, Cyp27a1 and the excretory proteins BSEP and MRP2 in the livers of B. fragilis-gavaged mice (Fig. 6d). Additionally, modulation of FXR signaling by the agonist GW4064 significantly attenuated B. fragilis-induced bile acid synthetic markers and elevated BSEP and MRP2 expression levels (Fig. 6e). We further observed that B. fragilis significantly upregulated bile acid synthesis genes expression and suppressed the expression of BSEP and MRP2 in WT mice, whereas no further effect was noted in FXR−/− mice (Fig. 6f). These data demonstrated that B. fragilis-mediated FXR signaling inhibition was responsible for excessive bile acid synthesis and interrupted hepatic bile excretion to ultimately promote the initiation of ICP. We further evaluated the relationship among B. fragilis, FXR signaling and bile acid synthesis in the ICP cohort. We found that the relative abundance of B. fragilis was negatively correlated with the levels of FGF19 (Fig. 6g), but positively related to the concentrations of C4 (Fig. 6h). Together, these clinical data suggest that B. fragilis suppressed FXR signaling to increase bile acid synthesis during ICP development, implying a promising intervention target for ICP treatment. Fig. 6B. fragilis induces excessive bile acid synthesis and inhibits hepatic bile acid excretion through suppression of FXR signaling to promote ICP.a Hepatic mRNA expression levels of bile acids synthetic and bile excretory genes in mice transplanted with fecal microbiota of ICP and healthy controls ($$n = 6$$ per group). P values were determined by two-tailed Student’s t-test. b Hepatic mRNA expression levels of bile acids synthetic and bile excretory genes in control group, B. fragilis group and EE2 group ($$n = 6$$ per group). c Hepatic mRNA expression levels of bile acids synthetic and bile excretory genes in each group ($$n = 6$$ per group). P values were determined by Welch ANOVA with Games-Howell’s multiple comparisons test for b, c. d IHC staining of hepatic bile acids synthetic and bile excretory proteins of mice colonized with B. fragilis together with GDCA or not. Scale bar: 20 μm. e Hepatic mRNA expression levels of bile acids synthetic and bile excretory genes in mice colonized with B. fragilis together with GW4064 (10 mg/kg/d) or not ($$n = 6$$ per group). f Hepatic mRNA expression levels of bile acids synthetic and bile excretory genes in WT mice or FXR−/− mice colonized with B. fragilis or not ($$n = 6$$ per group). P values were determined by Welch ANOVA with Games-Howell’s multiple comparisons test for e, f. g, h Correlations between B. fragilis abundance and FGF19 (g) or C4 (h) were determined by Spearman’s rank test. i Schematic mechanisms underlying the role of the B. fragilis-bile acid-FXR axis in regulating ICP. Data are presented as mean ± SEM for a-c, e, f. Source data are provided as a Source Data file.
## Discussion
ICP is emerging as a noteworthy pregnancy-related disease that is accompanied by an increased risk of adverse perinatal outcomes. As the mechanism remains poorly understood, therapies have been largely empiric, and there remains an urgent unmet clinical need. Herein, we found that the gut microbiota were changed in patients with ICP and that transplantation of gut microbiota from patients with ICP was sufficient to promote an ICP phenotype in mice. Notably, the microbiomes of patients with ICP were primarily characterized by B. fragilis, and B. fragilis was markedly increased in patients with severe ICP. B. fragilis suppressed FXR signaling by mediating bile acid metabolism through its BSH activity.
Studies have demonstrated that specific alterations in the gut microbiome during pregnancy are closely related to many pregnancy-related diseases, including gestational hypertension, GDM and metabolic syndrome22–24. A previous study revealed that normal pregnancy was characterized by a higher ratio of Bacteroidetes to Firmicutes25. Another study also showed that there was notable variation in Bacteroides and Firmicutes abundances between samples from the first (T1) to third (T3) trimesters17. It is possible that the gestational increase in oestrogen, progesterone and corticosteroids exerts microbial selection pressures on the growth of different gut bacteria31,32. A previous study reported that the abundance of Blautia, Citrobacter and *Streptococcus was* significantly higher in ICP patients27, and a higher proportion of the genera Escherichia_Shigella, Olsenella, and Turicibacter was observed in patients with severe ICP 28. In the present study, we observed a notable difference in microbial profiles between individuals with ICP and healthy controls. We found that the microbiomes of patients with ICP were primarily characterized by Bacteriodes fragilis, Klebsiella pneumoniae, Klebsiella variicola, Klebsiella quasipneumoniae, Weissella confusa, Citrobacter youngae, and Enterobacter cloacae, while the microbiomes of patients with severe ICP were mainly characterized by B. fragilis. We noticed that there were differences and discrepancies among the previous two studies and our study. We believe the following factors might have contributed to these discrepancies: [1] Previous studies used 16S rRNA sequencing, while we used metagenomic sequencing analysis. [ 2] Previous studies characterized taxonomic differences through LefSe analysis, while we used MaAsLin2 to analyze the bacterial taxonomic differences. [ 3] The stool samples for sequencing in the previous studies were from one area, while we recruited patients with ICP and healthy controls from different regions. We believe that there might also be other factors contributing to the discrepancies. Although we did not observe a significant difference in the microbial diversity between the mild and severe ICP groups, B. fragilis was notably enriched and increased in abundance in individuals with severe ICP compared to the mild group and was positively correlated with the core features of ICP. It is well known that bile acids and the gut microbiota maintain a complex and bidirectional relationship. Elevated levels of bile acids and Bacteroidetes have been found in normal pregnancy with advancing gestation25. Previous studies have identified B. fragilis as a bile-resistant bacteria 33, and have revealed that bile salts can enhance resistance to structurally unrelated antimicrobial agents, bacterial coaggregation, intestinal colonization and biofilm formation of B. fragilis34. Here, the present study showed that the abundance of B. fragilis was increased in patients with ICP and severe ICP. Thus, we believe that there is a possibility that the increased bile acids could cause increases in B. fragilis abundance and could enhance intestinal colonization.
Gut microbiota are able to participate in various metabolic processes, including those related to bile acids, short-chain fatty acids (SCFAs), trimethylamine-N-oxide (TMAO), endogenous ethanol, indole, and other metabolites, which are important in intestinal barrier function, the immune system, and host intestinal homeostasis35–37. We observed that conjugated bile acid levels were dramatically decreased after faecal transplantation from patients with ICP or after B. fragilis transplantation. Bile acid deconjugation is mediated by gut microbes via bile salt hydrolase (BSH) activity. In fact, functional BSH is present in some major bacterial groups, including Lactobacillus, Bifidobacterium, *Clostridium and* Bacteroides26. The human and mouse bile acid profiles are known to differ with the predominance of taurine conjugation in mice and glycine conjugation in humans. We observed that the levels of glycine-conjugated bile acids were much lower and that some were undetectable in the mouse small intestine. Bacterial BSH from the faeces of patients with ICP or with B. fragilis was able to convert GDCA (in humans) or TDCA (in mice) to DCA. Previous studies have reported that the BSH enzyme has substrate specificity, which might be affected by the optimal pH, substrate concentrations, different monomeric subunits of the enzyme, and so on38,39. The BSH enzyme in B. fragilis has been found to have the strongest binding affinities for the glycine or taurine conjugates of DCA (GDCA or TDCA) and the weakest binding affinities for glycine or taurine conjugates of CA (GCA or TCA)38,40. Moreover, BSH from B. fragilis shows lower activity using GCDCA and TCA as substrates and undetectable activity with glycine- or taurine-conjugated LCA as substrates38. We also found that B. fragilis had lower activity in conjugated CA or CDCA compared to conjugated DCA, further demonstrating the substrate specificity of the BSH enzyme in B. fragilis, which might contribute to the differences in bile acid species after B. fragilis colonization. Here, we observed that TDCA levels (corresponding to GDCA levels in humans) were much lower in mice transplanted with faeces from patients with ICP or with B. fragilis colonization. Although we did not observe a significant difference in DCA levels, there was an increasing trend, which might cause more damage to the liver and promote ICP development. GDCA is nearly undetectable in the mouse small intestine because taurine conjugation is predominant in mice. Since GDCA (corresponding to TDCA in mice) is predominant in humans and B. fragilis can deconjugate GDCA with higher activity, we focused on the role of GDCA in B. fragilis-induced ICP.
FXR is expressed in the ileum and liver and has been extensively explored for its important role in bile acid homeostasis41,42. FXR activation can lead to downregulation of bile acid synthesis, increase bile acid excretion, and decrease bile acid reabsorption43. A previous study showed that FXR function was reduced in pregnancy, together with reduced enterohepatic cycling and elevated serum bile acids25. We also observed a reduction in intestinal FXR signaling and enterohepatic feedback in patients with ICP and found that intestinal and hepatic FXR signaling was inhibited after B. fragilis colonization. We found that B. fragilis could cause cholestasis through inhibiting FXR signaling, while B. fragilis could not cause worse cholestasis in FXR knockout mice, suggesting that FXR is essential for B. fragilis-induced cholestasis. However, the whole body FXR knockout is limited in identifying the specific role of intestinal and hepatic FXR signaling in B. fragilis-induced cholestasis. It is known that alterations in the intestinal bile acid composition, resulting from bacterial bile acid modification, can impact enterocyte FXR induction in two ways: by impairing bile acid uptake by the enterocyte and/or by changing the ratio of agonistic to antagonist bile acid ligands25. We observed lower levels of conjugated bile acids, with a decrease in T-αMCA, T-βMCA, and TDCA in the mouse small intestine after B. fragilis colonization. Since T-αMCA and T-βMCA are FXR antagonists44, we hypothesized that the profiles of small intestinal bile acids in B. fragilis-colonized mice might not be the primary contributors to the impairment of intestinal FXR function, given that there were reductions in the antagonistic species. We believe that the impairment of intestinal FXR signaling might be attributed to the decrease in conjugated bile acids and reduced terminal ileal uptake because of the lower ASBT affinity of unconjugated bile acids in ileal uptake. Previous studies reported that B. fragilis also possessed 7α-hydroxysteroid dehydrogenase (7α-HSDH) activity, which could convert chenodeoxycholic acid (CDCA) into its 7-oxo derivative (7-oxo-lithocholic acid, 7-oxo-LCA)45,46. We observed a decreasing trend in CDCA levels in the small intestine of B. fragilis-colonized mice, which might further contribute to the reduced intestinal FXR signaling. In addition, we revealed that hepatic levels of T-αMCA and T-βMCA were significantly increased in B. fragilis-colonized mice, and given that T-αMCA and T-βMCA have been reported to be FXR antagonists44, these increases might contribute to the reduced hepatic FXR signaling.
A previous study reported that DCA could activate FXR, while GDCA or TDCA were inactive against FXR in CHO cells transfected with FXR47. Wang et al. reported that the glycine or taurine conjugates of CDCA and DCA could activate FXR in vivo but were inactive in certain cell lines, and further revealed that the conjugated bile acids could not readily cross cell membranes, leading to the inactivity of FXR in cell lines. The researchers cotransfected bile acid transporters in cell lines and observed that the glycine or taurine conjugates of CDCA and DCA were highly effective in activating FXR in cells coexpressing bile acid transporters48. Thus, we believe that the different research methods might have caused the discrepancy in the role of GDCA in activating FXR signaling observed in previous studies. In the present study, we applied the TR-FRET FXR coactivator assay and luciferase reporter assay in cells transfected with bile acid transporters to verify that GDCA could activate FXR in vitro. The primary bile acids CA and CDCA and their conjugated counterparts are the major bile acids that are elevated in the serum of patients with ICP, while DCA and GDCA are not the predominant bile acid species. However, we revealed that CA and CDCA or their conjugated counterparts were not changed, while GDCA or TDCA was changed significantly in faecal samples of patients with ICP or in mouse small intestines of ICP faecal transplantation or B. fragilis colonization mice. Thus, we focused on bile acid and its role in B. fragilis-induced ICP development. We revealed that GDCA administration could activate FXR signaling and alleviate B. fragilis-induced cholestasis in vivo. We observed that DCA and GCDCA could not mitigate B. fragilis-induced cholestasis and may even aggravate B. fragilis-induced cholestasis. Previous studies have reported that DCA is a toxic bile acid that could induce cholestasis, and GCDCA was previously reported to accumulate in human cholestasis and was able to induce the apoptosis of cholangiocytes and hepatocytes49–53. In the present study, we found that GDCA activated FXR signaling, inhibited bile acid synthesis, and mitigated B. fragilis-induced cholestasis. Since GCDCA was not the target bile acid altered by B. fragilis colonization and was able to further aggravate B. fragilis-induced cholestasis, we cannot simply say that conjugated bile acids are the key to inhibiting cholestasis induced by B. fragilis colonization.
Cholestasis is mainly characterized by bile flow impairment and excessive bile acid accumulation43,54. Bile acids are synthesized in the liver and secreted into the intestine through the biliary tract, and are mediated by negative feedback regulation through FXR42. CYP7A1, CYP8B1 and CYP27A1 are liver-specific microsomal cytochrome P450 enzymes that act as the rate-limiting step in bile acid synthesis42. Additionally, bile acid homeostasis mainly relies on transporters such as the bile salt export pump (BSEP) and multidrug resistance-associated protein 2 (MRP2) to excrete bile acids into the bile ducts55,56. In the liver, FXR activation can induce small heterodimer partner (SHP) expression, which then binds to liver receptor homologue-1 (LRH-1) and inhibits CYP7A1, CYP8B1 and CYP27A1 expression26,57. Moreover, FXR activation in the ileum induces FGF15 expression in mice (FGF19 in humans) to bind to FGF receptor 4 (FGFR4) and triggers the JNK$\frac{1}{2}$ and ERK$\frac{1}{2}$ signaling cascade, which then inhibits CYP7A1 expression58. Various studies have shown that CYP7A1, CYP8B1, CYP27A1, BSEP and MRP2 are direct targets for FXR59,60. Our data related to elevated fasting C4 indicated excessive bile acid synthesis in ICP, likely secondary to suppressed FXR signaling by ICP faecal transplantation or B. fragilis colonization. We found that hepatic levels of T-αMCA and T-βMCA were significantly increased in B. fragilis-colonized mice, which might account for the reduced hepatic FXR signaling. Intestinal bile acid composition alterations might affect intestinal FXR signaling by impairing bile acid uptake or by changing the proportion of agonistic and antagonistic bile acids. Bile acid binding to ASBT is the key step in bile acid uptake from the lumen to ileal enterocytes, and ASBT preferentially binds to conjugated bile acids25. Our data revealed significantly reduced conjugated bile acids after B. fragilis colonization, which might have reduced intestinal FXR induction secondary to the reduced bile acid uptake. Previous studies showed that activation of FXR by GW4064 protected animals from cholestasis61,62. Herein, we provide evidence that administration of either GDCA or GW4064 to activate ileum and hepatic FXR signaling attenuated B. fragilis-induced expression of CYP7A1, CYP8B1 and CYP27A1 and restored expression of BSEP and MRP2, suggesting that modulation of FXR signaling might be an effective strategy for drug development in ICP.
Currently, the mainstay of treatment for ICP is ursodeoxycholic acid (UDCA), which is produced by the gut microbiota63. UDCA is a polar bile acid without a direct FXR agonist effect, and it functions by decreasing the toxicity and hydrophobicity of the bile pool64,65. Although some previous studies have revealed that UDCA was effective in ameliorating pruritus and improving liver function in ICP patients4,7, the largest randomized controlled clinical trial on this agent, the PITCHES trial, found that there was no evidence that UDCA could reduce adverse perinatal outcomes64. A Cochrane database meta-analysis examining treatment with UDCA versus a placebo in ICP found that UDCA administered to patients with ICP probably resulted in a slight reduction in pruritus, with no significant differences in total bile acid levels6,7. Because UDCA is not a direct FXR agonist, potent agonists of FXR might be more effective than UDCA in lowering bile acid levels and improving outcomes in ICP patients. Because of the key role of FXR in bile acid synthesis, transport and excretion, the synthetic FXR agonist obeticholic acid (OCA) has been shown to promote bile acid efflux and reduce bile acid synthesis and has shown promising effects in the treatment of cholestatic conditions such as primary biliary cholangitis66,67. A previous study also showed that OCA administration could improve fetal hypercholanaemia in an ICP mouse model68. Therefore, bile acid modulation currently remains the mainstay of treatment for ICP. Although UDCA treatment of ICP has been shown to reduce maternal bile acid levels and improve liver function, it is not effective in all patients64, and a recent trial revealed no benefit for adverse perinatal outcomes64, leading to controversy about its use in ICP. Because of the risk of serious liver injury associated with obeticholic acid (Ocaliva), its use is restricted for some patients by the FDA. Thus, it is urgent to develop more effective and safer treatment strategies for ICP. Microbiota-based therapeutics have the potential to become a major therapeutic modality, encompassing those therapies targeting specific pathogens/pathobionts as well as those aiming to strategically restructure a given microbial community in a specific way. There are many microbial therapeutic strategies, including probiotics, faecal microbiota transplantation, those involving administration of metabolites derived from microbial sources or those based on non-bacterial micro-organisms such as bacteriophages69,70. Faecal microbiota transplantation has emerged as a treatment for multiple recurrent Clostridioides difficile infections (rCDIs) that are nonresponsive to standard therapy approaches recommended in multiple professional society guidelines71–73. Notably, the present study demonstrated that B. fragilis could suppress FXR signaling and promote ICP development, implying that manipulating the abundances of specific bacteria might be potentially effective. Therefore, although microbial therapy is not yet used in clinical practice for ICP management, it may have benefits and prospects in the future.
In conclusion, we demonstrated that the abundance of B. fragilis was substantially increased in individuals with ICP. B. fragilis colonization could suppress FXR signaling by mediating bile acid metabolism and result in excessive bile acid synthesis and disrupted bile acid excretion, highlighting the potential role of gut microbiota-bile acid-FXR axis in ICP development (Fig. 6i).
## Human samples
The study was approved by the Ethics Committee of Xinqiao Hospital, Army Medical University (Approved No. 2020-146-01). Written informed consents for participating this study were obtained from all participants. The authors affirm that human research participants provided written informed consent for publication of the potentially identifiable medical data included in this article. Participants didn’t receive cash remuneration. ICP was diagnosed according to the Guidelines for diagnosis and treatment of intrahepatic cholestasis of pregnancy from China with the following criteria: unexplainable pruritus; elevated serum bile acids (≥10 μmol/L); no identifiable cause for liver dysfunction; resolution of symptoms and laboratory values postpartum. Exclusion criteria were as follows: preeclampsia, low platelets (HELLP) syndrome, acute fatty liver of pregnancy, active viral hepatitis and primary biliary cirrhosis; patients receiving any antibiotic or probiotics treatment within 1 months; patients with other pregnant complications such as pregnancy diabetes and hypertensive disorders. All pregnant women with ICP were first-visit patients and did not receive any treatment. 50 individuals with ICP and 41 age, BMI and offspring gender matched healthy pregnant women were recruited from Chongqing and Guangdong province of China. There were 30 mild (TBA range 10–39.9 μmol/L) and 20 severe (TBA ≥ 40 μmol/L) ICP patients included. All the characteristics were summarized in Supplementary Tables 1 and 2.
Age, height, body weight, gestation week, birth weight and Apgar score were recorded, and the body mass index (BMI) was calculated. The gestational weeks were strictly matched within 1 week to reduce the impact of gestational week on gut microbiota. Fecal and blood samples were collected after fasting at least 8 h. Fecal samples were stored at −80 °C immediately until further processed. Biochemical parameters were detected by autoanalyzer.
## Animal study
All animal protocols were approved by the Animal Care and Use Committee of the Army Medical University and adhered to the Animal Ethics Statement (Approved No. AMUWEC2020197). C57BL/6 mice were from Vital River Laboratory (Beijing, China) and housed in a standard specific-pathogen-free environment.
For the fecal microbiota transplantation (FMT) study, the donor stools were randomly collected from six ICP patients and six healthy pregnant controls in the third trimester with matched gestational weeks (within 1 week), matched BMI and offspring gender (Supplementary Table 3). The donor stool samples from six ICP or healthy pregnant controls were mixed with saline solution and centrifuged to collect the supernatant. Then, the aliquot of the mixture was gavaged to the recipient mice, respectively. Before stool transplantation, female mice aged 6–8 weeks were administered antibiotics cocktail (vancomycin, 100 mg/kg; neomycin sulfate, metronidazole and ampicillin, 200 mg/kg) intragastrically once daily for five days. Then, 200 μl of the stool suspension from ICP or healthy controls was gavaged to mice twice a week. To avoid contamination from male mice, all the male mice underwent antibiotics treatment and stool transplantation as same as the female mice. Three weeks later, female mice were housed with male mice at a 2:1 ratio, and pregnancy was confirmed by the presence of vaginal spermatozoa. Various tissues and samples were collected at the 18th day of pregnancy (E18d).
Bacteroides fragilis (B. fragilis) was purchased from American Type Culture Collection (Cat# ATCC25285, ATCC) and cultured in modified chopped meat medium (ATCC-Medium 1490) in a 37 °C anaerobic incubator. Before B. fragilis transplantation, we performed a dose-response analysis to determine the optimal dose. We found that the effect could be dose-dependent, and the dose of 2 × 108 colony-forming units (cfu) could exert the maximum response (Supplementary Fig. 17), and we chose this dose for the following experiments. The male and female recipient mice were administered antibiotics cocktail once daily for five days and gavaged with B. fragilis at a dose of 2 × 108 cfu per 200 μl sterile PBS twice a week until end of the pregnancy. Gavage of the same dose of heat-killed B. fragilis was used as a negative control. Additionally, the recipient mice were gavaged with B. fragilis with or without caffeic acid phenethyl ester (CAPE) (75 mg/kg/d, Cat# HY-N0274, MCE) for 6 weeks to explore the role of BSH activity of B. fragilis. To evaluate the S1PR2 activation in B. fragilis colonized mice, the recipient mice were gavaged with B. fragilis with or without JTE-013 (Cat# HY-100675, MCE) (S1PR2 antagonist, 40 mg/kg/d). To evaluate the effect of GDCA in B. fragilis-induced ICP, the recipient mice were gavaged with B. fragilis with or without GDCA (30 mg/kg/d; once a day for 3 weeks) (Cat# IG2290, Solarbio). Furthermore, mice gavaging B. fragilis were administered intragastrically with GW4064 (10 mg/kg/d, Cat# HY-50108, MCE) to explore the role of FXR. Experimental intrahepatic cholestasis of pregnancy is induced by subcutaneous injections of 17α-ethynylestradiol (EE2) (5 mg/kg, Cat# E4876, Sigma-Aldrich) once daily for 6 days from the 12th day of pregnancy (E12d). The control group was injected daily with corn oil. B6.129 × 1(FVB)-Nr1h4tm1Gonz/J (FXR−/−) mice were obtained from Jackson Laboratory (Cat# 00724). Six- to eight-week-old female FXR−/− mice and wild type (WT) mice were gavaged with B. fragilis (2 × 108 cfu/200 μl) or same dose of heat-killed B. fragilis. To avoid contamination from male mice, all the male mice underwent antibiotics treatment and B. fragilis transplantation as same as the female mice.
Before the mice were killed, the animals from each group were measured weight. The serum was collected after 4 h fasting. Livers, placenta, fetal were weighed and the livers and placenta tissues were further processed for hematoxylin and eosin (HE) staining or immumohistochemical staining. Feces and intestines were stored at −80 °C for further use.
## DNA extraction and 16S rRNA sequencing
Fecal bacterial DNA was extracted using TIANamp Stool DNA Kit (Cat# DP328, TIANGEN Biotech Co. Ltd., China) according to the manufacture’s instruction. The quantity and quality were measured using a Nanodrop (Thermo Scientific, USA). The V3-V4 hypervariable regions of the 16S rRNA were amplified. PCR amplicons were purified and sequenced on the Illumina HiSeq platform (Illumina, San Diego, USA) by Magigene (Magigene Guangzhou, China). The Quantitative Insights into Microbial Ecology 2 (QIIME2, version 2019. 7) platform was used to process the sequencing data. The V3-V4 primers of pair-end fastq format sequence files were trimmed by using cutadapt 3.1 with Python 3.6.9 and imported into QIIME2. The beta diversity was calculated by “qiime diversity beta” command from the rarefied feature-table. The PCoA results were calculated by “qiime diversity pcoa” command and visualized by “qiime emperor plot” command.
## Metagenomic sequencing and analysis
Total microbial DNA were extracted using the QIAamp PowerFecal Pro DNA Kit (Cat#51804, QIAGEN). DNA concentration was measured. 1 μg DNA per sample was used as input. Sequencing libraries were generated using NEBNext® Ultra™ DNA Library Prep Kit (Cat# E7370L, NEB). DNA samples were fragmented by sonication to 350 bp, which were end-polished, A-tailed, and ligated. PCR products were purified. The clustering of the index-coded samples was performed on a cBot Cluster Generation System, and then sequenced on an Illumina Novaseq 6000 platform by Novogene (Novogene Tianjin, China).
QC process including trimming of low-quality bases, masking of human DNA contamination, and removal of duplicated reads were performed by using kneaddata (version v0.6.1). Human DNA contamination was identified by aligning all raw reads to the human reference genome (hg19) using bowtie2 (version 2.3.5.1). Taxonomic annotation of metagenome and the abundance quantification were performed by MetaPhlAn (version 2.0). Relative abundance of each clade was calculated at six levels (L2: phylum, L3: class, L4: order, L5: family, L6: genus, L7: species). Functional annotations were performed by using the data files from the HMP Unified Metabolic Analysis Network 3.0 (HUMAnN 3.0)74. The clean paired-end sequencing data were merged into a single fastq file. The HUMAnN 3.0 toolkit was run by using the “humann–input myseq*.fq–output humann3/–threads 32–memory-use maximum -r -v” command, which calls Bowtie275 to compare nucleic acid sequence and calls DIAMOND76 to compare protein sequences to complete gene and protein function annotation to obtain KEGG pathway annotation. Differences in bacterial abundance and functional pathway were analyzed using MaAslin277. Richness indices were calculated using the R Community Ecology Package vegan. Weighted Unifrac distance was calculated using Metaphlan3 R script “Unifrac_distance.r” and root-tree file “mpa_v30_CHOCOPhlAn_201901_species_tree.nwk”. The PCoA results were calculated and visualized using R build-in functions and the plot3D R package. The ANOSIM test was used to calculate the significance of dissimilarity using the R Community Ecology Package vegan. Pearson correlation and P values were evaluated using the rcorr function in the Hmisc R package.
## Cell culture
Caco-2 (Cat# HTB-37, ATCC) and HEK293 cells (Cat#CRL-3216, ATCC) were purchased from ATCC (Manassas, VA) and cultured in DMEM with $10\%$ FBS. Cells were then exposed to different concentrations of specific bile acids for 24 h. HEK293 cells were transfected with different vectors and used for luciferase reporter assay.
## Bile acid analysis
50 μL of serum was mixed with 150 μL of methanol. The mixture was vortexed for 2 min and centrifuged at 20,000 × g at 4 °C for 10 min. 160 μL of supernatant was vacuum-dried. The residue was redissolved with acetonitrile and water to a volume of 40 μL. Supernatant was used for UPLC-MS/MS analysis. 200 μL aliquot of methanol/water (1:1) was added to 10 mg of fecal samples. Samples were homogenized and centrifuged at 13,000 × g for 15 min. The supernatant was transferred into a tube and sample residue was extracted by methanol/acetonitrile (2:8). The extraction mixture was vortexed and centrifuged at 13,000 × g for 15 min. The supernatant was then used for further analysis.
Bile acid analysis was performed on the UPLC-MS/MS (Waters Corp., USA). The elution solvents were water + $0.01\%$ formic acid (A) and acetonitrile/methanol (19:1) + $0.01\%$ formic acid (B). The elution gradient at a flow rate of 450 μL/min was as follows: 0–2 min ($20\%$ B), 2–3 min (20–$25\%$ B), 3–6 min ($25\%$ B), 6–8 min (25–$35\%$ B), 8–11.5 min ($35\%$ B), 11.5–18 min (35–$99\%$ B), 18–19 min ($99\%$ B), and 19–20 min (99–$20\%$ B). The peak annotation and quantitation was performed by TargetLynx application manager. Multi Quant 2.1 software were used for bile acids data collection.
## BSH activity analysis
B. fragilis proteins were prepared using sonication. The incubation was carried out in 3 mM sodium acetate buffer containing 0.1 mg/ml protein and 0.1 mM d4-GDCA (Sigma-Aldrich, Cat#330271 W, 100 μg) with or without CAPE (MCE, Cat# HY-N0274, 10 mg). The mixtures were incubated at 37 °C and the reactions were stopped in dry ice. 100 μL of methanol was added and the mixtures were vortexed and centrifuged for 20 min. The supernatants were used for d4-GDCA quantification by UPLC-TQMS (Waters, Milford, MA, USA). The BSH activities were predicted by hydrolysis of d4-GDCA. For BSH activity assay in different conjugated bile acids, the mixture contained 2 mL 0.1 M phosphate buffer, 2 mg B. fragilis cells, and 500 μg of conjugated bile acids (GDCA, Cat#IG2290; TDCA, Cat#YS167267; GCDCA, Cat#YS175025; TCDCA, Cat#YZ110846; TCA, Cat#T8510; Solarbio) (GCA, Cat#475-31-0, Aladdin). The mixtures were incubated at 37 °C for up to 120 min. The reaction was terminated by adding 200 μL $15\%$ (w/v) trichloroacetic acid. The mixtures were centrifuged at 15,000 × g for 10 min to obtain the reaction samples. The deconjugated bile acids were detected using the UPLC-MS/MS (Waters Corp., USA).
## Molecular docking
The crystal structures of the complex of farnesoid X receptor (FXR) and GW 4064 were downloaded from RCSB Protein Data Bank (PDB ID: 3dct, https://www.rcsb.org/) and prepared by SYBYL-X 2.0. The docking analysis was performed using the Surflex-Dock GeomX (SFXC) in SYBYL-X 2.0. The binding interaction was generated using PyMOL and ligplot.
## TR-FRET FXR coactivator assay
Direct FXR activity was evaluated using the LanthaScreen™ TR-FRET Farnesoid X Receptor Coactivator Assay kit (Cat# PV4833, ThermoFisher Scientific). Briefly, prepare a 12-point 100× dilution series of GDCA (Cat#IG2290, Solarbio), CDCA (Cat#IC0300, Solarbio), GW4064 (Cat# HY-50108, MCE) and GUDCA (Cat#IG0840, Solarbio) in a 96-well plate by serial dilution, respectively. Dilute each 100× serial dilution to 2× using Complete Coregulator buffer G. Then, the 2× serial dilutions were mixed with FXR-LBD-glutathione S-transferase fusion protein, fluorecein-SRC2-2 coactivator peptide and Lantha-screen Tb anti-GST antibody (Cat# PV4833, ThermoFisher Scientific, 1:1500) in the 384-well assay plate. Mix the 384-well plate and the TR-FRET signal was evaluated in a Multi-Mode Microplate Reader (Varioskan Flash, Thermo Fisher). Calculate the TR-FRET ratio by dividing the emission signal at 520 nm by the emission signal at 495 nm. Generate a binding curve by plotting the emission ratio vs. [ligand].
## Luciferase reporter assay
pGL3-Basic-SHP firefly luciferase reporter vector, human FXR expression vector, and human ASBT expression vector were constructed by Sangon Biotech (Shanghai, China). HEK293 cells were cultured and co-transfected with human FXR expression vector, human ASBT expression vector, pGL3-Basic-SHP firefly luciferase reporter vector and the Renilla luciferase control vector (Promega, Madison, WI) using LipofectamineTM 3000 transfection reagent (Cat# L3000015, ThermoFisher Scientific). Luciferase assays were performed by Dual-Luciferase® Reporter Assay System (Cat# E1910, Promega), and Firefly and Renilla luciferase activities were measured by Microplate Reader (Varioskan Flash, Thermo Fisher).
## Measurement of FGF15, FGF19 and C4 in serum
Serum FGF15 levels were tested by the ELISA Kit (Cat# LS-F35359, LifeSpan BioSciences). Briefly, add 100 μL of standards, blank or samples to each well which was pre-coated with FGF15 antibody, and incubate for 90 min at 37 °C. Remove liquid and tap against clean absorbent paper for three times. Add 100 μL of biotinylated detection antibody solution to each well, and gently agitate to ensure thorough mixing. Total mixture was incubated at 37 °C, followed by wash buffer per well for three times. 100 μL of HRP-Streptavidin conjugate working solution was added to each well, incubate for 30 min at 37 °C and washed five times. 90 μL of TMB substrate was added to each well and incubated for 30 min at 37 °C. 50 μL of stop solution was added to each well. Determine the optical density (OD value) of each well using a microplate reader set to 450 nm. The standard stock solution was prepared to generate a standard curve.
Human serum FGF19 levels were quantified using the Human FGF19 SimpleStep ELISA® Kit (Cat#ab230943, Abcam). Briefly, add 50 µL of all samples or standard dilution series to each well and 50 µL of the Antibody Cocktail was added. Incubate the mixture for 1 h at room temperature. Wash each well with wash buffer. 100 µL of TMB solution was added to each well and incubate for 10 min. 100 µL of stop solution was added to each well. Record the OD at 450 nm, create a standard curve, and determine the concentration using the standard curve.
The level of serum 7α-hydroxy-4-cholesten-3-one (C4) was tested by a mass spectrometry-based (MS-based) method. Briefly, serum sample was mixed with acetonitrile containing 1 μM chlorpropamide. Sample injection and flow rate were set at 2 μL and 0.35 ml/min. The samples were separated using an ACQUITY BEH C18 column (1.7 μm, 100 mm × 2.1 mm) with a linear gradient of $0.1\%$ formic acid (FA) in water (A) and $0.1\%$ FA in acetonitrile (B). The eluate delivered into a 5600 TripleTOF (SCIEX, Framingham, MA).
## Real-time quantitative PCR
The tissues were homogenized and total RNA was isolated by Trizol Reagent (Invitrogen, USA). Concentration was measured using the NanoDrop (Thermo Fisher Scientific, USA). Real-time qPCR was performed using the ABI 7500 real-time PCR system (Applied Biosystems). Specific primers for quantitative PCR used in this study are shown in Supplementary Tables 4 and 5. The relative count of genes was calculated by normalizing to 18 S mRNA.
## Western blot analysis
Liver tissues of the mice with different treatment were harvested and lysed with the lysis buffer, respectively. The lysate was kept on ice, followed by the centrifugation at 13,400 × g at 4 °C for 20 min. The protein concentration was measured by the BCA assay (Cat# P0012, Beyotime). The protein samples were separated by $10\%$ SDS-PAGE, and then electronically transferred onto the PVDF membrane (Cat#IPVH00010, Millipore). After blocking with $5\%$ non-fat milk at room temperature, the membrane was incubated with the primary antibodies for p-ERK (Cat#4370, Cell Signaling, 1:1000), ERK (Cat#4695, Cell Signaling, 1:1000), p-AKT (Cat#4060, Cell Signaling, 1:1000), and AKT (Cat#9272, Cell Signaling, 1:1000) at 4 °C overnight. After washing, the membrane was incubated with the secondary antibody conjugated with horseradish peroxidase (Cat# A0208, Beyotime, 1: 2000) at 37 °C for 1 h. After the exposure development, the protein bands were imaged and analyzed.
## Histological analysis and immunohistochemistry
Liver and placenta tissues were fixed in $4\%$ paraformaldehyde, dehydrated and embedded in paraffin. The tissues were serially sectioned into 4 μm sections, and stained with hematoxylin and eosin. All sections were mounted onto a glass slide and observed under the light microscope and collected by NIS-Elements 3.2 (Nikon, Tokyo, Japan). Sections were examined by a qualified and blinded pathologist to evaluate the pathological changes.
Tissue sections were deparaffinized and rehydrated using ethanol and distilled water, and treated with $3\%$ H2O2. Sections were then rinsed twice and incubated with goat serum to block non-specific antibody binding. Immunohistochemistry was performed using the primary antibodies for Cyp7a1 (Cat#bs-21430R, Bioss, 1:100), Cyp8b1 (Cat#bs-14165R, Bioss, 1:100), Cyp27a1 (Cat#bs-5049R, Bioss, 1:100), MRP2 (Cat#bs-1092R, Bioss, 1:100) and BSEP (Cat#bs-12440R, Bioss, 1:100). After washing, sections were incubated with the secondary antibody (PV-6001; Zhongshan, China) for 30 min. The sections were stained with DAB, dehydrated with ethanol and xylene, and then sealed. The slides were photographed using a digital microscope camera and collected by NIS-Elements 3.2 (Nikon, Tokyo, Japan).
## Serum biochemical analysis
Mice blood samples were collected and centrifuged at 3000 × g for 10 min, the serum was then collected for the analysis. Serum levels of aspartate transaminase (AST) (Cat#C010-2-1), alanine aminotransferase (ALT) (Cat#C009-2-1), alkaline phosphatase (ALP) (Cat#A059-2-2), gamma-glutamyl transpeptidase (GGT) (Cat#C017-2-1) and total bile acid (TBA) (Cat#E003-2-1) were analyzed by commercially available kits (Jiancheng Institute of Biotechnology, Nanjing, China).
## Statistical analysis
Data were shown as the mean ± S.E.M. The statistical data were collected with Microsoft Excel [2013] and GraphPad Prism software (v9, GraphPad Software Inc., San Diego, USA). The sample distribution was determined by the Kolmogorov-Smirnov normality test. A two-tailed Student’s t-test was used to evaluate statistical significance between two groups for normal distribution. For the nonparametric tests, the two-tailed Mann-Whitney test was used to evaluate statistical significance between two groups. One-way analysis of variance (ANOVA) followed by Tukey’s correction was used to evaluate the statistical significance of differences among multiple groups with assumed equal variances. Brown-Forsythe test was used to test homoscedasticity. Welch ANOVA with Games-Howell’s multiple comparisons test was used to evaluate the statistical significance of differences among three or more groups if equal variances were not assumed. For the nonparametric tests among three or more groups, the significance was calculated by Kruskal-Wallis with Dunn’s multiple comparisons test. Correlation analysis was performed using Spearman’s rank test or Pearson correlation analysis. Multiple testing was corrected using the Benjamini-Hochberg method to control the false-discovery rate (FDR). P value or FDR-corrected P value < 0.05 was considered statistically significant. All data shown were representative results from at least three independent experiments.
## Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.
## Supplementary information
Supplementary Information Reporting Summary The online version contains supplementary material available at 10.1038/s41467-023-36981-4.
## Source data
Source data
## Peer review information
Nature Communications thanks Jasmohan Bajaj and the other, anonymous, reviewers for their contribution to the peer review of this work.
## References
1. Ibrahim SH. **Liver diseases in the perinatal period: interactions between mother and infant**. *Hepatology* (2020) **71** 1474-1485. DOI: 10.1002/hep.31109
2. Puljic A. **The risk of infant and fetal death by each additional week of expectant management in intrahepatic cholestasis of pregnancy by gestational age**. *Am. J. Obstet. Gynecol.* (2015) **212** 667.e661-665
3. Monrose E. **Burden of future liver abnormalities in patients with intrahepatic cholestasis of pregnancy**. *Am. J. Gastroenterol.* (2021) **116** 568-575. DOI: 10.14309/ajg.0000000000001132
4. Chappell LC. **Ursodeoxycholic acid versus placebo, and early term delivery versus expectant management, in women with intrahepatic cholestasis of pregnancy: semifactorial randomised clinical trial**. *BMJ* (2012) **344** e3799. DOI: 10.1136/bmj.e3799
5. Vasavan T. **Fetal cardiac dysfunction in intrahepatic cholestasis of pregnancy is associated with elevated serum bile acid concentrations**. *J. Hepatol.* (2021) **74** 1087-1096. DOI: 10.1016/j.jhep.2020.11.038
6. Walker KF, Chappell LC, Hague WM, Middleton P, Thornton JG. **Pharmacological interventions for treating intrahepatic cholestasis of pregnancy**. *Cochrane Database Syst. Rev.* (2020) **7** Cd000493. PMID: 32716060
7. Joutsiniemi T, Timonen S, Leino R, Palo P, Ekblad U. **Ursodeoxycholic acid in the treatment of intrahepatic cholestasis of pregnancy: a randomized controlled trial**. *Arch. Gynecol. Obstet.* (2014) **289** 541-547. DOI: 10.1007/s00404-013-2995-5
8. Tilg H, Zmora N, Adolph TE, Elinav E. **The intestinal microbiota fuelling metabolic inflammation**. *Nat. Rev. Immunol.* (2020) **20** 40-54. DOI: 10.1038/s41577-019-0198-4
9. Gong B. **A genetic variant in IL-6 lowering its expression is protective for critical patients with COVID-19**. *Signal Transduct. Target. Ther.* (2022) **7** 112. DOI: 10.1038/s41392-022-00923-1
10. Sonnenburg ED. **Diet-induced extinctions in the gut microbiota compound over generations**. *Nature* (2016) **529** 212-215. DOI: 10.1038/nature16504
11. Nuriel-Ohayon M. **Progesterone increases bifidobacterium relative abundance during late pregnancy**. *Cell Rep.* (2019) **27** 730-736.e733. DOI: 10.1016/j.celrep.2019.03.075
12. Gálvez EJC, Iljazovic A, Gronow A, Flavell R, Strowig T. **Shaping of intestinal microbiota in Nlrp6- and Rag2-deficient mice depends on community structure**. *Cell Rep.* (2017) **21** 3914-3926. DOI: 10.1016/j.celrep.2017.12.027
13. Gentile CL, Weir TL. **The gut microbiota at the intersection of diet and human health**. *Science* (2018) **362** 776-780. DOI: 10.1126/science.aau5812
14. Rooks MG, Garrett WS. **Gut microbiota, metabolites and host immunity**. *Nat. Rev. Immunol.* (2016) **16** 341-352. DOI: 10.1038/nri.2016.42
15. Liang L. **Metabolic dynamics and prediction of gestational age and time to delivery in pregnant women**. *Cell* (2020) **181** 1680-1692.e1615. DOI: 10.1016/j.cell.2020.05.002
16. Deshmukh H, Way SS. **Immunological basis for recurrent fetal loss and pregnancy complications**. *Annu. Rev. Pathol.* (2019) **14** 185-210. DOI: 10.1146/annurev-pathmechdis-012418-012743
17. Koren O. **Host remodeling of the gut microbiome and metabolic changes during pregnancy**. *Cell* (2012) **150** 470-480. DOI: 10.1016/j.cell.2012.07.008
18. Nyangahu DD. **Disruption of maternal gut microbiota during gestation alters offspring microbiota and immunity**. *Microbiome* (2018) **6** 124. DOI: 10.1186/s40168-018-0511-7
19. Zakaria ZZ. **Physiological changes and interactions between microbiome and the host during pregnancy**. *Front. Cell. Infect. Microbiol.* (2022) **12** 824925. DOI: 10.3389/fcimb.2022.824925
20. Kim S. **Maternal gut bacteria promote neurodevelopmental abnormalities in mouse offspring**. *Nature* (2017) **549** 528-532. DOI: 10.1038/nature23910
21. Chen X. **Gut dysbiosis induces the development of pre-eclampsia through bacterial translocation**. *Gut* (2020) **69** 513-522. DOI: 10.1136/gutjnl-2019-319101
22. Wang J. **Dysbiosis of maternal and neonatal microbiota associated with gestational diabetes mellitus**. *Gut* (2018) **67** 1614-1625. DOI: 10.1136/gutjnl-2018-315988
23. Crusell MKW. **Gestational diabetes is associated with change in the gut microbiota composition in third trimester of pregnancy and postpartum**. *Microbiome* (2018) **6** 89. DOI: 10.1186/s40168-018-0472-x
24. Gomez-Arango LF. **Increased systolic and diastolic blood pressure is associated with altered gut microbiota composition and butyrate production in early pregnancy**. *Hypertension* (2016) **68** 974-981. DOI: 10.1161/HYPERTENSIONAHA.116.07910
25. Ovadia C. **Enhanced microbial bile acid deconjugation and impaired ileal uptake in pregnancy repress intestinal regulation of bile acid synthesis**. *Hepatology* (2019) **70** 276-293. DOI: 10.1002/hep.30661
26. Wahlström A, Sayin SI, Marschall HU, Bäckhed F. **Intestinal crosstalk between bile acids and microbiota and its impact on host metabolism**. *Cell Metabolism* (2016) **24** 41-50. DOI: 10.1016/j.cmet.2016.05.005
27. Li R. **Characterization of gut microbiota associated with clinical parameters in intrahepatic cholestasis of pregnancy**. *BMC Gastroenterol.* (2020) **20** 395. DOI: 10.1186/s12876-020-01510-w
28. Zhan Q. **Alterations of the human gut microbiota in intrahepatic cholestasis of pregnancy**. *Front. Cell. Infect. Microbiol.* (2021) **11** 635680. DOI: 10.3389/fcimb.2021.635680
29. Xu G. **Glycocholic acid and glycodeoxycholic acid but not glycoursocholic acid inhibit bile acid synthesis in the rabbit**. *Gastroenterology* (1992) **102** 1717-1723. DOI: 10.1016/0016-5085(92)91735-M
30. Xu G. **Different feedback regulation of hepatic cholesterol and bile acid synthesis by glycodeoxycholic acid in rabbits**. *Gastroenterology* (1993) **105** 1192-1199. DOI: 10.1016/0016-5085(93)90967-H
31. Org E. **Sex differences and hormonal effects on gut microbiota composition in mice**. *Gut Microbes* (2016) **7** 313-322. DOI: 10.1080/19490976.2016.1203502
32. Mayneris-Perxachs J. **Gut microbiota steroid sexual dimorphism and its impact on gonadal steroids: influences of obesity and menopausal status**. *Microbiome* (2020) **8** 136. DOI: 10.1186/s40168-020-00913-x
33. Yamazaki E, Sugimoto K, Niwano K, Okada J. **A simple bile-disk method for the identification of Bacteroides fragilis**. *Microbiol. Immunol.* (1982) **26** 759-765. DOI: 10.1111/j.1348-0421.1982.tb00222.x
34. Pumbwe L. **Bile salts enhance bacterial co-aggregation, bacterial-intestinal epithelial cell adhesion, biofilm formation and antimicrobial resistance of Bacteroides fragilis**. *Microb. Pathog.* (2007) **43** 78-87. DOI: 10.1016/j.micpath.2007.04.002
35. Postler TS, Ghosh S. **Understanding the holobiont: how microbial metabolites affect human health and shape the immune system**. *Cell Metabolism* (2017) **26** 110-130. DOI: 10.1016/j.cmet.2017.05.008
36. Garrett WS. **Immune recognition of microbial metabolites**. *Nat. Rev. Immunol.* (2020) **20** 91-92. DOI: 10.1038/s41577-019-0252-2
37. Dai X. **Microbial metabolites: critical regulators in NAFLD**. *Front. Microbiol.* (2020) **11** 567654. DOI: 10.3389/fmicb.2020.567654
38. Stellwag EJ, Hylemon PB. **Purification and characterization of bile salt hydrolase from Bacteroides fragilis subsp. fragilis**. *Biochim. Biophys. Acta* (1976) **452** 165-176. DOI: 10.1016/0005-2744(76)90068-1
39. Gu XC. **Cloning and analysis of bile salt hydrolase genes from Lactobacillus plantarum CGMCC No. 8198**. *Biotechnol. Lett.* (2014) **36** 975-983. DOI: 10.1007/s10529-013-1434-9
40. Masuda N. **Deconjugation of bile salts by Bacteroids and Clostridium**. *Microbiol. Immunol.* (1981) **25** 1-11. DOI: 10.1111/j.1348-0421.1981.tb00001.x
41. Jia W, Xie G, Jia W. **Bile acid-microbiota crosstalk in gastrointestinal inflammation and carcinogenesis**. *Nat. Rev. Gastroenterol. Hepatol.* (2018) **15** 111-128. DOI: 10.1038/nrgastro.2017.119
42. Fiorucci S, Distrutti E, Carino A, Zampella A, Biagioli M. **Bile acids and their receptors in metabolic disorders**. *Prog. Lipid Res.* (2021) **82** 101094. DOI: 10.1016/j.plipres.2021.101094
43. Trauner M, Fuchs CD, Halilbasic E, Paumgartner G. **New therapeutic concepts in bile acid transport and signaling for management of cholestasis**. *Hepatology* (2017) **65** 1393-1404. DOI: 10.1002/hep.28991
44. Sayin SI. **Gut microbiota regulates bile acid metabolism by reducing the levels of tauro-beta-muricholic acid, a naturally occurring FXR antagonist**. *Cell Metabolism* (2013) **17** 225-235. DOI: 10.1016/j.cmet.2013.01.003
45. Bennett MJ, McKnight SL, Coleman JP. **Cloning and characterization of the NAD-dependent 7alpha-Hydroxysteroid dehydrogenase from Bacteroides fragilis**. *Curr. Microbiol.* (2003) **47** 475-484. DOI: 10.1007/s00284-003-4079-4
46. Fukiya S. **Conversion of cholic acid and chenodeoxycholic acid into their 7-oxo derivatives by Bacteroides intestinalis AM-1 isolated from human feces**. *FEMS Microbiol. Lett.* (2009) **293** 263-270. DOI: 10.1111/j.1574-6968.2009.01531.x
47. Chiang JY, Kimmel R, Weinberger C, Stroup D. **Farnesoid X receptor responds to bile acids and represses cholesterol 7alpha-hydroxylase gene (CYP7A1) transcription**. *J. Biol. Chem.* (2000) **275** 10918-10924. DOI: 10.1074/jbc.275.15.10918
48. Wang H, Chen J, Hollister K, Sowers LC, Forman BM. **Endogenous bile acids are ligands for the nuclear receptor FXR/BAR**. *Mol. Cell* (1999) **3** 543-553. DOI: 10.1016/S1097-2765(00)80348-2
49. Farooqui N, Elhence A, Shalimar A. **Current understanding of bile acids in chronic liver disease**. *J. Clin. Exp. Hepatol.* (2022) **12** 155-173. DOI: 10.1016/j.jceh.2021.08.017
50. Tanaka M. **Ursodeoxycholic acid prevents hepatic cytochrome P450 isozyme reduction in rats with deoxycholic acid-induced liver injury**. *J. Hepatol.* (1999) **31** 263-270. DOI: 10.1016/S0168-8278(99)80223-2
51. Sousa T. **Deoxycholic acid modulates cell death signaling through changes in mitochondrial membrane properties**. *J. Lipid Res.* (2015) **56** 2158-2171. DOI: 10.1194/jlr.M062653
52. González R. **Nitric oxide mimics transcriptional and post-translational regulation during α-tocopherol cytoprotection against glycochenodeoxycholate-induced cell death in hepatocytes**. *J. Hepatol.* (2011) **55** 133-144. DOI: 10.1016/j.jhep.2010.10.022
53. Rust C. **Bile acid-induced apoptosis in hepatocytes is caspase-6-dependent**. *J. Biol. Chem.* (2009) **284** 2908-2916. DOI: 10.1074/jbc.M804585200
54. Ibrahim SH, Kamath BM, Loomes KM, Karpen SJ. **Cholestatic liver diseases of genetic etiology: advances and controversies**. *Hepatology* (2022) **75** 1627-1646. DOI: 10.1002/hep.32437
55. van Wessel DBE. **Genotype correlates with the natural history of severe bile salt export pump deficiency**. *J. Hepatol.* (2020) **73** 84-93. DOI: 10.1016/j.jhep.2020.02.007
56. Ghosh S. **NF-κB regulation of LRH-1 and ABCG5/8 potentiates phytosterol role in the pathogenesis of parenteral nutrition-associated cholestasis**. *Hepatology* (2021) **74** 3284-3300. DOI: 10.1002/hep.32071
57. Clifford BL. **FXR activation protects against NAFLD via bile-acid-dependent reductions in lipid absorption**. *Cell Metabolism* (2021) **33** 1671-1684.e1674. DOI: 10.1016/j.cmet.2021.06.012
58. Hartmann P. **Modulation of the intestinal bile acid/farnesoid X receptor/fibroblast growth factor 15 axis improves alcoholic liver disease in mice**. *Hepatology* (2018) **67** 2150-2166. DOI: 10.1002/hep.29676
59. Chávez-Talavera O, Tailleux A, Lefebvre P, Staels B. **Bile acid control of metabolism and inflammation in obesity, type 2 diabetes, dyslipidemia, and nonalcoholic fatty liver disease**. *Gastroenterology* (2017) **152** 1679-1694.e1673. DOI: 10.1053/j.gastro.2017.01.055
60. Gonzalez FJ, Jiang C, Patterson AD. **An intestinal microbiota-farnesoid X receptor axis modulates metabolic disease**. *Gastroenterology* (2016) **151** 845-859. DOI: 10.1053/j.gastro.2016.08.057
61. El Kasmi KC. **Pharmacologic activation of hepatic farnesoid X receptor prevents parenteral nutrition-associated cholestasis in mice**. *Hepatology* (2022) **75** 252-265. DOI: 10.1002/hep.32101
62. Rizzolo D. **Bile acid homeostasis in female mice deficient in Cyp7a1 and Cyp27a1**. *Acta Pharm. Sin. B* (2021) **11** 3847-3856. DOI: 10.1016/j.apsb.2021.05.023
63. Ridlon JM, Bajaj JS. **The human gut sterolbiome: bile acid-microbiome endocrine aspects and therapeutics**. *Acta Pharm. Sin. B* (2015) **5** 99-105. DOI: 10.1016/j.apsb.2015.01.006
64. Chappell LC. **Ursodeoxycholic acid versus placebo in women with intrahepatic cholestasis of pregnancy (PITCHES): a randomised controlled trial**. *Lancet* (2019) **394** 849-860. DOI: 10.1016/S0140-6736(19)31270-X
65. Montano-Loza AJ, Corpechot C. **Definition and management of patients with primary biliary cholangitis and an incomplete response to therapy**. *Clin. Gastroenterol. Hepatol.* (2021) **19** 2241-2251.e2241. DOI: 10.1016/j.cgh.2020.06.062
66. **The diagnosis and management of patients with primary biliary cholangitis**. *J. Hepatol.* (2017) **67** 145-172. DOI: 10.1016/j.jhep.2017.03.022
67. Kjærgaard K. **Obeticholic acid improves hepatic bile acid excretion in patients with primary biliary cholangitis**. *J. Hepatol.* (2021) **74** 58-65. DOI: 10.1016/j.jhep.2020.07.028
68. Pataia V. **Obeticholic acid improves fetal bile acid profile in a mouse model of gestational hypercholanemia**. *Am. J. Physiol. Gastrointest. Liver Physiol.* (2020) **319** G197-g211. DOI: 10.1152/ajpgi.00126.2020
69. Li Y, Honda K. **Toward the development of defined microbial therapeutics**. *Int. Immunol.* (2021) **33** 761-766. DOI: 10.1093/intimm/dxab038
70. Sorbara MT, Pamer EG. **Microbiome-based therapeutics**. *Nat. Rev. Microbiol.* (2022) **20** 365-380. DOI: 10.1038/s41579-021-00667-9
71. McDonald LC. **Clinical Practice Guidelines for Clostridium difficile Infection in Adults and Children: 2017 Update by the Infectious Diseases Society of America (IDSA) and Society for Healthcare Epidemiology of America (SHEA)**. *Clin. Infect. Dis.* (2018) **66** 987-994. DOI: 10.1093/cid/ciy149
72. Kelly CR. **ACG Clinical Guidelines: Prevention, Diagnosis, and Treatment of Clostridioides difficile Infections**. *Am. J. Gastroenterol.* (2021) **116** 1124-1147. DOI: 10.14309/ajg.0000000000001278
73. Vaughn BP. **Effectiveness and safety of colonic and capsule fecal microbiota transplantation for recurrent Clostridioides difficile infection**. *Clin. Gastroenterol. Hepatol.* (2022) **S1542-3565** 00880-00881
74. Beghini F. **Integrating taxonomic, functional, and strain-level profiling of diverse microbial communities with bioBakery 3**. *eLife* (2021) **10** e65088. DOI: 10.7554/eLife.65088
75. Langmead B, Salzberg SL. **Fast gapped-read alignment with Bowtie 2**. *Nat. Methods* (2012) **9** 357-359. DOI: 10.1038/nmeth.1923
76. Buchfink B, Xie C, Huson DH. **Fast and sensitive protein alignment using DIAMOND**. *Nat. Methods* (2015) **12** 59-60. DOI: 10.1038/nmeth.3176
77. Mallick H. **Multivariable association discovery in population-scale meta-omics studies**. *PLoS Comput. Biol.* (2021) **17** e1009442. DOI: 10.1371/journal.pcbi.1009442
78. Chen T. **The genome sequence archive family: toward explosive data growth and diverse data types**. *Genomics Proteomics Bioinformatics* (2021) **19** 578-583. DOI: 10.1016/j.gpb.2021.08.001
79. **Database Resources of the National Genomics Data Center, China National Center for Bioinformation in 2022**. *Nucleic Acids Res.* (2022) **50** D27-D38. DOI: 10.1093/nar/gkab951
|
---
title: The electronic medical record management systems may improve monitoring and
control of disease activity in patients with ankylosing spondylitis
authors:
- Pei-Ju Huang
- Yi-Hsing Chen
- Wen-Nan Huang
- Yi-Ming Chen
- Kuo-Lung Lai
- Tsu-Yi Hsieh
- Wei-Ting Hung
- Ching-Tsai Lin
- Chih-Wei Tseng
- Kuo-Tung Tang
- Yin-Yi Chou
- Yi-Da Wu
- Chin-Yin Huang
- Chia-Wei Hsieh
- Yen-Ju Chen
- Yu-Wan Liao
- Yen-Tze Liu
- Hsin-Hua Chen
journal: Scientific Reports
year: 2023
pmcid: PMC9998629
doi: 10.1038/s41598-023-30848-w
license: CC BY 4.0
---
# The electronic medical record management systems may improve monitoring and control of disease activity in patients with ankylosing spondylitis
## Abstract
To investigate the impact of an electronic medical record management system (EMRMS) on disease activity and the frequency of outpatient visits among patients with ankylosing spondylitis (AS). We identified 652 patients with AS who were followed up for at least 1 year before and after the first Ankylosing Spondylitis Disease Activity Score (ASDAS) assessment and compared the number of outpatient visits and average visit time within 1 year before and after the initial ASDAS assessment. Finally, we analyzed 201 patients with AS who had complete data and received ≥ 3 continuous ASDAS assessments at an interval of 3 months, and we compared the results of the second and third ASDAS assessments with those of the first. The number of annual outpatient visits increased after ASDAS assessment (4.0 (4.0, 7.0) vs. 4.0 (4.0, 8.0), $p \leq 0.001$), particularly among those with a high initial disease activity. The average visit time was reduced within 1 year after ASDAS assessment (6.4 (8.5, 11.2) vs. 6.3 (8.3, 10.8) min, $$p \leq 0.073$$), especially among patients whose with an inactive disease activity was < 1.3 (ASDAS C-reactive protein (CRP) 6.7 (8.8, 11.1) vs. 6.1 (8.0, 10.3) min, $$p \leq 0.033$$; ASDAS erythrocyte sedimentation rate (ESR) 6.4 (8.7, 11.1) vs. 6.1 (8.1, 10.0) min, $$p \leq 0.027$$). Among patients who received at least three ASDAS assessments, the third ASDAS-CRP tended to be lower than the first (1.5 (0.9, 2.1) vs. 1.4 (0.8, 1.9), $$p \leq 0.058$$). The use of an EMRMS increased the frequency of ambulatory visits among AS patients with high and very high disease activity and reduced the visit time among those with an inactive disease. Continual ASDAS assessments may help control the disease activity of patients with AS.
## Introduction
Ankylosing spondylitis (AS) is a form of axial spondyloarthritis (axSpA), characterized by chronic back pain with articular and periarticular extraspinal features, including synovitis, enthesitis and dactylitis, and nonarticular features, including psoriasis, uveitis, and inflammatory bowel disease (IBD). AS is characterized by sacroiliitis and spinal abnormalities, strongly associated with human leukocyte antigen (HLA) B27 and often accompanied by elevated C-reactive protein (CRP)1.
The mean AS prevalence per 10,000 population (from 36 eligible studies) was $0.238\%$ in Europe, $0.167\%$ in Asia, $0.319\%$ in North America, $0.102\%$ in Latin America, and $0.074\%$ in Africa in 2014 systemic research2. The prevalence in *Taiwan is* $0.337\%$3. The risk of AS is higher in men and in individuals with a family history of AS. The guidelines of the European League Against Rheumatism recommend a range of treatment strategies for the optimal management of axSpA, including nonpharmacological treatment, pharmacological treatment, surgery, and lifestyle modifications4. The primary goal of AS treatment is to attenuate inflammation for relieving pain and stiffness, preventing or delaying complications and spinal deformity, reducing extraspinal and extra-articular manifestations and comorbidities, and maintaining effective psychosocial function. These AS treatment strategies, along with regular monitoring of disease activity, are generally applied by rheumatologists. The Ankylosing Spondylitis Disease Activity Score (ASDAS) is used as a measure of disease activity in patients with AS by using clinical laboratory data and a self-administered questionnaire. The management of axSpA in *Taiwan is* strongly influenced by the National Health Insurance reimbursement system and local health circumstances5. However, rheumatologists in Taiwan are usually operating at maximum capacity, and consequently, they are unable to assess disease activity in patients with AS. Therefore, a new integrated disease surveillance strategy must be developed to monitor disease activity in patients with AS.
With advances in mobile technologies, electric health (eHealth) and smartphone applications (known as apps) have been developed to facilitate the transmission of information related to infectious diseases in numerous low-income countries, such as some nations in Africa6. Furthermore, smart apps have been extensively used for standard clinical evaluations and monitoring diseases and changes in the health status of patients with chronic health conditions7, including asthma8, obesity, diabetes9, hypertension, cardiovascular disease10, and multiple sclerosis11. Such app-based data systems ensure complete and timely data collection12. However, a comprehensive, high-quality, evidence-based data app for disease management in patients with AS is lacking13.
Therefore, the purpose of this study was to investigate the effects of an interactive electronic medical record management system (EMRMS) intervention on the disease activity and frequency of outpatient visits of patients with AS in Taiwan.
## Ethics
The study protocol was approved by the Institutional Review Board (IRB) of Taichung Veterans General Hospital (TCVGH-IRB No.: CE20145B). All experiments were performed in accordance with relevant guildlines and regulations. Informed consent was waived by Institutional Review Board (II), TCVGH, because individual information had been anonymized, de-identify and de-link before analysis.
## Study design
This was a single center, retrospective, cross-sectional study.
## Data source
The EMRMS was established in November, 2016 to assist rheumatologists in conducting ASDAS assessments and comprehensively evaluating clinical outcomes in all patients with AS attending TCVGH. The EMRMS database contains information necessary to determite ASDAS, including CRP, level and erythrocyte sedimentation rate [ESR], patient comorbidities, patient history, and family history. The reliability and validity of the data have been verified14.Patients with AS were consecutively enrolled in the TCVGH-AS cohort after they received a confirmed AS diagnosis from a TCVGH rheumatologist according to the 1984 modified New York criteria10. The CRP and ESR data were automatically uploaded to the TCVGH healthcare information system (HIS) to reduce human error. The baseline information, which was collected by trained nurses during the initial visit, including clinical characteristics, onset age, comorbidities at presentation (hypertension, diabetes mellitus, hyperlipidemia, hepatitis B, hepatitis C, renal insufficiency, gout, coronary artery disease, stroke, periodontal disease, osteoporosis, and tuberculosis history), periarticular extraspinal features (synovitis, enthesitis, and dactylitis) and nonarticular manifestations (psoriasis, uveitis, and IBD), family history of autoimmune disease, and patient history of arthropathy, obtained through standardized questionnaires and worksheets to ensure reproducibility and adherence to good laboratory practice. The rheumatologist in charge then confirmed patients’ clinical characteristics, and nurses assisted the patients with AS to complete the self-assessment questionnaires for disease evaluation. The following measures were used: global assessment of disease activity on a numerical rating scale (NRS) of 0–10, back pain on an NRS of 0–10, duration of morning stiffness on an NRS of 0–10, and peripheral pain or swelling on an NRS of 0–10. Before every 3-month visiting clinic, the patient would first to have blood examination. Blood reports can be uploaded to EMRMS through the HIS system, trained nurses assist patient fills out the questionnaire on EMRMS, the assessment of disease activity completed before visiting the doctor. All laboratory data, including CRP and ESR, have been uploaded to the HIS. The IT at TCVGH help "feed-forward" the patient reported outcomes to HIS, and do the auto-calculation of ASDAS-ESR, ASDAS-CRP using the ESR, CRP data in HIS, then "feed-back" these data to both HIS and EMRMS, showing the data on the summary overview "dashboard" in the EMRMS, which was shown both in HIS and the devices (iPAD handled by a nurse in charge and smartphones of patients with AS).
## Definition of AS
Patients were defined as having AS if they received a diagnosis of AS (International Classification of Diseases, Ninth Revision, Clinical Modification [ICD-9-CM] code 720.0) according to the modified New *York criteria* for AS proposed in 1984 during at least three ambulatory visits and received AS treatment concurrently.
## Definition of AS disease activity
Four disease activity states for the ASDAS score have been defined: The ASDAS assessment measures four disease activity states: inactive, moderate, high, and very high. Disease status was evaluated on the basis of three cutoff values: 1.3, 2.1, and 3.5 units. The 3 cut-offs selected to separate these states were: < 1.3 between "inactive disease" and "moderate disease activity", < 2.1 between "moderate disease activity" and "high disease activity", and > 3.5 between "high disease activity" and "very high disease activity". The following cutoff values were selected to indicate improvement: a change of ≥ 1.1 units denoted a clinically important improvement, and a change of ≥ 2.0 units denoted a major improvement15–18.
## Study participants
A total of 652 eligible patients with AS with complete baseline demographic and assessment data who received an AS diagnosis before April 17, 2020, were enrolled to investigate the changes in their health-related behaviors after EMRMS implementation. Patients with AS with [1] incomplete details related to ASDAS-CRP, ASDAS-ESR, and age of symptom onset in the assessment questionnaires, [2] their first ASDAS assessment after February 01, 2019, [3] or no outpatient visits within 1 year before or after the first assessment were excluded from this study. Of the 652 patients with AS, 201 underwent three consecutive assessments of disease activity using the EMRMS.
## Outcome
The primary outcomes were the frequency and time of outpatient visits. The secondary outcome was changes in ASDAS after the EMRMS intervention.
## Statistical analysis
Continuous variables are reported as means ± standard deviations, and categorical variables are reported as percentages. Differences in continuous variables were assessed for the same patient at two time points by using a paired t test. Shapiro–Wilk/Kolmogorov–Smirnov test was used for normality test, non-parametric data was used Wilcoxon signed rank test to compare paired data. The results of a cross-validation analysis strongly supported the selected cutoff values15–18. Data were analyzed using SAS software (SAS Institute, Inc., Cary, NC, USA).
## Results
We enrolled 652 patients with AS who were followed up for at least 1 year before and after their first ASDAS assessment (Fig. 1) and compared the number of outpatient visits and average visit time within one year before and after the initial ASDAS assessment. We identified 201 AS patients who received ≥ 3 continuous ASDAS assessment with at an interval of 3 months and compared the results of the second and third ASDAS assessments with those of the first. Figure 1Flowchart of patient enrollment.
## Basline characterisitics
The mean age of the 652 eligible patients with AS at their first assessment on the EMRMS was 43.1 ± 13.7 years, and 475 patients were men ($72.9\%$). The AS onset age distribution was 26.8 ± 11.5 years, the disease duration was 16.4 ± 11.7 years, 430 patients were HLA-B27 positive, and 203 were undergoing biologic treatment ($31.1\%$). The most common comorbidity was hypertension (131 patients), the most common AS symptom was uveitis (168 patients), and the most common cause of past history was fracture (70 patients, $10.7\%$). The mean age of the 201 eligible patients with AS and more than three consecutive assessments on the EREMS was 43.3 ± 13.4 years. In all, 148 were men ($73.6\%$), the AS onset age distribution was 27.0 ± 11.2 years, the disease duration was 16.3 ± 11.6 years, 136 were HLA-B27 positive, and 49 were undergoing biologic treatment ($24.4\%$). The most common comorbidity was hypertension (43 patients), the most common AS symptom was uveitis (53 patients), and the most common cause of past history was fracture (15 patients) Table 1.Table 1Baseline characteristics of patients with AS obtained using the electronic medical record management system for ASDAS assessment. AS patients, $$n = 652$$Eligible AS patients with ≥ 3 continuous assessment, $$n = 201$$Age at first ASDAS assessment, years, Mean ± SD (n/N)43.1 ± 13.7 ($\frac{652}{652}$)43.3 ± 13.4 ($\frac{201}{201}$)Gender Female27.1 ($\frac{177}{652}$)26.4 ($\frac{53}{201}$) Male72.9 ($\frac{475}{652}$)73.6 ($\frac{148}{201}$)Age at AS diagnosis, years, mean ± SD (n/N)26.8 ± 11.5 ($\frac{652}{652}$)27.0 ± 11.2 ($\frac{200}{201}$)Disease duration, years, mean ± SD (n/N)16.4 ± 11.7 ($\frac{652}{652}$)16.3 ± 11.6 ($\frac{200}{201}$)HLA-B27 positive*88.7 ($\frac{430}{485}$)86.1 ($\frac{136}{158}$)Biologics therapy31.1 ($\frac{203}{652}$)24.4 ($\frac{49}{201}$)Comorbidities* Hypertension, % (n/N)20.2 ($\frac{131}{648}$)21.5 ($\frac{43}{200}$) Diabetes mellitus, % (n/N)7.4 ($\frac{48}{650}$)4.5 ($\frac{9}{200}$) Hyperlipidemia, % (n/N)14.9 ($\frac{96}{644}$)16.2 ($\frac{32}{197}$) Hepatitis B, % (n/N)10.9 ($\frac{71}{649}$)9.5 ($\frac{19}{200}$) Hepatitis C, % (n/N)2.3 ($\frac{15}{650}$)3.0 ($\frac{6}{200}$) Renal insufficiency, % (n/N)3.2 ($\frac{21}{651}$)2.0 ($\frac{4}{200}$) Gout, % (n/N)4.5 ($\frac{29}{651}$)3.0 ($\frac{6}{200}$) Coronary artery disease, % (n/N)3.2 ($\frac{21}{650}$)3.5 ($\frac{7}{200}$) Stroke, % (n/N)0.3 ($\frac{2}{651}$)0.5 ($\frac{1}{200}$) Periodontitis, % (n/N)18.7 ($\frac{121}{646}$)19.6 ($\frac{39}{199}$) Osteoporosis, % (n/N)6.7 ($\frac{42}{630}$)6.2 ($\frac{12}{193}$) Tuberculosis history, % (n/N)6.9 ($\frac{45}{648}$)6.1 ($\frac{12}{198}$)AS-related manifestations Uveitis, % (n/N)25.8 ($\frac{168}{651}$)26.8 ($\frac{53}{198}$) Psoriasis, % (n/N)6.8 ($\frac{44}{649}$)11.1 ($\frac{22}{199}$) Crohn's disease, % (n/N)0.0 ($\frac{0}{650}$)0.5 ($\frac{1}{200}$) Ulcerative colitis, % (n/N)0.5 ($\frac{3}{651}$)1.0 ($\frac{2}{199}$) Peripheral arthritis, % (n/N)19.3 ($\frac{125}{646}$)21.5 ($\frac{43}{200}$) Enthesitis, % (n/N)14.4 ($\frac{92}{641}$)18.3 ($\frac{36}{197}$) Dactylitis, % (n/N)2.3 ($\frac{15}{652}$)2.0 ($\frac{4}{199}$)Family history AS-First degree relatives, % (n/N)18.7 ($\frac{119}{636}$)15.5 ($\frac{30}{193}$) AS-Secondary degree relatives, % (n/N)29.0 ($\frac{185}{639}$)24.5 ($\frac{48}{196}$) Psoriasis, % (n/N)4.2 ($\frac{27}{643}$)5.1 ($\frac{10}{196}$) Psoriatic arthritis, % (n/N)0.6 ($\frac{4}{641}$)1.0 ($\frac{2}{194}$) Uveitis, % (n/N)4.7 ($\frac{30}{643}$)6.1 ($\frac{12}{196}$) Crohn’s disease, % (n/N)0.0 ($\frac{0}{641}$)0.0 ($\frac{0}{195}$) Ulcerative colitis, % (n/N)0.3 ($\frac{2}{641}$)0.5 ($\frac{1}{195}$) Rheumatoid arthritis, % (n/N)6.0 ($\frac{38}{638}$)5.7 ($\frac{11}{194}$) Systemic Lupus Erythematosus, % (n/N)3.0 ($\frac{19}{641}$)2.6 ($\frac{5}{195}$) Sicca syndrome, % (n/N)2.5 ($\frac{16}{641}$)3.1 ($\frac{6}{195}$)Past history Total hip replacement, % (n/N)3.8 ($\frac{25}{652}$)2.0 ($\frac{4}{200}$) Total knee replacement, % (n/N)0.6 ($\frac{4}{652}$)0.5 ($\frac{1}{200}$) Fracture, % (n/N)10.7 ($\frac{70}{652}$)7.5 ($\frac{15}{200}$) Palindromic rheumatism, % (n/N)1.1 ($\frac{7}{652}$)0.5 ($\frac{1}{200}$)Abbreviations: AS ankylosing spondylitis, ASDAS Ankylosing Spondylitis Disease Activity Score.
## Primary outcomes
After the first assessment on the EREMS within 1 year, the frequency of outpatient visits increased from 4.0 (4.0, 7.0) to 4.0 (4.0, 8.0) ($p \leq 0.001$), particularly in patients with a high disease activity (ASDAS-CRP, 4.0 (4.0, 7.0) vs 4.0 (4.0, 8.0), $$p \leq 0.001$$ and ASDAS-ESR, 3.0 (4.0, 7.0) vs 4.0 (4.0, 9.0), $p \leq 0.001$) and very high disease activity (ASDAS-CRP, 3.0 (5.5, 8.0) vs 5.0 (9.0, 12.0), $p \leq 0.001$ and ASDAS-ESR, 3.0 (4.0, 8.0) vs 5.0 (8.0, 12.0), $$p \leq 0.002$$; Table 2). The duration of outpatient visits decreased from 6.4 (8.5, 11.2) to 6.3 (8.3, 10.8), ($$p \leq 0.073$$), especially in those with an inactive disease (ASDAS-CRP, 6.7 (8.8, 11.1) vs 6.1 (8.0, 10.3) min, $$p \leq 0.033$$; ASDAS-ESR, 6.4 (8.7, 11.1) vs 6.1 (8.1, 10.0) min, $$p \leq 0.027$$; Table 3).Table 2Frequency of outpatient visits before and after the first EMRMS assessment ($$n = 652$$).(Min, Max)(mean ± SD)Median (P25, P75)p-value of normality test*p-value#Analysis populationASDASNumberOne year before assessmentOne year after assessmentOne year before assessmentOne year after assessmentOne year before assessmentOne year after assessmentCRP < 1.3211(1.0, 13.0)(1.0, 15.0)5.5 ± 3.45.6 ± 3.44.0 (4.0, 7.0)4.0 (4.0, 7.0) < 0.0100.3881.3—< 2.1245(1.0, 15.0)(1.0, 14.0)5.4 ± 3.45.6 ± 3.43.0 (4.0, 7.0)4.0 (4.0, 7.0) < 0.0100.4002.1—3.5170(1.0, 15.0)(1.0, 18.0)5.1 ± 3.25.9 ± 3.54.0 (4.0, 7.0)4.0 (4.0, 8.0) < 0.0100.001 > 3.526(1.0, 12.0)(1.0, 15.0)5.7 ± 3.48.7 ± 4.03.0 (5.5, 8.0)5.0 (9.0, 12.0)0.001 < 0.001ESR < 1.3195(1.0, 14.0)(1.0, 13.0)5.4 ± 3.25.4 ± 3.14.0 (4.0, 6.0)4.0 (4.0, 7.0) < 0.0100.8581.3—< 2.1258(1.0, 13.0)(1.0, 18.0)5.3 ± 3.45.5 ± 3.43.0 (4.0, 7.0)4.0 (4.0, 7.0) < 0.0100.1512.1—3.5174(1.0, 15.0)(1.0, 15.0)5.4 ± 3.56.2 ± 3.73.0 (4.0, 7.0)4.0 (4.0, 9.0) < 0.010 < 0.001 > 3.525(1.0, 12.0)(1.0, 15.0)5.6 ± 3.68.4 ± 3.93.0 (4.0, 8.0)5.0 (8.0, 12.0)0.0730.002Total652(1.0, 15.0)(1.0, 18.0)5.4 ± 3.45.8 ± 3.54.0 (4.0, 7.0)4.0 (4.0, 8.0) < 0.010 < 0.001*Normality test: When number is ≤ 50, the Shapiro–Wilk test is used; when number is > 50, the Kolmogorov–Smirnov test is used.#When normality test p value is > 0.05, the pair-t test is used; when normality test p value is < 0.05, the Wilcoxon Signed Rank test is used. Abbreviations: ASDAS Ankylosing Spondylitis Disease Activity Score, CRP C-reactive protein, ESR erythrocyte sedimentation rate. Table 3Time of outpatient visits before and after the first EMRMS assessment ($$n = 652$$).Time of outpatient visits (minutes)(Min, Max)Mean ± SDMedian (P25, P75)p-value of normality test*p-value#Analysis populationASDASNumberOne year before assessmentOne year after assessmentOne year before assessmentOne year after assessmentOne year before assessmentOne year after assessmentCRP < 1.3211(1.5, 41.8)(0.3, 22.5)9.2 ± 4.28.5 ± 3.36.7 (8.8, 11.1)6.1 (8.0, 10.3) < 0.0100.0331.3—< 2.1245(0.8, 31.6)(1.0, 37.0)9.3 ± 4.89.0 ± 4.46.3 (8.5, 11.2)6.4 (8.5, 11.0) < 0.0100.3952.1—3.5170(0.9, 22.0)(0.9, 21.5)8.7 ± 4.18.5 ± 3.46.0 (8.3, 11.3)6.3 (8.1, 10.5) < 0.0100.825 > 3.526(1.2, 17.3)(4.3, 15.0)9.6 ± 4.19.8 ± 2.76.7 (10.0, 11.6)8.8 (9.7, 11.5)0.2840.827ESR < 1.3195(1.4, 41.8)(1.7, 22.5)9.1 ± 4.28.4 ± 3.36.4 (8.7, 11.1)6.1 (8.1, 10.0) < 0.0100.0271.3—< 2.1258(0.8, 31.6)(0.3, 37.0)8.9 ± 4.28.5 ± 4.16.3 (8.4, 10.8)6.0 (7.7, 10.7)0.0150.3182.1—3.5174(1.1, 29.5)(0.9, 24.1)9.3 ± 4.69.4 ± 3.76.2 (8.4, 11.5)6.8 (9.0, 11.5) < 0.0100.623 > 3.525(3.3, 22.0)(3.2, 15.0)11.0 ± 4.59.3 ± 2.97.7 (11.0, 13.3)8.2 (9.3, 11.5)0.6100.124Total652(0.8, 41.8)(0.3, 37.0)9.2 ± 4.48.7 ± 3.86.4 (8.5, 11.2)6.3 (8.3, 10.8) < 0.0100.073*Normality test: When number is ≤ 50, the Shapiro–Wilk test is used; when number is > 50, the Kolmogorov–Smirnov test is used.#When normality test p value is > 0.05, the pair-t test is used; when normality test p value is < 0.05, the Wilcoxon Signed Rank test is used. Abbreviations: ASDAS Ankylosing Spondylitis Disease Activity Score, CRP C-reactive protein, ESR Erythrocyte sedimentation rate.
## Secondary outcomes
The third ASDAS-CRP and ASDAS-ESR tend to be lower than the first assessment (1.5 (0.9, 2.1), 1.4 (0.8, 1.9), $$p \leq 0.058$$ and 1.4 (1.1, 2.0), 1.4 (1.0, 1.9), $$p \leq 0.161$$, respectively), but both ASDAS-CRP and ASDAS-ESR did not reached significant difference (Table 4).Table 4ASDAS-CRP and ASDAS-ESR from three consecutive assessments† ($$n = 201$$).T1T2T3p-value#T1 versus T2T2 versus T3T1 versus T3ASDAS-CRP (Min, Max)(0.1, 4.1)(0.0, 4.7)(0.1, 4.3)0.2250.5800.058 Mean ± SD1.6 ± 0.81.5 ± 0.81.5 ± 0.8 Median (P25, P75)1.5 (0.9, 2.1)1.3 (0.9, 1.9)1.4 (0.8, 1.9)ASDAS-ESR (Min, Max)(0.3, 4.4)(0.3, 4.2)(0.4, 4.2)0.5100.4880.161 Mean ± SD1.6 ± 0.81.5 ± 0.71.5 ± 0.7 Median (P25, P75)1.4 (1.1, 2.0)1.5 (1.0, 2.0)1.4 (1.0, 1.9)Abbreviations: ASDAS Ankylosing Spondylitis Disease Activity Score, CRP C-reactive protein, ESR erythrocyte sedimentation rate, T time, SD standard deviation, P percentile.†Comorbidities were identified within 1 year before the index date.#Because the p-value of the normality test (Kolmogorov–Smirnov test) is < 0.05, the Wilcoxon Signed Rank test is used.†Time of the three consecutive assessment: T1: the earliest time was considered the first time; T2: the second time was 84 ± 7 days; T3: the third time was 168 ± 7 days.
## Discussion
To our knowledge, several meta-analyses have assessed the effectiveness of mHealth applications for monitoring AS disease activity in multiple centers5,6. However, no study had explored the influence of EMRMS intervention on disease activity and the time and frequency of outpatient visits among patients with AS in a single medical center. This study demonstrated changes in the health-related behaviors of patients with AS, including increased outpatient visit frequency among AS patients with high and very high disease activity, as well asdecreased outpatient visit time among AS patients with inactive disease activity.
The study findings indicate that the proposed smartphone-based management system is a time- and cost-effective disease management tool, achieving high ASDAS and the efficient detection of inflammatory markers in a Chinese AS cohort17,18. Self-report ASDAS questionnaires have been applied extensively to evaluate the disease activity of patients with AS in single medical center studies20–25. In contrast to previous studies, the current study linked the TCVGH HIS with smartphone applications, wherein laboratory data are automatically integrated into the app and trained nurses assist patients with completing assessments through an NRS on the app for ASDAS calculation. Nobody calls AS patients for additional visit. All treatment decisions, including frequency of follow-up, were left to the physician's discretion. Visits to assess study ASDAS were scheduled at least every 12 weeks. Given that the TCVGH is a center of excellence endorsed by the Asia Pacific League of Associations for Rheumatology, we assume that most rheumatologists at the TCVGH tried to apply the recommendations of treat to target published in 2017 on line first after monitoring of ASDAS in their usual care26. In addition, the results of the TICOSPA trial that aimed to assess the efficacy of a tight-control and treat-to-target strategy in axial spondyloarthritis showed that patients with 'usual care' also showed a significant improvement of disease activity27.
This system has a high interrater reliability, accuracy, and precision. Optimal treatment of AS must involve shared decision-making between patients and health professionals. Through this user-friendly EMRMS, patients with AS can more thoroughly understand their disease severity, which may in turn improve their treatment adherence rates and increase clinic visits. This study comprehensively furnished many disease information, such as comorbidities, medications, laboratory data, family history and past history. The study findings provide new insight into the use of apps for disease monitoring to reduce consultation times for individuals with a mild disease status, improve the treatment adherence rates of individuals with a severe disease status, and ameliorate AS disease activity.
Our study has some limitations. First, data were collected from a single medical center in Taiwan, which may have introduced selection bias. Second, we did not have the data of changes in disease activity and frequency of outpatient visits in AS patients who did not receive ASDAS assessment using the EMRMS. Therefore, we cannot compare our data of the VGHTC-AS cohort with those of a control group. Third, We need longer follow-up period to investigate the alteration of ASDAS-CRP status in patients who received continual regular assessment using the EMRMS. Finally, the results may not be generalizable to the entire population of Taiwan with AS.
## Conclusions
This is the first single medical center study in Taiwan to compare the treatment outcomes of patients with AS after EMRMS management. The EMRMS is an effective management system that offers satisfactory levels of usability; the data obtained are of a high quality, and the system enables a comprehensive analysis of patient function, helping patients more accurately understand their conditions and thus leading to improved patient treatment adherence. The EMRMS help increase the frequency of outpatient visits in those with high disease activity and may improve control of disease activity in patients with AS. Further research should explore the application of EMRMS for the management of other chronic diseases.
## References
1. Sepriano A, Ramiro S, van der Heijde D, van Gaalen F, Hoonhout P, Molto A, Saraux A, Ramonda R, Dougados M, Landewé R. **What is axial spondyloarthritis? A latent class and transition analysis in the SPACE and DESIR cohorts**. *Ann. Rheum. Dis.* (2020) **79** 324-331. DOI: 10.1136/annrheumdis-2019-216516
2. Dean LE, Jones GT, MacDonald AG, Downham C, Sturrock RD, Macfarlane GJ. **Global prevalence of ankylosing spondylitis**. *Rheumatol. (Oxf.)* (2014) **53** 650-657. DOI: 10.1093/rheumatology/ket387
3. Chou CT, Pei L, Chang DM, Lee CF, Schumacher HR, Liang MH. **Prevalence of rheumatic diseases in Taiwan: A population study of urban, suburban, rural differences**. *J. Rheumatol.* (1994) **21** 302-306. PMID: 8182641
4. Ramiro S, Nikiphorou E, Sepriano A, Ortolan A, Webers C, Baraliakos X, Landewé RBM, Van den Bosch FE, Boteva B, Bremander A, Carron P, Ciurea A, van Gaalen FA, Géher P, Gensler L, Hermann J, de Hooge M, Husakova M, Kiltz U, López-Medina C, van der Heijde D. **ASAS-EULAR recommendations for the management of axial spondyloarthritis: 2022 update**. *Ann. Rheum. Dis.* (2023) **82** 19-34. DOI: 10.1136/ard-2022-223296
5. Wei JC, Liu CH, Tseng JC, Hsieh LF, Chen CH, Chen HH, Chen HA, Chen YC, Chou CT, Liao HT, Lin YC, Luo SF, Yang DH, Yeo KJ, Tsai WC. **Taiwan Rheumatology Association consensus recommendations for the management of axial spondyloarthritis**. *Int. J. Rheum. Dis.* (2020) **23** 7-23. DOI: 10.1111/1756-185X.13752
6. El-Khatib Z, Shah M, Zallappa SN, Nabeth P, Guerra J, Manengu CT, Yao M, Philibert A, Massina L, Staiger CP, Mbailao R, Kouli JP, Mboma H, Duc G, Inagbe D, Barry AB, Dumont T, Cavailler P, Quere M, Willett B, Reeder B. **SMS-based smartphone application for disease surveillance has doubled completeness and timeliness in a limited-resource setting - evaluation of a 15-week pilot program in Central African Republic (CAR)**. *Confl. Health* (2018) **12** 42. DOI: 10.1186/s13031-018-0177-6
7. Moses JC, Adibi S, Shariful Islam SM, Wickramasinghe N, Nguyen L. **Application of smartphone technologies in disease monitoring: A systematic review**. *Healthc. (Basel, Switz.)* (2021) **9** 889. DOI: 10.3390/healthcare9070889
8. Marcano Belisario JS, Huckvale K, Greenfield G, Car J, Gunn LH. **Smartphone and tablet self management apps for asthma**. *Cochrane Datab. Syst. Rev.* (2013) **2013** 010013. DOI: 10.1002/14651858.CD010013.pub2
9. Wang Y, Xue H, Huang Y, Huang L, Zhang D. **A systematic review of application and effectiveness of mHealth interventions for obesity and diabetes treatment and self-management**. *Adv. Nutrit. (Bethesda, Md.)* (2017) **8** 449-462. DOI: 10.3945/an.116.014100
10. Coorey GM, Neubeck L, Mulley J, Redfern J. **Effectiveness, acceptability and usefulness of mobile applications for cardiovascular disease self-management: Systematic review with meta-synthesis of quantitative and qualitative data**. *Eur. J. Prev. Cardiol.* (2018) **25** 505-521. DOI: 10.1177/2047487317750913
11. Bonnechère B, Rintala A, Spooren A, Lamers I, Feys P. **Is mHealth a useful tool for self-assessment and rehabilitation of people with multiple sclerosis? A systematic review**. *Brain Sci.* (2021) **11** 1187. DOI: 10.3390/brainsci11091187
12. Debon R, Coleone JD, Bellei EA, De Marchi A. **Mobile health applications for chronic diseases: A systematic review of features for lifestyle improvement**. *Diabetes Metab. Syndr.* (2019) **13** 2507-2512. DOI: 10.1016/j.dsx.2019.07.016
13. Ji X, Wang Y, Ma Y, Hu Z, Man S, Zhang Y, Li K, Yang J, Zhu J, Zhang J, Huang F. **Improvement of disease management and cost effectiveness in Chinese patients with ankylosing spondylitis using a smart-phone management system: A prospective cohort study**. *Biomed. Res. Int.* (2019) **2019** 2171475. DOI: 10.1155/2019/2171475
14. Chen HH, Chen YM, Lai KL, Hsieh TY, Hung WT, Lin CT, Tseng CW, Tang KT, Chou YY, Wu YD, Huang CY, Hsieh CW, Huang WN, Chen YH. **Gender difference in ASAS HI among patients with ankylosing spondylitis**. *PLoS ONE* (2020) **15** e0235678. DOI: 10.1371/journal.pone.0235678
15. van der Linden S, Valkenburg HA, Cats A. **Evaluation of diagnostic criteria for ankylosing spondylitis. A proposal for modification of the New York criteria**. *Arthr. Rheum.* (1984) **27** 361-368. DOI: 10.1002/art.1780270401
16. van der Heijde D, Lie E, Kvien TK, Sieper J, Van den Bosch F, Listing J, Braun J, Landewé R. **ASDAS, a highly discriminatory ASAS-endorsed disease activity score in patients with ankylosing spondylitis**. *Ann. Rheum. Dis.* (2009) **68** 1811-1818. DOI: 10.1136/ard.2008.100826
17. Machado P, Landewé R, Lie E, Kvien TK, Braun J, Baker D, van der Heijde D. **Ankylosing Spondylitis Disease Activity Score (ASDAS): Defining cut-off values for disease activity states and improvement scores**. *Ann. Rheum. Dis.* (2011) **70** 47-53. DOI: 10.1136/ard.2010.138594
18. Machado PM, Landewé R, Heijde DV. **Ankylosing spondylitis disease activity score (ASDAS): 2018 update of the nomenclature for disease activity states**. *Ann. Rheum. Dis.* (2018) **77** 1539-1540. DOI: 10.1136/annrheumdis-2018-213184
19. Au YL, Wong WS, Mok MY, Chung HY, Chan E, Lau CS. **Disease activity assessment in ankylosing spondylitis in a Chinese cohort: BASDAI or ASDAS?**. *Clin. Rheumatol.* (2014) **33** 1127-1134. DOI: 10.1007/s10067-014-2729-5
20. Godfrin-Valnet M, Prati C, Puyraveau M, Toussirot E, Letho-Gyselink H, Wendling D. **Evaluation of spondylarthritis activity by patients and physicians: ASDAS, BASDAI, PASS, and flares in 200 patients**. *Joint Bone Spine* (2013) **80** 393-398. DOI: 10.1016/j.jbspin.2013.01.003
21. Wach J, Letroublon MC, Coury F, Tebib JG. **Fibromyalgia in spondyloarthritis: effect on disease activity assessment in clinical practice**. *J. Rheumatol.* (2016) **43** 2056-2063. DOI: 10.3899/jrheum.160104
22. López-Medina C, Garrido-Castro JL, Castro-Jiménez J, González-Navas C, Calvo-Gutiérrez J, Castro-Villegas MC, Ortega-Castro R, Escudero-Contreras A, Font-Ugalde P, Collantes-Estévez E. **Evaluation of quality of life in patients with axial spondyloarthritis and its association with disease activity, functionality, mobility, and structural damage**. *Clin. Rheumatol.* (2018) **37** 1581-1588. DOI: 10.1007/s10067-018-4112-4
23. Chiowchanwisawakit P, Thaweeratthakul P, Wattanamongkolsil L, Srinonprasert V, Koolvisoot A, Muangchan C, Nilganuwong S, Arromdee E, Katchamart W. **Relationship between health-related quality of life and patient acceptable symptom state with disease activity and functional status in patients with ankylosing spondylitis in Thailand**. *J. Clin. Rheumatol. Pract. Rep. Rheum. Musculoskelet. Dis.* (2019) **25** 16-23. DOI: 10.1097/RHU.0000000000000750
24. Bedaiwi MK, AlRasheed RF, Bin Zuair A, Alqurtas EM, Baeshen MO, Omair MA. **A cross-sectional study on clinical characteristics of Saudi axial spondylarthritis: Preliminary results**. *Eur. Rev. Med. Pharmacol. Sci.* (2021) **25** 5241-5247. DOI: 10.26355/eurrev_202108_26538
25. Rusman T, Nurmohamed MT, Hoekstra S, van Denderen CJ, van Vollenhoven RF, Boers M, Ter Wee MM, van der Horst-Bruinsma IE. **Disease activity in women with ankylosing spondylitis remains higher under Tumour Necrosis Factor inhibitor treatment than in men: a five-year observational study**. *Scand. J. Rheumatol.* (2021). DOI: 10.1080/03009742.2021.1967046
26. Smolen JS, Schöls M, Braun J, Dougados M, FitzGerald O, Gladman DD, Kavanaugh A, Landewé R, Mease P, Sieper J, Stamm T, Wit M, Aletaha D, Baraliakos X, Betteridge N, Bosch FVD, Coates LC, Emery P, Gensler LS, Gossec L, van der Heijde D. **Treating axial spondyloarthritis and peripheral spondyloarthritis, especially psoriatic arthritis, to target: 2017 update of recommendations by an international task force**. *Ann. Rheum. Dis.* (2018) **77** 3-17. DOI: 10.1136/annrheumdis-2017-211734
27. Molto A, López-Medina C, Van den Bosch FE, Boonen A, Webers C, Dernis E, van Gaalen FA, Soubrier M, Claudepierre P, Baillet A, Starmans-Kool M, Spoorenberg A, Jacques P, Carron P, Joos R, Lenaerts J, Gossec L, Pouplin S, Ruyssen-Witrand A, Sparsa L, Dougados M. **Efficacy of a tight-control and treat-to-target strategy in axial spondyloarthritis: results of the open-label, pragmatic, cluster-randomised TICOSPA trial**. *Ann. Rheum. Dis.* (2021) **80** 1436-1444. DOI: 10.1136/annrheumdis-2020-219585
|
---
title: Longitudinal uric acid has nonlinear association with kidney failure and mortality
in chronic kidney disease
authors:
- Mathilde Prezelin-Reydit
- Christian Combe
- Denis Fouque
- Luc Frimat
- Christian Jacquelinet
- Maurice Laville
- Ziad A. Massy
- Céline Lange
- Carole Ayav
- Roberto Pecoits-Filho
- Sophie Liabeuf
- Bénédicte Stengel
- Jérôme Harambat
- Karen Leffondré
- Natalia Alencar de Pinho
- Natalia Alencar de Pinho
- Yves-Edouard Herpe
- Christophe Pascal
- Joost Schanstra
- Oriane Lambert
- Marie Metzger
- Elodie Speyer
journal: Scientific Reports
year: 2023
pmcid: PMC9998636
doi: 10.1038/s41598-023-30902-7
license: CC BY 4.0
---
# Longitudinal uric acid has nonlinear association with kidney failure and mortality in chronic kidney disease
## Abstract
We investigated the shape of the relationship between longitudinal uric acid (UA) and the hazard of kidney failure and death in chronic kidney disease (CKD) patients, and attempted to identify thresholds associated with increased hazards. We included CKD stage 3–5 patients from the CKD-REIN cohort with one serum UA measurement at cohort entry. We used cause-specific multivariate Cox models including a spline function of current values of UA (cUA), estimated from a separate linear mixed model. We followed 2781 patients ($66\%$ men, median age, 69 years) for a median of 3.2 years with a median of five longitudinal UA measures per patient. The hazard of kidney failure increased with increasing cUA, with a plateau between 6 and 10 mg/dl and a sharp increase above 11 mg/dl. The hazard of death had a U-shape relationship with cUA, with a hazard twice higher for 3 or 11 mg/dl, compared to 5 mg/dl. In CKD patients, our results indicate that UA above 10 mg/dl is a strong risk marker for kidney failure and death and that low UA levels below 5 mg/dl are associated with death before kidney failure.
## Introduction
Chronic kidney disease (CKD) is recognized as a major public health problem and identification of modifiable determinants of CKD progression, other than hypertension or proteinuria1, is essential to develop effective strategies to slow disease progression. In the past two decades, uric acid (UA) has drawn attention in the nephrology community.
Previous cohort studies found conflicting results on the relationship between UA and CKD progression2–11, and did not well characterize the shape of this longitudinal relationship12. The discrepancy between results may be due to differences in CKD stages at baseline, duration of follow-up, definitions of CKD progression, the choice of adjustment factors13, or to the use of a single measure of UA assessed at baseline, i.e. entry into the cohort2–5,8–10. Indeed, entry into a cohort does not usually correspond to any relevant time point in patient’s course of CKD and varies across studies and from patient to patient. To estimate the association between UA and CKD progression, most previous studies thus compared the baseline uric acid values between two patients who potentially did not enter the cohort at the same moment of their CKD history, and fully ignored their respective subsequent evolution of uric acid between baseline and the current time when the risk was assessed. Yet, UA may change over time, and from a clinical point of view, it might be more relevant to compare the level of UA reached at the time when the risk is assessed. To our knowledge, only one single-center cohort study in Taiwan used longitudinal measures of UA14. They found that elevated UA trajectories had an increased hazard of dialysis initiation and death. However, the analysis based on classes of trajectories did not allow the examination of the shape of the relationship between the level of UA reached at a given time and the hazard of kidney failure or death at the same time, and thus the identification of potential critical thresholds of UA, which yet might be useful in monitoring patients with CKD.
The objective of the present study was to investigate, in the French Chronic Kidney Disease-Renal Epidemiology and Information Network (CKD-REIN) cohort15, the shape of the relationship between longitudinal UA and the hazard of kidney failure and death in CKD stage 3–5 patients, and identify potential thresholds of longitudinal UA associated with increased hazards.
## Study population
CKD-REIN is an ongoing prospective cohort of CKD Stage 3–5 patients receiving nephrologist-led care, without prior chronic dialysis or kidney transplantation15. The study included 3033 patients over 18 years of age. They were first selected and recruited in an enrolment phase in 40 nephrology centers located over all metropolitan France and representative of all centers with respect to legal status (public, private non-for-profit, and private for-profit), and then actively included into the cohort at a first visit (baseline) between 2013 and 2016, in the same centers, after obtaining inform consent15. The study protocol was conducted with adherence to the Declaration of Helsinki and approved by the institutional review board at the French National Institute of Health and Medical Research (INSERM; reference: IRB00003888). The study was registered at ClinicalTrials.gov (NCT03381950). In the present study, we included patients who had at least one serum UA and one creatinine measurement within 6 months of their inclusion into the cohort.
## Exposure and outcomes
The exposure was serum UA concentration, assessed in each center along with routine laboratory investigations, at baseline, annually per protocol and more frequently if considered necessary by the nephrologist.
The outcomes of interest were [1] kidney failure assessed by initiation of chronic dialysis or pre-emptive transplantation, and [2] death before kidney failure. To ensure complete collection of kidney failure, a record linkage was performed with the national REIN registry. Administrative censoring was performed on July 30, 2018.
## Covariates
Baseline characteristics were recorded by clinical research associates from medical files or by interview. Data included age, sex, body mass index (BMI), hypertension (patients having an office blood pressure greater than or equal to $\frac{140}{90}$ mmHg or an antihypertensive treatment), cardiovascular history (coronary artery disease, arrhythmic disorders, congestive heart failure, stroke, peripheral vascular disease and/or valvulopathy), diabetes (diabetes history or antidiabetic treatment or glycated hemoglobin ≥ $6.5\%$ or fasting glycemia ≥ 7 mmol/l or non-fasting glycemia ≥ 11), gout history, dyslipidemia, primary kidney disease, time since CKD diagnosis (time elapsed from the date of CKD diagnosis found in the medical record and the cohort entry), number of consultation in the previous year with nephrologist and dietician, treatment (urate-lowering therapy (ULT), diuretics, antiplatelet agents, renin-angiotensin system inhibitors (RASi)), laboratory data (serum creatinine, eGFR estimated by the CKD-EPI equation, serum UA, albuminemia, C-reactive protein and, albuminuria—or equivalent—classified according to the KDIGO 2012 guidelines16), salt intake (estimated by 24-h natriuresis) and protein intake (estimated by 24-h urinary urea)17, medication adherence according to the Girerd score in categories (good, minimal and poor)18, health literacy according to their need for help reading medical documents (never need vs always or partly need)19 and type of center (university, non-university hospital, private non-profit and private for-profit clinic).
## Statistical analyses
We described characteristics of all the patients in the CKD-REIN cohort, those included in the present study, as well as those who contributed to the estimation of the different statistical models. We also estimated the crude association between patients’ characteristics and UA at baseline.
To estimate the shape of the relationship between the level of UA reached at a given time, i.e. current level of UA (cUA), and the hazard of kidney failure or death at the same time, we used a two-stage statistical approach which have been evaluated previously20–22. At Stage 1, we estimated for each patient its full longitudinal trajectory of UA from a linear mixed model estimated using information on all patients. This allowed us to get cUA at any time point of follow-up for each patient. At Stage 2, we estimated the association between cUA and the hazards of kidney failure or death using cause-specific Cox models including cUA as a continuous time-dependent variable. More specifically, the linear mixed model (at Stage 1) included a 3-knot natural cubic spline function of time with random effects on each coefficient, and some selected baseline factors (age, sex, hypertension, eGFR, BMI, use of ULT and diuretics) to make the estimation of all individual UA trajectories more accurate. The cause-specific Cox models (at Stage 2) included a 2-knot natural cubic spline function of cUA (derived from Stage 1) to allow the estimation of nonlinear association and thus the detection of potential thresholds. The number of knots were selected using the Akaike Information Criterion (AIC)23. The Cox models were adjusted for a set of confounders which were selected from a directed acyclic graph (DAG), i.e. a diagram of causal pathways summarizing a priori hypothetical causal relationship between variables (Figs. S1 and S2). This approach allows the selection of optimal set of adjustment factors (i.e. factors associated both with the exposure and the outcome), avoiding adjustment for unnecessary factors, mediators, and colliders24. For kidney failure, this included age (in years), sex, primary kidney disease (diabetic, glomerular, hypertensive, vascular, tubulo-interstitial, polycystic or unknown nephropathy), hypertension (yes/no), diabetes (yes/no), cardiovascular disease (yes/no), dyslipidemia (yes/no), BMI (< 25, 25–30, ≥ 30 kg/m2), albuminuria (A1, A2, A3), CKD stages (5, 4, 3B, 3A or less), medication adherence (good, minimal, poor), RASi (yes/no), and ULT (yes/no) (Fig. S1, Model 1 for kidney failure). For death, we added spironolactone (yes/no) and anti-platelet agents (yes/no) (Fig. S2, Model 1 for death). All confounders (including renal function represented by CKD stages) were taken at baseline only to respect the temporal sequence between confounders and the exposure, i.e. the subsequent cUA which was the only time-dependent variable. The analyses thus accounted for the fact that UA may be consequence of a decreased renal function by adjusting the effect of cUA for CKD stage at baseline.
To investigate if the association between cUA and the hazard of kidney failure or death differed according to sex, we included interaction terms with the spline function of cUA, and test the interaction using the likelihood ratio test.
In a first sensitivity analysis, we further adjusted Model 1 for salt intake (< 95, 95–127, 128–170 and ≥ 170 mmol/day) and protein intake (in mmol/day) at baseline (called “eating habits” in the DAG) (Model 2) because patients with high salt or high protein intake may be at higher risk of hyperuricemia25,26 and CKD progression27,28. We perform this further adjustment in the subsample of patients having 24-h natriuresis and urinary urea at baseline. In a second sensitivity analysis, for comparison with previous studies, we estimated the association between baseline value of UA and outcomes using the same set of adjustment factors as Model 1 (Model 3).
In all Cox models, we accounted for correlation between patients of the same type of center using robust standard errors based on the sandwich estimator29. Proportional hazards assumption was checked using Schoenfeld residuals. Linearity of the effect of all adjusting quantitative variables was checked using 4-df penalized spline functions30, which was kept in the model if the effect was nonlinear31. All analyses were performed using R version 3.6.032.
## Patients’ selection and characteristics, UA distribution, and number of events
Among the 3033 patients enrolled in the CKD-REIN study, 2781 patients had a UA measurement within six months of inclusion into the cohort and were thus included in the present study (Fig. 1). At baseline, they had a median age of 69.0 years (interquartile range (IQR): 60.0–76.0), $65.5\%$ were men, $96.1\%$ had hypertension, $42.5\%$ diabetes, $53.4\%$ cardiovascular disease, $73.2\%$ dyslipidemia, and $22\%$ a gout history (Table 1, Included Population). About $40\%$ of patients had either hypertensive nephropathy or diabetic nephropathy. Median eGFR was 32 ml/min/1.73 m2 (IQR: 23–41), with $94\%$ of CKD stage 3 or 4 ($6\%$ of patients who had progressed to CKD stage 2 or 5 between their selection and their actual inclusion into the cohort). Median salt intake was 6 g/day (natriuresis of 128 mmol/day) and estimated median protein intake was 61.1 g/day (urinary urea of 305 mmol/day). A total of 938 patients ($33.7\%$) were prescribed ULT (Table 1, Included Population). Median UA at baseline was 7.1 mg/dl (IQR: 5.8–8.5), and was statistically significantly higher in younger men and in patients with higher BMI, advanced CKD stages, diabetes, cardiovascular history, no gout history, and in those receiving diuretics and no ULT (Table S1).Figure 1Included population, CKD-REIN, France, 2013–2018.Table 1Characteristics of included population compared to the characteristics of CKD-REIN population. CKD-REIN population ($$n = 3033$$)Included population ($$n = 2781$$)Population for Models 1 and 3 ($$n = 2344$$)Population for Model 2 ($$n = 1212$$)Nn (%) or median (IQR)Nn (%) or median (IQR)Nn (%) or median (IQR)Nn (%) or median (IQR)Age (years)303369.0 (60.0–76.0)278169.0 (60.0, 76.0)234468.0 (60.0–76.0)121268.5 (61.0–76.0)Male gender30331982 (65.3)27811821 (65.5)23441549 (66.1)1212817 (67.4)Body mass index (kg/m2)296827.8 (24.6–31.8)272627.8 (24.6, 31.6)234427.9 (24.7–31.8)121228.2 (25.0–32.1)Hypertension30262915 (96.3)27742666 (96.1)23442260 (96.4)12121168 (96.4)Cardiovascular history29911594 (53.3)27401464 (53.4)23441230 (52.5)1212640 (52.8)Diabetes30261301 (43.0)27781180 (42.5)23441025 (43.7)1212520 (42.9)Dyslipidemia30192223 (73.6)27192025 (73.2)23441759 (75.0)1212920 (75.9)Gout history2968618 (20.8)2768597 (22.0)2325525 (22.6)1205283 (23.5)Primary kidney disease3033278123441212 Diabetic nephropathy611 (20.1)545 (19.6)475 (20.3)219 (18.1) Glomerulopathy532 (17.5)505 (18.2)455 (19.4)257 (21.2) Hypertensive nephropathy633 (20.9)570 (20.5)476 (20.3)239 (19.7) Vascular nephropathy216 (7.1)203 (7.3)156 (6.7)93 (7.7) Tubulo-interstitial nephropathy377 (12.4)349 (12.5)297 (12.7)159 (13.1) Polykystic renal disease166 (5.5)157 (5.6)131 (5.6)74 (6.1) Other or unknown498 (16.4)452 (16.2)354 (15.1)171 (14.1)Glomerular filtration rate (ml/min/1.73 m2)302732.0 (23.2,41.4)278131.8 (23.2, 41.4)234431.9 (23.3–41.5)121231.5 (23.2–40.8)Chronic Kidney Disease stage302727812344 10 (0.0)0 (0.0)0 (0.0)0 (0.0) 265 (2.1)58 (2.1)46 (2.0)25 (2.1) 31612 (53.3)1464 (52.6)1234 (52.6)619 (51.1) 41233 (40.7)1146 (41.2)977 (41.7)522 (43.1) 5117 (3.9)113 (4.0)87 (3.7)46 (3.8)Time since Chronic Kidney *Disease diagnosis* (years)28915.1 (2.4, 8.0)26555.2 (2.5, 10.3)22665.2 (2.5–10.4)11685.4 (2.6–11.5)Uric acid (mg/dl)27257.1 (5.8, 8.4)27227.1 (5.8, 8.4)23107.2 (5.9–8.4)12047.2 (5.9–8.5)Uric acid (in categories)2725272223101204 < 4 mg/dl119 (4.4)119 (4.4)100 (4.3)43 (3.6) 4–5 mg/dl224 (8.2)223 (8.2)185 (8.0)105 (8.7) 5–6 mg/dl416 (15.3)416 (15.3)354 (15.3)173 (14.4) 6–8 mg/dl1048 (38.5)1048 (38.5)891 (38.6)455 (37.8) 8–10 mg/dl669 (24.6)668 (24.5)569 (24.6)319 (26.5) > 10 mg/dl249 (9.1)248 (9.1)211 (9.1)109 (9.1)Protein-to-creatinine ratio (mg/mmol)184935.7 (13.7, 115.7)172435.7 (13.5, 115.7)161135.8 (13.1–114.4)74336.2 (13.5–126.7)Albumin-to-creatinine ratio269325072344 < 3 mg/mmol742 (27.6)684 (27.3)642 (27.4)318 (26.2) 3–30 mg/mmol847 (31.5)788 (31.4)729 (31.1)391 (32.3) > 30 mg/mmol1104 (41.0)1035 (41.3)973 (41.5)503 (41.5)Natriuresis (mmol/day)1663128.0 (95.0–170.5)1602128.0 (95.0—170.0)1458128.0 (96.0–169.0)1212128.0 (96.0–167.2)*Urinary urea* (mmol/day)1403304.7 (233.2–389.5)1361305.0 (234.8—389.6)1233306.6 (236.4–389.6)1212306.2 (236.4–388.1)Albumin (µmol/l)2459585.5 (550.7–623.2)2354585.5 (550.7–623.2)2058584.1 (550.7–623.2)1119588.4 (550.7–623.2)C-reactive Protein (mg/l)11973.8 (1.7—7.3)11493.9 (1.7–7.3)10073.9 (1.8–7.4)5203.9 (1.7–7.3)Urate lowering therapy (febuxostat or allopurinol)3024999 (33.0)2774936 (33.7)2344831 (35.5)424 (35.0)Diuretics (all types)30241605 (53.1)27741455 (52.5)23441240 (52.9)1212618 (51.0)Spironolactone3024109 (3.6)277498 (3.5)234487 (3.7)121243 (3.5)Antiplatelet agents30241238 (40.7)27741133 (40.8)2344978 (41.7)1212493 (40.5)Renin-angiotensin inhibitors30242294 (75.9)27742091 (75.4)23441808 (77.1)1212956 (78.9)Medication adherence according to the Girerd score3002275323441212 Good (score equal to 0)1129 (37.6)1028 (37.3)875 (37.3)418 (34.5) Minimal (score equal to 1 or 2)1651 (55.0)1528 (55.5)1313 (56.0)723 (59.7) Poor (score ≥ 3)222 (7.4)197 (7.2)156 (6.7)71 (5.8)Health literacy according to their need for help reading medical documents3033278123441212 Never575 (19.0)519 (18.7)420 (17.9)219 (18.1) Rarely, sometimes, often, or always2458 (81.0)2262 (81.3)1924 (82.1)993 (81.9)Number of nephrological consultations in the year before inclusion2612241420561080 052 (2.0)48 (2.0)41 (2.0)20 (1.9) 1 or 21660 (63.6)1516 (62.8)1277 (62.1)676 (62.6) 3624 (23.9)587 (24.3)519 (25.2)269 (24.9) 4 or more276 (10.6)263 (10.9)219 (10.7)115 (10.6)Number of dietary consultations in the year before inclusion2476229519561017 01862 (75.2)1723 (75.1)1449 (74.1)730 (71.8) 1424 (17.1)394 (17.1)348 (17.8)200 (19.7) 2112 (4.5)105 (4.6)91 (4.7)50 (4.9) 3 or more78 (3.2)73 (3.2)68 (3.5)37 (3.6)Type of center2892270723441212University center1734 (60.0)1609 (59.4)1395 (59.5)764 (63.0)Hospital center577 (20.0)554 (20.5)482 (20.6)260 (21.5)Non profit institution119 (4.1)115 (4.2)113 (4.7)36 (3.0)For-profit institution462 (15.9)429 (15.8)354 (15.1)152 (12.5)N: available data. IQR: InterQuartile Range 25–75.Uric acid in mg/dl to µmol/l, × 59.48.Cardiovascular history defined as patients having coronary artery disease, arrhythmic disorders, congestive heart failure, stroke, peripheral vascular disease and/or valvulopathy. Diabetes defined as patients having diabetes history or antidiabetic treatment or glycated hemoglobin ≥ $6.5\%$ or fasting glycemia ≥ 7 mmol/l or non-fasting glycemia ≥ 11 mmol/l.Hypertension defined as patients having an office blood pressure greater than or equal to $\frac{140}{90}$ mmHg or an antihypertensive treatment.
Over the follow-up (Median 3.2, IQR: 2.6–3.8 years), the 2781 included patients had a median of 5 UA measures, with a median of 120 days between two consecutive measures (Table 2, Fig. S3A). This led to a total of 16 947 measures of UA which were mostly observed during the first three years of follow-up (Figure S3B) and were normally distributed in both men and women (Figure S4). Individual observed values of UA, as well as the true UA trajectory estimated from the linear mixed model, are shown in Figure S5 for some selected patients with extreme UA values. The overall goodness of fit of the linear mixed model is described in Figure S6, which shows how on average the individual predicted values of UA were close to the observed values all along the follow-up. Of the 2781 patients, 439 ($15.8\%$) initiated dialysis ($$n = 375$$) or received a pre-emptive transplant ($$n = 64$$) and 264 ($9.5\%$) died before kidney failure during the follow-up (Table 2).Table 2Description of repeated uric acid measures and outcomes over the follow-up in the population included in our analysis ($$n = 2781$$).Median (IQR 25–75)n (%)Total number of UA measures over follow-up16,947Number of UA measures by patient5 (3–8)Time interval between two consecutive UA measures (days)120 (63—197)Number of patients with Only one UA measure231 (8.3) Two UA measures272 (9.8) Three UA measures323 (11.6) More than three UA measures1955 (70.5)Number of UA measures over follow-up in each of the following UA value classes < 4 mg/dl808 (4.8) 4–6 mg/dl4203 (24.8) 6–8 mg/dl6509 (38.4) 8–10 mg/dl4022 (23.7) > 10 mg/dl1405 (8.3)Number of patients withKidney failure439 ($15.8\%$)Death before kidney failure264 ($9.5\%$)UA uric acid, IQR interquartile range.
Compared to the 3033 patients participating in the CKD-REIN cohort, the 2781 included patients (with UA measurement within six months of inclusion into the cohort) had similar baseline characteristics (Table 1). The 2344 patients used for Models 1 and 3 (with no missing data on adjustment factors) had a higher proportion of ULT use at baseline (Table 1). The 1212 patients used for Model 2 (with further 24-h urine collection at baseline) were more often followed-up in a University Hospital and tended to have more dietary consultations before inclusion (Table 1).
## UA and risk of kidney failure
The hazard of kidney failure increased with increasing cUA, with a plateau for cUA between 6 and 10 mg/dl (Fig. 2A). At any time after inclusion into the cohort, patients with a cUA of precisely 3 mg/dl, had a $59\%$ decreased hazard of kidney failure at that time compared to patients with a cUA of precisely 5 mg/dl at the same time (HR 0.41, $95\%$ confidence interval (CI): 0.31, 0.54, Model 1 in Table 3). The hazard of kidney failure was increased by $70\%$ for patients with a cUA of precisely 11 mg/dl compared to patients with a cUA of precisely 5 mg/dl at the same time (HR 1.70, $95\%$ CI: 1.18, 2.47, Model 1 in Table 3). The association between cUA and the hazard of kidney failure tended to be similar in men and women (p-value for interaction of 0.07). However, an increase in cUA from 3 to 7 mg/dl was associated with a moderately higher increase in the hazard of kidney failure in females than in males. Above 10 mg/dl, the increase in the hazard of kidney failure was detectable in men only (Figure S7), the data being too sparse in women over that range (Figure S4). Further adjustment for salt and protein intake at baseline weakly affected the association in the subsample of the 1212 patients with available information (Model 2 vs. Model 1 in Table 3) but adjusting or not for eating habits in this subsample produced very similar results (Table S2). Finally, the association was much weaker with baseline UA than with cUA (Model 3 in Table 3, Fig. 2B).Figure 2(A) Estimated effect of current uric acid value on the hazard of kidney failure in all patients ($$n = 2344$$ including 382 KRT, Model 1 in Table 3). ( B) Estimated effect of baseline uric acid value on the hazard of kidney failure in all patients ($$n = 2344$$, Model 3 in Table 3). Results from cause-specific Cox models using a spline function for uric acid, adjusted for age, sex, primary kidney disease, hypertension, diabetes, cardiovascular disease, dyslipidemia, body mass index, albuminuria, medication adherence, use of renin-angiotensin system inhibitors and urate lowering therapy, all at baseline. The reference value of uric acid for the HR indicated in the y-axis was arbitrarily chosen at 5 mg/dl, which corresponds to the midpoint of the normal range of uric acid (uric acid in mg/dl to µmol/l: × 59.48). CKD-REIN cohort, France, 2013–2018.Table 3Association between current or baseline value of uric acid and the hazard of kidney failure or death before kidney failure. Results from time-dependent cause-specific Cox models accounting for nonlinear effect uric acid. CKD-REIN cohort, France, 2013–2018.Current value of UA*Model 1 ($$n = 2344$$)Model 2 ($$n = 1212$$)Model 3 ($$n = 2344$$)HR$95\%$ CIHR$95\%$ CIHR$95\%$ CIKidney failure3 mg/dl0.410.31–0.540.360.18–0.690.980.77–1.255 mg/dl1117 mg/dl1.281.15–1.431.160.78–1.721.030.97–1.099 mg/dl1.231.03–1.460.980.64–1.511.060.88–1.2811 mg/dl1.701.18–2.471.641.20–2.241.090.83–1.42Death3 mg/dl1.801.47–2.191.430.93–2.200.920.62–1.385 mg/dl1117 mg/dl0.880.75–1.040.920.89–0.951.111.00–1.239 mg/dl1.250.90–1.761.230.95–1.601.220.94–1.5711 mg/dl2.301.84–2.872.231.73–2.891.230.97–1.56UA, uric acid; HR, hazard ratio; CI, confidence intervals. Uric acid in mg/dl to µmol/l, × 59.48.Model 1: Cox model with UA as a continuous time-dependent variable and adjusted for age, sex, CKD stage, primary kidney disease, hypertension, diabetes, cardiovascular disease, dyslipidemia, body mass index, albuminuria, medication adherence, use of renin-angiotensin system inhibitors and urate lowering therapy, all at baseline. HR of death were further adjusted for spironolactone and antiplatelet agents at baseline. Model 2: Model 1 further adjusted for salt and protein intake at baseline. Model 3: Cox model with UA as a continuous variable measured only at baseline and adjusted for the same factors as Model 1.*The listed values of UA are precise current values since uric acid was taken as a continuous time-dependent covariate in the Cox model. HR of 1.70 for example means that a patient with a current value of uric acid of precisely 11 mg/dl had a $70\%$ increased hazard of kidney failure at that time of follow-up compared to a patient with a value of uric acid of precisely 5 mg/dl at the same time.
## UA and risk of death before kidney failure
The hazard of death before kidney failure had a U-shape relationship with cUA (Fig. 3A), with the lowest mortality for a cUA of 6 mg/dl. At any time after inclusion into the cohort, patients with a cUA of precisely 3 mg/dl, had an $80\%$ increased mortality at that time compared to patients with a cUA of precisely 5 mg/dl at the same time (HR 1.80, $95\%$CI 1.47, 2.19, Model 1 in Table 3). Mortality was twice higher for patients with a cUA of precisely 11 mg/dl compared to patients with a cUA of precisely 5 mg/dl at the same time (HR 2.30, $95\%$CI 1.84, 2.87, Fig. 3A, Model 1 in Table 3). The U-shaped association was similar in men and women (p-value for interaction of 0.68), but the confidence intervals were large in women (Figure S8) because of much less deaths before kidney failure in them (80 vs. 184 in men). As for kidney failure, further adjustment for salt and protein intake at baseline weakly affected the association (Model 2 vs. Model 1 in Table 3) and the results were very similar after adjusting or not for eating habits in this subsample (Table S2). Finally, the association was also much weaker with baseline UA than with cUA (Model 3 in Table 3, Fig. 3B).Figure 3Estimated effect of current uric acid value on the hazard of death before kidney failure, Panel (A) Estimated effect of current uric acid value on the hazard of death before kidney failure adjusted for age, sex, primary kidney disease, hypertension, diabetes, cardiovascular disease, dyslipidemia, body mass index, albuminuria, CKD stage, medication adherence, use of renin-angiotensin system inhibitors, urate lowering therapy, spironolactone and anti-platelet agents, all at baseline (Model 1 in Table 3); Panel (B) Estimated effect of baseline uric acid value on the hazard of death before kidney failure, adjusted for the same factors as in (A) (Model 3 in Table 3). The reference value of uric acid for the HR indicated in the y-axis was arbitrarily chosen at 5 mg/gl, which corresponds to the midpoint of the normal range of uric acid (uric acid in mg/dl to µmol/l: × 59.48). Results from cause-specific Cox models using a spline function for uric acid. CKD-REIN cohort, France, 2013–2018 ($$n = 2344$$, including 218 death before kidney failure).
## Discussion
Using longitudinal data analysis, our results highlight the strong nonlinear association between longitudinal UA and both kidney failure and mortality in CKD patients. After adjustment for major risk factors for CKD progression, the hazard of kidney failure increased with increasing cUA, with a plateau between 6 and 10 mg/dl. By contrast, mortality before kidney failure had a U-shape relationship with cUA, with a minimum for cUA of 6 mg/dl, and a mortality twice higher for cUA of 3 or 11 mg/dl, compared to 5 mg/dl. The association with UA at inclusion was much weaker for both kidney failure and mortality.
Previous experimental studies showed that hyperuricemia may cause and accelerate CKD25,33, by mitochondrial dysfunction34, activation of the renin–angiotensin–aldosterone system35, induction of afferent arteriolar sclerosis36,37, pro-inflammation, or urate crystals deposition in the tubules38,39. Several epidemiologic studies also found that hyperuricemia was associated with CKD progression6,7,9–11,14,40,41, but most of them used baseline UA only, and only two examined the shape of the relationship between baseline UA and kidney failure and death10,14. As us, they found no or weak association between baseline UA and the hazard of kidney failure after adjustment for baseline eGFR. These results for baseline UA contrasted with our results for cUA suggesting a strong increased hazard of kidney failure for cUA above 11 mg/dl. The contrast of results between baseline and current UA value may explain why others studies did not find any association between UA and kidney failure2–5. The stronger association with cUA than with baseline UA may also suggest a stronger short term association than long term association, but this should be further explored using specific lags in the statistical analysis. Furthermore, the moderately stronger increase in the hazard of kidney failure in women than in men, associated with any increase of cUA till 7 mg/dl is consistent with previous studies which found a significant association between UA and CKD progression in women only41.
The U-shape of association between UA and death has already been found in a population of dialysis patients42 or in a Korean population without CKD at baseline43, but this was investigated using baseline UA and not cUA. Similarly, another study found a significant increased all-cause mortality in non-diabetic patients with severe CKD and UA below 5 mg/dl44. Two other studies rather found a J-shape association in non-dialysis CKD patients, with an increased mortality for any baseline UA above 9 mg/dl10 or above 11 mg/dl14, after adjustment for baseline eGFR. A potential explanation of our results suggesting an increased mortality at low cUA values, is that UA is involved in reducing oxidative stress and that a moderate increase in UA is needed to counteract oxidative damage, particularly in the context of arteriosclerosis. Indeed, in patients with atherosclerotic risks, hypouricemia has been associated with a higher hazard of all-cause and cardiovascular mortality45. Similarly, dialysis patients with hypouricemia are also at greater risk of all-cause and cardiovascular death than patients with normal uric acid levels46,47. In contrast, in non-diabetic CKD patients, Lee et al. did not find any association between hypouricemia and cardiovascular mortality, using an unique measure of UA at baseline44. It may therefore be of interest to replicate our analyzes in CKD patients by focusing on cardiovascular mortality. Another explanation for an increased mortality associated with low UA could be that patients with low UA had a poor nutritional status, as in hemodialysis patients48 or in elderly patients49. However, further adjustment for salt and protein intake did not change the magnitude of association between cUA and mortality.
Our study has several strengths. First, the CKD-REIN cohort is a large multicentric prospective cohort, based on a nationally representative sample of nephrology clinics which is likely to enable adequate statistical power and generalizability of our findings to all French patients with CKD under nephrology care15. Second, thanks to frequent measures of UA and the use of advanced statistical methods, we were able to account for changes in UA over time and to estimate the nonlinear association between the reached level of UA and the hazard of death and kidney failure, which was much stronger than with baseline UA only.
However, our study has also limitations. The major limitation is that it is based on observational data, so we cannot exclude residual confounding, despite the use of DAGs to identify appropriate sets of adjusting factors. In particular, we adjusted for diet (salt and protein intake in particular) in a sensitivity analysis only because it was available for only half of patients. The estimated effect of cUA was weaker after adjustment, but this was more likely due to a selection bias in the subsample of half patients, than to a strong confounding bias. Indeed, adjusting or not for salt and protein intake in the subsample produced very similar results. However, it would be important to further investigate the role of diet in general in the association. It would also be of interest to investigate if the association is similar in patients with and without mutations or common variants in UMOD. Another limitation of our study is the use of a two-stage statistical approach instead of a joint analysis of UA trajectory and hazards of kidney failure and death21,50. Such a joint analysis would have accounted for potential informative dropout from the study in the estimation of the current value of UA and for its uncertainty. However, at the time of our analysis, the R packages that allowed the estimation of joint models assumed linear effects of biomarkers and thus did not allow nonlinear effects to be investigated. Because of these limitations and the complex and bidirectional relationship between changes in UA and kidney function during the course of CKD, we acknowledge that our study does not fully elucidate the causal role of UA in CKD progression. This remains all the more unclear that a recent meta-analysis of published placebo-controlled clinical trials51, including the two most recent52,53, found no evidence of benefits of ULT on the risk of kidney failure. These findings, as well as our results confirming the increased mortality for low UA44,46,47,54, question the use of ULT to slow progression of CKD in patients with UA below 10 mg/dl.
To conclude, our study investigating longitudinal UA rather than baseline UA only, indicates a strong nonlinear monotonic association between the reached level of UA and the hazard of kidney failure, and confirms the U-shape relationship with mortality. Although the use of ULT to slow progression of CKD is not recommended, we believe that a UA above 10 mg/dl may be considered as a strong risk marker for kidney failure and death, and thus should encourage nephrologists to be stricter in controlling cardiovascular and nephroprotective factors.
## Supplementary Information
Supplementary Information 1.Supplementary Information 2.Supplementary Information 3.Supplementary Information 4.Supplementary Information 5.Supplementary Information 6.Supplementary Information 7.Supplementary Information 8.Supplementary Information 9.Supplementary Information 10. The online version contains supplementary material available at 10.1038/s41598-023-30902-7.
## References
1. 1.Modification of Diet in Renal Disease Study Group, Hunsicker LG, Adler S, Caggiula A, England BK, Greene T, et al. Predictors of the progression of renal disease in the Modification of Diet in Renal Disease Study. Kidney Int.51(6), 1908–1919 (1997).
2. 2.Sturm G, Kollerits B, Neyer U, Ritz E, Kronenberg F, MMKD Study Group. Uric acid as a risk factor for progression of non-diabetic chronic kidney disease? The mild to Moderate Kidney Disease (MMKD) Study. Exp. Gerontol.43(4), 347–352 (2008).
3. Madero M, Sarnak MJ, Wang X, Greene T, Beck GJ, Kusek JW. **Uric acid and long-term outcomes in CKD**. *Am. J. Kidney Dis.* (2009.0) **53** 796-803. DOI: 10.1053/j.ajkd.2008.12.021
4. Liu WC, Hung CC, Chen SC, Yeh SM, Lin MY, Chiu YW. **Association of hyperuricemia with renal outcomes, cardiovascular disease, and mortality**. *Clin. J. Am. Soc. Nephrol.* (2012.0) **7** 541-548. DOI: 10.2215/CJN.09420911
5. Nacak H, van Diepen M, Qureshi AR, Carrero JJ, Stijnen T, Dekker FW. **Uric acid is not associated with decline in renal function or time to renal replacement therapy initiation in a referred cohort of patients with Stage III, IV and V chronic kidney disease**. *Nephrol. Dial. Transpl.* (2015.0) **30** 2039-2045. DOI: 10.1093/ndt/gfv225
6. Chang WX, Asakawa S, Toyoki D, Nemoto Y, Morimoto C, Tamura Y. **Predictors and the subsequent risk of end-stage renal disease—usefulness of 30% decline in estimated GFR over 2 years**. *PLoS ONE* (2015.0) **10** e0132927. DOI: 10.1371/journal.pone.0132927
7. Uchida S, Chang WX, Ota T, Tamura Y, Shiraishi T, Kumagai T. **Targeting uric acid and the inhibition of progression to end-stage renal disease—a propensity score analysis**. *PLoS ONE* (2015.0) **10** e0145506. DOI: 10.1371/journal.pone.0145506
8. Altemtam N, Russell J, El Nahas M. **A study of the natural history of diabetic kidney disease (DKD)**. *Nephrol Dial Transpl.* (2012.0) **27** 1847-1854. DOI: 10.1093/ndt/gfr561
9. Nacak H, van Diepen M, de Goeij MC, Rotmans JI, Dekker FW. **Uric acid: association with rate of renal function decline and time until start of dialysis in incident pre-dialysis patients**. *BMC Nephrol.* (2014.0) **15** 91. DOI: 10.1186/1471-2369-15-91
10. Srivastava A, Kaze AD, McMullan CJ, Isakova T, Waikar SS. **Uric acid and the risks of kidney failure and death in individuals with CKD**. *Am. J. Kidney Dis.* (2018.0) **71** 362-370. DOI: 10.1053/j.ajkd.2017.08.017
11. Shi Y, Chen W, Jalal D, Li Z, Chen W, Mao H. **Clinical outcome of hyperuricemia in IgA nephropathy: A Retrospective Cohort Study and Randomized Controlled Trial**. *Kidney Blood Press Res.* (2012.0) **35** 153-160. DOI: 10.1159/000331453
12. Zoccali C, Mallamaci F. **Uric acid in chronic kidney disease: The quest for causality continues**. *Nephrol Dial Transpl.* (2018.0) **33** 193-195. DOI: 10.1093/ndt/gfx341
13. Bonino B, Leoncini G, Russo E, Pontremoli R, Viazzi F. **Uric acid in CKD: Has the jury come to the verdict?**. *J. Nephrol.* (2020.0) **33** 715-724. DOI: 10.1007/s40620-020-00702-7
14. Tsai CW, Chiu HT, Huang HC, Ting IW, Yeh HC, Kuo CC. **Uric acid predicts adverse outcomes in chronic kidney disease: a novel insight from trajectory analyses**. *Nephrol. Dial Transpl.* (2018.0) **33** 231-241. DOI: 10.1093/ndt/gfx297
15. Stengel B, Combe C, Jacquelinet C, Briancon S, Fouque D, Laville M. **The French chronic kidney disease-renal epidemiology and information network (CKD-REIN) cohort study**. *Nephrol Dial Transpl.* (2014.0) **29** 1500-1507. DOI: 10.1093/ndt/gft388
16. **Clinical practice guideline for the evaluation and management of chronic kidney disease**. *Kidney Int. Suppl.* (2012.0) **2013** 1-150
17. Ikizler TA, Burrowes JD, Byham-Gray LD, Campbell KL, Carrero JJ, Chan W. **KDOQI clinical practice guideline for nutrition in CKD: 2020 update**. *Am. J. Kidney Dis.* (2020.0) **76** S1-S107. DOI: 10.1053/j.ajkd.2020.05.006
18. Girerd X. **Evaluation de l’observance par l’interrogatoire au cours du suivi des hypertendus dans des consultations spécialisées**. *Arch. Mal. Coeur. Vaiss.* (2001.0) **94** 839-842. PMID: 11575214
19. Chew LD, Griffin JM, Partin MR, Noorbaloochi S, Grill JP, Snyder A. **Validation of screening questions for limited health literacy in a large VA outpatient population**. *J. Gen. Intern. Med.* (2008.0) **23** 561-566. DOI: 10.1007/s11606-008-0520-5
20. Dafni U, Tsiatis A. **Evaluating surrogate markers of clinical outcome when measured with error**. *Biometrics* (1998.0) **54** 1445-1462. DOI: 10.2307/2533670
21. Sweeting MJ, Thompson SG. **Joint modelling of longitudinal and time-to-event data with application to predicting abdominal aortic aneurysm growth and rupture**. *Biom J* (2011.0) **53** 750-763. DOI: 10.1002/bimj.201100052
22. Ye W, Lin X, Taylor JMG. **Semiparametric modeling of longitudinal measurements and time-to-event data—a two-stage regression calibration approach**. *Biometrics* (2008.0) **64** 1238-1246. DOI: 10.1111/j.1541-0420.2007.00983.x
23. Harrell FE, Harrell J, Frank E. **General aspects of fitting regression models**. *Regression Modeling Strategies: With Applications to Linear Models, Logistic and Ordinal Regression, and Survival Analysis* (2015.0) 13-44
24. Suttorp MM, Siegerink B, Jager KJ, Zoccali C, Dekker FW. **Graphical presentation of confounding in directed acyclic graphs**. *Nephrol. Dial. Transpl.* (2015.0) **30** 1418-1423. DOI: 10.1093/ndt/gfu325
25. Sanchez-Lozada LG, Rodriguez-Iturbe B, Kelley EE, Nakagawa T, Madero M, Feig DI. **Uric acid and hypertension: An update with recommendations**. *Am. J. Hypertens.* (2020.0) **33** 583-594. DOI: 10.1093/ajh/hpaa044
26. Hediger MA, Johnson RJ, Miyazaki H, Endou H. **Molecular physiology of urate transport**. *Physiology* (2005.0) **20** 125-133. DOI: 10.1152/physiol.00039.2004
27. Kalantar-Zadeh K, Fouque D. **Nutritional management of chronic kidney disease**. *N. Engl. J. Med.* (2017.0) **377** 1765-1776. DOI: 10.1056/NEJMra1700312
28. Ko GJ, Rhee CM, Kalantar-Zadeh K, Joshi S. **The effects of high-protein diets on kidney health and longevity**. *J. Am. Soc. Nephrol.* (2020.0) **31** 1667-1679. DOI: 10.1681/ASN.2020010028
29. Paul R, Zaihra T. **Interval estimation of risk difference for data sampled from clusters**. *Stat. Med.* (2008.0) **27** 4207-4220. DOI: 10.1002/sim.3289
30. Eilers PHC, Marx BD. **Flexible smoothing with B-splines and penalties**. *Stat. Sci.* (1996.0) **11** 89-121. DOI: 10.1214/ss/1038425655
31. Leffondre K, Jager KJ, Boucquemont J, Stel VS, Heinze G. **Representation of exposures in regression analysis and interpretation of regression coefficients: Basic concepts and pitfalls**. *Nephrol. Dial. Transpl.* (2014.0) **29** 1806-1814. DOI: 10.1093/ndt/gft500
32. 32.R Core Team (2019). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. Available from: https://www.R-project.org/
33. Johnson RJ, Kivlighn SD, Kim YG, Suga S, Fogo AB. **Reappraisal of the pathogenesis and consequences of hyperuricemia in hypertension, cardiovascular disease, and renal disease**. *Am. J. Kidney Dis.* (1999.0) **33** 225-234. DOI: 10.1016/S0272-6386(99)70295-7
34. Sánchez-Lozada LG, Lanaspa MA, Cristóbal-García M, García-Arroyo F, Soto V, Cruz-Robles D. **Uric acid-induced endothelial dysfunction is associated with mitochondrial alterations and decreased intracellular ATP concentrations**. *Nephron. Exp. Nephrol.* (2012.0) **121** e71-78. DOI: 10.1159/000345509
35. Mazzali M, Hughes J, Kim YG, Jefferson A, Kang DH, Gordon KL. **Elevated uric acid increases blood pressure in the rat by a novel crystal-independent mechanism**. *Hypertension* (2001.0) **38** 1101-1106. DOI: 10.1161/hy1101.092839
36. Mazzali M, Kanellis J, Han L, Feng L, Xia YY, Chen Q. **Hyperuricemia induces a primary renal arteriolopathy in rats by a blood pressure-independent mechanism**. *Am. J. Physiol. Renal. Physiol.* (2002.0) **282** F991-997. DOI: 10.1152/ajprenal.00283.2001
37. Sánchez-Lozada LG, Tapia E, Santamaría J, Avila-Casado C, Soto V, Nepomuceno T. **Mild hyperuricemia induces vasoconstriction and maintains glomerular hypertension in normal and remnant kidney rats**. *Kidney Int.* (2005.0) **67** 237-247. DOI: 10.1111/j.1523-1755.2005.00074.x
38. Bjornstad P, Maahs DM, Roncal CA, Snell-Bergeon JK, Shah VN, Milagres T. **Role of bicarbonate supplementation on urine uric acid crystals and diabetic tubulopathy in adults with type 1 diabetes**. *Diabetes Obes. Metab.* (2018.0) **20** 1776-1780. DOI: 10.1111/dom.13274
39. Bjornstad P, Roncal C, Milagres T, Pyle L, Lanaspa MA, Bishop FK. **Hyperfiltration and uricosuria in adolescents with type 1 diabetes**. *Pediatr. Nephrol. Berl. Ger.* (2016.0) **31** 787-793. DOI: 10.1007/s00467-015-3299-8
40. Iseki K, Ikemiya Y, Inoue T, Iseki C, Kinjo K, Takishita S. **Significance of hyperuricemia as a risk factor for developing ESRD in a screened cohort**. *Am. J. Kidney Dis.* (2004.0) **44** 642-650. DOI: 10.1016/S0272-6386(04)00934-5
41. Tsai CW, Lin SY, Kuo CC, Huang CC. **Serum uric acid and progression of kidney disease: A longitudinal analysis and mini-review**. *PLoS ONE* (2017.0) **12** e0170393. DOI: 10.1371/journal.pone.0170393
42. Zawada AM, Carrero JJ, Wolf M, Feuersenger A, Stuard S, Gauly A. **Serum uric acid and mortality risk among hemodialysis patients**. *Kidney Int. Rep.* (2020.0) **5** 1196-1206. DOI: 10.1016/j.ekir.2020.05.021
43. Cho SK, Chang Y, Kim I, Ryu S. **U-shaped association between serum uric acid level and risk of mortality**. *Arthritis Rheumatol.* (2018.0) **70** 1122-1132. DOI: 10.1002/art.40472
44. Lee CL, Tsai SF. **Association between mortality and serum uric acid levels in non-diabetes-related chronic kidney disease: An analysis of the National Health and Nutrition Examination Survey, USA, 1999–2010**. *Sci. Rep.* (2020.0) **10** 17585. DOI: 10.1038/s41598-020-74747-w
45. Cang Y, Xu S, Zhang J, Ju J, Chen Z, Wang K. **Serum uric acid revealed a U-shaped relationship with all-cause mortality and cardiovascular mortality in high atherosclerosis risk patients: The ASSURE study**. *Front. Cardiovasc. Med.* (2021.0) **8** 641513. DOI: 10.3389/fcvm.2021.641513
46. Li Q, Wu C, Kuang W, Zhan X, Zhou J. **Correlation analysis of low-level serum uric acid and cardiovascular events in patients on peritoneal dialysis**. *Int. Urol. Nephrol.* (2021.0) **53** 2399-2408. DOI: 10.1007/s11255-021-02902-x
47. Li M, Ye ZC, Li CM, Zhao WB, Tang H, Liu X. **Low serum uric acid levels increase the risk of all-cause death and cardiovascular death in hemodialysis patients**. *Ren. Fail.* (2020.0) **42** 315-322. DOI: 10.1080/0886022X.2020.1745234
48. Beberashvili I, Erlich A, Azar A, Sinuani I, Feldman L, Gorelik O. **Longitudinal study of serum uric acid, nutritional status, and mortality in maintenance hemodialysis patients**. *Clin. J. Am. Soc. Nephrol.* (2016.0) **11** 1015-1023. DOI: 10.2215/CJN.10400915
49. Tseng WC, Chen YT, Ou SM, Shih CJ, Tarng DC. **U-shaped association between serum uric acid levels with cardiovascular and all-cause mortality in the elderly: The role of malnourishment**. *J. Am. Heart Assoc.* (2018.0) **7** e007523. DOI: 10.1161/JAHA.117.007523
50. Rizopoulos D. *Joint Models for Longitudinal and Time-to-Event Data: With Applications in R* (2012.0)
51. Chen Q, Wang Z, Zhou J, Chen Z, Li Y, Li S. **Effect of urate-lowering therapy on cardiovascular and kidney outcomes: A systematic review and meta-analysis**. *Clin. J. Am. Soc. Nephrol.* (2020.0) **15** 1576-1586. DOI: 10.2215/CJN.05190420
52. Badve SV, Pascoe EM, Tiku A, Boudville N, Brown FG, Cass A. **Effects of allopurinol on the progression of chronic kidney disease**. *N. Engl. J. Med.* (2020.0) **382** 2504-2513. DOI: 10.1056/NEJMoa1915833
53. Doria A, Galecki AT, Spino C, Pop-Busui R, Cherney DZ, Lingvay I. **Serum urate lowering with allopurinol and kidney function in type 1 diabetes**. *N. Engl. J. Med.* (2020.0) **382** 2493-2503. DOI: 10.1056/NEJMoa1916624
54. Perez-Gomez MV, Bartsch LA, Castillo-Rodriguez E, Fernandez-Prado R, Kanbay M, Ortiz A. **Potential dangers of serum urate-lowering therapy**. *Am. J. Med.* (2019.0) **132** 457-467. DOI: 10.1016/j.amjmed.2018.12.010
|
---
title: N-acetylneuraminic acid links immune exhaustion and accelerated memory deficit
in diet-induced obese Alzheimer’s disease mouse model
authors:
- Stefano Suzzi
- Tommaso Croese
- Adi Ravid
- Or Gold
- Abbe R. Clark
- Sedi Medina
- Daniel Kitsberg
- Miriam Adam
- Katherine A. Vernon
- Eva Kohnert
- Inbar Shapira
- Sergey Malitsky
- Maxim Itkin
- Alexander Brandis
- Tevie Mehlman
- Tomer M. Salame
- Sarah P. Colaiuta
- Liora Cahalon
- Michal Slyper
- Anna Greka
- Naomi Habib
- Michal Schwartz
journal: Nature Communications
year: 2023
pmcid: PMC9998639
doi: 10.1038/s41467-023-36759-8
license: CC BY 4.0
---
# N-acetylneuraminic acid links immune exhaustion and accelerated memory deficit in diet-induced obese Alzheimer’s disease mouse model
## Abstract
Systemic immunity supports lifelong brain function. Obesity posits a chronic burden on systemic immunity. Independently, obesity was shown as a risk factor for Alzheimer’s disease (AD). Here we show that high-fat obesogenic diet accelerated recognition-memory impairment in an AD mouse model (5xFAD). In obese 5xFAD mice, hippocampal cells displayed only minor diet-related transcriptional changes, whereas the splenic immune landscape exhibited aging-like CD4+ T-cell deregulation. Following plasma metabolite profiling, we identified free N-acetylneuraminic acid (NANA), the predominant sialic acid, as the metabolite linking recognition-memory impairment to increased splenic immune-suppressive cells in mice. Single-nucleus RNA-sequencing revealed mouse visceral adipose macrophages as a potential source of NANA. In vitro, NANA reduced CD4+ T-cell proliferation, tested in both mouse and human. In vivo, NANA administration to standard diet-fed mice recapitulated high-fat diet effects on CD4+ T cells and accelerated recognition-memory impairment in 5xFAD mice. We suggest that obesity accelerates disease manifestation in a mouse model of AD via systemic immune exhaustion.
Obesity and aging increase Alzheimer’s disease (AD) risk. Here, using an AD mouse model and high-fat diet, we suggest that immune exhaustion links the two risk factors, and identify a metabolite that can hasten immune dysfunction and memory deficit.
## Introduction
Alzheimer’s disease (AD) is the most common form of dementia, characterized by progressive brain amyloidosis and neuronal loss. For decades regarded as neuron-centric, AD pathogenesis is now recognized as being strongly affected by the state of the whole organism1,2. In particular, a healthy immune system is required to support brain maintenance and repair3–11, whereas immune aging increases AD risk12–15. Among the many age-related changes in the immune system is immune exhaustion, marked by the progressive loss of immune effector functions, decrease in cytokine production, and increased inhibitory immune checkpoint signaling16–18. As such, exhausted immune cells have a reduced ability to protect the brain. In line with this contention, immune deficiency as well as immune suppression and exhaustion were linked to disease severity in mouse models of AD, while reducing or blocking immune suppression or exhaustion was found to attenuate pathological manifestations19–25. Based on this cumulative evidence, we envisioned that environmental conditions that are known as strong risk factors for AD might promote disease manifestations by negatively affecting immune functions. One such environmental condition is obesity, which is among the strongest AD risk factors and the most frequent AD comorbidity26–31. Of note, obesity is associated with persistent systemic inflammation and impaired immune responses that have been linked to increased risk of infections and other chronic conditions such as allergy and cancer32.
Here, we specifically focused on diet-induced obesity, and hypothesized that it might affect AD by deregulating systemic immunity. Using a transgenic mouse model carrying five human AD-linked mutations (5xFAD), we found that obesity induced by high-fat diet accelerated the onset of disease manifestations, including recognition-memory impairment, which was associated with increased splenic levels of exhausted CD4+ T effector memory cells, CD4+FOXP3+ regulatory T cells, and increased blood levels of the metabolite N-acetylneuraminic acid. In vitro and in vivo studies revealed that this metabolite could induce immune exhaustion and accelerate recognition-memory impairment in 5xFAD mice.
## Diet-induced obesity accelerated disease manifestations in 5xFAD mice
To test the impact of obesity on AD pathology, we generated a mouse model of obesity-AD comorbidity using 5xFAD mice, a transgenic model of amyloidosis33, that were fed with high-fat obesogenic diet (HFD) or standard control diet (CD). Wild-type (WT) mice fed with either HFD or CD were also included in the study. Both female and male mice were used. Since mid-life obesity increases the risk of AD29–31, we chose a long-term diet regimen starting from around 2 up to 8-9 months of age (mo) for a total of 24-28 weeks (Fig. 1a). Since the metabolic responses to HFD were similar between females and males (Supplementary Fig. 1a–h), we combined both sexes for subsequent analyses. We used the novel object recognition (NOR) test to assay recognition-memory performance (Fig. 1b), known to decline in the 5xFAD model34,35. The NOR test, which allows repeated measures, was used for a longitudinal follow-up to determine if and when the HFD accelerated cognitive impairment in 5xFAD mice. At the age of 6.5 mo, both CD and HFD-fed 5xFAD mice maintained intact performance in the NOR test (Supplementary Fig. 2a, b). At 8 mo, HFD-fed 5xFAD mice showed loss of NOR capability, whereas CD-fed 5xFAD mice still performed similarly to WT controls fed with CD or HFD (Fig. 1c; Supplementary Fig. 2c). HFD did not affect the performance in the NOR test of WT mice at either time point (Supplementary Fig. 2a; Fig. 1c). Locomotor activity was not different between groups (Supplementary Fig. 2d, e), whereas the time spent in the middle of the arena was lowest for HFD-fed WT mice, possibly indicating slightly higher anxiety levels in HFD-fed WT relative to CD-fed WT mice, but not in HFD-fed 5xFAD relative to CD-fed 5xFAD mice (Supplementary Fig. 2f, g).Fig. 1High-fat diet accelerated disease manifestations relative to control diet in 5xFAD mice.a Overview of the experimental strategy. a, c Reduction in novelty discrimination in HFD-fed 5xFAD mice. Evaluation of cognition with the NOR test (b) at 8 mo, 24 weeks of diet (wod; c). Mice from two independent experiments, sample n: WT CD = 13, WT HFD = 16, 5xFAD CD = 15, 5xFAD HFD = 12. Statistical analyses: two-way ANOVA followed by Fisher’s LSD post hoc test. d, Brain regions analyzed for histopathology. e, f Assessment of Aβ plaques load in the hippocampal hilus of 5xFAD mice. g, h Neuronal survival in the subiculum (Sub) of 5xFAD mice. i, j, Glial fibrillary acidic protein (GFAP)-immunoreactivity in the hippocampal hilus of 5xFAD mice. e, g, i Representative images, left: CD, right: HFD, scale bars: 70 μm. f, h, j Mice from two independent experiments, sample n: f, 5xFAD CD = 10, 5xFAD HFD = 10; h, 5xFAD CD = 9, 5xFAD HFD = 10; j, 5xFAD CD = 9, 5xFAD HFD = 9. h Data normalized by average WT CD value (Supplementary Fig. 3b). j Data normalized by average WT CD value (Supplementary Fig. 3c). Statistical analyses: two-tailed unpaired Student’s t test. c, f, h, j Box plots represent the minimum and maximum values (whiskers), the first and third quartiles (box boundaries), the median (box internal line), and the mean (cross). Statistical analyses: two-way ANOVA followed by Fisher’s LSD post hoc test. Source data are provided as a Source Data file.
After the last repeated NOR test, in which we detected that the HFD-fed 5xFAD mice had lower score than the CD-fed 5xFAD mice, we euthanized the mice and assessed if and how the obesogenic diet affected brain pathology. Specifically, we tested typical AD-related hallmarks in the 5xFAD model20–22 (Fig. 1d). No changes in neuronal survival or reactive astrogliosis were noticeable in HFD-fed WT mice compared to CD-fed controls (Supplementary Fig. 3a–c). In 5xFAD mice, the HFD did not affect the level of amyloid β oligomers (Aβ1-42; Supplementary Fig. 3d–f) nor amyloid β plaques (Fig. 1e, f; Supplementary Fig. 3g, h). However, compared to CD-fed 5xFAD mice, the HFD-fed 5xFAD mice displayed increased neuronal loss (subiculum, statistically significant, Fig. 1g, h; cortex layer V, trend, Supplementary Fig. 3i, j) and earlier manifestation of astrogliosis (Fig. 1i, j). Altogether, these results indicate that the HFD accelerated the onset of disease manifestations in 5xFAD mice but did not affect any of the other tested parameters in otherwise healthy age-matched WT mice.
## AD and HFD had largely independent effects on the mouse hippocampal cell fate
To closely analyze the brain’s cellular landscape, we performed single-nucleus RNA-sequencing of the hippocampus (sNuc-Seq, 10x genomics, Methods; Fig. 2a; Supplementary Fig. 4a–f, 5a–g). Briefly, we found effects of both morbidities, most prominently AD-associated alterations in microglia and astrocytes in 5xFAD mice, as previously reported36,37, and an HFD-associated increase in oligodendrocytes that was more prominent in WT mice (Supplementary Fig. 4f). In addition, we found a cluster of cells within the dentate gyrus (DG1) that was specifically over-represented in HFD-fed 5xFAD mice (Fig. 2b, c; Supplementary Fig. 5a). Gene set enrichment analysis of the differentially expressed genes in this cluster highlighted several pathways related to neuronal differentiation, integration, and growth (hypergeometric test, FDR < 0.050, Fig. 2d; Supplementary Data 1), indicative of an immature-like neuronal phenotype. Fig. 2AD and HFD-related changes in the cellular landscape of the mouse hippocampus.a UMAP embedding of 237,631 single nuclei profiles (sNuc-Seq), colored after post hoc cell type annotation. Mice from five independent experiments, sample $$n = 28$.$ For quantifications and statistical analyses, 219,237 nuclei were included from $$n = 26$$ samples: WT CD = 6, WT HFD = 7, 5xFAD CD = 6, 5xFAD HFD = 7 (Methods, Cell fraction estimations and statistics section). CA1–3, cornu Ammonis region 1–3; DG, dentate gyrus; ExN, excitatory neurons; GABA, GABAergic neurons; OPCs, oligodendrocyte precursor cells. b, c Sub-clustering analysis of the DG granule neurons (ExN DG). Sample n: see a. b UMAP embedding of sNuc-Seq profiles colored by cluster. c Changes in frequency of DG1 cluster across experimental conditions; DG2–4 clusters are shown in Supplementary Fig. 5a. d *Pathway analysis* of the genes associated with DG1 showing enrichment of pathways related to neuronal differentiation, integration, and growth (FDR-adjusted hypergeometric test P-value <0.050). e–j Sub-clustering analysis of the brain’s immune cells including microglia (e, f), astrocytes (g, h), and oligodendrocytes (i, j). Sample n: see a. e, g, i UMAP embedding of sNuc-Seq profiles colored by cluster. f, h, j Changes in frequency of cell types across experimental conditions. Abbreviations: AST1–3, astrocyte clusters 1–3; COPs, committed oligodendrocyte precursors; DAAs, disease-associated astrocytes; DAMs, disease-associated microglia; DOLs, disease-associated oligodendrocytes; HMG, homeostatic microglia; OLG1, 2, oligodendrocyte clusters 1, 2; PVMs, perivascular macrophages; RMG, replicating microglia. c, f, h, j Box plots represent the minimum and maximum values (whiskers), the first and third quartiles (box boundaries), the median (box internal line), and the mean (cross). Statistical analyses: two-way ANOVA followed by Fisher’s LSD post hoc test. Source data are provided as a Source Data file. k–m HFD-induced gene expression programs in microglia (k), astrocytes (l), and oligodendrocytes (m) in 5xFAD mice. Volcano plot representation of differentially expressed genes in HFD-fed 5xFAD mice ($$n = 7$$) compared to CD-fed 5xFAD controls ($$n = 6$$). Apoe, downregulated in both microglia and astrocytes, is highlighted in bold. X-axis: average log2 fold change (HFD relative to CD); Y-axis: FDR-adjusted MAST test P-value <0.010 (–log10).
Profiling of the major non-neuronal cell populations (microglia and other immune cell types, astrocytes, and oligodendrocytes) identified a range of glial cell states, including distinct states associated with AD and HFD (Fig. 2e–j). Sub-clustering analysis of microglia revealed the expected AD-associated decrease in homeostatic microglia (HMG, Cx3cr1highP2ry12high cells) and the elevation of disease-associated microglia36 (DAM, Trem2highCd9high cells), but no further HFD-related effects on microglial sub-states and frequencies (Fig. 2e, f). Within the non-microglia immune cell compartment, we detected a small population of perivascular macrophages (PVMs, Mrc1+F13a1+ cells) that significantly decreased with AD (Fig. 2e, f). In addition, we found a small population of T cells (Trbc2+Cd4+ cells; differentially upregulated genes compared to all other immune cell clusters in Supplementary Data 2) that was barely detectable in WT mice, but was found in 5xFAD mice and almost doubled in HFD-fed 5xFAD relative to CD-fed 5xFAD mice (Fig. 2e, f). In the astrocyte compartment, sub-clustering analysis revealed prominent AD-associated effects. In line with a previous report38, we identified two Gfaplow homeostatic states (AST1 and AST2) and two Gfaphigh states, including the previously described disease-associated astrocytes37 (DAAs, Serpina3n+ cells; Fig. 2g, h). AST1 cells (Mfge8highGarem1high) were markedly reduced in 5xFAD mice of both diet groups, concomitantly with the increase in DAAs (Fig. 2g, h). In the oligodendrocyte compartment, we identified AD-associated alterations of distinct cell subsets, such as the decrease in Ptgdshigh mature oligodendrocytes39,40 (OLG1) and the prominent increase in disease-associated oligodendrocytes41,42 (DOLs, C4b+ cells) and committed oligodendrocyte precursors (COPs; Fig. 2i, j).
We next tested the differentially expressed genes (DEGs) across all cell types between HFD-fed and CD-fed 5xFAD mice (Fig. 2k–m). The strongest response was found in astrocytes (Fig. 2l); microglia and oligodendrocytes also showed a differential transcriptional response to HFD, albeit comparably lower DEGs (Fig. 2k, m). Notably, both microglia and astrocytes of HFD-fed 5xFAD mice displayed reduced expression of Apoe, encoding apolipoprotein E (APOE; Fig. 2k, l, bold). In humans, APOE is involved in the clearance of amyloid β and other protein aggregates, and its variant APOE4, which is less efficient in clearing amyloid β, is the major genetic risk factor for AD43,44. While the observed transcriptional response of neuronal and glial cells might contribute to comorbidity, the overall HFD effect on the cellular landscape of the hippocampus did not seem to be robust to the extent that it could explain the accelerated disease progression in HFD-fed 5xFAD mice, which prompted us to search for a possible HFD-induced effect outside the brain.
## HFD induced immune exhaustion in 5xFAD mice
To test our working hypothesis that HFD-induced obesity might accelerate disease manifestations via imposing systemic immune deregulation, we first analyzed the circulating immune cell landscape by mass cytometry (CyTOF), and found that the major effect was a trend of an increase of CD4+ T cells in the HFD-fed group in both WT and 5xFAD mice (Supplementary Fig. 6a–c). In subsequent experiments, we focused on the lymphocyte compartment in the spleen. Of note, deregulation within the splenic CD4+ T-cell compartment in mice was previously linked to aging45–47 and neurodegeneration20,48. Thus, we analyzed freshly isolated splenocytes from mice culled after 28 weeks of diet by flow cytometry (Supplementary Fig. 7a). While CD4– T cells were unaffected (Supplementary Fig. 7b–d), we found shrinkage of the naive CD4+ T-cell population (Fig. 3a) and expansion of CD4+ T effector memory cells (TEMs; Fig. 3b) and CD4+FOXP3+ regulatory T cells (Tregs; Fig. 3c), all features associated with immune aging in mice45–47. Based on these findings, we used mass cytometry (CyTOF) to investigate more closely the profile of the CD4+ T-cell compartment, using mouse splenocytes that were cryopreserved at the end of the NOR follow-up (Supplementary Fig. 8a–c). Sub-clustering analysis of the CD4+CD44high TEM population identified 6 distinct clusters (Fig. 3d–f; Supplementary Fig. 9a). HFD-fed 5xFAD mice displayed the highest frequencies of exhausted TEMs (Cluster 6), characterized by increased expression of exhaustion markers, including the inhibitory immune checkpoint receptors PD-1, TIGIT and LAG-3 and the transcription factors TOX and EOMES, and reduced expression of tested cytokines16–18 (Fig. 3e, f). Furthermore, analysis of the expression of the same exhaustion markers over the entire CD4+ TEM compartment revealed their specific increase in HFD-fed 5xFAD mice (Supplementary Fig. 9b). Overall, the systemic immune system of 5xFAD mice showed augmented susceptibility to the HFD-induced challenges relative to age-matched WT mice, which resulted in increased splenic frequency of immune-suppressive Tregs and exhausted TEMs. Fig. 3Splenic CD4+ T-cell rearrangements in HFD-fed 5xFAD mice.a–c CD4+ T-cell immune deviations in HFD-fed 5xFAD mice. Flow cytometric quantification of splenic frequencies of CD4+ naive T cells (CD44lowCD62Lhigh; a), CD4+ TEMs (CD44highCD62Llow; b), and CD4+FOXP3+ Tregs (c). Mice from two independent experiments, sample n: WT CD = 16, WT HFD = 19, 5xFAD CD = 18, 5xFAD HFD = 18. Statistical analyses: two-way ANOVA followed by Fisher’s LSD post hoc test. d–f Characterization of the CD4+ TEM compartment by mass cytometry. Cryopreserved splenocytes from mice evaluated for both cognition (NOR test; Fig. 1c) and systemic immune phenotype (Fig. 3a–c) were used for CyTOF analysis of the CD4+ TEM compartment, sample n: WT CD = 5, WT HFD = 5, 5xFAD CD = 5, 5xFAD HFD = 5. d UMAP embedding of CD4+ TEM cell clusters (2000 cells, randomly selected from each animal). FlowSOM-based immune cell populations are overlaid as a color dimension. e Mean population expression levels of markers used for UMAP visualization and FlowSOM clustering of CD4+ TEMs. f Increased frequency of exhausted TEMs in HFD-fed 5xFAD mice. Sub-clustering analysis of the CD4+ TEM compartment identified six clusters; Cluster 6 only is shown here, Clusters 1 to 5 are shown in Supplementary Fig. 9a. Statistical analyses: two-way ANOVA followed by Fisher’s LSD post hoc test. a–c, f Box plots represent the minimum and maximum values (whiskers), the first and third quartiles (box boundaries), the median (box internal line), and the mean (cross). Source data are provided as a Source Data file.
## Plasma N-acetylneuraminic acid linked recognition-memory impairment and elevation of splenic Tregs in mice
We hypothesized that diet-induced metabolites might be responsible for the specific effect of HFD on the immune system of 5xFAD mice. To address this question, we carried out plasma metabolite profiling collected from the mice that were evaluated for both NOR and systemic immune phenotype. We identified a total of 229 metabolites (Supplementary Data 3). For each of these metabolites, a Cell Means Model was used to single out the metabolites that most closely associated with each experimental condition (WT:CD, WT:HFD, 5xFAD:CD, and 5xFAD:HFD; Methods). We identified 46 metabolites with significant differential levels between conditions (one-way omnibus ANOVA test P-value <0.050; Fig. 4a). Of these 46 metabolites, 22 were significant (P-value <0.050; Fig. 4a, asterisks) for the 5xFAD:HFD condition, and were further tested for potential association with the NOR discrimination index and the levels of Tregs in the spleen (Fig. 4b), a parameter previously found to associate with disease severity in mouse models of AD20,48. Of these 22 metabolites, free N-acetylneuraminic acid (NANA; Supplementary Fig. 10a, b), the predominant form of sialic acid in mammals49, was not only specifically elevated in the blood of HFD-fed 5xFAD mice (among other metabolites, Fig. 4a; Supplementary Fig. 10c), but it also showed the highest correlation with worsening NOR capability and with increased splenic Tregs frequency (Fig. 4b–d). Using liquid chromatography with tandem mass spectrometry to assess potential changes of free NANA levels in the hippocampus, we found no differences across groups (Supplementary Fig. 10d, e). Validation on a larger number of plasma samples, which included samples from the same animals used for metabolite profiling, confirmed that circulating free NANA was specifically elevated in HFD-fed 5xFAD mice (Fig. 4e). Using this larger cohort of samples, we found that the correlation between NANA levels and NOR discrimination index (Supplementary Fig. 10f) and between NANA levels and splenic Tregs frequency (Supplementary Fig. 10g) confirmed our findings from metabolite profiling. Based on these results, we further focused on NANA as the metabolite potentially driving the accelerated disease manifestations in HFD-fed 5xFAD mice. Fig. 4High plasma levels of free NANA in HFD-fed 5xFAD mice.a–d Metabolite profiling of plasma samples collected from some of the male mice that were evaluated for both cognition (NOR test; Fig. 1c) and systemic immune phenotype (Fig. 3a–c), sample n: WT CD = 4, WT HFD = 5, 5xFAD CD = 5, 5xFAD HFD = 6. a Heatmap representation of the plasma metabolites whose linear regression models had unadjusted one-way omnibus ANOVA test P-value <0.050 (46 out of 229 total identified metabolites). Each column represents one metabolite and each row one sample (mouse). Asterisks indicate the metabolites of interest (22 in total; Methods). The red box highlights the block of metabolites whose overall levels trended highest in HFD-fed 5xFAD mice, which include N-acetylneuraminic acid (NANA; cyan box and green arrowhead). Complete list of identified metabolites, regression coefficients, and exact P-values is provided in Supplementary Data 3, “cell means model” tab. b, ρ-ρ plot (Methods). The red diagonal discriminates between metabolites associated with high NOR discrimination index (DI) and low splenic Tregs abundance (% out of total CD4+ T cells; quadrant II) and metabolites associated with low NOR DI and high splenic Tregs % (quadrant IV). The position of NANA (inset) is indicated by the green arrow. c, d Simple linear regression (black line) and Spearman’s rank correlation (ρ coefficient, two-tailed P-value) between NANA levels, as quantified after plasma metabolite profiling, and NOR discrimination index (DI; c) and splenic Tregs abundance (d); arb. u., arbitrary unit (normalized peak area/100,000). e Quantification of plasma NANA using a fluorometric assay of both female and male mice that were evaluated for both cognition (NOR test; Fig. 1c) and systemic immune phenotype (Fig. 3a–c), also including the same animals described in (a–d), sample n: WT CD = 14, WT HFD = 17, 5xFAD CD = 16, 5xFAD HFD = 14. Statistical analyses: two-way ANOVA followed by Fisher’s LSD post hoc test. Box plots represent the minimum and maximum values (whiskers), the first and third quartiles (box boundaries), the median (box internal line), and the mean (cross). b–e *Source data* are provided as a Source Data file.
## Visceral fat is a potential source of NANA in HFD-fed 5xFAD mice
To identify a putative source of NANA, we studied the cellular landscape of the most likely target organ of the HFD in the periphery: the visceral adipose tissue (VAT). To this end, we analyzed the fate of the gonadal VAT by sNuc-Seq (10x genomics, Methods; Supplementary Fig. 11a–e, 12a–i). The most prominent effect among the four experimental groups was the obesity-driven expansion of macrophages, similarly in both genotype groups (Supplementary Fig. 11a–e; Fig. 5a, b). In particular, sub-clustering analysis revealed a macrophage population (MAC3) that selectively expanded with HFD, characterized by the simultaneous expression of Trem2 and several additional markers consistent with the previously described lipid-associated macrophages50 (LAMs; Supplementary Fig. 12g, h). In contrast, AD alone hardly showed discernible effects. Fig. 5Neu1-expressing macrophages in the mouse visceral adipose tissue are a potential source of NANA.a Cellular landscape of the mouse visceral adipose tissue (VAT) immune cells across all genotype and diet conditions. UMAP embedding of single nuclei profiles (sNuc-Seq), colored after post hoc cell type annotation. Mice from three independent experiments, sample n: WT CD = 10, WT HFD = 6, 5xFAD CD = 9, 5xFAD HFD = 13. b HFD increased the frequency of VAT macrophages (MACs). Changes in frequency of the other VAT immune cell types are shown in Supplementary Fig. 11e. Annotations are as in Supplementary Fig. 11d, e. Sample n: see a. c Dot plots featuring the expression of sialidase *Neu* genes (color scale) and the percentage of cells expressing them (dot size) in the overall VAT (left) and VAT immune compartment (right). Of the four mammalian *Neu* genes, only Neu1 and Neu3 transcripts were detected. d HFD increased the frequency of Neu1-expressing macrophages in the VAT. Sample n: see (a). b, d Statistical analyses: two-way ANOVA followed by Fisher’s LSD post hoc test. Box plots represent the minimum and maximum values (whiskers), the first and third quartiles (box boundaries), the median (box internal line), and the mean (cross). Source data are provided as a Source Data file.
Given the HFD-induced expansion of macrophages in the VAT, we evaluated the expression profiles of the four neuraminidase *Neu* genes51, encoding enzymes known to generate NANA. We found that Neu1 was most abundantly expressed by immune cells, and preferentially by macrophages; in contrast, Neu3 was hardly detectable, whereas Neu2 and Neu4 were below detection level (Fig. 5c). In obese mice, the majority of Neu1+ macrophages were MAC3/LAMs (Supplementary Fig. 12i), suggesting that neuraminidase activity might be an important component in the macrophage response associated with disrupted lipid homeostasis50. Remarkably, Neu1-expressing macrophages were particularly increased in HFD-fed 5xFAD mice (average 3.1-fold relative to CD-fed 5xFAD mice) compared to HFD-fed WT mice (average 1.7-fold relative to CD-fed WT mice; Fig. 5d). Taken together, these findings suggest that the increased proportion of Neu1-expressing macrophages in the VAT combined with the increase of adiposity might account, at least in part, for the specific elevation of circulating NANA levels in HFD-fed 5xFAD mice.
## NANA drove T-cell deregulation in vitro and in vivo and accelerated recognition-memory impairment in 5xFAD mice
Since the major HFD-related changes in 5xFAD mice were in the CD4+ T-cell compartment, we tested if NANA alone could recapitulate this effect, and if so whether it would also lead to an effect on NOR performance. To test the effect of NANA on T cells, we evaluated T-cell proliferation in both mouse splenic T cells (Supplementary Fig. 13a) and peripheral-blood human T cells (Fig. 6a), cultured in vitro in the presence or absence of NANA. In both mice (Supplementary Fig. 13b–g) and humans (Fig. 6b, c; Supplementary Fig. 14a–e), NANA suppressed the proliferation of CD4+, but not CD8+ T cells. In addition, NANA induced elevated levels of PD-1 expression in human CD4+ T cells (Fig. 6d, e). To gain mechanistic insight into the transcriptional programs linking NANA to immune deregulation, we performed bulk RNA-sequencing of the same human T-cell cultures (Supplementary Fig. 14f). We found that NANA significantly upregulated 258 genes and downregulated 484 genes (DESeq2, FDR < 0.050; Methods; Fig. 6f; Supplementary Data 4). *Upregulated* gene set enrichment analysis highlighted several pathways associated with T-cell activity, such as clathrin-dependent endocytosis, T-cell activation, and T-cell differentiation (hypergeometric test, FDR < 0.050, Fig. 6g; Supplementary Data 5). Consistent with the results from T-cell proliferation assays, the top-downregulated pathways were found to be related to cell proliferation (Fig. 6g; Supplementary Data 5). Moreover, we found prominent suppression of genes involved in cell metabolic and bioenergetic pathways, and specifically NAD metabolic process, biosynthesis of amino acids, and glycolysis/gluconeogenesis (Fig. 6g; Supplementary Data 5), thus suggesting that NANA impacts T-cell function, at least in part, via disrupted cell metabolism. Fig. 6NANA induced human T-cell exhaustion in vitro.a–e Effect of NANA on human T cells from peripheral blood cultured in vitro. a Schematic presentation of treatment regimen. b–e Assessment of proliferative ability (b, c) and PD-1 geometric mean fluorescence intensity (gMFI; d, e) in human CD4+ T cells. The shown experiment is one of three independent experiments where different NANA concentrations were tested (Methods, Data reporting section). Sample $$n = 4$$ individuals. From each individual, one aliquot of T cells was treated with NANA, and one with medium as control. Statistical analyses: paired two-tailed Student’s t test. b, d Black lines connect paired points. c, e Histograms of representative samples. f, g Bulk RNA-Seq of the same human T-cell cultures in (a–e). f Heatmap showing significantly (FDR-adjusted DESeq2 P-value <0.050) upregulated (red) and downregulated genes (blue). Each column represents one individual and each row one gene. For each individual, gene expression values are expressed as log-transformed fold-changes (NANA versus medium control). g Bar plot showing significantly (FDR-adjusted hypergeometric test P-value <0.050) upregulated (red) and downregulated (blue) pathways. b–e, g *Source data* are provided as a Source Data file.
To determine whether the increased levels of circulating NANA could contribute to immune rearrangements, we first tested the effects of NANA on the immune system in WT animals. To this end, we used both WT young-adult (6.5-9 mo) and middle-aged (11-14 mo) mice and treated them with repeated administrations of NANA, and subsequently analyzed their splenic T-cell compartment by flow cytometry one day after the last injection (Fig. 7a–c; Supplementary Fig. 15a). In young-adult mice, we did not observe any effect of NANA (CD4+ T cells, Fig. 7b; CD8+ T cells, Supplementary Fig. 15b). In contrast, treatment of middle-aged mice with NANA substantially recapitulated the CD4+ T-cell rearrangements observed in HFD-fed 5xFAD mice (Fig. 7c), whereas the effect on CD8+ T cells was marginal (Supplementary Fig. 15c). Therefore, it appears that NANA induced immune rearrangements in mice that are relatively susceptible to stress factors, as is the case with middle age and neurodegeneration52.Fig. 7NANA administration induced CD4+ T-cell deregulation in vivo and accelerated cognitive decline in 5xFAD mice.a–c Effect of NANA administration on the spleen immune profile of WT mice. a Schematic presentation of treatment regimen. NANA or PBS control was subcutaneously injected twice a day, once in the morning (white arrowheads) and once in the evening (black arrowheads), for 7 consecutive days. b, c Flow cytometric quantification of splenic frequencies of CD4+ naive T cells (CD44lowCD62Lhigh), CD4+ TEMs (CD44highCD62Llow), CD4+FOXP3+CD25+ Tregs, and CD4+PD1+ T cells in young-adult (b) and middle-aged mice (c). For both (b) and (c), data from two independent experiments, sample n: b, PBS = 7, NANA = 7; c, PBS = 7, NANA = 8. Statistical analyses: b, multiple two-tailed unpaired Student’s t tests; c, multiple two-tailed unpaired Student’s t tests with Welch’s correction (Methods, Statistical analyses section). d–f Effect of NANA on novelty discrimination and CD4+ T-cell profile in 5xFAD female mice. d Schematic presentation of treatment regimen. 5xFAD female mice were treated with NANA or PBS control as in (a); PBS-injected age-matched WT female controls were also included. Three weeks after the last injection, novelty discrimination was assessed using the NOR test. e Results of the NOR test. Data from three independent cohorts, age at cognitive assessment: 9, 10, and 11.5 mo, sample n: WT PBS = 7, 5xFAD PBS = 11, 5xFAD NANA = 13. Statistical analyses: one-way ANOVA followed by Fisher’s LSD post hoc test. f Spleen CD4+ T-cell profile. Data from two of the three cohorts described in (e), age at cognitive assessment: 9 and 11.5 mo, sample n: 5xFAD PBS = 6, 5xFAD NANA = 8. Statistical analyses: multiple two-tailed unpaired Student’s t tests. b, c, e, f Box plots represent the minimum and maximum values (whiskers), the first and third quartiles (box boundaries), the median (box internal line), and the mean (cross). Source data are provided as a Source Data file.
The effect of NANA administration on the immune system of middle-aged WT mice encouraged us to test whether administration of NANA to 5xFAD mice would lead to an effect of HFD on NOR performance (Fig. 7d). Using the NOR test 3 weeks after one week of repeated injections of NANA, we found loss of recognition memory at the time when the age-matched PBS-injected 5xFAD control mice hardly showed any impairment as compared to the age-matched PBS-injected WT mice (Fig. 7e; Supplementary Fig. 16a–c). Right after behavior assessment we euthanized the mice and analyzed the fate of their splenic CD4+ T-cell profile (Supplementary Fig. 16d). We found a significant reduction in naive cells and a significant elevation of TEMs in NANA-injected 5xFAD mice relative to PBS-injected 5xFAD controls (Fig. 7f). In one of the mouse cohorts, we also analyzed the profile of CD4+ T cells in the blood and found a stronger effect than in the spleen of the same mice, including a significant increase in circulating FOXP3+CD25+ Tregs and PD-1+ cells (Supplementary Fig. 16e). Of note, whereas the analyses in the WT mice described above were performed one day after the last NANA injection (Fig. 7a), here, in the 5xFAD mice, the immune profiling was performed 3-4 weeks after the last NANA injection (Fig. 7d). In conclusion, NANA administration could affect CD4+ T cells in vivo under conditions of reduced resilience, causing aging-like rearrangements compatible with terminal differentiation, immune-suppression, and metabolic dysfunction, and accelerating deterioration of novelty discrimination capability in neurodegeneration-prone animals.
## Discussion
In the present study, we found that a long-term obesogenic diet regimen accelerated disease manifestations in 5xFAD mice, with no effect on age-matched WT mice. We further show that the primary disease effect on the brain and adipose tissue were due to the neurodegenerative process and the diet, respectively. The comorbidity effect in HFD-fed 5xFAD mice was functionally linked to quantitative and qualitative changes in splenic CD4+ T cells and elevation of plasma levels of free NANA. In vitro, NANA was found to induce CD4+ T-cell deregulation, which was confirmed in independent in vivo experiments performed in mice with relatively low-resilience conditions (5xFAD mice and middle-aged WT mice), but not in young-adult WT mice.
There are several studies supporting the contention that conditions that dampen systemic immunity contribute to the accelerated progression of neurodegenerative diseases. Studies using immune-compromised mice have demonstrated that CD4+ T cells are specifically involved in multiple aspects of brain function, including microglia maturation53, adult hippocampal neurogenesis10, spatial navigation54, and emotional behavior55. Here we found reduced naive cells and increased TEMs and FOXP3+ Tregs in the splenic CD4+ T-cell compartment of HFD-fed 5xFAD mice. These changes mirror those occurring with immune aging in mice and humans45–47,56. Furthermore, we found increased proportions of exhausted CD4+ TEMs. Overall, our results are in line with studies on peripheral blood from human AD patients showing reduced CD4+ naive T cells and increased CD4+ TEMs57,58, and with the recent report of AD-associated increased PD-L1 expression in several circulating T effector-cell subsets, including CD4+ cells59. Findings on the proportions or the immune-suppressive capacity of circulating CD4+ Tregs in human AD patients are contradictory, with studies reporting increase, decrease, or no change58,60–66. Studies in AD mouse models demonstrated that homing of CD4+ Tregs to the brain is associated with disease amelioration20,67. However, the contribution of Tregs to withstanding chronic neurodegeneration might depend on their location (i.e. whether they are in the brain or in the periphery) and disease stage. This could explain the apparent contradiction of independent studies in AD mouse models showing, on the one hand, that early and repeated depletion of *Tregs is* detrimental68,69, whereas on the other hand, late and transient depletion promotes their homing into the brain and disease amelioration20. It is therefore conceivable that the increase of CD4+ Tregs in HFD-fed 5xFAD mice might initially reflect the attempt to restrain peripheral inflammation due to the concomitant expansion of T effector cells55,66. However, in the long term, persistently high levels of CD4+ Tregs in the periphery might interfere with the brain’s ability to recruit reparatory immune cells, such as bone marrow-derived myeloid cells as well as Tregs7,20,67,70–73.
Upon metabolite profiling, we identified free NANA as the diet-associated metabolite in the circulation that displayed the strongest association with both the decline of recognition memory and the elevation of splenic Tregs levels in HFD-fed 5xFAD mice. Since we did not test non-polar metabolites, we cannot rule out the possibility that additional HFD-related metabolites may have effects on disease manifestations. Our data suggest that neuraminidase Neu1-expressing macrophages in the VAT may be a putative source of NANA in obesity. Our findings are in line with previous reports of increased NEU1 enzymatic activity in the VAT of two strains of obese mice74 and attenuated weight gain and VAT inflammation in mice with diet-induced obesity treated with a pan-neuraminidase inhibitor75. Neu1 expression is required during monocyte-to-macrophage differentiation and for macrophage activation and phagocytosis76–78. Therefore, accumulation of Neu1-expressing VAT macrophages in obesity may contribute to chronic local inflammation, which may in turn negatively affect systemic immunity. Although we cannot rule out the possibility that NANA had direct effects on the brain, for example via microglia or astrocytes, as previously reported for other metabolites79,80, the amount of free NANA in the hippocampus was comparable across diet and genotype groups.
Overall, our data suggest that the effects of NANA on the immune cells in the periphery were the driving force of the accelerated disease manifestations in HFD-fed 5xFAD mice. Consistently, CD4+ T cells of both mice and humans were more susceptible to NANA than CD8+ T cells, in agreement with our finding that CD4+ T cells were specifically perturbed in HFD-fed 5xFAD mice. Furthermore, the effects of NANA on cultured human T cells, which encompass augmented expression of PD-1 protein and transcriptional changes associated with T-cell activation/differentiation, metabolic perturbations, and reduced cell proliferation, mirror our findings in HFD-fed 5xFAD mice of increased levels of CD4+ TEMs (i.e. differentiated cells) and expression of exhaustion markers, including PD-1. Of note, the negative impact of NANA might be dependent on some predisposing condition by the host. In line, we found that the immune system of relatively aged WT mice and 5xFAD mice was more vulnerable to the effects of NANA, and that 5xFAD mice treated with NANA displayed accelerated loss of recognition memory.
Taking together our in vivo and in vitro immune profiling results, we suggest that the accelerated disease manifestations in HFD-fed 5xFAD mice are caused by the HFD-accelerated aging of the CD4+ T-cell compartment, at least in part driven by NANA. The aging-like changes in the CD4+ T-cell compartment impair CD4+ T-cell functionality in the periphery56 and, thus, could contribute to the diminished ability of the immune system to support brain homeostasis and to facilitate coping with brain pathology81. In line with this contention are the recent studies demonstrating that aging of the immune system is sufficient to drive aging of non-lymphoid solid organs, including the brain82, and is linked to cognitive decline12; the general anti-aging effects of systemically targeting the immune exhaustion-related molecule PD-183; and several independent studies demonstrating both alleviated brain pathology and improved cognition in different mouse models of neurodegeneration after systemic blockade of the inhibitory immune checkpoints PD-1/PD-L121–24 and, more recently, ERMAP25. Of note, the lack of effect of PD-1 blockade on amyloid β plaques burden in the study of Latta-Mahieu et al. 201884 could be an outcome of the experimental conditions. Intriguingly, elevation of total sialic acid in the circulation has been observed in a wide range of AD comorbidities, such as aging85, obesity86, diabetes87 and cardiovascular disease88, in addition to AD itself89. Thus, we propose that sialic acid-driven systemic immune deregulation may be a risk factor not only in obesity, but generally in pathological states undermining organismic resilience, and especially in compound AD-comorbid states. Therefore, strategies rejuvenating the immune system or therapies targeting diet-induced metabolites like NANA75 may provide salutary benefits for dementia modification regardless of underlying etiology.
## Mice
All experiments detailed herein complied with the regulations formulated by the Institutional Animal Care and Use Committee (IACUC) of the Weizmann Institute of Science (application numbers: 03960618-3, 01200121-2, 03230322-2). Female and male mice were bred and maintained by the Animal Breeding Center of the Weizmann Institute of Science. Housing conditions were: 12-hour dark/light cycle (lights on at 8 am), temperature 22 °C, humidity 30-$70\%$. For comorbidity studies, heterozygous 5xFAD transgenic mice33 (line Tg6799, The Jackson Laboratory) on a C57/BL6-SJL background and age-matched wild-type (WT) controls were used. Genotyping was performed by PCR analysis of ear clipping DNA, as previously described33. Since the C57/BL6-SJL strain carries the retinal degeneration Pde6brd1 mutation, which causes visual impairment in homozygosis (https://www.jax.org/strain/100012), mice were further tested for presence of the allele, as previously described90. To avoid gut microbiota-related cage effects due to coprophagia91, 5xFAD and WT mice were housed together. For the study of the effects of NANA on the immune system in vivo, we used four cohorts of female and male WT mice, and specifically: three cohorts of C57/BL6-SJL mice, age 6.5, 9, and 14 mo; one cohort of C57/BL6 mice, age 11 mo. For the study of the effects of NANA on novelty discrimination, female C57/BL6-SJL 5xFAD and age-matched WT controls were used. To avoid NANA assimilation with coprophagia, NANA-injected mice were housed separately from the PBS-injected controls. All mice were provided with standard chow (calories from proteins: $24\%$; calories from carbohydrates: $58\%$; calories from fat: $18\%$; 2918, Teklad), placed on a hopper integrated with the cage lid, and water ad libitum, and housed in cages enriched with one paper shelter. For comorbidity studies, to induce obesity, at 6-9 weeks of age, mice were switched to a high-fat diet (HFD; calories from proteins: $18\%$; calories from carbohydrates: $22\%$; calories from fat: $60\%$; TD.06414, Teklad), and the food pellet checked twice a week for replenishment. Control mice were kept on standard chow (control diet, CD). Mice allocated for behavioral studies or NANA/PBS injections were switched to a 12-h reversed dark/light cycle (lights on at 8 pm) at least 7 days prior to behavior assessment, and maintained in the regimen until experimental endpoint.
## Glucose tolerance test
Prior to test, mice were fasted overnight for 16 h, and tail blood glucose measured for 0 time point (FreeStyle glucose meter and strips). Subsequently, mice were given 5 μL/g body weight of a 200 mg/mL glucose (Sigma-Aldrich)/water solution intraperitoneally, and blood glucose was measured at 15, 30, 60, 90, and 120 min after glucose injection.
## Novel Object Recognition (NOR) test
Mice that were homozygotes for the Pde6brd1 allele, which causes visual impairment, were not used. A previously published protocol92 was modified as follows. Mice were handled daily for at least 7 days prior to the initiation of NOR test. The NOR test spanned 2 days and included 3 trials: habituation trial, day 1 (20 min session in the empty arena); familiarization trial, day 2 (10 min session in the arena with two identical objects located in opposite corners of the floor, approximately 9 cm from the walls on each side); test trial, day 2, 60-70 min after familiarization (6 min session in the arena with one of the objects replaced by a novel one). Familiar and novel objects were visually and tactually distinct. To control for potential positional preference, the location of the novel object relative to that of the familiar object was randomized. Two identical arenas placed side by side were used simultaneously (one mouse in each arena). Each arena was a 41.5×41.5 cm gray plastic box. Mouse behavior was recorded and measured blindly using the EthoVision XT 11 automated tracking system (Noldus). Novel object preference was measured as discrimination index, expressed as time of interaction with the novel object relative to time of interaction with both objects (%) during the test trial. A discrimination index above $50\%$ indicates novelty recognition, with $50\%$ indicating no recognition. After each trial, the arenas and equipment were wiped with $10\%$ ethanol. Female and male mice were tested on different days. For longitudinal assessment of cognitive performance (comorbidity studies), objects were changed; in addition, to control for potential arena effect, mice tested in one arena at 6.5 mo were tested in the other arena at 8 mo. Each mouse was presented two different novel objects at 6.5 and 8 mo.
## Exogenous NANA administration
Mice were subcutaneously injected with 20 μL/g body weight of a 25 mg/mL NANA (Biosynth-Carbosynth)/PBS solution (pH adjusted to 7.4) twice a day (8-9 h interval) for 7 consecutive days. Control mice were given PBS.
## Mouse blood and tissue collection and processing
In all experiments, sacrifices were carried out in the morning (from 9 am to 1 pm). Mice were anesthetized, exsanguinated by heart puncture, and transcardially perfused with PBS. For plasma metabolite analysis, blood was collected in tubes containing 20 μL heparin and spun at 3,000 g for 15 min at 4 °C. Supernatant plasma was aliquoted and snap frozen in liquid nitrogen (LN). For immune profiling, peripheral blood mononuclear cells (PBMCs) were isolated by density gradient centrifugation protocol (Cytiva™ – density 1.084 ± 0.001 g/ml). After dissection, the following brain regions were excised: for histology, left hemisphere, post-fixed in $4\%$ paraformaldehyde (PFA)/PBS; for sNuc-Seq, left hippocampus, snap frozen in LN; for ELISA and biochemistry, right cortex and right hippocampus, snap frozen in LN. The spleen was mashed with the plunger of a syringe against a 70 mm strainer and treated with ammonium-chloride-potassium (ACK) lysis buffer (Gibco™) to remove erythrocytes. Splenocytes were then used immediately for flow cytometry and pan-T-cell cultures, while for CyTOF, aliquots were resuspended in cell freezing medium (Sigma-Aldrich) and frozen in a Mr. Frosty container (Thermo Fisher) at −80 °C. The left gonadal fat pad was snap frozen in LN for sNuc-Seq.
## Human samples
The study was approved by the Institutional Review Board of the Rambam and Galilee Medical Centers (application number: 0013-20-RMB/66756). All donors were informed on the purpose of the study and gave their consent. Blood was collected from four healthy volunteers (two male and two female subjects). All subjects, average 34 years, were free from acute infectious diseases and in good physical condition. PBMCs were freshly isolated from blood collected in EDTA-coated vials by layering diluted blood (1:1 in PBS) on top of an equal volume of Ficoll (GE Healthcare Life Sciences™ – density 1.077 ± 0.001 g/ml), followed by centrifugation and isolation of the buffy coat.
## Flow cytometry
Splenocytes or PBMCs were washed with ice-cold PBS and stained with LIVE/DEAD™ Fixable Aqua Dead (Thermo Fisher) according to the manufacturer’s instructions. After fragment crystallizable (Fc) blocking (Biolegend), cells were stained for surface antigens. For intranuclear staining, a FOXP3 staining kit was used (Invitrogen) according to the manufacturer’s instructions. The antibodies used, pre-conjugated to fluorophores, are reported in Supplementary Table 1. To assess apoptosis of human lymphocytes following in vitro exposure to NANA, the FITC Annexin V Apoptosis Detection Kit with PI (Biolegend) was used, according to the manufacturer’s instructions. Flow cytometry data were acquired with CytExpert on a CytoFLEX S system (Beckman Coulter) and analyzed using FlowJo software (v10; Tree Star). In each experiment, relevant negative, single-stained, and fluorescence-minus-one controls were used to identify the populations of interest.
## Lymphocyte cultures
Pan-T cells from mouse splenocytes and human PBMCs were isolated by negative selection using magnetic beads (Miltenyi) according to the manufacturer’s instructions. From each mouse or human individual, isolated cell aliquots were plated in at least duplicates for each treatment condition, except in the experiment presented in Supplementary Fig. 13c, where single technical replicates were used. Isolated cells were stained with CellTrace Violet Cell Proliferation Kit (Thermo Fisher) according to the manufacturer’s instructions, and 8×104 cells were plated into 96-well U-bottom and activated with anti-CD3/anti-CD28-coated Dynabeads (Thermo Fisher), as previously described93. Cells were then cultured for 96 h (mouse cell cultures) or 120 h (human cell cultures) in RPMI (BioSource International) supplemented with $5\%$ fetal bovine serum, 5 mM glutamine, 25 mM Hepes (Sigma-Aldrich), and $1\%$ antibiotics (Invitrogen), with or without NANA (pH adjusted to 7.4). Samples were acquired on CytoFLEX S (Beckman Coulter) and analyzed using FlowJo. To assess proliferative ability, the Proliferation Tool of FlowJo was used to estimate the proliferation index, i.e. the total number of cell divisions divided by the number of cells that underwent division.
## Mass cytometry
For intracellular staining, short-term reactivation of cryopreserved splenocytes and subsequent mass cytometry analysis were performed as described previously94. In short, cells were kept at −80 °C for less than 2 months before thawing in a 37 °C water bath. Cells were then immediately resuspended in cell culture medium supplemented with 1:10,000 benzonase (Sigma-Aldrich), and centrifuged at 300 g for 7 min at 24 °C. Samples were then left overnight at 37 °C before restimulation with 50 ng/mL phorbol 12-myristate 13-acetate (Sigma-Aldrich) and 500 ng/mL ionomycin (Sigma-Aldrich) in the presence of 1× brefeldin A (BD Biosciences) and 1× monensin (Biolegend) for 4 h at 37 °C. For splenocytes, one anchor sample to correct batch effect among different acquisitions and one non-stimulated control sample were also included. For both PBMCs and reactivated cryopreserved splenocytes, surface staining was performed for 30 min at 4 °C. To identify dead cells, 2.5 μM cisplatin (Sigma-Aldrich) was added for 5 min on ice. To minimize technical variability, an equal number of cells from each sample were barcoded using Cell-ID 20-Plex (Fluidigm). Cells from all samples were then combined in one single tube. The combined sample was then processed for Live/Dead and surface staining. For intracellular cytokine staining of reactivated cryopreserved splenocytes, after surface staining, cells were fixed and permeabilized with FOXP3 staining kit (Invitrogen) and stained for intracellular markers and cytokines. The antibodies used are reported in Supplementary Table 2. Because CD4 molecules are internalized during phorbol 12-myristate 13-acetate/ionomycin stimulation95, the anti-CD4 antibody was used in both the surface staining and the intracellular staining steps. After washing, the combined sample was incubated with $4\%$ PFA overnight at 4 °C. Prior to acquisition in a Helios™ II CyTOF® system, samples were washed with cell staining buffer and mass cytometry grade water. Multidimensional datasets were analyzed using FlowJo and R (R Core Team, 2017).
## Acquisition and data pre-processing
Quality control and tuning of the Helios™ II CyTOF® system (Fluidigm) was performed on a daily basis and before each acquisition. Samples were acquired in two separate days, and data were normalized using five-element beads (Fluidigm) that were added to the sample immediately before every acquisition, and using an anchor sample included in each reading, as previously described96. For analysis, live single cells were identified based on event length, DNA (191Ir and 193Ir) and live cell (195Pt) channels using FlowJo. Samples were then debarcoded using Boolean gating and mass cytometry data were transformed with an inverse hyperbolic sine (arsinh) function with cofactor 5 using the R environment.
## Algorithm-based high-dimensional analysis
After gating for CD4+ T cells using FlowJo, pre-processed data were considered for the analysis. All FlowSOM-based k-NN clustering was performed on the combined dataset to enable identification of small populations. For CD4+ TEMs, resulting nodes were meta-clustered with the indicated k-values (based on the elbow criterion) and annotated manually. FlowSOM k-NN clustering and two-dimensions UMAP projections were calculated using the CyTOF workflow package (v. 1.2)97.
## Plasma preparation for metabolite profiling
Plasma samples from some of the male mice included in the experiments described in Fig. 1c and in Fig. 3a–c were used. Extraction and analysis of polar metabolites were performed as previously described98,99, with some modifications: 100 μL of plasma were mixed with 1 mL of a pre-cooled (−20 °C) homogenous methanol (MetOH):methyl-tert-butyl-ether (MTBE) 1:3 (v/v) mixture. The tubes were vortexed and then sonicated for 30 min in ice-cold sonication bath (taken for a brief vortex every 10 min). Then, UPLC-grade water (DDW):MetOH (3:1, v/v) solution (0.5 mL) containing internal standards: 13C- and 15N-labeled amino acids standard mix (Sigma-Aldrich) was added to the tubes followed by centrifugation. The upper organic phase was transferred into 2 mL Eppendorf tube. The polar phase was re-extracted as described above, with 0.5 mL of MTBE. Both organic phases were combined and dried in speedvac and then stored at −80 °C until analysis. Lower, polar phase used for polar metabolite analysis was lyophilized, dissolved in 150 μL of 1:1 DDW:MetOH (v/v), centrifuged at 20,000 g for 5 min, transferred to a new tube, and centrifuged again. For analysis, 80 μL were transferred into HPLC vials.
## Human Aβ1-42 ELISA
Samples were homogenized in Tris Buffered Saline EDTA (TBSE) solution (50 mM Tris, 150 mM NaCl, and 2 mM EDTA, pH 7.4) with the addition of $1\%$ Protease Inhibitor Cocktail (Sigma-Aldrich) using a microtube homogenizer with plastic pestles (for hippocampus; 1 mL/100 mg tissue) or a glass homogenizer (for cortex; 1 mL/200 mg tissue). The homogenates were then centrifuged for 40 min at 350,000 g in 500 μL polycarbonate centrifuge tubes (Beckman Coulter) at 4 °C in an Optima MAX-XP Ultracentrifuge with a TLA 120.1 rotor (Beckman Coulter). The supernatant (TBSE-soluble fraction) was collected, aliquoted, and stored at –80 °C until use. BCA assay (Pierce BCA Protein Assay Kit) was performed to determine total protein amount for normalization. To quantify Aβ1-42 peptides, the human Aβ42 Ultrasensitive ELISA Kit (Invitrogen) was used according to the manufacturer’s instructions. Data were acquired using a Spark microplate reader (Tecan).
## LC-MS polar metabolite analysis
Metabolic profiling of polar phase was done as previously described99 with minor modifications described below. Briefly, analysis was performed using Acquity I class UPLC System combined with mass spectrometer Q Exactive Plus Orbitrap™ (Thermo Fisher) that was operated in a negative ionization mode. The LC separation was done using the SeQuant Zic-pHilic (150×2.1 mm) with the SeQuant guard column (20 × 2.1 mm; Merck). The composition of mobile phase B and A was acetonitrile, and 20 mM ammonium carbonate with $0.1\%$ ammonium hydroxide in DDW:acetonitrile (80:20, v/v), respectively. The flow rate was kept at 200 μL/min and gradient as follows: 0–2 min $75\%$ of B, 14 min $25\%$ of B, 18 min $25\%$ of B, 19 min $75\%$ of B, for 4 min.
## Polar metabolite data analysis
The data processing was done using TraceFinder (Thermo Fisher). The detected compounds were identified by accurate mass, retention time, isotope pattern, fragments, and verified using in-house-generated mass spectra library. Peaks were quantified by calculating the area under curve (AUC) and then normalizing the AUC values by internal standards and original sample volume. A total of 229 metabolites were identified.
## Metabolite selection for follow-up studies
For each of the identified metabolite, a linear regression model was built with metabolite plasma levels (scaled values) across all samples of all experimental conditions as the dependent variable, and the interaction between the two categorical “genotype” and “diet” variables (“genotype:diet” term) as the only predictor. To avoid collinearity, the intercept was set to 0. The overall approach is equivalent to a Cell Means Model, which fits an individual mean for each of the predictor levels. The metabolites whose models had unadjusted one-way omnibus ANOVA test P-value <0.050 and P-value for the 5xFAD:HFD condition <0.050 were considered “metabolites of interest”. The metabolites of interest were, in total, 22. For each metabolite of interest, the correlation between the metabolite’s levels and the NOR discrimination index or splenic Tregs abundance across all samples of all genotype:diet conditions was determined using the Spearman’s rank correlation. For each metabolite of interest, the coefficients of the Spearman’s rank correlations (ρ) between metabolite level and NOR discrimination index (DI), and between metabolite level and splenic Tregs abundance, across all samples of all genotype:diet combinations, were calculated and plotted against each other (ρ-ρ plot). The statistics related to the Cell Means Model and correlations, as well as the pairwise comparisons of each metabolite’s abundance across genotype and diet groups, are reported in Supplementary Data 3.
## Measurement of NANA in the hippocampus
NANA was measured in the TBSE-soluble fraction (see Human Aβ1-42 ELISA section) using liquid chromatography with tandem mass spectrometry (LC-MS/MS).
## Materials
Acetonitrile and formic acid of ULC/MS grade were from Bio-Lab (Israel). Water with resistivity 18.2 MΩ was obtained using Direct 3-Q UV system (Millipore). N-acetylneuraminic acid standard (NANA) was purchased from Biosynth-Carbosynth. 13C3-N-acetylneuraminic acid (13C3-NANA) from Omicron Biochemicals was used as internal standard.
## Sample preparation
To 100 μL of TBSE-soluble fraction from mouse hippocampi, 250 μL of ethanol and 10 μL of 10 μg/mL of internal standard were added, and the mixture was incubated in shaker (10 °C, 1,500 rpm, 3 h). The extracts were then centrifuged at 20,000 g for 20 min. The obtained supernatants were evaporated, then re-suspended in 100 μL of $20\%$-aqueous acetonitrile. For LC-MS/MS analysis, the samples were placed in 0.2-μm PTFE-filter vials (Thomson).
## LC-MS analysis
The LC-MS/MS instrument consisted of Acquity I-class UPLC system (Waters) and Xevo TQ-S triple quadrupole mass spectrometer (Waters) was used for the analysis. MassLynx and TargetLynx software (v.4.1, Waters) were applied for the acquisition and analysis of data. Chromatographic separation was done on a 100 ×2.1-mm i.d. 1.7-μm UPLC Atlantis Premier BEH C18 AX column (Waters) with $0.2\%$ formic acid as mobile phase A and $0.2\%$ formic acid in acetonitrile as B at a flow rate of 0.3 mL/min and column temperature 25 °C. A gradient was as follows: 0.5 min the column was hold at $4\%$ B, then linear increase from 1 to $15\%$ B in 1 min, then to $40\%$ B in 3 min, and to $100\%$ B in 0.5 min, and hold at $100\%$ B for 1 min. Then back to $1\%$ B during 0.5 min, and equilibration for 1 min. Samples kept at 8 °C were automatically injected in a volume of 1 μl. The mass spectrometer equipped with an electrospray ion source and operated in negative ion mode was used, with 0.10 mL/min of argon as a collision gas flow. The capillary voltage was set to 1.85 kV, cone voltage 30 V, source offset 12 V, source temperature 120 °C, desolvation temperature 500 °C, desolvation gas flow 600 L/hr, cone gas flow 150 L/hr. Analytes were detected using corresponding multiple reaction monitoring (MRM): for NANA, 308.1 > 170.1 and 308.1 > 87.1 m/z, with collision energies 14 and 15 eV, respectively, and for 13C3-NANA 311.0 > 173.0 and 311.0 > 90.0 m/z with collision energy 30 eV. The retention time of N-Acetylneuraminic acid peak was at 2.03 min. Sample NANA concentrations were calculated from the standard curve of 1–10,000 ng/mL and normalized by total protein amount (see Human Aβ1-42 ELISA section).
## Fluorometric assays on plasma samples
Assays were performed using commercially available kits (Abcam) according to the manufacturer’s instructions, but scaling reaction volumes 1:5 to allow running in 384-well flat-bottom black plates (Greiner). To measure total cholesterol, the cholesterol assay kit – HDL and LDL/VLDL (ab65390) was used. To measure free NANA, the sialic acid (NANA) assay kit (ab83375) was used, and reactions, including background controls, were run for 1 h at RT. Samples were run in duplicates. Data were acquired using an Infinite 200 PRO microplate reader (Tecan) or a Spark microplate reader (Tecan).
## Measurement of plasma leptin
The mouse leptin ELISA kit (Abcam, ab100718) was used according to the manufacturer’s instructions. Samples were run in duplicates. Data were acquired using an Infinite 200 PRO microplate reader (Tecan).
## Brain samples
Hippocampus tissue specimens were kept frozen at −80 °C until processing. Samples were batched in sets of four representing all experimental and sex groups. Working on ice throughout the nuclei isolation process, the frozen hippocampus tissue was transferred into a Dounce homogenizer (Sigma-Aldrich, D8938) with 2 mL of lysis buffer. Lysis buffer used was either EZ Lysis Buffer, as we previously used100 (Sigma-Aldrich, NUC101), or Igepal Lysis Buffer, containing $0.1\%$ IGEPAL® CA-630 (Sigma-Aldrich, I8896), 10 mM Tris HCL pH 7.5, 146 mM NaCl, 3 mM MgCl2, 40 U/mL of RNAse inhibitor (NEB M0314L, as we previously used37). Tissue was gently homogenized while on ice 15 times with pestle A followed by 15 times with pestle B, then transferred to a 15 mL conical tube. A further 3 mL of lysis buffer was added to a final volume of 5 mL, left on ice for 5 min, and then centrifuged in a swing bucket rotor at 500 g for 5 min at 4 °C. Samples were processed two at a time, the supernatant was removed, and the pellets were left on ice while processing the remaining tissues to complete a batch of 4 samples. The nuclei pellets were then resuspended in a wash buffer containing $0.02\%$ BSA (NEB B9000S) and 40 U/mL of RNAse inhibitor in PBS, and the volume adjusted to 5 mL by adding more wash buffer. The nuclei were centrifuged in a swing bucket rotor at 500 g for 5 mins at 4 °C. The supernatant was removed and the pellet was gently resuspended in 500 μL of wash buffer. Nuclei were filtered through a 30 μm MACS Smartstrainer (Miltenyi, 130-098-458) and counted using the LUNA-FL™ Dual Fluorescence Cell Counter (Logos Biosystems) after staining with Acridine Orange/Propidium Iodide Stain (Logos Biosystems, F23001) to differentiate between nuclei and cell debris. Sixteen thousand [16,000] nuclei were run on the 10x Single Cell RNA-Seq Platform using the Chromium Single Cell 3’ Reagent Kits v3. Libraries were made following the manufacturer’s protocol. Briefly, single nuclei were partitioned into nanoliter-scale Gel Bead-In-Emulsion (GEMs) in the Chromium controller instrument, where cDNA shares a common 10x barcode from the bead. Amplified cDNA was measured by Qubit HS DNA assay (Thermo Fisher, Q32851) and quality assessed by High Sensitivity D5000 ScreenTape [5067- 5592] with High Sensitivity D5000 Reagents [5067- 5593] on the 2200 TapeStation system (Agilent). The WTA (whole transcriptome amplified) material was diluted to <8 ng/mL and processed through v3 library construction according to the manufacturer’s protocol, and resulting libraries were quantified again by Qubit and TapeStation. Libraries from 4 channels were pooled and sequenced on 1 lane of NextSeq 550 (or 8 channels sequenced on two NextSeq 550 runs) at the Center for Genomic Technologies in the Institute of Life Sciences at The Hebrew University of Jerusalem, for a target coverage of around 150 million reads per channel.
## Visceral adipose tissue samples
Visceral adipose tissue (VAT) specimens were kept frozen at −80 °C until processing. Nuclei were isolated with a previously published protocol using salt-Tris (ST)-based buffers101. Frozen VAT was placed into a gentleMACSTM C tube (Miltenyi, 130-096-334) with 1 mL TST buffer and homogenized using the gentleMACSTM Octo dissociator (Miltenyi). The sample was removed and, with a further 1 mL of TST, transferred to a 15 mL conical tube on ice for 10 min. Next, the sample was filtered through a 40 μm FalconTM cell strainer (Thermo Fisher, 08-771-1), to which 3 mL of 1x ST buffer was also added through the filter. The resulting 5 mL sample was then centrifuged at 500 g for 5 min at 4 °C in a swinging bucket centrifuge, following which the supernatant was removed and the pellet resuspended in 1x ST buffer (volume determined by pellet size). The nuclei-containing solution was transferred to a 5 mL polystyrene tube through a 35 μm cell strainer cap (Thermo Fisher, 08-771-23), and nuclei were counted using a C-chip disposable hemocytometer (VWR, 82030-468). Eight thousand [8,000] single nuclei were loaded into each channel of the Chromium single cell 3’ chip, for V3 10x technology (10x Genomics). Single nuclei were partitioned into GEMs and incubated to generate barcoded cDNA by reverse transcription. Barcoded cDNA was next amplified by PCR prior to library construction. Libraries of paired-end constructs were generated using fragmentation, sample index and adaptor ligation, and PCR was run according to the manufacturer’s recommendations (10x Genomics). Libraries from four 10x channels were pooled together and sequenced on one lane of an Illumina HiSeq X (Illumina) by the Genomics Platform of the Broad Institute.
## Quality controls for sequencing and pre-processing of sNuc-Seq data
De-multiplexing of samples after Illumina sequencing was done using 10x Cellranger version 5.0.0 mkfastq to generate a Fastq file for each sample. Alignment to the mm10 transcriptome and unique molecular identifier (UMI)-collapsing were performed using the Cellranger count (version 5.0.0, mm10-2020-A_premrna transcriptome, single cell 3’ chemistry). Separate Fastq files of the same mouse sample were combined by running the CellRanger count with multiple fastqs input parameters. Since nuclear RNA includes roughly equal proportions of intronic and exonic reads, we built and aligned reads to a genome reference with pre-mRNA annotations, which account for both exons and introns.
## Technical artifacts and ambient RNA correction
To account for technical artifacts in the data, specifically correcting gene counts shifted due to ambient RNA, we ran the CellBender102 (version 2) program on each sample, which removes counts due to ambient RNA molecules and random barcode swapping from (raw) UMI-based single cell or nucleus RNA-Seq count matrices, and also determines which cell barcodes are valid nuclei libraries, excluding empty droplets and low-quality libraries. We used the CellBender output as input to downstream analysis.
## Data normalization
For every nucleus, we quantified the number of genes for which at least one read was mapped, and then excluded all nuclei with fewer than 100 detected genes. For visceral adipose tissue (VAT) data, nuclei with more than 5000 genes were also filtered out. Genes that were detected in fewer than 3 nuclei were excluded. Expression values Ei,j for gene i in cell j were calculated by dividing UMI counts for gene i by the sum of the UMI counts in nucleus j, to normalize for differences in coverage, and then multiplying by 10,000 to create TPM-like values, and finally computing log2(TP10K + 1) using the NormalizeData function from the Seurat package103–106 (version 4).
## Doublet detection
We annotated each nucleus with a doublet score – the nucleus’ probability of being a doublet, related to the fraction of artificially generated doublet neighbors (using an in-house optimization of DoubletFinder107 with the following parameters: PCs = 1-45, pN = 0.25, pK = 150/(#cells), pANN = False and sct = False). This score would later be considered for the removal of doublets. We first used a high-resolution clustering (1.3 for the hippocampus and 1.5 for the VAT, see description under Dimensionality reduction and clustering). We excluded clusters that had more than $50\%$ of cells that had over a high doublet score (0.35 for the hippocampus and 0.4 for the adipose tissue). Second, cells from other clusters that had over a high doublet score were excluded. In the VAT: 10,625 doublets were removed and 275,336 nuclei remained in the dataset. In the hippocampus: we excluded from the analysis cells from clusters classified as endothelial cells or OPCs, since the doublet detection failed for these cell types. For OPCs, these specifically include cells differentiating from oligodendrocytes. At the end of this stage in the hippocampus dataset, 38,060 doublets were removed and 269,503 nuclei remained in the data set (for $$n = 28$$ mice, across all mouse genotype and diet groups). The downstream analysis of sub-clustering of specific cell types included a second inspection for doublets.
## Identification of variable genes and scaling the data matrix
After data pre-processing, samples from the four conditions within each batch were merged into a single Seurat object. For each batch, variable genes were selected by using a variance-stabilizing transformation (using FindVariableFeatures method, Seurat). This method, first, fits a line to the relationship of log(variance) and log(mean) of each gene from the non-normalized data, using local polynomial regression (LOESS); next, it standardizes the feature values using the observed mean and expected variance (given by the fitted line). Feature variance is then calculated on the standardized values and the genes with the top values were selected as variable genes for downstream analysis (2,500 for the hippocampus and 2,000 for the adipose tissue). For sub-clustering analysis of macrophages in the adipose tissue, genes in the ambient RNA signature were removed from the variable gene list to prevent clustering based on ambient RNA expression. The data were scaled, yielding the relative expression of each variable gene by scaling and centering (using ScaleData method, Seurat).
## Data integration
After the identification of variable genes per batch, the batches (7 for the hippocampus and 2 for the VAT) were integrated into a single Seurat object (using Seurat v.4 integration workflow103,104), based on the CCA algorithm, using the methods FindIntegrationAnchors followed by IntegrateData. The integrated data matrix was then used for dimensionality reduction and clustering.
## Dimensionality reduction, clustering, and visualization
The integrated data matrix (restricted to the genes chosen as integration anchors) was then used for dimensionality reduction, visualization and clustering. Dimensionality reduction was done with principal component analysis (PCA, using RunPCA method Seurat). After PCA, significant principal components (PCs) were identified using the elbow method, plotting the distribution of standard deviation of each PC (ElbowPlot in Seurat). In the VAT analysis: 30 PCs were used. In the hippocampus analysis: 45 PCs for analysis of all cells, 20 PCs for astrocytes, 20 for microglia, and 10 for oligodendrocytes. Within the top PC space, transcriptionally similar nuclei were clustered together using a graph-based clustering approach. First, a k-nearest neighbor (k-NN) graph is constructed based on the Euclidean distance. For any two nuclei, edge weights were refined by the shared overlap of the local neighborhoods using Jaccard similarity (FindNeighbors method Seurat, with $k = 60$). Next, nuclei were clustered using the Louvain algorithm108 which iteratively grouped nuclei and located communities in the input k-NN graph (FindClusters method Seurat, with resolution 0.5). Note that for the doublet detection stage on all cell types, we first used 45 PCs with a higher resolution clustering of 1.3 on data matrices that were merged based on the batch (see Doublet detection section). The obtained clusters were hierarchically clustered and re-ordered (using BuildClusterTree method Seurat). For visualization, the dimensionality of the datasets was further reduced by UMAP, using the same top principal components as input to the algorithm (using the RunUMAP method Seurat). Note that the distribution of samples within each cluster was examined to eliminate that clusters were driven by batch or other technical effects. Clusters with low-quality cells (low number of genes detected, and missing or low-key cell-type marker genes and house-keeping genes such as Malat1), doublet clusters expressing markers of multiple cell types, and neuronal clusters from neighboring region of the hippocampal subiculum that appeared in an uneven form across samples, were removed from the analysis, leaving the hippocampus dataset with 237,631 (for $$n = 28$$ mice, across all mouse genotype and diet groups). Data visualization using UMAP showed that the clusters displayed a mixture of nuclei from all technical and biological replicates, with a variable number of genes, meaning the clustering was not driven by a technical effect.
## Sub-clustering analysis of cell types
Specific cell types (i.e. microglia, astrocytes, oligodendrocytes, DG neurons, macrophages in the adipose tissue) were subsetted from the main dataset for a high-resolution analysis. For each such subset another cycle of clean-up was performed, removing doublet clusters based on different thresholds. Cells were clustered in high-resolution and clusters were then annotated and merged based on marker expression.
## Identification of clusters’ cell types
Identification of cell types was done in the hippocampus using an in-house modification of a logistic regression model (linear_model. LogisticRegression from Python’s sklearn package). The modifications included calculating the classification probability of each cell, and eventually associating the cell with the original cluster, and classifying the cluster as the overall highest scoring cell type. The classifier was trained on $80\%$ of the nuclei (~50,000 nuclei) of our previously published and annotated sNuq-*Seq data* of the mouse hippocampus37, and tested on the remaining $20\%$ of the nuclei. Cell types were annotated according to the classification and further validated using known marker gene and as previously published37. In the VAT, identification of cell types was done based on known marker genes and the Bioconductor package SingleR109.
## Cell fraction estimations and statistics
The fraction of different cell populations (i.e. clusters) was separately computed, for each sample across all clusters, as the fraction of nuclei in each cluster out of the total number of nuclei, by a parameter of interest (e.g. diet, sex, and more). Correlations between these fractions of interest were calculated using Spearman’s rank correlation coefficient (with cor function from R’s base package, method = “spearman”). For quantifications and statistical analyses, 219,237 nuclei were included from $$n = 26$$ samples: WT CD = 6, WT HFD = 7, 5xFAD CD = 6, 5xFAD HFD = 7. Two mice of the original $$n = 28$$ samples were excluded from the statistics due to technical artifacts that hampered their annotation to a specific member of one of the experimental groups (mice identity numbers: 636 and 640). To assess whether there was a significant change between experimental groups across conditions (diet, genotype), two-way ANOVA followed by Fisher’s Least Significant Difference (LSD) post hoc test was used (see Statistical analyses section).
## Marker genes and differential expression analysis
*Marker* genes for glial and dentate gyrus (DG) granule neurons sub-clusters were found using the MAST test110 (using batch and sex as latent variables for glial cells in the brain to account for technical effects), which was run using the FindMarkers function in Seurat v.4 (using the assay set to RNA to use the normalized UMI counts values per gene). To find diet-dependent differentially expressed signatures, we used the same scheme as above using the MAST algorithms with batch and sex as latent variables, within each glial cell type, comparing all cells of HFD-fed and CD-fed 5xFAD mice; similarly, WT mice were compared between diet groups. P-values were adjusted for multiple hypothesis testing using Benjamini–Hochberg’s correction (FDR). Adjusted P-value threshold of 0.010 was used to report significant changes and a fold change threshold of 0.250.
## Bulk RNA-Seq library preparation and analysis
Cultured human T cells after treatment were flash frozen in Buffer RLT (Qiagen, 79216) and kept in −80 °C until processing. Total RNA was extracted with the NucleoSpin RNA kit (Macherey-Nagel, 740955). One microgram (1 μg) total RNA was used as an input for mRNA isolation (NEB E7490S), followed by library preparation using NEBNext® Ultra™ II Directional RNA Library Prep Kit for Illumina (NEB E7760). Quantification of the libraries was done by Qubit and TapeStation. Paired-end sequencing of the libraries was performed on Nextseq 550. De-multiplexing of samples was done with Illumina’s bcl2fastq software. The fastq files were next aligned to the human genome (hg38) using STAR111 and the transcriptome alignment and gene counts were obtained with HTseq112. *Aligned* gene counts were normalized to fragments per kilobase of transcript per million mapped reads (log10(FPKM)) using the fpkm function in DESeq2 R package113 (1.32.0). Genes with low counts (<10) were filtered before performing statistical analysis. Differential expression analysis of bulk RNA-*Seq data* between NANA-treated samples and control was performed using the DESeq2 R package (P-value adjusted to multiple hypothesis testing ≤0.050), while accounting for inter-sample differences. Significantly upregulated genes (adjusted P-value <0.050, log2-fold change >0.250) and downregulated genes (adjusted P-value <0.050, log2-fold change < –0.250) were determined. Up- and downregulated gene lists were separately functionally annotated with gene sets defined by the KEGG and Gene Ontology databases (org.Hs.eg.db, version 3.5.0), using enrichGO and enrichKEGG functions in R package clusterProfiler (3.6.0), using the hypergeometric P-value and FDR correction for multiple hypothesis (with a threshold of FDR < 0.050).
## Histology and immunohistochemistry
Paraffin-embedded tissue was sectioned with a thickness of 6 μm. One slide per animal was used for staining, each containing 5 equally spaced sections. Cresyl violet staining was performed to visualize neurons22. Antibody staining was performed as previously described114, except that all primary antibodies were incubated overnight at RT, followed by another overnight at 4 °C. For Aβ staining, the Mouse On Mouse detection kit (Vector labs) was used according to manufacturer’s instructions. The following primary antibodies were used: mouse anti-human Aβ (1:150; Covance); chicken anti-GFAP (1:150; Abcam). Cy2/Cy3-conjugated anti-mouse/chicken secondary antibodies (1:150; Jackson Immunoresearch) were used. For counterstaining, 4’,6-diamidino-2-phenylindole (1:5000; Biolegend) was used.
## Microscopic imaging and analysis
Images were acquired using a fluorescence microscope (E800, Nikon) equipped with a digital camera (DXM 1200 F, Nikon), and with a ×20 NA 0.50 objective lens (Plan Fluor, Nikon). Quantitative analyses were performed by an experimenter blind to the identity of the animals, and using either the Image-Pro Plus software (Media Cybernetics) or ImageJ (NIH). Neuronal survival on cresyl violet-stained sections and Aβ plaques quantification were performed as previously described22. GFAP intensity was measured using the ImageJ software by applying a segmentation algorithm to mask stained areas (Otsu’s method) and subsequently measuring average integrated density over 3-5 sections per animal. For each animal, stained sections’ quantified values were averaged. Representative images were optimized using ImageJ and processed equally for all experimental conditions displayed.
## Experimental design
No statistical method was used to predetermine sample sizes, which were chosen with adequate statistical power based on the literature and past experience21,22,67,115. The specific sample sizes and tests used to analyze each set of experiments are indicated in the Figure legends. Animals were randomly allocated to experimental groups balancing sex and genotype. Sample selection for subsequent analyses such as CyTOF, sNuc-Seq, and metabolite profiling was based on behavioral (NOR test), metabolic, flow-cytometric, and/or histological analyses. For in vitro experiments with both mouse and human lymphocytes, subjects were not divided in groups, but from each subject (mouse or human) one aliquot of cells was treated with NANA and another one with medium as control. Investigators were blind to animal identity during experiments and outcome assessment, except during behavior experiments in comorbidity studies, where diet groups, but not genotypes, were obvious. For in vitro experiments with both mouse and human lymphocytes, blinding was not relevant, as data were acquired under the same conditions using the software described in the Methods.
## Data reporting
Mice of both sexes were included in all experiments, except in the following cases: only male mice were used for plasma metabolite profiling (Fig. 4a–d and Supplementary Fig. 10a–c); only female mice were used for cognitive assessment and immune profiling following NANA/PBS administration to 5xFAD mice and age-matched WT controls (Fig. 7d–f and Supplementary Fig. 16a–e). When animals from different cohorts/experiments were merged for presentation, the number of cohorts/experiments considered is indicated in the Figure legends. For comorbidity studies (Figs. 1–5, Supplementary Figs. 1–12), the data herein presented originated from five independent cohorts. Animals used for the mass cytometry analysis of the blood immune profile (Supplementary Fig. 6a–c) were selected from the first cohort; animals used for the sNuc-Seq of the hippocampus (Fig. 2a–m, Supplementary Figs. 4 and 5) were selected from all five cohorts; animals used for the sNuc-Seq of the VAT (Fig. 5a–d, Supplementary Figs. 11 and 12) were selected from the first three cohorts; animals used for all other analyses were selected from the last two cohorts. For the study on the effects of NANA administration on the spleen immune profile of young-adult and middle-aged WT mice (Fig. 7a–c and Supplementary Fig. 15a–c), four independent experiments were conducted: two with young-adult mice (6.5-9 mo), and two with middle-aged (11-14 mo) mice. For the study of the effects of NANA administration on novelty discrimination (NOR test; Fig. 7d, e and Supplementary Fig. 16a–c), four independent experiments were conducted using 9-12-mo female mice. Of these four experiments, one could not be included due to complete loss of object discrimination in the PBS-injected 5xFAD controls at the time when the animals were tested (12 mo), which could not allow us to detect further exacerbation by the treatment. No animal was excluded from analyses, except those removed before experimental endpoint according to IACUC guidelines, or because of technical reasons detailed as follows: for microscopic image analysis, poorly stained or overstained sections or slides were not included; for flow and mass cytometry, samples with not enough cells to proceed with the analysis or samples in which the staining did not work were not included; for the analysis of the effects of NANA administration on the spleen immune profile of middle-aged WT mice (Fig. 7c and Supplementary Fig. 15c), one PBS-injected animal was not included due to its statistics across most immune cell populations (1.5x InterQuartile Range method); for the analysis of the effects of NANA administration on novelty discrimination (NOR test; Fig. 7e and Supplementary Fig. 16a–c), one PBS-injected WT mouse was not included in all tested behavioral parameters due to freezing during the test trial of the NOR assay; for the analysis of the effects of NANA administration on the spleen CD4+ T-cell profile of 5xFAD mice (Fig. 7f), one PBS-injected mouse was not included due to its statistics across all immune cell populations (1.5x InterQuartile Range method). For in vitro studies with mouse lymphocytes, two independent experiments were conducted: one simultaneously testing two concentrations of NANA, 1 and 5 mg/mL (Supplementary Fig. 13c); and one testing 1 mg/mL of NANA. For in vitro studies with human lymphocytes, we performed one pilot study using 10 and 25 mg/mL of NANA. We observed that 25 mg/mL of NANA suppressed both CD4+ and CD8+ T-cell proliferation, whereas 10 mg/mL of NANA had no effect. Since the data from mouse studies in vivo (HFD-fed 5xFAD mice, Fig. 3 and Supplementary Fig. 7; NANA-injected mice, Fig. 7a–c and Supplementary Fig. 15) and in vitro (NANA-treated pan-T-cell cultures, Supplementary Fig. 13) suggested a specific impact on CD4+ T cells, with only a minor effect on CD8+ T cells, in a subsequent single experiment we used an intermediate concentration of 12.5 mg/mL of NANA (Fig. 6 and Supplementary Fig. 14).
## Statistical analyses
The normality of data distribution was evaluated using Anderson-Darling’s, D’Agostino-Pearson’s (“omnibus K2”), Shapiro-Wilk’s, and Kolmogorov-Smirnov’s tests, and via visual assessment of quantile-quantile plots. Homogeneity of variance was tested using F-test, for two groups, and Spearman’s test for heteroscedasticity, followed by visual assessment of the homoscedasticity plot, for more than two groups. Data were analyzed using two-tailed Student’s t test to compare between two groups; Welch’s correction was applied for heteroscedastic groups. For lymphocyte cultures, paired Student’s t test or one-way within-subjects ANOVA followed by Fisher’s LSD post hoc test were used. For comorbidity studies, in the presence of two categorical independent variables (“genotype”, two levels: “WT” and “5xFAD”; “diet”, two levels: “CD” and “HFD”), two-way ANOVA followed by Fisher’s LSD post hoc test was used. For weight gain and glucose tolerance test, two-way within-subjects ANOVA followed by Fisher’s LSD post hoc test was used. For the analysis of novelty discrimination and locomotor activity/anxiety of NANA-injected 5xFAD mice and PBS-injected 5xFAD and WT controls, one-way ANOVA followed by Fisher’s LSD post hoc test was used. For null hypothesis testing, the test statistics with confidence intervals, degrees of freedom, and P-values are reported in the Source Data file. For Student’s t tests and post hoc tests, P-values <0.060 are reported in the graphs, rounded to three decimal digits; P-value <0.050 was considered significant. Statistical analyses were carried out using GraphPad Prism version 9.0, R, and Microsoft Excel. Graphs were generated with GraphPad Prism version 9.0 and R.
## Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.
## Supplementary information
Supplementary Information Description of Additional Supplementary Files Supplementary Data 1 Supplementary Data 2 Supplementary Data 3 Supplementary Data 4 Supplementary Data 5 Peer Review File Reporting Summary The online version contains supplementary material available at 10.1038/s41467-023-36759-8.
## Source data
Source Data
## Peer review information
Nature Communications thanks Chotima Boettcher and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Peer reviewer reports are available.
## References
1. Barnes DE, Yaffe K. **The projected effect of risk factor reduction on Alzheimer’s disease prevalence**. *Lancet. Neurol.* (2011.0) **10** 819-828. DOI: 10.1016/S1474-4422(11)70072-2
2. Norton S, Matthews FE, Barnes DE, Yaffe K, Brayne C. **Potential for primary prevention of Alzheimer’s disease: an analysis of population-based data**. *Lancet. Neurol.* (2014.0) **13** 788-794. DOI: 10.1016/S1474-4422(14)70136-X
3. Filiano AJ. **Unexpected role of interferon-γ in regulating neuronal connectivity and social behaviour**. *Nature* (2016.0) **535** 425-429. DOI: 10.1038/nature18626
4. Ziv Y. **Immune cells contribute to the maintenance of neurogenesis and spatial learning abilities in adulthood**. *Nat. Neurosci.* (2006.0) **9** 268-275. DOI: 10.1038/nn1629
5. Shi Y, Holtzman DM. **Interplay between innate immunity and Alzheimer disease: APOE and TREM2 in the spotlight**. *Nat. Rev. Immunol.* (2018.0) **18** 759-772. DOI: 10.1038/s41577-018-0051-1
6. Beers DR, Henkel JS, Zhao W, Wang J, Appel SH. **CD4+ T cells support glial neuroprotection, slow disease progression, and modify glial morphology in an animal model of inherited ALS**. *Proc. Natl. Acad. Sci. USA* (2008.0) **105** 15558-15563. DOI: 10.1073/pnas.0807419105
7. Shechter R. **Recruitment of beneficial M2 macrophages to injured spinal cord is orchestrated by remote brain choroid plexus**. *Immunity* (2013.0) **38** 555-569. DOI: 10.1016/j.immuni.2013.02.012
8. Kipnis J, Avidan H, Caspi RR, Schwartz M. **Dual effect of CD4+CD25+ regulatory T cells in neurodegeneration: a dialogue with microglia**. *Proc. Natl. Acad. Sci. USA* (2004.0) **101** 14663-14669. DOI: 10.1073/pnas.0404842101
9. Ron-Harel N. **Age-dependent spatial memory loss can be partially restored by immune activation**. *Rejuvenation Res.* (2008.0) **11** 903-913. DOI: 10.1089/rej.2008.0755
10. Wolf SA. **CD4-positive T lymphocytes provide a neuroimmunological link in the control of adult hippocampal neurogenesis**. *J. Immunol* (2009.0) **182** 3979-3984. DOI: 10.4049/jimmunol.0801218
11. Derecki NC. **Regulation of learning and memory by meningeal immunity: a key role for IL-4**. *J. Exp. Med.* (2010.0) **207** 1067-1080. DOI: 10.1084/jem.20091419
12. Minhas PS. **Restoring metabolism of myeloid cells reverses cognitive decline in ageing**. *Nature* (2021.0) **590** 122-128. DOI: 10.1038/s41586-020-03160-0
13. Zhao Y, Zhan J-K, Liu Y. **A perspective on roles played by immunosenescence in the pathobiology of Alzheimer’s disease**. *Aging Dis.* (2020.0) **11** 1594-1607. DOI: 10.14336/AD.2020.0205
14. Lutshumba J, Nikolajczyk BS, Bachstetter AD. **Dysregulation of systemic immunity in aging and dementia**. *Front. Cell. Neurosci.* (2021.0) **15** 652111. DOI: 10.3389/fncel.2021.652111
15. Da Mesquita S. **Functional aspects of meningeal lymphatics in ageing and Alzheimer’s disease**. *Nature* (2018.0) **560** 185-191. DOI: 10.1038/s41586-018-0368-8
16. Mogilenko DA, Shchukina I, Artyomov MN. **Immune ageing at single-cell resolution**. *Nat. Rev. Immunol.* (2022.0) **22** 484-498. DOI: 10.1038/s41577-021-00646-4
17. Miggelbrink AM. **CD4 T-cell exhaustion: does it exist and what are its roles in cancer?**. *Clin. Cancer Res.* (2021.0) **27** 5742-5752. DOI: 10.1158/1078-0432.CCR-21-0206
18. McLane LM. **Role of nuclear localization in the regulation and function of T-bet and Eomes in exhausted CD8 T cells**. *Cell Rep.* (2021.0) **35** 109120. DOI: 10.1016/j.celrep.2021.109120
19. Marsh SE. **The adaptive immune system restrains Alzheimer’s disease pathogenesis by modulating microglial function**. *Proc. Natl. Acad. Sci. USA* (2016.0) **113** E1316-E1325. DOI: 10.1073/pnas.1525466113
20. Baruch K. **Breaking immune tolerance by targeting Foxp3(+) regulatory T cells mitigates Alzheimer’s disease pathology**. *Nat. Commun.* (2015.0) **6** 7967. DOI: 10.1038/ncomms8967
21. Baruch K. **PD-1 immune checkpoint blockade reduces pathology and improves memory in mouse models of Alzheimer’s disease**. *Nat. Med.* (2016.0) **22** 135-137. DOI: 10.1038/nm.4022
22. Rosenzweig N. **PD-1/PD-L1 checkpoint blockade harnesses monocyte-derived macrophages to combat cognitive impairment in a tauopathy mouse model**. *Nat. Commun.* (2019.0) **10** 465. DOI: 10.1038/s41467-019-08352-5
23. Xing Z. **Influenza vaccine combined with moderate-dose PD1 blockade reduces amyloid-β accumulation and improves cognition in APP/PS1 mice**. *Brain. Behav. Immun.* (2021.0) **91** 128-141. DOI: 10.1016/j.bbi.2020.09.015
24. Zou Y. **Programmed cell death protein 1 blockade reduces glycogen synthase kinase 3β activity and tau hyperphosphorylation in Alzheimer’s disease mouse models**. *Front. Cell Dev. Biol.* (2021.0) **9** 769229. DOI: 10.3389/fcell.2021.769229
25. Liu H, Zhao J, Lin Y, Su M, Lai L. **Administration of anti-ERMAP antibody ameliorates Alzheimer’s disease in mice**. *J. Neuroinflammation* (2021.0) **18** 268. DOI: 10.1186/s12974-021-02320-x
26. Alford S, Patel D, Perakakis N, Mantzoros CS. **Obesity as a risk factor for Alzheimer’s disease: weighing the evidence**. *Obes. Rev.* (2018.0) **19** 269-280. DOI: 10.1111/obr.12629
27. Dake MD. **Obesity and brain vulnerability in normal and abnormal aging: a multimodal MRI study**. *J. Alzheimer’s Dis. Rep.* (2021.0) **5** 65-77. DOI: 10.3233/ADR-200267
28. O’Brien PD, Hinder LM, Callaghan BC, Feldman EL. **Neurological consequences of obesity**. *Lancet. Neurol.* (2017.0) **16** 465-477. DOI: 10.1016/S1474-4422(17)30084-4
29. Chuang Y-F. **Midlife adiposity predicts earlier onset of Alzheimer’s dementia, neuropathology and presymptomatic cerebral amyloid accumulation**. *Mol. Psychiatry* (2016.0) **21** 910-915. DOI: 10.1038/mp.2015.129
30. Singh-Manoux A. **Obesity trajectories and risk of dementia: 28 years of follow-up in the Whitehall II Study**. *Alzheimers. Dement.* (2018.0) **14** 178-186. DOI: 10.1016/j.jalz.2017.06.2637
31. Dahl AK, Hassing LB. **Obesity and cognitive aging**. *Epidemiol. Rev.* (2013.0) **35** 22-32. DOI: 10.1093/epirev/mxs002
32. Andersen CJ, Murphy KE, Fernandez ML. **Impact of obesity and metabolic syndrome on immunity**. *Adv. Nutr.* (2016.0) **7** 66-75. DOI: 10.3945/an.115.010207
33. Oakley H. **Intraneuronal beta-amyloid aggregates, neurodegeneration, and neuron loss in transgenic mice with five familial Alzheimer’s disease mutations: potential factors in amyloid plaque formation**. *J. Neurosci.* (2006.0) **26** 10129-10140. DOI: 10.1523/JNEUROSCI.1202-06.2006
34. Giannoni P. **Early administration of RS 67333, a specific 5-HT4 receptor agonist, prevents amyloidogenesis and behavioral deficits in the 5XFAD mouse model of Alzheimer’s disease**. *Front. Aging Neurosci.* (2013.0) **5** 96. DOI: 10.3389/fnagi.2013.00096
35. Frydman-Marom A. **Orally administrated cinnamon extract reduces β-amyloid oligomerization and corrects cognitive impairment in Alzheimer’s disease animal models**. *PLoS ONE* (2011.0) **6** e16564-e16564. DOI: 10.1371/journal.pone.0016564
36. Keren-Shaul H. **A unique microglia type associated with restricting development of Alzheimer’s disease**. *Cell* (2017.0) **169** 1276-1290.e17. DOI: 10.1016/j.cell.2017.05.018
37. Habib N. **Disease-associated astrocytes in Alzheimer’s disease and aging**. *Nat. Neurosci.* (2020.0) **23** 701-706. DOI: 10.1038/s41593-020-0624-8
38. Zeisel A. **Molecular architecture of the mouse nervous system**. *Cell* (2018.0) **174** 999-1014.e22. DOI: 10.1016/j.cell.2018.06.021
39. Floriddia EM. **Distinct oligodendrocyte populations have spatial preference and different responses to spinal cord injury**. *Nat. Commun.* (2020.0) **11** 5860. DOI: 10.1038/s41467-020-19453-x
40. Marques S. **Oligodendrocyte heterogeneity in the mouse juvenile and adult central nervous system**. *Science* (2016.0) **352** 1326-1329. DOI: 10.1126/science.aaf6463
41. Zhou Y. **Human and mouse single-nucleus transcriptomics reveal TREM2-dependent and TREM2-independent cellular responses in Alzheimer’s disease**. *Nat. Med.* (2020.0) **26** 131-142. DOI: 10.1038/s41591-019-0695-9
42. Kenigsbuch M. **A shared disease-associated oligodendrocyte signature among multiple CNS pathologies**. *Nat. Neurosci.* (2022.0) **25** 876-886. DOI: 10.1038/s41593-022-01104-7
43. Castellano JM. **Human apoE isoforms differentially regulate brain amyloid-β peptide clearance**. *Sci. Transl. Med.* (2011.0) **3** 89ra57. DOI: 10.1126/scitranslmed.3002156
44. Zhao N. **APOE4 exacerbates α-synuclein pathology and related toxicity independent of amyloid**. *Sci. Transl. Med.* (2020.0) **12** eaay1809. DOI: 10.1126/scitranslmed.aay1809
45. Han G-M, Zhao B, Jeyaseelan S, Feng J-M. **Age-associated parallel increase of Foxp3(+)CD4(+) regulatory and CD44(+)CD4(+) memory T cells in SJL/J mice**. *Cell. Immunol.* (2009.0) **258** 188-196. DOI: 10.1016/j.cellimm.2009.05.003
46. Elyahu Y. **Aging promotes reorganization of the CD4 T cell landscape toward extreme regulatory and effector phenotypes**. *Sci. Adv.* (2019.0) **5** eaaw8330. DOI: 10.1126/sciadv.aaw8330
47. Harpaz I, Bhattacharya U, Elyahu Y, Strominger I, Monsonego A. **Old mice accumulate activated effector CD4 T cells refractory to regulatory T cell-induced immunosuppression**. *Front. Immunol.* (2017.0) **8** 283. DOI: 10.3389/fimmu.2017.00283
48. Di Benedetto G. **Beneficial effects of curtailing immune susceptibility in an Alzheimer’s disease model**. *J. Neuroinflammation* (2019.0) **16** 166. DOI: 10.1186/s12974-019-1554-9
49. Rawal P, Zhao L. **Sialometabolism in brain health and Alzheimer’s disease**. *Front. Neurosci.* (2021.0) **15** 308. DOI: 10.3389/fnins.2021.648617
50. Jaitin DA. **Lipid-associated macrophages control metabolic homeostasis in a Trem2-dependent manner**. *Cell* (2019.0) **178** 686-698.e14. DOI: 10.1016/j.cell.2019.05.054
51. Miyagi T, Yamaguchi K. **Mammalian sialidases: physiological and pathological roles in cellular functions**. *Glycobiology* (2012.0) **22** 880-896. DOI: 10.1093/glycob/cws057
52. Gonzales MM. **Biological aging processes underlying cognitive decline and neurodegenerative disease**. *J. Clin. Invest.* (2022.0) **132** e158453. DOI: 10.1172/JCI158453
53. Pasciuto E. **Microglia require CD4 T cells to complete the fetal-to-adult transition**. *Cell* (2020.0) **182** 625-640.e24. DOI: 10.1016/j.cell.2020.06.026
54. Radjavi A, Smirnov I, Kipnis J. **Brain antigen-reactive CD4+ T cells are sufficient to support learning behavior in mice with limited T cell repertoire**. *Brain. Behav. Immun.* (2014.0) **35** 58-63. DOI: 10.1016/j.bbi.2013.08.013
55. Evans FL, Dittmer M, de la Fuente AG, Fitzgerald DC. **Protective and regenerative roles of T cells in central nervous system disorders**. *Front. Immunol.* (2019.0) **10** 2171. DOI: 10.3389/fimmu.2019.02171
56. Lefebvre JS, Haynes L. **Aging of the CD4 T cell compartment**. *Open Longev. Sci.* (2012.0) **6** 83-91. DOI: 10.2174/1876326X01206010083
57. Larbi A. **Dramatic shifts in circulating CD4 but not CD8 T cell subsets in mild Alzheimer’s disease**. *J. Alzheimers. Dis.* (2009.0) **17** 91-103. DOI: 10.3233/JAD-2009-1015
58. Pellicanò M. **Immune profiling of Alzheimer patients**. *J. Neuroimmunol.* (2012.0) **242** 52-59. DOI: 10.1016/j.jneuroim.2011.11.005
59. Wu C-T. **A change of PD-1/PD-L1 expression on peripheral T cell subsets correlates with the different stages of Alzheimer’s disease**. *Cell Biosci* (2022.0) **12** 162. DOI: 10.1186/s13578-022-00897-1
60. Faridar A. **Restoring regulatory T-cell dysfunction in Alzheimer’s disease through ex vivo expansion**. *Brain Commun.* (2020.0) **2** fcaa112. DOI: 10.1093/braincomms/fcaa112
61. Oberstein TJ. **Imbalance of circulating T(h)17 and regulatory T cells in Alzheimer’s disease: a case control study**. *Front. Immunol.* (2018.0) **9** 1213. DOI: 10.3389/fimmu.2018.01213
62. Ciccocioppo F. **The characterization of regulatory T-cell profiles in Alzheimer’s disease and multiple sclerosis**. *Sci. Rep.* (2019.0) **9** 8788. DOI: 10.1038/s41598-019-45433-3
63. Saresella M. **PD1 negative and PD1 positive CD4+ T regulatory cells in mild cognitive impairment and Alzheimer’s disease**. *J. Alzheimers. Dis.* (2010.0) **21** 927-938. DOI: 10.3233/JAD-2010-091696
64. Fu J. **Mild cognitive impairment patients have higher regulatory T-cell proportions compared with Alzheimer’s disease-related dementia patients**. *Front. Aging Neurosci.* (2020.0) **12** 624304. DOI: 10.3389/fnagi.2020.624304
65. Le Page A. **Differential phenotypes of myeloid-derived suppressor and T regulatory cells and cytokine levels in amnestic mild cognitive impairment subjects compared to mild Alzheimer diseased patients**. *Front. Immunol.* (2017.0) **8** 783. DOI: 10.3389/fimmu.2017.00783
66. Rosenkranz D. **Higher frequency of regulatory T cells in the elderly and increased suppressive activity in neurodegeneration**. *J. Neuroimmunol.* (2007.0) **188** 117-127. DOI: 10.1016/j.jneuroim.2007.05.011
67. Ben-Yehuda H. **Key role of the CCR2-CCL2 axis in disease modification in a mouse model of tauopathy**. *Mol. Neurodegener.* (2021.0) **16** 39. DOI: 10.1186/s13024-021-00458-z
68. Dansokho C. **Regulatory T cells delay disease progression in Alzheimer-like pathology**. *Brain* (2016.0) **139** 1237-1251. DOI: 10.1093/brain/awv408
69. Baek H. **Neuroprotective effects of CD4+CD25+Foxp3+ regulatory T cells in a 3xTg-AD Alzheimer’s disease model**. *Oncotarget* (2016.0) **7** 69347-69357. DOI: 10.18632/oncotarget.12469
70. Butovsky O, Kunis G, Koronyo-Hamaoui M, Schwartz M. **Selective ablation of bone marrow-derived dendritic cells increases amyloid plaques in a mouse Alzheimer’s disease model**. *Eur. J. Neurosci.* (2007.0) **26** 413-416. DOI: 10.1111/j.1460-9568.2007.05652.x
71. Koronyo Y. **Therapeutic effects of glatiramer acetate and grafted CD115**. *Brain* (2015.0) **138** 2399-2422. DOI: 10.1093/brain/awv150
72. Raposo C. **CNS repair requires both effector and regulatory T cells with distinct temporal and spatial profiles**. *J. Neurosci.* (2014.0) **34** 10141 LP-10110155. DOI: 10.1523/JNEUROSCI.0076-14.2014
73. Ito M. **Brain regulatory T cells suppress astrogliosis and potentiate neurological recovery**. *Nature* (2019.0) **565** 246-250. DOI: 10.1038/s41586-018-0824-5
74. Natori Y, Ohkura N, Nasui M, Atsumi G, Kihara-Negishi F. **Acidic sialidase activity is aberrant in obese and diabetic mice**. *Biol. Pharm. Bull.* (2013.0) **36** 1027-1031. DOI: 10.1248/bpb.b12-00995
75. Pilling D, Karhadkar TR, Gomer RH. **High-fat diet-induced adipose tissue and liver inflammation and steatosis in mice are reduced by inhibiting sialidases**. *Am. J. Pathol.* (2021.0) **191** 131-143. DOI: 10.1016/j.ajpath.2020.09.011
76. Liang F. **Monocyte differentiation up-regulates the expression of the lysosomal sialidase, Neu1, and triggers its targeting to the plasma membrane via major histocompatibility complex class II-positive compartments***. *J. Biol. Chem.* (2006.0) **281** 27526-27538. DOI: 10.1074/jbc.M605633200
77. Amith SR. **Dependence of pathogen molecule-induced Toll-like receptor activation and cell function on Neu1 sialidase**. *Glycoconj. J.* (2009.0) **26** 1197. DOI: 10.1007/s10719-009-9239-8
78. Seyrantepe V. **Regulation of phagocytosis in macrophages by neuraminidase 1**. *J. Biol. Chem.* (2010.0) **285** 206-215. DOI: 10.1074/jbc.M109.055475
79. Mayo L. **Regulation of astrocyte activation by glycolipids drives chronic CNS inflammation**. *Nat. Med.* (2014.0) **20** 1147-1156. DOI: 10.1038/nm.3681
80. Rothhammer V. **Microglial control of astrocytes in response to microbial metabolites**. *Nature* (2018.0) **557** 724-728. DOI: 10.1038/s41586-018-0119-x
81. Kipnis J, Gadani S, Derecki NC. **Pro-cognitive properties of T cells**. *Nat. Rev. Immunol.* (2012.0) **12** 663-669. DOI: 10.1038/nri3280
82. Yousefzadeh MJ. **An aged immune system drives senescence and ageing of solid organs**. *Nature* (2021.0) **594** 100-105. DOI: 10.1038/s41586-021-03547-7
83. Wang T-W. **Blocking PD-L1–PD-1 improves senescence surveillance and ageing phenotypes**. *Nature* (2022.0) **611** 358-364. DOI: 10.1038/s41586-022-05388-4
84. Latta-Mahieu M. **Systemic immune-checkpoint blockade with anti-PD1 antibodies does not alter cerebral amyloid-β burden in several amyloid transgenic mouse models**. *Glia* (2018.0) **66** 492-504. DOI: 10.1002/glia.23260
85. Mehdi MM, Singh P, Rizvi SI. **Erythrocyte sialic acid content during aging in humans: correlation with markers of oxidative stress**. *Dis. Markers* (2012.0) **32** 293429. DOI: 10.1155/2012/293429
86. Rajappa M. **Relationship of raised serum total and protein bound sialic acid levels with hyperinsulinemia and indices of insulin sensitivity and insulin resistance in non-diabetic normotensive obese subjects**. *Diabetes Metab. Syndr.* (2013.0) **7** 17-19. DOI: 10.1016/j.dsx.2013.02.030
87. K P. **Predictive value of serum sialic Acid in type-2 diabetes mellitus and its complication (nephropathy)**. *J. Clin. Diagn. Res.* (2013.0) **7** 2435-2437. PMID: 24392365
88. Sillanaukee P, Pönniö M, Jääskeläinen IP. **Occurrence of sialic acids in healthy humans and different disorders**. *Eur. J. Clin. Invest.* (1999.0) **29** 413-425. DOI: 10.1046/j.1365-2362.1999.00485.x
89. Yadav J. **Sialic acid associated with oxidative stress and total antioxidant capacity (TAC) expression level as a predictive indicator in moderate to severe Alzheimer’s disease**. *Exp. Gerontol.* (2020.0) **141** 111092. DOI: 10.1016/j.exger.2020.111092
90. Giménez E, Montoliu L. **A simple polymerase chain reaction assay for genotyping the retinal degeneration mutation (Pdeb(rd1)) in FVB/N-derived transgenic mice**. *Lab. Anim.* (2001.0) **35** 153-156. DOI: 10.1258/0023677011911525
91. Campbell JH. **Host genetic and environmental effects on mouse intestinal microbiota**. *ISME J* (2012.0) **6** 2033-2044. DOI: 10.1038/ismej.2012.54
92. Bevins RA, Besheer J. **Object recognition in rats and mice: a one-trial non-matching-to-sample learning task to study ‘recognition memory’**. *Nat. Protoc.* (2006.0) **1** 1306-1311. DOI: 10.1038/nprot.2006.205
93. Trickett A, Kwan YL. **T cell stimulation and expansion using anti-CD3/CD28 beads**. *J. Immunol. Methods* (2003.0) **275** 251-255. DOI: 10.1016/S0022-1759(03)00010-3
94. Galli E. **GM-CSF and CXCR4 define a T helper cell signature in multiple sclerosis**. *Nat. Med.* (2019.0) **25** 1290-1300. DOI: 10.1038/s41591-019-0521-4
95. Bigby M, Wang P, Fierro JF, Sy MS. **Phorbol myristate acetate-induced down-modulation of CD4 is dependent on calmodulin and intracellular calcium**. *J. Immunol* (1990.0) **144** 3111-3116. DOI: 10.4049/jimmunol.144.8.3111
96. Van Gassen S, Gaudilliere B, Angst MS, Saeys Y, Aghaeepour N. **CytoNorm: a normalization algorithm for cytometry data**. *Cytometry. A* (2020.0) **97** 268-278. DOI: 10.1002/cyto.a.23904
97. Nowicka M. **CyTOF workflow: differential discovery in high-throughput high-dimensional cytometry datasets**. *F1000Research* (2017.0) **6** 748. DOI: 10.12688/f1000research.11622.1
98. Malitsky S. **Viral infection of the marine alga Emiliania huxleyi triggers lipidome remodeling and induces the production of highly saturated triacylglycerol**. *New Phytol* (2016.0) **210** 88-96. DOI: 10.1111/nph.13852
99. Zheng L. **Fumarate induces redox-dependent senescence by modifying glutathione metabolism**. *Nat. Commun.* (2015.0) **6** 6001. DOI: 10.1038/ncomms7001
100. Habib N. **Massively parallel single-nucleus RNA-seq with DroNc-seq**. *Nat. Methods* (2017.0) **14** 955-958. DOI: 10.1038/nmeth.4407
101. Slyper M. **A single-cell and single-nucleus RNA-Seq toolbox for fresh and frozen human tumors**. *Nat. Med.* (2020.0) **26** 792-802. DOI: 10.1038/s41591-020-0844-1
102. 102.Fleming, S. J., Marioni, J. C. & Babadi, M. CellBender remove-background: a deep generative model for unsupervised removal of background noise from scRNA-seq datasets. bioRxiv 791699 10.1101/791699 (2019).
103. Stuart T. **Comprehensive Integration of Single-**. *Cell Data Cell* (2019.0) **177** 1888-1902.e21. PMID: 31178118
104. Butler A, Hoffman P, Smibert P, Papalexi E, Satija R. **Integrating single-cell transcriptomic data across different conditions, technologies, and species**. *Nat. Biotechnol.* (2018.0) **36** 411-420. DOI: 10.1038/nbt.4096
105. Satija R, Farrell JA, Gennert D, Schier AF, Regev A. **Spatial reconstruction of single-cell gene expression data**. *Nat. Biotechnol.* (2015.0) **33** 495-502. DOI: 10.1038/nbt.3192
106. Hao Y. **Integrated analysis of multimodal single-cell data**. *Cell* (2021.0) **184** 3573-3587.e29. DOI: 10.1016/j.cell.2021.04.048
107. 107.McGinnis, C. S., Murrow, L. M. & Gartner, Z. J. DoubletFinder: Doublet Detection in Single-Cell RNA Sequencing Data Using Artificial Nearest Neighbors. Cell Syst.8, 329–337.e4 (2019).
108. Waltman L, van Eck NJ. **A smart local moving algorithm for large-scale modularity-based community detection**. *Eur. Phys. J. B* (2013.0) **86** 471. DOI: 10.1140/epjb/e2013-40829-0
109. Aran D. **Reference-based analysis of lung single-cell sequencing reveals a transitional profibrotic macrophage**. *Nat. Immunol.* (2019.0) **20** 163-172. DOI: 10.1038/s41590-018-0276-y
110. Finak G. **MAST: a flexible statistical framework for assessing transcriptional changes and characterizing heterogeneity in single-cell RNA sequencing data**. *Genome Biol.* (2015.0) **16** 278. DOI: 10.1186/s13059-015-0844-5
111. Dobin A. **STAR: ultrafast universal RNA-seq aligner**. *Bioinformatics* (2013.0) **29** 15-21. DOI: 10.1093/bioinformatics/bts635
112. Anders S, Pyl PT, Huber W. **HTSeq—a Python framework to work with high-throughput sequencing data**. *Bioinformatics* (2015.0) **31** 166-169. DOI: 10.1093/bioinformatics/btu638
113. 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
114. Ziv Y, Avidan H, Pluchino S, Martino G, Schwartz M. **Synergy between immune cells and adult neural stem/progenitor cells promotes functional recovery from spinal cord injury**. *Proc. Natl. Acad. Sci. USA* (2006.0) **103** 13174-13179. DOI: 10.1073/pnas.0603747103
115. Dvir-Szternfeld R. **Alzheimer’s disease modification mediated by bone marrow-derived macrophages via a TREM2-independent pathway in mouse model of amyloidosis**. *Nat. Aging* (2022.0) **2** 60-73. DOI: 10.1038/s43587-021-00149-w
|
---
title: The role of position in consensus dynamics of polarizable networks
authors:
- Helge Giese
- Felix Gaisbauer
- Nico Gradwohl
- Ariana Strandburg-Peshkin
journal: Scientific Reports
year: 2023
pmcid: PMC9998643
doi: 10.1038/s41598-023-30613-z
license: CC BY 4.0
---
# The role of position in consensus dynamics of polarizable networks
## Abstract
Communication constraints often complicate group decision-making. In this experiment, we investigate how the network position of opinionated group members determines both the speed and the outcome of group consensus in 7-member communication networks susceptible to polarization. To this end, we implemented an online version of a color coordination task within experimentally controlled communication networks. In 72 networks, one individual was incentivized to prefer one of two options. In 156 networks, two individuals were incentivized to prefer conflicting options. The network positions of incentivized individuals were varied. In networks with a single incentivized individual, network position played no significant role in either the speed or outcome of consensus decisions. For conflicts, the incentivized individual with more neighbors was more likely to sway the group to their preferred outcome. Furthermore, consensus emerged more slowly when the opponents had the same number of neighbors, but could not see each other’s votes directly. These results suggest that the visibility of an opinion is key to wielding group influence, and that specific structures are sufficient to run communication networks into polarization, hindering a speedy consensus.
## Introduction
Both animal and human groups often need to find mutual agreement to effectively exploit their environment1 or to fulfill social needs2. For example, cohesively moving animal groups must come to consensus on when and where to move3, while human groups must come to consensus on topics ranging from where to go for lunch to political decisions. Consensus typically must be reached via communication among group members. Yet, in many cases communication is constrained such that individuals only communicate directly with specific neighbors rather than globally with all group members4. In animal groups, individuals are typically only able to observe and interact with group members who are sufficiently close to them in space4. In human groups, communication links are often also shaped by physical proximity, but especially with the rise of the internet and social media can also reflect other social, economic, or political ties. The resulting communication networks for both animal and human groups can vary dramatically in their efficiency at spreading information and behavior, depending on the structural arrangement of communication ties5–8 and the distribution of opinions across network positions9–14. In this study, we empirically investigate how communication network structures interact with the opinions of individuals populating the network to determine the speed and outcomes of group-wide consensus decisions.
To reach a consensus, it is often crucial for groups to effectively reconcile divergent opinions, information, and preferences15. In some cases, individuals with a preference for a particular outcome may successfully coordinate the diffusion of behaviors, resulting in the group selecting their preferred outcome16. Such opinionated individuals can also play a key role in determining the speed of consensus17,18. However, when two or more individuals have conflicting preferences, not only can such conflict slow down the consensus finding process, but it can also qualitatively change how a consensus is reached18: While increasing the density of communication channels (i.e. network ties) can often speed up group decision-making processes, this effect can be mitigated under conflicts of interest18. Likewise, the number and position of unopinionated individuals becomes more relevant in conflict scenarios18,19.
## Differences in influence based on network degree vs. information centrality
Prior theoretical and empirical work suggests that individuals more centrally located in a group’s communication network may play an outsized role in facilitating the coordination needed for consensus. However, little is known about the specific function of central individuals in consensus decision making. The different possible mechanisms underlying how an individual’s network position translates into influence may be best captured by the different definitions of network centrality.
On the one hand, individuals may exert more influence on the group outcome by displaying their opinions more visibly to others using their higher number of contacts, as captured by the notion of degree centrality20. In this vein, Kearns et al.13 showed that conflicting networks designed to have higher degree differences were considerably faster, and leaned toward opinions held by individuals with higher degrees. Other studies have shown that tweaking the degrees of opponents may even lead to false perceptions of which opinions are held by the majority, ultimately biasing consensus decisions10,11.
On the other hand, opinionated individuals may also influence the group by selecting which information they transmit21–23. This ability to influence networks through spreading information is reflected in centrality measures that capture more global information than degree centrality, for example information centrality—defined as the harmonic average of the inverse of all communication path lengths involving a node20. In this regard, Fitch and Leonard17 showed via theoretical models that individuals high in this form of centrality spread information most efficiently and are the most effective at leading groups to come to consensus on their preferred outcomes. Similar effects linked to these types of centrality can be observed in coordination games24,25. Yet, because these different types of centrality influence are typically conflated in random or naturally observed networks26, their potentially different roles in finding consensus, especially in conflict scenarios, remains unknown.
## Polarization of opinions
Natural networks typically show clusters of opinions27, also referred to as polarization. Such polarization can arise both due to the tendency to form ties to like-minded individuals (homophily) and due to the tendency to adopt the opinions of social ties. In this study specifically, we define polarization as the group average of the difference between the proportion of individual agreement to directly tied neighbors and non-neighbors (for further details and mathematical operationalization see Section 3 of the Supplement). Polarized opinion networks are widely regarded as problematic, as polarization slows down consensus formation and increases the likelihood of deadlocks11,13.
It is unclear whether problems in finding consensus primarily stem from disadvantageous distributions of individual opinions in a network, or whether polarization and associated problems can also be provoked merely by the setup of specific communication network structures. Centola6 indicated it may be beneficial to have reinforcing, clustered ties for the spread of information. In cases of conflict, however, densely-tied neighborhoods that receive little and indirect feedback from outside may quickly run into false majority beliefs and therefore be in danger of polarization and deadlocks11. As such, it may be crucial for the avoidance of such states that a potential conflict of interest is directly and not indirectly experienced, with a network structure allowing people with contradicting opinions to directly communicate.
## Current study
In this study, we explore how the positions of people with specific preferences in a communication network affect consensus formation in a color coordination game13,18, in which all members of a network have to coordinate to eventually agree on one of two colors by observing others color choices and voting within 50 trials. We study two scenarios: a “leadership” scenario where the group contains a single individual (henceforth “leader”) additionally incentivized to prefer one of the colors, and a “conflict” scenario where the group contains two individuals with conflicting incentivization (henceforth “opponents”). In particular, we aim to compare the effects of degree vs. information centrality and direct vs. indirect conflict on the speed and outcomes of consensus by randomly allocating the opinionated individuals to a position in 7-individual dumbbell-shaped communication network (Fig. 1) and testing whether any positioning will be superior for the consensus outcome or speed. Extending on classic group experiment structures, this dumbbell network allows us to explicitly contrast different centrality definitions, compare indirect and direct conflict, and to emulate a small-world structure25,28.Figure 1An example task screen (left) and the underlying network structure of the task (right). Left: A horizontal array of colored boxes represents the choices made in the last round. The choices of players not neighboring in the network are greyed out. In the example, players V, N, and R are visible to focal player E, who has been assigned the target to sway the group to blue. Only individuals incentivized to be opinionated about a specific target color saw the information about the extra reward and target color (dashed box). Right: Visualization of the underlying network structure. Note that this information was NOT visible to players. Node colors represent players’ choices in the last round, and the triangle node shape represents the incentivized individual.
We expect to replicate the finding that conflict slows down consensus18 and that under conditions of leadership, networks will form a consensus faster if that leader is more centrally located (e.g.12). We also test whether degree central (in particular, S and E in Fig. 1) or information central leaders (in particular, V in Fig. 1) and opponents wield more influence over group outcomes (see Fig. S1 for detailed differences between network positions in a set of centrality measures). In addition, we expect that direct conflicts between opponents (i.e. when two individuals with opposing incentives are directly tied/neighbors) should be resolved quickly because opinionated individuals experience imminent deadlocks first-hand. In contrast, in cases of indirect conflict (i.e. when opposing individuals are apart and cannot see one another directly), consensus should require more time due to the need for information to diffuse along neighborhoods of followers before reaching opponents.
## Speed of consensus in the two scenarios
To some degree, we could replicate previous findings that convergence is slower under conflict compared to leadership, yet the effect was marginally significant only (log-rank-χ2 ($$n = 228$$, df = 1) = 3.64, $$p \leq 0.056$$, Fig. 2).Figure 2Consensus speed by scenario. Shaded areas represent $95\%$ CI.
## Positional influence of a single opinionated individual (leadership)
In replication of Gaisbauer et al.18, we find that a single, opinionated individual tends to dictate group choice, thus effectively exerting leadership. 35 out of the 41 converged leadership networks agreed on the leader’s target color ($85\%$ of the converged networks involving a leader; $95\%$ CI [$71\%$, $94\%$], pbinom(41, 0.5) < 0.0001). The leader’s position in the network neither significantly predicted the outcome of consensus formation (χ2($$n = 72$$, df = 6) = 1.39, $$p \leq 0.967$$) nor the speed of the process (log-rank-χ2($$n = 72$$, df = 2) = 1.34, $$p \leq 0.511$$).
## Positional influence of opposing opinionated individuals (conflict)
Under conflict of interest, most networks converged and successfully averted deadlocks (78 out of 84 non-dropped networks). Convergence and drop-out did not differ significantly based on the position of the opponents (χ2($$n = 156$$, df = 12) = 11.62, $$p \leq 0.477$$), nor based on the combination of the position of opponents and their degree (χ2($$n = 156$$, df = 6) = 7.83, $$p \leq 0.251$$).
Out of the converged networks with differing degree centrality between opponents, opponents with higher degree centrality were more likely to win ($68\%$ of the converged networks involving opponents with different degree centrality; $95\%$ CI [$51\%$, $81\%$], pbinom(40, 0.5) = 0.0385), whereas information centrality did not yield any significant effect on the group outcome ($48\%$ of the converged networks involving opponents with different information centrality; $95\%$ CI [$32\%$, $63\%$]; pbinom(44, 0.5) = 0.880). Robustness checks revealed that these degree-centrality effects were also present when considering the majority of choices in the final round of each group before convergence, drop-out, or dead-lock (see Supplement).
## Positional effects of opponents on convergence speed and polarization
Generally, the speed of convergence differed marginally across all seven configurations (log-rank-χ2($$n = 156$$, df = 6) = 11.04, $$p \leq 0.087$$, for single configuration comparisons see Supplement). When collapsing conditions by centrality and neighborhood, there was a marginal effect of neighboring opponents showing that opponents being apart slows down convergence in general (b = – 0.431, HR = 0.649, $$p \leq 0.058$$; Table 1). However, this effect was qualified by the degree similarity of the individuals (Table 1): having opponents as neighbors rather than apart significantly improved speed particularly when opponents had the same degree ($b = 0.977$, HR = 2.657, $$p \leq 0.008$$), but not when they had different degrees (b = – 0.288, HR = 0.749, $$p \leq 0.413$$; Fig. 3). On the other hand, information centrality differences neither significantly predicted convergence speed nor significantly moderated the effects of neighbors (Table 1).Table 1Cox proportional hazards models of convergence speed in the conflict scenario. Model12a3a2b3bNeighbor (− 0.5 = neigh.) ( 0.5 = apart) − 0.431+ (0.228) − 0.327 (0.283) − 0.344 (0.246) − 0.449+ (0.252) − 0.851* (0.371)Degree sim. ( − 0.5 = diff.) ( 0.5 = same) − 0.175 (0.283) − 0.183 (0.246)Inf. centr. sim. ( − 0.5 = diff.) ( 0.5 = same) − 0.043 (0.253) − 0.074 (0.240)Neighbor x Degree sim. − 1.266* (0.523)Neighbor x Inf. centr. sim.0.686 (0.489)AIC601.08602.69599.2603.05603.14R20.0230.0250.0590.0230.035LR-χ2(df = 1)3.57 + 0.385.49*0.031.91SE in parenthesis. + $p \leq 0.1$; *$p \leq 0.05.$Figure 3Consensus speed and degree of polarization by position in the conflict scenario. Colors of the graph represent the relative density of the polarization level of groups in the game over the course of all 50 trials with 1(red) being the highest density observed per condition. The bars above the graph indicate the proportion of groups in the respective panel not converged or dropped out in the corresponding trial (scaled between 0 and 1). Numbers above the graph state the respective hazards of convergence (i.e., convergence speed inverse), bracketed numbers are the respective $95\%$ CI. For the definition of polarization and how patterns affect polarization, please refer to the Supplement. Densities were normalized by condition to avoid negligibly low densities across all observations and to account for uninformative differences in the number of observations per condition. Note that only a discrete number of polarization values was realized in our experiment because polarization was constrained by network configurations (the maximum polarization is 0.74, followed by 0.36; see Supplement Fig. S7).
Figure 3 additionally illustrates that networks particularly with non-neighboring, same-degree opponents were not only slow to find consensus, but also polarized to a larger extent for longer periods of time up to the maximal possible degree of polarization (one arm voting against the other arm; see Supplement for the possible patterns of polarization, also for a Figure for each of the seven opponents' positions separately).
## Discussion
Overall, groups under constrained communication are able to come to consensus in short times, while even settling competing interests within a group13,18. Under conflicts of interest about the group-wide decision outcome, power imbalances and relative position of opinionated opponents vis-à-vis each other can drive the speed with which conflicts are resolved and consensus is reached.
When only one member of the network was opinionated (leadership condition), networks converged quickly and mostly on the incentivized color, overall replicating the results of Gaisbauer et al.18. In cases of conflicting interests, more degree-central individuals were better able to wield their influence on the group, while information centrality did not seem to matter as much.
In addition, if the degrees of the opponents were the same, groups profited from direct interaction of the opponents with the unopinionated individual’s responses potentially serving as tie-breakers in voting. Generally, while individual decisions seem to require reinforcement in the local neighborhood—in line with majority rule and the “complex contagion” account of Centola6, overall consensus can be found more easily when dissent can be directly experienced—in line with the “small world” literature9,29. Thus, when coming to consensus in a group with conflicting interests, there seems to be an inherent efficiency tradeoff between the clustering of communication ties and having long ties to avoid polarization by directly interacting with opponents. Extending information propagation and leadership scenarios to situations in which varying degrees of conflicts are involved may therefore further help to reconcile two seemingly contradictive arms of research that advocate for opposing structures (clustered vs. small world designs) to achieve communication efficiency in consensus decision making.
As a cautionary note, given the very fast speed and high rate of convergence, the null-findings in the leadership scenario may be due to a ceiling effect. In addition, all non-significant results could also be due to a lack of power and thus should not be regarded as evidence for no effect. Therefore, we rather focus our interpretation on the effects found. Furthermore, the small size of our exemplary network and the small differences in centrality within may have generally limited the importance of positional effects. Accordingly, future studies should test generalizability to larger groups and to more difficult coordination tasks.
In sum, the positioning of conflicting opinions in neighborhoods that are only loosely tied to each other may be particularly susceptible to polarization—also showing the limits of a potential balancing role of uninformed group members in promoting group consensus18,19. Mainly visibility of invested people’s votes helps them to steer the group to their preferred outcome in cases of conflict of interests.
## Participants
Participants were recruited online via Amazon Mechanical Turk (MTurk). We ensured technically that MTurk workers did not participate multiple times or in an earlier study with a related design18 and data of that study are not used in the current one. Participants earned an average performance bonus of 0.35 USD (SD = 0.28 USD) in addition to a show-up fee (1.50 USD) and waiting bonus (up to 0.50 USD). They needed about 10 min to complete the whole study. 1596 participants in 228 networks started the main task ($56.33\%$ self-reported as female, Mage = 35.27 years, SDage = 10.62 years). Drop-out rates after experimental assignment ($7.1\%$) were in line with comparable online studies18,30. The study adhered to the Declaration of Helsinki, relevant laws, and institutional guidelines, as certified by IRB of the University of Konstanz. The University of Konstanz IRB approved the study. All participants gave informed consent.
## Procedures
The experiment was conducted online via the software oTree31 and the procedures are comparable to Gaisbauer et al.18. After informed consent, participants were presented a CAPTCHA to screen out automated scripts, followed by detailed instructions and a comprehension test.
For the main task, participants were placed in groups of 7 individuals and were instructed to find a group-wide consensus by unanimously choosing one out of two colors (blue vs. yellow) as fast as possible within 50 rounds: If the network converged on one color, everyone received a bonus fee which started at 0.50 USD and decreased by 0.01 USD increments per round. In case a participant dropped out of the study, we stopped the experiment for that group. The abandoned group members, and also all others completing the task, received a show-up fee of 1.50 USD regardless of success.
At the start of the task, each player was randomly assigned a position in the dumbbell network (Fig. 1), was given a random letter as name, and started without a pre-selected color. The network position determined which choices of other group members of the last trial were visible to whom. The task ended when all players had chosen the same color or when 50 rounds had passed without consensus. Participants received feedback on their total earnings after the main task, completed a short survey about the satisfaction with the task and potential strategies the implemented, and were debriefed. For a view of the participant screen in the task, see Fig. 1.
## Design
We varied the degree of conflict in this networked color coordination paradigm by incentivizing one or two individuals within the networks to prefer a specific color (i.e. to be opinionated; see13,18). In the leadership scenario, one individual (the “leader”) from each network was randomly assigned and instructed to receive an additional bonus of 0.50 USD if the network converged on a specific color. In the conflict scenario, two individuals (the “opponents”) were assigned to receive the bonus with conflicting target colors. These instructions were unknown to the other participants. The network positions of leaders and opponents were randomly determined.
## Data analysis
We used χ2-tests to assess how features of the network are associated with the different decision outcomes of the group (convergence, drop-out, non-convergence). To determine the role of degree and information centrality on who is more likely to win in conflict scenarios, we used binomial testing and confidence bands.
We analyzed time-until-consensus for networks with a Kaplan–Meier survival analysis with a LR-χ2-test using the R-package survival32. Observations were treated as right-censored if the network did not reach consensus within 50 rounds or if the network dropped out of the study. Likewise, after comparing the two scenarios, we analyzed the time-until consensus data separately for each scenario (leadership vs. conflict) to test for effects of positioning on the speed of consensus. That is, for the leadership scenario ($$n = 72$$ networks), we compared the speed of consensus between the networks with different leader positions. For the conflict scenario ($$n = 156$$), we compared the speed of consensus between all 7 configurations of how opponents can be positioned on the network. In addition, we evaluated, in a 2 × 2 design, whether consensus speed is affected by opponents being neighbors (vs. apart) and having the same (vs. different) centrality using Cox Proportional Hazard models with this information effect-coded as factors with interactions, using the R-package survival32 and LR-χ2-tests for testing differences between models.
We operationalized and tested positional influence of an opinionated individual by its centrality in the network with two measures: (a) degree centrality, defined as the number of direct neighbors, and (b) information centrality, a measure of the control of information flow, defined as the harmonic average of the inversely related path lengths between a focal node and all other nodes17,20. Please note that in this specific network set-up, more information central nodes are also the ones with higher betweenness and closeness centrality (see Fig. S1). To evaluate interactions, we used simple effects analysis of the hazards using the R package emmeans33.
The analyses and hypotheses were not preregistered. Given the number of sampled networks (72 and 156), the study is powered to detect medium effects in standard analyses with a power of 1–β = 0.8034.
## Supplementary Information
Supplementary Information. The online version contains supplementary material available at 10.1038/s41598-023-30613-z.
## References
1. Hastie R, Kameda T. **The robust beauty of majority rules in group decisions**. *Psychol. Rev.* (2005.0) **112** 494-508. DOI: 10.1037/0033-295X.112.2.494
2. Baumeister RF, Leary MR. **The need to belong: desire for interpersonal attachments as a fundamental human motivation**. *Psychol. Bull.* (1995.0) **117** 497-529. DOI: 10.1037/0033-2909.117.3.497
3. Conradt L, Roper TJ. **Deciding group movements: Where and when to go**. *Behav. Process.* (2010.0) **84** 675-677. DOI: 10.1016/j.beproc.2010.03.005
4. Conradt L, Roper TJ. **Consensus decision making in animals**. *Trends Ecol. Evol.* (2005.0) **20** 449-456. DOI: 10.1016/j.tree.2005.05.008
5. Christakis NA, Fowler JH. **The spread of obesity in a large social network over 32 years**. *N. Engl. J. Med.* (2007.0) **357** 370-379. DOI: 10.1056/NEJMsa066082
6. Centola D. **The spread of behavior in an online social network experiment**. *Science* (2010.0) **329** 1194-1197. DOI: 10.1126/science.1185231
7. Giese H, Stok FM, Renner B. **The role of friendship reciprocity in university freshmen’s alcohol consumption**. *Appl. Psychol. Health Well-Being* (2017.0) **9** 228-241. DOI: 10.1111/aphw.12088
8. Judd S, Kearns M, Vorobeychik Y. **Behavioral dynamics and influence in networked coloring and consensus**. *Proc. Natl. Acad. Sci.* (2010.0) **107** 14978-14982. DOI: 10.1073/pnas.1001280107
9. Watts DJ, Strogatz SH. **Collective dynamics of ‘small-world’ networks**. *Nature* (1998.0) **393** 440-442. DOI: 10.1038/30918
10. Lerman K, Yan X, Wu XZ. **The ‘majority illusion’ in social networks**. *PLoS One* (2016.0) **11** 1-13. DOI: 10.1371/journal.pone.0147617
11. Stewart AJ. **Information gerrymandering and undemocratic decisions**. *Nature* (2019.0) **573** 117-121. DOI: 10.1038/s41586-019-1507-6
12. Kearns M, Suri S, Montfort N. **An experimental study of the coloring problem on human subject networks**. *Science* (2006.0) **313** 824-827. DOI: 10.1126/science.1127207
13. Kearns M, Judd S, Tan J, Wortman J. **Behavioral experiments on biased voting in networks**. *Proc. Natl. Acad. Sci. U. S. A.* (2009.0) **106** 1347-1352. DOI: 10.1073/pnas.0808147106
14. Yang G, Cavaliere M, Zhu C, Perc M. **Strategically positioning cooperators can facilitate the contagion of cooperation**. *Sci. Rep.* (2021.0) **11** 1-12. PMID: 33414495
15. Couzin ID, Krause J, Franks NR, Levin SA. **Effective leadership and decision-making in animal groups on the move**. *Nature* (2005.0) **433** 513-516. DOI: 10.1038/nature03236
16. Centola D. **An experimental study of homophily in the adoption of health behavior**. *Science* (2011.0) **334** 1269-1272. DOI: 10.1126/science.1207055
17. 17.Fitch, K. & Leonard, N. E. Information centrality and optimal leader selection in noisy networks. In 52nd {IEEE} Conference on Decision and Control (2013).
18. Gaisbauer F, Strandburg-Peshkin A, Giese H. **Local majority-with-inertia rule can explain global consensus dynamics in a network coordination game**. *Soc. Netw.* (2022.0) **70** 218-227. DOI: 10.1016/j.socnet.2022.01.013
19. Couzin ID. **Uninformed individuals promote democratic consensus in animal groups**. *Science* (2011.0) **334** 1578-1580. DOI: 10.1126/science.1210280
20. Wasserman S, Faust K. *Social Network Analysis: Methods and Applications. Social Networks* (1994.0)
21. Bakshy E, Messing S, Adamic LA. **Exposure to ideologically diverse news and opinion on Facebook**. *Science* (2015.0) **348** 1130-1132. DOI: 10.1126/science.aaa1160
22. Moussaïd M, Brighton H, Gaissmaier W. **The amplification of risk in experimental diffusion chains**. *Proc. Natl. Acad. Sci. U. S. A.* (2015.0) **112** 5631-5636. DOI: 10.1073/pnas.1421883112
23. Giese H, Neth H, Moussaïd M, Betsch C, Gaissmaier W. **The echo in flu-vaccination echo chambers: Selective attention trumps social influence**. *Vaccine* (2020.0) **38** 2070-2076. DOI: 10.1016/j.vaccine.2019.11.038
24. Burt RS, Reagans RE, Volvovsky HC. **Network brokerage and the perception of leadership**. *Soc. Netw.* (2021.0) **65** 33-50. DOI: 10.1016/j.socnet.2020.09.002
25. Freeman LC, Roeder D, Mulholland RR. **Centrality in social networks: II Experimental results**. *Soc. Netw.* (1979.0) **2** 119-141. DOI: 10.1016/0378-8733(79)90002-9
26. Valente TW, Coronges K, Lakon C, Costenbader E. **How correlated are network centrality measures?**. *Connections* (2008.0) **28** 16-26. PMID: 20505784
27. McPherson M, Smith-Lovin L, Cook JM. **Birds of a Feather: Homophily in Social Networks**. *Annu. Rev. Sociol.* (2001.0) **27** 415-444. DOI: 10.1146/annurev.soc.27.1.415
28. Valente TW, Fujimoto K. **Bridging: Locating critical connectors in a network**. *Soc. Netw.* (2010.0) **32** 212-220. DOI: 10.1016/j.socnet.2010.03.003
29. Granovetter M. **The strength of weak ties: A network theory revisited**. *Soc. Theory* (1983.0) **1** 201-233. DOI: 10.2307/202051
30. Arechar AA, Gächter S, Molleman L. **Conducting interactive experiments online**. *Exp. Econ.* (2017.0). DOI: 10.1007/s10683-017-9527-2
31. Chen DL, Schonger M, Wickens C. **oTree–An open-source platform for laboratory, online, and field experiments**. *J. Behav. Exp. Financ.* (2016.0) **9** 88-97. DOI: 10.1016/j.jbef.2015.12.001
32. 32.Therneau, T. M. A Package for Survival Analysis in R (2022).
33. 33.Lenth, R. V. emmeans: Estimated Marginal Means, aka Least-Squares Means (2021).
34. Cohen J. **A power primer**. *Psychol. Bull.* (1992.0) **112** 155-159. DOI: 10.1037/0033-2909.112.1.155
|
---
title: Low-carbohydrate diets containing plant-derived fat but not animal-derived
fat ameliorate heart failure
authors:
- Satoshi Bujo
- Haruhiro Toko
- Kaoru Ito
- Satoshi Koyama
- Masato Ishizuka
- Masahiko Umei
- Haruka Yanagisawa-Murakami
- Jiaxi Guo
- Bowen Zhai
- Chunxia Zhao
- Risa Kishikawa
- Norifumi Takeda
- Kensuke Tsushima
- Yuichi Ikeda
- Eiki Takimoto
- Hiroyuki Morita
- Mutsuo Harada
- Issei Komuro
journal: Scientific Reports
year: 2023
pmcid: PMC9998649
doi: 10.1038/s41598-023-30821-7
license: CC BY 4.0
---
# Low-carbohydrate diets containing plant-derived fat but not animal-derived fat ameliorate heart failure
## Abstract
Cardiovascular disease (CVD) is a global health burden in the world. Although low-carbohydrate diets (LCDs) have beneficial effects on CVD risk, their preventive effects remain elusive. We investigated whether LCDs ameliorate heart failure (HF) using a murine model of pressure overload. LCD with plant-derived fat (LCD-P) ameliorated HF progression, whereas LCD with animal-derived fat (LCD-A) aggravated inflammation and cardiac dysfunction. In the hearts of LCD-P-fed mice but not LCD-A, fatty acid oxidation-related genes were highly expressed, and peroxisome proliferator-activated receptor α (PPARα), which regulates lipid metabolism and inflammation, was activated. Loss- and gain-of-function experiments indicated the critical roles of PPARα in preventing HF progression. Stearic acid, which was more abundant in the serum and heart of LCD-P-fed mice, activated PPARα in cultured cardiomyocytes. We highlight the importance of fat sources substituted for reduced carbohydrates in LCDs and suggest that the LCD-P-stearic acid-PPARα pathway as a therapeutic target for HF.
## Introduction
Cardiovascular disease (CVD) is a global health burden with increasing prevalence and mortality rates1. CVDs result from unhealthy lifestyles, as well as genetic factors, and some risk factors, such as smoking status, physical inactivity, obesity, poor dietary habits, high blood glucose levels, high serum cholesterol levels, and high blood pressure, can be changed through lifestyle modifications2,3. Dietary intervention is one of the most effective methods of reducing these risk factors. Low-carbohydrate diets (LCDs) have been used for weight loss4, and many randomized clinical trials show that LCDs induce beneficial changes in CVD risk factors, such as hypertension, dyslipidemia, and diabetes5–8. However, it has remained unclear whether LCDs reduce the incidence of CVDs and related death9–11. Recent clinical studies have suggested the importance of macronutrients substituted for reduced carbohydrates in LCDs12–14. LCDs with animal-derived fat and protein sources were associated with higher CVD mortality, whereas LCDs with plant-derived fat and protein sources were associated with lower CVD mortality. Despite these distinct effects of LCDs on CVDs, their pathological mechanisms have not yet been elucidated.
Peroxisome proliferator-activated receptor α (PPARα) is a transcription factor that regulates lipid metabolism genes by sensing fatty acids. PPARα promotes fatty acid oxidation (FAO) in the muscle, liver, and adipose tissue, increasing energy expenditure and reducing body fat. PPARα agonists have been clinically used to treat dyslipidemia. In addition to these effects of lowering the CVD risk factors, anti-inflammatory effects are also reported in liver and blood vessels15. Since heart failure (HF), a leading cause of CVD-related death, is characterized by low FAO rates with energy metabolism disruption, PPARα activation is expected to rescue the failing heart16. There were several studies concerning the effects of LCDs on HF. A lard diet significantly hastened death as compared with a high-carbohydrate diet, whereas a linoleate diet significantly delayed the death of spontaneously hypertensive HF rats17. Additionally, strict restriction of carbohydrates attenuated the progression of pathological hypertrophy and systolic dysfunction in a pressure-overload model, and LCDs with high-protein supplementation and high-fat supplementation showed cardioprotection through distinct mechanisms18. These findings suggest that the beneficial effects of LCDs on HF depend on the substitute supplementation in LCDs.
In this study, we investigated whether the effects of LCDs on HF differed depending on the fat sources of substitute supplementation in LCDs and explored the molecular mechanisms of these distinct effects.
## Differences in the effects of LCDs on cardiac function depending on the fat source
To examine whether the effects of LCDs on cardiac function depend on the fat source in LCDs, we prepared two types of LCDs ($12\%$ carbohydrate and $59\%$ fat of total energy) as follows: LCD with animal-derived fat (beef tallow; LCD-A) and LCD with plant-derived fat (cocoa butter; LCD-P). A high-carbohydrate standard diet ($59\%$ carbohydrate and $12\%$ fat of total energy; SD) was used as a control (Fig. 1a). There was no significant difference in calorie intake and serum triglyceride levels between mice fed these two types of LCDs (Supplementary Fig. 1a,b). We provided mice with SD, LCD-A, or LCD-P for 4 weeks starting from the day of pressure-overload or sham surgery (Fig. 1b). Pressure overload induced by transverse aortic constriction (TAC) increased the wall thickness and dimensions of the left ventricle (LV) and reduced cardiac systolic function in mice on SD (Fig. 1c,d). LCD-P attenuated these morphologic changes, including LV hypertrophy and LV dilatation, and ameliorated LV systolic dysfunction as compared with that observed in the SD group. By contrast, LV dilatation and systolic dysfunction on LCD-A were more prominent than those on SD, and we noted further differences in LV dimensions and systolic function relative to those on LCD-P. Additionally, heart weight (HW) and the HW:tibial length (TL; HW/TL) ratio were lower in mice on LCD-P than in those on SD or LCD-A, which agreed with the result of LV mass (LVM) calculated using echocardiographic data (Fig. 1d–f). The expression levels of Nppb, a well-known marker gene for hypertrophy and heart failure, were also decreased in the heart of LCD-P-fed mice more than that of SD- or LCD-A-fed mice (Fig. 1g).Figure 1Distinct effects of LCDs on cardiac hypertrophy by fat sources. ( a) Pie chart showing the proportions of calories from carbohydrate (C), fat (F), and protein (P) in each diet. SD, standard diet; LCD-A, low-carbohydrate diet with animal-derived fat; LCD-P, low-carbohydrate diet with plant-derived fat. ( b) Experimental outline. Wild-type mice were fed with the indicated diets for 4 weeks starting from the day of transverse aortic constriction (TAC) or sham surgery. Analysis was performed at 1- or 4-weeks post-surgery. ( c) Representative images of motion-mode (M-mode) echocardiography taken at 4-weeks post-surgery. Vertical scale bar, 1 mm. Transverse scale bar, 100 ms. ( d) Cardiac function was assessed by M-mode echocardiography at 4-weeks post-TAC or post-sham surgery ($$n = 8$$–10). IVSd diastolic interventricular-septum thickness, LVDd left ventricular end-diastolic dimension, LVDs left ventricular end-systolic dimension, LVM left ventricular mass, FS fractional shortening. ( e) Representative gross anatomies of the heart at 4-weeks post-TAC or post-sham surgery. Scale bar, 2 mm. ( f) Body weight (BW), heart weight (HW), HW:BW ratio, and HW:tibial length (TL) ratio at 4-weeks post-TAC or post-sham surgery ($$n = 8$$–10). ( g) mRNA level of Nppb in the heart at 1-week post-TAC or post-sham surgery ($$n = 3$$–5). Data represent the mean ± SEM. Statistical significance was analyzed by two-way ANOVA, followed by Holm–Sidak’s post-hoc test. * $P \leq 0.05$, **$P \leq 0.01$, ***$P \leq 0.001$, ****$P \leq 0.0001.$
Histochemical analyses revealed that LCD-A increased the number of inflammatory cells, such as F$\frac{4}{80}$-positive macrophages, in hypertrophied hearts at 1-week post-TAC, which was not observed in the SD or LCD-P groups (Fig. 2a,b). Moreover, LCD-A strongly upregulated the expression of inflammatory cytokines, such as Il6 and Tnf, which was not observed in the SD or LCD-P groups (Fig. 2c). These results indicated that LCD-P and LCD-A exerted distinct effects on inflammation and HF progression. Figure 2Effects of LCDs on cardiac inflammation induced by TAC surgery. ( a) Representative immunofluorescent staining of heart sections with anti-F$\frac{4}{80}$ antibody at 1-week post-TAC surgery. Scale bar, 100 μm. ( b) The number of F$\frac{4}{80}$-positive cells in heart sections at 1-week post-TAC or post-sham surgery ($$n = 6$$–9). ( c) mRNA levels of Il6 and Tnf in the heart at 1-week post-TAC or post-sham surgery ($$n = 4$$–10). Data represent the mean ± SEM. Statistical significance was analyzed by two-way ANOVA, followed by Holm–Sidak’s post-hoc test. * $P \leq 0.05$, **$P \leq 0.01$, ***$P \leq 0.001$, ****$P \leq 0.0001.$
## LCD-P increases the expression levels of PPARα target genes in hearts
To elucidate the mechanisms by which LCD-P exerts its beneficial effects on pressure-overloaded hearts, we examined gene expression by RNA-sequencing (RNA-seq) analysis, which detected 14,683 genes. Using a false discovery rate of < 0.05 and an absolute \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\sqrt{2}$$\end{document}2 -fold change expression as cut-offs, we identified 363 differentially expressed genes (DEGs) in TAC mouse hearts in the LCD-A group from those in the SD, and that LCD-P altered the expression of 139 genes relative to the SD group (Fig. 3a and Supplementary Tables 1 and 2). We obtained the top 15 Gene Ontology (GO) terms from the enrichment analysis of these DEGs (Fig. 3b and Supplementary Tables 3 and 4). LCD-A induced the expression of cell cycle-related genes, whereas FAO-related genes were upregulated in LCD-P hearts. We then used Ingenuity Pathway Analysis (IPA) to elucidate the upstream regulators of the DEGs. IPA identified Ppara, a regulator of FAO, as the most highly activated transcription factor among the DEGs in LCD-P hearts (Fig. 3c and Supplementary Table 5). Furthermore, five PPARα agonists, including pirinixic acid, gemfibrozil, and fenofibrate, were detected among the top 15 potential upstream regulators. Additionally, network analysis of the DEGs indicated that PPARα target genes play critical roles in the hearts of mice on LCD-P (Fig. 4a). We subsequently confirmed upregulated expression of PPARα target genes (such as Cpt1b, Lcad, Mcad, and Plin5) using quantitative reverse transcription polymerase chain reaction (RT-qPCR) (Fig. 4b). Interestingly, even after TAC, the expression levels of some PPARα-related genes were clearly upregulated in LCD-P hearts, which was not observed in LCD-A hearts. Moreover, the Gene Expression Omnibus (GEO) dataset GSE57338 revealed that PPARA expression is downregulated in human hearts with HF, especially related to dilated cardiomyopathy (Fig. 4c). These results suggested that LCD-P enhances FAO through PPARα activation, leading to increased energy efficiency in the failing heart. Figure 3RNA-seq results in the hearts of mice fed different LCDs. ( a) Volcano plot showing differentially expressed genes (DEGs) in the hearts of mice fed SD, LCD-A, or LCD-P diets at 1-week post-TAC surgery. ( b) Gene *Ontology analysis* of DEGs between SD-fed mice and LCD-A- or LCD-P-fed mice after TAC surgery. ( c) Top 15 upstream regulators of DEGs between SD-fed mice and LCD-P-fed mice after TAC surgery that were identified by Ingenuity Pathway Analysis. The red bar shows PPARα-related regulators. Figure 4PPARα expression in failing hearts in mice and humans. ( a) Significantly interacting genes among DEGs found in LCD-P-fed mice after TAC surgery. ( b) mRNA levels of PPARα target genes in the heart at 1-week post-TAC or post-sham surgery ($$n = 5$$–7). ( c) PPARA-expression levels in 136 non-failing (NF) hearts and 82 failing hearts with dilated cardiomyopathy (DCM), as determined by analyzing human Gene Expression Omnibus dataset GSE57338. Data represent the mean ± SEM. Statistical significance was analyzed by two-way ANOVA, followed by Holm–Sidak’s post-hoc test among multiple group comparisons. The Mann–Whitney U test was used for two-group comparisons. * $P \leq 0.05$, **$P \leq 0.01$, ***$P \leq 0.001.$
## Cardiac function is preserved by PPARα agonist and deteriorated by the loss of PPARα under stress
Since we suspected that PPARα plays a central role in LCD-P-mediated cardioprotection against pressure overload, we investigated the effects of PPARα on HF. First, we examined the role of PPARα in HF development using cardiomyocyte-specific Ppara-conditional knockout (cKO) mice. Although there was no change in LV size or function in control and cKO mice without pressure overload, LV dimension was larger and LV systolic function was lower in cKO mice after TAC as compared with control mice (Fig. 5a,b). These physiological changes were accompanied with higher Nppb levels in cKO heart than control after TAC, indicating heart failure status in cKO (Fig. 5c). Additionally, we found a higher number of inflammatory cells and elevated gene-expression levels of inflammatory cytokines in cKO mice relative to those in control mice (Fig. 5d,e). We then activated PPARα using pemafibrate, a selective PPARα modulator, and found that pemafibrate ameliorated the TAC-induced LV dilatation and systolic dysfunction (Fig. 5f,g) and reduced *Nppb* gene expression (Fig. 5h). Furthermore, pemafibrate significantly decreased the number of infiltrating inflammatory cells and reduced the gene-expression levels of inflammatory cytokines in hypertrophied hearts (Fig. 5i,j).Figure 5Effects of PPARα agonist and Ppara deletion on cardiac function. ( a) Representative motion-mode (M-mode) echocardiography images of Ppara-conditional-knockout (cKO) mice at 4-weeks post-TAC or post-sham surgery. Vertical scale bar, 1 mm. Transverse scale bar, 100 ms. ( b) Cardiac function of cKO mice, as assessed by M-mode echocardiography at 4-weeks post-TAC or post-sham surgery ($$n = 5$$–7). ( c) mRNA level of Nppb in the hearts of cKO mice at 1-week post-TAC or post-sham surgery ($$n = 4$$–9). ( d) The number of F$\frac{4}{80}$-positive cells per field of view in the hearts of cKO mice at 1-week post-TAC surgery ($$n = 5$$–6). ( e) mRNA levels of Il6 and Tnf in the hearts of cKO mice at 1-week post-TAC or post-sham surgery ($$n = 4$$–9). ( f) Representative M-mode echocardiography images of mice in the pemafibrate group at 4-weeks post-TAC or post-sham surgery. Vertical scale bar, 1 mm. Transverse scale bar, 100 ms. ( g) Cardiac function of mice in the pemafibrate group, as assessed by M-mode echocardiography at 4-weeks post- TAC or post-sham surgery ($$n = 4$$–10). ( h) mRNA level of Nppb in the hearts of mice in the pemafibrate group at 1-week post-TAC or post-sham surgery ($$n = 3$$–5). ( i) The number of F$\frac{4}{80}$-positive cells per field of view in the hearts of mice in the pemafibrate group at 1-week post-TAC surgery ($$n = 6$$). ( j) mRNA levels of Il6 and Tnf in the hearts of mice in the pemafibrate group at 1-week post-TAC or post-sham surgery ($$n = 3$$–5). Data represent the mean ± SEM. Statistical significance was analyzed by two-way ANOVA, followed by Holm–Sidak’s post-hoc test among multiple group comparisons. The two-tailed unpaired Student’s t test was used for two-group comparisons. * $P \leq 0.05$, **$P \leq 0.01$, ***$P \leq 0.001$, ****$P \leq 0.0001.$
## Loss of PPARα abolishes the cardioprotective effects of LCD-P
We then investigated whether LCD-P exerts beneficial effects on the heart via PPARα activation by using PPARα-cKO mice. In SD-fed cKO mice, TAC-induced cardiac hypertrophy and heart failure were observed as indicated by the increase in diastolic interventricular-septum thickness (IVSd), LV end-diastolic dimension (LVDd) and HW:body weight (BW; HW/BW) ratio and the decrease in LV fractional shortening (FS). In LCD-P-fed cKO mice, there were no significant differences in IVSd, LVDd, HW/BW, or FS compared to that of SD-fed cKO mice (Fig. 6a–c). Likewise, the expression levels of PPARα and its target gene expressions, which were upregulated by LCD-P in the wild mice (Fig. 4b), were unchanged by LCD-P in cKO mice (Fig. 6d), suggesting that the loss of cardioprotective effects of LCD-P in cKO heart were caused by the PPARα inactivity. Taken together, LCD-P exerts beneficial effects on the heart via PPARα activation. Figure 6Effects of LCD-P under Ppara deletion on cardiac hypertrophy. ( a) Representative motion-mode (M-mode) echocardiography images of Ppara-conditional-knockout (cKO) mice at 4-weeks post-TAC after 4-weeks of LCD-P feeding. Vertical scale bar, 1 mm. Transverse scale bar, 100 ms. ( b) Cardiac function of cKO mice after 4-weeks of LCD-P feeding, as assessed by M-mode echocardiography at 4-weeks post-TAC surgery ($$n = 6$$–12). ( c) Body weight (BW), heart weight (HW), HW:BW ratio, and HW:tibial length (TL) ratio at 4-weeks post-TAC surgery ($$n = 6$$–12). ( d) mRNA levels of Ppara and its target genes in the hearts of cKO mice after 1-week of LCD-P feeding ($$n = 4$$–6). Data represent the mean ± SEM. Statistical significance was analyzed by two-way ANOVA, followed by Holm–Sidak’s post-hoc test. * $P \leq 0.05$, **$P \leq 0.01$, ***$P \leq 0.001$, ****$P \leq 0.0001.$
## Stearic acid increases the expression levels of PPARα target genes
To elucidate the mechanism by which LCD-P activates PPARα, we first examined the fatty acid (FA) compositions of LCD-P and LCD-A. LCD-P was rich in saturated fatty acids (SFAs) and had much higher levels of stearic acid (SA) than LCD-A (Fig. 7a). By contrast, LCD-A had a much higher content of monounsaturated fatty acids (MUFAs) and oleic acid (OA) than LCD-P. Additionally, we examined the changes in FA profiles in the sera and hearts of mice at 1-week post-surgery, when there was no difference in cardiac function among the three groups. Each LCD resulted in a significant elevation of both serum SA and OA levels and reduced serum linoleic acid (LA) levels (Fig. 7b). Moreover, we observed a significant increase only in serum SA levels in LCD-P-fed mice as compared with those in LCD-A-fed mice, with similar patterns observed in SA, OA, and LA levels in the hearts of the respective mice (Fig. 7c). Although serum palmitic acid (PA) levels were higher in LCD-fed mice than in SD-fed mice, we found no significant difference between the LCD-A and LCD-P groups, and cardiac PA concentrations were similar among all three groups either with or without TAC. These findings suggested that LCD-P rich in SA might exert beneficial effects on hypertrophied hearts. Figure 7Selected FA compositions in the sera and hearts of mice. ( a) FA profiles of total lipids among the three diet groups. LA, linoleic acid; MUFA, monounsaturated fatty acid; OA, oleic acid; PA, palmitic acid; PUFA, polyunsaturated fatty acid; SA, stearic acid; SFA, saturated fatty acid. ( b) Selected FA compositions in the sera at 1-week post-TAC or post-sham surgery ($$n = 3$$–7). ( c) Selected FA compositions in the hearts at 1-week post-TAC or post-sham surgery ($$n = 5$$–11). Data represent the mean ± SEM. Statistical significance was analyzed by two-way ANOVA, followed by Holm–Sidak’s post-hoc test. * $P \leq 0.05$, **$P \leq 0.01$, ***$P \leq 0.001$, ****$P \leq 0.0001.$
Long-chain FAs, such as OA, PA, and SA, are physiological ligands of PPARα19. Thus, we added SA to the culture medium of neonatal rat cardiomyocytes (NRCMs). We identified SA-mediated elevations in the expression levels of PPARα target genes, such as Acaa2, Atgl, Cpt1a, Lcad, and Plin5, in a dose-dependent manner (Fig. 8a). Additionally, administration of phenylephrine (PE) reduced the expression levels of PPARα target genes, whereas this response was countered by SA stimulation (Fig. 8b). Consequently, SA reduced the expression levels of hypertrophy and heart failure marker genes Nppa and Nppb that were increased by PE. These results suggested that SA, which is abundant in the hearts of LCD-P-fed mice, plays an important role in cardioprotection by activating PPARα under the hypertensive state. Figure 8Expression levels of PPARα target genes in SA-stimulated cardiomyocytes. ( a) mRNA levels of Ppara and its target genes in neonatal rat cardiomyocytes (NRCMs) after stimulation with SA for 6 h ($$n = 6$$–7). The SA concentrations tested were 10 μM and 50 μM. (b) mRNA levels of hypertrophic markers and PPARα target genes under SA stimulation in NRCMs treated with 100 μM phenylephrine (PE) for 24 h ($$n = 9$$). The SA concentrations tested were 10 μM and 50 μM. Data represent the mean ± SEM. Statistical significance was analyzed by one-way ANOVA, followed by the Holm–Sidak’s post-hoc test. * $P \leq 0.05$, **$P \leq 0.01$, ***$P \leq 0.001$, ****$P \leq 0.0001.$
## Discussion
The results of recent cohort studies and meta-analyses suggest the importance of replacing carbohydrates in LCDs with fat and protein sources12–14. In the present study, we found that LCD-P ameliorated HF progression while LCD-A aggravated cardiac dysfunction and that the distinct effects of the LCDs on HF may depend on PPARα activation.
PPARα is a transcription factor abundantly expressed in the heart and regulates various physiological processes16,20. Its expression is downregulated in failing human and rodent hearts21–23, whereas mechanical unloading in failing human hearts increases the expression levels of PPARα and its target genes24. Global Ppara-KO mice demonstrate exacerbated cardiac dysfunction in response to pressure overload25,26, suggesting that PPARα is involved in HF development. Because PPARα is highly expressed in many other organs, including the liver, kidney, and skeletal muscle15,27, it is difficult to exclude any indirect effects of non-cardiac PPARα signaling on the heart. Furthermore, the effects of PPARα agonists on cardiac remodeling are controversial22,28–30. Therefore, to elucidate the role of cardiac PPARα, we generated cardiomyocyte-specific Ppara-cKO mice. Although there was no change in LV size and function in either control or cKO mice without pressure overload, cKO mice demonstrated more severe LV dysfunction and inflammation than control mice following TAC (Fig. 5b–e). Additionally, PPARα activation by pemafibrate, a selective PPARα modulator, ameliorated LV dysfunction and TAC-induced inflammation (Fig. 5g–j). The activation ability and selectivity of fibrates previously used as PPARα agonists might be low, and they could activate other PPAR isotypes, such as PPARγ and PPARδ, to some extent. By contrast, pemafibrate is a more potent and selective PPARα agonist than other fibrates31–33. The results of the loss- and gain-of-function experiments in the present study clearly indicated that PPARα activation was beneficial in preventing HF development and that the cardioprotective effects of LCD-P were PPARα dependent (Fig. 6b–d).
Cardiac hypertrophy accelerates the metabolic-flux shift from FAO to glucose oxidation, which may lead to energy insufficiency34. Several studies have shown improvement in cardiac hypertrophy and function by enhancing FAO35–37. Since PPARα reportedly activates FAO, LCD-P might exert beneficial effects on cardiac function in the presence of pressure overload by activating FAO via PPARα activation in order to meet the high cardiac energy demand16. In the present study using RNAseq, among 139 differentially expressed genes between SD and LCD-P, we found FAO-related genes being upregulated in LCD-P. Further analysis for upstream regulators of FAO revealed Ppara as the most highly activated transcription factor. Since the depletion of Ppara, specifically in cardiomyocytes, abolished the beneficial effects of LCD-P and the upregulation of FAO-related genes, we suggest that a possible mechanism by which LCD-P protects the heart is the enhancement of FAO through PPARα activation, leading to increased energy efficiency in the failing heart.
Inflammation reportedly causes cardiac dysfunction, and PPARα decreases the expression of proinflammatory genes such as Il6 and Tnf15,38. The present study exhibiting the exacerbation of cardiac inflammation by LCD-A but not by LCD-P suggested that LCD-P may have the potential to prevent the development of cardiac dysfunction by PPARα-mediated suppression of inflammation. Indeed, the PPARα cKO heart demonstrated a higher number of F$\frac{4}{80}$ positive inflammatory cells in the TAC heart (Fig. 5d) and, conversely, pemafibrate, a selective PPARα activator, successfully reduced the inflammation (Fig. 5i). These results suggest that LCD-P ameliorated HF through PPARα-mediated anti-inflammatory effects.
X-ray crystallography recently revealed SA and PA as physiological PPARα ligands39. In the present study, the results showed that SA dose-dependently upregulated the expression of PPARα target genes in cultured cardiomyocytes in vitro and restored the expression of genes downregulated by PE, a cardiomyocyte hypertrophy-inducing factor40 (Fig. 8a,b). These results suggest that abundant SA in LCD-P enhances PPARα activity in hypertrophied hearts. Additionally, SA reportedly improves mitochondrial function by increasing mitochondrial fusion in Drosophila and humans41,42 and exerts an anti-inflammatory response in cholestatic liver injury by reducing leukocyte accumulation and NF-κB activity in rats43. Moreover, a previous study reported the neuroprotective effects of SA against oxidative stress in the rat brain44. These results collectively suggest that SA protects various organs by activating multiple pathways, including the PPARα signaling pathway as shown in the heart.
This study has limitations. First, we used only one type of fat source for LCD-A and LCD-P, respectively; therefore, further studies are required to confirm whether the results of this study are dependent on animal- or plant-derived fat. Second, we did not provide precise mechanisms related to worsening HF by LCD-A. Exacerbation of cardiac inflammation and the associated elevation of inflammatory genes were the hallmarks of LCD-A-fed heart tissue, but no possible upstream regulators of these changes were discovered even with RNA-seq analysis. Well-used high-fat diets often consist of animal fats such as lard, which are similar to LCD-A and are associated with poorer cardiac outcomes during stress or aging45–47. Third, we could not measure the FAO or glucose oxidation rate, which remains to be clarified in future research. Finally, there are controversies in the relationship between SFA and CVD48. Given that individual SFAs have different biological effects and their long-term influence on health remains unclear, the duration, timing, and amount of LCDs containing plant-derived fat should be carefully optimized before clinical application.
In conclusion, these findings suggest that substituting reduced carbohydrates with plant-derived fat is beneficial to preventing HF development and that the LCD-P-SA-PPARα pathway may be a potential therapeutic target for treating HF.
## Mice
All animal experiments were approved by the Ethics Committee for Animal Experiments of the University of Tokyo (Tokyo, Japan) and adhered strictly to animal experimental guidelines and the ARRIVE guidelines. The procedures strictly adhered to the guidelines for animal experiments of the University of Tokyo and the National Institutes of Health guidelines for the Care and Use of Laboratory Animals. All mice were housed under a controlled temperature with a 12-h light/dark cycle and provided food and water ad libitum.
For the LCD experiments, 9- to 10-week-old male C57BL/6J mice were purchased from CLEA Japan (Tokyo, Japan). Littermate mice of similar ages and weights were randomized to experimental and control groups. The mice were anesthetized with $2\%$ isoflurane and subjected to TAC (26-gauge needle) or sham surgery, as previously described49. After the surgery, cages of the mice were randomized to the feeding protocols by another researcher who was blinded to the experiments and those analyses. They were fed either a standard diet (CE-2; $59\%$ carbohydrate and $12\%$ soybean-based fat of total energy, 344 kcal/100 g; CLEA Japan), an LCD-A diet ($12\%$ carbohydrate and $59\%$ beef tallow-based fat of total energy, 485 kcal/100 g; Oriental Yeast, Tokyo, Japan), or an LCD-P diet ($12\%$ carbohydrate and $59\%$ cocoa butter-based fat of total energy, 485 kcal/100 g; Oriental Yeast). The mice were fed the indicated diets ad libitum for 4 weeks starting from the day of surgery.
Pemafibrate was kindly provided by Kowa (Aichi, Japan). Specifically, 9- to 10-week-old male C57BL/6J mice received pemafibrate (1.0 mg/kg body weight) orally daily from 1 week before TAC (26-gauge needle) or sham surgery until the end of the study. Methylcellulose ($0.5\%$) was used as the vehicle.
Ppara-cKO mice were generated using the Cre/loxP system. Pparaflox/flox mice were provided by Dr. Frank J. Gonzalez (National Cancer Institute, Bethesda, MD, USA)50. αMHC-Cre mice (Tg [Myh6-cre] 1Jmk/J, #009074) were purchased from the Jackson Laboratory (Bar Harbor, ME, USA). Specifically, 9- to 10-week-old male αMHC-Cre+/−; Pparaflox/flox mice underwent TAC (25-gauge needle) or sham surgery, with littermate control mice (Pparaflox/flox) operated in the same way. As αMHC-Cre+/−; Pparaflox/flox mice were more vulnerable to TAC-induced heart failure and died soon after the surgery, we used a larger needle (25-gauge), not a standard one (26-gauge), leaving milder aortic constriction.
The mice were analyzed at 1- and 4-weeks post-surgery. Transthoracic echocardiography was performed on conscious mice using a Vevo2100 ultrasound system (FUJIFILM VisualSonics, Toronto, ON, Canada). Diastolic interventricular-septum thickness (IVSd), diastolic posterior-wall thickness (PWd), LV end-diastolic dimension (LVDd), and LV end-systolic dimension (LVDs) measurements were taken using motion-mode echocardiography. LV fractional shortening (FS) was calculated as follows: FS = (LVDd − LVDs)/LVDd × 100. LVM was calculated as follows: 1.05 × {(IVSd + LVDd + PWd)3 − (LVDd)3}. Blood samples were collected by intracardiac puncture under $2\%$ isoflurane anesthesia. Deeply anesthetized mice were euthanized by cervical dislocation. Serum samples were isolated by centrifugation at 3000 rpm for 15 min.
## Immunohistochemical analysis
Mouse hearts were fixed with $20\%$ formalin (Sakura Finetek Japan, Tokyo, Japan), embedded in paraffin, and sectioned at a 4-μm thickness. The sections were stained with anti-F$\frac{4}{80}$ antibodies (MCA497GA; Bio-Rad, Hercules, CA, USA) to evaluate cell infiltration. The number of infiltrating cells was counted in each heart section for the LCD experiments and in each visual field (200×) for the pemafibrate and Ppara-cKO mouse experiments. Visual fields were selected randomly under a BZ-X800 microscope (Keyence, Osaka, Japan), and the average number of infiltrating cells in five fields was calculated for each sample.
## RT-qPCR analysis
Total RNA was extracted from cells and tissues using an RNeasy Mini kit (Qiagen, Hilden, Germany) or a tissue total RNA Mini kit (Favorgen, Ping-Tung, Taiwan), respectively, according to manufacturer instructions. cDNA was generated using ReverTra Ace qPCR RT master mix (Toyobo, Osaka, Japan), and qPCR was performed using Thunderbird Next SYBR qPCR mix (Toyobo) with primers specific for each gene of interest (Supplementary Table 6). The resulting data were analyzed using a QuantStudio 5 instrument (Thermo Fisher Scientific, Waltham, MA, USA). *Relative* gene expression levels were determined using the relative standard curve method and normalized against the expression of 18S ribosomal RNA.
## Bulk RNA-seq analysis
Total RNA was extracted from mouse heart tissues at 1-week post-TAC or post-sham surgery ($$n = 3$$ mice/group) and used to generate RNA-seq libraries with a TruSeq stranded mRNA library prep kit (Illumina, San Diego, CA, USA). The Illumina HiSeq 2500 platform was used for sequencing. Raw reads were checked for quality using the FastQC program (version 0.11.15) and trimmed using Trimmomatic (version 0.36), where low-quality bases with Phred quality scores of < 33 were discarded. Adaptor sequences were removed using Cutadapt (version 1.14). STAR aligner (version 2.5.2b) was used to align clean reads to the mouse reference genome (mm9). RSubread-2.0.1-FeatureCounts software was employed for quantification, and DESeq2 software (version 1.28.0) was used for differential analysis using R software (version 3.5.1).
## GEO dataset analysis
We selected the GSE57338 dataset to analyze differences in gene expression associated with human HF. PPARA expression levels were compared in 136 non-failing hearts and 82 failing hearts with dilated cardiomyopathy. *The* gene expression levels were determined using the GEO2R web tool.
## FA analysis
Dietary FAs were measured by Japan Food Research Laboratories (Tokyo, Japan) using gas chromatography-mass spectrometry (GC–MS), as previously described51.
The FA compositions in serum and heart samples were measured by a central laboratory (BML, Tokyo, Japan). Lipids from the hearts were extracted as previously described by Folch et al.52 After spiking the samples with tricosanoic acid as an internal control, serum and heart lipids were methylated using boron trifluoride and methanol. The methylated FAs were then measured by GC–MS analysis (QC-2010; SHIMADZU, Kyoto, Japan)53.
## NRCM culture
NRCMs were prepared as previously described54. NRCMs were obtained from 0- to 2-day-old Wistar rats (Takasugi Experimental Animal Supply, Saitama, Japan), dispersed by collagenase digestion, and subjected to Percoll gradient centrifugation. Isolated cardiomyocytes were cultured for 48 h in Dulbecco’s modified *Eagle medium* (Nacalai Tesque, Kyoto, Japan) supplemented with $10\%$ fetal calf serum. After 24 h of serum starvation, NRCMs were stimulated with SA and PE.
## SA stimulation
SA (S4751; Sigma-Aldrich, St. Louis, MO, USA) was conjugated to FA-free bovine serum albumin (BSA; A6003; $20\%$ concentration; Sigma-Aldrich) at a molecular ratio of 2.2:1. Working stocks of SA (10 μM or 50 μM; used for stimulation) were prepared from a 5 mM stock solution. NRCMs were treated with SA for 6 h in the absence of PE. Cardiomyocyte hypertrophy was induced by treatment with 100 μM PE (P6126; Sigma-Aldrich) for 24 h. In the presence of PE, NRCMs were pre-stimulated with SA 2 h before PE treatment. Control cells were cultured in the presence of $0.03\%$ BSA alone.
## Statistical analysis
Data are shown as mean ± standard error of the mean (SEM). Statistical analyses were performed using GraphPad Prism software (version 9.2.0; GraphPad Software, San Diego, CA, USA). A two-tailed unpaired Student’s t-test or a Mann–Whitney U test was performed to compare the two groups. One-way or two-way ANOVA followed by Holm-Sidak’s post-hoc analysis was conducted to compare multiple groups. The threshold for statistical significance was set at $P \leq 0.05.$
## Supplementary Information
Supplementary Information. The online version contains supplementary material available at 10.1038/s41598-023-30821-7.
## References
1. Ziaeian B, Fonarow GC. **Epidemiology and aetiology of heart failure**. *Nat. Rev. Cardiol.* (2016) **13** 368-378. DOI: 10.1038/nrcardio.2016.25
2. Jessup M, Brozena S. **Heart failure**. *N. Engl. J. Med.* (2003) **348** 2007-2018. DOI: 10.1056/NEJMra021498
3. Pritchett AM, Foreyt JP, Mann DL. **Treatment of the metabolic syndrome: The impact of lifestyle modification**. *Curr. Atheroscler. Rep.* (2005) **7** 95-102. DOI: 10.1007/s11883-005-0030-4
4. Astrup PA, Meinert Larsen DT, Harper A. **Atkins and other low-carbohydrate diets: Hoax or an effective tool for weight loss?**. *Lancet* (2004) **364** 897-899. DOI: 10.1016/S0140-6736(04)16986-9
5. Kirkpatrick CF. **Review of current evidence and clinical recommendations on the effects of low-carbohydrate and very-low-carbohydrate (including ketogenic) diets for the management of body weight and other cardiometabolic risk factors: A scientific statement from the National Lipid Association Nutrition and Lifestyle Task Force**. *J. Clin. Lipidol.* (2019) **13** 689-711.e1. DOI: 10.1016/j.jacl.2019.08.003
6. Foster GD. **A randomized trial of a low-carbohydrate diet for obesity**. *N. Engl. J. Med.* (2003) **348** 2082-2090. DOI: 10.1056/NEJMoa022207
7. Samaha FF. **A low-carbohydrate as compared with a low-fat diet in severe obesity**. *N. Engl. J. Med.* (2003) **348** 2074-2081. DOI: 10.1056/NEJMoa022637
8. Seid H, Rosenbaum M. **Low carbohydrate and low-fat diets: What we don’t know and why we should know it**. *Nutrients* (2019) **11** 2749. DOI: 10.3390/nu11112749
9. Trichopoulou A, Psaltopoulou T, Orfanos P, Hsieh C-C, Trichopoulos D. **Low-carbohydrate-high-protein diet and long-term survival in a general population cohort**. *Eur. J. Clin. Nutr.* (2007) **61** 575-581. DOI: 10.1038/sj.ejcn.1602557
10. Mazidi M, Katsiki N, Mikhailidis DP, Sattar N, Banach M. **Lower carbohydrate diets and all-cause and cause-specific mortality: A population-based cohort study and pooling of prospective studies**. *Eur. Heart J.* (2019) **40** 2870-2879. DOI: 10.1093/eurheartj/ehz174
11. Nakamura Y. **Low-carbohydrate diets and cardiovascular and total mortality in Japanese: A 29-year follow-up of NIPPON DATA80**. *Br. J. Nutr.* (2014) **112** 916-924. DOI: 10.1017/S0007114514001627
12. Seidelmann SB. **Dietary carbohydrate intake and mortality: A prospective cohort study and meta-analysis**. *Lancet Public Health* (2018) **2667** 1-10
13. Fung TT. **Low-carbohydrate diets and all-cause and cause-specific mortality: Two cohort studies**. *Ann. Intern. Med.* (2010) **153** 289-298. DOI: 10.7326/0003-4819-153-5-201009070-00003
14. Li S. **Low carbohydrate diet from plant or animal sources and mortality among myocardial infarction survivors**. *J. Am. Heart Assoc.* (2014) **3** 1-12. DOI: 10.1161/JAHA.114.001169
15. Bougarne N. **Molecular actions of PPARα in lipid metabolism and inflammation**. *Endocr. Rev.* (2018) **39** 760-802. DOI: 10.1210/er.2018-00064
16. Montaigne D, Butruille L, Staels B. **PPAR control of metabolism and cardiovascular functions**. *Nat. Rev. Cardiol.* (2021) **18** 809-823. DOI: 10.1038/s41569-021-00569-6
17. Chicco AJ. **Linoleate-rich high-fat diet decreases mortality in hypertensive heart failure rats compared with lard and low-fat diets**. *Hypertension* (2008) **52** 549-555. DOI: 10.1161/HYPERTENSIONAHA.108.114264
18. Nakamura M. **Dietary carbohydrates restriction inhibits the development of cardiac hypertrophy and heart failure**. *Cardiovasc. Res.* (2021) **117** 2365-2376. DOI: 10.1093/cvr/cvaa298
19. Xu HE. **Molecular recognition of fatty acids by peroxisome proliferator-activated receptors**. *Mol. Cell* (1999) **3** 397-403. DOI: 10.1016/S1097-2765(00)80467-0
20. Dubois V, Eeckhoute J, Lefebvre P, Staels B. **Distinct but complementary contributions of PPAR isotypes to energy homeostasis**. *J. Clin. Invest.* (2017) **127** 1202-1214. DOI: 10.1172/JCI88894
21. Karbowska J, Kochan Z, Smoleński RT. **Peroxisome proliferator-activated receptor alpha is downregulated in the failing human heart**. *Cell. Mol. Biol. Lett.* (2003) **8** 49-53. PMID: 12655356
22. Young ME, Laws FA, Goodwin GW, Taegtmeyer H. **Reactivation of peroxisome proliferator-activated receptor alpha is associated with contractile dysfunction in hypertrophied rat heart**. *J. Biol. Chem.* (2001) **276** 44390-44395. DOI: 10.1074/jbc.M103826200
23. Barger PM, Brandt JM, Leone TC, Weinheimer CJ, Kelly DP. **Deactivation of peroxisome proliferator-activated receptor-alpha during cardiac hypertrophic growth**. *J. Clin. Invest.* (2000) **105** 1723-1730. DOI: 10.1172/JCI9056
24. Gupte AA. **Mechanical unloading promotes myocardial energy recovery in human heart failure**. *Circ. Cardiovasc. Genet.* (2014) **7** 266-276. DOI: 10.1161/CIRCGENETICS.113.000404
25. Smeets PJH. **Cardiac hypertrophy is enhanced in PPARα−/− mice in response to chronic pressure overload**. *Cardiovasc. Res.* (2008) **78** 79-89. DOI: 10.1093/cvr/cvn001
26. Luptak I. **Decreased contractile and metabolic reserve in peroxisome proliferator-activated receptor-alpha-null hearts can be rescued by increasing glucose transport and utilization**. *Circulation* (2005) **112** 2339-2346. DOI: 10.1161/CIRCULATIONAHA.105.534594
27. Jordan S. **Dietary intake regulates the circulating inflammatory monocyte pool**. *Cell* (2019) **178** 1102-1114.e17. DOI: 10.1016/j.cell.2019.07.050
28. Xu SC, Ma ZG, Wei WY, Yuan YP, Tang QZ. **Bezafibrate attenuates pressure overload-induced cardiac hypertrophy and fibrosis**. *PPAR Res.* (2017) **2017** 5789714. DOI: 10.1155/2017/5789714
29. Kaimoto S. **Activation of PPAR-α in the early stage of heart failure maintained myocardial function and energetics in pressure-overload heart failure**. *Am. J. Physiol. Heart Circ. Physiol.* (2017) **312** H305-H313. DOI: 10.1152/ajpheart.00553.2016
30. Duhaney T-AS. **Peroxisome proliferator-activated receptor alpha-independent actions of fenofibrate exacerbates left ventricular dilation and fibrosis in chronic pressure overload**. *Hypertension* (2007) **49** 1084-1094. DOI: 10.1161/HYPERTENSIONAHA.107.086926
31. Sasaki Y. **Gene expression profiles induced by a novel selective peroxisome proliferator-activated receptor α modulator (SPPARMα) pemafibrate**. *Int. J. Mol. Sci.* (2019) **20** 5682. DOI: 10.3390/ijms20225682
32. Fruchart J-C. **Pemafibrate (K-877), a novel selective peroxisome proliferator-activated receptor alpha modulator for management of atherogenic dyslipidaemia**. *Cardiovasc. Diabetol.* (2017) **16** 124. DOI: 10.1186/s12933-017-0602-y
33. Wei Z-L. **Isoxazolyl-serine-based agonists of peroxisome proliferator-activated receptor: Design, synthesis, and effects on cardiomyocyte differentiation**. *J. Am. Chem. Soc.* (2004) **126** 16714-16715. DOI: 10.1021/ja046386l
34. Abdurrachim D. **Good and bad consequences of altered fatty acid metabolism in heart failure: Evidence from mouse models**. *Cardiovasc. Res.* (2015) **106** 194-205. DOI: 10.1093/cvr/cvv105
35. Kolwicz SC. **Cardiac-specific deletion of acetyl CoA carboxylase 2 prevents metabolic remodeling during pressure-overload hypertrophy**. *Circ. Res.* (2012) **111** 728-738. DOI: 10.1161/CIRCRESAHA.112.268128
36. Guo Y. **Enhancing fatty acid utilization ameliorates mitochondrial fragmentation and cardiac dysfunction via rebalancing optic atrophy 1 processing in the failing heart**. *Cardiovasc. Res.* (2018) **114** 979-991. DOI: 10.1093/cvr/cvy052
37. Shao D. **Increasing fatty acid oxidation prevents high-fat diet-induced cardiomyopathy through regulating parkin-mediated mitophagy**. *Circulation* (2020) **142** 983-997. DOI: 10.1161/CIRCULATIONAHA.119.043319
38. Adamo L, Rocha-Resende C, Prabhu SD, Mann DL. **Reappraising the role of inflammation in heart failure**. *Nat. Rev. Cardiol.* (2020) **17** 269-285. DOI: 10.1038/s41569-019-0315-x
39. Kamata S. **PPARα ligand-binding domain structures with endogenous fatty acids and fibrates**. *iScience* (2020) **23** 101727. DOI: 10.1016/j.isci.2020.101727
40. Grimm M. **Key role of myosin light chain (MLC) kinase-mediated MLC2a phosphorylation in the alpha 1-adrenergic positive inotropic effect in human atrium**. *Cardiovasc. Res.* (2005) **65** 211-220. DOI: 10.1016/j.cardiores.2004.09.019
41. Senyilmaz D. **Regulation of mitochondrial morphology and function by stearoylation of TFR1**. *Nature* (2015) **525** 124-128. DOI: 10.1038/nature14601
42. Senyilmaz-Tiebe D. **Dietary stearic acid regulates mitochondria in vivo in humans**. *Nat. Commun.* (2018) **9** 3129. DOI: 10.1038/s41467-018-05614-6
43. Pan P-H. **Stearic acid attenuates cholestasis-induced liver injury**. *Biochem. Biophys. Res. Commun.* (2010) **391** 1537-1542. DOI: 10.1016/j.bbrc.2009.12.119
44. Wang Z-J, Li G-M, Tang W-L, Yin M. **Neuroprotective effects of stearic acid against toxicity of oxygen/glucose deprivation or glutamate on rat cortical or hippocampal slices**. *Acta Pharmacol. Sin.* (2006) **27** 145-150. DOI: 10.1111/j.1745-7254.2006.00259.x
45. Ikeda S. **Yes-associated protein (YAP) facilitates pressure overload-induced dysfunction in the diabetic heart**. *JACC Basic Transl. Sci.* (2019) **4** 611-622. DOI: 10.1016/j.jacbts.2019.05.006
46. Steinbusch LKM. **Absence of fatty acid transporter CD36 protects against Western-type diet-related cardiac dysfunction following pressure overload in mice**. *Am. J. Physiol. Endocrinol. Metab.* (2011) **301** E618-E627. DOI: 10.1152/ajpendo.00106.2011
47. Holzem KM, Marmerstein JT, Madden EJ, Efimov IR. **Diet-induced obesity promotes altered remodeling and exacerbated cardiac hypertrophy following pressure overload**. *Physiol. Rep.* (2015) **3** 12489. DOI: 10.14814/phy2.12489
48. Krauss RM, Kris-Etherton PM. **Public health guidelines should recommend reducing saturated fat consumption as much as possible: Debate consensus**. *Am. J. Clin. Nutr.* (2020) **112** 25-26. DOI: 10.1093/ajcn/nqaa134
49. Toko H. **Regulation of cardiac hypertrophic signaling by prolyl isomerase Pin1**. *Circ. Res.* (2013) **112** 1244-1252. DOI: 10.1161/CIRCRESAHA.113.301084
50. Brocker CN. **Hepatocyte-specific PPARA expression exclusively promotes agonist-induced cell proliferation without influence from nonparenchymal cells**. *Am. J. Physiol. Gastrointest. Liver Physiol.* (2017) **312** G283-G299. DOI: 10.1152/ajpgi.00205.2016
51. Morrison WR, Tan SL, Hargin KD. **Methods for the quantitative analysis of lipids in cereal grains and similar tissues**. *J. Sci. Food Agric.* (1980) **31** 329-340. DOI: 10.1002/jsfa.2740310402
52. Folch J, Lees M, Sloane Stanley GH. **A simple method for the isolation and purification of total lipides from animal tissues**. *J. Biol. Chem.* (1957) **226** 497-509. DOI: 10.1016/S0021-9258(18)64849-5
53. Nozue T. **Effects of statins on serum n-3 to n-6 polyunsaturated fatty acid ratios in patients with coronary artery disease**. *J. Cardiovasc. Pharmacol. Ther.* (2013) **18** 320-326. DOI: 10.1177/1074248412473202
54. Ishizuka M. **CXCR7 ameliorates myocardial infarction as a β-arrestin-biased receptor**. *Sci. Rep.* (2021) **11** 3426. DOI: 10.1038/s41598-021-83022-5
|
---
title: Central retinal microvasculature damage is associated with orthostatic hypotension
in Parkinson’s disease
authors:
- Jong Hyeon Ahn
- Min Chae Kang
- Dongyoung Lee
- Jin Whan Cho
- Kyung-Ah Park
- Jinyoung Youn
journal: NPJ Parkinson's Disease
year: 2023
pmcid: PMC9998652
doi: 10.1038/s41531-023-00480-6
license: CC BY 4.0
---
# Central retinal microvasculature damage is associated with orthostatic hypotension in Parkinson’s disease
## Abstract
Orthostatic hypotension (OH) is a common non-motor symptom in Parkinson’s disease (PD). OH can cause cerebral and retinal hypoperfusion and is associated with microvascular damage in PD. Optical coherence tomography angiography (OCTA) is a non-invasive technology that can be used to visualize the retinal microvasculature and detect microvascular damage in PD. In the present study, 51 PD patients (OH+, $$n = 20$$, 37 eyes; OH−, $$n = 32$$, 61 eyes) and 51 healthy controls (100 eyes) were evaluated. The Unified Parkinson’s Disease Rating Scale III, Hoehn and Yahr scale, Montreal Cognitive Assessment, levodopa equivalent daily dose, and vascular risk factors, including hypertension, diabetes, and dyslipidemia, were investigated. PD patients underwent a head-up tilt (HUT) test. The PD patients had a lower superficial retinal capillary plexus (SRCP) density in the central region than control patients. The PDOH+ group had lower vessel density in the SRCP of the central region compared with the control group and lower vessel density in the DRCP of the central region than the PDOH− and control groups. The changes in systolic and diastolic blood pressure during the HUT test in PD patients showed a negative correlation with the vessel density in the DRCP central region. The presence of OH was a critical factor associated with central microvasculature damage in PD. These findings indicate that OCTA can be a useful and non-invasive tool for detecting microvasculature damage in PD patients.
## Introduction
Orthostatic hypotension (OH) is a common non-motor symptom in Parkinson’s disease (PD), even in the early stage of the disease1. OH in PD is associated with a higher mortality rate, cognitive decline, higher fall risk, and poorer quality of life2. OH induces a recurrent episodic cerebral and retinal hypoperfusion and presents with various symptoms such as blurred vision, loss of peripheral vision (gray-out), color changes, and scotomas3. The underpinnings of the relationship between OH and clinical outcomes in PD are not well understood; however, a vascular hypothesis has been suggested as a possible pathophysiological mechanism2. The fluctuation of systemic blood pressure (BP) and impairment of autoregulation induce microvasculature damage in patients with PD4. However, this is difficult to evaluate, as investigating microvasculature damage in vivo is challenging.
Optical coherence tomography angiography (OCTA) is a novel non-invasive technology that can be used to visualize retinal microvasculature. The observation of ocular microcirculation provides an opportunity to evaluate how human circulation responds to stress stimuli5. OCTA is useful for detecting microvascular change induced in diabetes and glaucoma, as well as systemic hypotension5,6. Microvascular changes have been proposed as a potential contributor to PD, and OCTA has previously been used to detect retinal microvascular changes in PD7. PD patients were shown to have a lower macular vessel density than healthy controls, indicating that OCTA parameters can be potential diagnostic biomarkers8,9. In contrast, others reported no significant differences in the OCTA parameters of patients with PD compared with healthy controls or patients with essential tremors, and scientific evidence explaining the inconsistent results and pathophysiology of microvascular alterations in PD is lacking10–12. In the present study, we hypothesized that OH in PD might be associated with microvasculature damage and that OCTA might be a useful tool for detecting such damage in PD patients.
## Results
The analysis included 51 patients with PD (61 eyes) and 51 healthy controls (100 eyes). The demographics and clinical characteristics of the enrolled participants are described in Table 1. Significant differences in age, sex, vascular risk factors, and spherical equivalent refractive errors (SERs) were not found between the PD patients and controls. Twenty patients (37 eyes) had OH (PDOH+), 31 (61 eyes) did not (PDOH−), the prevalence of OH was $39.2\%$, and females were more affected. The mean age, sex, and SERs were comparable among the PDOH+, PDOH−, and control groups. Significant differences in the disease duration, Unified Parkinson’s Disease Rating Scale (UPDRS) III, Hoehn and Yahr (H&Y) scale, levodopa equivalent daily dose (LEDD), and Montreal Cognitive Assessment (MoCA) were not observed between the PDOH+ and PDOH− groups. The mean systolic BP (SBP) and diastolic BP (DBP) measured in the supine position were not significantly different between the PDOH+ and PDOH− groups. The mean decreases in SBP from supine to head-up tilt position in the PDOH+ and PDOH− groups was 33.3 and 8.9 mmHg, respectively. The mean DBP decrease from supine to head-up tilt position was also higher in the PDOH+ group than in the PDOH− group (14.2 vs. 3.4 mmHg). The prevalence of supine hypertension was $25.0\%$ in the PDOH+ group and $6.5\%$ in the PDOH− group. Table 1Demographics and clinical characteristics of the participants. PD ($$n = 51$$, 98 eyes)Controls ($$n = 51$$, 100 eyes)PDOH+ ($$n = 20$$, 37 eyes)PDOH− ($$n = 31$$, 61 eyes)p valuebp valuecAge (years)65.4 ± 6.365.3 ± 6.765.3 ± 6.365.3 ± 7.10.983>0.999Sex (male/female)$\frac{19}{3219}$/$\frac{324}{1615}$/16>0.9990.123Disease duration (months)51.1 ± 32.8–60.3 ± 40.145.2 ± 26.1–0.109UPDRS IIIa12.5 ± 7.6–14.5 ± 9.911.1 ± 5.5–0.123H&Y scalea1.7 ± 0.5–1.8 ± 0.51.6 ± 0.5–0.451LEDD (mg)384.1 ± 228.6–406.0 ± 241.8369.9 ± 222.6–0.587MoCA26.5 ± 2.0–26.8 ± 1.826.4 ± 2.1–0.667SER (diopters)b−0.10 ± 1.71−0.05 ± 1.62−0.24 ± 1.42−0.01 ± 1.860.8680.796HTN (%)29.423.530.029.00.5010.676DM (%)9.813.70.016.10.5390.180DL (%)13.721.610.016.10.1020.263Mean supine SBP (mmHg)128.8 ± 15.5128.0 ± 19.50.873Mean supine DBP (mmHg)69.2 ± 6.770.5 ± 11.60.867Mean change SBP (mmHg)33.3 ± 15.98.9 ± 8.4<0.001Mean change DBP (mmHg)14.2 ± 8.03.4 ± 3.8<0.001Supine HTN (%)13.725.06.50.060Data are presented as mean and standard deviation (SD).OH orthostatic hypotension, UPDRS Unified Parkinson’s Disease Rating Scale, H&Y Hoehn and Yahr, HTN hypertension (previously diagnosis), DM diabetes mellitus, DL dyslipidemia, LEDD levodopa equivalent daily dose, MoCA Montreal Cognitive Assessment, SER spherical equivalent refractive errors, SBP systolic blood pressure, DBP diastolic blood pressure.aEvaluated at the medication on state.bComparison of p and controls analyzed using the Student’s t test or chi-square test.cComparison of PDOH+, PDOH−, and controls analyzed using the analysis of covariance or chi-square test.
## Comparison of OCT and OCTA parameters between PD patients and healthy controls
The macular retinal thickness and peripapillary nerve fiber layer (pRNFL) thickness were measured using OCT. Significant differences in macular retinal thickness or pRNFL thickness were not found between the PD patients and controls. Macular retinal thickness nor pRNFL thickness differed among the three groups (Table 2).Table 2Comparison of retinal layer thickness and parafoveal vessel densities in patients with Parkinson’s disease and healthy controls. PD (98 eyes)Controls (100 eyes)PDOH+ (37 eyes)PDOH− (61 eyes)p valueap valuebpRNFL thickness (μm) Average100.1 ± 12.8101.8 ± 8.2101.7 ± 14.799.1 ± 11.5>0.999>0.999 Temporal74.5 ± 12.275.9 ± 11.376.4 ± 13.173.4 ± 11.6>0.999>0.999 Inferior131.0 ± 21.8128.7 ± 17.0134.1 ± 26.1129.1 ± 18.90.390>0.999 Nasal69.2 ± 13.471.2 ± 15.670.4 ± 12.968.5 ± 13.7>0.999>0.999 Superior125.5 ± 19.7126.2 ± 17.5126 ± 18.3125.2 ± 20.6>0.999>0.999Total macular thickness (μm) Central (1 mm)262.6 ± 19.9271.9 ± 18.2257.3 ± 21.8265.8 ± 18.00.0950.261 Average (3 mm)331.7 ± 15.9337.8 ± 13.3328 ± 13.4333.9 ± 16.9>0.999>0.999 Temporal (3 mm)324.8 ± 17.1330.0 ± 13.6321.5 ± 13.7326.8 ± 18.6>0.999>0.999 Inferior (3 mm)331.1 ± 16.5337.2 ± 13.9327.7 ± 13.2333.2 ± 17.9>0.999>0.999 Nasal (3 mm)336.3 ± 16.6342.4 ± 14.2331.3 ± 15.1339.4 ± 16.8>0.999>0.999 Superior (3 mm)334.6 ± 18.3341.8 ± 13.4331.8 ± 15.5336.4 ± 19.80.670>0.999 Average (6 mm)296.3 ± 14.6301.6 ± 15.0296.3 ± 9.9296.2 ± 16.8>0.999>0.999 Temporal (6 mm)281.5 ± 16.5287.1 ± 15.0281.1 ± 14.3281.7 ± 17.9>0.999>0.999 Inferior (6 mm)293.2 ± 19.2297.9 ± 26.8295.8 ± 20291.6 ± 18.7>0.999>0.999 Nasal (6 mm)306.9 ± 16.0314.1 ± 14.8304.1 ± 14.5308.6 ± 16.7>0.999>0.999 Superior (6 mm)302.7 ± 15.2307.4 ± 20.4302.6 ± 12.9302.7 ± 16.5>0.999>0.999FAZ area (mm3)0.329 ± 0.1050.274 ± 0.6670.361 ± 0.1300.309 ± 0.8070.0170.049cSRCP (%) Average47.6 ± 2.146.4 ± 2.548.1 ± 2.247.4 ± 2.10.4250.620 Central17.8 ± 4.519.6 ± 3.316.5 ± 4.618.6 ± 4.30.0070.008c Temporal46.6 ± 3.745.9 ± 3.047.2 ± 3.346.3 ± 3.9>0.999>0.999 Inferior48.7 ± 3.946.8 ± 4.149.4 ± 4.248.2 ± 3.70.6610.544 Nasal46.4 ± 2.745.1 ± 3.246.3 ± 2.846.5 ± 2.60.4240.674 Superior48.9 ± 2.847.7 ± 3.549.4 ± 2.648.5 ± 2.90.5260.444DRCP (%) Average49.2 ± 3.048.0 ± 2.549.8 ± 2.948.9 ± 3.10.1660.356 Central14.8 ± 4.616.0 ± 4.812.5 ± 3.616.2 ± 4.60.296<0.001c,d Temporal46.5 ± 4.146.6 ± 3.246.7 ± 3.746.4 ± 4.30.070>0.999 Inferior51.7 ± 5.849.5 ± 4.552.9 ± 6.251.0 ± 5.5>0.9990.094 Nasal48.0 ± 4.147.1 ± 3.748.2 ± 3.548.0 ± 4.4>0.999>0.999 Superior50.7 ± 3.649.0 ± 3.851.3 ± 3.350.3 ± 3.70.0760.141Data are presented as mean and standard deviation. The Bonferroni correction was performed for multiple comparison. PD Parkinson’s disease, OH orthostatic hypotension, PDOH+ Parkinson’s disease patients with orthostatic hypotension, PDOH− Parkinson’s disease patients without orthostatic hypotension, OCT optical coherence tomography, SRCP superficial retinal capillary plexus, DRCP deep retinal capillary plexus.aPD vs. control analyzed using the generalized estimating equation (GEE) analysis.bPDOH+, PDOH− and controls analyzed using the generalized estimating equation (GEE) analysis.cStatistically significant between the PDOH+ and control group.dStatistically significant between the PDOH+ and PDOH− groups.
The OCTA parameters of the participants are described in Table 2. The foveal avascular zone (FAZ) area was larger in the PD group than in the control group. The PD patients had a lower superficial retinal capillary plexus (SRCP) vessel density in the central region than the controls. In contrast, significant differences were not observed between the PD patients and controls in the temporal, inferior, nasal, or superior vessel densities in the SRCP and DRCP (Fig. 1b). To investigate the differences based on the presence or absence of OH in PD, we compared the OCTA parameters among the PDOH+, PDOH−, and control groups. The FAZ area was larger in the PDOH+ group than in the control group. The PDOH+ group showed a lower vessel density in the DRCP of the central region than the PDOH− and control groups. The central SRCP vessel density in the PDOH+ group was lower than in the control group but was not significant when compared with the PDOH− group. In contrast, none of the OCTA parameters showed significant differences between the PDOH− and control groups (Fig. 1c).Fig. 1Comparisons of OCTA parameters between patients with Parkinson’s disease and healthy controls. Schematic diagram and size of OCTA subregions (a). The Parkinson’s disease (PD) patients had lower superficial retinal capillary plexus (SRCP) vessel density in the central region than controls (b). The PD patients with orthostatic hypotension (PDOH+ group) showed a lower vessel density in the DRCP central region than the PD patients without orthostatic hypotension (PDOH−) and the control group (c). The SRCP vessel density in the central region in the PDOH+ group was lower than that in the control group but was non-significant compared with the PDOH− group (c). The asterisk indicates statistically significant. The hatched area represents statistical significance. The numbers represent the mean value of the vessel density. C center, T temporal quadrant, I inferior quadrant, N nasal quadrant, S superior quadrant.
## Relationship between the changes of blood pressure and central retinal vessel densities
To investigate the relationship between the central SRCP and central DRCP vessel densities and changes in blood pressure, we performed linear mixed model analysis. The results indicate that the central SRCP vessel density was not associated with changes in SBP (β = −0.054, standard error [SE] = 0.028, $$p \leq 0.218$$) or DBP (β = −0.128, SE = 0.058, $$p \leq 0.218$$). Central DRCP vessel density was associated with changes in SBP (β = −0.077, SE = 0.058, $$p \leq 0.019$$) and DBP (β = −0.180, SE = 0.057, $$p \leq 0.006$$) (Table 3 and Fig. 2).Fig. 2Correlations between the OCTA parameters and changes in blood pressure. The superficial retinal capillary plexus (DRCP) vessel density in the central region did not have a significant relationship with the changes in systolic blood pressure (SBP) and diastolic blood pressure (DBP) (a, b). The deep retinal capillary plexus (DRCP) vessel density in the central region had a negative correlation with changes in SBP and DBP (c, d).Table 3Results of the linear mixed model analysis. Beta coefficient (SE)$95\%$ CIp valueaSRCP centralΔSBP−0.054 (0.028)−0.108, 0.0010.218ΔDBP−0.128 (0.058)−0.014, −0.0140.113DRCP centralΔSBP−0.077 (0.027)−0.130, −0.0230.019ΔDBP−0.180 (0.057)−0.291, −0.0690.006Linear mixed model after adjustment for age, sex, disease duration, side of the eyes, spherical equivalent refractive errors, supine hypertension, hypertension, diabetes, and dyslipidemia. SE standard error, SRCP superficial retinal capillary plexus, DRCP deep retinal capillary plexus, ΔSBP change of systolic blood pressure during the head-up tilt test, ΔDBP change of diastolic blood pressure during the head-up tilt test.ap values were corrected for multiple comparisons using the Bonferroni correction.
## Discussion
This is the first study in which the change in OCTA parameters was investigated based on the presence of OH in PD patients. PD patients had significantly lower central retinal vessel densities than healthy controls, and the difference remained significant when comparing the PDOH+ group with the control group but was not significant in the PDOH− group. Furthermore, the amount of SBP and DBP changes correlated with the vessel density in the DRCP central area. These results indicate the central retinal vessel densities measured using OCTA are significantly associated with the presence of OH in PD patients.
Growing evidence shows that alterations of vasculature contribute to the development and aggravation of neurodegenerative diseases, including PD7,13,14. OCTA is an emerging non-invasive technology used to investigate in vivo retinal microvascular alterations in various neurological disorders15. Robbinson and colleagues suggested that OCTA parameters, especially the inner ring of the macula, are potential biomarkers for the diagnosis of PD9. In recent studies on PD, OCTA parameters were suggested as biomarkers for diagnosis and associated with disease progression and cognitive decline16,17. In contrast, Rascuna et al. reported that PD patients showed no significant differences in OCTA parameters in DRCP or SRCP compared with healthy controls10. When the OCTA parameters in PD, essential tremor, and healthy controls were compared, differences were not observed among the three groups11. These inconsistent results indicate that a range of factors can affect OCTA findings. Based on the results of the present study, the presence of OH is presumed to be a crucial factor associated with OCTA parameters in PD patients. When the PD patients were divided into two groups based on the presence or absence of OH, the PDOH+ group had lower retinal vessel densities in the SRCP and DRCP compared with the PDOH− and control groups (Fig. 2). The results indicated that central retinal microvascular changes could be associated with OH in addition to PD, and OCTA can be useful for their detection. In terms of the area of FAZ, it was larger in the PD group than in the control group. The difference was still significant when comparing the PDOH+ and the control groups, but it was not significant between the PDOH− and the control groups. In PD, several studies reported that there were no differences in FAZ area between patients and controls9,18–20, although two studies reported that PD patients had smaller FAZ area than controls16,21 Murueta-Goyena et al. suggested that smaller FAZ area was a distinguishable feature of PD patients, and it was negatively correlated with cognitive function21. Xu et al. suggested that foveal dopaminergic neuronal damage is an underlying mechanism of the reduction in FAZ area16. The patients included in the present study had normal cognitive function, milder motor symptoms, and lower LEDD compared with those of the previous studies21. The discrepancy between study results might be due to differences in baseline characteristics of the included patients, such as age distribution, PD phenotypes, disease severity, disease duration, cognitive function, and concomitant neurological or systemic conditions.
OH is a common non-motor symptom in PD that induces recurrent episodic cerebral and retinal hypoperfusion3. Recurrent episodic cerebral hypoperfusion may result in cerebral microvascular and white matter damage. PD patients with OH had increased white matter hyperintensities on brain magnetic resonance imaging (MRI) compared with the patients without OH22, and chronic cerebral hypoperfusion was associated with microvascular pathology in the brains of PD model mice23. In this context, OH may induce retinal microvascular damage in addition to cerebral microvascular damage. In terms of retinal vessels, microvascular damage may be associated with hypoperfusion induced by carotid artery stenosis as well as systemic hypoperfusion. For example, changes in OCTA metrics are associated with intradialytic hypotension episodes in chronic hemodialysis patients6. Furthermore, early changes in retinal microvasculature have predictive value regarding the development of systemic vascular disorders24,25. However, because a control group with OH and/or supine hypertension was not included in the present study, it remains unknown whether the changes are PD specific. Further studies that include controls with OH, supine hypertension, and patients with other diseases that show OH (drug-induced OH, pure autonomic failure, or multiple system atrophy) can aid in understanding the relationships among fluctuations in blood pressure, PD, and central retinal microvascular damage.
In the present study, significant changes were found in superficial and deep retinal vessel density only in the central macular area, indicating that the central retinal area is predominantly affected in patients with PD. The results are in agreement with previous studies in which the central retinal area was more greatly affected in PD patients9,12,18,19, indicating that the retinal vessels in the central area are more vulnerable to hypoperfusion and disease. The FAZ area, which is a central area highly sensitive to ischemic events such as diabetes26 and retinal vascular obstruction27, was increased in the PD group, particularly in the PDOH+ group. In addition, thinning of the macular inner retinal layers, an emerging biomarker of PD, reportedly occurs mainly in the parafoveal area in the early stages of PD28. This result supports the assumption that the central macular area is the most affected or vulnerable retinal area in PD patients28. However, there is a lack of evidence to explain the pathomechanism, and further research is needed to corroborate this hypothesis.
In addition, the increased reduction in SBP and DBP during the head-up tilt (HUT) test was significantly associated with lower vessel density in DRCP. The retinal artery consists of two parallel vascular networks. The superficial vascular plexus consists of approximately 75-µm diameter vessels supplied by the central retinal artery and smaller deep capillary plexuses (20-µm diameter) supplied by vertical anastomoses from the superficial vascular plexus29,30. The smaller vessels in DRCP may be more sensitive to hypoperfusion induced by OH than SRCP29. The relationship found between SBP, DBP, and retinal vessel density is in agreement with previous studies in which the association between BP and hypoperfusion or neuronal damage was analyzed. In previous studies, lower SBP and DBP were reportedly associated with the progression of normal tension glaucoma31, exacerbation of cerebral hypoperfusion and brain atrophy32, and leukoaraiosis33, and lower DBP also contributed to the development of dementia34–36. A similar mechanism could occur in the retina. In summary, both the SRCP and DRCP are affected by OH in PD. The DRCP is more likely to be affected, and the reduction of SBP and DBP might play an important role in damage to the microvasculature of the central retina in PD patients.
Increasing evidence shows the importance of the retina as a potential biomarker of early diagnosis and prognostication in PD28. In recent studies, parafoveal inner retinal change was shown to be detected in the early stages of PD, followed by progressive atrophy of pRNFL and macula over time28. In the present study, although the intraocular vessel density was significantly reduced in PD patients compared with controls, the pRNFL thickness and macular retinal thickness, which represent the degree of neuroaxonal damage, did not show a significant difference between PD patients and controls, even in the PDOH+ group. Rascuna et al. showed a positive correlation between retinal thicknesses (RNFL, ganglion cell layer, and inner plexiform layer) and microvascular density in the foveal region, and they suggested that microvascular change and macular atrophy reciprocally interact in PD37. In contrast, Robbinson and colleagues reported microvascular change without macular atrophy in PD patients9, suggesting microvascular change could be present prior to macular atrophy. Although macular atrophy is one of the most important potential causes of microvascular change, the results of the present study suggest that retinal microvascular dysfunction may occur primarily in PD rather than secondary to macular atrophy. Further large-scale studies on detailed retinal structures in various stages of the disease are needed to clarify the temporal relationship and precise mechanisms of early retinal and microvascular changes in PD.
The present study has several limitations. First, the HUT test was not performed in the control group; therefore, healthy controls with OH cannot be omitted, but healthy controls who had any neurological signs including dizziness, headache, or orthostatic symptoms were excluded. Second, various vascular risk factors were investigated in this study; however, carotid artery stenosis, obesity, and smoking history were not assessed. Third, the HUT test and OCTA were not performed on the same day, which might have influenced the results. Fourth, vessel density was measured using OCTA, in which the angiographic signal was based on movement. However, many other factors, such as blood flow velocity, morphology, and alterations in the vascular endothelial barrier, can compromise the measurement of perfusion. Therefore, false-positive findings cannot be excluded due to technical and methodological issues. Finally, the OCTA measurements in this study included large blood vessels. The changes in vessel density found in the present study were, therefore, a combination of changes in both the microvasculature and major vessels. The lack of statistical significance in the difference in parafoveal vessel densities between groups in this study could be attributed to those methodological limitations.
Recent research has demonstrated a strong correlation between the retina and PD, and there is growing interest in the role of retinal microvascular changes as a potential biomarker for the development and progression of PD. Despite its importance, in vivo examination of microvascular damage in PD patients is very limited. In the present study, the microvasculature damage in PD patients based on the presence of OH was investigated, and OH was a potentially critical factor associated with central retinal microvascular damage in PD. The results showed that central retinal microvascular damage measured using OCTA occurs prior to the development of macular thinning and pRNFL. Based on the results, OCTA can be a useful non-invasive method for detecting central retinal microvascular damage in PD patients.
## Participants and clinical assessment
Participants were recruited from the movement disorder clinic of the Samsung Medical Center. The Institutional Review Board of Samsung Medical Center approved this study, and all subjects provided written informed consent. Patients were enrolled if they were diagnosed with PD based on the United Kingdom Brain Bank Criteria for PD38. Patients with any of the following conditions were excluded: any neurologic disorder other than PD, systemic vasculitis, cardiovascular disease, musculoskeletal disease, end-stage renal disease, peripheral nervous system autonomic failure (diabetic neuropathy, Guillain-Barre syndrome, amyloid neuropathy, surgical sympathectomy, and pheochromocytoma, etc.), ocular pathology that could affect OCTA measurements (glaucoma, a refractive error >+6.0 diopters of spherical equivalent or <−6.0 diopters of spherical equivalent, astigmatism ≥ 3.0 diopters, epiretinal membrane, age-related macular degeneration, diabetic retinopathy, hypertensive retinopathy, retinal artery/vein occlusion, or optic neuropathy) or previous retinal surgery. Exact age- and sex-matched controls were recruited. The healthy controls were required to have normal visual acuity, normal intraocular pressure ≤21 mm Hg, and normal optic discs. The same exclusion criteria were applied to healthy controls and PD patients. Demographic and clinical data, including age, sex, and comorbid vascular risk factors (hypertension, diabetes mellitus, dyslipidemia), were collected for all enrolled participants. The UPDRS III39, H&Y scale40, LEDD41, and MoCA42 were investigated in all enrolled PD patients.
## OCT and OCTA
All included patients and healthy subjects underwent spectral-domain OCT (SD-OCT; Spectralis, Heidelberg Engineering, Heidelberg, Germany) that provided 40,000 A-scans per second with 7-μm optical and 3.5-μm digital axial resolution. An internal fixation target was used, and the patient’s other eye was covered during scanning. OCT peripapillary RNFL circular scans centered on the optic disc of each patient were obtained. In addition, macular thickness measurements in the central 1-mm area and in each quadrant in the 3- and 6-mm areas were obtained. Swept-source OCT (DRI OCT Triton Plus; Topcon Corporation, Tokyo, Japan) coupled with non-invasive OCTA technology was also completed for all PD patients and healthy subjects. The details have been previously described43,44. The SRCP slab was automatically segmented from 3 µm under the internal limiting membrane (ILM) to 15 µm below the IPL, and the DRCP slab was automatically segmented from 15 to 70 µm under the IPL following a formerly corroborated method by Park et al.45. The radial peripapillary capillary (RPC) segment ranged from the ILM to the posterior boundary of the RNFL. Vessel density was determined as the percentage of the total area occupied by vessels and microvasculature, quantitatively expressed as color-coded vessels in a localized region that was obtained by automatically applying an Early Treatment Diabetic Retinopathy Study (ETDRS) grid overlay containing the two inner rings of the ETDRS grid pattern to the fovea, which yielded a calculation of the density in each layer. The parafoveal ring divided the macular region into the temporal, nasal, inferior, and superior sections (Fig. 3). All participants completed both OCT and OCTA imaging within 1 day. The software generated TopQ image quality values for each OCTA scan and vessel density measurement. To assess scan quality, we included the scan images based on the quality assessment criteria suggested by Fenner et al.46. Expert graders reviewed and verified all images (D.L. and K.-A.P.). In patients with PD and healthy controls, bilateral eyes were analyzed except OCTA images, which are difficult to analyze due to motion artifacts or incomplete acquisitions. Fig. 3Representative fundus photography and OCTA images of the macular area in a healthy control. Macular OCTA measurement was obtained by automatically applying an Early Treatment Diabetic Retinopathy Study (ETDRS) grid overlay containing the two inner rings of the ETDRS grid pattern to the fovea, which yielded the vessel density in each layer. C center, T temporal quadrant, I inferior quadrant, N nasal quadrant, S superior quadrant.
## Head-up tilt test
Participants with PD underwent the HUT test. The patients discontinued medications that could affect the HUT test results at least 24 h before the test. In addition, the participants were prohibited from smoking and drinking beverages containing caffeine on the day of the test. The electrode and BP cuff were attached to the patient, and BP was continuously recorded using Finometer 1 (FMS, Amsterdam, the Netherlands) after a period of 5 min of supine rest. OH was defined as a reduction of at least 20 mmHg in SBP or a 10 mmHg fall in DBP within 3 min of HUT testing47. In patients with supine hypertension, a reduction in SBP of 30 mm Hg was applied48.
## Statistical analysis
The normality of the data was evaluated using the Shapiro–Wilk test. Clinical and demographic features were presented using mean and standard deviation (SD). Differences among the three groups were determined using the Student’s t test, analysis of variance (ANOVA), or chi-square test. The OCTA and OCT parameters of the participants were compared using a generalized estimating equation (GEE) analysis with an exchangeable correlation structure to account for the inclusion of both eyes from the same individual. The following variables were also included as confounding factors for the GEE: age, sex, SER, supine hypertension, hypertension, diabetes, and dyslipidemia. Linear mixed model analysis was performed to investigate the association between the OCTA parameters and changes in blood pressure after adjustment for age, sex, side of the eye, disease duration, SER, and vascular risk factors. Four models were established for two dependent variables (central vessel density of the SRCP or central vessel density of the DRCP) and two independent variables (changes in SBP or DBP during the HUT test) because of the collinearity between the variables. Bonferroni correction was applied for multiple comparisons. Results were considered significant if the p value was ≤0.05. Statistical analysis was performed with IBM SPSS (version 28.0; IBM Inc., USA) software for Windows.
## References
1. Bae HJ, Cheon SM, Kim JW. **Orthostatic hypotension in drug-naïve patients with Parkinson’s disease**. *J. Mov. Disord.* (2011.0) **4** 33-37. DOI: 10.14802/jmd.11005
2. McDonald C, Newton JL, Burn DJ. **Orthostatic hypotension and cognitive impairment in Parkinson’s disease: causation or association?**. *Mov. Disord.* (2016.0) **31** 937-946. DOI: 10.1002/mds.26632
3. Wieling W. **Symptoms and signs of syncope: a review of the link between physiology and clinical clues**. *Brain* (2009.0) **132** 2630-2642. DOI: 10.1093/brain/awp179
4. Xing Y. **Impaired cerebral autoregulation in Parkinson’s disease: an orthostatic hypotension analysis**. *Front. Neurol.* (2022.0) **13** 811698. DOI: 10.3389/fneur.2022.811698
5. de Carlo TE, Romano A, Waheed NK, Duker JS. **A review of optical coherence tomography angiography (OCTA)**. *Int. J. Retina Vitreous* (2015.0) **1** 5. DOI: 10.1186/s40942-015-0005-8
6. Coppolino G. **OCT angiography metrics predict intradialytic hypotension episodes in chronic hemodialysis patients: a pilot, prospective study**. *Sci. Rep.* (2021.0) **11** 7202. DOI: 10.1038/s41598-021-86609-0
7. Paul G, Elabi OF. **Microvascular changes in Parkinson’s disease- focus on the neurovascular unit**. *Front. Aging Neurosci.* (2022.0) **14** 853372. DOI: 10.3389/fnagi.2022.853372
8. Kwapong WR. **Retinal microvascular impairment in the early stages of Parkinson’s disease**. *Invest. Ophthalmol. Vis. Sci.* (2018.0) **59** 4115-4122. DOI: 10.1167/iovs.17-23230
9. Robbins CB. **Characterization of retinal microvascular and choroidal structural changes in Parkinson disease**. *JAMA Ophthalmol.* (2021.0) **139** 182-188. DOI: 10.1001/jamaophthalmol.2020.5730
10. Rascuna C. **Retinal thickness and microvascular pattern in early Parkinson’s disease**. *Front. Neurol.* (2020.0) **11** 533375. DOI: 10.3389/fneur.2020.533375
11. 11.Satue, M. et al. Ability of Swept-source OCT and OCT-angiography to detect neuroretinal and vasculature changes in patients with Parkinson disease and essential tremor. Eye10.1038/s41433-022-02112-4 (2022).
12. Shi C. **Characterization by fractal dimension analysis of the retinal capillary network in Parkinson disease**. *Retina* (2020.0) **40** 1483-1491. DOI: 10.1097/IAE.0000000000002641
13. Potashkin J. **Understanding the links between cardiovascular disease and Parkinson’s disease**. *Mov. Disord.* (2020.0) **35** 55-74. DOI: 10.1002/mds.27836
14. Guan J. **Vascular degeneration in Parkinson’s disease**. *Brain Pathol.* (2013.0) **23** 154-164. DOI: 10.1111/j.1750-3639.2012.00628.x
15. Wang L, Murphy O, Caldito NG, Calabresi PA, Saidha S. **Emerging applications of optical coherence tomography angiography (OCTA) in neurological research**. *Eye Vis.* (2018.0) **5** 11. DOI: 10.1186/s40662-018-0104-3
16. Xu B. **Retinal microvascular density was associated with the clinical progression of Parkinson’s disease**. *Front. Aging Neurosci.* (2022.0) **14** 818597. DOI: 10.3389/fnagi.2022.818597
17. Li Y. **Retinal microvascular impairment in Parkinson’s disease with cognitive dysfunction**. *Parkinsonism Relat. Disord.* (2022.0) **98** 27-31. DOI: 10.1016/j.parkreldis.2022.03.008
18. Zhou M. **Visual impairments are associated with retinal microvascular density in patients with Parkinson’s disease**. *Front. Neurosci.* (2021.0) **15** 718820. DOI: 10.3389/fnins.2021.718820
19. Zou J. **Combination of optical coherence tomography (OCT) and OCT angiography increases diagnostic efficacy of Parkinson’s disease**. *Quant. Imaging Med. Surg.* (2020.0) **10** 1930-1939. DOI: 10.21037/qims-20-460
20. Lin CW, Lai TT, Chen SJ, Lin CH. **Elevated alpha-synuclein and NfL levels in tear fluids and decreased retinal microvascular densities in patients with Parkinson’s disease**. *Geroscience* (2022.0) **44** 1551-1562. DOI: 10.1007/s11357-022-00576-6
21. Murueta-Goyena A. **Foveal remodeling of retinal microvasculature in Parkinson’s disease**. *Front. Neurosci.* (2021.0) **15** 708700. DOI: 10.3389/fnins.2021.708700
22. Oh YS, Kim JS, Lee KS. **Orthostatic and supine blood pressures are associated with white matter hyperintensities in Parkinson disease**. *J. Mov. Disord.* (2013.0) **6** 23-27. DOI: 10.14802/jmd.13006
23. Tang H. **Chronic cerebral hypoperfusion independently exacerbates cognitive impairment within the pathopoiesis of Parkinson’s disease via microvascular pathologys**. *Behav. Brain Res.* (2017.0) **333** 286-294. DOI: 10.1016/j.bbr.2017.05.061
24. Lee WH. **Retinal microvascular change in hypertension as measured by optical coherence tomography angiography**. *Sci. Rep.* (2019.0) **9** 156. DOI: 10.1038/s41598-018-36474-1
25. Carnevali A. **Optical coherence tomography angiography analysis of retinal vascular plexuses and choriocapillaris in patients with type 1 diabetes without diabetic retinopathy**. *Acta Diabetol.* (2017.0) **54** 695-702. DOI: 10.1007/s00592-017-0996-8
26. Tang FY. **Determinants of quantitative optical coherence tomography angiography metrics in patients with diabetes**. *Sci. Rep.* (2017.0) **7** 2575. DOI: 10.1038/s41598-017-02767-0
27. Koulisis N. **Quantitative microvascular analysis of retinal venous occlusions by spectral domain optical coherence tomography angiography**. *PLoS ONE* (2017.0) **12** e0176404. DOI: 10.1371/journal.pone.0176404
28. Lee JY. **Multimodal brain and retinal imaging of dopaminergic degeneration in Parkinson disease**. *Nat. Rev. Neurol.* (2022.0) **18** 203-220. DOI: 10.1038/s41582-022-00618-9
29. Savastano MC, Lumbroso B, Rispoli M. **In vivo characterization of retinal vascularization morphology using optical coherence tomography angiography**. *Retina* (2015.0) **35** 2196-2203. DOI: 10.1097/IAE.0000000000000635
30. Campbell JP. **Detailed vascular anatomy of the human retina by projection-resolved optical coherence tomography angiography**. *Sci. Rep.* (2017.0) **7** 42201. DOI: 10.1038/srep42201
31. Okumura Y, Yuki K, Tsubota K. **Low diastolic blood pressure is associated with the progression of normal-tension glaucoma**. *Ophthalmologica* (2012.0) **228** 36-41. DOI: 10.1159/000335978
32. Monteiro A, Castro P, Azevedo E, Zorond F, Polonia J. **Low ambulatory diastolic blood pressure is associated with brain atrophy in strongly treated hypertensive patients**. *J. Hypertens.* (2019.0) **37** e281. DOI: 10.1097/01.hjh.0000573580.96638.4a
33. Matsubayashi K. **Postural dysregulation in systolic blood pressure is associated with worsened scoring on neurobehavioral function tests and leukoaraiosis in the older elderly living in a community**. *Stroke* (1997.0) **28** 2169-2173. DOI: 10.1161/01.STR.28.11.2169
34. Verghese J, Lipton RB, Hall CB, Kuslansky G, Katz MJ. **Low blood pressure and the risk of dementia in very old individuals**. *Neurology* (2003.0) **61** 1667-1672. DOI: 10.1212/01.WNL.0000098934.18300.BE
35. van Dalen JW. **Association of systolic blood pressure with dementia risk and the role of age, u-shaped associations, and mortality**. *JAMA Intern. Med.* (2022.0) **182** 142-152. DOI: 10.1001/jamainternmed.2021.7009
36. Duschek S, Schandry R. **Reduced brain perfusion and cognitive performance due to constitutional hypotension**. *Clin. Auton. Res.* (2007.0) **17** 69-76. DOI: 10.1007/s10286-006-0379-7
37. Rascuna C. **Retinal thickness and microvascular pathway in idiopathic rapid eye movement sleep behaviour disorder and Parkinson’s disease**. *Parkinsonism Relat. Disord.* (2021.0) **88** 40-45. DOI: 10.1016/j.parkreldis.2021.05.031
38. Hughes AJ, Daniel SE, Kilford L, Lees AJ. **Accuracy of clinical diagnosis of idiopathic Parkinson’s disease: a clinico-pathological study of 100 cases**. *J. Neurol. Neurosurg. Psychiatry* (1992.0) **55** 181-184. DOI: 10.1136/jnnp.55.3.181
39. **The Unified Parkinson’s Disease Rating Scale (UPDRS): status and recommendations**. *Mov. Disord.* (2003.0) **18** 738-750. DOI: 10.1002/mds.10473
40. Hoehn MM, Yahr MD. **Parkinsonism: onset, progression and mortality**. *Neurology* (1967.0) **17** 427-442. DOI: 10.1212/WNL.17.5.427
41. Schade S, Mollenhauer B, Trenkwalder C. **Levodopa equivalent dose conversion factors: an updated proposal including opicapone and safinamide**. *Mov. Disord. Clin. Pract.* (2020.0) **7** 343-345. DOI: 10.1002/mdc3.12921
42. Nasreddine ZS. **The Montreal Cognitive Assessment, MoCA: a brief screening tool for mild cognitive impairment**. *J. Am. Geriatr. Soc.* (2005.0) **53** 695-699. DOI: 10.1111/j.1532-5415.2005.53221.x
43. Lee GI. **Parafoveal and peripapillary vessel density in pediatric and juvenile craniopharyngioma patients**. *Sci. Rep.* (2022.0) **12** 5355. DOI: 10.1038/s41598-022-09391-7
44. Lee GI, Park KA, Oh SY, Kong DS. **Changes in parafoveal and peripapillary perfusion after decompression surgery in chiasmal compression due to pituitary tumors**. *Sci. Rep.* (2021.0) **11** 3464. DOI: 10.1038/s41598-021-82151-1
45. Park JJ, Soetikno BT, Fawzi AA. **Characterization of the middle capillary plexus using optical coherence tomography angiography in healthy and diabetic eyes**. *Retina* (2016.0) **36** 2039-2050. DOI: 10.1097/IAE.0000000000001077
46. Fenner BJ. **Identification of imaging features that determine quality and repeatability of retinal capillary plexus density measurements in OCT angiography**. *Br. J. Ophthalmol.* (2018.0) **102** 509-514. DOI: 10.1136/bjophthalmol-2017-310700
47. Freeman R. **Consensus statement on the definition of orthostatic hypotension, neurally mediated syncope and the postural tachycardia syndrome**. *Clin. Auton. Res.* (2011.0) **21** 69-72. DOI: 10.1007/s10286-011-0119-5
48. Fanciulli A. **Consensus statement on the definition of neurogenic supine hypertension in cardiovascular autonomic failure by the American Autonomic Society (AAS) and the European Federation of Autonomic Societies (EFAS): endorsed by the European Academy of Neurology (EAN) and the European Society of Hypertension (ESH)**. *Clin. Auton. Res.* (2018.0) **28** 355-362. DOI: 10.1007/s10286-018-0529-8
|
---
title: Humans with inherited MyD88 and IRAK-4 deficiencies are predisposed to hypoxemic
COVID-19 pneumonia
authors:
- Ana García-García
- Rebeca Pérez de Diego
- Carlos Flores
- Darawan Rinchai
- Jordi Solé-Violán
- Àngela Deyà-Martínez
- Blanca García-Solis
- José M. Lorenzo-Salazar
- Elisa Hernández-Brito
- Anna-Lisa Lanz
- Leen Moens
- Giorgia Bucciol
- Mohamed Almuqamam
- Joseph B. Domachowske
- Elena Colino
- Juan Luis Santos-Perez
- Francisco M. Marco
- Claudio Pignata
- Aziz Bousfiha
- Stuart E. Turvey
- Stefanie Bauer
- Filomeen Haerynck
- Javier Gonzalo Ocejo-Vinyals
- Francisco Lendinez
- Seraina Prader
- Nora Naumann-Bartsch
- Jana Pachlopnik Schmid
- Catherine M. Biggs
- Kyla Hildebrand
- Alexandra Dreesman
- Miguel Ángel Cárdenes
- Fatima Ailal
- Ibtihal Benhsaien
- Giuliana Giardino
- Agueda Molina-Fuentes
- Claudia Fortuny
- Swetha Madhavarapu
- Daniel H. Conway
- Carolina Prando
- Laire Schidlowski
- María Teresa Martínez de Saavedra Álvarez
- Rafael Alfaro
- Felipe Rodríguez de Castro
- Gerhard Kindle
- Gerhard Kindle
- Nizar Mahlaoui
- Markus G. Seidel
- Lougaris Vassilios
- Mikko R.J. Seppänen
- Laurent Abel
- Laurent Abel
- Alessandro Aiuti
- Saleh Al-Muhsen
- Fahd Al-Mulla
- Mark S. Anderson
- Evangelos Andreakos
- Andrés A. Arias
- Hagit Baris Feldman
- Alexandre Belot
- Catherine M. Biggs
- Dusan Bogunovic
- Alexandre Bolze
- Anastasiia Bondarenko
- Ahmed A. Bousfiha
- Petter Brodin
- Yenan Bryceson
- Carlos D. Bustamante
- Manish J. Butte
- Giorgio Casari
- John Christodoulou
- Antonio Condino-Neto
- Stefan N. Constantinescu
- Megan A. Cooper
- Clifton L. Dalgard
- Murkesh Desai
- Beth A. Drolet
- Jamila El Baghdadi
- Sara Espinosa-Padilla
- Jacques Fellay
- Carlos Flores
- José Luis Franco
- Antoine Froidure
- Peter K. Gregersen
- Bodo Grimbacher
- Filomeen Haerynck
- David Hagin
- Rabih Halwani
- Lennart Hammarström
- James R. Heath
- Sarah E. Henrickson
- Elena W.Y. Hsieh
- Eystein Husebye
- Kohsuke Imai
- Yuval Itan
- Erich D. Jarvis
- Timokratis Karamitros
- Kai Kisand
- Cheng-Lung Ku
- Yu-Lung Lau
- Yun Ling
- Carrie L. Lucas
- Tom Maniatis
- Davood Mansouri
- László Maródi
- Isabelle Meyts
- Joshua D. Milner
- Kristina Mironska
- Trine H. Mogensen
- Tomohiro Morio
- Lisa F.P. Ng
- Luigi D. Notarangelo
- Antonio Novelli
- Giuseppe Novelli
- Cliona O'Farrelly
- Satoshi Okada
- Keisuke Okamoto
- Tayfun Ozcelik
- Qiang Pan-Hammarström
- Jean W. Pape
- Rebecca Perez de Diego
- David S. Perlin
- Graziano Pesole
- Anna M. Planas
- Carolina Prando
- Aurora Pujol
- Lluis Quintana-Murci
- Sathishkumar Ramaswamy
- Laurent Renia
- Igor Resnick
- Carlos Rodríguez-Gallego
- Vanessa Sancho-Shimizu
- Anna Sediva
- Mikko R.J. Seppänan
- Mohammed Shahrooei
- Anna Shcherbina
- Ondrej Slaby
- Andrew L. Snow
- Pere Soler-Palacín
- András N. Spaan
- Ivan Tancevski
- Stuart G. Tangye
- Ahmad Abou Tayoun
- Stuart E. Turvey
- K M Furkan Uddin
- Mohammed J. Uddin
- Diederik van de Beek
- Donald C. Vinh
- Horst von Bernuth
- Joost Wauters
- Mayana Zatz
- Pawel Zawadzki
- Helen C. Su
- Jean-Laurent Casanova
- Isabelle Meyts
- Fabian Hauck
- Anne Puel
- Paul Bastard
- Bertrand Boisson
- Emmanuelle Jouanguy
- Laurent Abel
- Aurélie Cobat
- Qian Zhang
- Jean-Laurent Casanova
- Laia Alsina
- Carlos Rodríguez-Gallego
journal: The Journal of Experimental Medicine
year: 2023
pmcid: PMC9998661
doi: 10.1084/jem.20220170
license: CC BY 4.0
---
# Humans with inherited MyD88 and IRAK-4 deficiencies are predisposed to hypoxemic COVID-19 pneumonia
## Abstract
MyD88- and IRAK-4–deficient patients have a higher risk of hypoxemic COVID-19 pneumonia than individuals of similar age in the general population, due to impaired TLR7-dependent type I IFN production.
X-linked recessive deficiency of TLR7, a MyD88- and IRAK-4–dependent endosomal ssRNA sensor, impairs SARS-CoV-2 recognition and type I IFN production in plasmacytoid dendritic cells (pDCs), thereby underlying hypoxemic COVID-19 pneumonia with high penetrance. We report 22 unvaccinated patients with autosomal recessive MyD88 or IRAK-4 deficiency infected with SARS-CoV-2 (mean age: 10.9 yr; 2 mo to 24 yr), originating from 17 kindreds from eight countries on three continents. 16 patients were hospitalized: six with moderate, four with severe, and six with critical pneumonia, one of whom died. The risk of hypoxemic pneumonia increased with age. The risk of invasive mechanical ventilation was also much greater than in age-matched controls from the general population (OR: 74.7, $95\%$ CI: 26.8–207.8, $P \leq 0.001$). The patients’ susceptibility to SARS-CoV-2 can be attributed to impaired TLR7-dependent type I IFN production by pDCs, which do not sense SARS-CoV-2 correctly. Patients with inherited MyD88 or IRAK-4 deficiency were long thought to be selectively vulnerable to pyogenic bacteria, but also have a high risk of hypoxemic COVID-19 pneumonia.
## Graphical Abstract
## Introduction
Less than $10\%$ of individuals infected with SARS-CoV-2 develop hypoxemic COVID-19 pneumonia, which may be severe (about $7\%$) or critical ($3\%$) (Zhang et al., 2022a). Age is the major epidemiological risk factor for hospitalization or death from COVID-19 pneumonia, the risk doubling with every 5 yr of age, from childhood onwards (Zhang et al., 2020a; Knock et al., 2021; Le Vu et al., 2021; O’Driscoll et al., 2021; Sah et al., 2021). The infection fatality rate in unvaccinated individuals is $0.001\%$ at 5 yr of age and $10\%$ at 85 yr of age (a 10,000-fold increase; Bennett et al., 2021; Knock et al., 2021; Le Vu et al., 2021; Navaratnam et al., 2021; O’Driscoll et al., 2021). Most children, adolescents, and young adults with SARS-CoV-2 infection are asymptomatic or present a benign upper respiratory tract disease (Brotons et al., 2021; Chua et al., 2021; Mantovani et al., 2020; Schober et al., 2022; Woodruff et al., 2022). The proportion of asymptomatic infections is greater in children than in adults (Sah et al., 2021). However, interindividual clinical variability remains vast, for all age categories. Various comorbid conditions operate as modest risk factors, with odds ratios (ORs) typically <1.5 and always <2. Men have a 1.5× higher risk of death than women, after correction for other risk factors (Zhang et al., 2020a, 2022a; Brodin, 2021). Likewise, the contribution of common genetic variants detected by genome-wide association studies is modest, the most robustly associated region being a Neanderthal haplotype on chromosome 3 conferring predisposition with an OR around 2 (2.7 in patients <60 yr and 1.5 in patients >60 yr; Nakanishi et al., 2021).
A first molecular explanation for critical COVID-19 pneumonia was provided by inborn errors of TLR3- and/or TLR7-dependent type I IFN immunity, including autosomal recessive (AR) IRF7 and IFNAR1 deficiencies, in about 1–$5\%$ of patients, this proportion being lower for individuals over 60 yr of age (Zhang et al., 2020a, 2020b; Asano et al., 2021). This led to the discovery of pre-existing autoantibodies (auto-Abs) against type I IFNs in about 15–$20\%$ of patients, with a higher proportion in patients over 70 yr of age (Bastard et al., 2021b, 2022; Bourgeois et al., 2021; Solanich et al., 2021b) and in patients with “breakthrough” hypoxemic COVID-19 pneumonia whose response to RNA vaccines was normal (Bastard et al., 2022). In particular, we found X-linked recessive (XR) TLR7 deficiency in about $1.8\%$ of male patients below the age of 60 yr and in $8.9\%$ of boys (<16 yr) in the COVID Human Genetic Effort (CHGE) consortium cohort (https://www.covidhge.com; Asano et al., 2021; Zhang et al., 2022b). The proportion of patients with combined AR and XR inborn errors of type I IFNs is particularly high in children in this cohort, accounting for $10\%$ of cases of hospitalization for COVID-19 pneumonia (Zhang et al., 2022b). TLR7 is a MyD88/IRAK-4–dependent endosomal receptor for single-stranded RNA in blood plasmacytoid dendritic cells (pDCs), which do not express TLR3 (Asano et al., 2021; Beutler, 2004; Diebold et al., 2004; Reizis, 2019). Conversely, TLR3 is an endosomal receptor of dsRNA in tissue respiratory epithelial cells (RECs), which do not express TLR7 (Zhang et al., 2007a; Guo et al., 2011; Kyung Lim et al., 2019). *This* genetic approach therefore suggested that both pDCs and tissue respiratory epithelial cells are crucial for type I IFN immunity to SARS-CoV-2 in the respiratory tract (Asano et al., 2021; Casanova and Abel, 2021, 2022; Zhang et al., 2022a).
MyD88 and IRAK-4 are crucial for signaling through the canonical Toll/IL-1 receptor pathway mediated by the 10 human TLRs (including TLR7) other than TLR3, and the IL-1Rs, IL-18R and IL-33R (Beutler, 2004; Kawai and Akira, 2011). Human-inherited MyD88 and IRAK-4 deficiencies are immunological and clinical phenocopies (Alsina et al., 2014; Picard et al., 2010). Affected patients are particularly prone to invasive staphylococcal and pneumococcal bacterial infections in childhood. However, infections become rarer after adolescence (Ku et al., 2007; von Bernuth et al., 2008, 2012, Picard et al., 2010, 2011). Remarkably, due to the abolition of responses driven by TLRs except TLR3, and by all IL-1Rs (in response to all IL-1 paralogs, IL-18, and IL-33), clinical and laboratory signs of inflammation develop slowly in these patients, even during bacterial disease (Picard et al., 2010). Surprisingly, unusually severe viral, fungal, and parasitic diseases have been reported only rarely in patients with MyD88 or IRAK-4 deficiency (Bucciol et al., 2022a; Nishimura et al., 2021; Picard et al., 2010; Tepe et al., 2022; Yang et al., 2005; Zhang et al., 2007a). The only virus reported to cause disease in more than one patient to date is HHV6 (Nishimura et al., 2021; Tepe et al., 2022). The apparent lack of severe viral disease in other known patients is particularly intriguing, as responses to TLR7, TLR8, and TLR9 were abolished in the cells of these patients (Casanova et al., 2011; Yang et al., 2005). Moreover, the four loci encoding the endosomal TLRs sensing nucleic acids—TLR3, TLR7, TLR8, and TLR9—are under stronger negative selection than other TLR loci (Quach et al., 2013).
Indeed, MyD88- and IRAK-4–deficient cells, including fibroblasts and leukocytes, do not produce type I IFN in response to the endosomal TLR7, TLR8, and TLR9 nucleic acid sensors, but they respond normally to TLR3 stimulation (Von Bernuth et al., 2008; Yang et al., 2005). Until recently, the production of type I IFN in response to specific viruses had never been tested for pDCs from patients with MyD88 or IRAK-4 deficiency. pDCs from an IRAK-4–deficient patient were recently shown not to produce type I IFN in response to SARS-CoV-2 (Onodi et al., 2021). Similar findings were obtained for an UNC-93B–deficient patient, whose cells failed to respond to TLR3, TLR7, TLR8, and TLR9 stimulation (Casrouge et al., 2006; Onodi et al., 2021). Nevertheless, the clinical impact of SARS-CoV-2 infection in patients with MyD88 or IRAK-4 deficiency is unclear. Only brief reports have emerged, of three patients in meeting abstracts (Mahmood et al., 2021) or five patients in case reports (Bucciol et al., 2022a; Deyà-Martínez et al., 2021; Goudouris et al., 2021; Milito et al., 2021), describing clinical phenotypes ranging from moderate to critical pneumonia. These patients would be predicted to be prone to hypoxemic COVID-19 pneumonia, due to their complete lack of TLR7-dependent SARS-CoV-2 sensing by pDCs (Asano et al., 2021; Onodi et al., 2021). However, their lack of TLR- and IL-1R–mediated inflammation might, perhaps, mitigate this vulnerability to some extent. We report here the natural course of SARS-CoV-2 infection in these and other patients, for a total of 22 patients from 17 kindreds and eight countries on three continents.
## Patients with inherited MyD88 or IRAK-4 deficiency
Following an international call for collaboration, we obtained data for 22 patients from 17 kindreds with inherited MyD88 (15 patients) or IRAK-4 (7 patients) deficiency, from Spain (7 patients and 6 kindreds), the USA (6 and 3), Belgium (2 and 2), Germany (2 and 2), Canada (2 and 1), Morocco [1], Italy [1], and Switzerland [1], all of whom were infected with SARS-CoV-2 before vaccination (Fig. 1 and Tables 1 and S1). SARS-CoV-2 infection was diagnosed by real-time PCR (RT-PCR; 18 patients) or antigenic assays (3 patients) on nasal swabs after the patients came into contact with a case and/or respiratory clinical manifestations had emerged, or by a serological test in one asymptomatic patient (P11) during routine hospital screening (Table S1). All 10 patients tested were seropositive for SARS-CoV-2–specific IgG/M 4–70 d after infection (Table S1). All but 2 of the 22 patients (the exceptions being P14 and P15) were known to suffer from MyD88 or IRAK-4 or deficiency before the start of the COVID-19 pandemic. All had an AR, complete deficiency, and the genotypes of 10 patients have been reported elsewhere (Yang et al., 2005; Cardenes et al., 2006; von Bernuth et al., 2006, 2008; Ku et al., 2007; Conway et al., 2010; Picard et al., 2010; Weller et al., 2012; Jia et al., 2020; Bucciol et al., 2022a). The IRAK4 or MYD88 genotypes of the remaining 12 patients are reported here (Fig. 1). The 22 patients were aged 2 mo to 24 yr (mean: 10.9 ± 6.8 yr). There were 18 ($81.8\%$) male patients and 4 ($18.2\%$) female patients (Fig. 1). 16 patients were receiving prophylaxis at the time of COVID-19 infection: 11 were on oral antibiotics, and 15 were on IgG replacement therapy (IgRT), with 10 patients on both (Table S1). Previous viral infections in these patients are summarized in Tables S2 and S3.
**Figure 1.:** *AR MyD88 and IRAK-4 deficiencies and SARS-CoV-2 infection in 17 kindreds. Pedigrees of the 17 kindreds containing 7 IRAK-4– and 15 MyD88-deficient patients with SARS-CoV-2 infection (P1–P22 are shown). Patients are identified by the number within the symbol for the individual concerned. The mutations are indicated above each pedigree, and the genotype of each individual is identified below the symbol (M, mutation; ND, no data). Kindred N contains five healthy sisters, indicated by the number “5” within the circle symbol.* TABLE_PLACEHOLDER:Table 1.
## Clinical manifestations of SARS-CoV-2 infection in the patients
16 of the 22 patients were hospitalized for pneumonia, as confirmed by x ray or computed tomography (CT) scan, including six with moderate pneumonia and 10 with hypoxemic pneumonia, which was critical and required admission to an intensive care unit (ICU) in six cases. Six patients had silent or mild infection, including one with x-ray data (P4, hospitalized as a precaution) and five without x-ray or CT scan data. None of the patients infected with the Omicron variant suffered from pneumonia, whereas patients infected with other viral variants had a wide spectrum of clinical manifestations, ranging from silent infection to death, indicating that the viral variant involved had a major impact on infection severity (Fig. 2 A). RT-PCR cycle threshold (Ct) values were available for 10 patients and were highly variable, even in patients with similar disease severity (Table S1). This is not surprising, because Ct values are dependent on both the RT-PCR platform and the RT-PCR assay used, and on the SARS-CoV-2 genomic variant (Amorim et al., 2022; Fomenko et al., 2022; Kogoj et al., 2022; Mohsin and Mahmud, 2022; Ong et al., 2022). Sex had no major impact on the severity of infection, as $44.4\%$ of male patients and $50\%$ of female patients developed hypoxemic pneumonia (Fig. 2 B). By contrast, age had a major effect on disease severity in patients with MyD88/IRAK-4 deficiency. Indeed, $60\%$ of patients under the age of 8 yr had mild infections, without pneumonia, whereas all patients over the age of 8 yr had pneumonia, which was hypoxemic in $50\%$ of cases, with one death (P17, 23 yr; Fig. 2 C). The chances of developing hypoxemic pneumonia were, therefore, much higher in older patients. Thus, the risk factors for the general population, including viral variant and age, also had a detectable impact on the severity of COVID-19 pneumonia in MyD88- or IRAK-4–deficient patients. The penetrance of hypoxemic pneumonia was higher in older patients infected with more virulent variants than in younger patients infected with the Omicron variant.
**Figure 2.:** *Severity of SARS-CoV-2 infection in patients with MyD88 or IRAK-4 deficiencies associated with risk factors. (A) Severity of the infections associated with viral variants. (B) Severity of infection as a function of sex. (C) Severity of infection as a function of age.*
## Clinical course of disease in the 16 patients with COVID-19 pneumonia
The time from first symptoms to hospital admission for these 16 patients was 3.3 ± 1.7 d. For the 12 patients for whom the mode of transmission was known (Table S1), the time from first contact with an individual with confirmed SARS-CoV-2 infection to hospitalization was 8.3 ± 4.7 d. The mean duration of hospitalization for patients with moderate, severe, or critical pneumonia was 5.0 (range: 3–7), 8.8 (range: 5–11), and 31.3 (range: 7–94) d, respectively. The six patients with critical pneumonia had a mean duration of ICU stay of 16.3 d (range: 1–28 d; Table 1). Mean time from first symptoms to oxygen therapy in our 10 patients with hypoxemic pneumonia was 4.0 (range: 2–7) d, and the patients were on oxygen therapy for 16.2 (range: 3–94) d (3.3 d, range: 3–4 d for severe pneumonia vs. 24.8 d, range: 5–94 d for critical pneumonia, $$P \leq 0.013$$). The 16 patients with confirmed pneumonia received antibiotics; 11 patients received systemic glucocorticoids, 7 received remdesivir, 1 received tenofovir, 4 received hydroxychloroquine, 2 received tocilizumab, 1 received baricitinib, and P19, who had moderate pneumonia, was treated with SARS-CoV-2–neutralizing mAbs (casirivimab plus imdevimab), 2 d after the onset of symptoms. One of the 22 patients (P17, 23 yr old) died (Table 1). The course of SARS-CoV-2 infection in patients with pneumonia was generally severe, albeit with some interindividual variability.
## Lack of other known genetic or immunological disorders in the patients
We screened the patients for other genetic or immunological disorders known to cause hypoxemic COVID-19 pneumonia (Bastard et al., 2020, 2021b, Zhang et al., 2020b, 2022b; Asano et al., 2021). Autoantibodies neutralizing IFN-α2, IFN-β, and IFN-ω were not detected in any of the 12 patients tested, 3 of whom had mild infections and 9 of whom had COVID-19 pneumonia (5 moderate, 2 severe, and 2 critical). We also sequenced the exomes of 13 patients and analyzed the available exomes of 2 other patients. We screened our 15 patients for rare (MAF < 10−3 according to the gnomAD database) predicted loss-of-function (pLOF) variants of the 478 genes known to underlie AR, autosomal dominant (AD), or XR inborn errors of immunity (IEIs; Tangye et al., 2022). Among the five patients with mild SARS-CoV-2 infection, three patients carried candidate variants potentially linked to an IEI (Table S4). Noteworthy, two brothers with mild COVID-19 (P20 and P21) carried a heterozygous variant (p.P301L) in the OAS1 gene. AR LOF mutations in OAS1 were recently shown to cause the multisystem inflammatory syndrome in children (Lee et al., 2022); and heterozygous OAS1 gain-of-function (GOF) variants cause a polymorphic autoinflammatory immunodeficiency (OPAID), characterized by recurrent fever, dermatitis, inflammatory bowel disease, pulmonary alveolar proteinosis, and hypogammaglobulinemia (Cho et al., 2018; Magg et al., 2021). Both patients and their parents (fully vaccinated against SARS-CoV-2) developed a mild SARS-CoV-2 infection and did not have any of the clinical or immunological signs of gain-of-function– or LOF-OAS1 deficiency.
Among 10 patients with COVID-19 pneumonia (5 moderate, 2 severe, and 3 critical), 4 patients carried variants that could potentially be linked to an IEI (Table S4). Two patients, P14 (critical pneumonia) and P18 (moderate pneumonia), carried heterozygous variants (p.R756W and p.E941K, respectively) in RTEL1. AR and, more rarely, AD mutations in RTEL1 are associated with dyskeratosis congenita, and AD variants in RTEL1 have also been associated with idiopathic pulmonary fibrosis, even with a late or very late onset (Borie et al., 2019; Moore et al., 2019; Newton et al., 2022; Stuart et al., 2015; Walne et al., 2013). Neither the patients nor their parents had a history of any of the cardinal features of dyskeratosis congenita or of idiopathic pulmonary fibrosis, although no CT scans were performed. Heterozygous variants in POL3RA and IFNAR1 were found in P15 (critical pneumonia). The variant in POLR3A found in P15 is the same as that observed in P3 and his mother, both unvaccinated against SARS-CoV-2 and with mild COVID-19 (Table S4 and Fig. 1) and no previous varicella-zoster virus infection or any pathological predisposition to viral infectious diseases. AD mutations in IFNAR1, which encodes a subunit of the type I IFN receptor, have been previously shown to cause critical COVID-19 pneumonia (Zhang et al., 2020b). The impact on expression and function of the observed p.Q80H mutation in IFNAR1 was studied and found to be neutral (Zhang et al., 2020b). Finally, P17 (critical pneumonia) was found to be homozygous for a p.E941K variant in APOL1. Many African individuals express APOL1 variants that, in heterozygosity, counteract resistance factors from human infective trypanosomes, enabling them to avoid sleeping sickness. In addition, the APOL1 variants that confer protection against trypanosomiasis are associated with chronic kidney disease, particularly in the context of virus-induced inflammation such as COVID-19 (Pays et al., 2022; Zhang et al., 2019). It is unlikely that these variants predisposed to critical COVID-19 pneumonia in our patient with no documented chronic kidney disease and normal serum creatinine levels in the course of his SARS-CoV-2 infection.
None of the patients carried candidate variants at any IEI loci (Tangye et al., 2022), suggesting that other known genetic disorders did not contribute to their poor control of SARS-CoV-2. Thus, we did not detect auto-Abs against type I IFN or additional genetic defects in the patients. However, these findings do not exclude the possibility that other genetic or acquired modifiers affected the outcome of SARS-CoV-2 infection in these patients. Collectively, however, they suggest that the MYD88 and IRAK4 genotypes were the main drivers of COVID-19 pneumonia in 15 of the 22 patients.
## More severe COVID-19 pneumonia in patients with MyD88/IRAK-4 deficiencies than in age-matched individuals from the general population
We compared the risks of hospitalization and critical COVID-19 pneumonia between our patients and a retrospective series of 167,262 SARS-CoV-2–infected children before the emergence of the Omicron variant (mean age: 11.9 yr; interquartile range [IQR]: 6.0–16.1 yr) within the National COVID Cohort Collaborative (NC3) cohort (total 1,068,410 children <19 yr of age, NC3, USA). We included only 19 patients from our cohort who were also infected with the original and Delta variants. These two cohorts had similar age distributions (mean age of patients with MyD88 or IRAK-4 deficiency: 13.0 yr [IQR 6.5–16.0] vs. NC3 series: 11.9 yr [IQR 6–16.1]), but there were significantly more male patients in our cohort ($84.2\%$) than in the NC3 cohort ($50.1\%$; $$P \leq 0.006$$; OR 5.3, $95\%$ confidence interval [CI] 1.5–18.1). In total, 10,245 ($6.1\%$) of the controls were hospitalized (Martin et al., 2022a). The risk of hospitalization was significantly higher in MyD88- or IRAK-4–deficient patients ($89.5\%$) than in the N3C series ($6.1\%$; $P \leq 0.001$; OR: 130.3; $95\%$ CI: 30.1–563.9; Fig. 3, A and B). Invasive mechanical ventilation was required for 796 of the 167,262 patients positive for SARS-CoV-2 in the NC3 series ($0.5\%$), and for 5 of our 19 patients ($26.3\%$; $P \leq 0.001$; OR: 74.7, $95\%$ CI: 26.8–207.8). Moreover, the infection fatality rate was significantly higher in our patients ($5\%$) than in the NC3 series ($0.08\%$; $P \leq 0.001$; OR: 70.9, $95\%$ CI: 9.4–535.2; Fig. 3, A and B). Overall, the clinical manifestations of SARS-CoV-2 infection were much more severe in MyD88- or IRAK-4–deficient patients than in age-matched individuals from the general population.
**Figure 3.:** *Susceptibility to severe COVID-19 in patients with MyD88 or IRAK-4 deficiency. (A and B) Severity (A) and OR (B) of SARS-CoV-2 infection in MyD88/IRAK-4–deficient patients relative to the age-matched controls from the NC3 cohort infected with the same viral variants. *, P < 0.001. (C and D) Severity (C) and OR (D) of SARS-CoV-2 infection in heterozygous relatives of MyD88- or IRAK-4–deficient patients relative to the Spanish general population between the ages of 20 and 49 yr after the first wave (April 2020).*
## Similar COVID-19 pneumonia severity in patients with MyD88/IRAK-4 deficiencies and patients with TLR7 deficiency
35 male patients with SARS-CoV-2 infection and experimentally confirmed TLR7 deficiency have been reported to date (van der Made et al., 2020; Asano et al., 2021; Fallerini et al., 2021; Solanich et al., 2021a; Zhang et al., 2022b). We compared 22 patients with MyD88/IRAK-4 deficiency with 35 patients with TLR7 deficiency. Globally, MyD88/IRAK-4–deficient patients were less likely to develop severe infections (mild infections: 27.3 vs. $5.7\%$; moderate: 27.3 vs. $2.9\%$; severe: 18.2 vs. $11.4\%$; critical: 27.3 vs. $80\%$; $$P \leq 0.0002$$). These differences cannot be explained by the sex of the patients because they remained significant when only the 18 male patients with MyD88 or IRAK-4 deficiency were considered ($$P \leq 0.0001$$). There may have been an ascertainment bias, because the IRAK-4– and MyD88-deficient patients in our cohort were recruited prospectively, whereas most of the TLR7-deficient patients were recruited retrospectively. We, therefore, also compared SARS-CoV-2 infection severity between our cohort and a subgroup of seven prospectively recruited TLR7-deficient patients (Asano et al., 2021; Solanich et al., 2021a; Zhang et al., 2022b). No significant differences in age were found between our 22 patients (mean 10.9 ± 6.8 yr) and these 7 TLR7-deficient patients (25.4 ± 16.9; $$P \leq 0.06$$). No statistically significant differences in disease severity were observed either (mild: 27.3 vs. $28.6\%$; moderate: 27.3 vs. $14.3\%$; severe: 18.2 vs. $14.3\%$; critical: 27.3 vs. $42.9\%$; $$P \leq 0.9$$). However, the group of prospectively recruited TLR7-deficient patients was small. Overall, COVID-19 severity appears to be similar in patients with MyD88/IRAK-4 deficiency and in those with TLR7 deficiency. The difference between these groups is clearly less pronounced than that between either group of patients and age-matched individuals from the general population.
## Lack of MYD88 or IRAK4 co-dominance for COVID-19 pneumonia
The bacterial infection phenotype of MyD88 or IRAK-4 deficiency is clearly recessive (Picard et al., 2010), but we analyzed possible co-dominance for the COVID-19 phenotype in 20 unvaccinated household relatives heterozygous for MYD88 or IRAK4 (mean age: 32.4 yr, SD: 12.3 yr, range: 10–52 yr). Given their age distribution and the Spanish origin of 6 of the 15 kindreds, we compared the hospitalization rate, ICU need, and mortality of this cohort with data for the *Spanish* general population between the ages of 20 and 49 yr after the first wave in April 2020 (Cannistraci et al., 2021; Ministerio de Sanidad, 2022). We found no significant differences in hospitalization rate (3 of 20, $15\%$, of MyD88/IRAK-4 heterozygous relatives vs. 8,823 of 35,583 ($24.8\%$) in the general population, $$P \leq 0.4$$; OR: 0.5, $95\%$ CI: 0.2–1.8), ICU admission ($\frac{2}{20}$–$10.0\%$ vs. $\frac{647}{35}$,583 [$1.8\%$], $$P \leq 0.06$$; OR: 6.0, $95\%$ CI: 1.4–25.9) or mortality ($\frac{1}{20}$ [$5.0\%$] vs. $\frac{161}{35}$,583 [$0.4\%$]; $$P \leq 0.2$$; OR: 11.6, $95\%$ CI: 1.5–87.0; Fig. 3, C and D). We also searched for an enrichment in rare (gnomAD frequency <10−3) pLOF variants of IRAK4 and MYD88 in 3,269 patients with critical COVID-19, and 1,373 controls with asymptomatic or mild SARS-CoV-2 infection from the CHGE (Matuozzo et al., 2022). We identified three heterozygous individuals among the patients with critical disease and one among the controls with mild disease ($$P \leq 0.09$$). Overall, our results suggest that heterozygous carriers may not have a higher risk of hypoxemic COVID-19, or that penetrance is very low in these individuals.
## Systemic inflammation during SARS-CoV-2 infection
In 7 of the 16 hospitalized patients for whom data were available, hypothermia (1 patient) or low-grade fever (37.5–38.3°C; 6 patients) was documented; the other 9 patients developed a fever with a body temperature between 39 and 40.9°C. Thus, most patients were able to mount a fever, contrasting with the poor or delayed fever mounted in response to pyogenic bacteria in such patients (Picard et al., 2010). We also studied the blood levels of inflammatory markers including C reactive protein (CRP), ferritin, lactate dehydrogenase (LDH), and absolute neutrophil count (ANC), in hospitalized patients upon admission, and the highest or lowest levels detected if multiple measurements were collected. Overall, most of the patients had high CRP, ferritin, and LDH levels (Fig. 4, A–C). The levels of ferritin and LDH seemed to be particularly high in the patients with the most severe disease, indicating a strong inflammatory response, at odds with previous studies of inflammation in the course of bacterial disease in these patients (Picard et al., 2003; von Bernuth et al., 2008; Picard et al., 2010). However, MyD88- or IRAK-4–deficient patients with critical COVID-19 had somewhat lower blood CRP, ferritin, and LDH levels than TLR7-deficient patients with critical COVID-19 pneumonia (Asano et al., 2021), hospitalized patients under the age of 21 yr from the general population (Bourgeois et al., 2021), and patients from the general population admitted to the ICU (Pierce et al., 2020), although this difference was not statistically significant (Fig. 4, A–C). High ANC is another marker of systemic inflammation often observed in patients with severe COVID-19 pneumonia (Pierce et al., 2020; Bennett et al., 2021; Martin et al., 2022a). However, neutropenia is rare in the acute phase of infection, even in critical cases (Manson et al., 2020; Pierce et al., 2020; Bennett et al., 2021; Bourgeois et al., 2021; Martin et al., 2022a). Surprisingly, we observed frequent neutropenia (ANC < 1,500/mm3) in MyD88- or IRAK-4–deficient patients (Fig. 4 D). Seven of the hospitalized MyD88- or IRAK-4–deficient patients developed neutropenia during the acute phase of infection (three of four moderate cases and four of six critical cases), whereas only two had a high ANC at a particular time point (ANC > 8,000/mm3; Fig. 4 D and Table S5). By contrast, only 1 of 12 TLR-7–deficient patients with critical disease developed neutropenia and 5 had a high ANC (Fig. 4 D). This phenotype was also observed in MyD88- and IRAK-4–deficient patients with pyogenic bacterial infections (Picard et al., 2003; Picard et al., 2010). Overall, IRAK-4– and MyD88-deficient patients were able to mount an inflammatory response to SARS-CoV-2 infection, with characteristic neutropenia potentially due to defective IL-1R signaling.
**Figure 4.:** *Inflammation markers in MyD88/IRAK-4-deficient patients during acute infections. (A) Left: CRP on admission (colored dots), highest level detected (upper bar), and lowest level detected (lower bar) in hospitalized patients with MyD88 or IRAK-4 deficiency. Right: CRP on admission in patients with TLR7 deficiency, hospitalized members of the general population <21 yr of age (Bourgeois et al., 2021), or members of the general population admitted to the ICU (Pierce et al., 2020). Dashed red line: normal range of CRP concentration (<10 mg/dl). (B and C) Ferritin (B) and LDH (C) concentrations in patients with MyD88 or IRAK-4 deficiency (left panels), or TLR7 deficiency, hospitalized members of the general population <21 yr of age (Bourgeois et al., 2021), or members of the general population admitted to the ICU (Pierce et al., 2020; right panels). Dashed red lines: normal range of ferritin (11–336 ng/ml) and LDH (<280 U/liter) concentrations. (D) Left: ANC on admission (colored dots), highest level detected (upper bar), and lowest level detected (lower bar) in hospitalized patients with MyD88 or IRAK-4 deficiency. Right: ANC on admission in patients with TLR7 deficiency, hospitalized members of the general population <21 yr of age (Bourgeois et al., 2021), or members of the general population admitted to the ICU (Pierce et al., 2020). Dashed red line: normal range of ANC (1,500–8,000 cells/mm3).*
## Blood transcriptome inflammatory signature in patients during COVID-19
We also performed a transcriptome analysis focused on the genes of the inflammatory response to infection. The four patients studied had COVID-19 pneumonia (two moderate and two severe cases). The inflammatory response has been shown to be correlated with COVID-19 severity in several studies (Bennett et al., 2021; Kim and Shin, 2021; Maleknia et al., 2022; Martin et al., 2022a). In our patients, the upregulation of genes involved in the inflammatory response, particularly those involved in the IL-1–mediated pathway, was in the range observed in the four patients with mild COVID-19. However, this upregulation was markedly lower than that observed in the patient with IRF9 deficiency, who also had mild COVID-19 (Fig. 5 A; Lévy et al., 2021). Thus, transcriptomic analysis demonstrated that the MyD88/IRAK-4–deficient patients were able to mount a systemic inflammatory response during the acute phase of the infection.
**Figure 5.:** *Transcriptome analysis of whole-blood samples from SARS-CoV-2–infected individuals. (A) Single-sample gene set enrichment analysis (Hänzelmann et al., 2013) was used to evaluate the IFN-α response, the IFN-γ response, TNF-α signaling through NF-κB, IL-6 JAK-STAT3 signaling, and the inflammatory response. There was one sample for each time-point and patient, and the assay was performed once for each sample. Dot heatmap representing pathway enrichment scores for individual samples. The enrichment score is represented by a colored spot, with red indicating an increase in abundance and blue indicating a decrease in abundance. The intensity of the spots reflects the enrichment score. (B) Time-dependent consistent changes in transcript abundance for type I IFN (red), ISGs with known antiviral functions (purple), other ISGs (yellow), and protein-coding genes (gray) are represented on a scatter plot for IRAK4- and MyD88-deficient patients and a non-infected healthy control. In parentheses, D indicates the number of days after positive RT-PCR for controls and the days after symptom onset for patients. Red arrows indicate the day of treatment with mAbs (casirivimab and imdevimab).*
## Blood type I IFN transcriptome signature in patients during COVID-19
We collected whole-blood samples from four patients with IRAK-4 (P19, with moderate disease) or MyD88 (P1 with moderate disease, and P2 and P5 with severe disease) deficiency in the course of primary SARS-CoV-2 infection. These samples were used for whole-blood RNA-seq. Ex vivo transcriptome analysis showed that IRAK-4– and MyD88-deficient patients were able to produce type I IFNs during the acute phase of SARS-CoV-2 infection, as shown by the induction of IFN-stimulated genes (ISGs) in leukocytes, especially those with known antiviral functions (Fig. 5 B). The type I IFN activity detected in our patients was, as expected, much higher than that observed in an 8-yr-old girl with mild COVID-19 and an AR complete deficiency of IRF9 (Lévy et al., 2021), which governs ISGF-3–dependent responses to type I and III IFNs (Fig. 5 A). Indeed, a strong upregulation of numerous ISGs has been observed in peripheral blood during the first few days after symptom onset in patients displaying progression to severe disease relative to patients with mild disease (Galani et al., 2020; Hadjadj et al., 2020; Lee et al., 2020; Zhu et al., 2020; Kim and Shin, 2021; Ng et al., 2021; Ren et al., 2021; Zhao et al., 2021; Unterman et al., 2022). The scores obtained for our patients were in the range for control individuals with mild COVID-19 and higher than that obtained for a non-infected healthy control (Fig. 5 A). We also observed some heterogeneity in type I IFN activity between patients, possibly due to the timing of sampling after disease onset (with samples collected between 2 and 9 d after disease onset), or disease severity, which ranged from moderate and severe. These data suggest that despite the deficit of type I IFN production by pDCs, as in TLR7-deficient patients (Asano et al., 2021), other cells can produce type I IFNs that can activate leukocytes (Chiale et al., 2022; Lucas et al., 2020). However, the type I IFN signature was weaker than expected given the clinical severity of disease and relative to the general population infected with SARS-CoV-2 and analyzed in the first few days after symptoms onset.
## mAb-mediated neutralization of SARS-CoV-2 in an IRAK-4–deficient child
P19, a 14-yr-old male with IRAK4 deficiency, was admitted 2 d after the onset of clinical manifestation, including cough, nasal congestion, and fever (39.4°C). On day 1 of admission, PCR on a nasal swab revealed a very high load of SARS-CoV-2 (Ct Gene S: 16.20, Ct Gene N: 20.98, Ct Gene RdRP: 17.56), and showed the patient to be infected with the P681R/L425R Delta variant (previously known as B.1.617.2). A chest x-ray showed increased interstitial markings in the left retrocardiac space, but the SpO2/FIO2 ratio was $99\%$, indicating a moderate, non-hypoxemic, pneumonia. He then received a single dose of intravenous casirivimab (4,000 mg) and imdevimab (4,000 mg), a combination of human IgG1 neutralizing the receptor-binding domain of the SARS-CoV-2 spike protein on this same first day of admission. His symptoms and signs disappeared on day 3, and he was discharged on day 4. Follow-up evaluation on day 14 was unremarkable and the patient was asymptomatic.
We performed RNA-seq analysis on longitudinal whole-blood samples obtained from P19 at various time points from day 1 to 14. The transcripts of genes involved in antiviral immunity, particularly ISGs with known antiviral functions, were readily detected on days 1 and 2. Their levels decreased sharply on day 3, 4 d after symptom onset, when clinical manifestations disappeared, and, on day 4, they were barely detectable, if at all, as in a non-infected healthy control (Fig. 5 A). This pattern of expression contrasts with that observed in a patient with IRF9 deficiency and mild COVID-19 with positive PCR results for SARS-CoV-2, who was treated with casirivimab and imdevimab on day 2 after hospital admission (Lévy et al., 2021; Fig. 5 A). Moreover, the heatmap of RNA-seq–quantified gene expression (z-score–scaled log2-normalized counts) for TNF-α signaling through NF-κB gene sets showed lower transcript levels on admission in P19 than in the IRF9-deficient patient. Transcript levels for the induced genes of these inflammatory pathways decreased to very low levels 2 d after treatment with anti–SARS-Cov-2 mAbs in both patients (Fig. 5 B). These data, together with the rapid resolution of clinical manifestations in both patients, demonstrate the safety and efficacy of antibody-mediated viral neutralization in patients with either of these deficiencies of type I IFN–mediated immunity.
## Discussion
We showed that unvaccinated patients with MyD88 or IRAK-4 deficiency infected with SARS-CoV-2 are at high risk of COVID-19 pneumonia, including hypoxemic and even critical forms. The risk is much higher than for members of the same age group in the general population (Knock et al., 2021; Le Vu et al., 2021; Martin et al., 2022a; Martin et al., 2022b). The risk is also higher than that of heterozygous relatives, attesting to a lack of detectable co-dominance, and confirming that these two closely related IEIs confer a recessive predisposition to life-threatening COVID-19 pneumonia, even in childhood or adolescence. These patients appear to have a risk similar to that of children and adults with autoimmune polyendocrinopathy syndrome type 1 (APS-1), who are vulnerable to SARS-CoV-2 due to the production of auto-antibodies neutralizing type I IFNs (Bastard et al., 2021a; Lemarquis et al., 2021; Meisel et al., 2021). Our findings for patients with MyD88 or IRAK-4 deficiency are also consistent with previous findings from our own and other studies indicating that XR TLR7 deficiency confers a high risk of severe or critical COVID-19 pneumonia (van der Made et al., 2020; Asano et al., 2021; Fallerini et al., 2021; Kosmicki et al., 2021; Pessoa et al., 2021; Solanich et al., 2021a). The mechanism of disease in patients with MyD88 or IRAK-4 deficiency probably involves an impairment of the TLR7-mediated sensing of the virus by pDCs, as demonstrated ex vivo (Asano et al., 2021; Onodi et al., 2021). The residual response of TLR7-deficient pDCs to SARS-CoV-2 ex vivo, contrasting with the abolition of this response in IRAK-4– and UNC-93B–deficient pDCs, may be due to signaling through TLR9, as pDCs do not express TLR8 (Aluri et al., 2021; Asano et al., 2021; Boisson and Casanova, 2021) and UNC-93B– and IRAK-4–deficient pDCs have defects of both TLR7 and TLR9 signaling (Onodi et al., 2021).
The TLR3-dependent induction of type I IFNs is intact in patients with MyD88 or IRAK-4 deficiency. Our findings, therefore, confirm that the TLR7-dependent induction of type I IFNs by pDCS is essential for host defense against SARS-CoV-2 in the respiratory tract. Patients with MyD88 or IRAK-4 deficiency seem to suffer from COVID-19 disease as severe as that in patients with TLR7 deficiency. Their more profound pDC defect, with the abolition of responses to both TLR7 and TLR9 agonists and a complete lack of type I IFN production upon stimulation with SARS-CoV-2 (Yang et al., 2005; Ku et al., 2007; von Bernuth et al., 2008; Alsina et al., 2014; Onodi et al., 2021), may be mitigated by other mechanisms, such as the abolition of responses to IL-1 paralogs, IL-18, and IL-33. The retrospective and prospective nature of the recruitment of TLR7- and MyD88/IRAK-4–deficient patients, respectively, may have led to an ascertainment bias. Nevertheless, the ISG response in our patients’ leukocytes was weak but detectable, probably due to activation of the TLR3-dependent or other pathways in infected lung epithelial cells. In addition, other sensors of viral RNA, such as RIG-1 and MDA-5, expressed in several leukocyte subsets, particularly in myeloid dendritic cells and monocytes, and at very low levels in resting pDCs, may have contributed to type I IFN production (Bencze et al., 2021; Liu and Gack, 2020). Noteworthy, transcriptome analyses have shown that the epithelial and immune cells of the upper airways of healthy children are preactivated and express significantly higher basal levels of the genes coding for RIG-I and MDA5 compared to adults, resulting in stronger innate antiviral responses upon SARS-CoV-2 infection (Loske et al., 2022). In addition, proteomic analyses, revealed that particularly RIG-I (also called DDX58) is among the most differentially detectable protein in circulation and in lung parenchymal tissues in patients with acute and fatal SARS-CoV-2 infection respectively vs. healthy controls (Filbin et al., 2021; Gisby et al., 2021; Russell et al., 2022). These data suggest that RIG-I may be a dominant pathway induced by the virus. On the other hand, there is a significant overlap between gene signatures for type I and type II IFN signaling. Our RNA-seq data do not allows to rule out that the IFN signature observed may be secondary, at least partially, to signaling by IFN-γ. Studies of the transcriptome of TLR3- or TLR7-deficient patients, and the identification of new IEIs underling hypoxemic COVID-19 pneumonia, would help to resolve these questions. Overall, impairment of the TLR7/MyD88/IRAK-4 pathway prevents pDCs from producing sufficient type I IFN in response to SARS-CoV-2 in the respiratory tract, accounting for the patients’ vulnerability to infections with this virus.
Since the first description of IRAK-4 and MyD88 deficiencies, the MyD88/IRAK-4–mediated pathway has been considered redundant for protective immunity against viruses in humans (Ku et al., 2007; Picard et al., 2010). EBV viremia without clinical repercussions was later reported in a MyD88-deficient patient (Chiriaco et al., 2019), and P5 was recently reported to have suffered from bilateral pneumonia caused by influenza A virus and the human cororonavirus NL63 (Bucciol et al., 2022a). An IRAK-4–deficient patient with a suspected reactivation of human herpesvirus 6 (HHV-6) infection has also been described (Nishimura et al., 2021), and genomic material from HHV-6 was also detected in three patients in our series (Table S2). A Turkish patient with severe HHV-6 disease has also recently been described (Tepe et al., 2022). The presence of the HHV-6 genome has, however, repeatedly been reported in normal brains, and HHV-6 reactivation may occur in healthy children without apparent illness or during acute illness (Caserta et al., 2004; Komaroff et al., 2020; Pandey et al., 2020; Santpere et al., 2020). Overall, the susceptibility of MyD88- and IRAK-4–deficient patients to HHV-6 is probable, but not formally proven. TLR7-deficient patients do not appear to be susceptible to common viral infections other than SARS-CoV-2, but further studies are required to confirm this. Indeed, TLR3 deficiency was initially reported in patients with herpes simplex encephalitis (Guo et al., 2011; Jouanguy et al., 2020; Zhang et al., 2007b), but patients with TLR3 deficiency were progressively found to be prone to other viral diseases too, including critical influenza pneumonia (Bucciol et al., 2022b; Kyung Lim et al., 2019), other types of viral encephalitis (Chen et al., 2021; Hautala et al., 2020; Kuo et al., 2022; Partanen et al., 2020), and hypoxemic COVID-19 pneumonia (Zhang et al., 2020b). Overall, our data suggest that MyD88 and IRAK-4 deficiencies underlie hypoxemic COVID-19 pneumonia. Together with the reports of HHV6 disease in such patients, and the occurrence of severe influenza pneumonia in two patients from our cohort, these data suggest that MyD88- and IRAK-4–deficient patients may be prone to other severe viral infections, perhaps with low penetrance.
## Cohort recruitment and consent
We recruited patients with MyD88 or IRAK-4 deficiency who had suffered COVID-19 between the start of the pandemic and January 2022. Data were collected through an anonymized survey sent to specialists in immunology or pediatrics with reported or unreported patients with these IEIs, and through clinicians caring for patients with IEIs identified from the European Society for Immunodeficiencies (ESID) Registry. Samples were obtained from the probands, parents and relatives with written informed consent. The study was approved by the French Ethics Committee “Comité de Protection des Personnes,” the French National Agency for Medicine and Health Product Safety, the “Institut National de la Santé et de la Recherche Médicale,” in Paris, France (protocol no. C10-13), the Rockefeller University Institutional Review Board in New York, NY, USA (protocol no. JCA-0700), the Committees for Ethical Research of the University Hospital of Gran Canaria Dr. Negrínn (protocol no. 2020-200-1 COVID-19) and Hospital San Joan de Deu (protocol no. PIC-173-21), and the Office of Research & Innovation at Drexel University, Philadelphia, PA, USA (protocol no. 2112008918).
## Definition of SARS-CoV-2 infection
SARS-CoV-2 infection was defined as a positive RT-PCR or antigenic test result for a nasopharyngeal sample for symptomatic patients, or as a positive serological test result for patients with no symptoms. The SARS-CoV-2 Ct value varied between patients. Viremia was not analyzed. The SARS-CoV-2 variants also differed between patients. In five patients, the viral variant was confirmed by molecular methods (two Alpha and three Delta variants). In the remaining patients, infection was suspected to be caused by the predominant variant in the country at the time of diagnosis (Our World in Data, 2023). 12 patients were infected between April 2020 and February 2021 when the variants of clades 20A and 20B, which replaced the original virus infecting humans (clade 19A), predominated. Four patients were infected in April–May 2021, when the variant of concern (VOC) Alpha (clade 20I, genetically confirmed in two) predominated. Three patients infected from November 2021 to January 2022 had genetically confirmed infections with the VOC Delta (clade 21A). Finally, three patients were infected in January 2022, when Omicron (clade 21M) was the predominant VOC in their countries of residence (Fig. 2 A and Table S1). For 15 patients, infection was confirmed to have resulted from household transmission from an infected relative (Table S1). In another two patients (P4 and P5), the infection was contracted from a visiting relative. The mode of infection of the remaining five patients is unknown. None of the patients were vaccinated against SARS-CoV-2 at the time of infection.
## Data regarding COVID-19 and medical history
Clinical, laboratory, and chest imaging data obtained during COVID-19, other risk factors for severe COVID-19 (Rodrigues et al., 2020; Zhang et al., 2020a, 2022a; Bennett et al., 2021; Bourgeois et al., 2021; Kooistra et al., 2021; Navaratnam et al., 2021; O’Driscoll et al., 2021; Westblade et al., 2021; Martin et al., 2022a; Schober et al., 2022; Woodruff et al., 2022), and family history were collected for each patient in the survey. Concomitant infections were also recorded, when supported by clinical suspicion, positive cultures, and/or chest x-ray images. Adults with a body mass index (BMI) over 25 were considered to be overweight, and those with a BMI over 30 were considered to be obese. Children aged between 5 and 19 yr were considered to be obese if their BMI-for-age-and-sex was more than 2 SD above the WHO Growth Reference median. Children under 5 yr of age were considered to be obese if their weight-for-height was more than 3 SD above the WHO Child Growth Standard median (WHO, 2021).
COVID-19 severity was assessed according to the Human Genetic Effort clinical score (Asano et al., 2021). SARS-CoV-2 infection was classified as mild/non-confirmed pneumonia (for patients who were asymptomatic, presented upper respiratory tract disease with no signs of pneumonia on x ray or with respiratory symptoms not suggestive of a lower respiratory tract infection and therefore not requiring x ray), moderate (non-hypoxemic pneumonia, not requiring oxygen therapy), severe (hypoxemic pneumonia requiring therapy with oxygen <6 liters O2/min, without meeting the criteria for critical pneumonia) or critical (hypoxemic pneumonia requiring high-flow oxygen >6 liters O2/min, ventilatory support with or without intubation, or ECMO [extracorporeal membrane oxygenation]).
Laboratory values were recorded when available. Normal ranges of laboratory values were reported according to age and are expressed in standard units (Hollowell et al., 2005; Mayo Clinic, 2022).
## Definition of positive contacts
Patients were considered to be positive SARS-CoV-2 contacts if they had been exposed (for more than 15 min, at a distance of <6 feet) to a person infected with the virus for whom an oral-nasopharyngeal sample had tested positive for specific SARS-CoV-2 RNA or antigen (CDC, 2022).
## Analysis of anti-type I IFN auto-Abs
The presence of auto-Abs able to neutralize high doses (10 ng/ml) of IFN-α2 and IFN-ω, and IFN-β, or lower, more physiological doses (100 pg/ml) of α2 and IFN-ω, was analyzed as previously reported (Bastard et al., 2021b) in plasma or serum samples from the patients.
## Next-generation sequencing
Genomic DNA was extracted from whole blood from all patients except P17, for whom DNA was obtained from SV40-transformed fibroblasts. The whole exome was sequenced at the Genomics Core Facility of the Imagine Institute (Paris, France), the Yale Center for Genome Analysis the New York Genome Center, and The American Genome Center (Uniformed Services University of the Health Sciences, Bethesda, MD, USA), and the Genomics Division–Institute of Technology and Renewable Energies of the Canarian Health System sequencing hub (Canary Islands, Spain), as previously reported (Asano et al., 2021). The whole-exome sequences of the patients were filtered against the complete International Union of Immunological Societies list of genes (Tangye et al., 2022), with the retention of variants with an allele frequency below 0.001. We excluded synonymous mutations, downstream, upstream, intron and non-coding transcript variants and intergenic variants. We also excluded variants predicted to be benign and we checked the quality of the exome sequences. The mutation significance cutoff (http://pec630.rockefeller.edu:8080/MSC/) was used to determine whether variants were likely to be damaging.
## RNA-seq
Whole blood for RNA-seq was collected in PAXgene Blood RNA tubes (BD Biosciences, samples collected from the IRF9-deficient patient, her mother, P5, and P19), Tempus Blood RNA tubes (Thermo Fisher Scientific, P1 and P2) or EDTA tubes (controls). Samples from P1 (moderate pneumonia) were obtained at hospital admission, 8 d after positive PCR and 2 d after symptom onset; samples from P2 (severe pneumonia) were collected on hospital admission, 4 d after PCR and symptom onset; samples from P5 (severe pneumonia) were obtained 6 d after hospital admission, 9 d after PCR and symptom onset; and the first sample from P19 was obtained before treatment with anti–SARS-CoV-2 mAbs, on the day of hospital admission, 2 d after symptom onset, when PCR was performed. We also collected longitudinal whole-blood samples from P19 at various time points from day 1 to 14. We compared our data with data obtained for samples from four controls with mild COVID-19 obtained 4 d after a positive PCR for SARS-CoV-2, a non-infected healthy control, and the longitudinal data (day 1 [hospitalization] to day 32), obtained from a previously reported IRF9-deficient patient with mild SARS-CoV-2 infection also treated with anti–SARS-CoV-2 mAbs (Lévy et al., 2021) the day after hospital admission.
There was one sample for each time-point and patient and the assay was performed once for each sample.
Blood samples were subjected to hemoglobin RNA depletion. Samples were sequenced on the Illumina NextSeq platform with a single-end 75-bp configuration. The RNA-seq fastq raw data were inspected to ensure that they were of high quality. The sequencing reads were mapped onto the human reference genome GRCh38 with STAR aligner v.2.7, and the mapped reads were then quantified with featureCounts v2.0.2 to determine gene-level read counts. *The* gene-level read counts were normalized and log2-transformed with DESeq2, to obtain the gene expression profile for all samples.
Single-sample gene set enrichment analysis (Hänzelmann et al., 2013) was used to evaluate the IFN-α response, IFN-γ response, TNF-α signaling through NF-κB, IL-6 JAK-STAT3 signaling, including the inflammatory response enrichment scores. The raw RNA-seq data generated from this study are deposited in the National Center for Biotechnology Information database under the National Center for Biotechnology Information- Sequence Read Archive project PRJNA916275.
## Statistical analysis
Categorical variables are expressed as percentages, and discrete variables as medians with the observed range, or as means ±$95\%$ CI. Fisher’s exact tests or Yates correction and ORs with $95\%$ CI were used for comparative analyses. Continuous variables are presented as the arithmetic mean ± SD, and Mann–Whitney U-tests were used for the comparative analysis. The analysis was performed with SPSS v.15.0 software (SPSS, Inc.) and graphs were performed by GraphPad Prism v.7.00 for Windows, GraphPad Software, with values of P ≤ 0.05 considered statistically significant.
## Online supplemental material
Table S1 provides information about baseline characteristics of MyD88 and IRAK-4–deficient patients, data on the diagnosis of COVID-19 and the lung conditions during the infection. Table S2 contains information about non–SARS-CoV-2 viral infections in the MyD88 and IRAK-4–deficient patients. Table S3 shows the serological results for antibodies against common viruses for two patients. Table S4 summarizes the pLOF variants of the 478 genes known to underlie AR, AD, or XR IEIs in our MyD88/IRAK-4–deficient patients. Table S5 includes the laboratory data of MyD88 and IRAK-4–deficient patients during SARS-CoV-2 infection.
## References
1. Alsina L., Israelsson E., Altman M.C., Dang K.K., Ghandil P., Israel L., Von Bernuth H., Baldwin N., Qin H., Jin Z.. **A narrow repertoire of transcriptional modules responsive to pyogenic bacteria is impaired in patients carrying loss-of-function mutations in MYD88 or IRAK4**. *Nat. Immunol.* (2014) **15** 1134-1142. DOI: 10.1038/ni.3028
2. Aluri J., Bach A., Kaviany S., Chiquetto Paracatu L., Kitcharoensakkul M., Walkiewicz M.A., Putnam C.D., Shinawi M., Saucier N., Rizzi E.M.. **Immunodeficiency and bone marrow failure with mosaic and germline TLR8 gain of function**. *Blood* (2021) **137** 2450-2462. DOI: 10.1182/blood.2020009620
3. Amorim V.M., Guardia G.D.A., Dos Santos F.F., Ulrich H., Galante P.A.F., De Souza R.F., Guzzo C.R., Silva De Souza A., Martins De Freitas Amorim V., Guardia G.D.A.. **Severe acute respiratory syndrome coronavirus 2 variants of concern: A perspective for emerging more transmissible and vaccine-resistant strains**. *Viruses* (2022) **14** 827. DOI: 10.3390/v14040827
4. Asano T., Boisson B., Onodi F., Matuozzo D., Moncada-Velez M., Maglorius Renkilaraj M.R.L., Zhang P., Meertens L., Bolze A., Materna M.. **X-linked recessive TLR7 deficiency in ∼1% of men under 60 years old with life-threatening COVID-19**. *Sci. Immunol.* (2021) **6** 65. DOI: 10.1126/sciimmunol.abl4348
5. Bastard P., Rosen L.B., Zhang Q., Michailidis E., Hoffmann H.H., Zhang Y., Dorgham K., Philippot Q., Rosain J., Béziat V.. **Autoantibodies against type I IFNs in patients with life-threatening COVID-19**. *Science* (2020) **370**. DOI: 10.1126/science.abd4585
6. Bastard P., Orlova E., Sozaeva L., Lévy R., James A., Schmitt M.M., Ochoa S., Kareva M., Rodina Y., Gervais A.. **Preexisting autoantibodies to type I IFNs underlie critical COVID-19 pneumonia in patients with APS-1**. *J. Exp. Med.* (2021a) **218** 29. DOI: 10.1084/jem.20210554
7. Bastard P., Gervais A., Le Voyer T., Rosain J., Philippot Q., Manry J., Michailidis E., Hoffmann H.H., Eto S., Garcia-Prat M.. **Autoantibodies neutralizing type I IFNs are present in ∼4% of uninfected individuals over 70 years old and account for ∼20% of COVID-19 deaths**. *Sci. Immunol.* (2021b) **6**. DOI: 10.1126/sciimmunol.abl4340
8. Bastard P., Vazquez S., Liu J., Laurie M.T., Wang C.Y., Gervais A., Le Voyer T., Bizien L., Zamecnik C., Philippot Q.. **Vaccine breakthrough hypoxemic COVID-19 pneumonia in patients with auto-Abs neutralizing type I IFNs**. *Sci. Immunol.* (2022) **19**. DOI: 10.1126/sciimmunol.abp8966
9. Bencze D., Fekete T., Pázmándi K.. **Type I interferon production of plasmacytoid dendritic cells under control**. *Int. J. Mol. Sci.* (2021) **22** 4190. DOI: 10.3390/ijms22084190
10. Bennett T.D., Moffitt R.A., Hajagos J.G., Amor B., Anand A., Bissell M.M., Bradwell K.R., Bremer C., Byrd J.B., Denham A.. **Clinical characterization and prediction of clinical severity of SARS-CoV-2 infection among US adults using data from the US national COVID cohort collaborative**. *JAMA Netw. Open* (2021) **4**. DOI: 10.1001/jamanetworkopen.2021.16901
11. Beutler B.. **Inferences, questions and possibilities in Toll-like receptor signalling**. *Nature* (2004) **430** 257-263. DOI: 10.1038/nature02761
12. Boisson B., Casanova J.L.. **TLR8 gain of function: A tall surprise**. *Blood* (2021) **137** 2420-2422. DOI: 10.1182/blood.2020010463
13. Borie R., Bouvry D., Cottin V., Gauvain C., Cazes A., Debray M.P., Cadranel J., Dieude P., Degot T., Dominique S.. **Regulator of telomere length 1 (RTEL1) mutations are associated with heterogeneous pulmonary and extra-pulmonary phenotypes**. *Eur. Respir. J.* (2019) **53** 1800508. DOI: 10.1183/13993003.00508-2018
14. Bourgeois F.T., Gutiérrez-Sacristán A., Keller M.S., Liu M., Hong C., Bonzel C.L., Tan A.L.M., Aronow B.J., Boeker M., Booth J.. **International analysis of electronic health records of children and youth hospitalized with COVID-19 infection in 6 countries**. *JAMA Netw. Open* (2021) **4**. DOI: 10.1001/jamanetworkopen.2021.12596
15. Brodin P.. **Immune determinants of COVID-19 disease presentation and severity**. *Nat. Med.* (2021) **27** 28-33. DOI: 10.1038/s41591-020-01202-8
16. Brotons P., Launes C., Buetas E., Fumado V., Henares D., de Sevilla M.F., Redin A., Fuente-Soro L., Cuadras D., Mele M.. **Susceptibility to severe acute respiratory syndrome coronavirus 2 infection among children and adults: A seroprevalence study of family households in the Barcelona metropolitan region, Spain**. *Clin. Infect. Dis.* (2021) **72** e970-e977. DOI: 10.1093/cid/ciaa1721
17. Bucciol G., Moens L., Corveleyn A., Dreesman A., Meyts I.. **A novel kindred with MyD88 deficiency**. *J. Clin. Immunol.* (2022a) **42** 885-888. DOI: 10.1007/s10875-022-01240-6
18. Bucciol G., Desmet L., Corveleyn A., Meyts I., Corveleyn A., Meyts I.. **Pathogenic P554S variant in TLR3 in a patient with severe influenza pneumonia**. *J. Clin. Immunol.* (2022b) **42** 430-432. DOI: 10.1007/s10875-021-01172-7
19. Cannistraci C.V., Valsecchi M.G., Capua I.. **Age-sex population adjusted analysis of disease severity in epidemics as a tool to devise public health policies for COVID-19**. *Sci. Rep.* (2021) **11** 1-8. DOI: 10.1038/s41598-021-89615-4
20. Cardenes M., von Bernuth H., García-Saavedra A., Santiago E., Puel A., Ku C.L., Emile J.F., Picard C., Casanova J.L., Colino E.. **Autosomal recessive interleukin-1 receptor-associated kinase 4 deficiency in fourth-degree relatives**. *J. Pediatr.* (2006) **148** 549-551. DOI: 10.1016/j.jpeds.2005.12.012
21. Casanova J.L., Abel L.. **Mechanisms of viral inflammation and disease in humans**. *Science* (2021) **374** 1080-1086. DOI: 10.1126/science.abj7965
22. Casanova J.L., Abel L.. **From rare disorders of immunity to common determinants of infection: Following the mechanistic thread**. *Cell* (2022) **185** 3086-3103. DOI: 10.1016/j.cell.2022.07.004
23. Casanova J.-L., Abel L., Quintana-Murci L.. **Human TLRs and IL-1Rs in host defense: natural insights from evolutionary, epidemiological, and clinical genetics**. *Annu. Rev. Immunol.* (2011) **29** 447-491. DOI: 10.1146/annurev-immunol-030409-101335
24. Caserta M.T., McDermott M.P., Dewhurst S., Schnabel K., Carnahan J.A., Gilbert L., Lathan G., Lofthus G.K., Hall C.B.. **Human herpesvirus 6 (HHV6) DNA persistence and reactivation in healthy children**. *J. Pediatr.* (2004) **145** 478-484. DOI: 10.1016/j.jpeds.2004.06.016
25. Casrouge A., Zhang S.Y., Eidenschenk C., Jouanguy E., Puel A., Yang K., Alcais A., Picard C., Mahfoufi N., Nicolas N.. **Herpes simplex virus encephalitis in human UNC-93B deficiency**. *Science* (2006) **314** 308-312. DOI: 10.1126/science.1128346
26. **Isolation and Precautions for People with COVID-19**. (2022)
27. Chen J., Jing H., Martin-Nalda A., Bastard P., Rivière J.G., Liu Z., Colobran R., Lee D., Tung W., Manry J.. **Inborn errors of tlr3-or mda5-dependent type i ifn immunity in children with enterovirus rhombencephalitis**. *J. Exp. Med.* (2021) **218**. DOI: 10.1084/jem.20211349
28. Chiale C., Greene T.T., Zuniga E.I.. **Interferon induction, evasion, and paradoxical roles during SARS-CoV-2 infection**. *Immunol. Rev.* (2022) **309** 12-24. DOI: 10.1111/imr.13113
29. Chiriaco M., Di Matteo G., Conti F., Petricone D., De Luca M., Di Cesare S., Cifaldi C., De Vito R., Zoccolillo M., Serafinelli J.. **First case of patient with two homozygous mutations in MYD88 and CARD9 genes presenting with pyogenic bacterial infections, elevated IgE, and persistent EBV viremia**. *Front. Immunol.* (2019) **10** 130. DOI: 10.3389/fimmu.2019.00130
30. Cho K., Yamada M., Agematsu K., Kanegane H., Miyake N., Ueki M., Akimoto T., Kobayashi N., Ikemoto S., Tanino M.. **Heterozygous mutations in OAS1 cause infantile-onset pulmonary alveolar proteinosis with hypogammaglobulinemia**. *Am. J. Hum. Genet.* (2018) **102** 480-486. DOI: 10.1016/j.ajhg.2018.01.019
31. Chua G.T., Wong J.S.C., Lam I., Ho P.P.K., Chan W.H., Yau F.Y.S., Rosa Duque J.S., Ho A.C.C., Siu K.K., Cheung T.W.Y.. **Clinical characteristics and transmission of COVID-19 in children and youths during 3 waves of outbreaks in Hong Kong**. *JAMA Netw. Open* (2021) **4**. DOI: 10.1001/jamanetworkopen.2021.8824
32. Conway D.H., Dara J., Bagashev A., Sullivan K.E.. **Myeloid differentiation primary response gene 88 (MyD88) deficiency in a large kindred**. *J. Allergy Clin. Immunol.* (2010) **126** 172-175. DOI: 10.1016/j.jaci.2010.04.014
33. Deyà-Martínez A., García-García A., Gonzalez-Navarro E.A., Yiyi L., Vlagea A., Jordan I., Fumadó V., Fortuny C., Español M., Launes C.. **COVID-19 in children and young adults with moderate/severe inborn errors of immunity in a high burden area in pre-vaccine era**. *Clin. Immunol.* (2021) **230** 108821. DOI: 10.1016/j.clim.2021.108821
34. Diebold S.S., Kaisho T., Hemmi H., Akira S., Reis e Sousa C., Sousa C.. **Innate antiviral responses by means of TLR7-mediated recognition of single-stranded RNA**. *Science* (2004) **303** 1529-1531. DOI: 10.1126/science.1093616
35. Fallerini C., Daga S., Mantovani S., Benetti E., Picchiotti N., Francisci D., Paciosi F., Schiaroli E., Baldassarri M., Fava F.. **Association of toll-like receptor 7 variants with life-threatening COVID-19 disease in males: Findings from a nested case-control study**. *eLife* (2021) **10**. DOI: 10.7554/eLife.67569
36. Filbin M.R., Mehta A., Schneider A.M., Kays K.R., Guess J.R., Gentili M., Fenyves B.G., Charland N.C., Gonye A.L.K., Gushterova I.. **Longitudinal proteomic analysis of severe COVID-19 reveals survival-associated signatures, tissue-specific cell death, and cell-cell interactions**. *Cell Rep. Med.* (2021) **2** 100287. DOI: 10.1016/j.xcrm.2021.100287
37. Fomenko A., Weibel S., Moezi H., Menger K., Schmucker C., Metzendorf M.I., Motschall E., Falcone V., Huzly D., Panning M.. **Assessing severe acute respiratory syndrome coronavirus 2 infectivity by reverse-transcription polymerase chain reaction: A systematic review and meta-analysis**. *Rev. Med. Virol.* (2022) **32**. DOI: 10.1002/rmv.2342
38. Galani I.E., Rovina N., Lampropoulou V., Triantafyllia V., Manioudaki M., Pavlos E., Koukaki E., Fragkou P.C., Panou V., Rapti V.. **Untuned antiviral immunity in COVID-19 revealed by temporal type I/III interferon patterns and flu comparison**. *Nat. Immunol.* (2020) **22** 32-40. DOI: 10.1038/s41590-020-00840-x
39. Gisby J., Clarke C.L., Medjeral-Thomas N., Malik T.H., Papadaki A., Mortimer P.M., Buang N.B., Lewis S., Pereira M., Toulza F.. **Longitudinal proteomic profiling of dialysis patients with COVID-19 reveals markers of severity and predictors of death**. *Elife* (2021) **10** 1-30. DOI: 10.7554/eLife.64827
40. Goudouris E.S., Pinto-Mariz F., Mendonça L.O., Aranda C.S., Guimarães R.R., Kokron C., Barros M.T., Anísio F., Alonso M.L.O., Marcelino F.. **Outcome of SARS-CoV-2 infection in 121 patients with inborn errors of immunity: A cross-sectional study**. *J. Clin. Immunol.* (2021) **41** 1479-1489. DOI: 10.1007/s10875-021-01066-8
41. Guo Y., Audry M., Ciancanelli M., Alsina L., Azevedo J., Herman M., Anguiano E., Sancho-Shimizu V., Lorenzo L., Pauwels E.. **Herpes simplex virus encephalitis in a patient with complete TLR3 deficiency: TLR3 is otherwise redundant in protective immunity**. *J. Exp. Med.* (2011) **208** 2083-2098. DOI: 10.1084/jem.20101568
42. Hadjadj J., Yatim N., Barnabei L., Corneau A., Boussier J., Smith N., Péré H., Charbit B., Bondet V., Chenevier-Gobeaux C.. **Impaired type I interferon activity and inflammatory responses in severe COVID-19 patients**. *Science* (2020) **369** 718-724. DOI: 10.1126/science.abc6027
43. Hänzelmann S., Castelo R., Guinney J.. **GSVA: gene set variation analysis for microarray and RNA-seq data**. *BMC Bioinformatics* (2013) **14** 7. DOI: 10.1186/1471-2105-14-7
44. Hautala T., Chen J., Tervonen L., Partanen T., Winqvist S., Lehtonen J., Saarela J., Kraatari M., Kuismin O., Vuorinen T.. **Herpes simplex virus 2 encephalitis in a patient heterozygous for a TLR3 mutation**. *Neurol. Genet.* (2020) **6**. DOI: 10.1212/NXG.0000000000000532
45. Hollowell J.G., van Assendelft O.W., Gunter E.W., Lewis B.G., Najjar M., Pfeiffer C.. **Hematological and iron-related analytes-reference data for persons aged 1 year and over: United States, 1988–94**. *Vital Health Stat. 11* (2005) 1-156. PMID: 15782774
46. Jia A., James E., Lu H.Y., Sharma M., Modi B.P., Biggs C.M., Hildebrand K.J., Chomyn A., Erdle S., Kular H., Turvey S.E.. **Clinical IRAK4 deficiency caused by homozygosity for the novel IRAK4 (c.1049delG, p.Gly350Glufs*15) variant**. *Cold Spring Harb. Mol. Case Stud.* (2020) **6** a005298. DOI: 10.1101/mcs.a005298
47. Jouanguy E., Béziat V., Mogensen T.H., Casanova J.L., Tangye S.G., Zhang S.Y.. **Human inborn errors of immunity to herpes viruses**. *Curr. Opin. Immunol.* (2020) **62** 106-122. DOI: 10.1016/j.coi.2020.01.004
48. Kawai T., Akira S.. **Toll-like receptors and their crosstalk with other innate receptors in infection and immunity**. *Immunity* (2011) **34** 637-650. DOI: 10.1016/j.immuni.2011.05.006
49. Kim Y.M., Shin E.C.. **Type I and III interferon responses in SARS-CoV-2 infection**. *Exp. Mol. Med.* (2021) **53** 750-760. DOI: 10.1038/s12276-021-00592-0
50. Knock E.S., Whittles L.K., Lees J.A., Perez-Guzman P.N., Verity R., FitzJohn R.G., Gaythorpe K.A.M., Imai N., Hinsley W., Okell L.C.. **Key epidemiological drivers and impact of interventions in the 2020 SARS-CoV-2 epidemic in England**. *Sci. Transl. Med.* (2021) **13** 4262. DOI: 10.1126/scitranslmed.abg4262
51. Kogoj R., Korva M., Knap N., Resman Rus K., Pozvek P., Avšič-Županc T., Poljak M.. **Comparative evaluation of six SARS-CoV-2 real-time RT-PCR diagnostic approaches shows substantial genomic variant-dependent intra- and inter-test variability, poor interchangeability of cycle threshold and complementary turn-around times**. *Pathogens* (2022) **11** 462. DOI: 10.3390/pathogens11040462
52. Komaroff A.L., Pellett P.E., Jacobson S.. **Human herpesviruses 6a and 6b in brain diseases: Association versus causation**. *Clin. Microbiol. Rev.* (2020) **34** 1-36. DOI: 10.1128/CMR.00143-20
53. Kooistra E.J., van Berkel M., van Kempen N.F., van Latum C.R.M., Bruse N., Frenzel T., van den Berg M.J.W., Schouten J.A., Kox M., Pickkers P.. **Dexamethasone and tocilizumab treatment considerably reduces the value of C-reactive protein and procalcitonin to detect secondary bacterial infections in COVID-19 patients**. *Crit. Care* (2021) **25** 281. DOI: 10.1186/s13054-021-03717-z
54. Kosmicki J.A., Horowitz J.E., Banerjee N., Lanche R., Marcketta A., Maxwell E., Bai X., Sun D., Backman J.D., Sharma D.. **Pan-ancestry exome-wide association analyses of COVID-19 outcomes in 586,157 individuals**. *Am. J. Hum. Genet.* (2021) **108** 1350-1355. DOI: 10.1016/j.ajhg.2021.05.017
55. Ku C.L., von Bernuth H., Picard C., Zhang S.Y., Chang H.H., Yang K., Chrabieh M., Issekutz A.C., Cunningham C.K., Gallin J.. **Selective predisposition to bacterial infections in IRAK-4-deficient children: IRAK-4-dependent TLRs are otherwise redundant in protective immunity**. *J. Exp. Med.* (2007) **204** 2407-2422. DOI: 10.1084/jem.20070628
56. Kuo C.Y., Ku C.L., Lim H.K., Hsia S.H., Lin J.J., Lo C.C., Ding J.Y., Kuo R.L., Casanova J.L., Zhang S.Y.. **Life-threatening enterovirus 71 encephalitis in unrelated children with autosomal dominant TLR3 deficiency**. *J. Clin. Immunol.* (2022) **42** 606-617. DOI: 10.1007/s10875-021-01170-9
57. Kyung Lim H., Huang S.X.L., Chen J., Kerner G., Gilliaux O., Bastard P., Dobbs K., Hernandez N., Goudin N., Hasek M.L.. **Severe influenza pneumonitis in children with inherited TLR3 deficiency**. *J. Exp. Med.* (2019) **216** 2038-2056. DOI: 10.1084/jem.20181621
58. Le Vu S., Jones G., Anna F., Rose T., Richard J.B., Bernard-Stoecklin S., Goyard S., Demeret C., Helynck O., Escriou N.. **Prevalence of SARS-CoV-2 antibodies in France: Results from nationwide serological surveillance**. *Nat. Commun.* (2021) **12** 1-7. DOI: 10.1038/s41467-021-23233-6
59. Lee J.S., Park S., Jeong H.W., Ahn J.Y., Choi S.J., Lee H., Choi B., Nam S.K., Sa M., Kwon J.S.. **Immunophenotyping of COVID-19 and influenza highlights the role of type I interferons in development of severe COVID-19**. *Sci. Immunol.* (2020) **5** 1554. DOI: 10.1126/sciimmunol.abd1554
60. Lee D., Le Pen J., Yatim A., Dong B., Aquino Y., Ogishi M., Pescarmona R., Talouarn E., Rinchai D., Zhang P.. **Inborn errors of OAS-RNase L in SARS-CoV-2-related multisystem inflammatory syndrome in children**. *Science* (2022) **1**. DOI: 10.1126/science.abo3627
61. Lemarquis A., Campbell T., Aranda-Guillén M., Hennings V., Brodin P., Kämpe O., Blennow K., Zetterberg H., Wennerås C., Eriksson K.. **Severe COVID-19 in an APS1 patient with interferon autoantibodies treated with plasmapheresis**. *J. Allergy Clin. Immunol.* (2021) **148** 96-98. DOI: 10.1016/j.jaci.2021.03.034
62. Lévy R., Zhang P., Bastard P., Dorgham K., Melki I., Hadchouel A., Hartoularos G.C., Neven B., Castelle M., Roy C.. **Monoclonal antibody-mediated neutralization of SARS-CoV-2 in an IRF9-deficient child**. *Proc. Natl. Acad. Sci. USA* (2021) **118**. DOI: 10.1073/pnas.2114390118
63. Liu G., Gack M.U.. **Distinct and orchestrated functions of RNA sensors in innate immunity**. *Immunity* (2020) **53** 26-42. DOI: 10.1016/j.immuni.2020.03.017
64. Loske J., Röhmel J., Lukassen S., Stricker S., Magalhães V.G., Liebig J., Chua R.L., Thürmann L., Messingschlager M., Seegebarth A.. **Pre-activated antiviral innate immunity in the upper airways controls early SARS-CoV-2 infection in children**. *Nat. Biotechnol.* (2022) **40** 319-324. DOI: 10.1038/s41587-021-01037-9
65. Lucas C., Wong P., Klein J., Castro T.B.R., Silva J., Sundaram M., Ellingson M.K., Mao T., Oh J.E., Israelow B.. **Longitudinal analyses reveal immunological misfiring in severe COVID-19**. *Nature* (2020) **584** 463-469. DOI: 10.1038/s41586-020-2588-y
66. Magg T., Okano T., Koenig L.M., Boehmer D.F.R., Schwartz S.L., Inoue K., Heimall J., Licciardi F., Ley-Zaporozhan J., Ferdman R.M.. **Heterozygous OAS1 gain-of-function variants cause an autoinflammatory immunodeficiency**. *Sci. Immunol.* (2021) **6**. DOI: 10.1126/sciimmunol.abf9564
67. Mahmood H.Z., Madhavarapu S., Almuqamam M.. **Varying illness severity in patients with MyD88 deficiency infected with coronavirus SARS-CoV-2**. *Pediatrics* (2021) **147** 453-454. DOI: 10.1542/PEDS.147.3MA5.453
68. Maleknia S., Tavassolifar M.J., Mottaghitalab F., Zali M.R., Meyfour A.. **Identifying novel host-based diagnostic biomarker panels for COVID-19: A whole-blood/nasopharyngeal transcriptome meta-analysis**. *Mol. Med.* (2022) **28** 86. DOI: 10.1186/s10020-022-00513-5
69. Manson J.J., Crooks C., Naja M., Ledlie A., Goulden B., Liddle T., Khan E., Mehta P., Martin-Gutierrez L., Waddington K.E.. **COVID-19-associated hyperinflammation and escalation of patient care: a retrospective longitudinal cohort study**. *Lancet Rheumatol* (2020) **2** e594-e602. DOI: 10.1016/S2665-9913(20)30275-7
70. Mantovani A., Rinaldi E., Zusi C., Beatrice G., Saccomani M.D., Dalbeni A.. **Coronavirus disease 2019 (COVID-19) in children and/or adolescents: A meta-analysis**. *Pediatr. Res.* (2020) **89** 733-737. DOI: 10.1038/s41390-020-1015-2
71. Martin B., DeWitt P.E., Russell S., Anand A., Bradwell K.R., Bremer C., Gabriel D., Girvin A.T., Hajagos J.G., McMurry J.A.. **Characteristics, outcomes, and severity risk factors associated with SARS-CoV-2 infection among children in the US national COVID cohort collaborative**. *JAMA Netw. Open* (2022a) **5**. DOI: 10.1001/jamanetworkopen.2021.43151
72. Martin B., DeWitt P.E., Russell S., Sanchez-Pinto L.N., Haendel M.A., Moffitt R., Bennett T.D.. **Acute upper airway disease in children with the omicron (B.1.1.529) variant of SARS-CoV-2-A report from the US national COVID cohort collaborative**. *JAMA Pediatr.* (2022b) **176** 819-821. DOI: 10.1001/jamapediatrics.2022.1110
73. Matuozzo D., Talouarn E., Marchal A., Manry J., Seeleuthner Y., Zhang Y., Bolze A., Chaldebas M., Milisavljevic B., Zhang P.. **Rare predicted loss-of-function variants of type I IFN immunity genes are associated with life-threatening COVID-19**. *MedRxiv* (2022). DOI: 10.1101/2022.10.22.22281221
74. **Tests and Procedures.**. (2022)
75. Meisel C., Akbil B., Meyer T., Lankes E., Corman V.M., Staudacher O., Unterwalder N., Kölsch U., Drosten C., Mall M.A.. **Mild COVID-19 despite autoantibodies against type I IFNs in autoimmune polyendocrine syndrome type 1**. *J. Clin. Invest.* (2021) **131**. DOI: 10.1172/JCI150867
76. Milito C., Lougaris V., Giardino G., Punziano A., Vultaggio A., Carrabba M., Cinetto F., Scarpa R., Delle Piane R.M., Baselli L.. **Clinical outcome, incidence, and SARS-CoV-2 infection-fatality rates in Italian patients with inborn errors of immunity**. *J. Allergy Clin. Immunol. Pract.* (2021) **9** 2904-2906.e2. DOI: 10.1016/j.jaip.2021.04.017
77. **Ministerio de Sanidad - Profesionales - Situación actual Coronavirus**. (2022)
78. Mohsin M., Mahmud S.. **Omicron SARS-CoV-2 variant of concern A review on its transmissibility, immune evasion, reinfection, and severity**. *Medicine* (2022) **101**. DOI: 10.1097/MD.0000000000029165
79. Moore C., Blumhagen R.Z., Yang I.V., Walts A., Powers J., Walker T., Bishop M., Russell P., Vestal B., Cardwell J.. **Resequencing study confirms that host defense and cell senescence gene variants contribute to the risk of idiopathic pulmonary fibrosis**. *Am. J. Respir. Crit. Care Med.* (2019) **200** 199-208. DOI: 10.1164/rccm.201810-1891OC
80. Nakanishi T., Pigazzini S., Degenhardt F., Cordioli M., Butler-Laporte G., Maya-Miles D., Bujanda L., Bouysran Y., Niemi M.E.K., Palom A.. **Age-dependent impact of the major common genetic risk factor for COVID-19 on severity and mortality**. *J. Clin. Invest.* (2021) **131** 49. DOI: 10.1172/JCI152386
81. Navaratnam A.V., Gray W.K., Day J., Wendon J., Briggs T.W.R.. **Patient factors and temporal trends associated with COVID-19 in-hospital mortality in england: An observational study using administrative data**. *Lancet Respir. Med.* (2021) **9** 397-406. DOI: 10.1016/S2213-2600(20)30579-8
82. Newton C.A., Oldham J.M., Applegate C., Carmichael N., Powell K., Dilling D., Schmidt S.L., Scholand M.B., Armanios M., Garcia C.K.. **The role of genetic testing in pulmonary fibrosis: A perspective from the pulmonary fibrosis foundation genetic testing work group**. *Chest* (2022) **162** 394-405. DOI: 10.1016/j.chest.2022.03.023
83. Ng D.L., Granados A.C., Santos Y.A., Servellita V., Goldgof G.M., Meydan C., Sotomayor-Gonzalez A., Levine A.G., Balcerek J., Han L.M.. **A diagnostic host response biosignature for COVID-19 from RNA profiling of nasal swabs and blood**. *Sci. Adv.* (2021) **7**. DOI: 10.1126/sciadv.abe5984
84. Nishimura S., Kobayashi Y., Ohnishi H., Moriya K., Tsumura M., Sakata S., Mizoguchi Y., Takada H., Kato Z., Sancho-Shimizu V.. **IRAK4 deficiency presenting with anti-NMDAR encephalitis and HHV6 reactivation**. *J. Clin. Immunol.* (2021) **41** 125-135. DOI: 10.1007/s10875-020-00885-5
85. Ong S.W.X., Chiew C.J., Ang L.W., Mak T.M., Cui L., Toh M.P.H.S., Lim Y.D., Lee P.H., Lee T.H., Chia P.Y.. **Clinical and virological features of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) variants of concern: A retrospective cohort study comparing B.1.1.7 (Alpha), B.1.351 (Beta), and B.1.617.2 (Delta)**. *Clin. Infect. Dis.* (2022) **75** e1128-e1136. DOI: 10.1093/cid/ciab721
86. Onodi F., Bonnet-Madin L., Meertens L., Karpf L., Poirot J., Zhang S.-Y., Picard C., Puel A., Jouanguy E., Zhang Q.. **SARS-CoV-2 induces human plasmacytoid predendritic cell diversification via UNC93B and IRAK4**. *J. Exp. Med* (2021) **218** e20201387. DOI: 10.1084/jem.20201387
87. O’Driscoll M., Ribeiro Dos Santos G., Wang L., Cummings D.A.T., Azman A.S., Paireau J., Fontanet A., Cauchemez S., Salje H.. **Age-specific mortality and immunity patterns of SARS-CoV-2**. *Nature* (2021) **590** 140-145. DOI: 10.1038/s41586-020-2918-0
88. **SARS-CoV-2 variants in analyzed sequences, Spain**. (2023)
89. Pandey U., Greninger A.L., Levin G.R., Jerome K.R., Anand V.C., Dien Bard J., Bard J.D.. **Pathogen or Bystander: Clinical significance of detecting human herpesvirus 6 in pediatric cerebrospinal fluid**. *J. Clin. Microbiol.* (2020) **58**. DOI: 10.1128/JCM.00313-20
90. Partanen T., Chen J., Lehtonen J., Kuismin O., Rusanen H., Vapalahti O., Vaheri A., Anttila V.J., Bode M., Hautala N.. **Heterozygous TLR3 mutation in patients with hantavirus encephalitis**. *J. Clin. Immunol.* (2020) **40** 1156-1162. DOI: 10.1007/s10875-020-00834-2
91. Pays E., Radwanska M., Magez S.. **The pathogenesis of african trypanosomiasis**. *Annu. Rev. Pathol.* (2022) **18** 19-45. DOI: 10.1146/annurev-pathmechdis-031621-025153
92. Pessoa N.L., Bentes A.A., de Carvalho A.L., de Souza Silva T.B., Alves P.A., de Sousa Reis E.V., Rodrigues T.A., Kroon E.G., Campos M.A.. **Case report: Hepatitis in a child infected with SARS-CoV-2 presenting toll-like receptor 7 Gln11Leu single nucleotide polymorphism**. *Virol. J.* (2021) **18** 180. DOI: 10.1186/s12985-021-01656-3
93. Picard C., Puel A., Bonnet M., Ku C.L., Bustamante J., Yang K., Soudais C., Dupuis S., Feinberg J., Fieschi C.. **Pyogenic bacterial infections in humans with IRAK-4 deficiency**. *Science* (2003) **299** 2076-2079. DOI: 10.1126/science.1081902
94. Picard C., von Bernuth H., Ghandil P., Chrabieh M., Levy O., Arkwright P.D., McDonald D., Geha R.S., Takada H., Krause J.C.. **Clinical features and outcome of patients with IRAK-4 and MyD88 deficiency**. *Medicine* (2010) **89** 403-425. DOI: 10.1097/MD.0b013e3181fd8ec3
95. Picard C., Casanova J.-L., Puel A.. **Infectious diseases in patients with IRAK-4, MyD88, NEMO, or IκBα deficiency**. *Clin. Microbiol. Rev.* (2011) **24** 490-497. DOI: 10.1128/CMR.00001-11
96. Pierce C.A., Preston-Hurlburt P., Dai Y., Aschner C.B., Cheshenko N., Galen B., Garforth S.J., Herrera N.G., Jangra R.K., Morano N.C.. **Immune responses to SARS-CoV-2 infection in hospitalized pediatric and adult patients**. *Sci. Transl. Med.* (2020) **12**. DOI: 10.1126/scitranslmed.abd5487
97. Quach H., Wilson D., Laval G., Patin E., Manry J., Guibert J., Barreiro L.B., Nerrienet E., Verschoor E., Gessain A.. **Different selective pressures shape the evolution of Toll-like receptors in human and African great ape populations**. *Hum. Mol. Genet.* (2013) **22** 4829-4840. DOI: 10.1093/hmg/ddt335
98. Reizis B.. **Plasmacytoid dendritic cells: Development, regulation, and function**. *Immunity* (2019) **50** 37-50. DOI: 10.1016/j.immuni.2018.12.027
99. Ren X., Wen W., Fan X., Hou W., Su B., Cai P., Li J., Liu Y., Tang F., Zhang F.. **COVID-19 immune features revealed by a large-scale single-cell transcriptome atlas**. *Cell* (2021) **184** 1895-1913.e19. DOI: 10.1016/j.cell.2021.01.053
100. Rodrigues T.S., de Sá K.S.G., Ishimoto A.Y., Becerra A., Oliveira S., Almeida L., Gonçalves A.V., Perucello D.B., Andrade W.A., Castro R.. **Inflammasomes are activated in response to SARS-CoV-2 infection and are associated with COVID-19 severity in patients**. *J. Exp. Med.* (2020) **218**. DOI: 10.1084/jem.20201707
101. Russell C.D., Valanciute A., Gachanja N.N., Stephen J., Penrice-Randal R., Armstrong S.D., Clohisey S., Wang B., Al Qsous W., Wallace W.A.. **Tissue proteomic analysis identifies mechanisms and stages of immunopathology in fatal COVID-19**. *Am. J. Respir. Cell Mol. Biol.* (2022) **66** 196-205. DOI: 10.1165/rcmb.2021-0358OC
102. Sah P., Fitzpatrick M.C., Zimmer C.F., Abdollahi E., Juden-Kelly L., Moghadas S.M., Singer B.H., Galvani A.P.. **Asymptomatic SARS-CoV-2 infection: A systematic review and meta-analysis**. *Proc. Natl. Acad. Sci. USA* (2021) **118**. DOI: 10.1073/pnas.2109229118
103. Santpere G., Telford M., Andrés-Benito P., Navarro A., Ferrer I.. **The presence of human herpesvirus 6 in the brain in health and disease**. *Biomolecules* (2020) **10** 1520. DOI: 10.3390/biom10111520
104. Schober T., Caya C., Barton M., Bayliss A., Bitnun A., Bowes J., Brenes-Chacon H., Bullard J., Cooke S., Dewan T.. **Risk factors for severe PCR-positive SARS-CoV-2 infection in hospitalised children**. *BMJ Paediatr. Open* (2022) **6**. DOI: 10.1136/bmjpo-2022-001440
105. Solanich X., Vargas-Parra G., van der Made C.I., Simons A., Schuurs-Hoeijmakers J., Antolí A., del Valle J., Rocamora-Blanch G., Setié F., Esteller M.. **Genetic screening for TLR7 variants in young and previously healthy men with severe COVID-19**. *Front. Immunol.* (2021a) **12** 719115. DOI: 10.3389/fimmu.2021.719115
106. Solanich X., Rigo-Bonnin R., Gumucio V.D., Bastard P., Rosain J., Philippot Q., Perez-Fernandez X.L., Fuset-Cabanes M.P., Gordillo-Benitez M.Á., Suarez-Cuartin G.. **Pre-existing autoantibodies neutralizing high concentrations of type I interferons in almost 10% of COVID-19 patients admitted to intensive care in Barcelona**. *J. Clin. Immunol.* (2021b) **41** 1733-1744. DOI: 10.1007/s10875-021-01136-x
107. Stuart B.D., Choi J., Zaidi S., Xing C., Holohan B., Chen R., Choi M., Dharwadkar P., Torres F., Girod C.E.. **Exome sequencing links mutations in PARN and RTEL1 with familial pulmonary fibrosis and telomere shortening**. *Nat. Genet.* (2015) **47** 512-517. DOI: 10.1038/NG.327810.1038/ng.3278
108. Tangye S.G., Al-Herz W., Bousfiha A., Cunningham-Rundles C., Franco J.L., Holland S.M., Klein C., Morio T., Oksenhendler E., Picard C.. **Human inborn errors of immunity: 2022 update on the classification from the international union of immunological societies expert committee**. *J. Clin. Immunol.* (2022) **42** 1473-1507. DOI: 10.1007/s10875-022-01289-3
109. Tepe Z.G., Yazıcı Y.Y., Tank U., Köse L.I., Özer M., Aytekin C., Belkaya S.. **Inherited IRAK-4 deficiency in acute human herpesvirus-6 encephalitis**. *J. Clin. Immunol.* (2022) **1** 1. DOI: 10.1007/S10875-022-01369-410.1007/s10875-022-01369-4
110. Unterman A., Sumida T.S., Nouri N., Yan X., Zhao A.Y., Gasque V., Schupp J.C., Asashima H., Liu Y., Cosme C.. **Single-cell multi-omics reveals dyssynchrony of the innate and adaptive immune system in progressive COVID-19**. *Nat. Commun.* (2022) **13** 440. DOI: 10.1038/s41467-021-27716-4
111. van der Made C.I., Simons A., Schuurs-Hoeijmakers J., van den Heuvel G., Mantere T., Kersten S., van Deuren R.C., Steehouwer M., van Reijmersdal S.V., Jaeger M.. **Presence of genetic variants among young men with severe COVID-19**. *JAMA* (2020) **324** 663-673. DOI: 10.1001/jama.2020.13719
112. von Bernuth H., Ku C.L., Rodriguez-Gallego C., Zhang S., Garty B.Z., Maródi L., Chapel H., Chrabieh M., Miller R.L., Picard C.. **A fast procedure for the detection of defects in Toll-like receptor signaling**. *Pediatrics* (2006) **118** 2498-2503. DOI: 10.1542/peds.2006-1845
113. von Bernuth H., Picard C., Jin Z., Pankla R., Xiao H., Ku C.L., Chrabieh M., Mustapha I.B., Ghandil P., Camcioglu Y.. **Pyogenic bacterial infections in humans with MyD88 deficiency**. *Science* (2008) **321** 691-696. DOI: 10.1126/science.1158298
114. von Bernuth H., Picard C., Puel A., Casanova J.L.. **Experimental and natural infections in MyD88- and IRAK-4-deficient mice and humans**. *Eur. J. Immunol.* (2012) **42** 3126-3135. DOI: 10.1002/eji.201242683
115. Walne A.J., Vulliamy T., Kirwan M., Plagnol V., Dokal I.. **Constitutional mutations in RTEL1 cause severe dyskeratosis congenita**. *Am. J. Hum. Genet.* (2013) **92** 448-453. DOI: 10.1016/j.ajhg.2013.02.001
116. Weller S., Bonnet M., Delagreverie H., Israel L., Chrabieh M., Maródi L., Rodriguez-Gallego C., Garty B.Z., Roifman C., Issekutz A.C.. **IgM+IgD+CD27+ B cells are markedly reduced in IRAK-4-, MyD88-, and TIRAP- but not UNC-93B-deficient patients**. *Blood* (2012) **120** 4992-5001. DOI: 10.1182/blood-2012-07-440776
117. Westblade L.F., Simon M.S., Satlin M.J.. **Bacterial coinfections in coronavirus disease 2019**. *Trends Microbiol.* (2021) **29** 930-941. DOI: 10.1016/j.tim.2021.03.018
118. **Obesity and overweight.**. (2021)
119. Woodruff R.C., Campbell A.P., Taylor C.A., Chai S.J., Kawasaki B., Meek J., Anderson E.J., Weigel A., Monroe M.L., Reeg L.. **Risk factors for severe COVID-19 in children**. *Pediatrics* (2022) **149**. DOI: 10.1542/peds.2021-053418
120. Yang K., Puel A., Zhang S., Eidenschenk C., Ku C.L., Casrouge A., Picard C., von Bernuth H., Senechal B., Plancoulaine S.. **Human TLR-7-, -8-, and -9-mediated induction of IFN-α/β and -λ Is IRAK-4 dependent and redundant for protective immunity to viruses**. *Immunity* (2005) **23** 465-478. DOI: 10.1016/j.immuni.2005.09.016
121. Zhang S.Y., Jouanguy E., Sancho-Shimizu V., von Bernuth H., Yang K., Abel L., Picard C., Puel A., Casanova J.L.. **Human Toll-like receptor-dependent induction of interferons in protective immunity to viruses**. *Immunol. Rev.* (2007a) **220** 225-236. DOI: 10.1111/j.1600-065X.2007.00564.x
122. Zhang S.Y., Jouanguy E., Ugolini S., Smahi A., Elain G., Romero P., Segal D., Sancho-Shimizu V., Lorenzo L., Puel A.. **TLR3 deficiency in patients with herpes simplex encephalitis**. *Science* (2007b) **317** 1522-1527. DOI: 10.1126/science.1139522
123. Zhang S.Y., Jouanguy E., Zhang Q., Abel L., Puel A., Casanova J.L.. **Human inborn errors of immunity to infection affecting cells other than leukocytes: From the immune system to the whole organism**. *Curr. Opin. Immunol.* (2019) **59** 88-100. DOI: 10.1016/j.coi.2019.03.008
124. Zhang Q., Bastard P., Bolze A., Jouanguy E., Zhang S.Y., Cobat A., Notarangelo L.D., Su H.C., Abel L., Casanova J.L.. **Life-threatening COVID-19: Defective interferons unleash excessive inflammation**. *Med* (2020a) **1** 14-20. DOI: 10.1016/j.medj.2020.12.001
125. Zhang Q., Bastard P., Liu Z., Le Pen J., Moncada-Velez M., Chen J., Ogishi M., Sabli I.K.D., Hodeib S., Korol C.. **Inborn errors of type I IFN immunity in patients with life-threatening COVID-19**. *Science* (2020b) **370** eabd4570. DOI: 10.1126/science.abd4570
126. Zhang Q., Bastard P., Cobat A., Casanova J.L.. **Human genetic and immunological determinants of critical COVID-19 pneumonia**. *Nature* (2022a) **603** 587-598. DOI: 10.1038/s41586-022-04447-0
127. Zhang Q., Matuozzo D., Le Pen J., Lee D., Moens L., Asano T., Bohlen J., Liu Z., Moncada-Velez M., Kendir-Demirkol Y.. **Recessive inborn errors of type I IFN immunity in children with COVID-19 pneumonia**. *J. Exp. Med.* (2022b) **219**. DOI: 10.1084/jem.20220131
128. Zhao X.N., You Y., Cui X.M., Gao H.X., Wang G.L., Zhang S.B., Yao L., Duan L.J., Zhu K.L., Wang Y.L.. **Single-cell immune profiling reveals distinct immune response in asymptomatic COVID-19 patients**. *Signal Transduct. Target. Ther.* (2021) **6** 342. DOI: 10.1038/s41392-021-00753-7
129. Zhu L., Yang P., Zhao Y., Zhuang Z., Wang Z., Song R., Zhang J., Liu C., Gao Q., Xu Q.. **Single-Cell sequencing of peripheral mononuclear cells reveals distinct immune response landscapes of COVID-19 and influenza patients**. *Immunity* (2020) **53** 685-696.e3. DOI: 10.1016/j.immuni.2020.07.009
|
---
title: Highly purified and functionally stable in vitro expanded allospecific Tr1
cells expressing immunosuppressive graft-homing receptors as new candidates for
cell therapy in solid organ transplantation
authors:
- Saúl Arteaga-Cruz
- Arimelek Cortés-Hernández
- Evelyn Katy Alvarez-Salazar
- Katya Rosas-Cortina
- Christian Aguilera-Sandoval
- Luis E. Morales-Buenrostro
- Josefina M. Alberú-Gómez
- Gloria Soldevila
journal: Frontiers in Immunology
year: 2023
pmcid: PMC9998667
doi: 10.3389/fimmu.2023.1062456
license: CC BY 4.0
---
# Highly purified and functionally stable in vitro expanded allospecific Tr1 cells expressing immunosuppressive graft-homing receptors as new candidates for cell therapy in solid organ transplantation
## Abstract
The development of new strategies based on the use of Tr1 cells has taken relevance to induce long-term tolerance, especially in the context of allogeneic stem cell transplantation. Although Tr1 cells are currently identified by the co-expression of CD49b and LAG-3 and high production of interleukin 10 (IL-10), recent studies have shown the need for a more exhaustive characterization, including co-inhibitory and chemokines receptors expression, to ensure bona fide Tr1 cells to be used as cell therapy in solid organ transplantation. Moreover, the proinflammatory environment induced by the allograft could affect the suppressive function of Treg cells, therefore stability of Tr1 cells needs to be further investigated. Here, we establish a new protocol that allows long-term in vitro expansion of highly purified expanded allospecific Tr1 (Exp-allo Tr1). Our expanded Tr1 cell population becomes highly enriched in IL-10 producers (> $90\%$) and maintains high expression of CD49b and LAG-3, as well as the co-inhibitory receptors PD-1, CTLA-4, TIM-3, TIGIT and CD39. Most importantly, high dimensional analysis of Exp-allo Tr1 demonstrated a specific expression profile that distinguishes them from activated conventional T cells (T conv), showing overexpression of IL-10, CD39, CTLA-4 and LAG-3. On the other hand, Exp-allo Tr1 expressed a chemokine receptor profile relevant for allograft homing and tolerance induction including CCR2, CCR4, CCR5 and CXCR3, but lower levels of CCR7. Interestingly, Exp-allo Tr1 efficiently suppressed allospecific but not third-party T cell responses even after being expanded in the presence of proinflammatory cytokines for two extra weeks, supporting their functional stability. In summary, we demonstrate for the first time that highly purified allospecific Tr1 (Allo Tr1) cells can be efficiently expanded maintaining a stable phenotype and suppressive function with homing potential to the allograft, so they may be considered as promising therapeutic tools for solid organ transplantation.
## Introduction
T regulatory Type one (Tr1) cells are a subset of CD4+ T cells initially described in patients with mixed chimerism who did not develop Graft vs Host Disease (GvHD), after hematopoietic stem cell transplantation (HSCT) [1, 2]. Several reports have shown that Tr1 cells can be abundantly found in specific anatomical regions with IL-10-rich microenvironments such as the intestinal mucosa [3], promoting the tolerance induction towards microbiota and diet-derived antigens [4]. Nonetheless, the presence of Tr1 cells can be identified in peripheral blood and spleen, indicating that they can exert their functions both locally and systemically [5].
Tr1 cells are characterized by the co-expression of the specific surface markers: CD49b and LAG-3 (lymphocyte activation gene-3) which has facilitated their identification and purification [6]. Another relevant aspect of these cells is their capacity to produce high and intermediate amounts of IL-10 and TGF-β respectively, variable levels of IFN-γ and IL-2 and low or null production of IL-4 (reviewed in [7]). Although the production of anti-inflammatory cytokines has been described as one of their main suppressive mechanisms, the ability to exert cell-cell contact suppression (PD1 and CTLA-4), metabolic disruption (CD39 and CD73), as well as cytotoxicity (Granzyme B and perforin release) are other important features which regulate effector immune responses (reviewed in [8]). Importantly, Brockmann and co-workers have recently described a subpopulation within the CD49b+LAG-3+CD4+ Tr1 cells which co-expresses the co-inhibitory receptors (CIR) TIGIT, TIM-3 and PD1, and produce higher levels of IL-10. Moreover, CCR5+ CIR+ Tr1 cells appear to have greater immunoregulatory capacity [9].
Tr1 cells had previously been used as therapy tools in different preclinical models of intestinal inflammation [10] and autoimmune diseases, where they contribute to resolve effector T cell responses and reduce the clinical score disease (11–13). However, most of their therapeutic potential has been described within the allograft context, where in vitro differentiated Tr1 cells were able to efficiently induce tolerance towards allogeneic pancreatic islets and prevented GvHD development in HSCT patients (reviewed in [14]). In this context, IL-10-anergized donor T cells (IL-10-DLI) were able to promote cellular reconstitution in HSCT transplanted patients while reducing the risk of GvHD [15]. Other protocols based on T10-cells, a clinical grade population obtained after stimulation of CD4+ T cells with irradiated IL-10 derived allogeneic dendritic cells (DC10), have been safely used in patients with kidney transplant. Importantly, circulating T10-cells showed a stable tolerogenic gene signature thirty-six weeks post-transplantation [16]. Recently, a new protocol focused on HSC transplant patients with hematological malignancies used a population obtained from CD4+ T cells cultured with allogeneic DC10 in presence of IL-10 (T-allo10) as cell therapy, showing the maintenance of a stable population after patient’s infusion [17].
One of the important aspects to consider for the use of cellular therapy in transplanted patients is to ensure the stability of infused regulatory T (Tregs) cells in the transplanted patient. In this context, it has been reported that activation of thymic and induced Tregs cells expressing FOXP3 in the presence of proinflammatory cytokines, can promote a downregulation of FOXP3, leading to a loss of their suppressive capacity, while acquiring characteristics from inflammatory T cell populations ([18] and reviewed in [19]). On the other hand, there is growing evidence that chemokine receptor expression in infiltrated allograft cells is closely related to either the rejection or maintenance of organ transplants, as previously reported for FOXP3+ Tregs cells [20]. However, to date, scarce information is available about the effect of inflammatory cytokines and chemokine receptor expression on Tr1 stability and homing, respectively. In addition, it has been demonstrated that specific chemokine receptors (CR) expressed on Tregs cells, play a key role in transplantation tolerance, by favoring their homing Tregs cells towards the allograft (CCR2, CXCR3 and CCR4) or draining lymph nodes (CCR7) [20, 21]. Initial studies reported the expression of CR related to both Type 1 helper (Th1) (CXCR3 and CCR5) and (Th2) subsets (CCR3, CCR4 and CCR8), while CCR5 has been proposed as a phenotypic marker for Tr1 cells [9]. However, a more detailed characterization is still needed to evaluate their functional relevance in allograft tolerance.
Although current protocols to differentiate alloantigen-specific Tr1 cells in vitro have shown satisfactory results (15–17), none of these protocols have successfully obtained pure allospecific Tr1 cells and, furthermore, new methodologies are needed aiming at large-scale Tr1 cell production. In the present work, we describe for the first time the feasibility of large scale allospecific Tr1 (Allo Tr1) cells expansion, with high purity and stable function as potential tools for solid organ transplantation.
## Healthy donor samples
Buffy coats products from healthy donors whole blood were collected in the Blood Bank of “Instituto Nacional de Enfermedades Respiratorias (INER) Ismael Cosío Villegas”, Mexico City, after written informed consent, in accordance with the local human ethics committee procedures. This protocol was approved by the Committees Medical Ethics at the Instituto de Investigaciones Biomédicas (UNAM) and the Instituto Nacional de Ciencias Médicas y Nutrición Salvador Zubirán (Reference #1831) and performed in accordance with the revised Declaration of Helsinki, the Declaration of Istanbul and Good Clinical Practice Guidelines.
## Reagents and antibodies
For flow cytometry analysis, APC-Cy7 anti-CD4, PE-Cy7 anti-CD8, APC anti-CD11c and FITC anti-CD14 (San Diego, CA, USA), PE Cy7 anti-LAG-3, PE-Cy5.5 anti-CD3 and PerCP-eFluor 710 anti-ILT4 were purchased from Invitrogen (Waltham, MA, USA). Alexa Fluor 488 anti-PD1, Alexa Fluor 488 anti-CCR5, PerCP Cy5.5 anti-CCR7, PerCP-Cy5.5 anti-TIM3, PE anti-CD25, PE Dazzle 594 anti-TIGIT, PE-Cy7 anti-HLA-G, APC anti-CD49b, APC Fire 750 anti-CD45RA, Brilliant Violet (BV) 711 anti-CD39, BV421 anti-CTLA-4, BV421 anti-CCR4, BV711 anti-CXCR3, Zombie NIR and Zombie Aqua™, were purchased from Biolegend (San Diego, CA, USA) (Supplementary Table 1).
For in vitro experiments, Ficoll® Paque Plus (Ficoll) and dimethyl sulfoxide (DMSO) were obtained from Sigma-Aldrich (San Luis, MO USA). Recombinant human (rh) GM-CSF, IFN-γ, IL-1β, IL-2, IL-4, IL-6, IL-10 and TNF-α cytokines were obtained from PeproTech (New Jersey, NY, USA). Carboxy Fluorescein Succinimidyl Ester (CFSE), CellTrace™ Violet (CTV), Dynabeads Human T-activator CD3/CD28 (anti-CD3/CD28-coated beads), DynaMag-5™ Magnet (DynaMag), CTS™ OpTmizer™ T Cell Expansion SFM medium (Expansion culture medium), RPMI 1640 medium, antibiotic-antimycotic 100X, L-glutamine (GlutaMAX™), sodium pyruvate (100 mM), MEM Non-Essential amino acids 100X and Fetal Bovine Serum (FBS) were purchased from Thermo Fisher Scientific (Waltham, MA, USA). Pooled human AB serum (AB-HS) was obtained from Gemini Bio Products (Sacramento, CA, USA). All culture mediums were supplemented with L-glutamine, sodium pyruvate, MEM-NEAA and antibiotic-antimycotic. T cell cultures were performed in round bottom 96-well culture plates (Corning, Avon, France).
## Peripheral blood mononuclear cells isolation
Peripheral blood mononuclear cells (PBMCs) were isolated from buffy coat preparations of adult normal healthy donors by density-gradient centrifugation over Ficoll according to the manufacturer’s instructions. Isolated cells were resuspended in a solution containing $90\%$ Fetal Bovine Serum (FBS) and $10\%$ Dimethyl Sulfoxide (DMSO), were stored at -70°C for 24 hours and then cryopreserved to -195°C in liquid nitrogen. For functional assays, cryopreserved cells were thawed in a 37°C water bath and washed three times with culture medium supplemented with $10\%$ FBS.
## Monocytes-derived dendritic cells and IL-10-producing dendritic cells in vitro differentiation
CD14+ monocytes cells were isolated from PBMCs using the Human CD14 MicroBeads kit (Miltenyi Biotec, Bergisch Gladbach, Germany) according to the manufacturer’s instructions. CD14+ monocytes were resuspended in RPMI 1640 culture medium supplemented with $10\%$ AB-HS, stimulated with rhIL-4 (50 ng/mL) and rhGM-CSF (50 ng/mL) in the presence (DC10) or absence (moDC) of rhIL-10 (10 ng/mL) and they were cultured (37°C/$5\%$ of CO2) for 8 days; on days 3 and 5, culture medium and cytokines (25 ng/mL rhIL-4, 25 ng/mL rhGM-CSF and 10 ng/mL rhIL-10) were refreshed. The moDC and DC10 phenotypes were evaluated by flow cytometry staining with Zombie Aqua (viability) and anti-CD11c, anti-CD14, anti-HLA-G and anti-ILT4 monoclonal antibodies (mAbs). Samples were acquired on a Attune NxT Cytometer (Thermo Fisher Scientific) and data were analyzed with FlowJo v10.8 software (BD Biosciences, CA, USA).
## In vitro differentiation and purification of allospecific Tr1 cells
CD4+CD25-CD45RA+ naïve T (nT) cells were purified from PBMCs using a FACS Area I sorter (BD Biosciences, CA, USA). Sorted nT cells were labeled with CTV according to the manufacturer’s instructions and co-cultured for 14 days with moDCs (conventional T cell condition) or DC10 (Tr1 cell condition) at a ratio of 10:1 (nT: DC10 or moDC) in presence or absence of rhIL-10 (10 ng/mL), respectively; on 7 day of co-culture, a restimulation with rhIL-10 (10 ng/mL) and rhIL-2 (50 U/mL) was performed. Tr1 cell phenotype (CD4+CD49b+LAG-3+ and IL-10 production) was evaluated on the 14th day of co-culture by flow cytometry. IL-10 production of T cells was determined using IL-10 secretion assay and CytoStim™ kits (Miltenyi Biotec), according to the manufacturer’s instructions, with some modifications.
After 14 days of co-culture, CD4+CD49b+LAG-3+ Allo Tr1 cells from nT:DC10 co-culture and CD4+ allospecific conventional T (T conv) cells from nT:moDC co-culture were purified from proliferating cells (CTV-) using a MoFlo XDP cell sorter (Beckman Coulter, Brea, CA, USA). Sorted Allo Tr1 cells and T conv were cultured for 2 days in Expansion culture medium (50x103/well) supplemented with $10\%$ AB-HS and IL-2 (50 U/mL) previous to the polyclonal expansion protocol.
## Polyclonal expansion and antibody staining of Tr1 and conventional T cells
FACS-sorter Allogenic Tr1 (Allo Tr1) cells (20x103/well) were cultured for 4 days (37°C/$5\%$ of CO2) in Expansion culture medium supplemented with $10\%$ AB-HS and stimulated with anti-CD3/CD28-coated beads at a 1:5 ratio (Beads: Tr1) plus rhIL-10 (10 ng/mL) and rhIL-2 (250 U/mL) (expansion period). After 4 days of stimulation, the anti-CD3/CD28-coated beads were removed using the DynaMag and Allo Tr1 cells (50x103) were rested for 3 days in Expansion culture medium supplemented with $10\%$ AB-HS and rhIL-2 (50 U/mL) (resting period). Three consecutive rounds of stimulation/resting (4 days of expansion and 3 days of resting each) were performed. A schematic representation of protocol is shown in Figure 1.
**Figure 1:** *General methodological scheme used for the expansion of Allo Tr1 regulatory type 1 cells. FACS-sorted CD4+CD25-CD45RA+ nT cells (labeled with CTV) were stimulated with allogeneic DC10 in the presence of exogenous IL-2 and IL-10 for 14 days. Proliferating Allo Tr1 cells (CD49b+LAG-3+CTV-) were purified by FACs from co-cultures (CD4+ nT cells:DC10) and polyclonally-expanded for three cycles of stimulation (4 days of culture with anti-CD3/anti-CD28, rhIL-2 and rhIL-10)/resting (3 days of culture with only rhIL-2) of 7 days each. Exp-allo Tr1 cell phenotype was evaluated at each cycle of stimulation. In vitro functional assays were performed after three rounds of polyclonal expansion. Finally, long-term Exp-allo Tr1 cells were re-stimulated for two additional cycles in the presence of proinflammatory cytokines for Tr1 stability assays. Created by BioRender.com.*
As a control group, allospecific CD4+ T conv cells (20x103/well) were polyclonally-expanded in parallel cultures with anti-CD3/anti-CD28 coated beads at 1:5 ratio (Beads:T conv) and rhIL-2 (100 U/mL) with the same rounds of stimulation/resting used for Allo Tr1 cells.
Throughout the ex vivo expansion protocol, IL-10 production of T cells was evaluated using IL-10 secretion assay and CytoStim™ kits (Miltenyi Biotec); next, the cells were stained with anti-CD4, anti-CD49b, anti-LAG-3, anti-TIM-3, anti-TIGIT, anti-PD1, anti-CTLA-4, anti-CD39 and Zombie Aqua™ for 30 min at 37°C in the dark and washed once with FACS buffer. The chemokine profile was evaluated with anti-CD4, anti-CD49b, anti-LAG-3, anti-CCR2, anti-CCR5, anti-CCR4, anti-CCR7, anti-CXCR3, and Zombie NIR™ and the samples were stained for 20 min at room temperature in the dark and washed once with FACS buffer. Sample acquisitions were performed in an Attune NxT cytometer (Thermo Scientific) and data were analyzed with FlowJo v10.8 software (BD Biosciences, CA, USA).
## High dimensional analysis of flow cytometry data
To capture the non-linear structure of our single cell data, we performed dimensionality reduction using a FlowJo implementation opt-tSNE [22]. To identify clusters within the opt-tSNE map, X-Shift was performed as previously described [23]. The X-Shift identified clusters were then phenotyped using FlowJo v10.8’s plug-in ClusterExplorer and ViolinBox (San Diego, BD).
## In vitro allo-specific suppression assay
Autologous CD3+ responder T cells (R.T.C.) were isolated from PBMCs using the Pan T Cell Isolation Kit (Miltenyi Biotec) and labeled with CFSE, following the manufacturer’s instructions. Expanded allospecific Tr1 (Exp-allo Tr1) cells expanded for three weeks were labeled with CTV, co-cultured with CFSE-labeled R.T.C. at different ratios CD3+ Exp-allo Tr1 (1:1, 1:3, 1:9 and 1:27) and stimulated with specific-allogeneic or third-party moDC in a ratio 1:2 (CD3+:moDC) in expansion culture medium with $10\%$ AB-HS. For some assays, co-cultures were stimulated in the presence or absence of 5 ng/mL of IFN-γ, IL-1β, IL-6 and/or TNF-α. After 4 days of co-culture, cells were recovered and stained with mAbs anti-CD3, anti-CD4 and anti-CD8, cells were acquired in an Attune NxT Cytometer and data were analyzed with FlowJo v10.8. software. The R.T.C. proliferation was determined by CFSE dilution on gated CD4+ or CD8+ T cells, and CTV-labeled Exp-allo Tr1 cells were excluded from the analysis. Results are presented as relative increments (RI) of the suppression obtained in control cultures without Exp-allo Tr1 cells. The relative increment was calculated using the following formula:
## Tr1 cells stability assays
After 3 weeks of expansion, Allo Tr1 cells were stimulated with 2 additional cycles of polyclonal expansion/resting in the presence or absence of rhIL-6 (5 ng/mL), rhTNF-α (5 ng/mL), rhIFN-γ (5 ng/mL) and/or IL-1β (5 ng/mL). Tr1 cell stability was evaluated by analyzing the expression of CD49b, LAG-3, IL-10, TIM-3, TIGIT, PD1, CTLA-4, CD39, as well as their alloantigen-specific suppressive function by flow cytometry.
## Cytokine production assay
For cytokine production analysis, CD4+ T cells (Exp-allo Tr1 and T conv) (2x104 cells/well) were stimulated with anti-CD3/anti-CD28 beads (beads:T cells at a ratio of 1:5) during 4 days. The concentrations of cytokines in the culture supernatants were measured using the LEGENDplex™ kit, Custom Human Panel (Biolegend), according to the manufacturer’s guidelines. The samples were acquired on the flow cytometer CytoFLEX S (Beckman Coulter) and data were analyzed using FlowJo (BD) and GraphPad Prism v8 softwares. Cytokine concentrations were determined using the standard curve generated in the same assay.
## Statistical analysis
Statistical analysis was performed using GraphPad Prism v8 software. The Shapiro–Wilk test was used to evaluate the distribution of the data. Differences between two groups were calculated using paired and unpaired t tests for normally distributed data; Wilcoxon’s rank sum test or Mann-Whitney test were used for non-normally distributed data. Graphs are expressed as Mean ± Standard Error of the Mean (SEM). Values with $p \leq 0.05$ were considered as statistically significant.
## In vitro-generated allogeneic DC10 efficiently induce Allo Tr1 cells
With the overall aim of increasing the yield and purity of the cellular products obtained with current Tr1-based methodologies, we designed a new experimental protocol that allows efficient long-term expansion of highly purified Allo Tr1 cells (Figure 1). Based on previous reports [24], we used in vitro differentiated DC10 as an efficient methodology to the differentiation of Allo Tr1 cells.
We first evaluated the phenotype of DC10, which presented a tolerogenic dendritic cell phenotype characterized by a significant increase of HLA-G+ILT4+ and CD14+ percentages compared to moDC (Figure S1). Next, CD4+CD25-CD45RA+ naïve T cells were sorted by FACS, CTV-labeled and co-cultured with allogeneic DC10 in the presence of rhIL-10 and rhIL-2 for 14 days and then analyzed the expression of specific Tr1 surface markers CD49b and LAG-3. As a negative control for Tr1 differentiation, nT cells were stimulated with moDC and only IL-2. We obtained a higher percentage of CD49b+LAG-3+ Allo Tr1 cells in co-cultures stimulated with allogeneic DC10 compared to allogeneic moDC ($46.5\%$ ± 19.1 vs $14\%$ ± 11) (Figures 2A, B). Moreover, Allo Tr1 cells differentiated with DC10 showed a significant increase in IL-10 production ($28.7\%$ ± 15 vs $6\%$ ± 5) (Figures 2C, D). Using our culture conditions, we were able to differentiate a high percentage of Allo Tr1 cell population with Tr1 cell phenotype.
**Figure 2:** *DC10 induces the differentiation of a high percentage of CD49b+LAG-3+ Allo Tr1 cells. nT (CD4+CD25-CD45RA+) cells were co-cultured with allogeneic moDC (Allo T conv) or DC10 (Allo Tr1 cells) for 14 days. (A) Representative contour-plot of CD49b and LAG-3 co-expression in proliferating cells from T conv (black) or Allo Tr1 cells (red). (B) Proportion of CD49b+LAG-3+ in Allo Tr1 (red, n= 11) condition culture compared to T conv (black, n= 9) condition. (C) Representative contour-plot of IL-10 production from CD49b+LAG-3+ population in T conv (black) and Allo Tr1 cells (red). (D) CD49b+LAG-3+ Allo Tr1 cells (red, n= 11) showed a significantly higher IL-10 production compared with CD49b+LAG-3+ T conv (black, n= 9). All experiments were performed in duplicates. Data are representative of six independent experiments. The results are shown as mean ± SEM. The statistical analysis was performed using the unpaired-t test or the Mann–Whitney U test.*
## Long-term polyclonally-expanded Alo-Tr1 cells maintain an enriched suppressor phenotype and high IL-10 production
To increase the purity and Allo Tr1 cell numbers, we isolated CD49b+LAG-3+ Allo Tr1 cells from co-cultures with DC10 by FACS and then we developed an efficient expansion protocol including cycles of polyclonal activation (4 days) followed by resting (3 days with only IL-2) for 21 days. As previously reported, Tregs cells express surface markers that are also present in highly activated T conv, therefore, we performed parallel cultures after purification of allogeneic T conv cells followed by polyclonal expansion under the similar stimulation conditions as Allo Tr1. As shown the Figure 3, the numbers of purified Allo Tr1 cells were significantly augmented, reaching an average fold increase of 840 times (Figure 3A). Moreover, after evaluating the phenotype of Exp-allo Tr1 cells at each polyclonal expansion cycle (days 7, 14 and 21), we observed a higher proportion of CD49b+LAG-3+ cells, compared with expanded T conv cells, (cycle I: $80.7\%$ ± 12.9 vs 38.1 ± 12.4; cycle II: $76.2\%$ ± 11.16 vs $50.5\%$ 15.3; cycle III: $82.1\%$ ± 8.4 vs $58.22\%$ ± 18.4) (Figures 3B, C). Importantly, Exp-allo Tr1 cells maintained significantly increased levels of IL-10 compared with T conv cells at each cycle of expansion (Figures 3D, E). Interestingly, the Median Fluorescence Intensity (MeFI) of LAG-3 was significantly higher in Exp-allo Tr1 compared to expanded T conv in each expansion cycle, while we did not observe differences in the expression levels of CD49b between both populations throughout the expansion protocol (Figures S2A, B).
**Figure 3:** *Purified Allo Tr1 cells can be long-term expanded, conserving their CD49b+LAG3+ phenotype and IL-10 production. FACS-sorter Allo Tr1 cells or T conv were polyclonally-expanded for 21 days in the presence of IL-10 plus IL-2 or only IL-2, respectively. (A) Left,
A
llo Tr1 cell proliferation reaching an average fold increase of 840 times at day 21 of stimulation (n= 6). Fold expansion was calculated by dividing the number of Tr1 cells obtained on the evaluated day by the number of Tr1 cells on day 1 of FACS-sorting. Right, viability of Exp-allo Tr1. Data are representative of four independent experiments. (B) Representative contour-plot of LAG-3 and CD49b co-expression in Exp-allo Tr1 cells (red) and T conv (black) at each cycle of in vitro expansion. (C) Proportion of CD49b+LAG-3+ in Exp-allo Tr1 cells (red, n= 6-12) compared to T conv (black, n= 6-12). (D) Representative histograms of IL-10 production in Exp-allo Tr1 cells and T conv. (E) Median Fluorescence Intensity (MeFI) of IL-10 in Exp-allo Tr1 cells (red) compared with T conv (black) throughout the expansion. All experiments were performed in duplicates. Data are representative of six independent experiments. The results are shown as mean ± SEM. The statistical analysis were performed using the unpaired-t test or the Mann–Whitney U test.*
To further investigate the phenotypic markers of Exp-allo Tr1 cells, we evaluated the expression of co-inhibitory receptors related to their suppressive function at 7, 14, and 21 days of polyclonal expansion. Importantly, both the percentage (Figures 4A, S3A) and expression levels (MeFi, Figure 4B) of TIM-3, TIGIT and PD-1 were maintained in Exp-allo Tr1 cells throughout the polyclonal expansion. Of note, these co-inhibitory receptors were similarly expressed in expanded T conv, as it would be expected in long-term stimulated T cells (Figures S3A, B). Interestingly, Exp-allo Tr1 cells showed significantly increased levels (MeFI) of CD39 and CTLA-4 (Figure 4B) at 21 days of expansion compared to expanded T conv cells.
**Figure 4:** *Long-term Exp-allo Tr1 cells maintain a high expression of co-inhibitory receptors. Purified Allo Tr1 cells were polyclonally expanded for 21 days and the expression of co-inhibitory receptors were evaluated. (A) Representative contour-plot of expression of co-inhibitory receptors in Exp-allo Tr1 cells (red) and Allo T conv (black). (B) Levels of expression (MeFI) of TIM-3, TIGIT, PD1, CTLA-4 and CD39 in Exp-allo Tr1 (red, n= 4-8) and Allo T conv (black, 4-8) at day 21 of stimulation. All experiments were performed in duplicates. Data are representative of six independent experiments. The results are shown as mean ± SEM. The statistical analysis were performed using an unpaired-t test or the Mann–Whitney U test.*
## High dimensional analysis of Exp-allo Tr1 confirms a distinct regulatory profile that differentiates from expanded T conv cells
Opt-tSNE was performed on live CD4+ cells from individual samples were down-sampled to 12,660 events per sample, individual samples were electronically barcoded, and finally concatenated for downstream analyses. A total of six samples post-expansion (from three individuals) were included in the analysis, three Exp-allo Tr1 cells and the corresponding T conv cells. Opt-tSNE was run using all compensated parameters except the previously gated live CD4 and prior parent populations. Two major islands were identified in the resulting opt-tSNE maps. These were pulled together predominantly according to Exp-allo Tr1 versus T conv expanded populations. Both major islands contained populations from all three patients (Figure 5A), demonstrating a non-patient bias in the dimensionality reduction.
**Figure 5:** *High dimensional analysis of Exp-allo Tr1 shows a differential expression profile compared to T conv cells. Purified Allo Tr1 cells were polyclonally expanded for 21 days, co-inhibitory receptors were evaluated, and high dimensional analysis was used to further identify and explore their expression profile. (A) Dimensionality reduction by Opt-SNE from live CD4+ Exp-allo Tr1 (n= 3) cells and T conv (n= 3) cells showed two major islands characterized by Tr1 (red) versus T conv (black) culturing conditions right panel). (B) X-Shift clustering analysis identified 11 clusters (excluding cluster 8, which represents 0.18% of the Tr1 population) corresponding to one major island composed of Exp-allo Tr1 (2, 3, 4, 6, 7, 9, 11 and 12) and the other with T conv (1, 5 and 10). The resulting heat map of median fluorescence expression from each cluster shows a distinct expression profile between Exp-allo Tr1 and T conv clusters. Clusters 3 and 10 were selected as representative of each culturing conditions clearly showing overexpression IL-10, CD39, LAG-3 and CTLA-4 wherein T conv were relatively low expressors for all these markers compared to Exp-allo Tr1 (C). Relative expression line graph of median fluorescence intensity shows the level expression of each co-inhibitory molecule from the underlying most representative populations: Exp-allo Tr1 (pink) vs T conv (green). Data are representative of two independent experiments.*
To identify cell clustering within our high-dimensional data visualized with opt-tSNE, we ran X-Shift on the live CD4+ population [23]. X-shift clustering identified 11 clusters (Figure S4), three clusters (Clusters 1, 5, 10) were found to have been composed exclusively from T conv cells and composing one of the two major opt-tSNE islands (Figure 5B). The remaining eight clusters (Clusters 2-4, 6-8, 9, 11-12) were found to have been composed exclusively of the Exp-allo Tr1 cells and composing the other major opt-tSNE island (Figure 5B). FlowJo’s ClusterExplorer and ViolinBox were used to phenotype the X-Shift clusters (Figure 5C). Importantly, only the eight clusters identified in the expanded Tr1 cells (Clusters 2-4, 6-8, 9, 11-12) expressed high IL-10, albeit at varying levels of expression. In contrast, T conv cell related clusters expressed lower levels of IL-10. We selected clusters 3 and 10, as the most representative of Exp-allo Tr1 ($46\%$) and T conv ($70\%$) respectively, and compared their expression profiles. As shown in Figures 5B, C, Cluster 3, in addition to IL-10, showed distinctive high levels of CD49b, CD39, CTLA-4 and LAG3 compared to Cluster 10, while both clusters expressed similar levels of TIGIT, TIM-3 and PD-1.
## Exp-allo Tr1 cells express a chemokine receptor profile relevant for solid organ allotransplantation homing
The phenotypical characterization of Tr1 cell populations candidates in cellular therapy, has mainly focused on the expression of molecules involved in their suppressive function, nevertheless, the evaluation of CR expression is key to ensure the migratory potential of infused Tr1 and the induction of effective tolerance towards de graft. To evaluate the homing potential of our Exp-allo Tr1 cells, we evaluated at day 21 of expansion the expression of CCR2, CXCR3, CCR4, CCR5 and CCR7 which are known to direct the migration of T cells to the allograft and draining lymph nodes, respectively (Figure 6). As previously reported for Tr1 cells [17], we showed a high percentage of Exp-allo Tr1 cells expressing CCR5 ($90.09\%$ ± 4.9) and this was significantly increased compared to T conv cells ($47.1\%$ ± $12.8\%$). On the other hand, expression of CXCR3 and CCR4 showed a trend towards a decrease in Exp-allo Tr1 cells compared to T conv cells (CXCR3 $60.3\%$ ± 25.51 vs $86.30\%$ ± 7.7, $$p \leq 0.0940$$; CCR4 $60.3\%$ ± 9.7 vs $93.2\%$ ± 5.2, $$p \leq 0.0759$$). By contrast, we found lower expression of CCR7 in both Exp-allo Tr1 cells ($26.08\%$ ± 19.4) and T conv cells ($18.6\%$ ± 18) compared to other receptors while high proportion of both populations were CCR2+ ($93.3\%$ ± 4 vs $95.8\%$ ± 1.1) (Figure 6B). Interestingly, when we compared the MeFI values of CR expression, we found significantly increased levels of CCR5, but decreased levels of CCR4, in Exp-allo Tr1 cells compared to T conv cells. Finally, CXCR3 was moderately upregulated on T conv compared to Exp-allo Tr1 cells and no significant differences were found in CCR2 and CCR7 MeFI values between both subpopulations (Figure 6C).
**Figure 6:** *Exp-allo Tr1 cells show a high expression of chemokines receptors relevant for homing to allografts. Purified Allo Tr1 cells were polyclonally expanded for 21 days and the expression of CR receptors were evaluated. (A) Representative contour-plot of chemokines receptors expression in Exp-allo Tr1 cells (red) and Allo T conv (black). (B) Percentages or (C) levels of expression (MeFI) of chemokine receptors CCR2, CXCR3, CCR4, CCR5 and CCR7 in Exp-allo Tr1 cells (red, n= 5) and T conv (black, n= 4) at day 21 of expansion. All experiments were performed in duplicates. Data are representative of three independent experiments. The results are shown as mean ± SEM. The statistical analysis was performed using the unpaired t-test or the Mann–Whitney U test.*
## Exp-allo Tr1 cells secrete cytokines with an anti-inflammatory profile
Another essential aspect to consider in the characterization of Tr1 cells is their cytokine production profile. We evaluated the cytokine secretion from Exp-allo Tr1 cell supernatant cultures, using a cytokine-based assay (CBA) (Figure 7). As we expected, the production of the anti-inflammatory cytokines IL-10 and TGF-β were significantly higher in Exp-allo Tr1 cultures compared to expanded T conv cells. By contrast, the production of pro-inflammatory cytokines TNF-α, and IL-1β and IL-4, by Exp-allo Tr1 was decreased when compared to T conv cells. Additionally, the production of IL-17, IL-6, and IL-9 were very low or undetectable in both T cell populations. The secretion of IFN-γ, granzyme B (GzMB) and perforin has been previously reported in Tr1 cells (reviewed in [7]); here, we found a similar production of GzMB and perforin but a trend towards to increase the IFN-γ production in Exp-allo Tr1 compared to T conv cells. Remarkably, the production of IL-2 was lower in T conv in comparison with Exp-allo Tr1. In conclusion, the cytokine production of the Exp-allo Tr1 cells fits with an anti-inflammatory profile.
**Figure 7:** *Prolonged expanded Allo Tr1 has a higher production of anti-inflammatory cytokines, but does not produce inflammatory cytokines. Exp-allo Tr1 cells (red, n= 3) and T conv (black, n= 3) cells expanded for three weeks were restimulated for 4 days, and supernatants were analyzed for cytokine production by cytometric bead array. The detection limits (pg/mL) for each cytokine are shown below each graph. All experiments were performed in duplicates. Data are representative of three independent experiments. The results are shown as mean ± SEM (n= 3). Statistical analysis was performed using an unpaired t-test or the Mann Whitney U test.*
## Exp-allo Tr1 cells suppress responder T cells proliferation efficiently
After phenotypical characterization of Exp-allo Tr1 cells and to ensure the functionality, we next evaluated their ability to suppress allospecifically CD4+ and CD8+ responder T cells (R.T.C.). As shown in Figure 8, we observed a reduction of the R.T.C. proliferation at all ratios evaluated, showing the most significant suppression at the 1:1 ratio for both CD4+ (Figures 8A, B) and CD8+ T cells (Figures 8C, D). Importantly, the suppressive capacity of Exp-allo Tr1 cells was significantly lower when T cells were stimulated with allogeneic moDC from a not related individual (Third-party) compared to the DCs towards which they were initially expanded (Allo) for both CD4+ (Figures 8E, F) and CD8+ (Figures 8G, H) proliferation.
**Figure 8:** *In vitro
Exp-allo Tr1 cells suppress the proliferation of T conv cells in an alloantigen-specific way. Exp-allo Tr1 were co-cultured with conventional CD3+ R.T.C. (labeled with CFSE) at different evaluated ratios and stimulated with allogeneic moDC from their respective donors or from non-related individuals (Third Party); on day 4 of culture, R.T.C. proliferation was evaluated by flow cytometry. (A–D)
Exp-allo Tr1 inhibited the proliferation of both CD4+
(A, B) and CD8+
(C, D) R.T.C. at all evaluated ratios (n= 5-6). (E–H)
Exp-allo Tr1 suppresses the proliferation of CD8+
(E, F) and CD4+
(G, H) R.T.C. cells only when they are stimulated with the allogeneic DCs with which they were initially expanded (Allogenic, red, n= 4-5), but do not suppress when they are stimulated with unrelated DCs (Third Party, blue, n= 4). Representative histograms are shown in (A), (C) and (E). All experiments were performed in duplicates. Data are representative of four independent experiments. The results are shown as mean ± SEM. Statistical analysis was performed using the unpaired-t test or one sample t and Wilcoxon test.*
Under inflammatory conditions, it has been demonstrated that some cytokines (IL-1β, TNF-α, IL-6) may be involved in FOXP3+ Treg suppression downmodulation [18, 25, 26], however this has not yet been extensively studied in Tr1 populations. Thus, we evaluated the suppression of Exp-allo Tr1 cells in the presence of an inflammatory environment. Markedly, Exp-allo Tr1 cells were able to suppress the CD4+ (Figures 9A, B) and CD8+ (Figures 9C, D) R.T.C proliferation even in the presence of exogenous IL-1β, IL-6, IFN-γ and TNF-α in the co-cultures, either alone or all together.
**Figure 9:** *Exp-allo Tr1 cells maintain their suppressive capacity in presence of inflammatory cytokines. Co-culture of Exp-allo Tr1 and conventional CD3+ T cells were stimulated with allogeneic moDC in the presence or absence of inflammatory cytokines and T cell proliferation was evaluated. (A–D)
Exp-allo Tr1 efficiently suppressed the proliferation of both CD8+
(A, B) and CD4+
(C, D) R.T.C. in the presence of inflammatory cytokines IL-1β, IL-6, IFN-γ, TNF-α or all together (n= 5). All experiments were performed in duplicates. Data are representative of four independent experiments. The results are shown as mean ± SEM. Statistical analysis was performed using the unpaired-t test or one sample t and Wilcoxon test.*
## Proinflammatory cytokines do not affect the phenotype and suppressive function in Exp-allo Tr1 cells
It has been widely reported that an inflammatory microenvironment can directly influence the phenotype or functionality of CD4+ T cells. Thus, Exp-allo Tr1 cells were further expanded for two extra weeks in the presence of IL-2 plus rhIL-1β, rhIL-6, rhIFN-γ, rhTNF-α, or all together, or with IL-10 (Exp-allo Tr1 control conditions). After these two extra expansion rounds, we did not observe significant difference in neither CD49b and LAG-3 (Figures 10A, B) nor in TIM-3, TIGIT, PD-1, CD39, CTLA-4 expression and in the IL-10 production under proinflammatory cytokine conditions (Figure 10C).
**Figure 10:** *Exp-allo Tr1 cells maintain their immunosuppressive phenotype after stimulation in an inflammatory microenvironment. Exp-allo Tr1 cells were stimulated with anti-CD3/anti-CD28 in the presence or absence of pro-inflammatory cytokines (IFN-γ, IL-6, IL-1β, TNF-α or all together) for two additional weeks. (A) Representative contour-plot of CD49b and LAG-3 expression in Exp-allo Tr1 activated in presence of inflammatory cytokines. (B) Proportion of CD49b+LAG-3+ expression in Exp-allo Tr1 cells activated in presence or absence of inflammatory cytokines (n= 5). (C) Relative increase (IR) of IL-10, TIM-3, TIGIT, PD-1, CD39 and CTLA-4 (MeFI). RI was calculated by dividing the value of all the conditions by the value of control Exp-allo Tr1 condition. All experiments were performed in duplicates. Data are representative of four independent experiments. The results are shown as mean ± SEM. Statistical analysis was performed using the unpaired-t test or one sample t and Wilcoxon test. No significant differences were observed.*
Once we demonstrated that phenotype and IL-10 production were not affected by pro-inflammatory cytokines, we next evaluated if Exp-allo Tr1 cells were able to maintain the capacity to suppress R.T.C. proliferation. As shown in Figure 11, we observed that Exp-allo Tr1 preserved their suppressive potential towards CD4 and CD8 R.T.C. proliferation, even after being expanded in the presence of all the pro-inflammatory cytokines together. ( Figures 11A, B). These results support our previous observation of the stability of Exp-allo Tr1 suppressive function in presence of proinflammatory cytokines (Figure 8).
**Figure 11:** *Exp-allo Tr1 cells maintain their suppressive function after stimulation in an inflammatory microenvironment. Exp-allo Tr1 activated in presence or absence of pro-inflammatory cytokines (IFN-γ, IL-6, IL-1β, TNF-α or all together) were co-cultured with conventional CD3+ R.T.C. (labeled with CFSE) at 1:1 ratio and stimulated with allogeneic moDC from their respective donors, on day 4 of co-culture. (A–D)
Exp-allo Tr1 efficiently suppressed the proliferation of both CD4+
(A, B) and CD8+
(C, D) R.T.C. after being activated in presence of IL-1β, IL-6, IFN-γ, TNF-α or all together (n= 2). All experiments were performed in duplicates. Data are representative of two independent experiments. The results are shown as mean ± SEM. Statistical analysis was performed using the unpaired-t test or one sample t and Wilcoxon test.*
## Discussion
In recent years new advances in the in vitro generation of Tr1 have demonstrated their potential as therapeutic tools for transplantation tolerance. IL-10 tolerogenic capacity was exploited to establish two experimental protocols; one used allogeneic CD3-depleted PBMC to generate IL-10 anergized T cells, containing Tr1 cells (IL-10-DLI) [27], and the other used tolerogenic DC10, with high production of IL-10 and co-expression HLA-G and ILT-4 [24], giving rise to T10 or T-allo 10 (reviewed in [7]. Both methodologies have been implemented using GMP grade standards showing promising results in clinical trials and a beneficial outcome in HSCT patients for improving immune reconstitution without increasing the risk of GvHD (ALT-TEN [15], NCT01656135 [16], and NCT03198234 [17].
As previously mentioned, we used monocyte-differentiated DC10 in ours protocol and we were able to generate a higher percentage of CD49b+LAG-3+ Allo Tr1 compared to previous works ($46.5\%$ ± 19.1 vs $6\%$ ± 3 [16] and $10.5\%$ [9 to $13\%$ interquartile range (IQR)] [17]. These differences might be because we performed some modifications in the original protocol, including a restimulation with rhIL-10 (10 ng/mL) and rhIL-2 (50 U/mL) in the co-cultures at day 7, compared to the work of Gregory et al, where only add rhIL–2 (20 U/mL) [24]. It has been reported that IL-10 signaling is required to promote Tr1 cell differentiation [28], while IL-2 signaling promotes the proliferation and survival of T cells [29]. Additionally, we used non-irradiated DC10 as stimulators which were able to induce a higher percentage of CD49b+LAG-3+ Allo Tr1 compared to irradiated DC10 ($46.5\%$ vs $12.0\%$, data not shown). These results suggest that in addition to IL-10 production and HLA-G/ILT4 signaling [24], another important stimulus provided by non-irradiated DC10 could be contributing to the Allo Tr1 differentiation.
When we evaluated IL-10 production by a flow cytometry-based IL-10 secretion assay at day 14 of co-culture (DC10:nT), we found an average of $28.7\%$ ± 15 of IL-10+ cells within the CD49b+LAG-3+ population, which identifies functional Tr1 cells. This method ensures a direct detection of IL-10 producing Allo Tr1 cells [9], without interference of other potential contaminating subpopulations in the co-culture, which cannot be avoided in the studies using ELISA-based methods previously reported [16, 17].
One of the main challenges in the field of Tr1 therapy is the isolation of a pure Tr1 cell population that could increase the efficiency of the treatment in vivo and reduce possible long-term side effects induced by contaminating effector cells in the final population. Additionally, cell therapy using FACS-based isolation protocols have shown safety in clinical trials in kidney transplant patients [30, 31]. With this aim in mind, we purified the CD49b+LAG-3+ cells from the initial allogeneic co-cultures by FACS sorting to ensure the obtention of an homogeneous Tr1 cell population. This is highly relevant for cell therapy purposes as it has been recently reported that Tr1 cells are the main functionally active component within T-allo 10 cell product displaying different suppressive mechanisms [17]. Thus, infusion of a highly purified Tr1 subpopulation may allow to reduce the total cell numbers required to maintain long term tolerance.
To date, few clinical trials have been reported using a heterogeneous population of tolerogenic cells that contain different proportions of Tr1 cells. These studies determined the number of cells to be infused per kg of body weight based on the characteristics of the cellular population. It has been suggested that the number of cells required for clinical trials ranges from 3 x 105-6/kg (CD3+ cells) [15] to 11-19 x 106/kg containing sufficient Tr1 cells for inducing efficient allograft tolerance. However, in these studies the actual numbers of Tr1 cells was not directly evaluated, but rather, the numbers of infused cells were adjusted based on their in vitro tolerogenic potential [16]. Considering the relatively low numbers of purified Tr1 cells obtained from our allogeneic co-cultures, we developed a polyclonal expansion protocol that allowed an increase of 840 times the initial numbers at day 21. In addition, our protocol allows us to determine the numbers of Tr1 cells transferred and may reduce the numbers Tr1 needed to achieve tolerance. To our knowledge this is the first protocol that enables allospecific Tr1 cell expansion. A different approach for total polyclonal Tr1 ex vivo expansion has recently reported a 70-fold increase after 18 days of expansion with anti-CD3 and anti-CD28 beads [32], although the authors reported less than $50\%$ of CD49b+LAG-3+ cells within the expanded population, while our expansion protocol maintained more than $80\%$ of Allo Tr1 for over three weeks of expansion.
Currently, co-expression of CD49b and LAG-3 molecules together with IL-10 production, are the current markers used to identify Tr1 cells. Interestingly, Exp-allo Tr1 cells showed higher expression levels of LAG-3 compared to T conv cells and furthermore, IL-10 production was maintained through the expansion cycles (Figure 2), demonstrating that LAG-3 and IL-10 are closely related to Exp-allo Tr1. In addition, the expression of other co-inhibitory molecules have also been related in different scenarios to the Tr1 cell identity as well as their capacity to suppress effector T cells. In this context, murine CD49b+LAG-3+IL-10+ Tr1 cells expressing TIM-3, TIGIT and PD-1 have been related with a higher suppressive function compared with “Tr1-like” subpopulations, showing low or null expression of these molecules [9]. Another report demonstrated that the ectoenzyme CD39 is important for Tr1 suppressive functions through the generation of adenosine [33]. While human Tr1 cells have shown a higher expression of CTLA-4 and PD-1 compared to non-Tr1 cells, moreover, both molecules were key for the suppressive function against effector cells [17]. The constant expression of TIM-3, TIGIT, PD1, CD39, and CTLA-4 during 21 days of expansion on Exp-allo Tr1 was in agreement with a high suppressive function. Notably, the higher expression of CTLA-4, CD39, LAG-3 and IL-10 in our Exp-allo Tr1 at day 21, compared to T conv, might be directly related to their suppressive potential.
In addition to expressing co-inhibitory receptors, CIR+ Tr1 were also reported to express CCR5, which is also displayed by our Exp-allo Tr1. The functional relevance of CCR5 and CXCR3 expression on T conv cells has been described to be pathogenic in mice and human solid organ transplantation [34]. In accordance, combined CXCR3 and CCR5 blockade were effective for prolonging allograft in a model of allogeneic heart transplantation [35], thus the expressión of these receptors in Tregs could suggest a tolerance induction as a result of direct suppression on T conv. However, detailed expression of chemokine receptors on in vitro differentiated Tr1 cells had not been yet evaluated. To aim for the potential clinical use of Exp-allo Tr1 cells to generate allograft tolerance, the evaluation of chemokine receptors with homing potential toward the allograft needs to be considered. In the context of kidney transplantation, expression of CXCR3, CCR2 and CCR4 on Foxp3+ Tregs was found to be key in preventing allograft rejection (reviewed in [20]). Therefore, expression of CXCR3, CCR2 and CCR4 in our Exp-allo Tr1 cells (Figure 6) strongly suggest their potential homing towards the kidney allograft.
As several co-inhibitor receptors are expressed by other T cell subpopulations, including long-term stimulated T cells, it was important a deeper phenotypic analysis that allowed us to distinguish our Exp-allo Tr1 cells from activated T conv cells, that may also display markers typically expressed by exhausted T cells [36, 37]. In this context, high dimensional analysis of multiparametric flow cytometry data has been employed for the characterization of tumor-infiltrating lymphoid populations [38] as well as in phenotypic analysis of Treg cell subpopulations [39]. Importantly, our opt-tSNE analysis demonstrated that Exp-allo Tr1 cells are clearly distinguishable from activated T conv cells, as they each group into distinct islands (Figure 5A) and with each island being differentiated by the expression of IL-10 (Figures 5B, C).
An important mechanism of Treg-mediated suppression is closely related to their cytokine secretion profile. As expected, and as previously reported in Tr1 cells, IL-10, TGF-β, IFN-γ, GzMB and perforin were significantly produced by our Exp-allo Tr1 confirming their identity (Figure 7). Besides, our Exp-allo Tr1 cells also produced low levels of IL-2 and IL-4 [8, 40, 41]. On the other hand, parallel cultures using expanded T conv cells showed moderate production of IL-2 and IFN-γ which is is in agreement with the long-term expansion conditions used and correlate with the expression of several markers typically expressed by exhausted T cells (TIGIT, TIM-3 and PD-1) [29, 37]. The secretion of IL-10 and TGF-β was higher in our Exp-allo Tr1 cells compared to T conv cells, according to their suppressive phenotype. In contrast, the pro-inflammatory cytokines TNF-α and the Th2 cytokine IL-4 were singularly produced by expanded T conv cells compared with Exp-allo Tr1 cells confirming their differential inflammatory versus regulatory phenotype. Of note, Exp-allo Tr1 cells did not produce detectable levels of IL-17 as it has been reported for effector T cells [42]. In addition, GzMB and perforin, that are important mechanisms for Tr1-mediated suppressive function, were significantly expressed to comparable levels with activated T conv cells [9, 43, 44], indicating their cytotoxic potential.
Most impressive are the findings of the suppressive function of Exp-allo Tr1 cells against allospecific CD4+ and CD8+ T cells, in accordance with the phenotypic characteristics of these cells. In contrast with other reports where allogeneic suppression was assessed at a ratio of 1:1, using an heterogeneous population of Tr1-containing cells [16, 17], we explored the Exp-allo Tr1 suppressive potential by comparing different Tr1:R.T.C. ratios, showing that Exp-allo Tr1 are able to suppress both CD4+ and CD8+ allogeneic T cells from 1:27 ratio, showing the highest potential at 1:1 ratio. This allows us to identify more precisely the suppressive function of a specific number of Tr1, and therefore extrapolate the effective Tr1 dose needed for in vivo tolerance. Remarkably, after polyclonal expansion our Exp-allo Tr1 remains allospecific, as proliferation towards third-party is significantly lower than to allo DCs.
One of the major concerns in Treg therapy is to ensure long term stability of the infused cells [45]. Several studies have reported that several cytokines, including IL-6, IL-23 TNF-α, IL-17 and IL-4, can promote a downregulation of FOXP3, leading to the loss of their suppressive capacity and acquisition of an inflammatory phenotype [18, 19]. Importantly, our group has recently reported that purified allospecific FOXP3+ Tregs and de novo generated CD4+CD25+FOXP3+ allospecific iTregs cells can be efficiently expanded in vitro maintaining their phenotype and antigen-specific suppressive function, even under a proinflammatory environment (IFN-γ, IL-4, TNF-α and IL-6) [46, 47]. However, it was not clear whether in vitro differentiated Allo Tr1 cells were able to maintain their phenotype and function under proinflammatory conditions. Moreover, an extensive profile of pro-inflammatory cytokines has been detected in the serum of transplanted patients and in animal models [48, 49]. Exceptionally, our findings indicate high stability of Exp-allo Tr1 both in their phenotypic markers (CD49b, LAG-3, IL-10, TIM-3, TIGIT, PD1, CD39, and, CTLA-4) as well as in the maintenance of the allogeneic suppressive function, after two expansion cycles in presence of proinflammatory cytokines IL-1β, IL-6, IFN-γ, and TNF-α. This is relevant since pro-inflammatory cytokines have been detected in serum of transplanted patients and animal models [48, 49].
On the other hand, the analysis of Tr1 in vivo stability in a mouse model of allogeneic pancreatic islet transplantation demonstrated that transferred congenic Tr1 cells expand and can be traced more than 3 months after transplantation and are sufficient to induce long-term tolerance [50]. Currently, it is hard to trace human transferred Tr1 cells, however, some reports suggest that in vivo Tr1 cells may be stable and promote the generation of new Tr1 cells (infectious tolerance) [51]. Following this concept, we wondered if long-term in vitro expansion of Exp-allo Tr1 in a pro-inflammatory environment could be negatively affected.
Among the issues that remain to be evaluated are the molecular mechanisms involved in Exp-allo Tr1 stability, including analysis of single cell transcriptomics which would allow us to identify the key molecules (transcription factors, epigenetic mechanisms) that are key for the establishment and maintenance of the Tr1 lineage. In this context, different transcription factors have been reported to regulate Tr1-like phenotype and function (c-Maf, BATF, Blimp-1, Eomes, AhR, Erg2, IRF1) [8, 52], however more extensive studies are needed in human cells. We believe that our Exp-allo Tr1 provides a good model to explore these mechanisms, as it achieves the obtention of large numbers of purified antigen-specific Tr1 cells, which could also be used for genetic manipulations, such as those proposed for improving stability of Foxp3+ Tregs [53, 54] aiming to ensure long term stability in vivo in cell therapy protocols used both in transplantation as other immune-associated diseases.
One interesting issue that has not yet been addressed is the potential relationship between Foxp3+ Tregs and Tr1. The relatively high abundance and co-localization of these Treg subpopulations in some tissues, such as the small and large intestine [55], as well as reports showing the induction of Tr1 cells after Tregs transfer in a mice model of pancreatic islet transplantation [56] suggest a potential interplay between these two subpopulations that could be considered in future human clinical trials.
In summary, our data demonstrate for the first time the feasibility of expanding a purified allospecific Tr1 population efficiently and at a large-scale, maintaining their characteristic phenotype (CD49b+LAG-3+) and IL-10 production. Moreover, Exp-allo Tr1 cells express an enriched suppressive phenotype TIM-3, TIGIT, PD-1, CD39 and CTLA-4, and produce significant levels of TGF-β, IFN-γ, GzMB and perforin which correlate with their ability of suppress CD4+ and CD8+ proliferation in an alloantigen-specific manner. Most importantly, Exp-allo Tr1 are functionally stable even in the presence of inflammatory cytokines including IL-1β, IL-6, IFN-γ, TNF-α. We propose that this highly purified Tr1 population could improve efficiency in vivo of Tr1-based immunotherapy and reduce the risk of potential side effects produced by heterogeneous populations in long term organ transplanted patients.
## Data availability statement
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.
## Ethics statement
The studies involving human participants were reviewed and approved by The Comité de Ëtica en Investigación" from the National Institute for Medical Sciences and Nutrition Salvador Zubirán, Reference 3886. Written informed consent for participation was not required for this study in accordance with the national legislation and the institutional requirements.
## Author contributions
GS and SA-C contributed to conception and design of the study. GS, SA-C, AC-H and KR-C contributed to the generation of data. GS, SA-C, AC-H, CA-S and EA-S contributed to the analysis and interpretation of data. GS, SA-C, AC-H, EA-S, CA-S and JA-G, wrote sections of the manuscript. GS, SA-C, AC-H, EA-S, KR-C, JA-G, CA-S and LM-B contributed to manuscript revision, read, and approved the submitted version.
## Conflict of interest
CA-S was employed by Accellix.
The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be constructed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu.2023.1062456/full#supplementary-material
## References
1. Roncarolo MG, Yssel H, Touraine JL, Betuel H, De Vries JE, Spits H. **Autoreactive T cell clones specific for class I and class ii hla antigens isolated from a human chimera**. *J Exp Med* (1988) **167**. DOI: 10.1084/jem.167.5.1523
2. Bacchetta R, Bigler M, Touraine JL, Parkman R, Tovo PA, Abrams J. **High levels of interleukin 10 production in vivo are associated with tolerance in scid patients transplanted with hla mismatched hematopoietic stem cells**. *J Exp Med* (1994) **179** 493-502. DOI: 10.1084/jem.179.2.493
3. Huber S, Gagliani N, Esplugues E, O'Connor W, Huber FJ, Chaudhry A. **Th17 cells express interleukin-10 receptor and are controlled by Foxp3**. *Immunity* (2011) **34**. DOI: 10.1016/j.immuni.2011.01.020
4. Bertolini TB, Biswas M, Terhorst C, Daniell H, Herzog RW, Piñeros AR. **Role of orally induced regulatory T cells in immunotherapy and tolerance**. *Cell Immunol* (2021) **359**. DOI: 10.1016/j.cellimm.2020.104251
5. Gagliani N, Jofra T, Valle A, Stabilini A, Morsiani C, Gregori S. **Transplant tolerance to pancreatic islets is initiated in the graft and sustained in the spleen**. *Am J Transplant* (2013) **13**. DOI: 10.1111/ajt.12333
6. Gagliani N, Magnani CF, Huber S, Gianolini ME, Pala M, Licona-Limon P. **Coexpression of Cd49b and lag-3 identifies human and mouse T regulatory type 1 cells**. *Nat Med* (2013) **19**. DOI: 10.1038/nm.3179
7. Gregori S, Roncarolo MG. **Engineered T regulatory type 1 cells for clinical application**. *Front Immunol* (2018) **9**. DOI: 10.3389/fimmu.2018.00233
8. Song Y, Wang N, Chen L, Fang L. **Tr1 cells as a key regulator for maintaining immune homeostasis in transplantation**. *Front Immunol* (2021) **12**. DOI: 10.3389/fimmu.2021.671579
9. Brockmann L, Soukou S, Steglich B, Czarnewski P, Zhao L, Wende S. **Molecular and functional heterogeneity of il-10-Producing Cd4+ T cells**. *Nat Commun* (2018) **9** 5457. DOI: 10.1038/s41467-018-07581-4
10. Yu H, Gagliani N, Ishigame H, Huber S, Zhu S, Esplugues E. **Intestinal type 1 regulatory T cells migrate to periphery to suppress diabetogenic T cells and prevent diabetes development**. *Proc Natl Acad Sci* (2017) **114**. DOI: 10.1073/pnas.1705599114
11. Barrat FJ, Cua DJ, Boonstra A, Richards DF, Crain C, Savelkoul HF. *J Exp Med* (2002) **195**. DOI: 10.1084/jem.20011629
12. Battaglia M, Stabilini A, Draghici E, Migliavacca B, Gregori S, Bonifacio E. **Induction of tolerance in type 1 diabetes**. *Diabetes* (2006) **55**. DOI: 10.2337/db05-1576
13. Clemente-Casares X, Blanco J, Ambalavanan P, Yamanouchi J, Singha S, Fandos C. **Expanding antigen-specific regulatory networks to treat autoimmunity**. *Nature* (2016) **530**. DOI: 10.1038/nature16962
14. Roncarolo MG, Gregori S, Bacchetta R, Battaglia M, Gagliani N. **The biology of T regulatory type 1 cells and their therapeutic application in immune-mediated diseases**. *Immunity* (2018) **49**. DOI: 10.1016/j.immuni.2018.12.001
15. Bacchetta R, Lucarelli B, Sartirana C, Gregori S, Lupo Stanghellini MT, Miqueu P. **Immunological outcome in haploidentical-hsc transplanted patients treated with il-10-Anergized donor T cells**. *Front Immunol* (2014) **5**. DOI: 10.3389/fimmu.2014.00016
16. Mfarrej B, Tresoldi E, Stabilini A, Paganelli A, Caldara R, Secchi A. **Generation of donor-specific Tr1 cells to be used after kidney transplantation and definition of the timing of their in vivo infusion in the presence of immunosuppression**. *J Trans Med* (2017) **15** 40. DOI: 10.1186/s12967-017-1133-8
17. Chen PP, Cepika AM, Agarwal-Hashmi R, Saini G, Uyeda MJ, Louis DM. **Alloantigen-specific type 1 regulatory T cells suppress through ctla-4 and pd-1 pathways and persist long-term in patients**. *Sci Trans Med* (2021) **13**. DOI: 10.1126/scitranslmed.abf5264
18. Guo H, Xun L, Zhang R, Hu F, Luan J, Lao K. **Stability and inhibitory function of treg cells under inflammatory conditions in vitro**. *Exp Ther Med* (2019) **18**. DOI: 10.3892/etm.2019.7873
19. Sawant DV, Vignali DA. **Once a treg, always a treg**. *Immunol Rev* (2014) **259**. DOI: 10.1111/imr.12173
20. Lamarche C, Levings MK. **Guiding regulatory T cells to the allograft**. *Curr Opin Organ Transplant* (2018) **23**. DOI: 10.1097/mot.0000000000000483
21. Alvarez CM, Opelz G, Garcia LF, Süsal C. **Expression of regulatory T–Cell-Related molecule genes and clinical outcome in kidney transplant recipients**. *Transplantation* (2009) **87**. DOI: 10.1097/TP.0b013e318199fa57
22. Belkina AC, Ciccolella CO, Anno R, Halpert R, Spidlen J, Snyder-Cappione JE. **Automated optimized parameters for T-distributed stochastic neighbor embedding improve visualization and analysis of Large datasets**. *Nat Commun* (2019) **10** 5415. DOI: 10.1038/s41467-019-13055-y
23. Samusik N, Good Z, Spitzer MH, Davis KL, Nolan GP. **Automated mapping of phenotype space with single-cell data**. *Nat Methods* (2016) **13**. DOI: 10.1038/nmeth.3863
24. Gregori S, Tomasoni D, Pacciani V, Scirpoli M, Battaglia M, Magnani CF. **Differentiation of type 1 T regulatory cells (Tr1) by tolerogenic dc-10 requires the Il-10–dependent Ilt4/Hla-G pathway**. *Blood* (2010) **116**. DOI: 10.1182/blood-2009-07-234872
25. Deknuydt F, Bioley G, Valmori D, Ayyoub M. **Il-1β and il-2 convert human treg into Th17 cells**. *Clin Immunol* (2009) **131** 298-307. DOI: 10.1016/j.clim.2008.12.008
26. Valencia X, Stephens G, Goldbach-Mansky R, Wilson M, Shevach EM, Lipsky PE. **Tnf downmodulates the function of human Cd4+Cd25hi T-regulatory cells**. *Blood* (2006) **108**. DOI: 10.1182/blood-2005-11-4567
27. Bacchetta R, Gregori S, Serafini G, Sartirana C, Schulz U, Zino E. **Molecular and functional characterization of allogantigen-specific anergic T cells suitable for cell therapy**. *Haematologica* (2010) **95**. DOI: 10.3324/haematol.2010.025825
28. Brockmann L, Gagliani N, Steglich B, Giannou AD, Kempski J, Pelczar P. **Il-10 receptor signaling is essential for Tr1 cell function in vivo**. *J Immunol (Baltimore Md 1950)* (2017) **198**. DOI: 10.4049/jimmunol.1601045
29. Spolski R, Li P, Leonard WJ. **Biology and regulation of il-2: From molecular mechanisms to human therapy**. *Nat Rev Immunol* (2018) **18**. DOI: 10.1038/s41577-018-0046-y
30. Chandran S, Tang Q, Sarwal M, Laszik ZG, Putnam AL, Lee K. **Polyclonal regulatory T cell therapy for control of inflammation in kidney transplants**. *Am J Transplant* (2017) **17**. DOI: 10.1111/ajt.14415
31. Trzonkowski P, Bieniaszewska M, Juścińska J, Dobyszuk A, Krzystyniak A, Marek N. **First-in-Man clinical results of the treatment of patients with graft versus host disease with human ex vivo expanded Cd4+Cd25+Cd127- T regulatory cells**. *Clin Immunol (Orlando Fla)* (2009) **133**. DOI: 10.1016/j.clim.2009.06.001
32. Cook L, Stahl M, Han X, Nazli A, MacDonald KN, Wong MQ. **Suppressive and gut-reparative functions of human type 1 T regulatory cells**. *Gastroenterology* (2019) **157**. DOI: 10.1053/j.gastro.2019.09.002
33. Mascanfroni ID, Takenaka MC, Yeste A, Patel B, Wu Y, Kenison JE. **Metabolic control of type 1 regulatory T cell differentiation by ahr and Hif1-A**. *Nat Med* (2015) **21**. DOI: 10.1038/nm.3868
34. Melter M, Exeni A, Briscoe DM. **Chemokines and their receptors in human clinical solid organ transplantation**. *Curr Opin Organ Transplant* (2002) **7** 77-84. DOI: 10.1097/00075200-200203000-00016
35. Schnickel GT, Hsieh GR, Garcia C, Shefizadeh A, Fishbein MC, Ardehali A. **Role of Cxcr3 and Ccr5 in allograft rejection**. *Transplant Proc* (2006) **38**. DOI: 10.1016/j.transproceed.2006.10.164
36. Yi JS, Cox MA, Zajac AJ. **T-Cell exhaustion: Characteristics, causes and conversion**. *Immunology* (2010) **129**. DOI: 10.1111/j.1365-2567.2010.03255.x
37. Wherry EJ, Kurachi M. **Molecular and cellular insights into T cell exhaustion**. *Nat Rev Immunol* (2015) **15**. DOI: 10.1038/nri3862
38. Faccani C, Rotta G, Clemente F, Fedeli M, Abbati D, Manfredi F. **Workflow for high-dimensional flow cytometry analysis of T cells from tumor metastases**. *Life Sci Alliance* (2022) **5**. DOI: 10.26508/lsa.202101316
39. Mason GM, Lowe K, Melchiotti R, Ellis R, de Rinaldis E, Peakman M. **Phenotypic complexity of the human regulatory T cell compartment revealed by mass cytometry**. *J Immunol* (2015) **195**. DOI: 10.4049/jimmunol.1500703
40. Groux H, O'Garra A, Bigler M, Rouleau M, Antonenko S, de Vries JE. **A Cd4+ T-cell subset inhibits antigen-specific T-cell responses and prevents colitis**. *Nature* (1997) **389**. DOI: 10.1038/39614
41. Fillatreau S, O'Garra A. **Interleukin-10 in health and disease**. *Curr Top Microbiol Immunol* (2014)
42. Kimura A, Naka T, Kishimoto T. **Il-6-Dependent and -independent pathways in the development of interleukin 17-producing T helper cells**. *Proc Natl Acad Sci* (2007) **104**. DOI: 10.1073/pnas.0705268104
43. Gruarin P, Maglie S, De Simone M, Häringer B, Vasco C, Ranzani V. **Eomesodermin controls a unique differentiation program in human il-10 and ifn-Γ coproducing regulatory T cells**. *Eur J Immunol* (2019) **49** 96-111. DOI: 10.1002/eji.201847722
44. Guipouy D, Gertner-Dardenne J, Pfajfer L, German Y, Belmonte N, Dupré L. **Granulysin- and granzyme-dependent elimination of myeloid cells by therapeutic ova-specific type 1 regulatory T cells**. *Int Immunol* (2019) **31**. DOI: 10.1093/intimm/dxy083
45. Safinia N, Grageda N, Scottà C, Thirkell S, Fry LJ, Vaikunthanathan T. **Cell therapy in organ transplantation: Our experience on the clinical translation of regulatory T cells**. *Front Immunol* (2018) **9**. DOI: 10.3389/fimmu.2018.00354
46. Alvarez-Salazar EK, Cortés-Hernández A, Arteaga-Cruz S, Alberú-Gómez J, Soldevila G. **Large-Scale generation of human allospecific induced tregs with functional stability for use in immunotherapy in transplantation**. *Front Immunol* (2020) **11**. DOI: 10.3389/fimmu.2020.00375
47. Cortés-Hernández A, Alvarez-Salazar EK, Arteaga-Cruz S, Rosas-Cortina K, Linares N, Alberú Gómez JM. **Highly purified alloantigen-specific tregs from healthy and chronic kidney disease patients can be long-term expanded, maintaining a suppressive phenotype and function in the presence of inflammatory cytokines**. *Front Immunol* (2021) **12**. DOI: 10.3389/fimmu.2021.686530
48. Mota AP, Menezes CA, Alpoim PN, Cardoso CN, Martins SR, Alves LV. **Regulatory and pro-inflammatory cytokines in Brazilian living-related renal transplant recipients according to creatinine plasma levels**. *Nephrol (Carlton Vic)* (2018) **23**. DOI: 10.1111/nep.13114
49. Lattmann T, Hein M, Horber S, Ortmann J, Teixeira MM, Souza DG. **Activation of pro-inflammatory and anti-inflammatory cytokines in host organs during chronic allograft rejection: Role of endothelin receptor signaling**. *Am J Transplant* (2005) **5**. DOI: 10.1111/j.1600-6143.2005.00807.x
50. Gagliani N, Jofra T, Stabilini A, Valle A, Atkinson M, Roncarolo MG. **Antigen-specific dependence of Tr1-cell therapy in preclinical models of islet transplant**. *Diabetes* (2010) **59**. DOI: 10.2337/db09-1168
51. Jofra T, Di Fonte R, Galvani G, Kuka M, Iannacone M, Battaglia M. **Tr1 cell immunotherapy promotes transplant tolerance**. *Eur J Immunol* (2018) **48**. DOI: 10.1002/eji.201747316
52. Solé P, Santamaria P. **Re-programming autoreactive T cells into T-regulatory type 1 cells for the treatment of autoimmunity**. *Front Immunol* (2021) **12**. DOI: 10.3389/fimmu.2021.684240
53. Sato Y, Passerini L, Piening BD, Uyeda MJ, Goodwin M, Gregori S. **Human-engineered treg-like cells suppress Foxp3-deficient T cells but preserve adaptive immune responses in vivo**. *Clin Trans Immunol* (2020) **9**. DOI: 10.1002/cti2.1214
54. Goodwin M, Lee E, Lakshmanan U, Shipp S, Froessl L, Barzaghi F. **Crispr-based gene editing enables Foxp3 gene repair in ipex patient cells**. *Sci Adv* (2020) **6**. DOI: 10.1126/sciadv.aaz0571
55. Maynard CL, Harrington LE, Janowski KM, Oliver JR, Zindl CL, Rudensky AY. **Regulatory T cells expressing interleukin 10 develop from Foxp3+ and Foxp3– precursor cells in the absence of interleukin 10**. *Nat Immunol* (2007) **8**. DOI: 10.1038/ni1504
56. Gagliani N, Gregori S, Jofra T, Valle A, Stabilini A, Rothstein DM. **Rapamycin combined with anti-Cd45rb mab and il-10 or with G-csf induces tolerance in a stringent mouse model of islet transplantation**. *PloS One* (2011) **6**. DOI: 10.1371/journal.pone.0028434
|
---
title: Visualizing the distribution of flavonoids in litchi (Litchi chinenis) seeds
through matrix-assisted laser desorption/ionization mass spectrometry imaging
authors:
- Yukun Liu
- Xiaofei Nie
- Jilong Wang
- Zhenqi Zhao
- Zhimei Wang
- Fang Ju
journal: Frontiers in Plant Science
year: 2023
pmcid: PMC9998689
doi: 10.3389/fpls.2023.1144449
license: CC BY 4.0
---
# Visualizing the distribution of flavonoids in litchi (Litchi chinenis) seeds through matrix-assisted laser desorption/ionization mass spectrometry imaging
## Abstract
Flavonoids are one of the most important bioactive components in litchi (*Litchi chinensis* Sonn.) seeds and have broad-spectrum antiviral and antitumor activities. Litchi seeds have been shown to inhibit the proliferation of cancer cells and induce apoptosis, particularly effective against breast and liver cancers. Elucidating the distribution of flavonoids is important for understanding their physiological and biochemical functions and facilitating their efficient extraction and utilization. However, the spatial distribution patterns and expression states of flavonoids in litchi seeds remain unclear. Herein, matrix-assisted laser desorption/ionization mass spectrometry imaging (MALDI-MSI) was used for in situ detection and imaging of the distribution of flavonoids in litchi seed tissue sections for the first time. Fifteen flavonoid ion signals, including liquiritigenin, apigenin, naringenin, luteolin, dihydrokaempferol, daidzein, quercetin, taxifolin, kaempferol, isorhamnetin, myricetin, catechin, quercetin 3-β-d-glucoside, baicalin, and rutin, were successfully detected and imaged in situ through MALDI-MSI in the positive ion mode using 2-mercaptobenzothiazole as a matrix. The results clearly showed the heterogeneous distribution of flavonoids, indicating the potential of litchi seeds for flavonoid compound extraction. MALDI-MS-based multi-imaging enhanced the visualization of spatial distribution and expression states of flavonoids. Thus, apart from improving our understanding of the spatial distribution of flavonoids in litchi seeds, our findings also facilitate the development of MALDI-MSI-based metabolomics as a novel effective molecular imaging tool for evaluating the spatial distribution of endogenous compounds.
## Introduction
Litchi (*Litchi chinensis* Sonn.; order Sapindales, family Sapindaceae), also known as Lizhi, Danli, and Liguo, is a subtropical fruit tree with a cultivation history in China of more than 2,300 years (Hu et al., 2021). It is the only species of Litchi (Yao et al., 2021). Litchi is an important fruit crop in southern China and is planted on more than 550,000 ha with an annual output of more than 2.2 million tons (Hu et al., 2021). The cultivation area and output of litchi in China account for more than $60\%$ of global production (Li et al., 2020). Litchi seeds are a major product, but only a small portion is processed for biological utilization, and many litchi seeds are discarded as waste. The abandonment of fruit seed residues is not only a considerable problem for the environment but also a waste of global resources. Litchi seeds are rich in various bioactive compounds, such as flavonoids, saponins, volatile oils, polyols, alkaloids, steroids, coumarins, fatty acids, amino acids, and sugars (Dong et al., 2019; Punia and Kumar, 2021), resulting in a variety of biological functions, including antiviral and anti-oxidation activities, reducing the degree of liver damage and lowering blood glucose levels (Choi et al., 2017; Dong et al., 2019; Punia and Kumar, 2021). Accumulating evidence has confirmed the antitumor/anticancer effects of litchi seed extracts (Emanuele et al., 2017; Tang et al., 2018; Zhao et al., 2020).
Flavonoids are polyphenolic compounds and endogenous bioactive components, which act as secondary metabolites with extensive pharmacological activities. Flavonoids exert important pharmacological properties, including cardioprotective, anticancer, anti-inflammatory, and anti-allergic activities (Maleki et al., 2019; Ciumărnean et al., 2020; Liskova et al., 2021; Rakha et al., 2022). Regarding anticancer activity, many preclinical studies indicated the antiproliferative effects of flavonoids on lung (Berk et al., 2022), prostate (Vue et al., 2016), colorectal (Park et al., 2012; Li et al., 2018b), and breast (Pan et al., 2012) cancers. Furthermore, flavonoids have anticancer effects on breast tumors through multiple mechanisms (Martinez-Perez et al., 2014; Magne Nde et al., 2015; Zhang et al., 2018; Sudhakaran et al., 2019). Flavonoids can inhibit procarcinogen bioactivation and estrogen-producing and estrogen-metabolizing enzymes (Surichan et al., 2012; Miron et al., 2017), as well as breast cancer resistance protein (BCRP) (Fan et al., 2019). Administering flavonoids could inhibit inflammation, proliferation, tumor growth, and metastasis (Peluso et al., 2013; Khan et al., 2021; Guo et al., 2022). Although many studies have shown the pharmacological effects of flavonoids widely distributed in litchi seeds, almost all such studies were based on the extraction, enrichment, and separation of bioactive components, and few have focused on the spatial distribution and expression states of flavonoids. In fact, the precise reveal of the distribution of these flavonoids in litchi seeds is important for understanding the physiological and biochemical functions of these compounds and facilitating their extraction and utilization.
Matrix-assisted laser desorption/ionization mass spectrometry imaging (MALDI-MSI) has emerged as a molecular-imaging tool for simultaneously detecting and characterizing the spatial distribution and relative abundance of endogenous and exogenous compounds, such as lipids, proteins, metabolites, peptides, and drugs (Van De Plas et al., 2015; Qin et al., 2018; Piehowski et al., 2020). Although MALDI‐MSI has been used in plant science with endogenous molecular profiling to determine the spatial distribution of small molecules in plant tissues (Zaima et al., 2010; Taira et al., 2015; Huang et al., 2016; Li et al., 2018a), to the best of our knowledge, no previous study has utilized MALDI-MSI to characterize the spatial distribution of flavonoids in litchi seeds.
This study is the first to use MALDI-MSI for the in situ detection and imaging of flavonoids in litchi seed tissues. The results clearly showed the heterogeneous distribution of flavonoids in litchi seeds, indicating the potential of litchi seeds as a source for flavonoid extraction. MALDI-MS-based multi-imaging enhanced the visualization of spatial distribution and expression states of flavonoids. Our findings provide insights into the spatial distribution of flavonoids in litchi seeds and support the development of MALDI-MSI-based metabolomics as an appealing and credible molecular imaging technique for evaluating the spatial distribution of endogenous compounds.
## Materials and reagents
Fresh litchi fruit was collected from the Yongfuda litchi orchard (Haikou, Hainan, China) in June 2022. Haikou is located on Hainan Island in China. It has a typical tropical marine climate and annual sunshine duration of over 2,000 h. The climate is humid, the temperature rises fast, and the average annual precipitation is approximately 260 mm. The Yongfuda litchi orchard is located in a volcanic rock soil planting area. Once harvested, the peel and flesh of the litchi were immediately removed, and the litchi seeds were flash-frozen with liquid nitrogen by slow immersion to prevent seed shattering and endogenous compound changes. The commonly used MALDI matrix, 2-mercaptobenzothiazole (2-MBT), was obtained from Sigma-Aldrich (St. Louis, MO, USA). Amino acid and oligopeptide standards, including His, Gly-Gly-Leu (tripeptide), Ala-His-Lys (tripeptide), Leu-Leu-Tyr (tripeptide), and Arg-Gly-Asp-dTyr-Lys (pentapeptide), were purchased from Bankpeptide Biological Technology Co., Ltd. (Hefei, Anhui, China). Trifluoroacetic acid (TFA) and liquid chromatography–mass spectrometry (LC-MS)-grade methanol and ethanol were obtained from Merck & Co., Inc. (Darmstadt, Germany). Ultrapure water in the whole process of the experiments was prepared using a Millipore Milli-Q system (Bedford, MA, USA). All other reagents and chemicals were purchased from Merck, unless otherwise noted.
## Tissue sectioning
For tissue sectioning, a Leica CM1860 cryostat (Leica Microsystems Inc., Wetzlar, Germany) was used. The frozen litchi seeds were cryo-sectioned into 12-μm-thick slices at a temperature of −20°C, and then the cryo-sectioned samples were thaw-mounted instantly on the conductive indium tin oxide films of microscope glass slides purchased from Bruker Daltonics (Bremen, Germany) (Figures 1A, B).
**Figure 1:** *Schematic diagram of MALDI-MSI procedure for imaging flavonoids in litchi seeds. (A) Whole litchi seeds were used for transection into 12-μm-thick slices in a cryostat microtome. (B) Serial tissue sections were immediately thaw-mounted on the conductive sides of indium tin oxide (ITO)-coated microscope glass slides. Optical images of the litchi seed section were obtained using a scanner. (C) To assist ionization, the sections were coated with the organic matrix. (D) MALDI-TOF-MS was used to detect analytes in situ on the surface of litchi seed tissue sections. The mass spectra of ionized analytes were acquired at each detected pixel point. (E) MS images of analytes were reconstructed from the MS spectra obtained at each laser spot using specific imaging reconstruction software. MALDI-MSI, matrix-assisted laser desorption/ionization mass spectrometry imaging; TOF, time of fight; MS, mass spectrometry.*
## Matrix coating
After being air-dried, the serial litchi seed tissue sections were used for MALDI matrix coating (Figure 1C). A 2-MBT matrix solution was prepared at an optimal concentration of 15 mg/ml and dissolved in methanol/water/TFA (80:20:0.2, v/v/v). Air-dried tissue sections were coated with the 2-MBT matrix solution by a GET-Sprayer (III) (HIT Co., Ltd, Beijing, China). Briefly, the 2-MBT matrix solution 15 cycles (5 s spray, 10 s incubation, and 20 s drying time) was sprayed on the surface of the tissue sections to pre-seed a thin layer of the 2-MBT matrix. After the tissue sections were completely air-dried in a vented fume hood, the matrix solution was evenly sprayed for 50 more of the same cycles.
## Histological staining
In order to obtain the histological images of litchi seed tissue sections, a slightly modified hematoxylin and eosin staining method was carried out based on an established procedure (Casadonte and Caprioli, 2011). Briefly, the tissue sections were washed in a series of ethanol solutions ($100\%$, $95\%$, $80\%$, and $70\%$ aqueous ethanol; 15 s/wash). After 10-s ultrapure water washing, tissue sections were stained with hematoxylin solution for 2 min and then washed with ultrapure water and $70\%$ and $95\%$ aqueous ethanol for 30 s each. The eosin solution was applied for another 1 min. Then, all tissue sections were washed with $95\%$ and $100\%$ ethanol and xylene for 30-s dehydration.
## Optimal image acquisition
Optical images of the tissue sections were acquired using an Epson Perfection V550 photo scanner (Seiko Epson Corp, Suwa, Japan) according to previous studies (Wu et al., 2021; Shi et al., 2022).
## MALDI-MS
An Autoflex Speed MALDI time-of-fight (TOF)/TOF mass spectrometer (Bruker Daltonics) with a MALDI source equipped with a 2,000-Hz solid-state Smartbeam Nd: YAG UV laser (355 nm, Azura Laser AG, Berlin, Germany) was used for profiling and imaging (Figure 1D).
To acquire in situ (+) MS profiling data of flavonoids from the tissue sections, all mass spectra were obtained over the m/z range of 100 to 700, each mass spectrum included an accumulation of 50 laser scans, and each scan was amassed from 500 laser shots. Three biological replicates of the sample and three technical replicates of each biological replicate were performed for MALDI-MS data acquisition ($$n = 3$$ × 3). To acquire the images of flavonoids, a 250-μm laser raster step-size was utilized for flavonoid in situ detection in tissues, and each pixel (scan spot) included 300 laser shots. With the use of FlexImaging 4.1 (Bruker Daltonics), the three “teaching points” for the correct positioning of the solid-state UV laser (Smartbeam Nd: YAG) for spectral acquisition were marked around a tissue section using a white ink correction pen. The m/z values of the compound ions that can be used for external mass calibration were listed as follows: His ([M+H]+, m/z 156.0768), Gly-Gly-Leu (tripeptide, [M+H]+, m/z 246.1448), Ala-His-Lys (tripeptide, [M+H]+, m/z 355.2088), Leu-Leu-Tyr (tripeptide, [M+H]+, m/z 408.2493), and Arg-Gly-Asp-dTyr-Lys (pentapeptide, [M+H]+, m/z 620.3151). Gly-Gly-Leu (tripeptide, [M+H]+, m/z 246.1448) and Arg-Gly-Asp-dTyr-Lys (pentapeptide, [M+H]+, m/z 620.3151) ions were selected in combination with the matrix ion of 2-MBT([M+H]+, m/z 167.9942) for internal mass calibration in the cubic enhanced mode. For the MALDI-TOF-MS analysis, MS/MS spectra were acquired in collision-induced dissociation (CID) mode, and argon was used as the collision gas. The flavonoid fragment ions were acquired under the following condition: ion source 1, 19.0 kV; ion source 2, 17.4 kV; lens, 8.8 kV; reflector 1, 21.0 kV; reflector 2, 9.8 kV; and accelerating voltage, 20.0 kV. The UV laser power ranged from $65\%$ to $90\%$. MS/MS spectra were recorded based on no less than 5,000 laser shots over the m/z range of 0 to 100 with a sampling rate of 2.00 G/s, a detector gain of 9.5×, and an electronic gain of 100 mV.
## Data analysis
For the MS profiling and MS/MS data analysis, Bruker FlexAnalysis 3.4 (Bruker Daltonics) was used for the preliminary viewing and processing of the mass spectra. Once the monoisotopic peak list was generated and exported, two metabolome databases (METLIN and HMDB) (Tautenhahn et al., 2012; Wishart et al., 2022) were used for the search of the detected m/z values of precursor ions and CID fragment ions against potential metabolite identities within an acceptable mass error of ±5 ppm. Three ion adduct forms (i.e., [M + H]+, [M + Na]+, and [M + K]+) were considered for the database search. For MALDI tissue imaging, Bruker FlexImaging 4.1 software was used for the reconstitution of the ion maps of the detected flavonoids (Figure 1E). For the generation of the ion images using FlexImaging, the mass filter width was set at 5 ppm.
## Flavonoid extraction and identification by LC-MS/MS
Flavonoids were extracted from the seeds of litchi for LC-MS/MS analysis. The details of the extraction of the flavonoids from litchi seeds and the procedure of LC-MS/MS analysis for the identification and structural confirmation of the flavonoids can be found in the Supplementary Material.
## Morphological characteristics of litchi seeds
As shown in Figures 2A, B, under a light microscope, the litchi seed showed the following structures: testa, micropyle, embryo, cotyledon, and cotyledon gap. Among these structures, the testa was dark coffee-colored, the embryo was brown, and the cotyledon was oyster white. In addition, a gap was observed in the middle of the cotyledon. After hematoxylin and eosin staining, litchi seeds were observed again under a light microscope (Figure 2C). The anatomical structure of the litchi seeds is illustrated in Figure 2D.
**Figure 2:** *Optimal images of litchi seed tissue sections. (A, B) Photos of litchi seed tissue sections. (C) An H&E-stained litchi seed tissue section. (D) A cartoon of anatomical structure of litchi seed tissue section. H&E, hematoxylin and eosin.*
## Flavonoids detected in situ by MALDI-TOF-MS
As shown in Figure 3, many flavonoid-related signals were detected in the m/z range of 100–700. These compounds were confirmed by comparing the m/z values and MS/MS spectra with those obtained by LC-MS/MS (Table 1). According to collision-induced dissociation, 15 flavonoids compounds were identified through MALDI-TOF-MS: liquiritigenin (m/z 257.081, [M+H]+), apigenin (m/z 271.060, [M+H]+), naringenin (m/z 273.076, [M+H]+), daidzein (m/z 293.020, [M+Na]+), luteolin (m/z 287.056, [M+H]+), dihydrokaempferol (m/z 289.071, [M+H]+), catechin (m/z 329.043, [M+H]+), quercetin (m/z 303.051, [M+H]+), kaempferol (m/z 309.036, [M+Na]+), isorhamnetin (m/z 317.066, [M+H]+), myricetin (m/z 319.046, [M+H]+), quercetin 3-β-d-glucoside (m/z 465.102, [M+H]+), baicalin (m/z 469.073, [M+Na]+), rutin (m/z 649.118, [M+K]+), and taxifolin (m/z 305.065, [M+H]+).
**Figure 3:** **Mass spectrum* of flavonoids detected in situ in a litchi seed tissue section using MALDI-TOF-MS and 2-MBT as the matrix in the positive ion mode. Two peptide standard ions, Gly-Gly-Leu (tripeptide, [M+H]+, m/z 246.145) and Arg-Gly-Asp-dTyr-Lys (pentapeptide, [M+H]+, m/z 620.322), and one matrix ion, 2-MBT ([M+H]+, m/z 167.994), were used as the reference peaks and are labeled with black triangle “▲”. Three biological repetitions and three technical repetitions were performed ($$n = 3$$ × 3). MALDI, matrix-assisted laser desorption/ionization; TOF, time-of-fight; MS, mass spectrometry; 2-MBT, 2-mercaptobenzothiazole.* TABLE_PLACEHOLDER:Table 1
## MALDI-MS imaging of flavonoids
MALDI-MSI can provide a snapshot of the distribution of molecules at a specific location on a tissue surface. We present the mass spectrometry images of all 15 flavonoids and performed our classification analysis in Figure 4.
**Figure 4:** *Ion images of 15 detectable flavonoids in litchi seed tissue sections from MALDI-TOF-MS using 2-MBT as the matrix in positive ion mode. MS imaging was acquired at 250-μm spatial resolution. MALDI, matrix-assisted laser desorption/ionization; TOF, time of fight; MS, mass spectrometry; 2-MBT, 2-mercaptobenzothiazole.*
Ion images of the 15 flavonoids indicated that they can be broadly classified into four types. Four compounds were distributed mainly in the embryo: liquiritigenin (m/z 257.081, [M+H]+), luteolin (m/z 287.056, [M+H]+), dihydrokaempferol (m/z 289.071, [M+H]+), and kaempferol (m/z 309.036, [M+Na]+). Luteolin was highly concentrated in the embryo and less concentrated in other parts, while kaempferol was distributed at low abundance in the cotyledons and more in the embryo. Myricetin (m/z 319.046, [M+H]+), baicalin (m/z 469.073, [M+Na]+), and rutin (m/z 649.118, [M+K]+) were primarily distributed in the cotyledons. Baicalin and rutin were distributed at the periphery of the cotyledons, and myricetin was distributed to one side of the cotyledon gap. Most of the compounds were distributed in both cotyledons and embryos, including naringenin (m/z 273.076, [M+H]+), apigenin (m/z 271.060, [M+H]+), daidzein (m/z 293.020, [M+Na]+), quercetin (m/z 303.051, [M+H]+), isorhamnetin (m/z 317.066, [M+H]+), catechin (m/z 329.043, [M+H]+), and quercetin 3-β-d-glucoside (m/z 465.102, [M+H]+). Naringenin and catechin were concentrated throughout the litchi seed, their distribution being more homogeneous and without obvious tissue specificity. Quercetin, quercetin 3-β-d-glucoside, and apigenin were distributed at the periphery of the cotyledons and in the embryo. The compound daidzein was uniformly distributed, whereas isorhamnetin was more distributed at the apical part of the cotyledons. Finally, the taxifolin (m/z 305.065, [M+H]+) content was low and mainly distributed in the inner seed testa.
Four compounds were mainly distributed in the embryo: liquiritigenin, luteolin, dihydrokaempferol, and kaempferol. As the embryo is the most important part of the seed in plant development, these flavonoids may provide essential substances for growth and development and improve seed resistance. Luteolin was highly concentrated in the embryo and less concentrated in other parts. Luteolin, through inducing root nodulation, plays an important role in nitrogen metabolism in nitrogen-fixing plants and enhanced plant stress tolerance by promoting its nitrogen enrichment (Peters et al., 1986). Liquiritigenin was also mainly concentrated in the embryo and lesser in the cotyledons close to the embryo. Liquiritigenin increases ultraviolet irradiation, indicating its anti-radiation function (Sun et al., 2012). Dihydrokaempferol and kaempferol were interconvertible; therefore, both had similar distribution characteristics and are distributed in the cotyledons as well as the embryo. Many studies have demonstrated that kaempferol, as a precursor of ubiquitin-ketone (coenzyme Q) biosynthesis, is an atypical node between primary and specialized metabolism (Soubeyrand et al., 2018; Berger et al., 2022). Kaempferol is involved in plant defense and signaling in response to stressful conditions (Soubeyrand et al., 2018; Jan et al., 2022). Dihydrokaempferol is involved in plant growth and development. As a precursor of orange pelargonidin-type anthocyanins, dihydrokaempferol plays a role in flower coloring (Johnson et al., 2001). Liquiritigenin rapidly inactivates the PI3K/AKT/mTOR pathway. In vivo studies demonstrated that liquiritigenin can significantly inhibit tumor growth, increase cell autophagy, and accelerate cell apoptosis. In addition, it attenuates the malignant-like biological behaviors in triple-negative breast cancer cells through its induction of autophagy-related apoptosis via the PI3K/AKT/mTOR pathway (Ji et al., 2021), decreased DNMT activity, and elevated BRCA1 expression and transcriptional activity (Liang et al., 2021). Dihydrokaempferol has strong anti-inflammatory and antioxidant activities, which can improve the inflammatory performance and oxidative stress state of acute pancreatitis (Liang et al., 2020; Zhang et al., 2021). In contrast, kaempferol shows more pharmacological activities, such as anti-bacterial (Yeon et al., 2019), anti-inflammatory (Yeon et al., 2019), anti-oxidant (Chen and Chen, 2013), antitumor (Calderón-Montaño et al., 2011), and anti-diabetic activities (Yang et al., 2021b), and are cardio-protective (Chen et al., 2022b) and neuro-protective (Wang et al., 2020). Currently, kaempferol is also commonly used in cancer chemotherapy (Ren et al., 2019). The mechanisms of kaempferol’s anticancer include apoptosis, cell cycle arrest at the G2/M phase, downregulation of epithelial–mesenchymal transition-related markers, and repression of overactivation of the phosphatidylinositol 3-kinase/protein kinase B signaling pathway (Imran et al., 2019; Wang et al., 2019). Luteolin sensitizes cancer cells to treatment-induced cytotoxicity via suppressing cell survival pathways and enhancing apoptosis pathways, including the apoptosis pathway of the tumor suppressor protein p53 (Lin et al., 2008). These compounds can be extracted from the embryo of litchi seeds, which is convenient for obtaining a higher content of target substances for pharmaceutical and mass production in the future.
Myricetin, baicalin, and rutin were mainly found in the cotyledons of litchi seeds. Myricetin was mainly concentrated on one side of the cotyledon gap, while rutin and baicalin were mainly distributed at the periphery (Figure 4). From a physiological point of view, flavonoids such as myricetin and baicalin assist in the reinforcement of plant tissues, maintenance of seed dormancy, and longevity of seeds during storage (Shirley, 1998). Rutin may participate in strengthening the plant’s defense system against environmental stresses, including UV exposure, low-temperature stress, drought stress, and bacterial pathogen infection (Suzuki et al., 2015; Yang et al., 2016). Myricetin has therapeutic effects on a variety of diseases, such as inflammation, cerebral ischemia, Alzheimer’s disease (AD), cancer, diabetes, pathogenic microorganism infection, thrombosis, and atherosclerosis (Song et al., 2021). Furthermore, myricetin has been reported to regulate the expression of STAT3, PI3K/AKT/mTOR, AChE, IκB/NF-κB, BrdU/NeuN, Hippo, eNOS/NO, ACE, MAPK, Nrf2/HO-1, TLR, and GSK-3β (Song et al., 2021). Rutin shows clear antioxidant and anticancer effects, including a strong ability to inhibit tumors in breast cancer, especially triple-negative breast cancer (Iriti et al., 2017; Liang et al., 2021). Baicalin, similar to rutin and myricetin, has inhibitory effects on lung, breast, and bladder cancers, through different signaling pathways and mechanisms (Ge et al., 2021; Kong et al., 2021; Zhao et al., 2021). Owing to their important pharmacological effects, our study of their spatial distribution provided a basis for the precise extraction of flavonoids for developing drugs.
Seven flavonoids, i.e., naringenin, apigenin, daidzein, quercetin, isorhamnetin, catechin, and quercetin-3-β-d-glucoside, were mainly found in both the cotyledon and embryo of litchi seeds. Among these compounds, catechin, naringenin, daidzein, apigenin, and quercetin-3-β-d-glucoside have homogeneous distributions with relatively high abundance. Isorhamnetin was mainly distributed in the radicle and tip of the cotyledon, while quercetin was distributed at the periphery of the cotyledon. Flavonoids are secondary metabolites in plants that play a critical role in impairing ultraviolet irradiation, regulating the oxidative stress response, and influencing the transport of plant hormones, flower coloring, and pathogen resistance (Buer et al., 2010; Chen et al., 2022a). Naringenin plays various roles in plant–microbe interactions (An et al., 2021). Lignin biosynthesis and coenzyme ligase (4CL) are involved in plant growth, and naringenin is one of the metabolites in this pathway that inhibit enzymes such as 4-CL (Deng et al., 2004). Apigenin (4′,5,7-trihydroxyflavone) is a bioactive compound that belongs to the flavone class, and it is the aglycone of many naturally occurring glycosides. It ameliorates the damaging effects of salinity on rice seedlings, presumably by regulating selective ion uptake by roots and translocation to shoots, thus maintaining the higher K+/Na+ ratio critical for normal plant growth under salinity stress (Mekawy et al., 2018). Daidzein, as an isoflavonoid, plays crucial roles in the expression of the nod genes of rhizobial bacteria. The expression of this compound in roots will increase the synthesis and secretion of nodulation factors, promoting a series of physiological changes in plant cells and initiating the formation of nodules (Bosse et al., 2021). Quercetin promotes a series of physiological and biochemical processes in plants, including seed germination, pollen growth, photosynthesis, and antioxidant machinery, thus facilitating proper plant growth and development (Singh et al., 2021). In addition, quercetin is an antioxidant that enhances plant resistance to some biotic and abiotic stresses. Quercetin-3-β-d-glucoside is a quercetin-derived compound with attached glucose instead of the 3-OH group of quercetin. Isorhamnetin is a methylated flavonoid derived from quercetin. Catechins, as a type of flavonoid, also belong to phenolic compounds. Making up more than $70\%$ of polyphenols, catechins consist of ester and non-ester catechins. The multifunctional catechins contribute to decreased reactive oxygen species and better adaptability of plants to the environment (Jiang et al., 2020). Some of these flavonoids have been previously extracted from litchi seeds, for example, catechin and naringenin (Zhu et al., 2019). Similar to other flavonoids, most of these compounds have many pharmacological effects, including anti-inflammatory, antioxidant, and antidiabetic activities. In particular, since the start of the COVID-19 epidemic, antiviral activity has been reported for catechin (Mishra et al., 2021) and quercetin (Bernini and Velotti, 2021). The antitumor effects of flavonoids have also been extensively studied, with the following mechanisms reported: inducing oxidative stress (Souza et al., 2017), enhancing chemotherapy drug effect (Yang et al., 2021a), and regulating signaling pathways (Amado et al., 2014). Notably, daidzein is a phytohormone similar to estrogens and thus may have a therapeutic effect on estrogen-dependent diseases (Meng et al., 2017). Therefore, flavonoid compounds are useful for developing drug-based therapies, and exploring the distribution of flavonoids will facilitate efficient extraction and utilization.
Although taxifolin was successfully detected in sections in situ using MALDI-MSI, the abundance of this compound was low. As shown in Figure 4, taxifolin was mainly found in the testa and peripheral part of the cotyledons, indicating that the compound can protect seed embryos from external biotic and abiotic factors, such as soil microbes (e.g., fungi and bacteria) and saline-alkali abiotic stress, thus improving seed vitality and germination rate (Ninfali et al., 2020; Wan et al., 2020). By regulating the aromatic hydrocarbon receptor/cytochrome P450 1A1 (CYP1A1) signaling pathway, taxifolin can significantly inhibit the proliferation, migration, invasion, and viability of gastric cancer cells (Xie et al., 2021). Similarly, the same effect of taxifolin has been observed on breast cancer by promoting mesenchymal-to-epithelial transition (EMT) through β-catenin signaling (Von Minckwitz et al., 2019).
## Conclusion
MALDI-MSI was used for in situ detection and imaging of flavonoid distribution in litchi seeds for the first time. Overall, 15 flavonoids were successfully imaged. Among them, four (dihydrokaempferol, liquiritigenin, luteolin, and kaempferol) were distributed in the seed embryo, three (rutin, baicalin, and myricetin) were mainly found in the cotyledons, seven (quercetin, naringenin, isorhamnetin, daidzein, apigenin, catechin, and quercetin 3-β-d-glucoside) were enriched in both the embryo and cotyledons, and one (taxifolin) was mainly detected in the inner testa. Our MALDI-MSI results showed clear tissue distribution heterogeneity for the different flavonoid compounds in litchi seeds. Such information will be important for further study to understand the physiological and chemical functions of such flavonoid compounds. Furthermore, our study provides a basis for further improving the efficiency of extracting and utilizing bioactive compounds from litchi seeds.
## Data availability statement
The original contributions presented in the study are included in the article/ Supplementary Material. Further inquiries can be directed to the corresponding author.
## Author contributions
Concepts: YL and XN. Design: ZW, FJ, and YL. Literature search: JW, ZZ, and YL. Data acquisition and analysis: ZW, XN, YL, and ZZ. Writing—original draft: YL, XN, and JW. Writing—review and editing: FJ, ZW, YL, and XN. Funding acquisition: YL, XN, and JW. Supervision: FJ and ZW. All authors contributed to the article and approved the submitted version.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2023.1144449/full#supplementary-material
## References
1. Amado N. G., Predes D., Fonseca B. F., Cerqueira D. M., Reis A. H., Dudenhoeffer A. C.. **Isoquercitrin suppresses colon cancer cell growth**. *J. Biol. Chem.* (2014) **289** 35456-35467. DOI: 10.1074/jbc.M114.621599
2. An J., Kim S. H., Bahk S., Vuong U. T., Nguyen N. T., Do H. L.. **Naringenin induces pathogen resistance against pseudomonas syringae through the activation of NPR1 in arabidopsis**. *Front. Plant Sci.* (2021) **12**. DOI: 10.3389/fpls.2021.672552
3. Berger A., Latimer S., Stutts L. R., Soubeyrand E., Block A. K., Basset G. J.. **Kaempferol as a precursor for ubiquinone (coenzyme q) biosynthesis: An atypical node between specialized metabolism and primary metabolism**. *Curr. Opin. Plant Biol.* (2022) **66**. DOI: 10.1016/j.pbi.2021.102165
4. Berk Ş., Kaya S., Akkol E. K., Bardakçı H.. **A comprehensive and current review on the role of flavonoids in lung cancer-experimental and theoretical approaches**. *Phytomedicine* (2022) **98**. DOI: 10.1016/j.phymed.2022.153938
5. Bernini R., Velotti F.. **Natural polyphenols as immunomodulators to rescue immune response homeostasis: Quercetin as a research model against severe COVID-19**. *Molecules* (2021) **26**. DOI: 10.3390/molecules26195803
6. Bosse M. A., Silva M. B. D., Oliveira N., Araujo M. A., Rodrigues C., Azevedo J. P.. **Physiological impact of flavonoids on nodulation and ureide metabolism in legume plants**. *Plant Physiol. Biochem.* (2021) **166** 512-521. DOI: 10.1016/j.plaphy.2021.06.007
7. Buer C. S., Imin N., Djordjevic M. A.. **Flavonoids: New roles for old molecules**. *J. Integr. Plant Biol.* (2010) **52** 98-111. DOI: 10.1111/j.1744-7909.2010.00905.x
8. Calderón-Montaño J. M., Burgos-Morón E., Pérez-Guerrero C., López-Lázaro M.. **A review on the dietary flavonoid kaempferol**. *Mini Rev. Med. Chem.* (2011) **11** 298-344. DOI: 10.2174/138955711795305335
9. Casadonte R., Caprioli R.M.. **Proteomic analysis of formalin-fixed paraffin-embedded tissue by MALDI imaging mass spectrometry**. *Nat. Protoc.* (2011) **6** 1695-1709. DOI: 10.1038/nprot.2011.388
10. Chen A. Y., Chen Y. C.. **A review of the dietary flavonoid, kaempferol on human health and cancer chemoprevention**. *Food Chem.* (2013) **138** 2099-2107. DOI: 10.1016/j.foodchem.2012.11.139
11. Chen J. H., Hou N., Xv X., Zhang D., Fan T. Q., Zhang Q. X.. **Flavonoid synthesis and metabolism during the fruit development in hickory (Carya cathayensis)**. *Front. Plant Sci.* (2022) **13**. DOI: 10.3389/fpls.2022.896421
12. Chen M., Xiao J., El-Seedi H. R., Woźniak K. S., Daglia M., Little P. J.. **Kaempferol and atherosclerosis: From mechanism to medicine**. *Crit. Rev. Food Sci. Nutr.* (2022) **1-19**. DOI: 10.1080/10408398.2022.2121261
13. Choi S.-A., Lee J. E., Kyung M. J., Youn J. H., Oh J. B., Whang W. K.. **Anti-diabetic functional food with wasted litchi seed and standard of quality control**. *Appl. Biol. Chem.* (2017) **60** 197-204. DOI: 10.1007/s13765-017-0269-9
14. Ciumărnean L., Milaciu M. V., Runcan O., Vesa Ș C., Răchișan A. L., Negrean V.. **The effects of flavonoids in cardiovascular diseases**. *Molecules* (2020) **25**. DOI: 10.3390/molecules25184320
15. Deng F., Aoki M., Yogo Y.. **Effect of naringenin on the growth and lignin biosynthesis of gramineous plants**. *Weed Biol. Manage.* (2004) **4** 49-55. DOI: 10.1111/j.1445-6664.2003.00119.x
16. Dong X., Huang Y., Wang Y., He X.. **Anti-inflammatory and antioxidant jasmonates and flavonoids from lychee seeds**. *J. Funct. Foods* (2019) **54** 74-80. DOI: 10.1016/j.jff.2018.12.040
17. Emanuele S., Lauricella M., Calvaruso G., D'anneo A., Giuliano M.. **Litchi chinensis as a functional food and a source of antitumor compounds: An overview and a description of biochemical pathways**. *Nutrients* (2017) **9**. DOI: 10.3390/nu9090992
18. Fan X., Bai J., Zhao S., Hu M., Sun Y., Wang B.. **Evaluation of inhibitory effects of flavonoids on breast cancer resistance protein (BCRP): From library screening to biological evaluation to structure-activity relationship**. *Toxicol. In Vitro* (2019) **61**. DOI: 10.1016/j.tiv.2019.104642
19. Ge A., Liu L., Deng X., Luo J., Xu Y.. **Exploring the mechanism of baicalin intervention in breast cancer based on MicroRNA microarrays and bioinformatics strategies**. *Evid. Based Complement. Alternat. Med.* (2021) **2021**. DOI: 10.1155/2021/7624415
20. Guo Y., Ding S. J., Ding X., Liu Z., Wang J. L., Chen Y.. **Effects of selected flavonoids oncellproliferation and differentiation of porcine muscle stem cells for cultured meat production**. *Food Res. Int.* (2022) **160**. DOI: 10.1016/j.foodres.2022.111459
21. Hu F.-C., Chen Z., Wang X.-H., Wang J.-B., Fan H.-Y., Qin Y.-H.. **Construction of high-density SNP genetic maps and QTL mapping for dwarf-related traits in litchi chinensis sonn**. *J. Integr. Agr.* (2021) **20** 2900-2913. DOI: 10.1016/s2095-3119(20)63387-1
22. Huang L., Tang X., Zhang W., Jiang R., Chen D., Zhang J.. **Imaging of endogenous metabolites of plant leaves by mass spectrometry based on laser activated electron tunneling**. *Sci. Rep.* (2016) **6**. DOI: 10.1038/srep24164
23. Imran M., Salehi B., Sharifi-Rad J., Aslam Gondal T., Saeed F., Imran A.. **Kaempferol: A key emphasis to its anticancer potential**. *Molecules* (2019) **24**. DOI: 10.3390/molecules24122277
24. Iriti M., Kubina R., Cochis A., Sorrentino R., Varoni E. M., Kabała-Dzik A.. **Rutin, a quercetin glycoside, restores chemosensitivity in human breast cancer cells**. *Phytother. Res.* (2017) **31** 1529-1538. DOI: 10.1002/ptr.5878
25. Jan R., Khan M., Asaf S., Lubna S., Asif K. M.. **Bioactivity and therapeutic potential of kaempferol and quercetin: New insights for plant and human health**. *Plants (Basel)* (2022) **11**. DOI: 10.3390/plants11192623
26. Ji Y., Hu W., Jin Y., Yu H., Fang J.. **Liquiritigenin exerts the anti-cancer role in oral cancer**. *Bioengineered* (2021) **12** 6070-6082. DOI: 10.1080/21655979.2021.1971501
27. Jiang C. K., Ma J. Q., Liu Y. F., Chen J. D., Ni D. J., Chen L.. **Identification and distribution of a single nucleotide polymorphism responsible for the catechin content in tea plants**. *Hortic. Res.* (2020) **7**. DOI: 10.1038/s41438-020-0247-y
28. Johnson E. T., Ryu S., Yi H., Shin B., Cheong H., Choi G.. **Alteration of a single amino acid changes the substrate specificity of dihydroflavonol 4-reductase**. *Plant J.* (2001) **25** 325-333. DOI: 10.1046/j.1365-313x.2001.00962.x
29. Khan H., Ullah H., Martorell M., Valdes S. E., Belwal T., Tejada S.. **Flavonoids nanoparticles in cancer: Treatment, prevention and clinical prospects**. *Semin. Cancer Biol.* (2021) **69** 200-211. DOI: 10.1016/j.semcancer.2019.07.023
30. Kong N., Chen X., Feng J., Duan T., Liu S., Sun X.. **Baicalin induces ferroptosis in bladder cancer cells by downregulating FTH1**. *Acta Pharm. Sin. B.* (2021) **11** 4045-4054. DOI: 10.1016/j.apsb.2021.03.036
31. Li H., Huang D., Ma Q., Qi W., Li H.. **Factors influencing the technology adoption behaviours of litchi farmers in China**. *Sustainability* (2020) **12**. DOI: 10.3390/su12010271
32. Li B., Neumann E. K., Ge J., Gao W., Yang H., Li P.. **Interrogation of spatial metabolome of ginkgo biloba with high-resolution matrix-assisted laser desorption/ionization and laser desorption/ionization mass spectrometry imaging**. *Plant Cell Environ.* (2018) **41** 2693-2703. DOI: 10.1111/pce.13395
33. Li Y., Zhang T., Chen G. Y.. **Flavonoids and colorectal cancer prevention**. *Antioxidants (Basel)* (2018) **7**. DOI: 10.3390/antiox7120187
34. Liang X., Hu C., Liu C., Yu K., Zhang J., Jia Y.. **Dihydrokaempferol (DHK) ameliorates severe acute pancreatitis (SAP)**. *Life Sci.* (2020) **261**. DOI: 10.1016/j.lfs.2020.118340
35. Liang F., Zhang H., Gao H., Cheng D., Zhang N., Du J.. **Liquiritigenin decreases tumorigenesis by inhibiting DNMT activity and increasing BRCA1 transcriptional activity in triple-negative breast cancer**. *Exp. Biol. Med. (Maywood)* (2021) **246** 459-466. DOI: 10.1177/1535370220957255
36. Lin Y., Shi R., Wang X., Shen H. M.. **Luteolin, a flavonoid with potential for cancer prevention and therapy**. *Curr. Cancer Drug Targets* (2008) **8** 634-646. DOI: 10.2174/156800908786241050
37. Liskova A., Samec M., Koklesova L., Brockmueller A., Zhai K., Abdellatif B.. **Flavonoids as an effective sensitizer for anti-cancer therapy: Insights into multi-faceted mechanisms and applicability towards individualized patient profiles**. *Epma. J.* (2021) **12** 155-176. DOI: 10.1007/s13167-021-00242-5
38. Magne Nde C. B., Zingue S., Winter E., Creczynski-Pasa T. B., Michel T., Fernandez X.. **Flavonoids, breast cancer chemopreventive and/or chemotherapeutic agents**. *Curr. Med. Chem.* (2015) **22** 3434-3446. DOI: 10.2174/0929867322666150729115321
39. Maleki S. J., Crespo J. F., Cabanillas B.. **Anti-inflammatory effects of flavonoids**. *Food Chem.* (2019) **299**. DOI: 10.1016/j.foodchem.2019.125124
40. Martinez-Perez C., Ward C., Cook G., Mullen P., Mcphail D., Harrison D. J.. **Novel flavonoids as anti-cancer agents: mechanisms of action and promise for their potential application in breast cancer**. *Biochem. Soc. Trans.* (2014) **42** 1017-1023. DOI: 10.1042/BST20140073
41. Mekawy A. M. M., Abdelaziz M. N., Ueda A.. **Apigenin pretreatment enhances growth and salinity tolerance of rice seedlings**. *Plant Physiol. Biochem.* (2018) **130** 94-104. DOI: 10.1016/j.plaphy.2018.06.036
42. Meng H., Fu G., Shen J., Shen K., Xu Z., Wang Y.. **Ameliorative effect of daidzein on cisplatin-induced nephrotoxicity in mice**. *Oxid. Med. Cell. Longev.* (2017) **2017**. DOI: 10.1155/2017/3140680
43. Miron A., Aprotosoaie A. C., Trifan A., Xiao J.. **Flavonoids as modulators of metabolic enzymes and drug transporters**. *Ann. N. Y. Acad. Sci.* (2017) **1398** 152-167. DOI: 10.1111/nyas.13384
44. Mishra C. B., Pandey P., Sharma R. D., Malik M. Z., Mongre R. K., Lynn A. M.. **Identifying the natural polyphenol catechin as a multi-targeted agent against SARS-CoV-2 for the plausible therapy of COVID-19: An integrated computational approach**. *Brief Bioinform.* (2021) **22** 1346-1360. DOI: 10.1093/bib/bbaa378
45. Ninfali P., Panato A., Bortolotti F., Valentini L., Gobbi P.. **Morphological analysis of the seeds of three pseudocereals by using light microscopy and ESEM-EDS**. *Eur. J. Histochem.* (2020) **64**. DOI: 10.4081/ejh.2020.3075
46. Pan H., Zhou W., He W., Liu X., Ding Q., Ling L.. **Genistein inhibits MDA-MB-231 triple-negative breast cancer cell growth by inhibiting NF-kappaB activity**. *Int. J. Mol. Med.* (2012) **30** 337-343. DOI: 10.3892/ijmm.2012.990
47. Park K. I., Park H. S., Nagappan A., Hong G. E., Lee D. H., Kang S. R.. **Induction of the cell cycle arrest and apoptosis by flavonoids isolated from Korean citrus aurantium l. @ in non-small-cell lung cancer cells**. *Food Chem.* (2012) **135** 2728-2735. DOI: 10.1016/j.foodchem.2012.06.097
48. Peluso I., Raguzzini A., Serafini M.. **Effect of flavonoids on circulating levels of TNF-alpha and IL-6 in humans: a systematic review and meta-analysis**. *Mol. Nutr. Food Res.* (2013) **57** 784-801. DOI: 10.1002/mnfr.201200721
49. Peters N. K., Frost J. W., Long S. R.. **A plant flavone, luteolin, induces expression of rhizobium meliloti nodulation genes**. *Science* (1986) **233** 977-980. DOI: 10.1126/science.3738520
50. Piehowski P. D., Zhu Y., Bramer L. M., Stratton K. G., Zhao R., Orton D. J.. **Automated mass spectrometry imaging of over 2000 proteins from tissue sections at 100-mum spatial resolution**. *Nat. Commun.* (2020) **11**. DOI: 10.1038/s41467-019-13858-z
51. Punia S., Kumar M.. **Litchi (Litchi chinenis) seed: Nutritional profile, bioactivities, and its industrial applications**. *Trends Food Sci. Technol.* (2021) **108** 58-70. DOI: 10.1016/j.tifs.2020.12.005
52. Qin L., Zhang Y., Liu Y., He H., Han M., Li Y.. **Recent advances in matrix-assisted laser desorption/ionisation mass spectrometry imaging (MALDI-MSI) for**. *Phytochem. Anal.* (2018) **29** 351-364. DOI: 10.1002/pca.2759
53. Rakha A., Umar N., Rabail R., Butt M. S., Kieliszek M., Hassoun A.. **Anti-inflammatory and anti-allergic potential of dietary flavonoids: A review**. *Biomed. Pharmacother.* (2022) **156**. DOI: 10.1016/j.biopha.2022.113945
54. Ren J., Lu Y., Qian Y., Chen B., Wu T., Ji G.. **Recent progress regarding kaempferol for the treatment of various diseases (Review)**. *Exp. Ther. Med.* (2019) **18** 2759-2776. DOI: 10.3892/etm.2019.7886
55. Shi Y., Hu H., Hao Q., Wu R., Wang L., Qin L.. **Michler's ethylketone as a novel negative-ion matrix for the enhancement of lipid MALDI tissue imaging**. *Chem. Commun. (Camb.)* (2022) **58** 633-636. DOI: 10.1039/d1cc05718a
56. Shirley B. W.. **Flavonoids in seeds and grains: Physiological function, agronomic importance and the genetics of biosynthesis**. *Seed Sci. Res.* (1998) **8** 415-422. DOI: 10.1017/s0960258500004372
57. Singh P., Arif Y., Bajguz A., Hayat S.. **The role of quercetin in plants**. *Plant Physiol. Biochem.* (2021) **166** 10-19. DOI: 10.1016/j.plaphy.2021.05.023
58. Song X., Tan L., Wang M., Ren C., Guo C., Yang B.. **Myricetin: A review of the most recent research**. *Biomed. Pharmacother.* (2021) **134**. DOI: 10.1016/j.biopha.2020.111017
59. Soubeyrand E., Johnson T. S., Latimer S., Block A., Kim J., Colquhoun T. A.. **The peroxidative cleavage of kaempferol contributes to the biosynthesis of the benzenoid moiety of ubiquinone in plants**. *Plant Cell* (2018) **30** 2910-2921. DOI: 10.1105/tpc.18.00688
60. Souza R. P., Bonfim-Mendonca P. S., Gimenes F., Ratti B. A., Kaplum V., Bruschi M. L.. **Oxidative stress triggered by apigenin induces apoptosis in a comprehensive panel of human cervical cancer-derived cell lines**. *Oxid. Med. Cell. Longev.* (2017) **2017**. DOI: 10.1155/2017/1512745
61. Sudhakaran M., Sardesai S., Doseff A. I.. **Flavonoids: New frontier for immuno-regulation and breast cancer control**. *Antioxidants (Basel)* (2019) **8**. DOI: 10.3390/antiox8040103
62. Sun R., Hikosaka S., Goto E., Sawada H., Saito T., Kudo T.. **Effects of UV irradiation on plant growth and concentrations of four medicinal ingredients in Chinese licorice (Glycyrrhiza uralensis)**. *Acta Hortic.* (2012) **956** 643-648. DOI: 10.17660/ActaHortic.2012.956.77
63. Surichan S., Androutsopoulos V. P., Sifakis S., Koutala E., Tsatsakis A., Arroo R. R.. **Bioactivation of the citrus flavonoid nobiletin by CYP1 enzymes in MCF7 breast adenocarcinoma cells**. *Food Chem. Toxicol.* (2012) **50** 3320-3328. DOI: 10.1016/j.fct.2012.06.030
64. Suzuki T., Morishita T., Kim S.-J., Park S.-U., Woo S.-H., Noda T.. **Physiological roles of rutin in the buckwheat plant**. *Japan Agric. Res. Quarterly: JARQ* (2015) **49** 37-43. DOI: 10.6090/jarq.49.37
65. Taira S., Tokai M., Kaneko D., Katano H., Kawamura-Konishi Y.. **Mass spectrometry imaging analysis of location of procymidone in cucumber samples**. *J. Agric. Food Chem.* (2015) **63** 6109-6112. DOI: 10.1021/acs.jafc.5b00957
66. Tang Y., Yu C., Wu J., Chen H., Zeng Y., Wang X.. **Lychee seed extract protects against neuronal injury and improves cognitive function in rats with type II diabetes mellitus with cognitive impairment**. *Int. J. Mol. Med.* (2018) **41** 251-263. DOI: 10.3892/ijmm.2017.3245
67. Tautenhahn R., Cho K., Uritboonthai W., Zhu Z., Patti G. J., Siuzdak G.. **An accelerated workflow for untargeted metabolomics using the METLIN database**. *Nat. Biotechnol.* (2012) **30** 826-828. DOI: 10.1038/nbt.2348
68. Van De Plas R., Yang J., Spraggins J., Caprioli R. M.. **Image fusion of mass spectrometry and microscopy: A multimodality paradigm for molecular tissue mapping**. *Nat. Methods* (2015) **12** 366-372. DOI: 10.1038/nmeth.3296
69. Von Minckwitz G., Huang C. S., Mano M. S., Loibl S., Mamounas E. P., Untch M.. **Trastuzumab emtansine for residual invasive HER2-positive breast cancer**. *N. Engl. J. Med.* (2019) **380** 617-628. DOI: 10.1056/NEJMoa1814017
70. Vue B., Zhang S., Chen Q. H.. **Flavonoids with therapeutic potential in prostate cancer**. *Anticancer Agents Med. Chem.* (2016) **16** 1205-1229. DOI: 10.2174/1871520615666151008122622
71. Wan L., Lei Y., Yan L., Liu Y., Pandey M. K., Wan X.. **Transcriptome and metabolome reveal redirection of flavonoids in a white testa peanut mutant**. *BMC Plant Biol.* (2020) **20** 161. DOI: 10.1186/s12870-020-02383-7
72. Wang J., Mao J., Wang R., Li S., Wu B., Yuan Y.. **Kaempferol protects against cerebral ischemia reperfusion injury through intervening oxidative and inflammatory stress induced apoptosis**. *Front. Pharmacol.* (2020) **11**. DOI: 10.3389/fphar.2020.00424
73. Wang X., Yang Y., An Y., Fang G.. **The mechanism of anticancer action and potential clinical use of kaempferol in the treatment of breast cancer**. *Biomed. Pharmacother.* (2019) **117**. DOI: 10.1016/j.biopha.2019.109086
74. Wishart D. S., Guo A., Oler E., Wang F., Anjum A., Peters H.. **HMDB 5.0: The human metabolome database for 2022**. *Nucleic Acids Res.* (2022) **50** D622-D631. DOI: 10.1093/nar/gkab1062
75. Wu R., Qin L., Chen L., Ma R., Chen D., Liu H.. **Copper adhesive tape attached to the reverse side of a non-conductive glass slide to achieve protein MALDI-imaging in FFPE-tissue sections**. *Chem. Commun. (Camb.)* (2021) **57** 10707-10710. DOI: 10.1039/d1cc03629g
76. Xie J., Pang Y., Wu X.. **Taxifolin suppresses the malignant progression of gastric cancer by regulating the AhR/CYP1A1 signaling pathway**. *Int. J. Mol. Med.* (2021) **48**. DOI: 10.3892/ijmm.2021.5030
77. Yang L., Gao Y., Bajpai V. K., El-Kammar H. A., Simal-Gandara J., Cao H.. **Advance toward isolation, extraction, metabolism and health benefits of kaempferol, a major dietary flavonoid with future perspectives**. *Crit. Rev. Food Sci. Nutr.* (2021) 1-17. DOI: 10.1080/10408398.2021.1980762
78. Yang C., Song J., Hwang S., Choi J., Song G., Lim W.. **Apigenin enhances apoptosis induction by 5-fluorouracil through regulation of thymidylate synthase in colorectal cancer cells**. *Redox Biol.* (2021) **47**. DOI: 10.1016/j.redox.2021.102144
79. Yang W., Xu X., Li Y., Wang Y., Li M., Wang Y.. **Rutin-mediated priming of plant resistance to three bacterial pathogens initiating the early SA signal pathway**. *PLos One* (2016) **11**. DOI: 10.1371/journal.pone.0146910
80. Yao P., Gao Y., Simal-Gandara J., Farag M. A., Chen W., Yao D.. **Litchi (Litchi chinensis sonn.): A comprehensive review of phytochemistry, medicinal properties, and product development**. *Food Funct.* (2021) **12** 9527-9548. DOI: 10.1039/d1fo01148k
81. Yeon M. J., Lee M. H., Kim D. H., Yang J. Y., Woo H. J., Kwon H. J.. **Anti-inflammatory effects of kaempferol on helicobacter pylori-induced inflammation**. *Biosci. Biotechnol. Biochem.* (2019) **83** 166-173. DOI: 10.1080/09168451.2018.1528140
82. Zaima N., Goto-Inoue N., Hayasaka T., Setou M.. **Application of imaging mass spectrometry for the analysis of oryza sativa rice**. *Rapid Commun. Mass Spectrom.* (2010) **24** 2723-2729. DOI: 10.1002/rcm.4693
83. Zhang H. W., Hu J. J., Fu R. Q., Liu X., Zhang Y. H., Li J.. **Flavonoids inhibit cell proliferation and induce apoptosis and autophagy through downregulation of PI3Kγ mediated PI3K/AKT/mTOR/p70S6K/ULK signaling pathway in human breast cancer cells**. *Sci. Rep.* (2018) **8** 11255. DOI: 10.1038/s41598-018-29308-7
84. Zhang J., Hu C., Li X., Liang L., Zhang M., Chen B.. **Protective effect of dihydrokaempferol on acetaminophen-induced liver injury by activating the SIRT1 pathway**. *Am. J. Chin. Med.* (2021) **49** 705-718. DOI: 10.1142/S0192415X21500324
85. Zhao L., Wang K., Wang K., Zhu J., Hu Z.. **Nutrient components, health benefits, and safety of litchi (Litchi chinensis sonn.): A review**. *Compr. Rev. Food Sci. Food Saf.* (2020) **19** 2139-2163. DOI: 10.1111/1541-4337.12590
86. Zhao F., Zhao Z., Han Y., Li S., Liu C., Jia K.. **Baicalin suppresses lung cancer growth phenotypes**. *Bioengineered* (2021) **12** 1699-1707. DOI: 10.1080/21655979.2021.1922052
87. Zhu X., Wang H., Sun J., Yang B., Duan X., Jiang Y.. **Pericarp and seed of litchi and longan fruits: Constituent, extraction, bioactive activity, and potential utilization**. *J. Zhejiang Univ. Sci. B* (2019) **20** 503-512. DOI: 10.1631/jzus.B1900161
|
---
title: 'Myalgic Encephalomyelitis/Chronic Fatigue Syndrome is common in post-acute
sequelae of SARS-CoV-2 infection (PASC): Results from a post-COVID-19 multidisciplinary
clinic'
authors:
- Hector Bonilla
- Tom C. Quach
- Anushri Tiwari
- Andres E. Bonilla
- Mitchell Miglis
- Phillip C. Yang
- Lauren E. Eggert
- Husham Sharifi
- Audra Horomanski
- Aruna Subramanian
- Liza Smirnoff
- Norah Simpson
- Houssan Halawi
- Oliver Sum-ping
- Agnieszka Kalinowski
- Zara M. Patel
- Robert William Shafer
- Linda N. Geng
journal: Frontiers in Neurology
year: 2023
pmcid: PMC9998690
doi: 10.3389/fneur.2023.1090747
license: CC BY 4.0
---
# Myalgic Encephalomyelitis/Chronic Fatigue Syndrome is common in post-acute sequelae of SARS-CoV-2 infection (PASC): Results from a post-COVID-19 multidisciplinary clinic
## Abstract
### Background
The global prevalence of PASC is estimated to be present in 0·43 and based on the WHO estimation of 470 million worldwide COVID-19 infections, corresponds to around 200 million people experiencing long COVID symptoms. Despite this, its clinical features are not well-defined.
### Methods
We collected retrospective data from 140 patients with PASC in a post-COVID-19 clinic on demographics, risk factors, illness severity (graded as one-mild to five-severe), functional status, and 29 symptoms and principal component symptoms cluster analysis. The Institute of Medicine (IOM) 2015 criteria were used to determine the Encephalomyelitis/Chronic Fatigue Syndrome (ME/CFS) phenotype.
### Findings
The median age was 47 years, $59.0\%$ were female; $49.3\%$ White, $17.2\%$ Hispanic, $14.9\%$ Asian, and $6.7\%$ Black. Only $12.7\%$ required hospitalization. Seventy-two ($53.5\%$) patients had no known comorbid conditions. Forty-five ($33.9\%$) were significantly debilitated. The median duration of symptoms was 285.5 days, and the number of symptoms was 12. The most common symptoms were fatigue ($86.5\%$), post-exertional malaise ($82.8\%$), brain fog ($81.2\%$), unrefreshing sleep ($76.7\%$), and lethargy ($74.6\%$). Forty-three percent fit the criteria for ME/CFS, majority were female, and obesity (BMI > 30 Kg/m2) ($$P \leq 0.00377895$$) and worse functional status ($$P \leq 0.0110474$$) were significantly associated with ME/CFS.
### Interpretations
Most PASC patients evaluated at our clinic had no comorbid condition and were not hospitalized for acute COVID-19. One-third of patients experienced a severe decline in their functional status. About $43\%$ had the ME/CFS subtype.
## Introduction
While most patients recover within weeks of SARS-CoV-2 infection, others experience debilitating symptoms that persist beyond the acute period [1]. The overall global prevalence of post-COVID-19 conditions is estimated at 0.43 of acute cases (hospitalized 0.54 and non-hospitalized 0.36) [2]. These post-COVID conditions, collectively known as a Post-Acute Sequelae of SARS-CoV-2 infection (PASC), or long COVID, are increasingly recognized even in patients who experience asymptomatic or mild SARS-CoV-2 infection [3]. The Center for Disease Control and Prevention (CDC) defines post-COVID conditions as symptoms persisting beyond 28 days after infection [4], while the UK National Institute for Health and Care Excellence (NICE) [5] and the World Health Organization (WHO) define this syndrome as symptoms persisting beyond 12 weeks after infection [6]. The American Academic of Physical Medicine and Rehabilitation (AAPMR&R) estimates that there are more than 24 million cases of long COVID as of May, 2022 [7].
PASC is most often characterized by extreme fatigue exacerbated by exertion, referred to as post-exertional malaise, difficulty with concentration and memory often referred to as brain fog, sleep disturbances, headaches, chest pain, and shortness of breath [2, 8]. PASC can range from mild to severe and incapacitating, interfering with patients' daily activities and work requirements [2, 8].
In an online, multi-national survey of 3,762 participants with confirmed or suspected COVID-19, fatigue, post-exertional malaise, and brain fog were the most frequently reported symptoms six months after SARS-CoV-2 infection [8]. This cluster of symptoms shares similar features with Encephalomyelitis/Chronic Fatigue Syndrome (ME/CFS), an often-debilitating disease that has a worldwide prevalence close to one percent, is also believed to frequently arise following a viral infection [9, 10]. Women are affected 1.5 to 2 times more often than men [2]. In two small cohorts of 41 and 42 COVID-19 patients ~$45\%$ met the ME/CFS diagnostic criteria [11, 12]. The pathobiological mechanism of ME/CFS is not known, but the large number of PASC patients presenting with features of this syndrome may allow us to better understand both conditions. The goals of our study are 2-fold: 1. Characterize our clinical cohort in terms of demographics and clinical presentations. 2. Estimate the prevalence of ME/CFS phenotype in PASC based on the IOM ME/CFS criteria [9].
## Study design and participants
The Stanford referral Post-Acute COVID-19 Syndrome (PACS) *Clinic is* a multidisciplinary center designed to provide clinical expertise in post-COVID-19 conditions, standardization of data collection and clinical management, and integration of research efforts. We standardized the clinical assessment by developing a clinic template embedded in the Electronic Health Records (EHR), allowing the extraction of the same data retrospectively. We used Research Electronic Data Capture (REDCap) and Microsoft Excel platforms for data collection and analysis. The study was approved by the Stanford University Institutional Review Board.
One hundred-forty consecutive adult patients with a history of COVID-19 were seen in the Stanford PACS clinic between May 18, 2021, and February 1, 2022. Referral criteria included a history of symptomatic SARS-CoV-2 infection, a diagnostic test for SARS-CoV-2 by either PCR, antigen detection, or a positive serology before SARS-CoV-2 vaccination, and persistent symptoms for at least 28 days following infection (CDC definition). In addition, each patient's initial symptoms during the acute infection was obtained from the questionnaire before their scheduled clinic visit and from the patient's EHR. The questionnaire includes: (i) the 29 symptoms reported to occur commonly in patients with acute COVID-19 (Supplementary material) to capture the different COVID-19 conditions such as ME/CFS, autonomic, respiratory, cardiac, neurological, psychiatric, olfactory, and gastrointestinal disorders; (ii) the severity of each symptom based on the Likert scale (1 mild symptom and 5 severe) (iii) SARS-CoV-2 vaccination status; (iv) the modified Post COVID-19 Functional Status Scale (FSS) which classifies patients as either asymptomatic (level 1), symptomatic without limitations (level 2), symptomatic with reduced daily activity (level 3), symptomatic with a struggle to perform daily activities (level 4), or incapacitated and bedridden (level 5) [2, 13]. From the EHR, we extracted the following: [1] demographics including age, sex, and self-identified race/ethnicity; [2] laboratory results and radiological data obtained before the initial office visit; and [3] vital signs, body mass index (BMI), oxygen saturation, orthostatic blood pressure, heart rate measurements obtained at the initial visit.
A second and identical questionnaire was sent to each patient before the clinic visit, collecting the same information in the past seven days as we did for the acute infection (Supplementary material). The Institute of Medicine (IOM) 2015 diagnostic criteria was used to identify the ME/CFS cohort; the patients must have persistent symptoms for at least 6 months; severe and incapacitating fatigue (Likert severity scale 4, or 5), unrefreshing sleep, post-exertional malaise (PEM), and orthostatic intolerance (lightheadedness), or brain fog [9]. The ME/CFS cohort were patients who fit the IOM diagnostic criteria, and their clinic records were review to exclude those with a history of fatigue symptom before COVID-19 (Figure 1, flow chart).
**Figure 1:** *Flow chart. *Positive SARC-CoV-2 test and over 28 days with symptoms. **Severe fatigue, unrefreshing sleep, PEM, and brain fog or orthostatic intolerance.*
## Statistical analysis
To explore how the symptoms correlated with each other in our study cohort, we utilized R (version 3.6.1) cluster analysis algorithms in 134 post-COVID-19 patients based on principal components analysis (PCA) to visualize the distribution of symptoms [14, 15]. We used the Wilcoxon Rank Sum Test to calculate the difference in the number of symptoms between males and females, Chi-square to calculate the difference in two categorical variables symptoms, T-test to compare the differences between groups and the Spearman Coefficient Correlation (rho) to compare the frequency of the symptoms and the median severity.
## Results
Among the 140 patients referred to the Stanford PASC clinic, six patients were excluded because of lack of a diagnostic test for SARS-CoV-2 infection or an incomplete questionnaire. For the remaining 134 patients, the median age was 47 years old (IQR: 30–60; range: 20–88). Seventy-nine ($59\%$) were female, 66 self-identified as White ($49.3\%$), 23 as Hispanic ($17.2\%$), 20 as Asian ($14.9\%$), and 9 as Black ($6.7\%$). Seventeen ($12.7\%$) required hospitalization for their acute infection and were considered to have had severe disease. Two of these patients required Intensive Care Unit (ICU) admission. Sixty-two ($46.3\%$) patients had a comorbid condition known to predispose to severe disease including obesity (BMI > 30 kg/m2) 47 ($35.1\%$), hypertension 24 ($17.9\%$), chronic lung disease 16 ($11.9\%$), diabetes 9 ($6.7\%$), immunosuppressive therapy 5 ($3.7\%$) and/or cardiovascular disease 4 ($3\%$). One hundred and nine ($82\%$) patients had some degree of functional limitation (i.e., FSS of 3, 4, or 5), including 45 ($33.9\%$) who exhibited significantly compromised wellbeing (FSS of 4 or 5) (Table 1).
**Table 1**
| Unnamed: 0 | PASC cohort (N = 134) | Chronic fatigue cohort (N = 105) persistent symptoms over 6 months | Chronic fatigue cohort (N = 105) persistent symptoms over 6 months.1 | Chronic fatigue cohort (N = 105) persistent symptoms over 6 months.2 |
| --- | --- | --- | --- | --- |
| Characteristics | | ME/CFS | No ME/CFS | Total— P -value |
| Age (years) median (range) | 47 (20–88) | 50 (23–75) | 47 (21–88) | 47 (21–88), 0.26755835 |
| Sex: female/male (female %) | 79/55 (59%) | 26/19 (58%) | 42/18 (70%) | 68/38, 0.48175231 |
| Days post-COVID median (range) | 285.5 (34–792) | 318 (183–686) | 310 (149–792) | 310 (183–792), 0.13233956 |
| Diagnostic test N (%) | | | | |
| Molecular/PCR | 119 (88.8%) | 38 (84%) | 54 (90%) | 92 |
| Antigen | 5 (3.7%) | 2 (6%) | 3 (5%) | 5 |
| Antibodies (+) before Vaccination | 10 (7.5%) | 5 (10%) | 3 (5%) | 8 |
| Treatment | | | | |
| Ambulatory | 117/134 (87.3%) | 44 (98%) | 57 (95%) | 101 (96%) |
| Hospital | 17/134 (12.7%) | 1 (2%) | 3 (5%) | 4 (4%) |
| Functional status (%) | | | | 0.0110474* |
| V (%) | 9 (6.8%) | 5 (11%) | 3 (5%) | 8 |
| IV (%) | 36 (27.1%) | 19 (42%) | 13 (22%) | 32 |
| III (%) | 64 (48.1%) | 20 (44%) | 27 (45%) | 47 |
| 1–II (%) | 20 (15%) | 1 (3%) | 17 (28%) | 18 |
| Co-morbid condition N (%) | | | | 0.34714359 |
| | 72 (53.7%) | 17 (38%) | 36 (60%) | 53 |
| One | 37 (27.6%) | 13 (29%) | 18 (30%) | 31 |
| Two | 15 (11.2%) | 7 (15%) | 5 (8%) | 12 |
| Three | 7 (5.2%) | 9 (18%) | 1 (2%) | 10 |
| BMI kg/m2 (median, range) | | | | 0.00377895* |
| All | 26.2 (16.9–46.6) | 29.23 (18.9–47.9) | 25.6 (18.3–41.1) | 27.2 (16.9–46.6) |
| >30 (%) | 31 (33.6%) | 22 (49%) | 17 (28%) | 39 |
| >35 (%) | 21 (15%) | 10 (22%) | 8 (13%) | 19 |
| Race/ethnicity | | | | |
| White | 66 (49.3%) | 17 (38%) | 34 (57%) | 51 |
| Hispanic | 23 (17.2%) | 13 (28%) | 7 (12%) | 20 |
| Black | 9 (6.7%) | 2 (4%) | 4 (7%) | 6 |
| Asian | 20 (14.9%) | 4 (8%) | 8 (13%) | 12 |
| Others | 2 (1.4%) | 1 (2%) | 1 (1%) | 2 |
| No data | 15 (11%) | 8 (18%) | 6 (10%) | 14 |
| Total | 134 (100%) | 45 (43%) | 60 (57%) | 105 (100%) |
The duration of the symptoms at the time of the initial office visit ranged from 34 to 792 days with a median duration of 285.5 days. Of the 29 symptoms assessed in the study, the median number per patient was 12 (range: 1–25 symptoms). The most common symptoms were fatigue ($86.5\%$), post-exertional malaise ($82.8\%$), brain fog ($81.2\%$), unrefreshing sleep ($76.7\%$), and daytime sleepiness ($74.6\%$). The median number of symptoms in female patients was greater than the median number in male patients (13 vs. 10, $$p \leq 0.005$$). The most common symptoms significantly higher in the female population were fatigue, insomnia, and change in taste or dysgeusia (P-values = 0.001, 0.044, and 0.01, respectively) (Table 2). There was a significant correlation between the frequency and the median severity of the symptom (Correlation coefficient is 0.86, $P \leq 0.001$).
**Table 2**
| Symptom | Rates by sex (%) | Rates by sex (%).1 | P -value | Rates Likert scale 4–5** | Total (%)** |
| --- | --- | --- | --- | --- | --- |
| Symptom | Female | Male | | | |
| Fatigue | 88.6 | 81.8 | 0.0011* | 61.8 | 86.5 |
| Post exertional malaise | 84.8 | 80 | 0.4675 | 66.6 | 82.8 |
| Brain fog | 81 | 80 | 0.8840 | 50.0 | 81.2 |
| Unrefreshing sleep | 81 | 69.1 | 0.1113 | 49.0 | 76.7 |
| Lethargic | 78.5 | 69.1 | 0.2191 | 57.0 | 74.6 |
| Insomnia | 74.7 | 58.2 | 0.0441* | 44.0 | 65.9 |
| Headache | 65.8 | 63.6 | 0.9129 | 31.4 | 62.1 |
| Anxiety/depression | 57 | 67.3 | 0.3226 | 39.5 | 52.6 |
| Lightheadedness | 57 | 45.5 | 0.1895 | 32.8 | 50.8 |
| Gastrointestinal | 57 | 40 | 0.1289 | 20.9 | 48.9 |
| Shortness of breath | 51.9 | 41.8 | 0.3145 | 30.1 | 43.2 |
| Nasal congestion | 49.4 | 32.7 | 0.07590 | 21.4 | 42.0 |
| Changes of smell | 49.4 | 32.7 | 0.1024 | 32.7 | 41.2 |
| Changes of taste | 46.8 | 27.3 | 0.01* | 31.5 | 39.2 |
| Cough | 44.3 | 29.1 | 0.074 | 23.5 | 36.6 |
In the PCA analysis, the direction of arrows in the correlation circle illustrates the extent to which symptoms were likely to occur in the same patient while the factor map indicates the contribution of different symptoms to the first five components (Figure 3). The analysis suggests that fatigue, PEM, daytime sleepiness (lethargy), brain fog, and unrefreshing sleep were likely to occur together. These symptoms were major contributors to the first component while anosmia and nasal congestion were major contributors to the second component. Similar analysis was performed on 72 patients without pre-COVID comorbidities two main groups were identified one with anosmia, headache, and congestion and a second that includes unrefreshing sleep, fatigue, and lethargy.
We assessed the prevalence of the ME/CFS phenotype in 105 patients ($69\%$) who had persistent symptoms longer than 6 months. The most common symptoms were clustered in the ME/CFS (Fatigue, PEM, brain fog, unrefreshing sleep, and lethargy) and autonomic dysfunction categories (lightheadedness and gastrointestinal) (Figure 3). The top six most common and severe symptoms were fatigue, PEM, brain fog, unrefreshing sleep, daytime sleepiness, and insomnia (Figure 2). Forty-five ($43\%$) of the study cohort fulfill the IOM criteria for ME/CFS (Table 1). Like our PASC cohort, the ME/CFS cohort was predominantly female, non-hospitalized, and healthy individuals and obesity was the most common risk factor (BMI > 30 Kg/m2) (Table 1). But obesity and worse FSS (FFS 4 and 5) were significantly higher in the ME/CFS population ($$P \leq 0.00377895$$; $$P \leq 0.0110474$$, respectively) (Table 1).
**Figure 2:** *Distribution of the frequency and severity of the symptoms on the PASC clinic intake questionnaire for subgroup of 105 patients with persistent symptoms for six or more months. Symptom severity was measured on the Likert scale with a severity score of 5 being the most severe. CP, Chest pain; GISx, Gastrointestinal symptoms; Anex/Depr, Anxiety/Depression; HA, Headache; UnRefreSleep, Unrefreshed sleep; PEM, Post-Exertional Malaise; NCongest, Nasal congestion.*
## Discussion
This manuscript presents the characteristics of 134 patients referred to a long COVID clinic who had who have history of SARS-CoV-2 infection and more than 28 more days of symptoms. The median age in our cohort was 47 years and ranged from 20 to 88 years, with a female predominance of $59\%$. Similar age and sex distribution was reported in several other studies [2, 16]. In contrast to hospitalized patients, females PASC were represented in a lower proportion, less severe symptoms, and lower mortality than males (17–19). These differences in sex distribution could be explained by the variations in the immune response between males and females, and the fact that female patients have more robust inflammatory, antiviral, and humoral immune responses, biological differences on sex hormones, and expression and regulation of angiotensin-converting enzyme 2 (ACE2) [18, 20]. Our study cohort, similar to other PASC studies, the subjects were predominantly white females, with obesity [17, 21, 22] characteristics associated with lower likelihood for full recovery [23]. In contrast, some racial and ethnic minority groups, such as Native American Indians, Alaska Natives, Hispanic and Black, have been shown to have a disproportionately higher risk for infection, severity of illness, hospitalization, and deaths. Those groups were underrepresented in our study [24].
The large majority ($87\%$) of our PASC patients had not been hospitalized or required oxygen for COVID-19 disease. The prevalence of PASC in asymptomatic patients with the mild disease is reported between 30 and $60\%$ [3, 25]. We can postulate that unknown factors besides hypoxemia and hospitalization are the drivers of PASC symptoms such as virus persistence, overactivation of the immune system, amyloid fibrin microclots, auto-antibodies, virus reactivation and others, may play a role in the pathobiology of this illness [26].
In our study, the duration of symptoms ranged from 34 to 792 days, with a median duration of 285.5 days. Females experienced more symptoms than males. The most common symptoms were fatigue, post-exertional malaise, brain fog, unrefreshing sleep, and daytime sleepiness; and the frequency of the symptoms correlated with severity (Figure 2, Table 2) and the ME/CFS symptoms clustered with one another (Figure 3). A reliable and straightforward long COVID score system is needed in post-COVID research and clinical care. In a population-based cohort study of confirmed SARS-CoV-2 infection, higher severity scores correlate with lower health quality of life [27]. In a meta-analysis which found that fatigue/weakness, myalgia/arthralgia, depression, anxiety, memory loss, concentration difficulties, dyspnea, and insomnia, were the most prevalent symptoms [2, 28]. Fatigue was the most prevalent symptom across the PASC studies [2, 28]. Post-viral fatigue was commonly reported after a viral infection such as Influenza, Severe acute respiratory syndrome/Middle East respiratory syndrome coronavirus (SARS/MERS), and Ebola [29].
**Figure 3:** *Quality of representation of 13 most common symptoms mapped to the first five dimensions (A) and principal components analysis correlation circle (B). Insomnia and ageusia were not included in the analysis. The closer the variable to the correlation circle, the better the representation on the factor map. The quality of the representation of the symptom on the first five dimensions is measured by the squared cosine between the symptom vector and its projection on the dimension. The proportions on the left side of the factor map represent a color scale (14). CP, Chest pain; GISx, Gastrointestinal symptoms; Anex/Depr, Anxiety/Depression; HA, Headache; Lethargy, daytime sleepiness; UnRefrSleep, Unrefreshed sleep; PEM, Post-Exertional Malaise.*
When we compare the frequency of the symptoms and severity, we found a significant correlation between the frequency and the severity of the symptom (Likert scale 4 and 5). A reliable and simple long COVID score system is tool need in post COVID research and clinical care. It will be important to access clinical outcomes and evaluated therapeutics interventions. In a population-based cohort study confirmed SARS-CoV-2 infection, Bahmer et al., the higher score in severity of COVID symptoms correlated with lower health quality of life [27].
PCA is often used to transform a large set of variables into a smaller one that contains most of the information in the large set. Smaller data sets are easier to explore and visualize and facilitate analyzing data much easier and faster [15]. We used it to cluster symptoms and severity. We were able to visualize a group that resembles ME/CFS.
In a study on 233 SARS survivors, 40.3 % reported having chronic fatigue, and $27.1\%$ met the criteria for ME/CFS; after influenza with H1N1, ME/CFS has reported 2.08 cases/100,000 person-month [29]. In our study, $43\%$ of the selected cohort fulfilled all the ME/CFS criteria, which is similar to $45\%$ reported by Mancini et al. [ 11] and Kedor et al. [ 12]. Like ME/CFS, the ME/CFS-PASC phenotype was more prevalent among the non-hospitalized female population (Table 1). The clinical similarities in our study cohort between ME/CFS and ME/CFS-PASC allow us to suggest common pathobiology. Those similarities include a preceding a viral illness [30], increase in inflammatory cytokines, neuroinflammatory change, mitochondria dysfunction, and alteration in NK cell function [31]. Obesity (BMI > 30 Kg/m2) was in our PASC study cohort, and the PASC-ME/CFS is the most common risk factor for this illness that has also been recognized in multiple studies as a risk factor for Long COVID [32, 33]. In addition, PASC-ME/CFS was an indicator worse functional status. However, we were not able to confirm that all symptoms resulted from prior SARS-CoV-2 infection and that it is possible that many of the reported symptoms, including chronic fatigue had other contributing causes. We believe there is an important relationship between Obstructive Sleep Apnea (OSA) and COVID-19, with significant symptomatic overlap between these two conditions and emerging data suggesting COVID is a risk factor for OSA, increasing the risk for severity and hospitalization [34, 35]. For these reasons, we agree that the relationship between long COVID and OSA deserves to be further explored, though this exploration is beyond the intended scope of this study.
Our study has several limitations. First, this study only represents the experience of a single center in Northern California located in a generally affluent area with a bias toward specific populations. The selection of this cohort may skew our population in the follow ways: [1] this is a referred population with multiple and more severe symptoms [2] our clinics have a lower proportion of underrepresented minority populations. Therefore, a multicenter study that includes a more diverse and larger population is necessary to corroborate our findings.
In summary, about half of the PASC patients with more than 6 months of symptoms fulfilled the ME/CFS criteria, and PASC-ME/CFS is an indicator of worse functional status; the majority of patients seen in our PACS clinic were female, with a median age of 47 years old, with obesity as the most common comorbidity. The majority of the patients had mild to moderate acute infection and were healthy prior to their COVID infection. Fatigue, post-exertional malaise, brain fog, unrefreshing sleep, and daytime sleepiness were the most prevalent and severe symptoms. This commonality between ME/CFS and ME/CFS-PASC may suggest a shared pathobiology. Therefore, defining specific subtypes within the umbrella of PASC/post-COVID conditions can help us understand different pathogenic mechanisms to tailor treatment.
## Data availability statement
The original contributions presented in the study are included in the article/Supplementary material, further inquiries can be directed to the corresponding author.
## Ethics statement
The studies involving human participants were reviewed and approved by Stanford University IRB. Written informed consent for participation was not required for this study in accordance with the national legislation and the institutional requirements.
## Author contributions
HB: conception and design of the study, data collection, analysis, draft, editing of the manuscript, full access to the whole data, and final version approval. TQ, AT, and AB: data collection, full access to the whole data, analysis, draft, and editing of the manuscript. MM, PY, LE, AS, LS, NS, HH, ZP, HS, AH, OS-p, and AK: conception and design of the study, full access to the whole data, and editing of the manuscript. RS and LG: conception and design of the study, full access to the whole data, data collection, analysis, draft, editing of the manuscript, and final version approval. All authors contributed to the article and approved the submitted version.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher's note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fneur.2023.1090747/full#supplementary-material
## References
1. Tenforde MW, Kim SS, Lindsell CJ, Billig Rose E, Shapiro NI, Files DC. **Symptom duration and risk factors for delayed return to usual health among outpatients with COVID-19 in a Multistate Health care systems network - United States, March-June 2020**. *MMWR Morb Mortal Wkly Rep.* (2020) **69** 993-8. DOI: 10.15585/mmwr.mm6930e1
2. Chen C, Haupert SR, Zimmermann L, Shi X, Fritsche LG, Mukherjee B. **Global prevalence of post COVID-19 condition or long COVID: a meta-analysis and systematic review**. *J Infect Dis* (2022) **226** 1593-1607. DOI: 10.1101/2021.11.15.21266377
3. Malkova A, Kudryavtsev I, Starshinova A, Kudlay D, Zinchenko Y, Glushkova A. **Post COVID-19 syndrome in patients with asymptomatic/mild form**. *Pathogens* (2021) **10** 1408. DOI: 10.3390/pathogens10111408
4. 4.CDC. Long COVID or Post-COVID Conditions. Atlanta, GA: CDC (2022).. *Long COVID or Post-COVID Conditions* (2022)
5. 5.COVID-19 Rapid Guideline: Managing the Long-Term Effects of COVID-19. London: National Institute for Health and Care Excellence: Clinical Guidelines (2020).. *COVID-19 Rapid Guideline: Managing the Long-Term Effects of COVID-19* (2020)
6. Soriano JB, Murthy S, Marshall JC, Relan P, Diaz JV. **A clinical case definition of post-COVID-19 condition by a Delphi consensus**. *Lancet Infect Dis* (2022) **22** e102-7. DOI: 10.1016/S1473-3099(21)00703-9
7. 7.Aapmr&r. Post-Acute Sequelae of SARS-CoV-2 Infections (PASC) Estimates and Insights. (2022). 10.51648/jac.57. *Post-Acute Sequelae of SARS-CoV-2 Infections (PASC) Estimates and Insights.* (2022). DOI: 10.51648/jac.57
8. Davis HE, Assaf GS, McCorkell L, Wei H, Low RJ, Re'em Y. **Characterizing long COVID in an international cohort: 7 months of symptoms and their impact**. *EClinicalMedicine* (2021) **38** 101019. DOI: 10.1016/j.eclinm.2021.101019
9. **The National Academies Collection: Reports funded by National Institutes of Health**. *Beyond Myalgic Encephalomyelitis/Chronic Fatigue Syndrome: Redefining an Illness* (2015)
10. Lim EJ, Ahn YC, Jang ES, Lee SW, Lee SH, Son CG. **Systematic review and meta-analysis of the prevalence of chronic fatigue syndrome/myalgic encephalomyelitis (CFS/ME)**. *J Transl Med* (2020) **18** 100. DOI: 10.1186/s12967-020-02269-0
11. Mancini DM, Brunjes DL, Lala A, Trivieri MG, Contreras JP, Natelson BH. **Use of cardiopulmonary stress testing for patients with unexplained dyspnea post-coronavirus disease**. *JACC Heart Fail* (2021) **9** 927-37. DOI: 10.1016/j.jchf.2021.10.002
12. Kedor C, Meyer-Arndt L, Wittke K, Zoller T, Steinbeis F, Haffke M. **Chronic COVID-19 Syndrome and Chronic Fatigue Syndrome (ME/CFS) following the first pandemic wave in Germany: a first analysis of a prospective observational study**. *medRxiv* (2021). DOI: 10.1101/2021.02.06.21249256
13. Klok FA, Boon G, Barco S, Endres M, Geelhoed JJM, Knauss S. **The Post-COVID-19 functional status scale: a tool to measure functional status over time after COVID-19**. *Eur Respir J.* (2020) **56** 2001494. DOI: 10.1183/13993003.01494-2020
14. Lê S JJ, Husson F. **FactoMineR: an R package for multivariate analysis**. *J Stat Softw* (2008) **25** 1-18. DOI: 10.18637/jss.v025.i01
15. Kassambara A. *Practical Guide To Principal Component Methods in R* (2017)
16. Fernandez-de-Las-Penas C, Palacios-Cena D, Gomez-Mayordomo V, Florencio LL, Cuadrado ML, Plaza-Manzano G. **Prevalence of post-COVID-19 symptoms in hospitalized and non-hospitalized COVID-19 survivors: a systematic review and meta-analysis**. *Eur J Intern Med.* (2021) **92** 55-70. DOI: 10.1016/j.ejim.2021.06.009
17. Ganesh R, Grach SL, Ghosh AK, Bierle DM, Salonen BR, Collins NM. **The female-predominant persistent immune dysregulation of the post-COVID syndrome**. *Mayo Clin Proc* (2022) **97** 454-64. DOI: 10.1016/j.mayocp.2021.11.033
18. Viveiros A, Rasmuson J, Vu J, Mulvagh SL, Yip CYY, Norris CM. **Sex differences in COVID-19: candidate pathways, genetics of ACE2, and sex hormones**. *Am J Physiol Heart Circ Physiol* (2021) **320** H296-304. DOI: 10.1152/ajpheart.00755.2020
19. Sharma G, Volgman AS, Michos ED. **Sex differences in mortality from COVID-19 pandemic: are men vulnerable and women protected?**. *JACC Case Rep* (2020) **2** 1407-10. DOI: 10.1016/j.jaccas.2020.04.027
20. Ciarambino T, Para O, Giordano M. **Immune system and COVID-19 by sex differences and age**. *Womens Health* (2021) **17** 17455065211022262. DOI: 10.1177/17455065211022262
21. Bell ML, Catalfamo CJ, Farland LV, Ernst KC, Jacobs ET, Klimentidis YC. **Post-acute sequelae of COVID-19 in a non-hospitalized cohort: results from the Arizona CoVHORT**. *PLoS ONE* (2021) **16** e0254347. DOI: 10.1371/journal.pone.0254347
22. Pfaff ER, Girvin AT, Bennett TD, Bhatia A, Brooks IM, Deer RR. **Identifying who has long COVID in the USA: a machine learning approach using N3C data**. *Lancet Digit Health* (2022) **4** e532-41. DOI: 10.1016/S2589-7500(22)00048-6
23. Sanchis-Gomar F, Lavie CJ, Mehra MR, Henry BM, Lippi G. **Obesity and outcomes in COVID-19: when an epidemic and pandemic collide**. *Mayo Clin Proc* (2020) **95** 1445-53. DOI: 10.1016/j.mayocp.2020.05.006
24. Romano SD, Blackstock AJ, Taylor EV, El Burai Felix S, Adjei S, Singleton CM. **Trends in racial and ethnic disparities in COVID-19 hospitalizations, by region - United States, March-December 2020**. *MMWR Morb Mortal Wkly Rep* (2021) **70** 560-5. DOI: 10.15585/mmwr.mm7015e2
25. Jacobson KB, Rao M, Bonilla H, Subramanian A, Hack I, Madrigal M. **Patients with uncomplicated coronavirus disease 2019 (COVID-19) have long-term persistent symptoms and functional impairment similar to patients with severe COVID-19: a cautionary tale during a global pandemic**. *Clin Infect Dis* (2021) **73** e826-9. DOI: 10.1093/cid/ciab103
26. Couzin-Frankel J. **Clues to long COVID**. *Science* (2022) **376** 1261-5. DOI: 10.1126/science.add4297
27. Bahmer T, Borzikowsky C, Lieb W, Horn A, Krist L, Fricke J. **Severity, predictors and clinical correlates of Post-COVID syndrome (PCS) in Germany: a prospective, multi-centre, population-based cohort study**. *EClinicalMedicine* (2022) **51** 101549. DOI: 10.1016/j.eclinm.2022.101549
28. Han Q, Zheng B, Daines L, Sheikh A. **Long-term sequelae of COVID-19: a systematic review and meta-analysis of one-year follow-up studies on post-COVID symptoms**. *Pathogens* (2022) **11** 269. DOI: 10.3390/pathogens11020269
29. Poenaru S, Abdallah SJ, Corrales-Medina V, Cowan J. **COVID-19 and post-infectious myalgic encephalomyelitis/chronic fatigue syndrome: a narrative review**. *Ther Adv Infect Dis* (2021) **8** 20499361211009385. DOI: 10.1177/20499361211009385
30. Hickie I, Davenport T, Wakefield D, Vollmer-Conna U, Cameron B, Vernon SD. **Post-infective and chronic fatigue syndromes precipitated by viral and non-viral pathogens: prospective cohort study**. *BMJ* (2006) **333** 575. DOI: 10.1136/bmj.38933.585764.AE
31. Komaroff AL, Lipkin WI. **Insights from myalgic encephalomyelitis/chronic fatigue syndrome may help unravel the pathogenesis of postacute COVID-19 syndrome**. *Trends Mol Med* (2021) **27** 895-906. DOI: 10.1016/j.molmed.2021.06.002
32. Vimercati L, De Maria L, Quarato M, Caputi A, Gesualdo L, Migliore G. **Association between long COVID and overweight/obesity**. *J Clin Med* (2021). DOI: 10.3390/jcm10184143
33. Colaneri M, Novelli V, Cutti S, Muzzi A, Resani G, Monti MC. **The experience of the health care workers of a severely hit SARS-CoV-2 referral Hospital in Italy: incidence, clinical course and modifiable risk factors for COVID-19 infection**. *J Public Health* (2021) **43** 26-34. DOI: 10.1093/pubmed/fdaa195
34. Rognvaldsson KG, Eythornorsson ES, Emilsson OI, Eysteinsdottir B, Palsson R, Gottfreethsson M. **Obstructive sleep apnea is an independent risk factor for severe COVID-19: a population-based study**. *Sleep* (2022) **45** zsab272. DOI: 10.1093/sleep/zsab272
35. Strausz S, Kiiskinen T, Broberg M, Ruotsalainen S, Koskela J, Bachour A. **Sleep apnoea is a risk factor for severe COVID-19**. *BMJ Open Respir Res* (2021) **8** e000845. DOI: 10.1136/bmjresp-2020-000845
|
---
title: Gut microbiota facilitates adaptation of the plateau zokor (Myospalax baileyi)
to the plateau living environment
authors:
- Bin Hu
- Jiamin Wang
- Ying Li
- Jin Ge
- Jinchao Pan
- Gaojian Li
- Yongcai He
- Haishun Zhong
- Bo Wang
- Yanyi Huang
- Shuyi Han
- Yanan Xing
- Hongxuan He
journal: Frontiers in Microbiology
year: 2023
pmcid: PMC9998695
doi: 10.3389/fmicb.2023.1136845
license: CC BY 4.0
---
# Gut microbiota facilitates adaptation of the plateau zokor (Myospalax baileyi) to the plateau living environment
## Abstract
Gut microbiota not only helps the hosts to perform many key physiological functions such as food digestion, energy harvesting and immune regulation, but also influences host ecology and facilitates adaptation of the host to extreme environments. Plateau zokors epitomize successful physiological adaptation to their living environment in the face of the harsh environment characterized by low temperature, low pressure and hypoxia in the Tibetan plateau region and high concentrations of CO2 in their burrows. Therefore, here we used a metagenomic sequencing approach to explore how gut microbiota contributed to the adaptive evolution of the plateau zokor on the Qinghai-Tibet Plateau. Our metagenomic results show that the gut microbiota of plateau zokors on the Tibetan plateau is not only enriched in a large number of species related to energy metabolism and production of short-chain fatty acids (SCFAs), but also significantly enriched the KO terms that involve carbohydrate uptake pathways, which well address energy uptake in plateau zokors while also reducing inflammatory responses due to low pressure, hypoxia and high CO2 concentrations. There was also a significant enrichment of tripeptidyl-peptidase II (TPPII) associated with antigen processing, apoptosis, DNA damage repair and cell division, which may facilitate the immune response and tissue damage repair in plateau zokors under extreme conditions. These results suggest that these gut microbiota and their metabolites together contribute to the physiological adaptation of plateau zokors, providing new insights into the contribution of the microbiome to the evolution of mammalian adaptation.
## Introduction
Gut microbiota can help the host perform many key physiological functions such as food digestion, energy harvesting and immune regulation (Tsuchida et al., 2004; Tremaroli and Bäckhed, 2012; Amato et al., 2014; Zhang et al., 2015). It may also strongly affect host ecology, for example, commensal microbes in animal’s guts often help to exclude bacterial pathogens (Łukasik et al., 2013; Steele et al., 2021), and increase their tolerance to extreme environments (Zhang et al., 2018). For example, one study showed that the rumen microbiome produced more methane and volatile fatty acids to help Tibetan sheep (Ovis aries) and yak (Bos grunniens) adapt to the harsh high-altitude environment (Zhang et al., 2016). The Qinghai-Tibet *Plateau is* the highest plateau in the world, known as the “roof of the world” and the “third pole” (Xu et al., 2011). Environmental factors such as temperature and topography in the Plateau have had great influences on the structure and evolution of animal populations in the region, leading to the evolution of unique species or specific subgroups with different physiological and genetic adaptations to lower altitude species (Yang et al., 2009; Zeng et al., 2020; Zhang G. et al., 2021). Although some native mammals on the Tibetan Plateau may be well adapted to extreme climates how the structure and function of animal gut microbial communities contribute to adaptation of the host to extreme environments is not fully understood.
The plateau zokor (Myospalax baileyi) is a small underground rodent endemic to China, mainly inhabiting alpine meadows, alpine grasslands, scrub and farmland on the Tibetan Plateau and surrounding high altitude areas, living alone in sealed underground burrows at altitudes of 2,000–4,200 m (Su et al., 2015; Zhang et al., 2022). Due to the confines of underground burrows and the effects of seasonal plant dieback, plateau zokor feed primarily on the underground parts of roots, rhizomes and other weeds, even preferring some common poisonous weeds such as *Oxytropis kansuensis* and Stellera chamaejasme (Liu D. et al., 2021; Kang et al., 2022). They are the epitome of successful physiological adaptation in the face of the harsh environment characterized by low temperatures, low pressure, lack of oxygen at high-altitude and high concentrations of CO2 in caves, as well as complete darkness (Pu et al., 2019a; Kang et al., 2020). The adaptations of plateau zokors to low oxygen conditions in the Tibetan plateau have been partially studied (Xu et al., 2021; Zhang T. et al., 2021).
Grassland small mammals often span large geographical areas and utilize different food resources, so host gut microbiota may show marked differences between habitats. Many animals have microbial communities in their habitats that play a key role in host biology, which influence many aspects of host health and have the potential to adapt to harsh environmental (West et al., 2019; Liu et al., 2021). While 16S ribosomal RNA (16S rRNA) sequencing allowed reliable taxonomic resolution down to the species level, it did not provide information on functional characteristics. Metagenomic research goes far beyond traditional 16S rRNA microbiome sequencing, enabling not only more accurate species classification, but also in-depth bioinformatics analysis (Mack et al., 2020), which provides a unique opportunity to explore the composition and function of the gut microbiota in response to environmental adaptation of the host.
In this study, we selected plateau zokors (Myospalax baileyi) and plateau pikas (Ochotona curzoniae) from alpine meadow ecosystems on the Tibetan Plateau and Brandt’s vole (Lasiopodomys brandtii), Mongolian gerbil (Meriones unguiculatus) and Daurian ground squirrel (Spermophilus dauricus) from typical grassland ecosystems in the Inner Mongolian grasslands as our subjects. In evolutionary terms, unlike the other four host animal species, which all belong to the Rodentia, the plateau pika belongs to the Ochotonidae of the Lagomorpha. The plateau zokor belongs to the Spalacidae of the Rodentia, the Brandt’s vole and the Mongolian gerbil belong to the Circetidae of the Rodentia, and the Daurian ground squirrel belongs to the Sciuridae of the Rodentia. These species represent the best adapted ecological populations of small mammals inhabiting grassland ecosystems at different altitudes. The gut contents of these target animal species were then analyzed by metagenome sequencing to assess the characteristics of the gut microbiota and to explore the relationship between adaptations of the host to its living environment and the gut microbiota.
## Ethics statement
This study was conducted in accordance with the Guidelines for the Care and Use of Animals in Research published by the Institute of Zoology, Chinese Academy of Sciences. This study was reviewed and approved by the Animal Ethics Committee of the Institute of Zoology, Chinese Academy of Sciences (2019FY100300-03).
The animal study was reviewed and approved by this study was conducted in accordance with the Guidelines for the Care and Use of Animals in Research published by the Institute of Zoology, Chinese Academy of Sciences. This study was reviewed and approved by the Animal Ethics Committee of the Institute of Zoology, Chinese Academy of Sciences (2019FY100300-03).
## Sample selection and sampling location information
The samples were collected between 15 July and 15 August 2021. Wild plateau zokor (Mb) and plateau pika (Oc) were captured from Xunhua County (35°38′46.3″N 102°15′02.0″E), Qinghai Province, on the Qinghai-Tibet Plateau. During the same time period that we sampled the Tibetan Plateau, we collected the best adapted ecological populations of small mammals at three different lower elevations in the grasslands of Inner Mongolia. We captured samples of Brandt’s vole (Lb) at New Barag Right Banner (45°33′48.0″N 116°59′12.7″E), Mongolian gerbil (Mu) at East Ujimqin Banner (48°48′32.9″N 116°51′03.5″E), and then zaiTaibus Banner for Daur’s gopher (Sd) (41°44′38.9″N 115°03′51.8″E), respectively. The species and location information are shown in Figure 1; Table 1. Rope trapping method was used to trap plateau zokors, and then four other species were sampled using cage trapping. Captured Animal samples were euthanized with isoflurane and then taken back to the local laboratory for risk assessment. After confirming that there was no potential biosafety risk, the autopsy was performed and recorded information such as location, age, sex, and weight. The cecal contents were collected into 2 ml sterilized storage tubes, immediately stored in liquid nitrogen, immersed in dry ice during transport, and then stored in a laboratory freezer at –80°C. Then, cecum feces of three adult male individuals were selected for subsequent metagenomic sequencing analysis.
**Figure 1:** *Sampling sites and host animal species names.* TABLE_PLACEHOLDER:Table 1
## Extraction of genomic DNA
Total fecal genomic DNA was extracted from the cecum feces using the Fecal Genomic DNA Extraction Kit (TianGen) following the manufacturer’s protocol. DNA was analyzed for purity and integrity using agarose gel electrophoresis (AGE); DNA concentration was accurately quantified by Qubit 2.0. Then it was sent to Tianjin Nuohezhiyuan Bio-Information Technology Co., Ltd. for purification and sequencing.
## Library construction and on-machine sequencing
The total genomic DNA samples were randomly broken into fragments of approximately 350 bp in length using a Covaris ultrasonic breaker, and the libraries were prepared by end-repair, A-tailing, sequencing junction, purification and Polymerase Chain Reaction (PCR) amplification. After library construction, the library was initially quantified using Qubit 2.0 and diluted to 2 ng/μL, and then the insert size of the library was checked using Agilent 2,100. After the insert size meets expectations, use the Q-PCR method to determine the effective concentration of the library. Accurate quantification (library effective concentration > 3 nM) to ensure library quality. After ensuring that the quality of the library is qualified, the different libraries are pooled according to the requirements of effective concentration and target data volume, and then Illumina PE150 sequencing is performed.
## Data quality control
In order to ensure the accuracy and reliability of the results of the subsequent information analysis, the raw data should first be filtered by quality control and hosts to obtain Clean Data. Remove reads containing a certain percentage of low-quality bases (quality <=38) or more (default is set to 40 bp). Remove reads with Adapter removes reads containing more than a certain percentage (default 40 bp) of low-quality bases (mass < =38). Remove the reads whose N base reaches a certain proportion (the default is 10 bp). Remove the reads whose overlap with the Adapter exceeds a certain threshold (the default is 15 bp).
## Gene assembly and prediction
Metagenome assembly was performed from the Clean Data of each sample after quality control. ORF (Open Reading Frame) was performed from the scaftigs (≥500 bp) of single sample assembly using MetaGeneMark (Li et al., 2014; Oh et al., 2014; Qin et al., 2014), and from the prediction results, information with length less than 100 nt (Qin et al., 2010; Nielsen et al., 2014) was filtered out. The ORF prediction results of each sample assembly were de-redundant using CD-HIT software to obtain a non-redundant initial gene catalog, using Bowtie2 to compare the Clean Data of each sample to the initial gene catalog and to calculate the number of reads of the genes in each sample. *The* genes supporting reads <= 2 in each sample were filtered out to obtain the final gene catalog (Unigenes) for subsequent analysis, and from the number of reads and gene lengths in the comparison reads and gene length, the abundance of each gene in each sample was calculated (Cotillard et al., 2013; Villar et al., 2015).
## Species annotation
Using the gene catalog for comparison in the MicroNR library to obtain species annotation information for each gene (Unigene), and combined with the gene abundance tables to obtain species abundance tables for different taxonomic levels. *The* genes were compared with each functional database using the DIAMOND software (Buchfink et al., 2015). Unigenes were compared (blastp, evalue <= 1e-5) with Bacteria, Fungi, Archaea and Virus sequences extracted from the NCBI NR (Version: 2018.01) database (Karlsson et al., 2013).
## Metagenomics data analysis
Functional annotation and abundance analysis of metabolic pathways in the Kyoto Encyclopedia of Genes and Genomes (KEGG) database using gene catalog data. Based on the species abundance tables at different taxonomic levels, we performed Principal Components Analysis (PCA) and Non-metric Multidimensional Scaling (NMDS) analyses, where the more similar the species composition of the samples, the closer they were in the PCA and NMDS plots. Based on the Bray–Curtis distance, Principal Co-ordinates Analysis (PCoA) analysis was performed, and the principal coordinate combination with the largest contribution rate was selected for graph display.
## LEfSe analysis of “biomarkers” between groups
The featured microbial taxa between groups were screened using Linear discriminant analysis Effect Size (LEfSe) differential analysis. First, the rank sum test was used to detect the differential species among different groups, and Linear Discriminant Analysis (LDA) was used to achieve dimension reduction and evaluate the impact of the differential species, namely, LDA score (Segata et al., 2011) was obtained. Featured microbial taxa with LDA Score greater than the set value (LDA score = 4.0) was defined as biomarkers with statistical differences between groups.
## Metastat analysis of functional differences between groups
In order to investigate the functions that differed significantly between groups, the functional abundance data were analyzed using the Metastats method, starting from a table of relative abundance of functions at different levels. The p-values were first obtained by hypothesis testing, and the q-values were obtained by correcting the p-values. Finally, the functions with significant differences were screened according to the q-values and box plots of the abundance distribution of the different functions between groups were drawn.
## Annotation of resistance genes
The Comprehensive Antibiotic Resistance Database (CARD) is a new database of resistance genes that has emerged in recent years. *The* gene catalog is annotated with the CARD, which provides information on the distribution of resistance gene abundance and the species affiliation and resistance mechanisms of these resistance genes.
## Gut microbiota diversity in grassland small mammals
After quality control, a total of 4,874,600 genes were obtained from the gene catalog, with 1,721,464 complete genes (both start and stop codons), representing $35.31\%$ of the total number of genes. *For* gene richness characterization a sparse analysis was performed (Table 2). The estimated gene richness values were almost close to saturation in all groups, indicating that the sequencing data had sufficient coverage and that only a very small number of genes may have been undetected. In this study, a total of 233,591 unique genes were identified in the cecum feces of the plateau zokor (Supplementary Figure S1A). Furthermore, we analyzed the gut microbiota composition of plateau zokors under the phylum and genus levels. The top 10 microbial taxa in terms of maximum relative abundance in the gut microbiota of each sample are shown in Figure 2A, with Firmicutes and Bacteroidetes being the predominant phylum in all sites. Inter-group analyses revealed that the Firmicutes ($51.87\%$) and the Bacteroidetes ($11.94\%$) were found under phylum levels in the gut of plateau zokors, respectively (Supplementary Table S1). Figure 2B shows the composition of the gut microbiota under the genus level, with Clostridium ($5.57\%$), Prevotella ($3.54\%$), Ruminococcus ($2.92\%$), Bacteroides ($2.56\%$), Eubacterium ($2.36\%$,) Alistipes ($0.26\%$), Lactobacillus ($0.26\%$), Desulfovibrio ($0.21\%$), and Chlamydia ($0.13\%$) in the plateau zokors. The heat map of gene number and abundance clustering under the species level is shown in Supplementary Figure S2.
## Differences in the composition of gut microbial in plateau zokor
Specific bacterial flora may influence the adaptation of the plateau zokor to the living environment, and we continued to look for “biomarkers” with statistical differences. Firstly, by using the method of PCA, NMDS, and PCoA under the phylum level, the results showed significant differences in the gut microbial composition of the five groups (Figures 3A–C). The results of the interspecific relative thermogram analysis among the Mb, Oc, Mu, Lb, and Sd groups (Figure 3D) showed that the relative abundance of three species of bacteria was high in the gut microbial of plateau zokors, namely Pseudoflavonifractor sp., Marseille-P3106, Clostridium sp. CAG:590, and Cellulosilyticum lentocellum. In order to better explore the adaptation mechanisms of plateau zokors in the plateau region, Biomarker was used to screen for species with significant differences between groups using LEfSe analysis. In order to distinguish which biomarker with statistical differences are unique to plateau zokors and which may be common to plateau species, we did not directly choose the all five groups to LEfSe analysis (LDA score > 4.0), but instead took an indirect comparative approach, specifically using these two plateau species separately from the other three species (Lb, Mu, and Sd groups) for the differential analysis, selecting the featured microbial taxa that are present in both plateau species (Figure 3E; Supplementary Figure S3). The two plateau species (Mb and Oc groups) were then chosen to LEfSe analysis, from which the featured microbial taxa unique to plateau zokors could be obtained. We found the following bacteria were enriched in both plateau animal species, Clostridium sp. CAG 253, Clostridium sp. ASF502, Clostridium sp. CAG 230, Cellulosilyticum lentocellum, *Lachnospiraceae bacterium* AB2028, *Lachnospiraceae bacterium* NK4A179, Eubacterium sp. CAG 115, Pseudoflavonifractor genus, Flavonifractor genus, Candidatus saccharibacteria, Betaretrovirus genus. In addition to this, there are also significant abundance differences among Pseudoflavonifractor sp. Marseille-P3106, *Lachnospiraceae bacterium* XPB1003, *Lachnospiraceae bacterium* NC2008, viruses kingdom, and unclassified_viruses. LEfSe analysis (LDA score > 4.0) was then carried out for plateau zokors and plateau pikas and the distribution of LDA score for the differential species is shown that (Figure 3F) the Clostridium sp. CAG 127and the Pseudoflavonifractor sp. Marseille P3106 may be the unique Biomarker to plateau zokors.
**Figure 3:** *Analysis of gut microbiota with significant differences. (A) PCA analysis among the Mb, Oc, Mu, Lb, and Sd groups under phyla level. The horizontal coordinates indicate one principal component, vertical coordinates indicate the second principal component, percentages indicate the contribution of the principal component to sample variation, each point in the figure represents a sample, and the samples are represented by the same color. (B) NMDS analysis among the Mb, Oc, Mu, Lb, and Sd groups under phyla level (Stress = 0.119). Each point in the graph represents a sample, and the distance between points represents the degree of difference. When the Stress is less than 0.2, it indicates that the NMDS analysis has certain reliability. (C) PCoA analysis among the Mb, Oc, Mu, Lb, and Sd groups under phyla level The horizontal coordinate indicates one principal component, the vertical coordinate indicates another principal component, and the percentage indicates the contribution of the principal component to the variance of the samples. (D) Relative abundances of bacterial species among the Mb, Oc, Mu, Lb, and Sd groups. (E) The featured microbial taxa of each animal species according to the results of LEfSe differential analysis. The bar chart of the distribution of LDA score shows microbial taxa with an LDA Score greater than a set value (LDA score > 4.0), i.e., Biomarkers that are statistically different between the Mb, Mu, Lb, and Sd groups, and the length of the bar chart represents the size of the effect of the differential species. (F) The featured microbial taxa of each animal species according to the results of LEfSe differential analysis. The differential analysis was conducted between the Mb and Oc groups (LDA score > 4.0). (G) Relative abundances of bacterial taxa (Biomarker with statistical differences) varied across the Mb, Mu, Lb, and Sd samples according to the results of LEfSe (LDA score > 4.0).*
In order to investigate the abundance of statistically different Biomarkers among different groups, p-values were obtained by hypothesis testing using the Metastats method, and q-values were obtained by correcting the p-values. As shown in Figure 3G, the featured microbial taxa of each animal species were Flavonifractor, Pseudoflavonifractor, Cellulosilyticum, and Betaretrovirus.
## The metabolic function of gut microbial facilitates adaptation of the plateau zokor to the plateau environment
In order to better understand the functional differences in the gut microbiota of plateau zokors, DIAMOND software was used to compare Unigenes with various functional databases, and the relative abundance of different functional levels was calculated from the comparison results. After *Metastats analysis* using the KEGG, it was found that the metabolic functions of the samples from different geographic regions were statistically significantly different. We found that the Organismal Systems: Sensory System metabolic pathway has obvious abundance enrichment in plateau zokors (Supplementary Figure S4). The difference analysis of K05685 pathway (Figure 4A) and E.C found that 3.4.14.10 tripeptidyl peptidase II (TPPII) (Figures 4B,D) was also significantly enriched in plateau zokor. In addition, we performed LEfSe tests to detect KEGG pathways based on significantly different abundances between groups. According to the results of the LEFSe test, K07079, K03406, K02027, K10117 were significantly enriched in both plateau zokors and plateau pikas (Figure 4C).
**Figure 4:** *Functional metagenomic comparison of the gut microbiota in different groups. (A) KO terms with significant differences between groups. The horizontal axis is the group name and the vertical axis is the relative abundance of the corresponding species. Pairwise statistical analysis was done by Metastats. * and ** denote q-value < 0.05 and q-value<0.01, respectively. (B) EC Number with significant differences between groups according to the results of Pairwise statistical analysis by Metastats. (C) Heatmap of KEGG ortholog pathways showing different enrichments according to the results of LEfSe differential analysis (LDA score > 3.0). (D) Heatmap of KEGG EC Number showing different enrichments according to the results of Pairwise statistical analysis by Metastats.*
## Resistance gene analysis
The CARD resistance gene database was used to align the gene sequences to annotate the resistance genes. The core component of the database is the Antibiotic Resistance Ontology (ARO), which integrates information on sequences, antibiotic resistance, mechanisms of action, and associations between AROs, and provides an online interface between AROs and databases such as PDB and NCBI. The results showed that the genetic resistance of the animal host in the plateau region differed significantly from those in other grassland regions (Supplementary Figure S5C), and the types and absolute abundance of resistance genes were lower than those in other regions (Supplementary Figures S5A,B). The resistance genes of plateau zokors and plateau pika were mainly APH6-Ic, vanF, MexS, vanTN and LRA-13 (Supplementary Figure S5D).
## Discussion
The plateau zokor, a typical subterranean rodent inhabiting the Tibetan plateau, has to cope with the complex environment of high humidity, limited oxygen, high CO2 concentration, low temperature and food scarcity (Shao et al., 2015; Su et al., 2015), and there have been studies on its adaptive evolution (Pu et al., 2019b; Zhang T. et al., 2021). However, compared with the study on the adaptation of the plateau pika, less study has been done on plateau zokor in terms of the unique lifestyle (Li et al., 2018). Considering the similarity of gut microbiota among sympatric species, we first chose plateau pika as a sympatric control species when conducting the study. Then the animal host species that is most suited to the local grassland environment was selected as a control species in other altitudinal regions, so that it was easier to find gut microbiota with the ability to significantly facilitate the adaptation of the host to the local environment. Therefore, based on the results of field sampling, we selected Mongolian gerbil, and Daurian ground squirrel from other regions as control species.
The first problem that plateau zokors have to overcome when faced with the harsh living conditions is energy intake, as the low oxygen levels and extreme cold weather at high altitudes require more energy intake to maintain the animal’s body temperature (Qiu et al., 2012; Wang et al., 2021; Du et al., 2022).
In this study, we discovered that the abundance of unigenes in the gut microbiota of plateau zokors and plateau pikas was higher than that of the other three grassland small mammals, and that a more varied gut microbiota might more effectively control energy metabolism (Wang et al., 2020). For example, the gut microbiota of brown bear with higher diversity may regulate energy metabolism and promote fat storage, whereas less diverse gut flora may slow host metabolism (Sommer et al., 2016). Compared with low-altitude mammals, the relative abundance of Firmicutes and Bacteroidetes in the gut microbiota plateau zokor and plateau pika of high-altitude was higher. For example, in this study, a large amount of *Cellulosilyticum lentocellum* was enriched in plateau zokor and plateau pika in the Qinghai-Tibet Plateau, which has cellulolytic properties and can hydrolyze cellulose and xylan (van der Wielen et al., 2002; Cai and Dong, 2010; Baba et al., 2019). We hypothesize that this may be related to the fact that microbial communities in plateau species may have a higher ability to utilize high fiber forage to help them meet their energy requirements in cold and high altitude habitats, which may help the host maintain gut homeostasis, energy homeostasis, and core body temperature in harsh environments (Liu et al., 2021). The concept that hypoxia induces inflammation has been generally accepted from studies of hypoxic signaling pathways (Eltzschig and Carmeliet, 2011), so plateau zokors also have to cope with the inflammatory response caused by low pressure, hypoxia and high concentrations of CO2 in the burrow. Vascular leakage, accumulation of inflammatory cells in multiple organs and elevated serum cytokine levels have been shown to occur in humans and mice following short-term exposure to low oxygen concentrations (Thompson et al., 2004; Eltzschig et al., 2005). High concentrations of CO2 in the environment can also cause tissue damage in the body and produce an inflammatory response (Thom et al., 2017; da Costa and Val, 2020). Plant-based fiber intake has been shown to increase microbiota diversity and reduce markers of inflammation (Ma et al., 2021; Wastyk et al., 2021). Additionally, gut microbes produce metabolites such as short-chain fatty acids (SCFAs), volatile fatty acids (VFA), essential amino acids and vitamins through their collective metabolic activities, which contribute to host to evolve adaptations (Nicholson et al., 2012; Mukherjee et al., 2020). For example, butyrate has been shown to have several beneficial effects, including being an excellent nutrient for epithelial cells, and having immunomodulatory and anti-inflammatory properties (Meehan and Beiko, 2014; Berger et al., 2021). Current studies have shown that SCFAs inhibit inflammation mainly by inhibiting the NF-κB pathway and/or histone deacetylase function (HDACi), thereby downregulating pro-inflammatory cytokines (Zhang et al., 2019). In this study, both plateau species were found to be enriched with a large number of SCFAs-producing strains. For example, both plateau zokors and plateau pikas were enriched with Lachnospiraceae. In addition, plateau zokors were enriched with *Lachnospiraceae bacterium* XPB1003 and *Lachnospiraceae bacterium* NC2008 alone. Furthermore, the enrichment of Clostridium spp. in plateau zokors may not only enable these hosts to obtain more energy from their food (Ma et al., 2019), but also convert dietary fiber into SCFAs such as butyric acid (Schwiertz et al., 2010; West et al., 2019). Flavonifractor spp. can also produce SCFAs (Chen et al., 2020; Gao et al., 2021). Studies have shown that *Flavonifractor plautii* can exhibit lower levels of inflammation and that the active component of FP’s lipoteichoic acid mediates strong inhibition of interleukin (IL)-17 signaling (Mikami et al., 2020). Eubacterium spp. provides butyrate-mediated protection and is considered a new generation of “potentially beneficial microorganisms” whose presence in the gut is largely associated with increased dietary fiber intake (Kanauchi et al., 2006; Duncan et al., 2007; Vermeiren et al., 2012; Geirnaert et al., 2017). Pseudoflavonifractor is also a bacterium that can produce butyrate (Kläring et al., 2013; Sakamoto et al., 2018). At present, there are few studies on Candidatus saccharibacteria, which are directly involved in the degradation of hydrocarbons (Figueroa-Gonzalez et al., 2020; Nie et al., 2022). Similar phenomena have also been found in other plateau species. For example, a study used a multi-omics approach to examine fecal samples from high-and low-altitude humans and pigs, and direct evidence of consistent results linking genes to metabolites suggests that gut microbiota from high-altitude Tibetan pigs may produce more short-or long-chain fatty acids (Zeng et al., 2020). These evidences indicate that the plateau zokor has a large number of bacteria that decompose dietary fiber and produce SCFAs, which may facilitate the adaptation of plateau zokor to the plateau environment.
In this study, we used KEGG database to predict the function of *Metagenomic data* of plateau zokors. Our results suggest that the estimated gene functional profile of the microbiome is significantly influenced by altitude. Some metabolic pathways were significantly enriched in both plateau zokors and plateau pikas in the Qinghai-Tibet Plateau region. Most strikingly, those genes involved in Organismal Systems: Sensory system in the KEGG level 2 pathway were enhanced in plateau zokors. This might be due to plateau zokors’ protracted burrowing habits, which maximize their capacity to assist them in perceiving environmental changes in the face of degraded vision. K05685, K10117, and K02027 are all involved in catalytic carbohydrate uptake and directly involved in ATP production (Schneider, 2001; Webb Alexander et al., 2008; Modali and Zgurskaya, 2011; Cerisy et al., 2019). K03406 is involved in bacterial chemotaxis, which is essential for host colonization and virulence of many pathogenic bacteria causing human, animal and plant diseases (Wadhams and Armitage, 2004; Rosenberg et al., 2007; Salah Ud-Din and Roujeinikova, 2017). The function of the K07079 signaling pathway is unknown, and it may also play a role in promoting host adaptation, which needs to be further verified by experiments.
In high-altitude environments, low oxygen and high UV radiation may lead to DNA and protein damage, while genes associated with replication and repair may help to reduce damage to biomolecules (Dosek et al., 2007; Sinha et al., 2010). In addition, the harsh environmental stress on the Tibetan plateau can also put the organism in a state of oxidative stress (Cui et al., 2016). As altitude increases in mountainous areas, the production of reactive oxygen species (ROS) accelerates, which may lead to severe oxidative tissue damage, capable of damaging proteins, nucleic acids, polysaccharides and lipids, thereby inducing apoptosis (Strapazzon et al., 2016; Debevec et al., 2017; Mrakic-Sposta et al., 2022). More importantly, our KEGG analysis results show that 3.4.14.10 tripeptidyl peptidase II (TPPII) is significantly enriched in the intestine of plateau zokors. TPPII has demonstrated independent enzymatic activity involved in a wide range of activities, including antigen processing, apoptosis, DNA damage repair and cell division (Rockel et al., 2012; Tan et al., 2016). Studies have shown that in several malignant cell lines, TPPII translocates into the nucleus after γ-irradiation and ROS production and is involved in DNA repair (Preta et al., 2009, 2010). Involvement in antigen processing is probably the most studied aspect of the potential physiological role of TPP II (Tomkinson, 2019). TPPII is involved in cancer and antigen processing by MHC-I presentation and the presented antigens can be detected by CD8+ T cells, a process that is essential for the detection and destruction of cancer cells or cells infected by viruses (Seifert et al., 2003; Diekmann et al., 2009; Embgenbroich and Burgdorf, 2018). Although there is still controversy over antigen processing, TPPII has been shown to be associated with immunodeficiency, autoimmunity and neurodevelopmental retardation (Stepensky et al., 2015), implying that TPPII may be crucial to the immune system of plateau zokors. Therefore, we hypothesize that TPPII may contribute to the adaptation of plateau zokors to survive in confined burrows at high altitudes. However, our results are based on predicted metagenomics only and may not accurately reflect how TPPII functions in plateau zokors. Further studies should be conducted for experimental validation to explore the role of TPPII in the environmental adaptation of plateau zokors.
In addition, the complex topography and physical barriers of the Tibetan plateau not only significantly reduce the dispersal of organisms, but may affect patterns of gene flow, which could ultimately affect the current spatial distribution of endemic plateau species and their genetic diversity (Hewitt, 2000; Toju, 2008). Compared with some ground-moving rodents, the plateau zokor has a lower activity time and frequency (Kang et al., 2020). The unique geographical location and habits of plateau zokors also result in a low probability of anthropogenic and pathogenic infestation, with the potential for a large number of unknown pathogenic microorganisms within the population (Cao et al., 2014; Zhao et al., 2014). As the largest population of mammals, rodents transmit a variety of infectious zoonotic diseases to humans (Morand et al., 2015). Our results suggest that plateau zokors have a large number of unknown viruses (Supplementary Figure S1B), and maybe a potential source of zoonotic transmission, increasing the risk of pathogenic spillover. Furthermore, resistance genes are widespread in the environment and increased antibiotic consumption directly leads to environmental pollution by antibiotics, which can threaten the balance of ecosystems and human health (Polianciuc et al., 2020). Although the abundance of resistance genes in plateau zokors was lower and the species varied significantly compared to other animal species, it also indicated that resistance genes are prevalent in the Tibetan plateau region. Therefore, it is necessary to increase active surveillance and reverse pathogenic research on special wildlife in the Tibetan Plateau region, to explore the pathogenic mechanisms of newly present pathogens, assess the risk of disease outbreaks and to provide early warning.
## Conclusion
Different geographical environments have significant effects on the composition of mammalian gut microbiota, while at the same time the gut microbiota can facilitate adaptation of a variety of small mammals to different geographical environments. Our study demonstrates that the gut microbiota of plateau zokors is highly enriched with the phylum Firmicutes and Bacteroidetes, among which *Cellulosilyticum lentocellum* and other bacteria play a role in the breakdown of dietary fiber. This, along with the enrichment of some signaling pathways related to energy uptake, ensure the energy supplementation of the plateau zokor in a low oxygen and low-pressure environment. The gut microbiota enriched with Lachnospiraceae, Pseudoflavonifractor, Eubacterium, Flavonifractor, Clostridium, and other flora, which produce SCFAs, may help plateau zokors to reduce the inflammatory response caused by low pressure, hypoxia, and high concentrations of CO2 in the burrow as they adapt to the high-altitude environment. The gut microbial metabolite TPPII may help plateau zokors adapt to the tissue damage and immune response caused by the high-altitude confined burrow environment. These unique gut microbiota and metabolites together contribute to the adaptation of plateau zokors to the plateau environment.
## Data availability statement
The data presented in the study are deposited in the National Genomics Data Center (NGDC), accession number PRJCA014883.
## Author contributions
HH and BH contributed to the conception of the study. JW, YL, JP, GL, YoH, and HZ contributed significantly to collect samples and perform the experiment. JG, YX, and SH contributed significantly to analysis and manuscript preparation. BW and YaH helped to perform the analysis with constructive discussions. BH, YL, and JW performed the data analyses and wrote the manuscript. All authors contributed to the article and approved the submitted version.
## Funding
This study was supported by the National Key Research and Development Program of China (No. 2022YFC2601602), the Major Program of National Natural Science Foundation of China (No. 32090023), and National Forestry and Grassland Administration, China.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb.2023.1136845/full#supplementary-material
## References
1. Amato K. R., Leigh S. R., Kent A., Mackie R. I., Yeoman C. J., Stumpf R. M.. **The role of gut microbes in satisfying the nutritional demands of adult and juvenile wild, black howler monkeys (**. *Am. J. Phys. Anthropol.* (2014) **155** 652-664. DOI: 10.1002/ajpa.22621
2. Baba Y., Matsuki Y., Takizawa S., Suyama Y., Tada C., Fukuda Y.. **Pretreatment of lignocellulosic biomass with cattle rumen fluid for methane production: fate of added rumen microbes and indigenous microbes of methane seed sludge**. *Microbes Environ.* (2019) **34** 421-428. DOI: 10.1264/jsme2.ME19113
3. Berger K., Burleigh S., Lindahl M., Bhattacharya A., Patil P., Stålbrand H.. **Xylooligosaccharides increase Bifidobacteria and Lachnospiraceae in mice on a high-fat diet, with a concomitant increase in short-chain fatty acids, especially butyric acid**. *J. Agric. Food Chem.* (2021) **69** 3617-3625. DOI: 10.1021/acs.jafc.0c06279
4. Buchfink B., Xie C., Huson D. H.. **Fast and sensitive protein alignment using DIAMOND**. *Nat. Methods* (2015) **12** 59-60. DOI: 10.1038/nmeth.3176
5. Cai S., Dong X.. *Int. J. Syst. Evol. Microbiol.* (2010) **60** 845-849. DOI: 10.1099/ijs.0.014712-0
6. Cao Y.-F., Nie X.-H., Zhang T.-Z., Du S.-Y., Duszynski D. W., Bian J.-H.. **Four new coccidia (Apicomplexa: Eimeriidae) from the plateau zokor,**. *China Syst. Parasitol.* (2014) **87** 181-186. DOI: 10.1007/s11230-013-9466-z
7. Cerisy T., Iglesias A., Rostain W., Boutard M., Pelle C., Perret A.. **ABC transporters required for hexose uptake by**. *J. Bacteriol.* (2019) **201** e00241-19. DOI: 10.1128/jb.00241-19
8. Chen H., Zhang F., Zhang J., Zhang X., Guo Y., Yao Q.. **A holistic view of Berberine inhibiting intestinal carcinogenesis in conventional mice based on microbiome-metabolomics analysis**. *Front. Immunol.* (2020) **11** 588079. DOI: 10.3389/fimmu.2020.588079
9. Cotillard A., Kennedy S. P., Kong L. C., Prifti E., Pons N., Le Chatelier E.. **Dietary intervention impact on gut microbial gene richness**. *Nature* (2013) **500** 585-588. DOI: 10.1038/nature12480
10. Cui G., Wei X., Degen A. A., Wei X., Zhou J., Ding L.. **Trolox-equivalent antioxidant capacity and composition of five alpine plant species growing at different elevations on the Qinghai–Tibetan plateau**. *Plant Ecol. Divers.* (2016) **9** 387-396. DOI: 10.1080/17550874.2016.1261952
11. da Costa J. C., Val A. L.. **Extreme climate scenario and parasitism affect the Amazonian fish**. *Sci. Total Environ.* (2020) **726** 138628. DOI: 10.1016/j.scitotenv.2020.138628
12. Debevec T., Millet G. P., Pialoux V.. **Hypoxia-induced oxidative stress modulation with physical activity**. *Front. Physiol.* (2017) **8** 84. DOI: 10.3389/fphys.2017.00084
13. Diekmann J., Adamopoulou E., Beck O., Rauser G., Lurati S., Tenzer S.. **Processing of two latent membrane protein 1 MHC class I epitopes requires tripeptidyl peptidase II involvement**. *J. Immunol.* (2009) **183** 1587-1597. DOI: 10.4049/jimmunol.0803441
14. Dosek A., Ohno H., Acs Z., Taylor A. W., Radak Z.. **High altitude and oxidative stress**. *Respir. Physiol. Neurobiol.* (2007) **158** 128-131. DOI: 10.1016/j.resp.2007.03.013
15. Du X., Li F., Kong F., Cui Z., Li D., Wang Y.. **Altitude-adaption of gut microbiota in Tibetan chicken**. *Poult. Sci.* (2022) **101** 101998. DOI: 10.1016/j.psj.2022.101998
16. Duncan S. H., Belenguer A., Holtrop G., Johnstone A. M., Flint H. J., Lobley G. E.. **Reduced dietary intake of carbohydrates by obese subjects results in decreased concentrations of butyrate and butyrate-producing bacteria in feces**. *Appl. Environ. Microbiol.* (2007) **73** 1073-1078. DOI: 10.1128/aem.02340-06
17. Eltzschig H. K., Abdulla P., Hoffman E., Hamilton K. E., Daniels D., Schönfeld C.. **HIF-1-dependent repression of equilibrative nucleoside transporter (ENT) in hypoxia**. *J. Exp. Med.* (2005) **202** 1493-1505. DOI: 10.1084/jem.20050177
18. Eltzschig H. K., Carmeliet P.. **Hypoxia and inflammation**. *N. Engl. J. Med.* (2011) **364** 656-665. DOI: 10.1056/NEJMra0910283
19. Embgenbroich M., Burgdorf S.. **Current concepts of antigen cross-presentation**. *Front. Immunol.* (2018) **9** 1643. DOI: 10.3389/fimmu.2018.01643
20. Figueroa-Gonzalez P. A., Bornemann T. L. V., Adam P. S., Plewka J., Révész F., von Hagen C. A.. **Saccharibacteria as organic carbon sinks in hydrocarbon-fueled communities**. *Front. Microbiol.* (2020) **11** 587782. DOI: 10.3389/fmicb.2020.587782
21. Gao J., Guo X., Wei W., Li R., Hu K., Liu X.. **The Association of Fried Meat Consumption with the gut microbiota and fecal metabolites and its impact on glucose homoeostasis, intestinal endotoxin levels, and systemic inflammation: a randomized controlled-feeding trial**. *Diabetes Care* (2021) **44** 1970-1979. DOI: 10.2337/dc21-0099
22. Geirnaert A., Calatayud M., Grootaert C., Laukens D., Devriese S., Smagghe G.. **Butyrate-producing bacteria supplemented in vitro to Crohn’s disease patient microbiota increased butyrate production and enhanced intestinal epithelial barrier integrity**. *Sci. Rep.* (2017) **7** 11450. DOI: 10.1038/s41598-017-11734-8
23. Hewitt G.. **The genetic legacy of the quaternary ice ages**. *Nature* (2000) **405** 907-913. DOI: 10.1038/35016000
24. Kanauchi O., Fukuda M., Matsumoto Y., Ishii S., Ozawa T., Shimizu M.. *World J. Gastroenterol.* (2006) **12** 1071-1077. DOI: 10.3748/wjg.v12.i7.1071
25. Kang Y., Liu Q., Yao B., Hou Q., Su J.. **Pleistocene climate-driven diversification of plateau zokor (Eospalax baileyi) in the eastern Qinghai-Tibet plateau**. *Ecol. Indic.* (2022) **139** 108923. DOI: 10.1016/j.ecolind.2022.108923
26. Kang Y., Su J., Yao B., Ji W., Hegab I. M., Hanafy A. M.. **Geometric morphometric analysis of the plateau zokor (**. *Zoology* (2020) **140** 125779. DOI: 10.1016/j.zool.2020.125779
27. Karlsson F. H., Tremaroli V., Nookaew I., Bergström G., Behre C. J., Fagerberg B.. **Gut metagenome in European women with normal, impaired and diabetic glucose control**. *Nature* (2013) **498** 99-103. DOI: 10.1038/nature12198
28. Kläring K., Hanske L., Bui N., Charrier C., Blaut M., Haller D.. *Int. J. Syst. Evol. Microbiol.* (2013) **63** 4606-4612. DOI: 10.1099/ijs.0.051441-0
29. Li J., Jia H., Cai X., Zhong H., Feng Q., Sunagawa S.. **An integrated catalog of reference genes in the human gut microbiome**. *Nat. Biotechnol.* (2014) **32** 834-841. DOI: 10.1038/nbt.2942
30. Li H., Qu J., Li T., Wirth S., Zhang Y., Zhao X.. **Diet simplification selects for high gut microbial diversity and strong fermenting ability in high-altitude pikas**. *Appl. Microbiol. Biotechnol.* (2018) **102** 6739-6751. DOI: 10.1007/s00253-018-9097-z
31. Liu D., Song P., Yan J., Wang H., Cai Z., Xie J.. **Gut microbiome changes in captive plateau Zokors (**. *Evol. Bioinform. Online* (2021) **17** 1176934321996353. DOI: 10.1177/1176934321996353
32. Liu K., Yang J., Yuan H.. **Recent progress in research on the gut microbiota and highland adaptation on the Qinghai-Tibet plateau**. *J. Evol. Biol.* (2021) **34** 1514-1530. DOI: 10.1111/jeb.13924
33. Łukasik P., van Asch M., Guo H., Ferrari J., Godfray H. C.. **Unrelated facultative endosymbionts protect aphids against a fungal pathogen**. *Ecol. Lett.* (2013) **16** 214-218. DOI: 10.1111/ele.12031
34. Ma Y., Ma S., Chang L., Wang H., Ga Q., Ma L.. **Gut microbiota adaptation to high altitude in indigenous animals**. *Biochem. Biophys. Res. Commun.* (2019) **516** 120-126. DOI: 10.1016/j.bbrc.2019.05.085
35. Ma W., Nguyen L. H., Song M., Wang D. D., Franzosa E. A., Cao Y.. **Dietary fiber intake, the gut microbiome, and chronic systemic inflammation in a cohort of adult men**. *Genome Med.* (2021) **13** 102. DOI: 10.1186/s13073-021-00921-y
36. Mack A., Bobardt J. S., Haß A., Nichols K. B., Schmid R. M., Stein-Thoeringer C. K.. **Changes in gut microbial metagenomic pathways associated with clinical outcomes after the elimination of malabsorbed sugars in an IBS cohort**. *Gut Microbes* (2020) **11** 620-631. DOI: 10.1080/19490976.2019.1686322
37. Meehan C. J., Beiko R. G.. **A phylogenomic view of ecological specialization in the Lachnospiraceae, a family of digestive tract-associated bacteria**. *Genome Biol. Evol.* (2014) **6** 703-713. DOI: 10.1093/gbe/evu050
38. Mikami A., Ogita T., Namai F., Shigemori S., Sato T., Shimosato T.. **Oral administration of**. *Front. Nutr.* (2020) **7** 610946. DOI: 10.3389/fnut.2020.610946
39. Modali S. D., Zgurskaya H. I.. **The periplasmic membrane proximal domain of mac a acts as a switch in stimulation of ATP hydrolysis by mac B transporter**. *Mol. Microbiol.* (2011) **81** 937-951. DOI: 10.1111/j.1365-2958.2011.07744.x
40. Morand S., Jittapalapong S., Kosoy M.. **Rodents as hosts of infectious diseases: biological and ecological characteristics**. *Vector Borne Zoonotic Dis.* (2015) **15** 1-2. DOI: 10.1089/vbz.2015.15.1.intro
41. Mrakic-Sposta S., Montorsi M., Porcelli S., Marzorati M., Healey B., Dellanoce C.. **Effects of prolonged exposure to hypobaric hypoxia on oxidative stress: overwintering in Antarctic Concordia Station**. *Oxidative Med. Cell. Longev.* (2022) **2022** 4430032-4430014. DOI: 10.1155/2022/4430032
42. Mukherjee A., Lordan C., Ross R. P., Cotter P. D.. **Gut microbes from the phylogenetically diverse genus**. *Gut Microbes* (2020) **12** 1802866. DOI: 10.1080/19490976.2020.1802866
43. Nicholson J. K., Holmes E., Kinross J., Burcelin R., Gibson G., Jia W.. **Host-gut microbiota metabolic interactions**. *Science* (2012) **336** 1262-1267. DOI: 10.1126/science.1223813
44. Nie J., Utter D. R., Kerns K. A., Lamont E. I., Hendrickson E. L., Liu J.. **Strain-level variation and diverse host bacterial responses in Episymbiotic Saccharibacteria**. *mSystems* (2022) **7** e0148821. DOI: 10.1128/msystems.01488-21
45. Nielsen H. B., Almeida M., Juncker A. S., Rasmussen S., Li J., Sunagawa S.. **Identification and assembly of genomes and genetic elements in complex metagenomic samples without using reference genomes**. *Nat. Biotechnol.* (2014) **32** 822-828. DOI: 10.1038/nbt.2939
46. Oh J., Byrd A. L., Deming C., Conlan S., Kong H. H., Segre J. A.. **Biogeography and individuality shape function in the human skin metagenome**. *Nature* (2014) **514** 59-64. DOI: 10.1038/nature13786
47. Polianciuc S. I., Gurzău A. E., Kiss B., Ştefan M. G., Loghin F.. **Antibiotics in the environment: causes and consequences**. *Med. Pharm. Rep.* (2020) **93** 231-240. DOI: 10.15386/mpr-1742
48. Preta G., de Klark R., Chakraborti S., Glas R.. **MAP kinase-signaling controls nuclear translocation of tripeptidyl-peptidase II in response to DNA damage and oxidative stress**. *Biochem. Biophys. Res. Commun.* (2010) **399** 324-330. DOI: 10.1016/j.bbrc.2010.06.133
49. Preta G., de Klark R., Glas R.. **A role for nuclear translocation of tripeptidyl-peptidase II in reactive oxygen species-dependent DNA damage responses**. *Biochem. Biophys. Res. Commun.* (2009) **389** 575-579. DOI: 10.1016/j.bbrc.2009.09.021
50. Pu P., Lu S., Niu Z., Zhang T., Zhao Y., Yang X.. **Oxygenation properties and underlying molecular mechanisms of hemoglobins in plateau zokor ()**. *Am. J. Physiol. Regul. Integr. Comp. Physiol.* (2019a) **317** R696-R708. DOI: 10.1152/ajpregu.00335.2018
51. Pu P., Lu S., Niu Z., Zhang T., Zhao Y., Yang X.. **Oxygenation properties and underlying molecular mechanisms of hemoglobins in plateau zokor (**. *Am. J. Physiol. Regul. Integr. Comp. Physiol.* (2019b) **317** R696-R708. DOI: 10.1152/ajpregu.00335.2018
52. Qin J., Li R., Raes J., Arumugam M., Burgdorf K. S., Manichanh C.. **A human gut microbial gene catalogue established by metagenomic sequencing**. *Nature* (2010) **464** 59-65. DOI: 10.1038/nature08821
53. Qin N., Yang F., Li A., Prifti E., Chen Y., Shao L.. **Alterations of the human gut microbiome in liver cirrhosis**. *Nature* (2014) **513** 59-64. DOI: 10.1038/nature13568
54. Qiu Q., Zhang G., Ma T., Qian W., Wang J., Ye Z.. **The yak genome and adaptation to life at high altitude**. *Nat. Genet.* (2012) **44** 946-949. DOI: 10.1038/ng.2343
55. Rockel B., Kopec K. O., Lupas A. N., Baumeister W.. **Structure and function of tripeptidyl peptidase II, a giant cytosolic protease**. *Biochim. Biophys. Acta* (2012) **1824** 237-245. DOI: 10.1016/j.bbapap.2011.07.002
56. Rosenberg E., Koren O., Reshef L., Efrony R., Zilber-Rosenberg I.. **The role of microorganisms in coral health, disease and evolution**. *Nat. Rev. Microbiol.* (2007) **5** 355-362. DOI: 10.1038/nrmicro1635
57. Sakamoto M., Iino T., Yuki M., Ohkuma M.. *Int. J. Syst. Evol. Microbiol.* (2018) **68** 2074-2081. DOI: 10.1099/ijsem.0.002800
58. Salah Ud-Din A. I. M., Roujeinikova A.. **Methyl-accepting chemotaxis proteins: a core sensing element in prokaryotes and archaea**. *Cell. Mol. Life Sci.* (2017) **74** 3293-3303. DOI: 10.1007/s00018-017-2514-0
59. Schneider E.. **ABC transporters catalyzing carbohydrate uptake**. *Res. Microbiol.* (2001) **152** 303-310. DOI: 10.1016/S0923-2508(01)01201-3
60. Schwiertz A., Taras D., Schäfer K., Beijer S., Bos N. A., Donus C.. **Microbiota and SCFA in lean and overweight healthy subjects**. *Obesity (Silver Spring)* (2010) **18** 190-195. DOI: 10.1038/oby.2009.167
61. Segata N., Izard J., Waldron L., Gevers D., Miropolsky L., Garrett W. S.. **Metagenomic biomarker discovery and explanation**. *Genome Biol.* (2011) **12** R60. DOI: 10.1186/gb-2011-12-6-r60
62. Seifert U., Marañón C., Shmueli A., Desoutter J.-F., Wesoloski L., Janek K.. **An essential role for tripeptidyl peptidase in the generation of an MHC class I epitope**. *Nat. Immunol.* (2003) **4** 375-379. DOI: 10.1038/ni905
63. Shao Y., Li J.-X., Ge R.-L., Zhong L., Irwin D. M., Murphy R. W.. **Genetic adaptations of the plateau zokor in high-elevation burrows**. *Sci. Rep.* (2015) **5** 1-11. DOI: 10.1038/srep17262
64. Sinha S., Dutta A., Singh S. N., Ray U. S.. **Protein nitration, lipid peroxidation and DNA damage at high altitude in acclimatized lowlanders and native highlanders: relation with oxygen consumption**. *Respir. Physiol. Neurobiol.* (2010) **171** 115-121. DOI: 10.1016/j.resp.2010.03.002
65. Sommer F., Ståhlman M., Ilkayeva O., Arnemo J. M., Kindberg J., Josefsson J.. **The gut microbiota modulates energy metabolism in the hibernating brown bear**. *Cell Rep.* (2016) **14** 1655-1661. DOI: 10.1016/j.celrep.2016.01.026
66. Steele M. I., Motta E. V. S., Gattu T., Martinez D., Moran N. A.. **The gut microbiota protects bees from invasion by a bacterial pathogen**. *Microbiol. Spectr.* (2021) **9** e0039421. DOI: 10.1128/Spectrum.00394-21
67. Stepensky P., Rensing-Ehl A., Gather R., Revel-Vilk S., Fischer U., Nabhani S.. **Early-onset evans syndrome, immunodeficiency, and premature immunosenescence associated with tripeptidyl-peptidase II deficiency**. *Blood* (2015) **125** 753-761. DOI: 10.1182/blood-2014-08-593202
68. Strapazzon G., Malacrida S., Vezzoli A., Dal Cappello T., Falla M., Lochner P.. **Oxidative stress response to acute hypobaric hypoxia and its association with indirect measurement of increased intracranial pressure: a field study**. *Sci. Rep.* (2016) **6** 32426. DOI: 10.1038/srep32426
69. Su J., Aryal A., Nan Z., Ji W.. **Climate change-induced range expansion of a subterranean rodent: implications for rangeland management in Qinghai-Tibetan plateau**. *PLoS One* (2015) **10** e0138969. DOI: 10.1371/journal.pone.0138969
70. Tan Q., Ma S., Hu J., Chen X., Yu Y., Zang G.. **Role of tripeptidyl peptidase II in MHC class I antigen presentation: biological characteristics, cellular crosstalk and signaling**. *Biomed. Pharmacother.* (2016) **84** 1954-1958. DOI: 10.1016/j.biopha.2016.11.004
71. Thom S. R., Bhopale V. M., Hu J., Yang M.. **Inflammatory responses to acute elevations of carbon dioxide in mice**. *J. Appl. Physiol. (1985)* (2017) **123** 297-302. DOI: 10.1152/japplphysiol.00343.2017
72. Thompson L. F., Eltzschig H. K., Ibla J. C., Van De Wiele C. J., Resta R., Morote-Garcia J. C.. **Crucial role for ecto-5′-nucleotidase (CD73) in vascular leakage during hypoxia**. *J. Exp. Med.* (2004) **200** 1395-1405. DOI: 10.1084/jem.20040915
73. Toju H.. **Fine-scale local adaptation of weevil mouthpart length and camellia pericarp thickness: altitudinal gradient of a putative arms race**. *Evolution* (2008) **62** 1086-1102. DOI: 10.1111/j.1558-5646.2008.00341.x
74. Tomkinson B.. **Tripeptidyl-peptidase II: update on an oldie that still counts**. *Biochimie* (2019) **166** 27-37. DOI: 10.1016/j.biochi.2019.05.012
75. Tremaroli V., Bäckhed F.. **Functional interactions between the gut microbiota and host metabolism**. *Nature* (2012) **489** 242-249. DOI: 10.1038/nature11552
76. Tsuchida T., Koga R., Fukatsu T.. **Host plant specialization governed by facultative symbiont**. *Science* (2004) **303** 1989. DOI: 10.1126/science.1094611
77. van der Wielen P. W., Rovers G. M., Scheepens J. M., Biesterveld S.. *Int. J. Syst. Evol. Microbiol.* (2002) **52** 921-925. DOI: 10.1099/00207713-52-3-921
78. Vermeiren J., Van den Abbeele P., Laukens D., Vigsnaes L. K., De Vos M., Boon N.. **Decreased colonization of fecal**. *FEMS Microbiol. Ecol.* (2012) **79** 685-696. DOI: 10.1111/j.1574-6941.2011.01252.x
79. Villar E., Farrant G. K., Follows M., Garczarek L., Speich S., Audic S.. **Ocean plankton. Environmental characteristics of Agulhas rings affect interocean plankton transport**. *Science* (2015) **348** 1261447. DOI: 10.1126/science.1261447
80. Wadhams G. H., Armitage J. P.. **Making sense of it all: bacterial chemotaxis**. *Nat. Rev. Mol. Cell Biol.* (2004) **5** 1024-1037. DOI: 10.1038/nrm1524
81. Wang X., Zhou S., Wu X., Wei Q., Shang Y., Sun G.. **High-altitude adaptation in vertebrates as revealed by mitochondrial genome analyses**. *Ecol. Evol.* (2021) **11** 15077-15084. DOI: 10.1002/ece3.8189
82. Wang Y., Zhou R., Yu Q., Feng T., Li H.. **Gut microbiome adaptation to extreme cold winter in wild plateau pika (**. *FEMS Microbiol. Lett.* (2020) **367** fnaa134. DOI: 10.1093/femsle/fnaa134
83. Wastyk H. C., Fragiadakis G. K., Perelman D., Dahan D., Merrill B. D., Yu F. B.. **Gut-microbiota-targeted diets modulate human immune status**. *Cells* (2021) **184** 4137-4153. DOI: 10.1016/j.cell.2021.06.019
84. Webb Alexander J., Homer Karen A., Hosie Arthur H. F.. **Two closely related ABC transporters in**. *J. Bacteriol.* (2008) **190** 168-178. DOI: 10.1128/JB.01509-07
85. West A. G., Waite D. W., Deines P., Bourne D. G., Digby A., McKenzie V. J.. **The microbiome in threatened species conservation**. *Biol. Conserv.* (2019) **229** 85-98. DOI: 10.1016/j.biocon.2018.11.016
86. Xu S., Li S., Yang Y., Tan J., Lou H., Jin W.. **A genome-wide search for signals of high-altitude adaptation in Tibetans**. *Mol. Biol. Evol.* (2011) **28** 1003-1011. DOI: 10.1093/molbev/msq277
87. Xu D., Yang C., Shen Q., Pan S., Liu Z., Zhang T.. **A single mutation underlying phenotypic convergence for hypoxia adaptation on the Qinghai-Tibetan plateau**. *Cell Res.* (2021) **31** 1032-1035. DOI: 10.1038/s41422-021-00517-6
88. Yang S., Dong H., Lei F.. **Phylogeography of regional fauna on the Tibetan plateau: a review**. *Prog. Nat. Sci.* (2009) **19** 789-799. DOI: 10.1016/j.pnsc.2008.10.006
89. Zeng B., Zhang S., Xu H., Kong F., Yu X., Wang P.. **Gut microbiota of Tibetans and Tibetan pigs varies between high and low altitude environments**. *Microbiol. Res.* (2020) **235** 126447. DOI: 10.1016/j.micres.2020.126447
90. Zhang T., Chen J., Zhang J., Guo Y.-T., Zhou X., Li M.-W.. **Phenotypic and genomic adaptations to the extremely high elevation in plateau zokor (**. *Mol. Ecol.* (2021) **30** 5765-5779. DOI: 10.1111/mec.16174
91. Zhang T., Lei M.-L., Zhou H., Chen Z.-Z., Shi P.. **Phylogenetic relationships of the zokor genus**. *Zool. Res.* (2022) **43** 331. DOI: 10.24272/j.issn.2095-8137.2022.045
92. Zhang G., Nan Z., Yin Z., Zhao L.. **Isolating the contributions of seasonal climate warming to permafrost thermal responses over the Qinghai-Tibet plateau**. *J. Geophys. Res. Atmos.* (2021) **126** e2021JD035218. DOI: 10.1029/2021JD035218
93. Zhang J., Song L., Wang Y., Liu C., Zhang L., Zhu S.. **Beneficial effect of butyrate-producing Lachnospiraceae on stress-induced visceral hypersensitivity in rats**. *J. Gastroenterol. Hepatol.* (2019) **34** 1368-1376. DOI: 10.1111/jgh.14536
94. Zhang H., Sparks J. B., Karyala S. V., Settlage R., Luo X. M.. **Host adaptive immunity alters gut microbiota**. *ISME J.* (2015) **9** 770-781. DOI: 10.1038/ismej.2014.165
95. Zhang X. Y., Sukhchuluun G., Bo T. B., Chi Q. S., Yang J. J., Chen B.. **Huddling remodels gut microbiota to reduce energy requirements in a small mammal species during cold exposure**. *Microbiome* (2018) **6** 103. DOI: 10.1186/s40168-018-0473-9
96. Zhang Z., Xu D., Wang L., Hao J., Wang J., Zhou X.. **Convergent evolution of rumen microbiomes in high-altitude mammals**. *Curr. Biol.* (2016) **26** 1873-1879. DOI: 10.1016/j.cub.2016.05.012
97. Zhao F., Ma J.-Y., Cai H.-X., Su J.-P., Hou Z.-B., Zhang T.-Z.. **Molecular identification of**. *Dongwuxue Yanjiu* (2014) **35** 313-318. DOI: 10.13918/j.issn.2095-8137.2014.4.313
|
---
title: Impact of genetically predicted characterization of mitochondrial DNA quantity
and quality on osteoarthritis
authors:
- Houpu Liu
- Bingyue Cai
- Ruicheng Gong
- Ye Yang
- Jing Wang
- Dan Zhou
- Min Yu
- Yingjun Li
journal: Frontiers in Genetics
year: 2023
pmcid: PMC9998702
doi: 10.3389/fgene.2023.1130411
license: CC BY 4.0
---
# Impact of genetically predicted characterization of mitochondrial DNA quantity and quality on osteoarthritis
## Abstract
Background: Existing studies have indicated that mitochondrial dysfunction may contribute to osteoarthritis (OA) development. However, the causal association between mitochondrial DNA (mtDNA) characterization and OA has not been extensively explored.
Methods: Two-sample Mendelian randomization was performed to calculate the impact of mitochondrial genomic variations on overall OA as well as site-specific OA, with multiple analytical methods inverse variance weighted (IVW), weighted median (WM), MR-Egger and MR-robust adjusted profile score (MR-RAPS).
Results: Genetically determined mitochondrial heteroplasmy (MtHz) and mtDNA abundance were not causally associated with overall OA. In site-specific OA analyses, the causal effect of mtDNA abundance on other OA sites, including hip, knee, thumb, hand, and finger, had not been discovered. There was a suggestively protective effect of MtHz on knee OA IVW OR = 0.632, $95\%$ CI: 0.425–0.939, p-value = 0.023. No causal association between MtHz and other different OA phenotypes was found.
Conclusion: MtHz shows potential to be a novel therapeutic target and biomarker on knee OA development. However, the variation of mtDNA abundance was measured from leukocyte in blood and the levels of MtHz were from saliva samples rather than cartilage or synovial tissues. Genotyping samples from synovial and cartilage can be a focus to further exploration.
## 1 Introduction
Osteoarthritis (OA) is a common joint disease with degenerative changes in articular cartilage and remains major impact on public health worldwide (Nelson, 2018). According to the estimation of the Global Burden of Disease Study (GBD), nearly 300 million individuals throughout the world are suffering from OA, and the number is still increasing. OA is one of the leading causes of pain, disability, and great socioeconomic burden in developed countries (Glyn-Jones et al., 2015). However, the mechanism of OA remains unclear. Identifying risk factors such as a disorder-related biomarker is essential to understand OA pathogenesis, decrease the incidence of the disease and develop efficiency prevention strategies.
Mitochondria are complex multifunctional organelles involved in various cell functions, e.g., heat regulation, calcium homeostasis, reactive oxygen species (ROS) production, and proinflammatory cytokines production, the damage of which may affect chondrocyte health at an extent (Blanco et al., 2011; Fernandez-Moreno et al., 2017; Blanco et al., 2018). They have their own genome (mitochondrial DNA, mtDNA) which contain 37 genes and encode 13 proteins, 22 transfer RNAs (tRNAs) and two ribosomal RNAs (rRNAs) (Anderson et al., 1981). As multicopy genome, sequence mutation and copy number variation of mtDNA are prevalent in population (McArdle et al., 2013; McGuire, 2019). As one of the mtDNA quantity characteristics, mtDNA abundance is recognized as a rough estimation for the number of mitochondria (Castellani et al., 2020). Previous studies provided supportive evidence for extensive pathogenicity of mtDNA abundance (Frahm et al., 2005; Hu et al., 2016). As one of the mtDNA quality characteristics, mitochondrial herteroplasmy (MtHz), where mtDNA with distinct sequences coexist, has been found in a large spectrum of human disease, including classical mitochondrial diseases and complex disorders (Ye et al., 2014). MtDNA abundance and MtHz can integrate many aspects of mitochondrial function and serve as promising biomarkers in probing interactions between mitochondria and disorders (Castellani et al., 2020; Tian et al., 2021).
In a recent review, OA was recognized as a potential mitochondrial disease considering the impact of mitochondrial dysfunction on cartilage degradation (Fernandez-Moreno et al., 2022). However, the causality between mitochondria and OA has rarely been investigated. Therefore, the goal of our study was to probe causal effects of characterization of mtDNA quantity and quality on OA development which is crucial to understand the role of mitochondria in OA etiology. A two-sample Mendelian randomization (MR) analysis was performed to investigate potential causal associations between mtDNA abundance and MtHz with OA using summary statistics from large-scale genome-wide association study (GWAS). MR-Steiger test was applied to ascertain whether variation of mtDNA characterization is a cause or a consequence of OA development.
## 2.1 Study design overview
The design of our research is displayed in Figure 1. We adopted a two-sample MR design to compute the causal effect of the characterization of mtDNA quantity and quality on overall OA and site-specific OA separately (Boer et al., 2021). Genetic association estimates for mtDNA abundance were derived from the United Kingdom biobank study ($$n = 291$$,950) (Hagg et al., 2021). Single Nucleotide Polymorphisms (SNPs) associated with MtHz were obtained from the 23andME research program ($$n = 982$$,072), a personal genomics and biotechnology company (Nandakumar et al., 2021). The participants included in both GWASs were of European ancestry.
**FIGURE 1:** *The assumptions of MR and how we tested these assumptions in our analyses.*
## 2.2 Selection of genetic instruments
SNPs serving as instrumental variables (IVs) for MtHz and mtDNA abundance, all reach genome-wide significance ($p \leq 5$ × 10–8) (Hagg et al., 2021; Nandakumar et al., 2021). Twenty SNPs were related to MtHz and accounted for $32\%$ observed SNP-heritability, 64 SNPs were included to estimate the genetic liability of mtDNA abundance and explained approximately $8.3\%$ SNP-heritability. We performed linkage disequilibrium (LD) pruning in PLINK with 1,000 Genomes Europeans as the reference panel to ascertain whether these genetic variants are independent of each other (r 2 < 0.001) (Burgess et al., 2017). After LD test, five genetic variants linked to MtHz were removed, and 17 IVs related to mtDNA abundance were excluded. These SNPs were then matched in summary GWASs of OA in subsequent analysis and those not available in the outcome GWAS were removed or replaced by the proxy SNPs. F-statistic was calculated to filter the weak instruments. The threshold of F-statistic that was sufficient for identifying causal effect was 10 (Burgess et al., 2011). The selected genetic variants are demonstrated in Supplementary Table S1, S2.
## 2.3 Osteoarthritis and data sources
Summary statistics data on overall OA and its specific sites, comprising knee, hip, spine, thumb, and hand OA, were obtained from the latest publicly available GWAS of OA. This GWAS covered nine populations with up to 826,690 participants (177,517 OA patients) (Boer et al., 2021). The definition of OA satisfied the criteria of Genetics of Osteoarthritis (GO) including self-reported status, hospital diagnosed, ICD10 codes or radiographic as defined by the TREAT-OA consortium (Boer et al., 2021). All studies contributing data to our analyses were approved by the relevant ethics committees, and all study participants in these studies provided written, informed consent.
## 2.4 Statistical analysis
Two-sample MR analysis was executed in R software (R version 4.3.0) with the TwoSampleMR (version 0.5.6), MR-PRESSO (version 1.0) and mr. raps (version 0.2) R packages (Hemani et al., 2018; Verbanck et al., 2018). The methods applied in our research are presented in Figure 2. MR-steiger test was applied to infer causal direction between traits under investigation (Hemani et al., 2017). Four methodologies including inverse variance weighted (IVW), weighted median (WM), MR-Egger and MR-robust adjusted profile score (MR-RAPS) were employed to estimate causality between mtDNA abundance and MtHz and OA. IVW was taken as a primary MR analysis method in our research, which is based on the hypothesis that all selected genetic instruments are valid and give an overall causal estimate strengthening causal inference (Lawlor et al., 2008; Burgess et al., 2013). MR-Egger regression takes presence of directional pleiotropy into consideration and measures horizontal pleiotropy with regression intercept, whereas the results of this method are susceptible to outlying genetic variants (Bowden et al., 2015). Weighted Median based on hypothesis that at least $50\%$ of the variants are valid, improves power of causal effect detection but reduces precision (Bowden et al., 2016). MR-RAPS applies robust adjusted profile scores to correct for pleiotropy and makes our results more reliable (Hartwig et al., 2017).
**FIGURE 2:** *Flowchart of Mendelian randomization framework in this study.*
Several sensitivity analyses were applied to detect and correct for heterogeneity and pleiotropy. MR-PRESSO was conducted to detect the existence of horizontal pleiotropy and correct the causal estimate affected by possible pleiotropic outliers (Verbanck et al., 2018). Corresponding p-values were derived based on 1,000 simulations (Verbanck et al., 2018). And the estimation of MR Egger regression intercept was also employed to reflect presence of pleiotropy ($p \leq 0.05$) (Burgess and Thompson, 2017). IVW method was utilized to investigate heterogeneity. The level of heterogeneity was quantified by Cochran Q statistics (Carnegie et al., 2020). Leave-one-out sensitivity analysis was performed to identify possibly influential SNPs, which repeated MR analysis with each SNP excluded in turn (Carnegie et al., 2020). As five separate outcomes were tested in our study, main results had statistical significance at p-value<0.01 ($\frac{0.05}{5}$) after Bonferroni correction.
## 3.1 Causal effect of mitochondrial heteroplasmy on OA
Fifteen LD-independent genetic variants were taken for repeated MR analysis (Supplementary Table S1). We extracted data of above-mentioned SNPs from summary GWAS of outcome traits (overall OA and specific-site OA) and one SNP (rs2286639) was removed in all outcomes except finger OA due to the effect of non-concordant alleles (e.g., A/G vs A/C). In finger OA, two SNP (rs2286639, rs3702096) were excluded for absence of proxy SNPs on the online platform SNiPA (https://snipa.helmholtzmuenchen.de/snipa3/). The mean F-statistic was 36.357 which was above the threshold F value of 10. In total, there were 14 IVs for overall, knee, hip, thumb, and hand OA and 13 IVs for finger OA.
Results of the casual association between MtHz and OA are summarized in Table 1. The primary IVW analysis provided no evidence for the casual association between MtHz and overall OA [odds ratio (OR) = 0.852, $95\%$ confidence interval (CI): 0.664–1.093, $$p \leq 0.208$$]. The results of WM method, MR-RAPS and MR-Egger were consistent with the result of IVW. Both MR-Egger and MR-PRESSO reported no existence of horizontal pleiotropy (Egger-intercept: $p \leq 0.05$; MR-PRESSO global heterogeneity: $p \leq 0.05$). Cochran’s Q test did not detect heterogeneity in overall OA ($$p \leq 0.495$$ > 0.05).
**TABLE 1**
| Outcome traits | MR methods | Mitochondrial heteroplasmy | Mitochondrial heteroplasmy.1 | Mitochondrial heteroplasmy.2 | Mitochondrial heteroplasmy.3 | Mitochondrial heteroplasmy.4 | Mitochondrial heteroplasmy.5 |
| --- | --- | --- | --- | --- | --- | --- | --- |
| Outcome traits | MR methods | OR (95%CI) | SE | MR p-value | Heterogeneity test | Pleiotropy test | MR-stegier test |
| Outcome traits | MR methods | OR (95%CI) | SE | MR p-value | Cochran’s Q (p) | p intercept | MR-stegier test |
| Overall OA | MR-Egger | 0.922 (0.619,1.373) | 0.203 | 0.697 | 12.407 (0.495) | 0.625 | Direction: TRUEp-value < 0.0001 |
| Overall OA | Inverse variance weighted | 0.852 (0.664,1.093) | 0.127 | 0.208 | 12.407 (0.495) | 0.625 | Direction: TRUEp-value < 0.0001 |
| Overall OA | MR-RAPS | 0.850 (0.662,1.093) | 0.128 | 0.206 | 12.407 (0.495) | 0.625 | Direction: TRUEp-value < 0.0001 |
| Overall OA | Weighted median | 0.867 (0.640,1.174) | 0.155 | 0.355 | 12.407 (0.495) | 0.625 | Direction: TRUEp-value < 0.0001 |
| Overall OA | MRPRESSO | 0.851 (0.670,1.082) | 0.123 | 0.209 | 12.407 (0.495) | 0.625 | Direction: TRUEp-value < 0.0001 |
| Knee OA | MR-Egger | 0.667 (0.358,1.243) | 0.318 | 0.226 | 11.272 (0.588) | 0.831 | Direction: TRUEp-value < 0.0001 |
| Knee OA | Inverse variance weighted | 0.632 (0.425,0.939) | 0.202 | 0.023 | 11.272 (0.588) | 0.831 | Direction: TRUEp-value < 0.0001 |
| Knee OA | MR-RAPS | 0.630 (0.422,0.939) | 0.204 | 0.023 | 11.272 (0.588) | 0.831 | Direction: TRUEp-value < 0.0001 |
| Knee OA | Weighted median | 0.637 (0.402,1.009) | 0.235 | 0.055 | 11.272 (0.588) | 0.831 | Direction: TRUEp-value < 0.0001 |
| Knee OA | MRPRESSO | 0.631 (0.442,0.903) | 0.182 | 0.024 | 11.272 (0.588) | 0.831 | Direction: TRUEp-value < 0.0001 |
| Hip OA | MR-Egger | 0.708 (0.320,1.568) | 0.406 | 0.412 | 10.208 (0.676) | 0.897 | Direction: TRUEp-value < 0.0001 |
| Hip OA | Inverse variance weighted | 0.738 (0.447,1.221) | 0.257 | 0.237 | 10.208 (0.676) | 0.897 | Direction: TRUEp-value < 0.0001 |
| Hip OA | MR-RAPS | 0.734 (0.442,1.220) | 0.259 | 0.234 | 10.208 (0.676) | 0.897 | Direction: TRUEp-value < 0.0001 |
| Hip OA | Weighted median | 0.711 (0.391,1.293) | 0.310 | 0.272 | 10.208 (0.676) | 0.897 | Direction: TRUEp-value < 0.0001 |
| Hip OA | MRPRESSO | 0.736 (0.466,1.161) | 0.233 | 0.208 | 10.208 (0.676) | 0.897 | Direction: TRUEp-value < 0.0001 |
| Thumb OA | MR-Egger | 0.708 (0.320,1.568) | 0.406 | 0.412 | 10.161 (0.681) | 0.288 | Direction: TRUEp-value < 0.0001 |
| Thumb OA | Inverse variance weighted | 1.270 (0.506,3.188) | 0.469 | 0.611 | 10.161 (0.681) | 0.288 | Direction: TRUEp-value < 0.0001 |
| Thumb OA | MR-RAPS | 1.270 (0.502,3.216) | 0.474 | 0.615 | 10.161 (0.681) | 0.288 | Direction: TRUEp-value < 0.0001 |
| Thumb OA | Weighted median | 0.711 (0.391,1.293) | 0.305 | 0.264 | 10.161 (0.681) | 0.288 | Direction: TRUEp-value < 0.0001 |
| Thumb OA | MRPRESSO | 1.267 (0.576,2.791) | 0.403 | 0.566 | 10.161 (0.681) | 0.288 | Direction: TRUEp-value < 0.0001 |
| Hand OA | MR-Egger | 0.967 (0.333,2.803) | 0.543 | 0.951 | 7.958 (0.846) | 0.432 | Direction: TRUEp-value < 0.0001 |
| Hand OA | Inverse variance weighted | 1.357 (0.687,2.681) | 0.347 | 0.379 | 7.958 (0.846) | 0.432 | Direction: TRUEp-value < 0.0001 |
| Hand OA | MR-RAPS | 1.350 (0.678,2.686) | 0.351 | 0.392 | 7.958 (0.846) | 0.432 | Direction: TRUEp-value < 0.0001 |
| Hand OA | Weighted median | 1.132 (0.503,2.546) | 0.413 | 0.764 | 7.958 (0.846) | 0.432 | Direction: TRUEp-value < 0.0001 |
| Hand OA | MRPRESSO | 1.348 (0.769,2.362) | 0.286 | 0.315 | 7.958 (0.846) | 0.432 | Direction: TRUEp-value < 0.0001 |
| Finger OA | MR-Egger | 0.139 (0.010,1.909) | 1.336 | 0.168 | 8.157 (0.773) | 0.251 | Direction: TRUEp-value < 0.0001 |
| Finger OA | Inverse variance weighted | 0.483 (0.090,2.588) | 0.857 | 0.395 | 8.157 (0.773) | 0.251 | Direction: TRUEp-value < 0.0001 |
| Finger OA | MR-RAPS | 0.481 (0.088,2.621) | 0.865 | 0.398 | 8.157 (0.773) | 0.251 | Direction: TRUEp-value < 0.0001 |
| Finger OA | Weighted median | 0.364 (0.049,2.688) | 1.021 | 0.322 | 8.157 (0.773) | 0.251 | Direction: TRUEp-value < 0.0001 |
| Finger OA | MRPRESSO | 0.483 (0.121,1.927) | 0.707 | 0.323 | 8.157 (0.773) | 0.251 | Direction: TRUEp-value < 0.0001 |
In site-specific OA analyses, the nominally significant results of IVW analysis suggested that MtHz was a potentially protective factor for knee OA (OR = 0.632, $95\%$ CI: 0.425–0.939, $$p \leq 0.023$$, Table 1). MR-RAPS showed similar results with IVW analysis (OR = 0.629, $95\%$ CI: 0.422–0.939, $$p \leq 0.023$$, Table 1). However, no relationship between MtHz and other OA sites, including hip OA (IVW OR = 0.738, $95\%$ CI: 0.447–1.221, $$p \leq 0.237$$), hand OA (IVW OR = 1.357, $95\%$ CI: 0.687–2.681, $$p \leq 0.379$$), thumb OA (IVW OR = 1.270, $95\%$ CI: 0.506–3.188, $$p \leq 0.611$$) or finger OA (IVW OR = 0.483, $95\%$ CI: 0.090–2.588, $$p \leq 0.395$$) was observed in IVW model (Table 1). MR-PRESSO was performed to test for horizontal pleiotropy, and no outliers was identified. The results of MR-Egger intercept were consistent with the MR-PRESSO results (intercept $p \leq 0.05$). The heterogeneity was tested by Cochran’s Q test and MR-PRESSO global heterogeneity test, providing no evidence about the existence of heterogeneity. The MR-Steiger results supported that these SNPs were more predictive of the exposure than of the outcome ($p \leq 0.05$, Supplementary Table S3). The results of leave-one-out sensitivity analysis and forest plots demonstrated that our study in genetically prediction was robust (Supplementary Figure S1–S3).
## 3.2 Causal effect of mitochondrial abundance on OA
After LD test, 47 independent genetic variants were chosen for two-sample MR analysis. *Several* genetic variants could not be matched in summary statistic GWASs of knee (rs35734242), finger (rs1065853), and hand OA (rs12451555). We searched their proxy SNPs on SNiPA and those whose proxy SNPs were absent on SNiPA were excluded. The mean F-statistic for IVs of mtDNA abundance was 93.380, which was above the threshold 10 (Supplementary Table S2). Finally, the SNPS were selected as IVs for thumb, hip, overall OA was 47, and 46 SNPS for hand, finger, knee. Regarding overall OA, MR-PRESSO identified two outliers s (rs59488041, rs12924138) and therefore removed them for repeated MR analysis. In the presence of heterogeneity (Cochran’s Q: $$p \leq 0.033$$) and absence of pleiotropy (Egger-intercept: $$p \leq 0.687$$ > 0.05), the results of WM analysis on overall OA were nominally significant (OR = 1.108, $95\%$ CI: 1.001–1,226, $$p \leq 0.048$$, Table 2). IVW method in random effects model was utilized to correct results potentially impacted by heterogeneity, and the results suggested that genetically elevated mtDNA abundance was not casually associated with overall OA (OR = 1.040, $95\%$ CI: 0.963–1.125, $$p \leq 0.318$$). MR-RAPS analysis agreed with the results of IVW analysis (OR = 1.041, $95\%$ CI: 0.974–1.112, $$p \leq 0.228$$, Table 2).
**TABLE 2**
| Outcome traits | MR methods | MtDNA abundance | MtDNA abundance.1 | MtDNA abundance.2 | MtDNA abundance.3 | MtDNA abundance.4 | MtDNA abundance.5 |
| --- | --- | --- | --- | --- | --- | --- | --- |
| Outcome traits | MR methods | OR (95%CI) | SE | MR p-value | Heterogeneity test | Pleiotropy test | MR-stegier test |
| Outcome traits | MR methods | OR (95%CI) | SE | MR p-value | Cochran’s Q (p) | p intercept | MR-stegier test |
| Overall OA | MR-Egger | 1.091 (0.856,1.391) | 0.124 | 0.485 | 62.748 (0.033) | 0.687 | Direction: TRUE |
| Overall OA | Inverse variance weighted | 1.040 (0.963,1.125) | 0.040 | 0.318 | 62.748 (0.033) | 0.687 | Direction: TRUE |
| Overall OA | MR-RAPS | 1.041 (0.974,1.112) | 0.034 | 0.229 | 62.748 (0.033) | 0.687 | p-value < 0.0001 |
| Overall OA | Weighted median | 1.108 (1.001,1,226) | 0.052 | 0.048 | 62.748 (0.033) | 0.687 | p-value < 0.0001 |
| Overall OA | MRPRESSO | 1.040 (0.963,1.125) | 0.040 | 0.323 | 62.748 (0.033) | 0.687 | p-value < 0.0001 |
| Knee OA | MR-Egger | 0.823 (0.562,1.206) | 0.195 | 0.324 | 65.050 (0.027) | 0.225 | Direction: TRUE |
| Knee OA | Inverse variance weighted | 1.034 (0.915,1.168) | 0.062 | 0.592 | 65.050 (0.027) | 0.225 | Direction: TRUE |
| Knee OA | MR-RAPS | 1.035 (0.934,1.145) | 0.052 | 0.516 | 65.050 (0.027) | 0.225 | p-value < 0.0001 |
| Knee OA | Weighted median | 1.056 (0.903,1.236) | 0.080 | 0.494 | 65.050 (0.027) | 0.225 | p-value < 0.0001 |
| Knee OA | MRPRESSO | 1.034 (0.915,1.168) | 0.062 | 0.594 | 65.050 (0.027) | 0.225 | p-value < 0.0001 |
| Hip OA | MR-Egger | 0.616 (0.399,0.950) | 0.221 | 0.034 | 60.396 (0.062) | 0.027 | Direction: TRUE |
| Hip OA | Inverse variance weighted | 0.992 (0.857,1.152) | 0.076 | 0.931 | 60.396 (0.062) | 0.027 | Direction: TRUE |
| Hip OA | MR-RAPS | 0.993 (0.873,1.130) | 0.066 | 0.919 | 60.396 (0.062) | 0.027 | p-value < 0.0001 |
| Hip OA | Weighted median | 0.921 (0.763,1.113) | 0.096 | 0.397 | 60.396 (0.062) | 0.027 | p-value < 0.0001 |
| Hip OA | MRPRESSO | 0.993 (0.857,1.152) | 0.076 | 0.931 | 60.396 (0.062) | 0.027 | p-value < 0.0001 |
| Thumb OA | MR-Egger | 0.819 (0.364,1.844) | 0.414 | 0.633 | 61.709 (0.061) | 0.392 | Direction: TRUE |
| Thumb OA | Inverse variance weighted | 1.148 (0.880,1.499) | 0.136 | 0.309 | 61.709 (0.061) | 0.392 | Direction: TRUE |
| Thumb OA | MR-RAPS | 1.151 (0.912,1.454) | 0.119 | 0.235 | 61.709 (0.061) | 0.392 | p-value <0.0001 |
| Thumb OA | Weighted median | 1.135 (0.790,1.631) | 0.185 | 0.494 | 61.709 (0.061) | 0.392 | p-value <0.0001 |
| Thumb OA | MRPRESSO | 1.148 (0.880,1.499) | 0.136 | 0.314 | 61.709 (0.061) | 0.392 | p-value <0.0001 |
| Hand OA | MR-Egger | 1.028 (0.601,1.759) | 0.274 | 0.920 | 40.089 (0.680) | 0.663 | Direction: TRUE |
| Hand OA | Inverse variance weighted | 1.152 (0.969,1.369) | 0.088 | 0.108 | 40.089 (0.680) | 0.663 | Direction: TRUE |
| Hand OA | MR-RAPS | 1.154 (0.967,1.376) | 0.090 | 0.109 | 40.089 (0.680) | 0.663 | p-value < 0.0001 |
| Hand OA | Weighted median | 1.144 (0.887,1.476) | 0.130 | 0.299 | 40.089 (0.680) | 0.663 | p-value < 0.0001 |
| Hand OA | MRPRESSO | 1.152 (0.979,1.356) | 0.083 | 0.096 | 40.089 (0.680) | 0.663 | p-value < 0.0001 |
| Finger OA | MR-Egger | 0.876 (0.223,3.442) | 0.698 | 0.850 | 51.688 (0.229) | 0.837 | Direction: TRUE |
| Finger OA | Inverse variance weighted | 1.004 (0.647,1.556) | 0.224 | 0.987 | 51.688 (0.229) | 0.837 | Direction: TRUE |
| Finger OA | MR-RAPS | 1.003 (0.662,1.520) | 0.212 | 0.986 | 51.688 (0.229) | 0.837 | p-value < 0.0001 |
| Finger OA | Weighted median | 1.193 (0.641,2.221) | 0.317 | 0.577 | 51.688 (0.229) | 0.837 | p-value < 0.0001 |
| Finger OA | MRPRESSO | 1.004 (0.647,1.556) | 0.224 | 0.987 | 51.688 (0.229) | 0.837 | p-value < 0.0001 |
With respect to site-specific OA, we observed that genetically determined MtHz was not causally associated with knee OA (OR = 1.034, $95\%$ CI: 0.915–1.168, $$p \leq 0.592$$), thumb OA (OR = 1.148, $95\%$ CI: 0.880–1.499, $$p \leq 0.309$$), hand OA (OR = 1.152, $95\%$ CI: 0.969–1.369, $$p \leq 0.108$$) and finger OA (OR = 1.004, $95\%$ CI: 0.647–1.556, $$p \leq 0.987$$) in IVW model. The results are presented in Table 2. Both MR-Egger and MR-PRESSO reported no existence of horizontal pleiotropy (Egger-intercept: $p \leq 0.05$; MR-PRESSO global heterogeneity: $p \leq 0.05$). Cochran’s Q test did not detect heterogeneity in above-mentioned outcome traits ($p \leq 0.05$). For the analysis of hip OA, rs16978036 were excluded from MR-PRESSO analysis. In presence of horizontal pleiotropy ($$p \leq 0.027$$) and absence of heterogeneity ($$p \leq 0.062$$), the nominally significant result of MR-Egger method was found (OR = 0.616, $95\%$ CI: 0.399–0.950, $$p \leq 0.034$$). However, no evidence about causal relationship between the exposure and hip OA was observed with IVW method and MR-RAPS method (IVW: OR = 0.992, $95\%$ CI: 0.857–1.152, $$p \leq 0.931$$; MR-RAPS: OR = 0.993, $95\%$ CI: 0.873–1.130, $$p \leq 0.919$$) (Table 2). The results of MR-Stegier test were in Supplementary Table S3.
Scatter plots, forest plots and leave-on-out sensitivity analysis plots were displayed in Supplementary Figure S4–6. The results of leave-one-out analysis implicated that the selected genetic variants potentially impacted the pooled results, which suggested that careful interpretations for the results was crucial.
## 4 Discussion
The causal roles of MtHz and mtDNA abundance in OA pathogenesis were poorly studied in previous works. To our best knowledge, this is the first MR study to evaluate the causal association between mitochondrial genome traits and OA. Four MR methods were employed to estimate the causal association between mtDNA characterization and OA. The results of MR steiger test verified the causal direction of our research (MtHz and mtDNA abundance were exposure and OA were outcome). No effect of mtDNA abundance and MtHz on overall OA was detected. In subgroup analysis, a suggestively protective role of MtHz on knee OA was observed, but not on other sites. And we did not find that mtDNA abundance was causally associated with any site-specific OA.
Although mitochondria is widely recognized as an important factor of OA development (Blanco et al., 2011), the role of MtHz has not been thoroughly investigated (Suliman and Piantadosi, 2016). Indeed, heteroplasmic mutations in mtDNA are often pathogenetic (Ye et al., 2014). An animal study has indicated that the state of heteroplasmy itself was deleterious when the two mtDNA sequences contain no pathogenic variants (Sharpley et al., 2012). However, there is potentially a particular level threshold for MtHz (McCormick et al., 2020). For instance, when the A3243G mutation in mitochondrial DNA is present in more than $10\%$, patients can manifest Type 2 diabetes (Wallace, 2005). And low-frequency mtDNA variants ($0.2\%$–$2\%$ heteroplasmy) are extensively presented in healthy subjects (Payne et al., 2013). Furthermore, MtHz can be beneficial in health promotion as an intermediate state in emergence of novel mtDNA haplogroups (Wallace, 2016). In previous study, MtHz was found to be significantly relevant to several haplogroups (haplogroup H, J, K, T, U and X) with different characterizations among haplogroups (Nandakumar et al., 2021). Mitochondrial haplogroup J has been extensively found to mediate the development of OA [5]. A Spanish cohort-study (ncase = 457; ncontrol = 262) had reported that haplogroup J was associated with a decreased risk of knee OA (OR = 0.460, $95\%$ CI: 0.282–0.748, $$p \leq 0.002$$) (Rego-Perez et al., 2008). Furthermore, a meta-analysis in European cohorts also suggested that haplogroup J was associated with a lower risk of knee OA (HR = 0.702, $95\%$ CI: 0.541–0.912, $$p \leq 0.008$$) (Fernandez-Moreno et al., 2017). However, the association between mitochondrial DNA variants and OA has not been verified in a large sample observational study (Hudson et al., 2013). The study has only explored the causal relationship in terms of single mutation, but we take MtHz (variation at a whole mtDNA level) as an exposure which contribute to understand the causal role of mitochondria in OA development. In terms of biological mechanism, MtHz has been reported a regulation role of metabolic and epigenomic changes. Kopinski PK et al. had found the levels of mitochondrially drove acetyl-CoA decreased at high heteroplasmy (Kopinski et al., 2019). And the reduction of acetyl-CoA levels might influence histone acetylation and activate AMP-activated protein kinase (AMPK) to protect from OA development and progression (Chen et al., 2018). Besides, MtHz could play a role in other processes involved in OA pathogenesis to reduce the risk of OA, including regulating energy production, maintaining mitochondrial proteostasis, suppressing matrix metalloproteinase expression, reducing ROS generation, and promoting mitophagy (Blanco et al., 2018). In conjunction with prior observational findings and mechanism studies, our epidemiologic and genetic findings provided supportive evidence that MtHz may have therapeutic value in knee OA and can be regarded as a candidate biomarker after precisely identifying the relationship between MtHz and each haplogroup (Blanco et al., 2018).
However, the reason that MtHz showed diverse effects on different sites of OA remains unclear. So far, we did not find relevant research that investigated the relationship between MtHz and hand, finger, and thumb OA. A case-control study suggested that mitochondrial haplogroups was associated with hip OA (Rego et al., 2010) (OR = 0.661, $95\%$ CI: 0.440–0.993, $$p \leq 0.045$$), however, with no evidence on the causality. In addition, the data sources and sample size used in MR analysis varied in site-specific OA, which mainly included more knee OA cases and less other skeletal joints cases. More MR studies and additional GWAS covering more other site-specific OAs are needed to determine a role of MtHz in the risk of different OA sites.
MtDNA abundance has been regarded as a potential biomarker of mitochondrial function and plays a role in several human diseases (Castellani et al., 2020; Clyde, 2022). Existing literature has suggested that mtDNA abundance may involve in production of inflammatory mediators and regulation of immune function that can influence OA development (Blanco et al., 2018; Zhan et al., 2020). In addition, mtDNA abundance is also associated with sex, advanced age, and elevated BMI, which are also risk factors for OA [16]. However, there are few studies to estimate the effects of mtDNA abundance on OA. Only two case-control studies in Asian population were found that reported an association between mtDNA copy number and OA, but their findings were inconsistent (Fang et al., 2014; Zhan et al., 2020). One in Thailand (ncase = 204; ncontrol = 169) had found mtDNA abundance in the OA group was significantly lower than that in the control group ($p \leq 0.0001$), whereas the results were not adjusted by sex and age (Zhan et al., 2020). Another case-control study carried out in southern Chinese (ncase = 187; ncontrol = 420) observed a general increase of mtDNA abundance in OA patients ($$p \leq 0.019$$), but obesity was not adjusted in the analysis (Fang et al., 2014). These findings from case-control studies could not determine causal relationships between mtDNA abundance and OA, and the potential confounding variables were not comprehensively considered. Our MR analysis overcame these shortcomings and suggested that genetically determined mtDNA abundance was unrelated to OA. Furthermore, previous study has implicated that the relative contribution of mtDNA abundance might differ between different ethnic groups (Ruiz-Pesini et al., 2004). Considering that our studies have been carried out in European population, the comparisons with results from other ethnic groups such as Asian ancestry require careful considerations.
Our research applies MR methods to investigate causal relationships between mitochondrial genome characterization and OA. However, there are some limitations in our analysis. Firstly, we did not estimate causal effects of mitochondrial genome traits on OA stratified by gender. Mitochondrial genome are maternally inherited, and females may have lower MtHz than males from the perspective of mitochondrial inheritance (Nandakumar et al., 2021). Therefore, the effects of MtHz and mtDNA abundance on OA may differ in gender. Secondly, only participants of European descent are included in the study, but the impact of specific mtDNA variants on diseases could vary in different ethnic groups (Ruiz-Pesini et al., 2004). Additional MR studies on other ethnic groups are needed to probe a causal association between mtDNA characterization and OA. Thirdly, the genetic instruments for mtDNA abundance explain a relatively small amount of phenotypic variance ($8.3\%$), IVs that can account for more variance of mtDNA abundance are warranted to draw robust conclusions. Besides, considering the Bonferroni correction of multiple independent tests, our findings about the association between MtHz and knee OA are deemed suggestive evidence of possible associations (0.01 < $p \leq 0.05$). Furthermore, the variation of mtDNA abundance and the levels of MtHz varied in different tissues and cell types. And in original GWASs, mtDNA abundance was measured from leukocyte in blood and MtHz from saliva samples rather than cartilage or synovial tissues. Considering that OA is mainly characterized by progressive loss of cartilage and synovial hyperplasia, the application of data from blood samples and saliva samples would limit the explanation ability of our study to a certain extent (Sellam and Berenbaum, 2010). Observational epidemiology studies exploring an association between concrete levels of MtHz and a risk of OA are needed to improve causal inference.
## 5 Conclusion
In conclusion, our MR analyses elucidated that MtHz is a suggestively protective factor of knee OA, implying that MtHz could be a genetically prediction factor and a therapeutic target in the development of knee OA. No causal association was found between mtDNA abundance and OA. Additional MR analyses are warranted to probe causal relationships between mitochondrial genome traits and OA stratified by gender. Moreover, GWAS covering more than one ethnic population are needed to detect the effect of characterization of mtDNA quantity and quality in different ethnic groups.
## Data availability statement
The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding authors.
## Author contributions
HL mainly designed analysis and wrote the manuscript; BC, RG, DZ performed experiments and analysis; YY verified data and analysis results, JW, MY, YL supervised the entire project. All authors have read, provided critical feedback on intellectual content, and approved the final manuscript.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fgene.2023.1130411/full#supplementary-material
## References
1. Anderson S., Bankier A. T., Barrell B. G., de Bruijn M. H., Coulson A. R., Drouin J.. **Sequence and organization of the human mitochondrial genome**. *Nature* (1981) **290** 457-465. DOI: 10.1038/290457a0
2. Blanco F. J., Rego I., Ruiz-Romero C.. **The role of mitochondria in osteoarthritis**. *Nat. Rev. Rheumatol.* (2011) **7** 161-169. DOI: 10.1038/nrrheum.2010.213
3. Blanco F. J., Valdes A. M., Rego-Perez I.. **Mitochondrial DNA variation and the pathogenesis of osteoarthritis phenotypes**. *Nat. Rev. Rheumatol.* (2018) **14** 327-340. DOI: 10.1038/s41584-018-0001-0
4. Boer C. G., Hatzikotoulas K., Southam L., Stefansdottir L., Zhang Y., Coutinho de Almeida R.. **Deciphering osteoarthritis genetics across 826,690 individuals from 9 populations**. *Cell* (2021) **184** 6003-6005. DOI: 10.1016/j.cell.2021.11.003
5. Bowden J., Davey Smith G., Burgess S.. **Mendelian randomization with invalid instruments: Effect estimation and bias detection through egger regression**. *Int. J. Epidemiol.* (2015) **44** 512-525. DOI: 10.1093/ije/dyv080
6. Bowden J., Davey Smith G., Haycock P. C., Burgess S.. **Consistent estimation in mendelian randomization with some invalid instruments using a weighted median estimator**. *Genet. Epidemiol.* (2016) **40** 304-314. DOI: 10.1002/gepi.21965
7. Burgess S., Butterworth A., Thompson S. G.. **Mendelian randomization analysis with multiple genetic variants using summarized data**. *Genet. Epidemiol.* (2013) **37** 658-665. DOI: 10.1002/gepi.21758
8. Burgess S., Small D. S., Thompson S. G.. **A review of instrumental variable estimators for Mendelian randomization**. *Stat. Methods Med. Res.* (2017) **26** 2333-2355. DOI: 10.1177/0962280215597579
9. Burgess S., Thompson S. G., Collaboration C. C. G.. **Avoiding bias from weak instruments in Mendelian randomization studies**. *Int. J. Epidemiol.* (2011) **40** 755-764. DOI: 10.1093/ije/dyr036
10. Burgess S., Thompson S. G.. **Interpreting findings from Mendelian randomization using the MR-Egger method**. *Eur. J. Epidemiol.* (2017) **32** 377-389. DOI: 10.1007/s10654-017-0255-x
11. Carnegie R., Zheng J., Sallis H. M., Jones H. J., Wade K. H., Evans J.. **Mendelian randomisation for nutritional psychiatry**. *Lancet Psychiatry* (2020) **7** 208-216. DOI: 10.1016/S2215-0366(19)30293-7
12. Castellani C. A., Longchamps R. J., Sun J., Guallar E., Arking D. E.. **Thinking outside the nucleus: Mitochondrial DNA copy number in health and disease**. *Mitochondrion* (2020) **53** 214-223. DOI: 10.1016/j.mito.2020.06.004
13. Chen L. Y., Lotz M., Terkeltaub R., Liu-Bryan R.. **Modulation of matrix metabolism by ATP-citrate lyase in articular chondrocytes**. *J. Biol. Chem.* (2018) **293** 12259-12270. DOI: 10.1074/jbc.RA118.002261
14. Clyde D.. **Genomic insights into conifer evolution**. *Nat. Rev. Genet.* (2022) **23** 136. DOI: 10.1038/s41576-022-00452-1
15. Fang H., Liu X., Shen L., Li F., Liu Y., Chi H.. **Role of mtDNA haplogroups in the prevalence of knee osteoarthritis in a southern Chinese population**. *Int. J. Mol. Sci.* (2014) **15** 2646-2659. DOI: 10.3390/ijms15022646
16. Fernandez-Moreno M., Rego-Perez I., Blanco F. J.. **Is osteoarthritis a mitochondrial disease? What is the evidence**. *Curr. Opin. Rheumatol.* (2022) **34** 46-53. DOI: 10.1097/BOR.0000000000000855
17. Fernandez-Moreno M., Soto-Hermida A., Vazquez-Mosquera M. E., Cortes-Pereira E., Relano S., Hermida-Gomez T.. **Mitochondrial DNA haplogroups influence the risk of incident knee osteoarthritis in OAI and CHECK cohorts. A meta-analysis and functional study**. *Ann. Rheum. Dis.* (2017) **76** 1114-1122. DOI: 10.1136/annrheumdis-2016-210131
18. Frahm T., Mohamed S. A., Bruse P., Gemund C., Oehmichen M., Meissner C.. **Lack of age-related increase of mitochondrial DNA amount in brain, skeletal muscle and human heart**. *Mech. Ageing Dev.* (2005) **126** 1192-1200. DOI: 10.1016/j.mad.2005.06.008
19. Glyn-Jones S., Palmer A. J., Agricola R., Price A. J., Vincent T. L., Weinans H.. *Osteoarthr. Lancet.* (2015) **386** 376-387. DOI: 10.1016/S0140-6736(14)60802-3
20. Hagg S., Jylhava J., Wang Y., Czene K., Grassmann F.. **Deciphering the genetic and epidemiological landscape of mitochondrial DNA abundance**. *Hum. Genet.* (2021) **140** 849-861. DOI: 10.1007/s00439-020-02249-w
21. Hartwig F. P., Davey Smith G., Bowden J.. **Robust inference in summary data Mendelian randomization via the zero modal pleiotropy assumption**. *Int. J. Epidemiol.* (2017) **46** 1985-1998. DOI: 10.1093/ije/dyx102
22. Hemani G., Tilling K., Davey Smith G.. **Orienting the causal relationship between imprecisely measured traits using GWAS summary data**. *PLoS Genet.* (2017) **13** e1007081. DOI: 10.1371/journal.pgen.1007081
23. Hemani G., Zheng J., Elsworth B., Wade K. H., Haberland V., Baird D.. **The MR-Base platform supports systematic causal inference across the human phenome**. *Elife* (2018) **7** e34408. DOI: 10.7554/eLife.34408
24. Hu L., Yao X., Shen Y.. **Altered mitochondrial DNA copy number contributes to human cancer risk: Evidence from an updated meta-analysis**. *Sci. Rep.* (2016) **6** 35859. DOI: 10.1038/srep35859
25. Hudson G., Panoutsopoulou K., Wilson I., Southam L., Rayner N. W., Arden N.. **No evidence of an association between mitochondrial DNA variants and osteoarthritis in 7393 cases and 5122 controls**. *Ann. Rheum. Dis.* (2013) **72** 136-139. DOI: 10.1136/annrheumdis-2012-201932
26. Kopinski P. K., Janssen K. A., Schaefer P. M., Trefely S., Perry C. E., Potluri P.. **Regulation of nuclear epigenome by mitochondrial DNA heteroplasmy**. *Proc. Natl. Acad. Sci. U. S. A.* (2019) **116** 16028-16035. DOI: 10.1073/pnas.1906896116
27. Lawlor D. A., Harbord R. M., Sterne J. A., Timpson N., Davey Smith G.. **Mendelian randomization: Using genes as instruments for making causal inferences in epidemiology**. *Stat. Med.* (2008) **27** 1133-1163. DOI: 10.1002/sim.3034
28. McArdle M. A., Finucane O. M., Connaughton R. M., McMorrow A. M., Roche H. M.. **Mechanisms of obesity-induced inflammation and insulin resistance: Insights into the emerging role of nutritional strategies**. *Front. Endocrinol. (Lausanne)* (2013) **4** 52. DOI: 10.3389/fendo.2013.00052
29. McCormick E. M., Lott M. T., Dulik M. C., Shen L., Attimonelli M., Vitale O.. **Specifications of the ACMG/AMP standards and guidelines for mitochondrial DNA variant interpretation**. *Hum. Mutat.* (2020) **41** 2028-2057. DOI: 10.1002/humu.24107
30. McGuire P. J.. **Mitochondrial dysfunction and the aging immune system**. *Biol. (Basel)* (2019) **8** 26. DOI: 10.3390/biology8020026
31. Nandakumar P., Tian C., O'Connell J., andMe Research T., Hinds D., Paterson A. D.. **Nuclear genome-wide associations with mitochondrial heteroplasmy**. *Sci. Adv.* (2021) **7** eabe7520. DOI: 10.1126/sciadv.abe7520
32. Nelson A. E.. **Osteoarthritis year in review 2017: Clinical**. *Osteoarthr. Cartil.* (2018) **26** 319-325. DOI: 10.1016/j.joca.2017.11.014
33. Payne B. A., Wilson I. J., Yu-Wai-Man P., Coxhead J., Deehan D., Horvath R.. **Universal heteroplasmy of human mitochondrial DNA**. *Hum. Mol. Genet.* (2013) **22** 384-390. DOI: 10.1093/hmg/dds435
34. Rego I., Fernandez-Moreno M., Fernandez-Lopez C., Gomez-Reino J. J., Gonzalez A., Arenas J.. **Role of European mitochondrial DNA haplogroups in the prevalence of hip osteoarthritis in Galicia, Northern Spain**. *Ann. Rheum. Dis.* (2010) **69** 210-213. DOI: 10.1136/ard.2008.105254
35. Rego-Perez I., Fernandez-Moreno M., Fernandez-Lopez C., Arenas J., Blanco F. J.. **Mitochondrial DNA haplogroups: Role in the prevalence and severity of knee osteoarthritis**. *Arthritis Rheum.* (2008) **58** 2387-2396. DOI: 10.1002/art.23659
36. Ruiz-Pesini E., Mishmar D., Brandon M., Procaccio V., Wallace D. C.. **Effects of purifying and adaptive selection on regional variation in human mtDNA**. *Science* (2004) **303** 223-226. DOI: 10.1126/science.1088434
37. Sellam J., Berenbaum F.. **The role of synovitis in pathophysiology and clinical symptoms of osteoarthritis**. *Nat. Rev. Rheumatol.* (2010) **6** 625-635. DOI: 10.1038/nrrheum.2010.159
38. Sharpley M. S., Marciniak C., Eckel-Mahan K., McManus M., Crimi M., Waymire K.. **Heteroplasmy of mouse mtDNA is genetically unstable and results in altered behavior and cognition**. *Cell* (2012) **151** 333-343. DOI: 10.1016/j.cell.2012.09.004
39. Suliman H. B., Piantadosi C. A.. **Mitochondrial quality control as a therapeutic target**. *Pharmacol. Rev.* (2016) **68** 20-48. DOI: 10.1124/pr.115.011502
40. Tian Q., Moore A. Z., Oppong R., Ding J., Zampino M., Fishbein K. W.. **Mitochondrial DNA copy number and heteroplasmy load correlate with skeletal muscle oxidative capacity by P31 MR spectroscopy**. *Aging Cell* (2021) **20** e13487. DOI: 10.1111/acel.13487
41. Verbanck M., Chen C. Y., Neale B., Do R.. **Detection of widespread horizontal pleiotropy in causal relationships inferred from Mendelian randomization between complex traits and diseases**. *Nat. Genet.* (2018) **50** 693-698. DOI: 10.1038/s41588-018-0099-7
42. Wallace D. C.. **A mitochondrial paradigm of metabolic and degenerative diseases, aging, and cancer: A dawn for evolutionary medicine**. *Annu. Rev. Genet.* (2005) **39** 359-407. DOI: 10.1146/annurev.genet.39.110304.095751
43. Wallace D. C.. **Genetics: Mitochondrial DNA in evolution and disease**. *Nature* (2016) **535** 498-500. DOI: 10.1038/nature18902
44. Ye K., Lu J., Ma F., Keinan A., Gu Z.. **Extensive pathogenicity of mitochondrial heteroplasmy in healthy human individuals**. *Proc. Natl. Acad. Sci. U. S. A.* (2014) **111** 10654-10659. DOI: 10.1073/pnas.1403521111
45. Zhan D., Tanavalee A., Tantavisut S., Ngarmukos S., Edwards S. W., Honsawek S.. **Relationships between blood leukocyte mitochondrial DNA copy number and inflammatory cytokines in knee osteoarthritis**. *J. Zhejiang Univ. Sci. B* (2020) **21** 42-52. DOI: 10.1631/jzus.B1900352
|
---
title: Prognostic factors associated with improvements in patient-reported outcomes
in idiopathic adhesive capsulitis
authors:
- Paul V. Romeo
- Aidan G. Papalia
- Matthew G. Alben
- Neil Gambhir
- Dhruv Shankar
- Andrew S. Bi
- Joseph D. Zuckerman
- Mandeep S. Virk
journal: JSES International
year: 2022
pmcid: PMC9998727
doi: 10.1016/j.jseint.2022.12.007
license: CC BY 4.0
---
# Prognostic factors associated with improvements in patient-reported outcomes in idiopathic adhesive capsulitis
## Body
Idiopathic adhesive capsulitis (IAC) or frozen shoulder is a common shoulder pathology affecting $2\%$-$5\%$ of the general population.34 IAC is characterized by pain and decreased range of motion (ROM) due to inflammation and fibrosis of the synovial capsule, classically evolving through 3 overlapping phases: the painful freezing stage, the stiff frozen stage, and the resolving thawing stage.34 While the etiology of primary IAC remains largely unknown, several associated risk factors associated with the diagnosis of IAC have been established including female gender, thyroid disorders, and diabetes.9,20,23,30,36 However, prognostic risk factors for successful nonoperative treatment of IAC are not as well defined.
IAC has a long natural history with substantial pain and limited shoulder function for several months, resulting in significant morbidity in those afflicted and a significant socioeconomic burden to the healthcare system.2,11 Furthermore, due to the idiopathic nature of the condition and a disabling, protracted course, patients often look for answers regarding their prognosis and recovery via nonsurgical treatment. Although IAC is a self-resolving condition, it is increasingly important to identify factors predictive of improved patient outcomes to guide clinical management and set patient expectations.2,22 Patient-reported outcome measures (PROMs) are crucial tools in measuring treatment efficacy and disease progression in this population but there exists a paucity in the literature of studies investigating factors associated with PROMs in IAC.22 The purpose of this study is to identify prognostic factors associated with improvements in PROMs in primary IAC, particularly patient-perceived upper extremity function and reductions in patient-perceived pain related to daily tasks. We hypothesize that patients’ perception of pain and their functional limitations are multifactorial, incorporating both modifiable and nonmodifiable factors.
## Abstract
### Background
The purpose of this study was to identify prognostic factors that are associated with improvements in patient-reported outcomes measures (PROMs) related to upper extremity function and pain in those suffering from idiopathic adhesive capsulitis.
### Methods
All patients treated conservatively for primary idiopathic adhesive capsulitis were identified from our institutional database between 2019 and 2021. Exclusion criteria included any patients treated surgically, follow-up less than one year, or incomplete survey results. PROMs including Patient-Reported Outcomes Measurement Information System (PROMIS) Upper Extremity Computer Adaptive Test Version 2.0 (P-UE), Pain Interference (P-Interference), Pain Intensity (P-Intensity), and visual analog scale (VAS) pain scores. They were obtained at initial consultation and at one year to assess patient-perceived impact of their condition. Multiple linear and multivariable logistic regressions were performed to identify factors associated with improvement in patient-perceived pain and shoulder function using final PROM scores and difference in PROM scores from initial consultation. An independent t-test was used to compare baseline and one-year minimum follow-up PROMs. Odds ratios and their $95\%$ confidence intervals were calculated for each factor; a P value of <.05 was considered statistically significant.
### Results
A total of 56 patients (40 females and 16 males) were enrolled in the study with an average age of 54.7 ± 7.7 years. A significant improvement ($P \leq .001$) was demonstrated at one-year minimum outcomes for P-UE, P-Interference, P-Intensity, and VAS scores. With respect to comorbid conditions, hypothyroidism [P-UE (β: 9.57, $$P \leq .006$$)] was associated with greater improvements in PROMs, while hyperlipidemia [P-UE (β: −4.13, $$P \leq .01$$) and P-Intensity (β: 2.40, $$P \leq .02$$)] and anxiety [P-UE (β: −4.13, $$P \leq .03$$)] were associated with poorer reported changes in PROMs. Female sex [P-UE (β: 4.03, $$P \leq .007$$) and P-Interference (β: −2.65, $$P \leq .04$$)] and employment in manual labor professions [P-Interference (β: −3.07, $$P \leq .01$$), P-Intensity (β: −2.92, $$P \leq .006$$), and VAS (β: −0.66, $$P \leq .03$$)] were associated with significantly better patient-perceived outcomes. Hispanic heritage was associated with higher reported changes of P-Intensity (β: 8.45, $$P \leq .004$$) and VAS (β: 2.65, $$P \leq .002$$).
### Conclusion
Patient-perceived improvements in PROMIS score during the natural history of adhesive capsulitis are likely multifactorial, with anxiety, hyperlipidemia, increased body mass index, and Hispanic heritage associated with reduced improvement in PROMIS scores.
## Study ethics
An internal Institutional Review Board Approval was granted for this study with all subjects providing informed consent prior to enrollment (s20-00287).
## Study design and cohort selection
This was a retrospective review conducted on a prospectively enrolled database of consecutive patients treated for IAC using International Classification of Diseases 10 (M75.00, M75.01, and M75.02) codes between August 16, 2019 and January 1, 2021. Baseline and final follow-up PROMs were assessed using Patient-Reported Outcomes Measurement Information System (PROMIS) scores obtained by telephone, e-mail, or in person during office follow-up visits as per patient preference.
Subjects were enrolled in this study if they met the following inclusion criteria: [1] age of 18 years or more at time of initial consultation, [2] underwent nonsurgical treatment for idiopathic adhesive capsulitis, [3] minimum of one-year follow-up from initial to final consultation, [4] completed the required PROMs at initial consultation, and [5] were able to provide informed consent. Subjects were excluded from the study if they were [1] deceased or lost to follow-up, [2] had adhesive capsulitis from a secondary etiology (eg, surgery, trauma), [3] were unable to communicate in English, [4] underwent any surgery on the affected shoulder, or [5] were unable to provide informed consent or complete the study surveys.
## Diagnosis and treatment
All patients underwent plain radiographic imaging (anteroposterior, axillary, and scapular Y view) to rule out causes of secondary adhesive capsulitis (eg, glenohumeral arthritis). The diagnosis of adhesive capsulitis was made by the treating physician based on pertinent clinical history and physical examination. All patients presenting with shoulder pain and global loss of ROM (both active and passive) and radiographs showing no signs of arthritis or calcific tendinitis were diagnosed with adhesive capsulitis.
The natural history of adhesive capsulitis was explained to each patient and the benefits and risks for each treatment option were discussed in detail. However, treatment was guided through shared decision-making based on patient preference. Patients were offered nonsurgical treatments in the form of oral anti-inflammatory medications (over the counter or prescription strength nonsteroidal anti-inflammatory medications, oral steroids), home exercise program or outpatient-supervised physical therapy, and intra-articular steroid injection. The final determination for intra-articular steroid injection was made by the patient.
## Patient-reported outcomes and factors measured
Subjects’ active ROM was measured individually by 2 trained research fellows using a goniometer. In addition, subjects were required to complete PROMIS Upper Extremity Computer Adaptive Test Version 2.0 (P-UE), PROMIS Pain Interference (P-Interference), and PROMIS Pain Intensity (P-Intensity) and a visual analog pain scale (VAS) at initial consultation and one-year minimum follow-up. Normalized across a scale of 0-100 with a mean value of 50 (SD ± 10) in reference to the general United States population, PROMIS has been established to accurately assess a patient's perceived symptomatology across all facets of orthopedics.4,7,19,22,31 *Using a* combination of item response theory and computer adaptive testing, PROMIS can reliably capture patient outcomes in less questions as compared to legacy scores.8,29,37 PROMIS-based PROMs are derived from 1 of 4 domains (physical, mental, global, and social health), each of which is further divided into a multitude of separate subdomains. In respect to P-UE, P-Interference, and P-Intensity, each of these belong to the physical health domain which is further comprised of subdomains such as physical function, physical activity, pain, fatigue, and sleep disturbances. As patients initially presenting for IAC are typically limited in ROM with concomitant pain in their affected shoulder, these 3 PROMs allow for accurate assessment of the two most common complaints in those afflicted. Specifically, P-UE evaluates one’s ability to use their shoulder, while P-Interference and P-Intensity assess the limitations brought about by their shoulder pain and the severity of pain they are currently experiencing, respectively. While higher scores in P-UE indicate superior upper extremity function, lower scores in PROMIS pain instruments (P-Interference, P-Intensity) are indicative of a patient experiencing less pain.
To determine which factors may influence improvement in PROMs, we evaluated a multitude of factors including previously reported risk factors for primary IAC.1, 2, 3,6,12,14,15,18,20,21,24,25,28,30,33,34,36 Existing literature revealed prior trauma, HLA-B27 positivity, age more than 40 years, female sex, thyroid disease, obesity, and autoimmune diseases to be predisposing risk factors for developing IAC.13,20,23,26,32,35,36 We in turn evaluated age, gender, body mass index (BMI), PROMIS scores and ROM at initial consultation (flexion, internal rotation, and external rotation), concomitant medical conditions (hypertension, hyperlipidemia, hypothyroidism, diabetes, anxiety, and depression), smoking status, marital status, ethnicity, manual versus nonmanual labor, dominant arm involvement, number of corticosteroid injections received, and time from symptom onset to first visit as these factors could all be ascertained from patient medical records and were investigated for a possible impact on the investigated outcomes.
## Statistical methods
All statistical analysis was performed in R-studio version 4.0.3 (R Studio, Boston, MA, USA) and SAS Studio Version 9.4 (SAS Institute, Cary, NC, USA). An independent t-test was used to compare baseline and one-year minimum follow-up PROMs. Multiple linear and multivariable logistic regressions were conducted to determine which factors were associated with patients’ improvement in pain and function as determined by the final PROM score and change in PROM scores from initial to final visit. Stepwise selection was applied to each regression model to identify variables that were the strongest predictors of each outcome modeled. Regression coefficients (β) were calculated for each predictor selected for a given model and predictors with $P \leq .05$ were considered statistically significant.
A post hoc analysis was performed using G∗Power version 3.1.9.7 (Heinrich Heine Universität, Düsseldorf, Germany) to determine achieved power on multivariable regression analysis. Our analysis was performed using postoperative PROMIS-UE as the outcome and the aforementioned factors as independent variables in a mixed multivariable regression. Using the partial R2 of 0.58, the effect size was determined to be 1.38. Using an F-test family parameter for multivariable regression models, the calculated effect size of 1.38, an alpha error probability of 0.05, and a total sample size of 57 patients, the achieved power for our analysis was determined to be 0.98.
## Cohort characteristics
A total of 56 patients were included in the study with an average age of 54.7 ± 7.7 (range 43-74 years) years. There were 40 females ($71.4\%$) and 16 males ($28.6\%$). Mean BMI was 27.1 (range 18.97-44.61). Throughout the course of treatment, $\frac{39}{56}$ ($69.6\%$) patients received a glenohumeral steroid injection with an average of 1.2 ± 0.4 (range 0-2) and the dominant arm was affected in $46.4\%$ ($\frac{26}{56}$) of subjects. A complete list of patient demographics is provided in Table I.Table IPatient demographics. Overall ($$n = 56$$)AgeBMI Mean (SD)54.7 (±7.7) Mean (SD)27.1 (±6.6)SexConcomitant medical conditions Female40 ($71.4\%$) Diabetes mellitus12 ($21.4\%$) Male16 ($28.6\%$) Hypertension16 ($28.6\%$)Race Hyperlipidemia19 ($33.9\%$) Asian9 ($16.1\%$) Hypothyroidism2 ($3.6\%$) Black or African American8 ($14.3\%$) Anxiety7 ($12.5\%$) White31 ($57.4\%$) Depression7 ($12.5\%$) Unknown/Declined to report8 ($14.3\%$) Manual laborer13 ($23.2\%$)Hispanic ethnicity4 ($7.1\%$) Nonmanual laborer43 ($76.8\%$)Smoking StatusCorticosteroid Injection39 ($69.6\%$) Never29 ($51.8\%$) Mean (SD)1.2 (±0.4) Current5 ($8.9\%$)Preoperative Range of Motion Former22 ($39.3\%$) Flexion (SD)119 (±31)Affected Shoulder External Rotation (SD)40 (±17) Dominant26 ($46.4\%$) Internal Rotation∗ (SD)3 (±1) Left28 ($50\%$)Baseline PROMIS Scores Right28 ($50\%$) Upper Extremity34.6 (±9.1) Bilateral0 ($0\%$) Pain Interference58.1 (±7.1)Time From Symptom Onset to First Visit Pain Intensity51.4 (±6.2) Mean Days (sd)104.1 (±99.3)Baseline VAS (SD)6.4 (±2.2)Time Between First Visit and Final Survey Mean Days (SD)528.8 (±128.9)SD, standard deviation; BMI, body mass index; PROMIS, Patient Reported Outcome Measure Instrument Survey; VAS, visual analog scale.∗Internal rotation reported as per the standardized surgeon measure of active internal rotation scale by Mollon et al.
## Prognostic factors of patient-reported outcomes at final follow-up
Comparison of baseline PROM scores with one-year minimum follow-up outcomes can be seen in Table II. A significant improvement was found when comparing these 2 time points for each PROM: P-UE (34.6-46.4, $P \leq .001$), P-Interference (58.1-47.0, $P \leq .001$), P-Intensity (51.4-36.9, $P \leq .001$), and VAS (6.4-1.6, $P \leq .001$). Factors associated with changes in final PROM values can be seen in Table III. Furthermore, factors with an associated impact on the change in PROM score from baseline are demonstrated in Table IV.Table IIPatient-reported outcomes measure. PROM testPROM score ± standard deviationSignificanceBaseline PROMIS UE34.6 (±9.1)$P \leq .001$Follow-up PROMIS UE46.4 (±12.5) Baseline PROMIS P-Interference58.1 (±7.1)$P \leq .001$ Follow-up PROMIS P-Interference47.0 (±9.9) Baseline PROMIS P-Intensity51.4 (±6.2)$P \leq .001$ Follow-up PROMIS P-Intensity36.9 (±8.5) Baseline VAS6.4 (±2.2)$P \leq .001$ Follow-up VAS1.6 (±2.4)PROMIS UE, Patient-Reported Outcomes Measurement Information System: Upper Extremity; PROMIS P-Interference, Patient-Reported Outcomes Measurement Information System: Pain Interference; PROMIS P-Intensity, Patient-Reported Outcomes Measurement Information System: Pain Intensity; VAS, visual analog scale. Table IIIRisk factors associated with changes in final PROMIS score. PROMIS testPredictorBeta coefficient (β)P valuePROMIS UE Follow-up ScoreUE score initial visit0.29.048BMI−0.49.04Hyperlipidemia−4.13.01Anxiety−4.19.03Hypothyroidism9.57.006Female4.03.007PROMIS P-Interference Follow-up ScoreBMI0.52.004Dominant side involved2.30.05Female−2.67.03Manual labor−3.07.01PROMIS P-Intensity Follow-up ScoreBMI0.41.01Dominant side involved2.07.05Manual labor−2.92.006VAS Follow-up ScoreBMI0.11.02Dominant side involved0.79.01Manual labor−0.66.03BMI, body mass index; PROMIS UE, Patient-Reported Outcomes Measurement Information System Upper Extremity; PROMIS P-Interference, Patient-Reported Outcomes Measurement Information System: Pain Interference; PROMIS P-Intensity, Patient-Reported Outcomes Measurement Information System: Pain Intensity; UE, upper extremity; VAS, visual analog scale. Table IVRisk factors associated with changes in the degree of change of PROMIS scores. PROMIS testPredicatorBeta coefficient (β)P valuePROMIS UE Change in ScoreUE score Initial Visit−0.71<.001∗BMI−0.49.04∗Hyperlipidemia−4.13.01∗Anxiety−4.13.03∗Hypothyroidism9.57.006∗Female4.03.007∗PROMIS P-Interference Change in ScoreP-Interference initial visit−0.84<.001∗BMI0.59.002∗Female−2.65.04∗Manual labor−2.64.03∗PROMIS P-Intensity Change in ScoreP-Intensity initial visit−0.78<.001∗Hyperlipidemia2.40.02∗Hispanic8.45.004∗Manual labor−2.25.03∗VAS Change in ScoreVAS initial visit−0.91<.001∗Hyperlipidemia1.04<.001∗Hispanic2.65.002∗BMI, body mass index; PROMIS UE, Patient-Reported Outcomes Measurement Information System Upper Extremity; PROMIS P-Interference, Patient-Reported Outcomes Measurement Information System: Pain Interference; PROMIS P-Intensity, Patient-Reported Outcomes Measurement Information System: Pain Intensity; UE, upper extremity; VAS, visual analog scale.∗Significant values.
Higher BMI was associated with greater patient P-Intensity (β: 0.41, $$P \leq .01$$), P-Interference (β: 0.52, $$P \leq .004$$), VAS score (β: 0.11, $$P \leq .02$$), and a lower P-UE score (β: −0.49, $$P \leq .04$$). Dominant side involvement was also shown to increase patient perception of interference to daily tasks and pain level [P-Interference (β: 2.30, $$P \leq .05$$), P-Intensity (β: 2.07, $$P \leq .05$$), VAS (β: 0.79, $$P \leq .01$$)]. Conversely, manual labor professions were found to be a protective factor, demonstrating reduced perception of pain intensity and interference [P-interference (β: −3.07, $$P \leq .01$$), P-Intensity (β: −2.92, $$P \leq .006$$), and VAS (β: −0.66, $$P \leq .03$$)]. Similarly, female gender was associated with a reduction in P-Interference (β: −2.67, $$P \leq .03$$) and an increase in P-UE (β: 4.03, $$P \leq .007$$) scores. After accounting for confounding effects, higher P-UE baseline scores (β: 0.29, $$P \leq .048$$) and history of hypothyroidism (β: 9.57, $$P \leq .006$$) were associated with more favorable outcomes, while history of hyperlipidemia (β: −4.13, $$P \leq .01$$) and anxiety (β: −4.19, $$P \leq .03$$) were associated with worse outcomes (Table III). Age, preoperative ROM, diabetes, depression, smoking status, marital status, race, and number of corticosteroid injections received did not demonstrate a significant impact associated with reported final score outcomes.
## Prognostic factors associated with improvement in patient-reported outcomes from baseline
We evaluated the impact that established risk factors had on the overall change in a patient’s PROMIS scores from initial evaluation to final follow-up. As a factor, baseline PROMIS scores were determined to be a significant predictor of the magnitude of change for final follow-up PROMs [P-UE (β: −0.71, $P \leq .001$), P-Interference (β: −0.84, $P \leq .001$), P-Intensity (β: −0.78, $P \leq .001$), and VAS (β: −0.91, $P \leq .001$)]. Hyperlipidemia demonstrated a significant detrimental impact on the change of multiple PROM scores [P-UE (β: −4.13, $$P \leq .01$$), P-Intensity (β: 2.40, $$P \leq .02$$), and VAS (β: 1.04, $P \leq .001$)]. Similarly, BMI demonstrated a negative impact on the magnitude of change in P-UE (β: −0.49, $$P \leq .04$$) and P-Interference score (β: 0.59, $$P \leq .002$$). As compared to other races, Hispanic patients demonstrated a significantly smaller magnitude of change in P-Intensity (β: 8.45, $$P \leq .004$$) and VAS scores (β: 2.65, $$P \leq .002$$) throughout the course of the study. Female sex demonstrated improvement in initial to final P-UE (β: 4.03, $$P \leq .007$$) and P-Interference scores (β: −2.65, $$P \leq .04$$). Similarly, hypothyroidism demonstrated a likewise association for P-UE (β: 9.57, $$P \leq .006$$) (Tables III and IV). Age, preoperative ROM, depression, diabetes, hypertension, smoking status, marital status, and number of corticosteroid injections received did not demonstrate significant impact associations with preoperative to postoperative outcomes.
## Discussion
In our study, we found one-year follow-up PROM scores and the difference from their baseline scores are influenced by a variety of patient-related factors. Notably, increased BMI, hyperlipidemia, and dominant side involvement were associated with a decrease in shoulder functionality and an increase in shoulder pain, while involvement in manual labor professions, lower initial PROM scores, and female sex were associated with an increase in shoulder function and decrease in shoulder pain.
Knowledge of these modifiable and nonmodifiable factors and their associated effects provide physicians with a better understanding of a patient’s expected outcomes, thereby allowing physicians to better anticipate a patient’s perception and set more realistic treatment expectations. BMI, a modifiable risk factor, was shown in our study to be consistently associated with decreased shoulder function and increased pain. Evidenced by our study, patients with higher BMI experience worse function and increased pain once afflicted by IAC, but this population is also at an increased odds of suffering from IAC, with one study of 2190 patients finding an increased odds ratio of 1.26 ($P \leq .001$).20 However, these results are not universally agreed upon with another study of 87 patients showing decreased BMI to be associated with an increased overall risk of suffering from IAC ($$P \leq .02$$), more specifically a $3\%$ increased risk of IAC for every kilogram of lower weight.36 Hyperlipidemia, another modifiable risk factor, has proven through our study to detrimentally impact shoulder function and increase pain. Similar to BMI, hyperlipidemia has previously been associated with an overall increased risk of suffering from IAC.26,35 A study of 28,748 records from The National Health Insurance Research Database of Taiwan showed hyperlipidemia to have a crude hazard ratio of 1.7 ($95\%$ confidence interval [CI] 1.61-1.79; $P \leq .001$) and an adjusted hazard ratio of 1.50 ($95\%$ CI, 1.41-1.59; $P \leq .001$).35 Further support of the correlative risk of IAC related to hyperlipidemia was confirmed through multivariate analysis performed by Lo et al on 1 million patients from the Taiwan National Health Insurance database. They found hyperlipidemia to be an independent risk factor associated with IAC, having a hazard ratio of 1.29 ($95\%$ CI 1.11-1.49, $P \leq .001$).26 Unfortunately, the external validity of these studies is limited due to the limited demographic diversity of the patient population.
In our study, baseline PROM scores, a nonmodifiable factor, proved to be a risk factor for final PROM scores and impactful on the magnitude of change among PROMs. Previous studies have investigated this relationship with the Simple Shoulder Test (SST), showing patients without diabetes with a higher initial SST score were more likely to have scored higher on their final SST ($P \leq .05$).28 Both their study, although more limited in patient population, and ours proved to show less favorable outcomes for those with higher patient-reported shoulder limitations at initial visit.28 Interestingly, sex, another nonmodifiable risk factor, was identified as a prognostic risk factor for a reduction in P-Interference and an increase in P-UE while also associated with a favorable change in P-UE and P-Interference. Like BMI and hyperlipidemia, this factor has been established as a risk factor for developing IAC, with up to $70\%$ of patients with IAC being female.32 Loosely related, Candela et al’s study of 278 patients found there to be a significant difference between males and females with regards to initial pain intensity measured through VAS, favoring higher pain in females.5 However, contradictory to our finding, a smaller study ($$n = 47$$) conducted by Fernandes et al showed that female gender was independently associated with higher disability of the arm, shoulder, and hand scores ($$P \leq .0004$$).16 Our study does not come without limitations that must be considered. First, patient responses to surveys such as P-UE, P-Interference, P-Intensity, and VAS can vary over time due to recent events and a patient’s health status that are beyond the scope of the condition in question.10,17,27 Second, PROMs were administered in the same order for each patient which can introduce a level of survey burden. In an effort to control for this, further studies can introduce randomization of survey order. Third, our study was limited to only English-speaking patients. Accommodating for this limitation is possible; however, it would require the computer-adaptive surveys be translated into various languages and further studies be performed to ensure the internal validity of the surveys is retained after translation. Fourth, our study was limited to a single surgeon at a single institution which may limit generalizability.
## Conclusion
Patient-perceived improvements in PROMIS score during the natural history of adhesive capsulitis are likely multifactorial, with anxiety, hyperlipidemia, increased BMI, and Hispanic heritage associated with reduced improvement in PROMIS scores.
## Disclaimers
Funding: No outside funding or grants were received in support of the completion of this study.
Conflicts of interest: The authors, their immediate families, and any research foundation with which they are affiliated have not received any financial payments or other benefits from any commercial entity related to the subject of this article.
## References
1. Austin D.C., Gans I., Park M.J., Carey J.L., Kelly I.V.J.D.. **The association of metabolic syndrome markers with adhesive capsulitis**. *J Shoulder Elbow Surg* (2014) **23** 1043-1051. DOI: 10.1016/j.jse.2013.11.004
2. Bagheri F., Ebrahimzadeh M.H., Moradi A., Bidgoli H.F.. **Factors associated with pain, disability and quality of life in patients suffering from frozen shoulder**. *Arch Bone Jt Surg* (2016) **4** 243. PMID: 27517070
3. Barbosa F., Swamy G., Salem H., Creswell T., Espag M., Tambe A.. **Chronic adhesive capsulitis (frozen shoulder): comparative outcomes of treatment in patients with diabetes and obesity**. *J Clin Orthop Trauma* (2019) **10** 265-268. DOI: 10.1016/j.jcot.2018.02.015
4. Baron J.E., Parker E.A., Wolf B.R., Duchman K.R., Westermann R.W.. **PROMIS versus legacy patient-reported outcome measures for sports medicine patients undergoing arthroscopic knee, shoulder, and hip interventions: a systematic review**. *Iowa Orthop J* (2021) **41** 58. PMID: 34924871
5. Candela V., Giannicola G., Passaretti D., Venditto T., Gumina S.. **Adhesive capsulitis of the shoulder: pain intensity and distribution**. *Musculoskelet Surg* (2017) **101** 153-158. DOI: 10.1007/s12306-017-0488-6
6. Cao W., Chen J., Pu J., Fan Y., Cao Y.. **Risk factors for the onset of frozen shoulder in middle-aged and elderly subjects within 1 year of discharge from a hospitalization that involved intravenous infusion: a prospective cohort study**. *Front Med (Lausanne)* (2022) **9**. DOI: 10.3389/fmed.2022.911532
7. Chen R.E., Papuga M.O., Nicandri G.T., Miller R.J., Voloshin I.. **Preoperative Patient-Reported Outcomes Measurement Information System (PROMIS) scores predict postoperative outcome in total shoulder arthroplasty patients**. *J Shoulder Elbow Surg* (2019) **28** 547-554. DOI: 10.1016/j.jse.2018.08.040
8. Cheung E.C., Moore L.K., Flores S.E., Lansdown D.A., Feeley B.T., Zhang A.L.. **Correlation of PROMIS with orthopaedic patient-reported outcome measures**. *JBJS Rev* (2019) **7** e9. DOI: 10.2106/JBJS.RVW.18.00190
9. Cho C.H., Jin H.J., Kim D.H.. **Comparison of clinical outcomes between idiopathic frozen shoulder and diabetic frozen shoulder after a single ultrasound-guided intra-articular corticosteroid injection**. *Diagnostics (Basel)* (2020) **10** 370. DOI: 10.3390/diagnostics10060370
10. Cook C.E., Wright A., Wittstein J., Barbero M., Tousignant-Laflamme Y.. **Five recommendations to address the limitations of patient-reported outcome measures**. *J Orthop Sports Phys Ther* (2021) **51** 562-565. DOI: 10.2519/jospt.2021.10836
11. Date A., Rahman L.. **Frozen shoulder: overview of clinical presentation and review of the current evidence base for management strategies**. *Future Sci OA* (2020) **6** FSO647. DOI: 10.2144/fsoa-2020-0145
12. Ding H., Tang Y., Xue Y., Yang Z., Li Z., He D.. **A report on the prevalence of depression and anxiety in patients with frozen shoulder and their relations to disease status**. *Psychol Health Med* (2014) **19** 730-737. DOI: 10.1080/13548506.2013.873814
13. D’Orsi G.M., Via A.G., Frizziero A., Oliva F.. **Treatment of adhesive capsulitis: a review**. *Muscles Ligaments Tendons J* (2012) **2** 70. PMID: 23738277
14. Ebrahimzadeh M.H., Moradi A., Bidgoli H.F., Zarei B.. **The relationship between depression or anxiety symptoms and objective and subjective symptoms of patients with frozen shoulder**. *Int J Prev Med* (2019) **10** 38. DOI: 10.4103/ijpvm.IJPVM_212_17
15. Erickson B.J., Shishani Y., Bishop M.E., Romeo A.A., Gobezie R.. **Adhesive capsulitis: demographics and predictive factors for success following steroid injections and surgical intervention**. *Arthrosc Sports Med Rehabil* (2019) **1** e35-e40. DOI: 10.1016/j.asmr.2019.07.005
16. Fernandes M.R.. **Patient-reported measures of quality of life and functional capacity in adhesive capsulitis**. *Rev Assoc Med Bras* (2017) **63** 347-354. DOI: 10.1590/1806-9282.63.04.347
17. Gordon D., Pines Y., Ben-Ari E., Rokito A.S., Kwon Y.W., Zuckerman J.D.. **Minimal clinically important difference, substantial clinical benefit, and patient acceptable symptom state of PROMIS upper extremity after total shoulder arthroplasty**. *JSES Int* (2021) **5** 894-899. DOI: 10.1016/j.jseint.2021.05.003
18. Green H.D., Jones A., Evans J.P., Wood A.R., Beaumont R.N., Tyrrell J.. **A genome-wide association study identifies 5 loci associated with frozen shoulder and implicates diabetes as a causal risk factor**. *PLoS Genet* (2021) **17**. DOI: 10.1371/journal.pgen.1009577
19. Hung M., Baumhauer J.F., Latt L.D., Saltzman C.L., SooHoo N.F., Hunt K.J.. **Validation of PROMIS® Physical Function computerized adaptive tests for orthopaedic foot and ankle outcome research**. *Clin Orthop Relat Res* (2013) **471** 3466-3474. DOI: 10.1007/s11999-013-3097-1
20. Kingston K., Curry E.J., Galvin J.W., Li X.. **Shoulder adhesive capsulitis: epidemiology and predictors of surgery**. *J Shoulder Elbow Surg* (2018) **27** 1437-1443. DOI: 10.1016/j.jse.2018.04.004
21. Ko Y.W., Park J.H., Youn S.M., Rhee Y.G., Rhee S.M.. **Effects of comorbidities on the outcomes of manipulation under anesthesia for primary stiff shoulder**. *J Shoulder Elbow Surg* (2021) **30** e482-e492. DOI: 10.1016/j.jse.2020.11.007
22. Kolade O., Ghosh N., Luthringer T.A., Rosenthal Y., Kwon Y.W., Rokito A.S.. **Correlation of patient reported outcome measurement Information system (PROMIS) with American shoulder and Elbow surgeon (ASES), and Constant (CS) scores in idiopathic adhesive capsulitis**. *J Shoulder Elbow Surg* (2021) **30** 554-560. DOI: 10.1016/j.jse.2020.05.040
23. Le H.V., Lee S.J., Nazarian A., Rodriguez E.K.. **Adhesive capsulitis of the shoulder: review of pathophysiology and current clinical treatments**. *Shoulder Elbow* (2017) **9** 75-84. DOI: 10.1177/1758573216676786
24. Lesevic M., Awowale J.T., Moran T.E., Diduch D.R., Brockmeier S.F., Werner B.C.. **Immediate pain relief at time of corticosteroid injection for idiopathic adhesive capsulitis as a predictor of eventual outcomes**. *Orthop J Sports Med* (2021) **9**. DOI: 10.1177/23259671211019353
25. Linaker C.H., Walker-Bone K.. **Shoulder disorders and occupation**. *Best Pract Res Clin Rheumatol* (2015) **29** 405-423. DOI: 10.1016/j.berh.2015.04.001
26. Lo S.F., Chu S.W., Muo C.H., Meng N.H., Chou L.W., Huang W.C.. **Diabetes mellitus and accompanying hyperlipidemia are independent risk factors for adhesive capsulitis: a nationwide population-based cohort study (version 2)**. *Rheumatol Int* (2014) **34** 67-74. DOI: 10.1007/s00296-013-2847-4
27. Louwerens J.K., van den Bekerom M.P., van Royen B.J., Eygendaal D., van Noort A., Sierevelt I.N.. **Quantifying the minimal and substantial clinical benefit of the Constant-Murley score and the Disabilities of the Arm, Shoulder and Hand score in patients with calcific tendinitis of the rotator cuff**. *JSES Int* (2020) **4** 606-611. DOI: 10.1016/j.jseint.2020.05.001
28. Rill B.K., Fleckenstein C.M., Levy M.S., Nagesh V., Hasan S.S.. **Predictors of outcome after nonoperative and operative treatment of adhesive capsulitis**. *Am J Sports Med* (2011) **39** 567-574. DOI: 10.1177/0363546510385403
29. Rose M., Bjorner J.B., Becker J., Fries J.F., Ware J.E.. **Evaluation of a preliminary physical function item bank supported the expected advantages of the Patient-Reported Outcomes Measurement Information System (PROMIS)**. *J Clin Epidemiol* (2008) **61** 17-33. DOI: 10.1016/j.jclinepi.2006.06.025
30. Sarasua S.M., Floyd S., Bridges W.C., Pill S.G.. **The epidemiology and etiology of adhesive capsulitis in the US Medicare population**. *BMC Musculoskelet Disord* (2021) **22** 1-2. DOI: 10.1186/s12891-021-04704-9
31. Schneider M.B., Kung J.E., Zhang T., Rocca M.S., Foster M.J., Meredith S.J.. **Patient perception of being “completely better” after Anterior cruciate Ligament Reconstruction**. *Am J Sports Med* (2022) **50** 1215-1221. DOI: 10.1177/03635465221074331
32. Sheridan M.A., Hannafin J.A.. **Upper extremity: emphasis on frozen shoulder**. *Orthop Clin* (2006) **37** 531-539. DOI: 10.1016/j.ocl.2006.09.009
33. Sung C.M., Jung T.S., Park H.B.. **Are serum lipids involved in primary frozen shoulder?: a case-control study**. *J Bone Joint Surg Am* (2014) **96** 1828-1833. DOI: 10.2106/JBJS.M.00936
34. Toprak M., Erden M.. **Sleep quality, pain, anxiety, depression and quality of life in patients with frozen shoulder**. *J Back Musculoskelet Rehabil* (2019) **32** 287-291. DOI: 10.3233/BMR-171010
35. Wang J.Y., Liaw C.K., Huang C.C., Liou T.H., Lin H.W., Huang S.W.. **Hyperlipidemia is a risk factor of adhesive capsulitis: real-world evidence using the Taiwanese National health Insurance research database**. *Orthop J Sports Med* (2021) **9**. DOI: 10.1177/2325967120986808
36. Wang K., Ho V., Hunter-Smith D.J., Beh P.S., Smith K.M., Weber A.B.. **Risk factors in idiopathic adhesive capsulitis: a case control study**. *J Shoulder Elbow Surg* (2013) **22** e24-e29. DOI: 10.1016/j.jse.2012.10.049
37. Wong L.H., Meeker J.E.. **The promise of computer adaptive testing in collection of orthopaedic outcomes: an evaluation of PROMIS utilization**. *J Patient Rep Outcomes* (2022) **6** 1-5. DOI: 10.1186/s41687-021-00407-w
|
---
title: Fenretinide inhibits obesity and fatty liver disease but induces Smpd3 to increase
serum ceramides and worsen atherosclerosis in LDLR−/− mice
authors:
- Dawn Thompson
- Shehroz Mahmood
- Nicola Morrice
- Sarah Kamli-Salino
- Ruta Dekeryte
- Philip A. Hoffmann
- Mary K. Doherty
- Philip D. Whitfield
- Mirela Delibegović
- Nimesh Mody
journal: Scientific Reports
year: 2023
pmcid: PMC9998859
doi: 10.1038/s41598-023-30759-w
license: CC BY 4.0
---
# Fenretinide inhibits obesity and fatty liver disease but induces Smpd3 to increase serum ceramides and worsen atherosclerosis in LDLR−/− mice
## Abstract
Fenretinide is a synthetic retinoid that can prevent obesity and improve insulin sensitivity in mice by directly altering retinol/retinoic acid homeostasis and inhibiting excess ceramide biosynthesis. We determined the effects of Fenretinide on LDLR−/− mice fed high-fat/high-cholesterol diet ± Fenretinide, a model of atherosclerosis and non-alcoholic fatty liver disease (NAFLD). Fenretinide prevented obesity, improved insulin sensitivity and completely inhibited hepatic triglyceride accumulation, ballooning and steatosis. Moreover, Fenretinide decreased the expression of hepatic genes driving NAFLD, inflammation and fibrosis e.g. Hsd17b13, Cd68 and Col1a1. The mechanisms of Fenretinide’s beneficial effects in association with decreased adiposity were mediated by inhibition of ceramide synthesis, via hepatic DES1 protein, leading to increased dihydroceramide precursors. However, Fenretinide treatment in LDLR−/− mice enhanced circulating triglycerides and worsened aortic plaque formation. Interestingly, Fenretinide led to a fourfold increase in hepatic sphingomyelinase Smpd3 expression, via a retinoic acid-mediated mechanism and a further increase in circulating ceramide levels, linking induction of ceramide generation via sphingomyelin hydrolysis to a novel mechanism of increased atherosclerosis. Thus, despite beneficial metabolic effects, Fenretinide treatment may under certain circumstances enhance the development of atherosclerosis. However, targeting both DES1 and Smpd3 may be a novel, more potent therapeutic approach for the treatment of metabolic syndrome.
## Introduction
Obesity has reached epidemic proportions worldwide and contributes to the pathophysiology of many disease states including type 2 diabetes, cardiovascular disease (CVD), and non-alcoholic fatty liver disease (NAFLD) collectively referred to as metabolic syndrome. Indeed, many of these diseases have overlapping secondary pathologies and are risk factors for the development of further complications. For example, defective insulin signalling associated with dyslipidaemia and chronic low-grade inflammation results in increased risk of developing type2 diabetes, NAFLD and atherosclerosis1–3. Specifically, NAFLD, is now the most common liver disease in the Western world, characterised by an accumulation of triglycerides that can develop from simple steatosis to non-alcoholic steatohepatitis (NASH) and progress to cirrhosis and hepatocellular carcinoma4,5. Although there are therapeutics available for type 2 diabetes and CVD, there are no approved treatments currently available for NAFLD other than dietary and lifestyle intervention/changes. Since it is becoming increasingly common for patients to present with multiple overlapping co-morbidities, there is an urgent need for new therapeutics targeting the mechanistic causes of these diseases to halt the predicted rise in cases.
Increased lipotoxicity and accumulation of the bioactive mediator ceramide has been attributed to be a major player in the progression of obesity-associated metabolic diseases6. Ceramide belongs to the sphingolipid class of lipid mediators, the generation of which is tightly controlled by a series of enzymes through either de novo synthesis, sphingomyelin hydrolysis or through the salvage pathway. Indeed, there have been numerous studies reporting high-fat diet feeding leading to increased de novo sphingolipid synthesis resulting in the accumulation of ceramide in several tissues such as liver, adipose tissue, skeletal muscle and the heart7. Non-human primates fed a western diet to induce obesity and type 2 diabetes exhibited increased circulating ceramides8,9. Circulating ceramides are also considered important risk factors for cardiovascular disease10. Mechanistically, excess ceramide has been shown to lead to defective insulin signalling due to an impairment of the downstream effector cascades such as activation of Akt or GSK3B7,11,12. Therefore, inhibiting ceramide accumulation is an attractive target for manipulation.
Fenretinide (FEN, also known as N-(4-hydroxyphenyl)retinamide or 4-HPR)) is a synthetic derivative of retinoic acid and has been investigated as a potential therapeutic for metabolic syndrome. In previous studies FEN treatment led to decreased weight gain and adiposity and improved glucose homeostasis and insulin sensitivity in association with decreased accumulation of hepatic triglycerides in high-fat diet fed mice13–18. The mechanism of FEN action has been attributed to alterations in retinol homeostasis and retinoic acid signalling and prevention of lipotoxicity by directly inhibiting the elevation of several ceramide species both in vitro19 and in vivo, in adipose14, liver tissue17 and skeletal muscle20. Interestingly, Smpd3 encodes for a type 2-neutral sphingomyelinase (nSMase2) that has been identified as transcriptionally induced by retinoic acid21,22. Smpd3/nSMase2 is key to an alternative ceramide generation pathway (via sphingomyelin hydrolysis) that has recently been linked to atherosclerosis via regulation of serum ceramide levels23.
There have been several models, both dietary and genetic, used to study the effect of obesity on either type 2 diabetes, atherosclerosis or NAFLD, and some of these also to determine the mechanism of FEN action. Since many of the diseases associated with obesity have overlapping pathologies, we sought to determine if FEN could improve several disorders in LDLR−/− mice, traditionally a model for atherosclerosis with a similar lipoprotein profile to humans24. On a high-fat/high-cholesterol diet, LDLR−/− mice become obese, develop insulin resistance and accumulate hepatic triglycerides, with inflammation and hepatic fibrosis associated with NAFLD progression to NASH25,26. Given previous research has shown blocking ceramide biosynthesis is beneficial6,9, we hypothesized that FEN could be used as a novel intervention for NAFLD/NASH and atherosclerosis, in addition to its beneficial effects of decreased adiposity and improved insulin sensitization, via prevention of excess ceramide accumulation.
## Animal studies
All animal procedures were performed under a project licence (PPL P94B395E0) approved by the U.K. Home Office under the Animals (Scientific Procedures) Act 1986 and the University of Aberdeen ethics review board. Studies were performed following the recommendations in the ARRIVE guidelines under guidance by the Veterinary Surgeon and Animal Care and Welfare Officers of the institutional animal research facility. Thus, all methods were performed in accordance with the relevant guidelines and regulations. Male LDLR−/− mice, aged 4–6 weeks, were purchased from The Jackson Laboratory (supplied by Charles River UK Ltd), male and female ApoE−/− mice were bred in-house (University of Aberdeen). All mice were fed chow diet until 12 weeks of age then placed into three groups and fed the following diets (all Research Diets Inc.) to induce atherogenesis and NAFLD for 14 weeks: control ($10\%$ kCal fat D14121001) or high-fat/high-cholesterol diet (HFD, $40\%$ kCal fat from cocoa butter and soybean oil, $34.5\%$ kcal and $5.5\%$ kcal respectively, plus $1.25\%$ cholesterol, Clinton/Cybulsky D12108C) + /- $0.04\%$ Fenretinide (FEN-HFD, D18061502,16,27–29). Mice were maintained at 22–24 °C on 12-h light/dark cycle with free access to food/water. At week 14, mice were fasted for 5 h and injected intraperitoneally with either saline or insulin (10 mU/g body weight) for 10 min prior to CO2-induced anaesthesia followed by cervical dislocation. Heart and aortic tissues were collected for histological analysis. Peripheral metabolic tissues (liver, muscle and white adipose tissue (WAT)) were frozen in liquid nitrogen and stored at − 80 °C until subsequent analysis.
## Glucose and insulin tolerance tests
Mice were fasted for 5 h prior to commencement of glucose or insulin tolerance tests (GTT and ITT, respectively). Briefly, baseline glucose levels were sampled from tail blood using glucose meters (AlphaTRAK, Abbott Laboratories, Abbot Park, IL, USA). Subsequently mice were injected intraperitoneally with $20\%$ glucose (w/v) or insulin (0.75 mU/g body weight) and blood glucose measured at 15-, 30-, 60- and 90-min post-injection.
## Body fat mass composition
The body composition of mice was analysed using an Echo MRI 3-in-1 scanner (Echo MRI, Houston, TX, USA).
## Immunoblotting
Frozen liver tissues were homogenised in 400 µl of ice-cold RIPA buffer (10 mM Tris–HCl pH 7.4, 150 mM NaCl, 5 mM EDTA pH 8.0, 1 mM NaF, $0.1\%$ SDS, $1\%$ Triton X-100, $1\%$ Sodium Deoxycholate with freshly added 1 mM NaVO4 and protease inhibitors) using a PowerGen 125 homogeniser and lysates normalised to 1 µg per 1 µl. Proteins were separated on a 4–$12\%$ Bis–Tris gel by SDS-PAGE and transferred onto nitrocellulose membrane.
Membranes were probed for the following; phospho-AKT (Ser 473, cat: 4060), total Akt (cat: 4691), phospho-S6 (Ser $\frac{235}{236}$, cat: 4858), total S6 (cat: 2217), phospho-AMPK (Thr 172, cat: 2535), total AMPK (cat: 5832) and GAPDH (cat: 5174) (all Cell Signaling Technology), DEGS1 (cat: ab185237, Abcam), RBP4 (Dako), or IR β-chain (Santa Cruz Biotechnology). ApoB 48, ApoB 100 (Meridian Life Sciences UK, cat: K23300R) and Vinculin (Cell Signaling Technology, cat: 13901) were separated on a $6\%$ Tris–Glycine gel. Anti-rabbit and anti-mouse horse radish peroxidase (HRP) conjugated antibodies were from Anaspec. Primary and secondary antibodies were used at 1:1000 and 1:5000 respectively.
Blots used in figures are all compliant with the digital image and integrity policies of Nature publishing and Scientific Reports journal. Western blot membranes were cut at approximate molecular weight (± 20 kDa) of target protein before incubation of primary antibodies. Equal numbers of representative samples from all treatment groups were run on multiple gels/blots to accommodate all samples. Images obtained were minimally processed. Image analysis and quantification with normalisation to loading control protein was performed within the same membrane and then data combined for graphical representation. No direct quantitative comparisons between samples on different gels/blots were performed.
## RNA extraction and qPCR
Frozen tissues were lysed in TRIzol reagent (Sigma, U.K.) and RNA isolated using phenol/chloroform extraction according to manufacturer’s instructions. RNA was then synthesized into cDNA (tetrokit, Bioline) and subjected to qPCR analysis using SYBR green and LightCycler 480 (Roche). Gene expression was determined relative to the reference gene Nono or Ywaz. Details of primer sequences can be found in Supplemental Table 1.
## Histology
Liver tissues were sectioned and stained to assess steatosis (Haemotoxylin and Eosin (H&E)) or fibrosis (picrosirius red). Immediately following cervical dislocation, hearts with attached aortic root were immersed in formalin and stored at 4 °C for 24 h, before being transferred to PBS until further analysis. Hearts were bisected to remove the lower ventricles, frozen in OCT and subsequently sectioned at 5 µm intervals until the aortic sinus was reached. Sections of aortic roots of comparable anatomical position were obtained by NHS Grampian pathology unit. A single section from each mouse ($$n = 4$$–5) was mounted and stained with oil red O to assess plaque formation. The descending aorta was prepared for en face staining. Briefly aortas were trimmed of perivascular adipose tissue, cut longitudinally, and stained with Sudan IV to assess plaque formation. Images were captured using a light microscope and plaque formation quantified using Image J software. Plaque formation in aortic root sections was total area measured, whereas for en face plaque staining was calculated as a percentage of the total surface area of the vessel.
## Liver triglyceride assay
50–100 mg of frozen liver tissue was homogenised in 1 ml of PBS and frozen in liquid nitrogen to enable further cell lysis. Samples were thawed, centrifuged briefly (15 s at 7500 rpm) and the supernatant (including the lipid layer on top) transferred to a fresh tube. Total triglycerides were measured in homogenates according to manufacturer’s instructions (Sigma, cat: MAK266).
## Serum Analysis
Blood was collected during terminal procedures after fasting (5 h) and spun to isolate serum, then stored at − 80 °C. Serum samples were subsequently analysed for total cholesterol and triglycerides (Sigma, cat: MAK043 and MAK266 respectively) or Insulin and Leptin (Crystal Chem, cat 90080 and 90030 respectively) according to manufacturer’s instructions.
## Quantification of liver dihydroceramides and ceramides
Extraction of liver lipids was performed according to the method described by Folch et al.30. Dihydroceramides and ceramides and were isolated by solid phase extraction chromatography using C12:0 dihydroceramide and C17:0 ceramide (Avanti Polar Lipids, Alabaster, Al, USA) as internal standards. Samples were analysed by liquid chromatography-mass spectrometry (LC–MS) using a Thermo Exactive Orbitrap mass spectrometer (Thermo Scientific, Hemel Hempsted, UK) equipped with a heated electrospray ionization (HESI) probe and coupled to a Thermo Accela 1250 ultra-high-pressure liquid chromatography (UHPLC) system. Samples were injected onto a Thermo Hypersil Gold C18 column (2.1 mm by 100 mm; 1.9 μm) maintained at 50 °C. Mobile phase A consisted of water containing 10 mM ammonium formate and $0.1\%$ (vol/vol) formic acid. Mobile phase B consisted of a 90:10 mixture of isopropanol-acetonitrile containing 10 mM ammonium formate and $0.1\%$ (vol/vol) formic acid. The initial conditions for analysis were $65\%$ mobile phase A, $35\%$ mobile phase B and the percentage of mobile phase B was increased from 35 to $65\%$ over 4 min, followed by $65\%$ to $100\%$ over 15 min, with a hold for 2 min before reequilibration to the starting conditions over 6 min. The flow rate was 400 μl/min and samples were analyzed in positive ion mode. The LC–MS data were processed with Thermo Xcalibur v2.1 (Thermo Scientific) with signals corresponding to the accurate mass-to-charge ratio (m/z) values for dihydroceramide and ceramide molecular species extracted from raw data sets with the mass error set to 5 ppm. Quantification was achieved by relating the peak area of the dihydroceramide and ceramide lipid species to the peak area of their respective internal standard. All values were normalised to the wet weight of liver.
## Statistical analysis
Data are presented as mean+/± S.E.M. Group sizes were determined by performing a power calculation to lead to an $80\%$ chance of detecting a significant difference (P ≤ 0.05). For both in vivo and ex vivo data, each n value corresponds to a single mouse. Statistical analyses were performed by using one-way or two-way ANOVA followed by Bonferroni multiple-comparison tests to compare the means of three or more groups. Variances were similar between groups. In all figures, */#p ≤ 0.05, **/##p ≤ 0.01, ***/###p ≤ 0.001, ****/####p ≤ 0.0001. All analyses were performed using GraphPad Prism (GraphPad Software).
## Data and resource availability
The data sets generated and analyzed during the current study are available from the corresponding author upon reasonable request including RNA-*Seq data* referred to briefly here in the discussion, previously17 and manuscript in preparation. Fenretinide used in the current study is available from the corresponding author upon reasonable request and for non-commercial purposes17.
## Fenretinide prevents diet-induced obesity and improves insulin sensitivity in LDLR−/− mice fed an atherogenic diet
Male LDLR-/- mice were fed an obesogenic plus atherogenic, high-fat/high-cholesterol diet (HFD) + /- $0.04\%$ FEN (FEN-HFD) or control diet for 14 weeks. All mice gained body weight until about week 8 when HFD mice continued to gain body weight but FEN-HFD mice and control mice body weights reached a similar plateau for the remainder of the study (Fig. 1A). This inhibition of body weight gain was due specifically to an inhibition of adiposity in FEN-HFD mice and not due to alterations in lean mass (Fig. 1B,C). Serum leptin levels were markedly elevated in HFD mice whereas in FEN-HFD mice levels were similar to control mice (Table 1).Figure 1Fenretinide prevents diet-induced obesity and improves insulin sensitivity in the LDLR−/− mice. ( A) Weekly body weights of LDLR−/− mice fed either control ($$n = 14$$, $10\%$ kCal fat) or high-fat/high-cholesterol diet ± $0.04\%$ Fenretinide (HFD, $40\%$ kCal fat plus $1.25\%$ cholesterol or FEN-HFD, both $$n = 18$$). Body composition total body fat mass (B) and lean mass (C) at week 5 and week 13 diet. ( D) Serum RBP4 levels ($$n = 6$$ per group). ( E) Insulin tolerance test (ITT) at week 12 ($$n = 10$$ per group). At 14 weeks diet, liver tissue western blot analysis (F, representative image) and quantification of bands (G, $$n = 6$$ per group) from LDLR-/- mice injected with insulin (10 mU/g body weight, 10 min). Equal numbers of representative samples from all treatment groups were run on multiple gels/blots to accommodate all samples. Image analysis and quantification with normalisation to loading control protein was performed within the same membrane and then data combined for graphical representation. Data are all compliant with digital image and integrity policies. ( H) Glucose tolerance tests (GTT) at week 11. ( I) Hepatic expression of RA/RAR target genes ($$n = 8$$ per group). Data are represented as mean ± S.E.M. and analysed by either Two-way (A–C,E,H) or one-way ANOVA (D,G,I) followed by Bonferroni multiple comparison t-tests where *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001 and ****p ≤ 0.0001 (control compared to HFD) or #p ≤ 0.05 and ##p ≤ 0.01 and ###p ≤ 0.001 (HFD compared to FEN-HFD).
As expected, FEN treatment decreased serum RBP4 levels compared to levels in control and HFD LDLR−/− mice (Fig. 1D). However, several classic molecular markers of functional white adipose tissue (e.g. PPARγ, GLUT4) were largely unaltered in LDLR−/− mice fed HFD + /- FEN, although there was a near $50\%$ decrease in white adipose PEPCK, adiponectin, resistin and RBP4 (Supplemental Fig. 1). HFD induced physiological insulin resistance (Fig. 1E) and decreased acute hepatic insulin signalling to Akt (Fig. 1F,G). Whereas FEN treatment, resulted in improved insulin sensitivity and rescued hepatic Akt phosphorylation in response to insulin (Fig. 1F,G).
Despite these markedly beneficial physiological effects (decreased adiposity and improved insulin sensitivity) other parameters of glucose homeostasis were not similarly improved with FEN treatment. Basal serum glucose and serum insulin levels (in the 5-h fasted state) were similar in all three LDLR−/− groups and FEN treatment increased glucose intolerance compared to both HFD and control LDLR−/− mice (Table 1 and Fig. 1H). This effect of FEN was not attributable to changes in hepatic PEPCK, a known RA/RAR target gene (Fig. 1I) despite induction of hepatic Cyp26A1 and LRAT, classic RA/RAR target genes (Fig. 1I). In skeletal muscle, acute hepatic insulin signalling to Akt was not altered by HFD + /- FEN. FEN-HFD fed mice had significantly less total IR protein when compared to HFD (Supplemental Fig. 1). These initial findings suggested that the beneficial effects FEN treatment in LDLR−/− mice may have been more attributed to action in the liver than other insulin sensitive tissues. Table 1Serum and Tissue measurements.controlHFDFEN-HFDSerum levels Basal Glucose Week 11 (mmol/L)9.32 ± 0.579.51 ± 0.7611.12 ± 0.51 Basal Glucose Week 12 (mmol/L)7.65 ± 0.658.67 ± 0.579.75 ± 0.56 Insulin (ng/ml)1.23 ± 0.370.82 ± 0.120.69 ± 0.07 Leptin (ng/ml)6.94 ± 3.0315.79 ± 3.757.56 ± 1.43 Cholesterol (µg)5.96 ± 1.0017.21 ± 1.60 ****17.19 ± 0.57 **** Triglyceride (mg/ml)1.76 ± 0.1954.41 ± 0.43 ***7.95 ± 0.60 ****/####Tissue levels Liver triglyceride (µg/mg)14.24 ± 1.4735.28 ± 6.34 **13.82 ± 0.97##Data are represented as mean ± S.E.M. ($$n = 8$$ per group) and analysed by One-way ANOVA followed by Bonferroni multiple comparison t-tests where **p ≤ 0.01, ***p ≤ 0.001 and ****p ≤ 0.0001 (control compared to HFD) or ## p ≤ 0.01 and #### p ≤ 0.0001 (HFD compared to FEN-HFD).
## Fenretinide inhibits hepatic triglyceride accumulation and development of steatosis and alters hepatic metabolic gene expression in LDLR−/− mice fed an atherogenic diet
LDLR−/− mice are a recognised model of NAFLD, with high-fat feeding known to increase hepatic triglyceride accumulation26. HFD resulted in a 2.5-fold increase in triglyceride content in the livers of LDLR−/− mice (Fig. 2A). FEN treatment completely prevented intrahepatic triglyceride accumulation to levels similar to those in control mice. HFD feeding also led to a profound change in hepatic morphology compared to control mice including substantial lipid droplet accumulation (Fig. 2B). Whereas, FEN-HFD mice exhibited normal liver histology with the absence of lipid droplet accumulation within hepatocytes. Hepatic lipid homeostasis is maintained via a network of key transcription factors such as PPARα, LXR and SREBP which regulate the expression of genes involved in fatty acid synthesis, oxidation and transport. Dysregulation of this network in response to excess nutrition or genetic perturbations causes excess hepatic lipid accumulation and thus NALFD.Figure 2Fenretinide inhibits hepatic triglyceride accumulation and development of steatosis. ( A) Hepatic triglyceride levels in LDLR−/− mice. ( B) Representative H&E staining of hepatic tissues. Scale bar is 20 µm. ( C–F) Gene expression in liver tissue ($$n = 8$$ per group). Data are represented as mean + S.E.M. and analysed by one-way ANOVA followed by Bonferroni multiple comparison t-tests where *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001 and ****p ≤ 0.0001 (control compared to HFD) or #p ≤ 0.05 and ##p ≤ 0.01 (HFD compared to FEN-HFD).
HFD (including high cholesterol) feeding ± FEN of LDLR−/− mice increased hepatic expression of LXR target genes, Srebp1c, Abca1 and Abcg1 and PPARα target genes Mogat, Vldr, Cd36 and Abcc3 (Mrp3) compared to control diet mice (Fig. 2C). HFD ± FEN did not affect the expression of PPARα, LXR or RXR transcription factors in liver (Fig. 2D). Hepatic Dgat1, Acadm and Acox1, also all PPARα target genes, were unaffected by HFD but were modestly increased by FEN treatment. Cpt1 was not altered by either diet. However, FEN suppressed the statin target Hmgcr in liver (Fig. 2E) without affecting serum cholesterol levels (Table 1). HFD increased Hmgcr and Abcc3 in white adipose tissue and FEN trended to prevent these increases, in addition, FEN suppressed adipose Cd36 suggesting FEN also decreased adipose cholesterol in association with decreased adiposity (Supplemental Fig. 1).
Several human SNP/GWAS studies and more recent multi-omics data analyses have identified genes that have been described as key drivers of NAFLD31. Of these, HFD did not affect the gene expression of hepatic Pklr, Pnpla3, Tm6sf2 or Hsd17b13 compared to control LDLR−/− mice. However, FEN treatment resulted in a significant decrease in both Tm6sf2 and Hsd17b13 expression when compared to control mice (Fig. 2F). FEN had no effect on Pklr or Pnpla3 in this disease model (Fig. 2F see discussion).
## Fenretinide alters hepatic inflammatory and fibrotic gene expression
Persistent excess lipid accumulation is associated with a pro-inflammatory environment and the activation of hepatic stellate cells, the development of fibrosis and the progression to NASH, a more severe disease state. Indeed, HFD resulted in an increase in expression of the pro-inflammatory cytokine TNFα, the macrophage marker Cd68 and profibrogenic signalling factor TGF-β that participates in hepatic stellate cell activation32,33. FEN treatment significantly inhibited the increase in Cd68 and trended to inhibit TNFα and TGF-β thus suggestive of a less pro-inflammatory, pro-fibrogenic environment (Fig. 3A). HFD did not alter the expression of IL-6 or Mcp-1 compared to control LDLR−/− however, FEN treatment resulted in approximately 2.5-fold and 8-fold increase in gene expression respectively (Fig. 3A). HFD ± FEN did not alter expression of anti-inflammatory gene, IL-10.Figure 3Fenretinide alters pro-inflammatory and fibrotic gene expression in liver. Hepatic mRNA expression of (A) inflammation and (B) fibrosis genes in LDLR−/− mice ($$n = 8$$ per group). ( C) *Representative picrosirius* red staining of hepatic tissue. Scale bar is 50 µm. Data are represented as mean + S.E.M. and analysed by one-way ANOVA followed by Bonferroni multiple comparison t-tests where *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001 and ****p ≤ 0.0001 (control compared to HFD) or #p ≤ 0.05 and ##p ≤ 0.01 and ####p ≤ 0.0001 (HFD compared to FEN-HFD).
HFD resulted in significant increases in the expression of genes driving fibrosis and tissue remodelling such as collagen (Col1a1, Col4a1), matrix metalloproteinases (Mmp2) and the tissue inhibitors of Mmps (Timp1, Timp2; Fig. 3B). FEN treatment almost completely inhibited the expression of all these genes to levels similar to those in control LDLR−/− mice. A similar decrease in pro-inflammatory, macrophage and fibrosis genes with FEN treatment was determined in ApoE−/− mice treated with HFD ± FEN (Supplemental Fig. 2). In contrast to the inhibition of these genes (TNFα, Col1a1etc), FEN induced a fivefold increase in Mmp9 (Fig. 3B) which is indicative of retinoid-specific signalling and increased clearance of pro-fibrotic proteins34. However, despite these improvements in response to FEN treatment, there were no differences in hepatic fibrosis between all three diet groups in LDLR−/− mice when determined by histologic stain picrosirius red (Fig. 3C).
## Fenretinide treatment inhibits de novo ceramide synthesis and lipotoxicity
Enzymes involved in ceramide biosynthesis, dihydroceramide desaturase, (DES1) and ceramide synthase (CerS)-6 have been implicated with increased ceramide production mediating obesity associated metabolic dysregulation in mice and humans35. Hence, we next examined whether FEN treatment could inhibit the enzymes controlling de novo ceramide synthesis and thus lipotoxicity in the development of NAFLD/NASH in LDLR−/− mice.
HFD increased DES 1 in LDLR−/− mice (Fig. 4A,B), whereas FEN treatment prevented this increase so that protein levels were comparable to those control mice. Gene expression of hepatic dihydroceramide desaturase, Degs1, was unchanged with diet (Fig. 4C). However, HFD did trend to increase the hepatic Cers6 and Cers2 and FEN significantly decreased expression of Cers6 (Fig. 4C).Figure 4Fenretinide inhibits de novo ceramide synthesis via DES1 protein. ( A) Western blot of liver tissue for DES1 (upper panel) and GAPDH (lower panel) used as a loading control, in LDLR−/− mice. ( B) Quantification of data shown in (A) (control $$n = 10$$, HFD $$n = 14$$, FEN-HFD $$n = 14$$). Equal numbers of representative samples from all treatment groups were run on multiple gels/blots to accommodate all samples. Image analysis and quantification with normalisation to loading control protein was performed within the same membrane and then data combined for graphical representation. Data are all compliant with digital image and integrity policies. ( C) Hepatic mRNA expression of ceramide synthesis genes ($$n = 8$$ per group). Quantification of ceramide (D), dihydroceramide (E) species in liver. ( F) Total ceramide, dihydroceramide and ratio of total dihydroceramide: total ceramide in hepatic tissues (left, middle and right panels respectively). Data are represented as mean + S.E.M. and analysed by one-way ANOVA (D–F) followed by Bonferroni multiple comparison t-tests where *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001 and ****p ≤ 0.0001 (control compared to HFD) or #p ≤ 0.05 and ##p ≤ 0.01, ###p ≤ 0.001 and ####p ≤ 0.001 (HFD compared to FEN-HFD).
HFD increased several acyl ceramide species e.g., C18:0, C18:1 and C20:0 but total ceramide levels were not increased with HFD ± FEN compared to control LDLR−/−. FEN treatment specifically decreased C26:0 ceramide but not any other species (Fig. 4D). However, FEN treatment increased all species of dihydroceramides measured from C16:0 to C26:1 (Fig. 4E) and total dihydroceramide levels by 4.7 to 8.9-fold compared to HFD and control mice respectively (Fig. 4F). Similar results were obtained in male and female ApoE−/− mice (Supplemental Fig. 3).
HFD also elevated levels of the ER stress protein GRP78/BIP (Supplemental Fig. 4) and FEN almost completely inhibited this increase. eIF2α phosphorylation and CHOP protein expression trended to be altered similarly, but HFD ± FEN did not affect levels of autophagy proteins beclin-1 and p38 (Supplemental Fig. 4). Taken together, these results suggest that FEN mediated inhibition of DES1 protein and thus inhibition of excess ceramide biosynthesis and lipotoxicity may be part of the mechanism of preventing insulin resistance and NAFLD/NASH.
## Fenretinide worsens hypertriglyceridemia and accelerates atherogenesis in LDLR−/− mice
The liver packages triglycerides into lipoproteins together with cholesterol and apolipoproteins which are then transported in the circulation. Thus, next we investigated this system to determine the effects of HFD ± FEN on development of dyslipidemia and atherosclerosis. HFD caused a major increase in circulating triglycerides and total cholesterol compared to control LDLR−/− mice (Fig. 5A,B). Surprisingly, FEN did not prevent the increase in serum cholesterol and caused a further increase in serum triglyceride when compared to HFD mice. HFD elevated circulating apolipoprotein B (ApoB) 48 levels, but FEN-HFD resulted in increased ApoB100 protein in both, liver and serum (Fig. 5C–E) suggestive of unique effects respectively on further increasing very low-density lipoprotein (VLDL) and/or LDL levels in LDLR−/− mice. Figure 5Fenretinide increases circulating triglyceride and apolipoprotein B 100 levels. Serum triglyceride (A) and total cholesterol (B) in LDLR−/− mice ($$n = 8$$ per group). ( C) Western blot of ApoB 48 and ApoB 100 in serum (upper panel) and hepatic tissues (middle panel). Quantification of hepatic (D) and serum levels (E) shown in (C). Hepatic levels were normalised to vinculin. Data are represented as mean + S.E.M. ($$n = 4$$–5 per group) and analysed by one-way ANOVA followed by Bonferroni multiple comparison t-tests where **p ≤ 0.01, ***p ≤ 0.001 and ****p ≤ 0.0001 (control compared to HFD) or #p ≤ 0.05 and ####p ≤ 0.0001 (HFD compared to FEN-HFD). Representative samples from all treatment groups were run on one gel/blot. Image analysis and quantification with normalisation to loading control protein was performed within the same membrane for graphical representation. Data are all compliant with digital image and integrity policies.
Since elevated circulating triglyceride, ApoB-containing lipoproteins and the ratio of ApoB100 to ApoB48 are major risk factors for the development of CVD36, we next investigated the effect of HFD ± FEN on atherosclerotic plaque formation. HFD resulted in atherosclerotic plaque formation in the aortic root, in the aortic arch and the descending aorta in LDLR−/− mice (Fig. 6A–D). FEN-treated mice had a similar level of plaque formation compared to HFD mice in the aortic root and in the aortic arch (Fig. 6A–C), but considerably more atherosclerotic plaque throughout the descending aorta (Fig. 6B–D). To determine if this was the case in another commonly used model for atherosclerosis, we examined plaque formation in ApoE−/− mice and found it to be accelerated in this background too. FEN-HFD resulted in significantly greater plaque accumulation in the descending aorta of female mice (Supplemental Fig. 5).Figure 6Fenretinide accelerates atherogenesis in LDLR−/− mice. ( A) Plaque formation (upper panels, representative images) in heart aortic roots sections stained with Oil Red O and overall structure with H&E staining (lower panels). ( B) Plaque formation in the descending aorta prepared en face and stained with Sudan IV. ( C,D) Quantification ($$n = 5$$) of plaque area shown in (A) and (B) respectively. ( E) Hepatic expression of genes encoding neutral sphingomyelinases, Smpd1 and Smpd3 and WAT expression of Smpd3 ($$n = 8$$ per group). ( F) Hepatic *Smpd3* gene expression following acute retinoic acid (RA) or FEN intraperitoneally injection in C57BL/6 mice (on chow diet, ad lib fed, 50 mg/kg for 2 or 6 hours, $$n = 8$$ per group). Data are represented as mean + S.E.M. and analysed by One-way ANOVA followed by Bonferroni multiple comparison t-tests where **p ≤ 0.01, ***p ≤ 0.001 and ****p ≤ 0.0001 (compared to control) or ###p ≤ 0.001 and ####p ≤ 0.0001 (HFD compared to FEN-HFD).
In addition to the role of excess de novo ceramide synthesis in the pathogenesis of metabolic diseases, ceramide generation via sphingomyelin hydrolysis has also been linked to atherosclerosis (see Choi et al. for recent review10). FEN treatment in LDLR−/− mice lead to a striking four-fold increase in hepatic Smpd3 expression, the gene encoding neutral sphingomyelinase-2, recently shown to contribute to the development of atherosclerosis in ApoE−/− mice. ( Fig. 6E). Similar results were obtained in our ApoE−/− mice (Supplemental Fig. 5). Smpd3 expression was not altered in white adipose tissue (Fig. 6E). We tested whether Smpd3 could be induced by FEN directly via RAR-signalling, to understand the mechanism behind this alteration. Acute RA injection led to a potent increase in Smpd3 expression at 6 h but not earlier at 2 h in the livers of lean C57/BL6 mice. FEN treatment also led to an increase in Smpd3 expression at 6 h, but the effect was not as striking as with RA treatment (Fig. 6F).
Next, we examined whether an increase in hepatic Smpd3 expression could result in an increase in circulating ceramides and thereby contribute to the increased development of atherosclerosis with FEN treatment in LDLR−/− mice. FEN increased total serum ceramide levels 1.6-fold more than in HFD mice (Fig. 7A). We determined an increase in a number of ceramide species with FA acyl groups 18:0 to 26:0 (Fig. 7B–E). FEN also increased total serum dihydroceramide levels eight-fold higher than in HFD mice with increases in every species measured (Fig. 7A,F–J).Figure 7Fenretinide increases serum ceramide and dihydroceramide levels in LDLR-/- mice. Quantification of ceramide (A,B–E) and dihydroceramide (A,F–J) species in serum. Data are represented as mean + S.E.M. and analysed by one-way ANOVA followed by Bonferroni multiple comparison t-tests where *p ≤ 0.05, **p ≤ 0.01 and ***p ≤ 0.001 (control compared to HFD) or #p ≤ 0.05, ##p ≤ 0.01 and ###p ≤ 0.001 (HFD compared to FEN-HFD).
Thus, overall, these data suggest that FEN treatment was beneficial in the treatment of pathologies associated with an obesogenic diet and excess fat gain, thereby attenuating the development of insulin resistance, lipotoxicity and NAFLD/NASH, but at the detriment of the cardiovascular system, at least in genetic mouse models lacking LDLR−/− or ApoE−/−. Mechanistically, FEN treatment results in retinoic acid signalling mediated induction of sphingomyelinase gene Smpd3 and an increase in circulating ceramides and thereby may contribute to the increased development of atherosclerosis in LDLR−/− mice (as illustrated in Supplemental Fig. 7).
## Discussion
The multiple overlapping secondary pathologies in response to diet-induced obesity such as the development of metabolic syndrome, type 2 diabetes, NAFLD, atherosclerosis and CVD have interconnected underlying molecular drivers1,2,5,37. There is increasing evidence that dysregulation of lipid metabolism e.g., excess ceramide biosynthesis that can lead to lipotoxicity and insulin resistance, may be a tractable target for novel pharmacological interventions that can treat these co-morbidities. Currently there is no treatment for NAFLD or NASH and therapies are urgently needed. In this study we have demonstrated that inhibition of adiposity and inhibition of DES1, the final step in ceramide biosynthesis, by FEN can prevent hepatic triglyceride accumulation and steatosis in the LDLR−/− mouse model of diet-induced NAFLD and atherosclerosis. This comes at the expense of the cardiovascular system however, as FEN treatment results in augmented atherogenesis accompanied by alterations in hepatic and serum ApoB 100 species of lipoproteins and induction of Smpd3 (as illustrated in Supplemental Fig. 7).
It has previously been demonstrated that genetic deletion of enzymes involved in ceramide production and pharmacological treatment with myriocin, a natural fungal metabolite that inhibits serine palmitoyltransferase can attenuate obesity and improve glucose homeostasis11,38. Many of these studies have also shown an associated attenuation of hepatic steatosis13,15. We have previously demonstrated that FEN can partially prevent diet-induced obesity and associated metabolic pathologies, including a partial attenuation of fatty liver15,16. *In* genetically obese (leptin-signalling deficient) mice, FEN markedly improved glucose homeostasis without a decrease in body weight or liver triglyceride17. Thus, it is even more striking that FEN can prevent hepatic triglyceride accumulation and steatosis in LDLR−/− mice, a more robust model of dyslipidemia and NAFLD compared to obesity-prone C57BL/6 mice26. FEN did not improve glucose homeostasis in LDLR−/− mice, despite inhibition of ceramide biosynthesis and improvement in insulin action. Thus together, these findings suggest overlapping pathologies of metabolic syndrome can be dissociated depending on the animal model investigated and this can affect the beneficial effects observed with therapeutic treatments such as FEN.
*Hepatic* gene expression alterations that are linked with excess hepatic lipid accumulation revealed a unique pattern of beneficial improvement with FEN treatment including downregulation of Tm6sf2 and Hsd17b13 but not Mogat or Vldr. The biological roles of Tm6sf2 and Hsd17b13 appear to involve apolipoprotein secretion and lipid droplet homeostasis, respectively, but these are not clear and require further investigation to understand their role in normal lipid metabolism and dyslipidemia3. FEN treatment also decreased genes involved in deposition of excess extracellular matrix such as Col1a1 and Timp2, but increased Mmp9 expression34. Mmp9 has been reported to be a directly retinoid-responsive gene and confirmed by us to be directly regulated by RA and FEN treatment by RNA-seq and qPCR methods (not shown, manuscript in preparation). This induction of Mmp9 may directly contribute to a retinoid-specific effect to increase degradation of extracellular matrix and thus prevent the progression of hepatic fibrosis34.
FEN markedly increased aortic atherosclerotic plaque formation in LDLR−/− mice, despite all the beneficial effects of FEN and the demonstration that myriocin-mediated inhibition of sphingolipid biosynthesis decreased atherosclerosis in ApoE−/− mice39. In those studies, myriocin also decreased plasma triglycerides in hyperlipidemic ApoE−/− mice, whereas here FEN elevated circulating triglyceride, ApoB-containing lipoproteins and the ratio of ApoB100 to ApoB48. Since these are major risk factors for the development of atherosclerosis, this may contribute to the mechanism of increased aortic plaque formation. Some retinoic acid derivatives (that act primarily via retinoic X receptor, RXRs), have been used as an acne medication and in some patients results in hypertriglyceridemia via regulation of lipoproteins and thus careful monitoring is required40–42. Here, FEN treatment increased levels of ApoB 100 in both, the liver and serum, but not hepatic gene expression of other apolipoprotein species and moreover FEN and RA act via RARs not RXRs. FEN may regulate apolipoprotein B secretion via down-regulation of TM6SF2, which was recently identified as a ChREBP target in mouse liver but there is no evidence that FEN or other synthetic RA derivatives can regulate ChREBP activity17 and manuscript under preparation).
Our data suggest the putative mechanism of increased atherosclerosis with FEN treatment is via the upregulation of the gene encoding type 2-neutral sphingomyelinase (nSMase 2, gene Smpd3), an alternative ceramide generation pathway (via sphingomyelin hydrolysis). Genetic deficiency of nSMase2 (in mutant Smpd3fro/fro mice) or pharmacological inhibition of nSMase2 activity significantly reduced the size of atherosclerotic lesions in ApoE−/− mice23,43. nSMase2 is a key enzyme of sphingolipid metabolism and Smpd3 expression is highest in the brain but also significant in the liver43. Interestingly, Smpd3 has been identified as a RA induced gene in MCF7 breast carcinoma cells and mouse embryonic stem cells treated for 12–24 h with retinoic acid and nSMase2 activity is regulated a number of factors eg. pro-inflammatory cytokines and phosphorylation21,22,44. More recently, Jiang et al.27 reported that upregulation of intestinal Smpd3 induces intestinal ceramide production and secretion to increase circulating ceramide levels resulting in accelerated atherosclerosis. The upregulation of intestinal Smpd3 was attributed to farnesoid X receptor (FXR), a ligand-activated nuclear receptor that regulates cholesterol and bile acid metabolism. Although there may be a complex inter-relationship between FXR- and RAR-signalling and bile acid and vitamin A homeostasis, there is no evidence of FEN directly regulating FXR-mediated transcription. Moreover, both studies reported no influence on serum cholesterol or triglyceride levels23,27,45.
Increased aortic atherosclerosis with FEN treatment was recapitulated in the ApoE−/− mice (Supplemental material) and most recently by Chiesa and co-workers28, but in contrast they reported a decrease in circulating triglyceride and lipoprotein levels. In humans, FEN treatment over two years also prevented an increase in circulating triglyceride levels (associated with an increase in HOMA) in normal weight women46. It is currently unclear whether FEN, either via retinoid signalling or inhibition of ceramide biosynthesis can cause hypertriglyceridemia. The effect of DES1 genetic knockout has not been studied in ApoE−/− or LDLR−/− and may help to clarify this. Chiesa and co-workers study of FEN in ApoE−/− mice also reported splenomegaly and haematological alterations28. In our study, although FEN treatment also resulted in splenomegaly in both male and female ApoE−/− mice, it did not in LDLR−/− mice (Supplemental Fig. 6). Thus, we do not attribute increased atherosclerotic lesions to haematological defects. Our potential mechanism of FEN treatment resulting in induction of Smpd3 and an increase in circulating ceramides and thereby increased atherosclerosis in LDLR−/− mice and ApoE−/− mice appears to be applicable to both sexes, at least in ApoE−/− mice23,27,28.
Although FEN-treated mice had more atherosclerotic plaque formation throughout the descending aorta in both male LDLR−/− mice and female ApoE−/− mice, we did not measure a difference in the aortic root with FEN treatment in our study. This contrasts with the results obtained in studies of others with FEN treatment or Smpd3/nSMase2 genetic or pharmacological interventions23,27,28. This may be because we obtained a single section of the aortic root at comparable anatomical positions from each mouse whereas others serially sectioned the aortic root (5 or 10 um intervals) from the appearance of the aortic valve to the ascending aorta (until the valve cusps are no longer visible). This may be considered a limitation of our study.
We have previously examined the effect of FEN-$10\%$ fat control diet in C57Bl/6 mice for 22 weeks and FEN-HFD fed for only 7 days in C57Bl/6 mice and have determined there were no major alterations in physiology (e.g. no change in body weight)16,17. The model of atherosclerosis in LDLR−/− mice requires the supplementation of a high-fat/high-cholesterol diet to induce atherogenesis and NAFLD. Therefore, we did not have a technical rationale for inclusion of the FEN-control diet in LDLR−/− mice experimental control group since we hypothesised a beneficial effect of FEN treatment. Since FEN and RA can acutely induce hepatic Smpd3 in normal C57Bl/6 mice it may be beneficial to understand if FEN can increase both circulating ceramides and atherosclerosis in LDLR−/− mice without a high-fat/high-cholesterol diet. This may be considered a limitation of our study.
In summary, the present study has demonstrated that FEN treatment can prevent hepatic triglyceride accumulation, steatosis and fibrosis in addition to prevention of obesity in LDLR−/− mice. Part of this favourable effect is via prevention of obesity and also inhibition of ceramide biosynthesis and improvement in insulin action. Despite these beneficial metabolic effects, a clear worsening of atherosclerosis was established in this atherosclerosis‐prone mouse model, which appears to involve an alternative ceramide generation pathway (via sphingomyelin hydrolysis) regulated by Smpd3/nSMase2. Since excess ceramide production causes lipotoxicity, metabolic dysregulation and atherogenesis, dual targeting of both DES1 and Smpd3/nSMase2 may be a novel strategy for treatment of metabolic syndrome and importantly deadly co-morbidities.
## Supplementary Information
Supplementary Information. The online version contains supplementary material available at 10.1038/s41598-023-30759-w.
## References
1. Yazıcı D, Sezer H, Engin AB, Engin A. **Insulin resistance, obesity and lipotoxicity**. *Obesity and Lipotoxicity* (2017) 277-304
2. Hotamisligil GS. **Inflammation, metaflammation and immunometabolic disorders**. *Nature* (2017) **542** 177-185. DOI: 10.1038/nature21363
3. Pafili K, Roden M. **Nonalcoholic fatty liver disease (NAFLD) from pathogenesis to treatment concepts in humans**. *Mol. Metab.* (2021) **50** 101122. DOI: 10.1016/j.molmet.2020.101122
4. Kawano Y, Cohen DE. **Mechanisms of hepatic triglyceride accumulation in non-alcoholic fatty liver disease**. *J. Gastroenterol.* (2013) **48** 434-441. DOI: 10.1007/s00535-013-0758-5
5. Bessone F, Razori MV, Roma MG. **Molecular pathways of nonalcoholic fatty liver disease development and progression**. *Cell Mol. Life Sci. CMLS* (2019) **76** 99-128. DOI: 10.1007/s00018-018-2947-0
6. Rodriguez-Cuenca S, Barbarroja N, Vidal-Puig A. **Dihydroceramide desaturase 1, the gatekeeper of ceramide induced lipotoxicity**. *Biochim. Biophys. Acta.* (2015) **1851** 40-50. DOI: 10.1016/j.bbalip.2014.09.021
7. Longato L, Tong M, Wands JR, de la Monte SM. **High fat diet induced hepatic steatosis and insulin resistance: Role of dysregulated ceramide metabolism**. *Hepatol. Res.* (2012) **42** 412-427. DOI: 10.1111/j.1872-034X.2011.00934.x
8. Brozinick JT, Hawkins E, Hoang BH, Kuo MS, Tan B, Kievit P. **Plasma sphingolipids are biomarkers of metabolic syndrome in non-human primates maintained on a Western-style diet**. *Int. J. Obes.* (2013) **37** 1064-70. DOI: 10.1038/ijo.2012.191
9. Holland WL, Brozinick JT, Wang LP, Hawkins ED, Sargent KM, Liu Y. **Inhibition of ceramide synthesis ameliorates glucocorticoid-, saturated-fat-, and obesity-induced insulin resistance**. *Cell Metab.* (2007) **5** 167-179. DOI: 10.1016/j.cmet.2007.01.002
10. Choi RH, Tatum SM, Symons JD, Summers SA, Holland WL. **Ceramides and other sphingolipids as drivers of cardiovascular disease**. *Nat. Rev. Cardiol.* (2021) **18** 701-711. DOI: 10.1038/s41569-021-00536-1
11. Summers SA. **Ceramides in insulin resistance and lipotoxicity**. *Prog. Lipid Res.* (2006) **45** 42-72. DOI: 10.1016/j.plipres.2005.11.002
12. Thrush AB, Brindley DN, Chabowski A, Heigenhauser GJ, Dyck DJ. **Skeletal muscle lipogenic protein expression is not different between lean and obese individuals: a potential factor in ceramide accumulation**. *J. Clin. Endocrinol. Metab.* (2009) **94** 5053-5061. DOI: 10.1210/jc.2008-2565
13. Koh I, Jun HS, Choi JS, Lim JH, Kim WH, Yoon JB. **Fenretinide ameliorates insulin resistance and fatty liver in obese mice**. *Biol. Pharm. Bull.* (2012) **35** 369-75. DOI: 10.1248/bpb.35.369
14. Mcilroy GD, Tammireddy SR, Maskrey BH, Grant L, Doherty MK, Watson DG. **Fenretinide mediated retinoic acid receptor signalling and inhibition of ceramide biosynthesis regulates adipogenesis, lipid accumulation, mitochondrial function and nutrient stress signalling in adipocytes and adipose tissue**. *Biochem. Pharmacol.* (2016) **100** 86-97. DOI: 10.1016/j.bcp.2015.11.017
15. Preitner F, Mody N, Graham TE, Peroni OD, Kahn BB. **Long-term Fenretinide treatment prevents high-fat diet-induced obesity, insulin resistance, and hepatic steatosis**. *Am. J. Physiol. Endocrinol. Metab.* (2009) **297** E1420-1429. DOI: 10.1152/ajpendo.00362.2009
16. Mcilroy GD, Delibegovic M, Owen C, Stoney PN, Shearer KD, McCaffery PJ. **Fenretinide treatment prevents diet-induced obesity in association with major alterations in retinoid homeostatic gene expression in adipose, liver, and hypothalamus**. *Diabetes* (2013) **62** 825-836. DOI: 10.2337/db12-0458
17. Morrice N, Mcilroy GD, Tammireddy SR, Reekie J, Shearer KD, Doherty MK. **Elevated Fibroblast growth factor 21 (FGF21) in obese, insulin resistant states is normalised by the synthetic retinoid Fenretinide in mice**. *Sci. Rep.* (2017) **7** 43782. DOI: 10.1038/srep43782
18. Shearer KD, Morrice N, Henderson C, Reekie J, Mcilroy GD, McCaffery PJ. **Fenretinide prevents obesity in aged female mice in association with increased retinoid and estrogen signaling**. *Obesity* (2015) **23** 1655-1662. DOI: 10.1002/oby.21164
19. Rahmaniyan M, Curley RW, Obeid LM, Hannun YA, Kraveka JM. **Identification of dihydroceramide desaturase as a direct in vitro target for fenretinide**. *J. Biol. Chem.* (2011) **286** 24754-24764. DOI: 10.1074/jbc.M111.250779
20. Bikman BT, Guan Y, Shui G, Siddique MM, Holland WL, Kim JY. **Fenretinide prevents lipid-induced insulin resistance by blocking ceramide biosynthesis**. *J. Biol. Chem.* (2012) **287** 17426-17437. DOI: 10.1074/jbc.M112.359950
21. Clarke CJ, Shamseddine AA, Jacob JJ, Khalife G, Burns TA, Hannun YA. **ATRA transcriptionally induces nSMase2 through CBP/p300-mediated histone acetylation**. *J. Lipid Res.* (2016) **57** 868-881. DOI: 10.1194/jlr.M067447
22. Moutier E, Ye T, Choukrallah MA, Urban S, Osz J, Chatagnon A. **Retinoic acid receptors recognize the mouse genome through binding elements with diverse spacing and topology**. *J. Biol. Chem.* (2012) **287** 26328-26341. DOI: 10.1074/jbc.M112.361790
23. Lallemand T, Rouahi M, Swiader A, Grazide MH, Geoffre N, Alayrac P. **nSMase2 (Type 2-Neutral Sphingomyelinase) Deficiency or Inhibition by GW4869 Reduces Inflammation and Atherosclerosis in Apoe-/- Mice**. *Arterioscler. Thromb. Vasc. Biol.* (2018) **38** 1479-1492. DOI: 10.1161/ATVBAHA.118.311208
24. Ishibashi S, Perrey S, Chen Z, Osuga JI, Shimada M, Ohashi K. **Role of the low density lipoprotein (LDL) receptor pathway in the metabolism of chylomicron remnants. A quantitative study in knockout mice lacking the LDL receptor, apolipoprotein E, or both**. *J. Biol. Chem.* (1996) **271** 22422-7. DOI: 10.1074/jbc.271.37.22422
25. Subramanian S, Goodspeed L, Wang S, Kim J, Zeng L, Ioannou GN. **Dietary cholesterol exacerbates hepatic steatosis and inflammation in obese LDL receptor-deficient mice**. *J. Lipid Res.* (2011) **52** 1626-1635. DOI: 10.1194/jlr.M016246
26. Bieghs V, Van Gorp PJ, Wouters K, Hendrikx T, Gijbels MJ, van Bilsen M. **LDL receptor knock-out mice are a physiological model particularly vulnerable to study the onset of inflammation in non-alcoholic fatty liver disease**. *PLoS ONE* (2012) **7** e30668. DOI: 10.1371/journal.pone.0030668
27. Wu Q, Sun L, Hu X, Wang X, Xu F, Chen B. **Suppressing the intestinal farnesoid X receptor/sphingomyelin phosphodiesterase 3 axis decreases atherosclerosis**. *J. Clin. Invest.* (2021) **131** 142865. DOI: 10.1172/JCI142865
28. Busnelli M, Manzini S, Bonacina F, Soldati S, Barbieri SS, Amadio P. **Fenretinide treatment accelerates atherosclerosis development in apoE-deficient mice in spite of beneficial metabolic effects**. *Br. J. Pharmacol.* (2020) **177** 328-345. DOI: 10.1111/bph.14869
29. Oppi S, Lüscher TF, Stein S. **Mouse models for atherosclerosis research—which is my line?**. *Front. Cardiovasc. Med.* (2019). DOI: 10.3389/fcvm.2019.00046
30. Folch J, Lees M, Sloane Stanley GH. **A simple method for the isolation and purification of total lipides from animal tissues**. *J. Biol. Chem.* (1957) **226** 497-509. DOI: 10.1016/S0021-9258(18)64849-5
31. Chella Krishnan K, Kurt Z, Barrere-Cain R, Sabir S, Das A, Floyd R. **Integration of multi-omics data from mouse diversity panel highlights mitochondrial dysfunction in non-alcoholic fatty liver disease**. *Cell Syst.* (2018) **6** 103-115.e7. DOI: 10.1016/j.cels.2017.12.006
32. Fabregat I, Moreno-Càceres J, Sánchez A, Dooley S, Dewidar B, Giannelli G. **TGF-β signalling and liver disease**. *FEBS J.* (2016) **283** 2219-32. DOI: 10.1111/febs.13665
33. Dewidar B, Meyer C, Dooley S, Meindl-Beinker AN. **TGF-β in hepatic stellate cell activation and liver fibrogenesis-updated 2019**. *Cells* (2019) **8** E1419. DOI: 10.3390/cells8111419
34. Zaragozá R, Gimeno A, Miralles VJ, García-Trevijano ER, Carmena R, García C. **Retinoids induce MMP-9 expression through RARalpha during mammary gland remodeling**. *Am. J. Physiol. Endocrinol. Metab.* (2007) **292** E1140-1148. DOI: 10.1152/ajpendo.00463.2006
35. Turpin SM, Nicholls HT, Willmes DM, Mourier A, Brodesser S, Wunderlich CM. **Obesity-induced CerS6-dependent C16:0 ceramide production promotes weight gain and glucose intolerance**. *Cell Metab.* (2014) **20** 678-686. DOI: 10.1016/j.cmet.2014.08.002
36. Powell-Braxton L, Véniant M, Latvala RD, Hirano KI, Won WB, Ross J. **A mouse model of human familial hypercholesterolemia: markedly elevated low density lipoprotein cholesterol levels and severe atherosclerosis on a low-fat chow diet**. *Nat. Med.* (1998) **4** 934-938. DOI: 10.1038/nm0898-934
37. Ortega FB, Lavie CJ, Blair SN. **Obesity and cardiovascular disease**. *Circ. Res.* (2016) **118** 1752-1770. DOI: 10.1161/CIRCRESAHA.115.306883
38. Siddique MM, Li Y, Wang L, Ching J, Mal M, Ilkayeva O. **Ablation of dihydroceramide desaturase 1, a therapeutic target for the treatment of metabolic diseases, simultaneously stimulates anabolic and catabolic signaling**. *Mol. Cell Biol.* (2013) **33** 2353-2369. DOI: 10.1128/MCB.00226-13
39. Park TS, Panek RL, Mueller SB, Hanselman JC, Rosebury WS, Robertson AW. **Inhibition of sphingomyelin synthesis reduces atherogenesis in apolipoprotein E-knockout mice**. *Circulation* (2004) **110** 3465-3471. DOI: 10.1161/01.CIR.0000148370.60535.22
40. Bershad S, Rubinstein A, Paterniti JR, Le NA, Poliak SC, Heller B. **Changes in plasma lipids and lipoproteins during isotretinoin therapy for acne**. *N. Engl. J. Med.* (1985) **313** 981-985. DOI: 10.1056/NEJM198510173131604
41. Vu-Dac N, Gervois P, Torra IP, Fruchart JC, Kosykh V, Kooistra T. **Retinoids increase human apo C-III expression at the transcriptional level via the retinoid X receptor. Contribution to the hypertriglyceridemic action of retinoids**. *J. Clin. Invest.* (1998) **102** 625-32. DOI: 10.1172/JCI1581
42. Cisneros FJ, Gough BJ, Patton RE, Ferguson SA. **Serum levels of albumin, triglycerides, total protein and glucose in rats are altered after oral treatment with low doses of 13-cis-retinoic acid or all-trans-retinoic acid**. *J. Appl. Toxicol.* (2005) **25** 470-478. DOI: 10.1002/jat.1082
43. Stoffel W, Jenke B, Blöck B, Zumbansen M, Koebke J. **Neutral sphingomyelinase 2 (smpd3) in the control of postnatal growth and development**. *Proc. Natl. Acad. Sci. USA* (2005) **102** 4554-4559. DOI: 10.1073/pnas.0406380102
44. Filosto S, Ashfaq M, Chung S, Fry W, Goldkorn T. **Neutral sphingomyelinase 2 activity and protein stability are modulated by phosphorylation of five conserved serines**. *J. Biol. Chem.* (2012) **287** 514-522. DOI: 10.1074/jbc.M111.315481
45. Saeed A, Hoekstra M, Hoeke MO, Heegsma J, Faber KN. **The interrelationship between bile acid and vitamin A homeostasis**. *Biochim. Biophys. Acta. BBA.* (2017) **1862** 496-512. DOI: 10.1016/j.bbalip.2017.01.007
46. Johansson H, Gandini S, Guerrieri-Gonzaga A, Iodice S, Ruscica M, Bonanni B. **Effect of fenretinide and low-dose tamoxifen on insulin sensitivity in premenopausal women at high risk for breast cancer**. *Cancer Res.* (2008) **68** 9512-9518. DOI: 10.1158/0008-5472.CAN-08-0553
|
---
title: 'Association of myosteatosis with treatment response and survival in patients
with hepatocellular carcinoma undergoing chemoembolization: a retrospective cohort
study'
authors:
- Kittipitch Bannangkoon
- Keerati Hongsakul
- Teeravut Tubtawee
- Natee Ina
- Ply Chichareon
journal: Scientific Reports
year: 2023
pmcid: PMC9998862
doi: 10.1038/s41598-023-31184-9
license: CC BY 4.0
---
# Association of myosteatosis with treatment response and survival in patients with hepatocellular carcinoma undergoing chemoembolization: a retrospective cohort study
## Abstract
Patients with hepatocellular carcinoma (HCC) have poor prognosis and have frequent treatment-related toxicities resulting in cancer-associated cachexia. This study aimed to determine the association of myosteatosis and sarcopenia on mortality in patients with HCC treated with transarterial chemoembolization (TACE). Six hundred and eleven patients diagnosed with HCC and underwent TACE at a tertiary care center between 2008 and 2019 were included. Body composition was assessed using axial CT slices at level L3 to calculate the skeletal muscle density for myosteatosis and skeletal muscle index for sarcopenia. The primary outcome was overall survival while the secondary outcome was TACE response. Patients with myosteatosis had a poorer TACE response than patients without myosteatosis ($56.12\%$ vs. $68.72\%$, adjusted odds ratio [OR] 0.49, $95\%$ confidence interval [CI] 0.34–0.72). The rate of TACE response in patients with sarcopenia was not different from those without sarcopenia ($60.91\%$ vs. $65.22\%$, adjusted OR 0.79, $95\%$ CI 0.55–1.13). Patients with myosteatosis had shorter overall survival than without myosteatosis (15.9 vs. 27.1 months, $P \leq 0.001$). In the multivariable Cox regression analysis, patients with myosteatosis or sarcopenia had higher risk of all-cause mortality than their counterparts (adjusted hazard ratio [HR] for myosteatosis versus no myosteatosis 1.66, $95\%$ CI 1.37–2.01, adjusted HR for sarcopenia versus no sarcopenia 1.26, $95\%$ CI 1.04–1.52). Patients with both myosteatosis and sarcopenia had the highest 7 year mortality rate at $94.45\%$, while patients with neither condition had the lowest mortality rate at $83.31\%$. The presence of myosteatosis was significantly associated with poor TACE response and reduced survival. Identifying patients with myosteatosis prior to TACE could allow for early interventions to preserve muscle quality and might improve prognosis in HCC patients.
## Introduction
Hepatocellular carcinoma (HCC) is an aggressive type of malignancy that is the third leading cause of cancer-related mortality worldwide1. Transarterial chemoembolization (TACE) displays a good response rate and clinical benefit for patients with inoperable HCC2,3. In some patients, TACE enables the conversion of unresectable and locally advanced liver cancer to operable cancer leading to an increase in survival outcome4. While TACE is a commonly used treatment for HCC, many patients do not respond well to it and have poor survival outcomes5. This is important for patients who receive chemoembolization because, although chemotherapy can give survival benefits to patients, it also causes liver toxicity and can lead to physical inactivity.
Skeletal muscle density (SMD) is known as a poor prognostic factor in patients with malignancies and is highly associated with tumor progression and mortality6–8. In principle, fat infiltration in muscle tissue appears as a lower density on CT scans compared to regular muscle tissue measured in Hounsfield units (HU)9. Myosteatosis (low SMD) indicates intramuscular fat deposition and low-grade skeletal muscle, which is correlated to poor muscle strength10.
Studies on body compositions in patients with HCC undergoing TACE are lacking and frequently focus on sarcopenia, not myosteatosis11–13. To the best of our knowledge, the association between myosteatosis and response to chemoembolization in HCC patients has not been well established. Since an abdominal CT scan is part of the routine pre-and postoperative evaluation for patients with liver cancer14, assessment of skeletal muscle could be employed in clinical practice. In this study, we aimed to determine the association of myosteatosis with TACE response and survival outcome in patients with HCC. The results can be used to provide an early screening modality in HCC patients to identify patients at risk of not responding well to TACE. Furthermore, early preventive strategies may potentially improve the outcome for patients with myosteatosis.
## Ethical approval
This study adhered to the standards of the Declaration of Helsinki and current ethical guidelines. Ethical approval was obtained by the institutional review board of the Faculty of Medicine, Prince of Songkla University and Songklanagarind Hospital (REC.65-317-7-1). The requirement for informed consent for this study was waived by the Institutional Review Board of the Faculty of Medicine, Prince of Songkla University and Songklanagarind Hospital as the study was a retrospective study.
## Patient population
Patients diagnosed with HCC and underwent TACE from January 2008 to December 2019 were included and analyzed. The inclusion criteria were as follows: [1] age greater than 18 years; [2] HCC diagnosis by imaging or histological findings according to the American Association for the Study of Liver Disease guidelines15; [3] initial treatment with conventional TACE; [4] HCC with Barcelona Clinic Liver Cancer (BCLC) stage A, B, or C (subsegmental or segmental portal vein tumor thrombosis); [5] available medical records; and [6] Child–Pugh class A or B. The exclusion criteria were as follows: [1] absence of imaging data; [2] inability to measure the skeletal muscle mass; [3] concomitant malignancies; and [4] history of HCC rupture.
## TACE protocol and treatment schedule
All patients with HCC were treated using conventional TACE by two experienced interventional radiologists who had at least 10 years of experience. We administered a mixture of iodized oil (range: 4–16 mL) and doxorubicin hydrochloride (range: 5–50 mg) or mitomycin-C (range: 10–20 mg) via the tumor-feeding hepatic arteries. We finished the procedure when the tumor feeding branch was completely embolized by gelatin sponge particles. The decision to repeat TACE session was made on demand at an interval of 6–12 weeks in patients with favorable liver function and performance status.
We evaluated baseline CT scans before TACE and 1-month post-TACE to evaluate TACE responses. The treatment response was assessed based on the imaging studies of the patients, which were either 4-phase contrast-enhanced CT scan or dynamic magnetic resonance imaging within 1 month after the initial TACE. The modified Response Evaluation Criteria in Solid Tumors (mRECIST) was used to assess radiological changes of HCC after treatment16. The criteria have four categories; complete response (CR); partial response (PR); stable disease (SD); and progressive disease (PD). Complete or partial response in the imaging study at 1-month post-TACE was classified as TACE response whereas stable or progressive disease was defined as no response. Assessment of tumor response was reviewed independently by two radiologists with expertise in liver imaging to minimize variability. In cases of disagreement, the final decision was obtained by consensus.
## Measurement and definition of body composition
CT scans within 1 month prior to TACE or in the first post-TACE were selected to measure body composition. Pre-TACE scans were preferentially chosen. When these were unavailable, the earliest post-TACE scans were used in the study. The CT images at the level of the third lumbar vertebra (L3) were carefully chosen and archived as Digital Imaging and Communications in Medicine (DICOM) data. All DICOM data calculated body composition using in-house software developed by MATLAB (The MathWorks, Natick, MA, USA) and freeware Python 3.6.13 (Anaconda, Inc.), to generate the measurement model based on neural network architecture also known as UNet. The valid accuracy of the model was $99.17\%$ and validity of the intersect over union co-efficiency was $89.40\%$17.
The L3 skeletal muscle index (SMI) is used to identify sarcopenia and is calculated by dividing the cross-sectional area of the muscle by the square of the patient's height (cm2/m2). Sarcopenia was defined as SMI ≤ 36.2 cm2/m2 and ≤ 29.6 cm2/m2 for males and females, respectively11. The areas of the abdominal wall and back muscles were used to calculate the SMD based on the areas of the pixels with attenuation between − 29 and + 150 HU. Myosteatosis was defined as SMD ≤ 44.4 HU or ≤ 39.3 HU in males and females, respectively11. In addition, patients were classified into four groups according to their sarcopenia and myosteatosis status (Group A—neither sarcopenia nor myosteatosis, Group B—sarcopenia without myosteatosis, Group C—myosteatosis without sarcopenia, and Group D—sarcopenia with myosteatosis).
## Data collection
The following data were collected: demographic information (age, sex, body mass index; clinical history (hepatitis B or C virus carriers, alcohol consumption, diabetes, hypertension, cardiovascular disease, pulmonary disease, chronic kidney disease, Child–Pugh class, and BCLC staging); laboratory data (levels of aspartate transaminase, alanine transaminase, total bilirubin, albumin, platelet count, and serum alpha-fetoprotein); tumor factors (size and number of tumors); and imaging response within 1 month after the initial TACE. The up-to-seven criteria were calculated by the summation of the largest tumor diameter in cm and the number of tumors. Two strata of the scores included the summation of scores ≤ 7 or > 718. The clinical and laboratory information were collected from the electronic medical records.
## Statistical analysis
Data analyses were performed using R software, version 4.2.0 (R foundation, Vienna, Austria). Baseline characteristics are shown as mean ± SD for normally distributed continuous variables or median (interquartile range) for those with a skewed distribution. Discrete variables are shown as counts (percentages). Differences between groups were evaluated using t-test or continuous variables and Pearson’s chi-square test or Fisher’s exact test for categorical data.
The cumulative overall survival (OS) rate was estimated using the Kaplan–Meier method and significant differences in survival distributions were tested by the log-rank test. Vital status (death or alive) was obtained from the civil registry up to December 31, 2021. Survival time was defined as the interval between the first TACE session for HCC and death or December 31, 2021 if the patient was alive.
A scatter plot was used to show the correlation between SMD and SMI. The degree of correlation was quantified with the correlation coefficient (R2). The association between myosteatosis or sarcopenia and mortality was assessed in multivariable Cox regression models. Adjusted variables included age, chronic lung disease, and chronic kidney disease since these variables had prognostic impact on mortality and were associated with myosteatosis and sarcopenia19–23. Multivariable logistic regression analysis was used to determine the association between myosteatosis or sarcopenia and TACE response (response versus no response). The adjusted variables in the logistic regression model were the variables used in the adjusted Cox model. P values less than 0.05 were considered to be statistically significant.
## Patient and tumor characteristics
A total of 611 patients with HCC were included in this study (Fig. 1). The patients and tumor characteristics are shown in Table 1. The mean patient age was 61.4 ± 10.9 years and males were predominant in this cohort ($72.8\%$). Half of the patients ($50.0\%$) had normal weight. Hepatitis B virus infection was the most common cause of liver cirrhosis ($49.3\%$). Diabetes ($27.7\%$) and hypertension ($27.2\%$) were the two most common comorbidities. The BCLC stage distribution was as follows: stage A, 172 ($28.0\%$); stage B, 406 ($66.4\%$); and stage C, 33 ($5.4\%$).Figure 1Flow of study patients. Table 1Baseline characteristics of the study participants who underwent TACE.CharacteristicsValuePatients, N611Age, mean ± SD61.43 ± 10.90Gender male, N (%)445 (72.8)Body mass index (kg/m2) < 20.0 (underweight), N (%)106 (17.3) 20.0–24.9 (normal weight), N (%)305 (50.0) ≥ 25.0 (overweight/obese), N (%)200 (32.7)Etiology, N (%) HBV/HCV/HBV + HCV/Alcohol/others301 (49.3)/135 (22.1)/7 (1.1)/65 (10.6)/103 (16.9)Comorbidity, N (%) Diabetes169 (27.7) Hypertension166 (27.2) Cardiovascular disease39 (6.4) Pulmonary disease35 (5.7) Chronic kidney disease26 (4.3)Child–Pugh class, N (%) A/B465 (76.1)/146 (23.9)AST (IU/L), median (IQR)61.0 (42.0,88.0)ALT (IU/L), median (IQR)38.0 (25.0,61.0)Total bilirubin (mg/dL), median (IQR)0.81 (0.53,1.32)Albumin (g/dL), mean ± SD3.53 ± 0.54Platelet count (× 103/mm3), median (IQR)116 [74,189]BCLC-staging, N (%) A/B/C172 (28.2)/406 (66.4)/33 (5.4)Alpha-fetoprotein (ng/mL), N (%) < 200, ≥ 200418 (68.4)/193 (31.6)Tumor size (cm), N (%) ≤ 3, 3–5, > 5171 (28.0)/173 (28.3)/267 (43.7)Number of tumors, N (%) 1, 2–3, > 3257 (42.0)/188 (30.8)/166 (27.2)TACE sessions, median (IQR)2 (1–4)SMI (cm2/m2), median (IQR) Male39.9 (34.8,44.6) Female30.5 (27.0,34.3)SMD (HU), median (IQR) Male46.0 (41.9,50.2) Female39.7 (35.0,43.3)SD standard deviation, IQR interquartile range, HBV hepatitis B virus, HCV hepatitis C virus, AST aspartate transaminase, ALT alanine transaminase, BCLC Barcelona Clinic Liver Cancer, TACE transarterial chemoembolization, SMI skeletal muscle index, SMD skeletal muscle density.
## Correlation between SMI and SMD
The prevalences of low SMI and low SMD were 197 ($32.2\%$) and 237 ($38.8\%$) patients. The correlation between SMI and SMD is shown in the scatter plot (Fig. 2). A positive correlation between SMI and SMD was demonstrated (R2 = 0.238, $P \leq 0.001$).Figure 2Correlation between skeletal muscle index (SMI) and skeletal muscle density (SMD).
## Associations between sarcopenia and myosteatosis and TACE response and complications
Treatment response and complications after TACE were evaluated according to sarcopenia and myosteatosis status (Table 2). Among the 611 patients, 390 responded well to TACE while 221 had a poor response. Patients with myosteatosis had a lower rate of TACE response than patients without myosteatosis ($56.1\%$ vs. $68.7\%$, adjusted odds ratio [OR] 0.49, $95\%$ confidence interval [CI] 0.34–0.72). The rate of TACE response in the patients with sarcopenia was not different from those without sarcopenia ($60.9\%$ vs. $65.2\%$, adjusted OR 0.79, $95\%$ CI 0.55–1.13). Postembolization syndrome and liver decompensation were numerically, although not statistically, higher in the myosteatosis group compared to patients without myosteatosis ($21.5\%$ vs. $19.3\%$, $$P \leq 0.564$$ and $5.5\%$ vs. $4.0\%$, $$P \leq 0.515$$). Patients in group B had a similar chance for TACE response to those in group A (adjusted OR 0.73, $95\%$ CI 0.43–1.24). Compared with group A, the chance for TACE response was significantly lower in group C (adjusted OR 0.42, $95\%$ CI 0.27–0.67) and group D (adjusted OR 0.50, $95\%$ CI 0.31–0.81).Table 2TACE response and complication according to myosteatosis or sarcopenia. ParameterNo sarcopenia ($$n = 414$$)Sarcopenia ($$n = 197$$)P valueNo myosteatosis ($$n = 374$$)Myosteatosis ($$n = 237$$)P valueTACE response120 (60.9)270 (65.2)0.345133 (56.1)257 (68.7)0.002Postembolization syndrome76 (18.4)47 (23.9)0.14072 (19.3)51 (21.5)0.564Liver decompensation17 (4.1)11 (5.6)0.54215 (4.0)13 (5.5)0.515TACE transarterial chemoembolization.
## Association between sarcopenia or myosteatosis and survival
The median OS time for the cohort was 22.1 months ($95\%$ CI 18.7–25.1 months). The 1-, 3-, and 5-year OS rates were $67.8\%$, $32.5\%$, and $19.0\%$, respectively. Patients with myosteatosis also had a lower OS time than patients without myosteatosis (15.9 vs. 27.1 months, $P \leq 0.001$). Patients with sarcopenia had a lower OS time than patients without sarcopenia (16.6 vs. 23.8 months, $$P \leq 0.011$$).
The overall all-cause mortality rate at seven years following TACE was $87.13\%$. Patients with myosteatosis had a higher rate of mortality than patients without myosteatosis ($93.18\%$ vs. $83.58\%$, $P \leq 0.0001$) (Fig. 3). The rates of 7-year all-cause mortality were $90.2\%$ and $85.59\%$ in patients with and without sarcopenia, respectively ($$P \leq 0.011$$). In the multivariable Cox regression analysis, patients with myosteatosis or sarcopenia had a higher risk of all-cause mortality than their counterparts (adjusted hazard ratio [HR] for myosteatosis versus no myosteatosis 1.66, $95\%$ CI 1.37–2.01, adjusted HR for sarcopenia versus no sarcopenia 1.26, $95\%$ CI 1.04–1.52).Figure 3All-cause mortality of patients (A) with or without myosteatosis and (B) with or without sarcopenia. * adjusted by age, chronic lung disease, and chronic kidney disease.
Patients in group D (myosteatosis and sarcopenia) had the highest mortality rate at seven years ($94.45\%$), whereas group A (neither myosteatosis nor sarcopenia) had the lowest mortality rate ($83.31\%$) (Fig. 4). The mortality rate was $84.82\%$ in group B (sarcopenia without myosteatosis) while it was $91.62\%$ in group C (myosteatosis without sarcopenia). The all-cause mortality rate was significantly different among the four groups ($P \leq 0.0001$). Compared with group A, the adjusted risk for all-cause mortality was significantly higher in group C (adjusted HR 1.70, $95\%$ CI 1.33–2.17) and group D (adjusted HR 1.75, $95\%$ CI 1.36–2.23). The risk of all-cause mortality in group B was similar to group A (adjusted HR 1.19, $95\%$ CI 0.90–1.57).Figure 4All-cause mortality according to myosteatosis and sarcopenia status.
## Discussion
We evaluated the potential impact of myosteatosis on outcome in patients with HCC who received TACE as their initial treatment. Our cohort revealed that myosteatosis, as assessed by SMD, was associated with a poor response to TACE and lower survival rates in patients with HCC. The early identification of myosteatosis and the implementation of early preventive strategies may potentially improve the prognosis of patients with HCC.
Several previous studies have demonstrated that sarcopenia and myosteatosis were negative prognostic factors for oncological patients6,24–26. HCC is normally concomitant with chronic liver disease and cirrhosis, which have a robust relationship with alterations of body composition27. Moreover, most patients with HCC are frequently unfit for curative surgical treatment due to underlying cirrhosis and comorbidities; therefore, TACE remains the treatment of choice in the management of these patients28. Although previous cohorts demonstrated that frailty, malnutrition, and loss of muscle mass and function were related to poor results in HCC patients, these patients were usually highly heterogeneous with a variety of available treatments that primarily focused on sarcopenia11–13.
In our cohort, we demonstrated that myosteatosis is superior to sarcopenia in being associated with survival outcomes in patients with HCC who underwent TACE. Myosteatosis, rather than sarcopenia, was also recognized as an independent factor associated with mortality after adjusting for clinically significant covariates. The results of our study were similar to previous studies that reported that SMD was a better prognostic factor than SMI in terms of statistical significance in renal, pancreatic, gastric, and breast cancers6,8,29,30, which indicated that myosteatosis is a more reliable indicator of association with survival outcomes compared to sarcopenia status. *In* general, CT-based results permit for early recognition of decreases in SMD even when the SMI has not changed29. Therefore, SMD deterioration is identified earlier than a change in SMI.
Sarcopenia is a progressive loss of skeletal muscle mass and strength. Myosteatosis is characterized by the fatty infiltration of muscle tissue that can be identified as low muscle density on CT images and is a contributing component to sarcopenia31. SMD and SMI were positively correlated in our cohort, which was consistent with the results of some previous studies8,32,33, but contrasted with other studies34,35. However, there were some differences in terms of the cut-off values, type of cancer, and stage of disease. There is no consensus on a cutoff value for body composition that is applicable in Asian countries. Thus, we used the cut-off values for SMD and SMI from a study in Japan by Fujiwara et al.11, not from a Western study24. Furthermore, our study focused particularly on HCC patients in the intermediate stage with a relatively poor prognosis. We found that $19.5\%$ of HCC patients who underwent TACE exhibited a concomitant presence of myosteatosis and sarcopenia with the worst overall survival compared with the other groups (14.8 vs. 23.6 months, $P \leq 0.001$).
In the present study, myosteatosis was significantly associated with TACE response ($$P \leq 0.002$$); however, sarcopenia failed to stratify our patients based on tumor responsiveness. It has been theorized that adipocytes (intermuscular fat) in myosteatosis release inflammatory adipokines that cause impaired nutritive blood flow to muscle, which worsens insulin diffusion capacity and contributes to insulin resistance36. As a result, myosteatosis may reduce body immunities, stimulate cancer growth, and affect unfavorable treatment outcomes37.
Myosteatosis was shown in this study to be associated with worse overall survival in patients with HCC who underwent chemoembolization. Therefore, it is important to identify patients with higher skeletal muscle fat accumulation before treatment in order to implement early preventive strategies that aim to maintain muscle quality and improve prognosis. Currently, there is no established treatment strategy specifically for myosteatosis in cancer patients; however, some studies have suggested that intensifying perioperative and postoperative exercise, including resistance training, may help maintain postoperative physical strength and lead to earlier resumption of daily activities38,39. Additionally, nutritional support that includes vitamin D, omega-3 fatty acids, and β-hydroxy β-methyl butyrate may also help improve muscle mass and quality in cancer patients40. Nevertheless, it is important to note that more research is needed to fully understand the relationship between myosteatosis and cancer survival, as well as to determine the optimal treatment strategies for patients with this condition.
This study has the following limitations. First, this study was a single-center retrospective design. While most of the patients in our study received a CT scan before surgery, only one-third of patients had their body composition evaluated one month after TACE, which possibly led to variations in the imaging data. Second, this study was conducted at a tertiary referral center for cancer patients in southern Thailand; therefore, survival rates may vary between specialized and non-specialized centers as well as between different countries and healthcare systems. Further studies should be conducted in other populations to validate the findings of the current study. Finally, the lack of consensus definitions for myosteatosis and sarcopenia is a significant limitation in the field of body composition research in cancer patients. Further research is necessary to establish the optimal threshold for these definitions. This study had a few strengths. Our dataset had a relatively large number of patients and a homogeneous type of treatment modality, which was conventional TACE. Moreover, we not only determined the association of body composition with survival outcome but also established response to TACE in patients with HCC as useful clinical information to aid treatment decisions. Moreover, appropriate perioperative management strategies, such as preoperative respiratory exercise, protein supplementation, and other techniques that enhance skeletal muscle health, can be implemented to lower mortality rates in patients with HCC.
## Conclusion
The presence of myosteatosis was significantly associated with poor TACE response and reduced survival. Identifying patients with myosteatosis and implementing early preventive strategies may preserve muscle quality and potentially improve the prognosis of patients with HCC.
## References
1. Sung H, Ferlay J, Siegel RL, Laversanne M, Soerjomataram I. **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. **EASL clinical practice guidelines: Management of hepatocellular carcinoma**. *J. Hepatol.* (2018) **69** 182-236. DOI: 10.1016/j.jhep.2018.03.019
3. Reig M, Forner A, Rimola J, Ferrer-Fàbrega J, Burrel M. **BCLC strategy for prognosis prediction and treatment recommendation: The 2022 update**. *J. Hepatol.* (2022) **76** 681-693. DOI: 10.1016/j.jhep.2021.11.018
4. Sun HC, Zhu XD. **Downstaging conversion therapy in patients with initially unresectable advanced hepatocellular carcinoma: An overview**. *Front. Oncol.* (2021) **11** 772195. DOI: 10.3389/fonc.2021.772195
5. Chen L, Yu CX, Zhong BY, Zhu HD, Jin ZC. **Development of TACE refractoriness scores in hepatocellular carcinoma**. *Front. Mol Biosci.* (2021) **8** 615133. DOI: 10.3389/fmolb.2021.615133
6. Antoun S, Lanoy E, Iacovelli R, Albiges-Sauvin L, Loriot Y. **Skeletal muscle density predicts prognosis in patients with metastatic renal cell carcinoma treated with targeted therapies**. *Cancer* (2013) **119** 3377-3384. DOI: 10.1002/cncr.28218
7. Shachar SS, Deal AM, Weinberg M, Nyrop KA, Williams GR. **Skeletal muscle measures as predictors of toxicity, hospitalization, and survival in patients with metastatic breast cancer receiving taxane-based chemotherapy**. *Clin. Cancer Res.* (2017) **23** 658-665. DOI: 10.1158/1078-0432.CCR-16-0940
8. Kim IH, Choi MH, Lee IS, Hong TH, Lee MA. **Clinical significance of skeletal muscle density and sarcopenia in patients with pancreatic cancer undergoing first-line chemotherapy: A retrospective observational study**. *BMC Cancer* (2021) **21** 77. DOI: 10.1186/s12885-020-07753-w
9. Aubrey J, Esfandiari N, Baracos VE, Buteau FA, Frenette J. **Measurement of skeletal muscle radiation attenuation and basis of its biological variation**. *Acta Physiol. (Oxf).* (2014) **210** 489-497. DOI: 10.1111/apha.12224
10. Miljkovic I, Zmuda JM. **Epidemiology of myosteatosis**. *Curr. Opin .Clin. Nutr. Metab. Care* (2010) **13** 260-264. DOI: 10.1097/MCO.0b013e328337d826
11. Fujiwara N, Nakagawa H, Kudo Y, Tateishi R, Taguri M. **Sarcopenia, intramuscular fat deposition, and visceral adiposity independently predict the outcomes of hepatocellular carcinoma**. *J Hepatol.* (2015) **63** 131-140. DOI: 10.1016/j.jhep.2015.02.031
12. Von Hessen L, Roumet M, Maurer MH, Lange N, Reeves H. **High subcutaneous adipose tissue density correlates negatively with survival in patients with hepatocellular carcinoma**. *Liver Int.* (2021) **41** 828-836. DOI: 10.1111/liv.14755
13. Li Q, Zhang L, Hou ZH, Zhao DX, Li JB. **High visceral adipose tissue density correlates with unfavorable outcomes in patients with intermediate-stage hepatocellular carcinoma undergoing transarterial chemoembolization**. *Front. Cell Dev. Biol.* (2021) **9** 710104. DOI: 10.3389/fcell.2021.710104
14. Llovet JM, Lencioni R. **mRECIST for HCC: Performance and novel refinements**. *J. Hepatol.* (2020) **72** 288-306. DOI: 10.1016/j.jhep.2019.09.026
15. Heimbach JK, Kulik LM, Finn RS, Sirlin CB, Abecassis MM. **AASLD guidelines for the treatment of hepatocellular carcinoma**. *Hepatology* (2018) **67** 358-380. DOI: 10.1002/hep.29086
16. Lencioni R, Llovet JM. **Modified RECIST (mRECIST) assessment for hepatocellular carcinoma**. *Semin. Liver Dis.* (2010) **30** 52-60. DOI: 10.1055/s-0030-1247132
17. Cheewakul J, Yanaihap N, Yue-roh T, Kepan N, Thiangsook W. **Automated segmentation of skeletal muscle at the third lumbar vertebra using deep learning in computed tomography images**. *J. Med. Imaging Radiat. Sci.* (2022) **53** S37. DOI: 10.1016/j.jmir.2022.10.122
18. Mazzaferro V, Llovet JM, Miceli R, Bhoori S, Schiavo M. **Predicting survival after liver transplantation in patients with hepatocellular carcinoma beyond the Milan criteria: A retrospective, exploratory analysis**. *Lancet Oncol.* (2009) **10** 35-43. DOI: 10.1016/S1470-2045(08)70284-5
19. Ligthart-Melis GC, Luiking YC, Kakourou A, Cederholm T, Maier AB. **Frailty, sarcopenia, and malnutrition frequently (co-)occur in hospitalized older adults: A systematic review and meta-analysis**. *J. Am. Med. Dir. Assoc.* (2020) **21** 1216-1228. DOI: 10.1016/j.jamda.2020.03.006
20. Sepúlveda-Loyola W, Osadnik C, Phu S, Morita AA, Duque G. **Diagnosis, prevalence, and clinical impact of sarcopenia in COPD: A systematic review and meta-analysis**. *J. Cachexia Sarcopenia Muscle.* (2020) **11** 1164-1176. DOI: 10.1002/jcsm.12600
21. Shu X, Lin T, Wang H, Zhao Y, Jiang T. **Diagnosis, prevalence, and mortality of sarcopenia in dialysis patients: A systematic review and meta-analysis**. *J. Cachexia Sarcopenia Muscle.* (2022) **13** 145-158. DOI: 10.1002/jcsm.12890
22. Lim J, Kim KW, Ko Y, Jang IY, Lee YS. **The role of muscle depletion and visceral adiposity in HCC patients aged 65 and over undergoing TACE**. *BMC Cancer* (2021) **21** 1164. DOI: 10.1186/s12885-021-08905-2
23. Westreich D, Greenland S. **The table 2 fallacy: Presenting and interpreting confounder and modifier coefficients**. *Am. J. Epidemiol.* (2013) **177** 292-298. DOI: 10.1093/aje/kws412
24. Martin L, Birdsell L, Macdonald N, Reiman T, Clandinin MT. **Cancer cachexia in the age of obesity: Skeletal muscle depletion is a powerful prognostic factor, independent of body mass index**. *J. Clin. Oncol.* (2013) **31** 1539-1547. DOI: 10.1200/JCO.2012.45.2722
25. Choi Y, Oh DY, Kim TY, Lee KH, Han SW. **Skeletal muscle depletion predicts the prognosis of patients with advanced pancreatic cancer undergoing palliative chemotherapy, independent of body mass index**. *PLoS ONE* (2015) **10** e0139749. DOI: 10.1371/journal.pone.0139749
26. Park I, Choi SJ, Kim YS, Ahn HK, Hong J. **Prognostic factors for risk stratification of patients with recurrent or metastatic pancreatic adenocarcinoma who were treated with gemcitabine-based chemotherapy**. *Cancer Res. Treat.* (2016) **48** 1264-1273. DOI: 10.4143/crt.2015.250
27. Nachit M, Leclercq IA. **Emerging awareness on the importance of skeletal muscle in liver diseases: Time to dig deeper into mechanisms!**. *Clin. Sci.* (2019) **133** 465-481. DOI: 10.1042/CS20180421
28. Lurje G, Bednarsch J, Czigany Z, Amygdalos I, Meister F. **Prognostic factors of disease-free and overall survival in patients with hepatocellular carcinoma undergoing partial hepatectomy in curative intent**. *Langenbecks Arch. Surg.* (2018) **403** 851-861. DOI: 10.1007/s00423-018-1715-9
29. Hayashi N, Ando Y, Gyawali B, Shimokata T, Maeda O. **Low skeletal muscle density is associated with poor survival in patients who receive chemotherapy for metastatic gastric cancer**. *Oncol. Rep.* (2016) **35** 1727-1731. DOI: 10.3892/or.2015.4475
30. Rier HN, Jager A, Sleijfer S, van Rosmalen J, Kock MCJM. **Low muscle attenuation is a prognostic factor for survival in metastatic breast cancer patients treated with first line palliative chemotherapy**. *Breast* (2017) **31** 9-15. DOI: 10.1016/j.breast.2016.10.014
31. Zamboni M, Gattazzo S, Rossi AP. **Myosteatosis: A relevant, yet poorly explored element of sarcopenia**. *Eur. Geriatr. Med.* (2019) **10** 5-6. DOI: 10.1007/s41999-018-0134-3
32. Chu MP, Lieffers J, Ghosh S, Belch A, Chua NS. **Skeletal muscle density is an independent predictor of diffuse large B-cell lymphoma outcomes treated with rituximab-based chemoimmunotherapy**. *J. Cachexia Sarcopenia Muscle.* (2017) **8** 298-304. DOI: 10.1002/jcsm.12161
33. Caan BJ, Cespedes Feliciano EM, Prado CM, Alexeeff S, Kroenke CH. **Association of muscle and adiposity measured by computed tomography with survival in patients with nonmetastatic breast cancer**. *JAMA Oncol.* (2018) **4** 798-804. DOI: 10.1001/jamaoncol.2018.0137
34. Miyata H, Sugimura K, Motoori M, Fujiwara Y, Omori T. **Clinical assessment of sarcopenia and changes in body composition during neoadjuvant chemotherapy for esophageal cancer**. *Anticancer Res.* (2017) **37** 3053-3059. PMID: 28551644
35. Weinberg MS, Shachar SS, Muss HB, Deal AM, Popuri K. **Beyond sarcopenia: Characterization and integration of skeletal muscle quantity and radiodensity in a curable breast cancer population**. *Breast J.* (2018) **24** 29
36. Miljkovic I, Kuipers AL, Cvejkus R, Bunker CH, Patrick AL. **Myosteatosis increases with aging and is associated with incident diabetes in African ancestry men**. *Obesity (Silver Spring)* (2016) **24** 476-482. DOI: 10.1002/oby.21328
37. Miljkovic I, Cauley JA, Wang PY, Holton KF, Lee CG. **Osteoporotic Fractures in Men (MrOS) Research Group. Abdominal myosteatosis is independently associated with hyperinsulinemia and insulin resistance among older men without diabetes**. *Obesity (Silver Spring)* (2013) **21** 2118-2125. DOI: 10.1002/oby.20346
38. Adams SC, Segal RJ, McKenzie DC, Vallerand JR, Morielli AR. **Impact of resistance and aerobic exercise on sarcopenia and dynapenia in breast cancer patients receiving adjuvant chemotherapy: A multicenter randomized controlled trial**. *Breast Cancer Res. Treat.* (2016) **158** 497-507. DOI: 10.1007/s10549-016-3900-2
39. Davis MP, Panikkar R. **Sarcopenia associated with chemotherapy and targeted agents for cancer therapy**. *Ann. Palliat. Med.* (2019) **8** 86-101. DOI: 10.21037/apm.2018.08.02
40. Chen LK, Woo J, Assantachai P, Auyeung TW, Chou MY. **Asian working group for sarcopenia: 2019 consensus update on sarcopenia diagnosis and treatment**. *J. Am. Med. Dir. Assoc.* (2020) **21** 300-307.e2. DOI: 10.1016/j.jamda.2019.12.012
|
---
title: The toxic natural product tutin causes epileptic seizures in mice by activating
calcineurin
authors:
- Qing-Tong Han
- Wan-Qi Yang
- Caixia Zang
- Linchao Zhou
- Chong-Jing Zhang
- Xiuqi Bao
- Jie Cai
- Fangfei Li
- Qinyan Shi
- Xiao-Liang Wang
- Jing Qu
- Dan Zhang
- Shi-Shan Yu
journal: Signal Transduction and Targeted Therapy
year: 2023
pmcid: PMC9998865
doi: 10.1038/s41392-023-01312-y
license: CC BY 4.0
---
# The toxic natural product tutin causes epileptic seizures in mice by activating calcineurin
## Abstract
Tutin, an established toxic natural product that causes epilepsy in rodents, is often used as a tool to develop animal model of acute epileptic seizures. However, the molecular target and toxic mechanism of tutin were unclear. In this study, for the first time, we conducted experiments to clarify the targets in tutin-induced epilepsy using thermal proteome profiling. Our studies showed that calcineurin (CN) was a target of tutin, and that tutin activated CN, leading to seizures. Binding site studies further established that tutin bound within the active site of CN catalytic subunit. CN inhibitor and calcineurin A (CNA) knockdown experiments in vivo proved that tutin induced epilepsy by activating CN, and produced obvious nerve damage. Together, these findings revealed that tutin caused epileptic seizures by activating CN. Moreover, further mechanism studies found that N-methyl-D-aspartate (NMDA) receptors, gamma-aminobutyric acid (GABA) receptors and voltage- and Ca2+- activated K+ (BK) channels might be involved in related signaling pathways. Our study fully explains the convulsive mechanism of tutin, which provides new ideas for epilepsy treatment and drug development.
## Introduction
Epilepsy, a common paroxysmal chronic nervous system disease, is accompanied by various features, such as cognitive dysfunction and neuronal necrosis.1,2 *It is* thought that epilepsy is induced by a dysfunctional imbalance between excitatory and inhibitory neurotransmitters. The etiology of epilepsy is complex, and its pathogenesis has not been completely clarified. Increasing evidence suggests that multiple interacting factors influence seizure intiation.3 Previous studies have shown that pyramidal neuronal loss and degeneration are common features in various regions of the hippocampus in animal models of epilepsy.4,5 Anti-epileptic drug development has been hindered since the etiology of epilepsy has not yet fully elucidated. Therefore, uncovering the pathogenesis and molecular potential targets of epilepsy is an urgent need with great significance for the treatment of epilepsy.
Natural products are crucial sources of the discovery of new drugs, and increasing numbers of studies have suggested that authenticating underlying targets of natural products is a vital approach to fully elucidate the mechanism of their actions. Recently, more and more target proteins of natural products have been screened using chemical proteomic approaches.6,7 For example, researchers have conducted in-depth research on the mechanism of metformin through chemical biology methods, and found that metformin plays its role by targeting adenosine monophosphate-activated protein kinase (AMPK) signaling pathway via presenilin enhancer 2 (PEN2).8 Molecular targets found in epilepsy caused by toxic natural products may also provide new ideas and strategies for the treatment of the disease. Currently, the natural products that can induce epilepsy include pilocarpine and Coriaria lactone (CL). CL, extracted from Coriariaceae, has been used to treat schizophrenia with seizures. Unsurprisingly, toxicity studies indicated that CL-treated animals exhibited symptoms of seizures, muscle spasticity and respiratory paralysis.9,10 Tutin, an established toxic compound, was first isolated and identified as a convulsive poison in the leaves and seeds of the New Zealand species of Coriaria.11,12 *Tutin is* the main epileptogenic component of CL, and perfusion of it induced or increased discharge from nucleus tractus solitarius neurons in vitro.13 Tutin overstimulates the nervous system, leading to hyperactivity and seizures.14 Previous studies on the tutin mechanism of action found that tutin-induced epilepsy may be related to inhibition of gamma-aminobutyric acid (GABA) receptor and glycine, but the specific mechanism has not been clarified.9,15,16 To discover the molecular targets of epilepsy and prevent the occurrence of the disease, the toxicity mechanism of tutin urgently needs to be elucidated, which will contribute to the development of antiepileptic drugs.
In this study, a series of experiments were designed to identify the mechanism of tutin action. It was found that tutin induced seizures through targeting the catalytic subunit A (CNA) of calcineurin (CN) via thermal proteome profiling-temperature range (TPP-TR) approach and hydrogen deuterium exchange mass spectrometry (HDX-MS) methods. These findings further validated that CN might be a possible potential target of tutin by employing a CN inhibitor (FK506) and CNA knockdown in vivo.
## Epilepsy induced by tutin is antagonized by diazepam and MK-801 in mice
Tutin is a well-known epileptogenic agent whose chemical structure is shown in Fig. 1a. We evaluated its half convulsive dose (CD50) and half lethal dose (LD50), and verified that tutin can cause seizures in mice (Supplementary Fig. S1a, b), Moreover, the electroencephalography (EEG) signal of tutin-treated mice was detected in the present study. Consistent with previous studies,13 tutin-treated mice exhibited typical epileptic EEG (Fig. 1b). Microdialysis with LC-MS/MS experiment was used to investigate changes of seizure-related neurotransmitters. This result indicated that tutin induced increased glutamate (Glu)/GABA ratio in the intercellular space (Supplementary Fig. S2). Meanwhile, Glu/GABA ratio was higher in epileptic mice, supporting that brain hyperexcitability is a vital feature of epilepsy induced by tutin. To explore the underlying mechanism that causes epilepsy, we selected antiepileptic drugs with different mechanisms, including diazepam, MK-801, retigabine and carbamazepine, and observed their therapeutic effects on epilepsy induced by tutin (2 mg/kg, intraperitoneally injected [i.p.]). It is known that Diazepam alleviates seizures by enhancing GABA-mediated inhibitory neurotransmission, and MK-801 antagonizes N-methyl-D-aspartate (NMDA) receptor.17,18 Retigabine exerts anti-epileptic effects through inhibition of neuronal excitation via voltage-gated KCNQ2-5 potassium channel activation,19 and carbamazepine is a voltage-dependent sodium channel blocker.20–22 The results showed that pretreatment with diazepam or MK-801 significantly decreased the maximal seizure score in 2 h (Fig. 1c, d), indicating that diazepam and MK-801 alleviated the epileptic behavior of mice induced by tutin. However, the effects of retigabine and carbamazepine on seizures were not obvious. In addition, martentoxin, a voltage- and Ca2+- activated K+ (BK) channel blocker,23,24 alleviated tutin-induced severe seizures in mice (Fig. 1e). Therefore, we speculated that the mechanisms of tutin-induced epilepsy in mice might be related to NMDA receptors, GABA receptors and BK channels. Fig. 1Epilepsy caused by tutin is antagonized by diazepam and MK-801 in mice. a Chemical structure of tutin. b EEG (1-min) tracings during the pre-injection baseline and tutin injection. c Effects of antiepileptic drugs on maximum Racine score within 2 h after tutin injection. Results were represented as mean ± SD with $$n = 12$$, **$P \leq 0.01$, ***$P \leq 0.001$ vs. model group. d Effects of antiepileptic drugs on the intensity of the seizure attack in mice ($$n = 12$$). * $P \leq 0.05$, **$P \leq 0.01$ vs. model group. e Effect of martentoxin on the intensity of the seizure attack in mice (*$P \leq 0.05$ vs. sham group, $$n = 12$$)
## Identification of tutin-targeting proteins in primary cultured hippocampal neurons of rats is conducted via TPP-TR approach
Activity-based protein profiling (ABPP) is regarded as an important approach to identify the potential targets of natural products.25–27 We firstly performed structural modification of tutin, such as hydroxyl modification into probe with terminal alkyne group (Supplementary Figs. S3-S5), ethylene oxide opening, and lactone ring opening.28 However, the ability of this modified tutin to induce epilepsy was decreased significantly, which makes this method unsuitable for the targets study of tutin targets (Supplementary Tab. S1). Hence, TPP as an alternative modification-free strategy, was used to study the target of tutin in primary hippocampal neurons.29,30 The primary hippocampal neurons were incubated in situ with 5 μM of tutin or PBS for 3 h, and then exposed to differnent temperatures for 3 min. The isolated proteins were labeled by tandem mass tag (TMT),31 which were exposed to UPLC fractionation and assayed by LC-MS/MS32 (Fig. 2a). Tm shifts were calculated on the basis of two replicates of tutin compared to PBS treatment and visualized after filtering (Fig. 2b). The minimum temperature shift of 1 °C was identified for 36 proteins in both the tutin- and PBS-treated replicates. Only diphosphomevalonate decarboxylase (Mvd) met additional significance thresholds (see the blue dot and melting curve in Supplementary Fig. S6). In addition to the significance thresholds, four additional proteins satisfied all other conditions (Supplementary Tab. S2): serine/threonine-protein phosphatase 2B catalytic subunit alpha isoform (ppp3ca), glutathione S-transferase LANCL1 (Lancl1), aldehyde dehydrogenase family 3 member B1 (Aldh3b1) and hydroxysteroid dehydrogenase-like protein 2 (Hsdl2). As determined with an extensive literature review, Mvd, Lancl1, Aldh3b1 and Hsdl2 have not been found to be closely related to epilepsy. CNA has been regarded as a potential molecule of epilepsy, and reported to participate jointly in epilepsy and regulate the activities of GABA and NMDA receptors.33,34 Moreover, we have found that GABA and NMDA receptors may be involved in epilepsy induced by tutin since antiepileptic drugs (diazepam and MK-801) showed significant antagonism effects (Fig. 1c, d). Collectively, all the data give us a hint that CNA may be a possible target of tutin. Fig. 2Target identification of tutin in primary hippocampal neurons is conducted using TPP-TR. a Schematic procedure of TPP-TR. b Scatterplot of Tm shifts of tutin vs. PBS treatment. The TPP R package was used for calculating the thermal response curve fitting and melting point
## Tutin binds to CN
CN is composed of catalytic subunits (atalytic domain, CNB binding domain, calmodulin-binding domain and autoinhibitory domain) and regulatory subunits.35,36 Mutants lacking the latter calmodulin-binding domain and autoinhibitory domain exhibit constitutively active phosphatase activity independent of Ca2+/calmodulin.37 *In this* study, different methods were applied to investigate whether tutin interacts with CN. Firstly, Western blot-based cellular thermal shift assay was applied to analyze the protein-tutin interactions in cells. Both temperature- and dose-dependent cellular thermal shift assay (CETSA) data revealed that tutin influcences the CNA thermal stability (Fig. 3a, b), but not the β-actin (Supplementary Fig. S7), demonstrating that there was an interaction between CN and tutin. The isothermal titration calorimetry (ITC) analysis showed insufficient heat absorption or release, indicating that the binding force of tutin and CN was low and that cocrystallization and cryo-scanning electron microscopy (cryo-SEM) were not suitable for the detection of CN-tutin interactions. Therefore, microscale thermophoresis (MST) experiments were further applied to investigate the interaction between tutin and CN, and the data demonstrated that tutin directly bound to CN with a dissociation constant (KD) of 0.28 ± 0.27 μM (Fig. 3c). To explore their possible binding sites of CN and tutin, we compared the profiles of CN alone and CN with tutin through HDX-MS.38 Interestingly, the HDX-MS results suggested that CN bound tutin. All peptide deuterium uptake profiles were assayed, and seven peptides were identified using LC-MS/MS. Treatment with tutin altered the rate of H/D exchange of seven peptides (peptides 12-46, 62-72, 81-86, 81-95, 230-259, 232-242, and 243-258) (Fig. 3d, Supplementary Tab. S3). Among these seven peptides, three obvious changed peptides [230-259, 232-242, 243-258] were located in the active site of the catalytic CN subunit, indicating that tutin bound to the active site of the catalytic subunit of CN.Fig. 3Tutin binds to CN. a Tutin increases the CNA thermal stability in living cells by temperature-dependent CETSA ($$n = 3$$). b Tutin increases the CNA thermal stability in living cells by concentration-dependent CETSA. Results were expressed as mean ± SD with $$n = 3$.$ c The MST dose-response curve shows the interaction between tutin and CN. Results were expressed as mean ± SD with $$n = 3$.$ d Deuterium uptake of peptides targeted by tutin. HDX graphs of CN alone (black curves) and bound both to tutin (red curves). e Structural overview of a predicted CN-tutin complex model (CNA: gray, CNB: orange, tutin: yellow, residues: green). Zoom-in view of the predicted CN-tutin interface. f Interface residues (Arg 254 and Ala 283) in CN are shown and labeled by the names and positions of residues Considering the results of the HDX-MS, we performed active-pocket molecular docking with Schrodinger’s soft. To further predict how tutin interacts with CN, we modeled tutin binding to CN crystal structure (PDB ID: 6NUU). The results showed that tutin exhibited excellent docking to CN. Tutin may anchor in a CN binding site via the following interactions. The hydrogen bonds formed between the ethylene oxide moiety in tutin and Arg 254 or Ala 283 in CN (Fig. 3e, f). To further confirm this conclusion, two mutations (CN-R254K and CN-A283V) have been created, and MST experiments were carried out. The MST data revealed that the binding force was significantly decreased after the mutations. The KD of CN-R254K and tutin was 7.64 ± 4.48 μM and the KD of CN-A283V and tutin was 70.8 ± 45.6 μM respectively (Supplementary Fig. S8), both are significantly higher than the KD of wild type CN and tutin (0.28 ± 0.27 μM, showed in Fig. 3c). This result further confirms that tutin binds to Arg254 and Ala283 in CN.
## Tutin activates CN in vitro and in vivo experiments
We have confirmed that tutin binds with the catalytic domain of CN. However, whether tutin affects the enzyme activity of CN was unclear. Therefore, whether tutin influences CN activity was measured in vitro. The results showed that tutin significantly activated CN in a dose-dependent manner in vitro (Fig. 4a). Moreover, CN activities in the hippocampal and cortical regions of mice was measured (0.5, 1, 2, 6, 12, 24 and 72 h) after seizures induced by tutin. The data revealed that significant increases in both basal and maximal CN activities were observed after the seizures, peaking at 0.5-6 h, and decreased to basal levels in the hippocampus and cortex after 24 h (Fig. 4b, c). Besides, to better visualize the dose-dependent effect of tutin on CN activity, mice were injected with tutin (0, 1.6, 1.8, 2.0, 2.2 mg/kg). And then CN activities in the hippocampus and cortex were observed at 2 h, 12 h and 24 h after seizures. The results indicated that tutin increased CN activity in a dose-dependent manner in vivo (Fig. 4d, e). Moreover, it was unclear whether the identified CN activity was consistent with its expression. Western blot results showed that tutin did not affect CNA level in the hippocampus (Fig. 4f) or cortex (Fig. 4g), suggesting that tutin affects CN activity, not its expression. Collectively, the data confirmed that tutin activates CN in vitro and in vivo. Fig. 4Tutin activates CN activity in vitro and in vivo. a Tutin activates CN in vitro, and results were presented as mean ± SD with $$n = 3$.$ b, c Changes in CN activity after tutin-induced epilepsy. Basal and max CN activity of hippocampus or cortex were assayed (0.5, 1, 2, 6, 12, 24, 72 h). (* $P \leq 0.05$, **$P \leq 0.01$ vs. control group (basal). # $P \leq 0.05$, ##$P \leq 0.01$ vs. control group (max.)). Results were represented as mean ± SD with $$n = 9$.$ d, e Changes in CN activity of hippocampus or cortex after tutin-induced epilepsy. Mice were injected with different doses (0, 1.6, 1.8, 2.0, 2.2 mg/kg) of tutin, CN activity of hippocampus or cortex was assayed (2, 12, 24 h) (*$P \leq 0.05$, **$P \leq 0.01$ vs. saline group). Results were expressed as mean ± SD with $$n = 8$.$ f, g CNA expression in hippocampus or cortex by Western blot. Results were represented as mean ± SD with $$n = 8$$
## CN inhibitor FK506 antagonizes epilepsy induced by tutin in vivo
The CN enzyme-specific inhibitor FK506, was used to confirm that inhibiting CN activity reduces the stage of seizure induced by tutin. Seizure episodes induced by tutin showed typical increases in intensity. Compared with that in the tutin-challenged mice, pretreatment with FK506 obviously reduced the percentage of stage V seizures (Fig. 5a, b). Moreover, FK506 decreased the frequency of seizures and significantly reduced the duration of single seizure in EEG analysis (Supplementary Fig. S9). Since epilepsy can cause damage to neurons, particularly in the hippocampus,5,39,40 Nissl staining was performed in this study.41 The Nissl staining data showed that neurons were lost in both the hippocampus and cortex in the tutin-inuced mice, and FK506 significantly alleviated the injury of neurons in the hippocampal CA1 and CA3, while neurons in cortex were not significantly improved by FK506 (Fig. 5c, d). Other CN inhibitors had also been investigated, including Pimecrolimus and Cyclosporin. Consistent with FK506, Pimecrolimus significantly decreased the intensity of tutin-induced epilepsy. Cyclosporin showed similar effect on alleviating tutin-induced epilepsy in mice though without significance (Supplementary Tab. S4). Altogether, these data suggested that inhibition of CN activity may antagonize epilepsy induced by tutin, indicating that CN is a target of tutin during seizure induction. Fig. 5The CN inhibitor FK506 alleviates the intensity of seizures and reduces neuronal loss. a Percentage of tutin-induced seizures in mice with pre-treatment with FK506 ($$n = 40$$, *$P \leq 0.05$). b FK506 reduced the number of mice reached stage 5 epilepsy by tutin ($$n = 40$$, *$P \leq 0.05$). c Nissl staining of hippocampal CA1, CA3 and cortical neurons after tutin-induced seizures (scale bar: 100 μm). d Results of nissl staining were represented as mean ± SD with $$n = 8$$, **$P \leq 0.01$ vs. Control group, #$P \leq 0.05$ vs. Tutin group
## Knockdown of the CNA gene expression reduces tutin-induced epilepsy and neuronal damage in vivo
To further verify that CN is a target of tutin in epilepsy, we knocked down CNA gene expression in the brains of mice. Our preliminary experiment indicated that CNA-short interfering RNA (siRNA)-1 and CNA-siRNA-3 could effectively decrease the level of CNA gene expression in N2a cells (Supplementary Fig. S11). Therefore, CNA-siRNA-($\frac{1}{3}$) was applied to knock down CNA expression in vivo. Mice were microinjected with adeno‐associated virus (AAV)‐shRNA‐CNA into the left lateral ventricle, and sham control mice were injected with an empty AAV vector. After 30 days, the mice were injected with tutin, and seizure-like behaviors were observed (Fig. 6a). The expression of CNA in the hippocampus of the mice injected with adenovirus was significantly decreased (Fig. 6b). Behavior data showed that CNA-knockdown mice exhibited decreased intensity of the seizure attacks compared with sham mice (Fig. 6c). Moreover, the seizure score of the modified Racine scale was markedly decreased in the mice with CNA expression knocked down (Fig. 6d). The EEG results showed that CNA knockdown also inhibited the epileptic form discharges and significantly decreased the frequency of seizures (Supplementary Fig. S10). We then used Nissl staining to observe whether the reduction in CNA expression exerted a protective effect on neurons. The Nissl staining results showed that the neurons in the tutin-challenged mice were severely damaged in hippocampus (CA1, CA3) and cortex (Fig. 6e, f), while CNA knockdown significantly reduced the loss of neurons. These data further demonstrated that CN is a molecular target of tutin because knocking down CNA expression alleviated seizure severity and attenuated neuronal loss. Fig. 6CNA knockdown reduces the intensity of seizures and neuronal loss. a Experimental flowchart was shown. b CNA expression in hippocampus by Western blot. Quantitative analysis of CNA in hippocampus of mice, and results were expressed as mean ± SD with $$n = 6$$, **$P \leq 0.01$, ***$P \leq 0.001$, ****$P \leq 0.0001.$ c Behavioral episodes in mice with CNA knockdown induced by tutin ($$n = 20$$, *$P \leq 0.05$, **$P \leq 0.01$). d Distribution of maximum Racine score within 2 h after seizures, and results were represented as mean ± SD with $$n = 20$$ (*$P \leq 0.05$, **$P \leq 0.01$). e Nissl staining of hippocampal CA1, CA3 and cortical neurons after seizures (scale bar: 100 μm). f *Statistical analysis* of Nissl-stained cells of hippocampal CA1, CA3 and cortical neurons after seizures; Data are shown as mean ± SD with $$n = 8$$, *$P \leq 0.05$, **$P \leq 0.01$ vs.Tutin [-] -shRNA [-] group, #$P \leq 0.05$, ##$P \leq 0.01$ vs. Tutin (+) -shRNA [-] group
## Discussion
Targeted studies of natural products are critical for elucidating their mechanisms of action.31,42,43 As one of most importantly toxic molecules that cause epilepsy, tutin has been investigated to identify its neurotoxic targets through chemical proteomic approaches in this study. In our study, we demonstrated that tutin caused epilepsy in mice by activating CN with high neurotoxicity. Tutin bound CN by forming two hydrogen bonds in the active site to trigger CN activation in vitro and in vivo. It was confirmed that FK506 reduced the severity of epileptic seizures and degree of neuronal loss in mice treated with tutin. Furthermore, when CNA expression was knocked down in vivo, tutin-induced severe seizure rate was significantly reduced and nerve damage in mice with epilepsy was ameliorated. Therefore, we speculate that CN is an important target of tutin induction of seizures.
CN plays a central role in brain synaptic plasticity. Diazepam facilitates the inhibitory effects of GABA via targeting GABA receptors.33,44 Diazepam can significantly reverse tutin-induced seizures in mice, and it has been reported that CN downregulates GABAA receptor function and that activated CN may interact with synaptic GABAA receptors in the hippocampus CA1 inhibitory synapses.34 Injection of MK-801, a selective NMDA receptor antagonist, has been previously shown to block increases in CN activity, suggesting a role for calcium influx mediated by NMDA receptors in increased CN activity.33 In addition, martentoxin, a selective inhibitor for BK channel, exerted an antagonistic effect on epilepsy induced by tutin, suggesting that BK channel might be involved in tutin-induced epilepsy. In summary, tutin induces epileptic seizures in mice by activating CN, and NMDA receptors, GABA receptors and BK channels might be involved in related signaling pathways.
By employing various approaches, including TPP, CN inhibition and CNA knockdown, we verified that CN was, indeed, an important target of tutin during epileptic seizure induction. In neurons, CN plays an important role in epileptic seizures. During epileptiform activity, activation of CN has been reported to be related to α2 subunit containing GABAA receptors. In the present experiments, we found that tutin activated CN in vitro and in vivo, and MST results suggested tutin bound to CN directly. The HDX-MS experiments, the molecular docking and MST analysis of two mutations indicated that tutin binds to the active site of the CN catalytic subunit. However, in contrast to the CN inhibitor FK506 binding site, the FK506-FKBP12 binary complex did not contact the phosphatase active site in CN.45 To better characterize the direct association between tutin and CN, further studies with other methods are needed.
Although our study verified that CN might be a possible potential target of tutin, we could not completely exclude the role of the other four proteins in the TPP-TR experiment. Literature search found that Aldh7a1 and Aldh3b1 are the members of Aldh, and Aldh7a1 has been linked to pyridoxine-dependent epilepsy.46 To explore the possible relationship between Aldh3b1 and tutin, the effects of pyridoxine were tested on tutin-treated mice. The results showed that pre-treatment with pyridoxine only had a certain auxiliary effect in epilepsy induced by tutin (Supplementary Fig. S12), suggesting that Aldh3b1 might not be a main target of tutin. The functions of Mvd, Lancl1, Aldh3b1 and Hsdl2 in epilepsy are still worthy to be explored in our future studies.
In summary, we explained, for the first time, the epileptogenic mechanism of the neurotoxic molecule tutin. We look forward to further exploring the activation mode of CN and the triggering mechanism of epilepsy to provide new strategies for the development of antiepileptic drugs in the future.
## Reagents
Tutin (C15H18O6) was isolated from the roots of *Coriaria nepalensis* as described previously.47 Molecular weight of tutin was 294.11 and the purity of tutin was above $95\%$. The structure was identified by comparing its spectroscopic data (NMR and HR-ESI-MS).
## Primary cultured hippocampal neurons
Primary hippocampal cells were isolated from rats (Sprague Dawley, 24 h) (SiBeiFu Biotechnology, Beijing). The hippocampus was dissected out, and incubated with $0.05\%$ trypsin (Gibco, USA) for 15 min at cell incubator, and then the suspension was filtrated through 70 μm filters. The cells were plated on 6-well plates (1 × 106 cells) with Dulbecco’s modified Eagle’s medium (Gibco, USA), $10\%$ fetal bovine serum (Gibco, USA), 100 U/ml penicillin and 100 μg/mL streptomycin (Invitrogen) overnight at cell incubator. After that, the cell medium was replaced by *Neurobasal medium* and cultured for 10 days. On day 10, monoclonal antibodies against NeuN (ABclonal, China) and MAP-2 (ABclonal, China) were used to assess the cell populations in the culture. Ten-day-old cultures composed of >$95\%$ neurons were used for this study (Supplementary Fig. S13).
## TPP
TPP experiments were performed as previous studies with minor modifications.31 Primary neurons were supplemented with tutin (5 μM) or PBS. Primary neurons were digested with trypsin, and suspended in PBS. Neurons were pelleted (600 × g, 5 min), and the supernatants were discarded. Neurons were resuspended in PBS (5 mL) and centrifuged (600 × g, 5 min), which were reconsistuted in 1 mL PBS. All the samples were separated into 10 fractions, and then heated at the following temperatures (37 °C-67 °C) for 3 min by Eppendorf Thermomixers. After that the samples were incubated at RT for 3 min, neurons were lysed with four freeze-thaw cycles (incubation at 35 °C for 30 min and snap-freezing by liquid nitrogen), and centrifuged (15,000 × g, 30 min) and then the supernatants were obtained for MS.
The concentration of 37 °C sample was measured and applied to normalize the volume for TMT labeling. After calculating the volume of each sample at 25 µg, 100 mM DTT (15.43 mg/mL) (8 M urea preparation) was added at 56 °C for 30 min. The protein samples were transferred to a pre-labeled 10 KDa ultrafiltration tube and centrifuged (15,000 × g, 20 min, 4 °C), and NH4HCO3 solution (150 µL, 50 mM) was added and centrifuged (15,000× g, 20 min, 4 °C). Iodoacetamide (20 mM dissolved in 50 mM NH4HCO3) and samples were incubated for 30 min, and then centrifugated for 20 min (15,000 g, 4 °C). NH4HCO3 solution (150 µL, 50 mM) was added and centrifuged (15,000 × g, 20 min, 20 °C), and then TEAB solution (150 µL, 100 mM) was added and centrifuged. TEAB solution was repeatedly added and centrifuged. Proteins were incubated with trypsin (1:50) for 12-14 h at 37 °C.
Peptides labeling was carried out by TMT reagents (Thermo Scientific), and anhydrous acetonitrile (41 µL) was added to dissolve TMT (0.8 mg). TMT solution (10 µL) was added to the samples (20 °C, 600 rpm, 1 h), and hydroxylamine (5 %, 8 µL) was added to quench the reaction, and then the labeled peptides were dried under vacuum centrifugation.
Peptide fractionation was carried out using an ACQUITY Arc Bio system (Waters) equipped with a Waters Bridge column (3.5 μm, 150 × 2.1 mm). The solvent A was $98\%$ H2O $2\%$ MeCN and $0.1\%$ ammonia while the solvent B was $98\%$ MeCN $2\%$ H2O and ammonia. A 51 min gradient procedure was described as follows: 10 min $0\%$ B, 0.1 min to $5\%$ B, 1.9 min to $8\%$ B, 11 min to $16\%$ B, 21 min to $32\%$ B, 1 min to $95\%$ B, 1 min $95\%$ B, 2 min to $15\%$ B and 3 min $15\%$ B, 400 μL/min, 214 nm. Fractions were dried, reconstituted in $1\%$ FA in ddH2O, centrifuged (15,000 g, 4 °C, 30 min), and subjected to LC-MS/MS analysis. The experiment was performed in duplicates.
The peptides were tested by ultimate 3000 system coupled with an Orbitrap Fusion Lumos Mass spectrometer (Thermo Fisher Scientific, USA). The column of LC was an analytical column (50 μm, 15 cm) packed with 2 μm RSLC C18 (Mobile phase-A: water with $0.1\%$ FA, B: $80\%$ acetonitrile with $0.1\%$ FA). The gradient was as follows: 5 min of $4\%$ B, $4\%$-$30\%$ B for 65 min; $30\%$-$80\%$ B for 5 min; $80\%$ B for 5 min; $80\%$-$4\%$ B for 5 min and $4\%$ B for 5 min, 300 nL/min. The Orbitrap analyzer with 60,000 resolution (FWHM) was used to acquire the full scan MS spectra (m/z 350 to 1500). The elucidation of data was analyzed by Proteome Discoverer (2.3) workstation.
## CETSA
Neurons were heated at different temperatures (50-70 °C) for 3 min and lysed, which were separated by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) and incubated with CNA antibody.
## CN activity assay in vitro
CN activity in vitro was determined according to the instructions of recombinant/purified CN activity kits (Abcam). Briefly, after calcineurin and calcineurin assay buffer were mixed, different concentrations of tutin are added for 5 min. The calcineurin substrate was added to start the reaction. This assay was then terminated by thegreen assay reagent (100 μL). The absorbance of OD620 was detected and the CN activity was calculated.
## Animals and treatments
C57/BL mice (male, 22–26 g) were supplied by SiBeiFu (Beijing) Biotechnology Co., Ltd. (China). Mice were raised at 24 °C with 12 h light/dark cycles, which is accessible to food and water freely. Mice were adapted for 7 days to before the experiments. All the experiments, especially epileptic seizure score experiments, antiepileptic drug therapy experiments, enzyme activity test, CN inhibitor assay, CNA knockdown experiments and Nissl staining tests were performed with the standard of double-blinded way. In antiepileptic drug therapy experiment, the mice were injected with tutin (2 mg/kg, i.p.) after oral administration of MK-801 (0.5 mg/kg), Retigabine (60 mg/kg), Diazepam (3 mg/kg), Carbamazepine (20 mg/kg). In the antiepileptic evaluation experiment of martentoxin, martentoxin (0.05 μg in 1 μL saline) was microinjected into the region of hippocampus, and then the mice were injected with tutin. In CN inhibitor assay, CN inhibitor FK506 (0.5 mg/kg, i.p.) was administrated to mice 1 h before tutin injection. The animal trails were carried out following the rules of the Institutional Animal Care and Use Committee of Peking Union Medical College. All procedures were carried out according to “three Rs” principle.
## Stereotaxic injection
Adeno‐associated virus (AAV)‐shRNA‐CNA were intracerebroventricularly (icv.) injected into the left lateral ventricle (Y: +0.5 mm, X: +1.0 mm Z: −2.5 mm) as previously reported. Anesthetized mice were fastened to a brain stereotaxic apparatus (RWD Life Science, China). For a single icv. injection of shRNA-CNA, 2 μL of adenovirus were injected and the microsyringe was kept for 5 min and slowly retracted. The wound was sutured, and then closed in layer with penicillin powder.
## Assessment of seizure
Epileptic activity of mice induced by tutin was scored for 2 h using a modified Racine scale as follows48: Stage 0: no reaction; Stage 1: facial clonus, including blinking, locomotor whiskers, rhythmic mastication, etc. Stage 2: including stage 1 and rhythmic nodding; Stage 3: fore clonus in addition to stage 2; Stage 4: including stage 3 and standing with hind legs; Stage 5: fall or jump, repeated convulsions or convulsions resulting in death on a stage 4 basis.
## EEG analysis
Male C57BL/6 mice were applied for EEG monitoring. Two electrodes were implanted, and one screw serving was inserted as the ground electrode into the skull through a drilled hole. The electrodes and screws were fixed with bone cement. Mice had a seven-day postoperative rest before the drug study. The mice were housed individually. A burst of high-amplitude EEG activity represents a seizure.
The baseline was measured for 0.5 h. The EEG signals in mice injected with vehicle or FK506 were measured for 1 h. Acute seizures were induced by intraperitoneal injection of tutin and the EEG signals were recorded for 150 min. The EEG signals in sham/knockdown mice injected with tutin were measured for 2 h.
## Microdialysis analysis
After fixation of mice, the microdialysis-guided cannula was inserted into the hippocampus (coordinate: A, −1.8; L, +1.5; H, −1.0 mm from bregma) and the cortex (coordinate: A, +1.5; L, +1.5; H, −1.0 mm from bregma) of mice. Microdialysis studies were conducted 7 days later. CMA 7 Metal Free probe (CMA, Sweden) with a 1-mm and 6-kDa-cutoff regenerated cellulose membrane was inserted gently through the CMA 7 guide cannula. The probe was equilibrated with artificial cerebrospinal fluid at a flow rate of 1 μL/min for 1 h prior to initiation of dialysis. After that, baseline samples were collected into vials for 30 min, then mice were injected with tutin and dialysates were collected every 30 min for 1 h. Microdialysis samples were dried under vacuum centrifugation, and re-dissolved with deionized water. After benzoylation reaction, the standard or samples were analyzed by LC-MS.
## CN activity assay
CN activity was measured following a modification of the procedures as previous studies.49 Briefly, the basal wells contained 25 mM MOPS (pH 7.0), 2 mM p-nitrophenol phosphate (pNPP), 1 mM DTT, 2 mM EGTA, and 2 mM EDTA. Other than that, the maximum wells contained 2 mM MnCl2. The final volumes of all wells were 100 μL. Brain region homogenate (200 μg/mL) was added to start the reaction at 37 °C for 0.5 h. The reaction was then terminated by placing on ice and OD was measured at 405 nm. The absorbance units were converted to the concentrations of p-nitrophenol (pNP) through comparing with a standard absorption curve of pNP concentration.
## Western blot
Brains were homogenized in lysis buffer and separated by $10\%$ SDS-PAGE, and then transferred into PVDF membranes (Merck Millipore, USA). Membranes were incubated with skimmed milk ($5\%$) for 1.5 h and then incubated with the an anti-rabbit antibody of CNA (ab109412, 1:1000, Abcam, USA) and β-actin overnight at 4 °C, which were incubated with the secondary antibody (1:20000, Santa Cruz Biotechnology, USA) for 2 h. All blots were developed with enhanced chemiluminescence regents (Merck Millipore, USA) and analyzed by Image J 1.53 software.
## Nissl staining
Mouse brains were fixed in the $4\%$ paraformaldehyde, and then fixed with paraffin. After that, 3 mm brain slices were immersed in $1\%$ cresyl violet (50 °C, 1 h) and dehydrated with different ethanol solution, and then brain sections were cleared with xylene. Nissl-staining cells of cortex, hippocampal regions were imaged by light microscope (NIKON E600, Japan) and analyzed by Image-Pro Plus.
## RNA interference
Neuro-2a cells were cultured in 12 well plates (104/well). Negative control (NC) or CNA-siRNA were transfected by LipofectamineTM RNAiMAX transfection reagent for 48 h according to the instruction. CNA-siRNA sequence as follows: siRNA-1: 5′-GCCGTTCCATTTCCACCAAdTdT-3′; siRNA-2: 5′-GCGCTACTGTTGAGGCTATdTdT-3′; siRNA-3: 5′-GCAGTAATAGCAGCAATATdTdT-3′.
## CN expression and purification
DNA coding sequence of the human CNA/CNB (CNA: M1-N370, CNB: 16-170),50 was subcloned into a pET-15b (+) (Huada Gene BGI, Shenzhen) expression vector. The CN protein was expressed as described. Briefly, protein was expressed using E. coli strain Rosetta (DE3) (Tiangen Biotech (Beijing) Co., Ltd.) in 5 L Luria-*Bertani medium* induced with isopropyl-β-D-thiogalactoside (1 mM). Protein purification used a Ni-NTA Agarose column (Lot #163026181, Qiagen). Conditions for Ni-NTA affinity chromatography were 20 mM Tris-HCl, 10 mM imidazole, 150 mM NaCl, pH 7.5, as for the elution buffer, the concentration of imidazole was raised to 300 mM. The protein concentration and purity (>$85\%$) were estimated using NanoDrop™ 2000 spectrophotometers (Thermo Fisher Scientific) and SDS-PAGE (Bio-Rad) respectively (Supplementary Fig. S14). CN was concentrated and stored at -80 °C.
## HDX-MS analysis
Before 2 h of HDX analysis, the compound tutin (200 μM) was added into the sample, with the control sample adding an equal volume of tutin buffer. For deuterium labeling, CN (4 μM) in the buffer (20 mM Tris-HCl, 1 mM CaCl2, 0.5 mM TCEP, and 150 mM NaCl, in H2O, pH 7.5) in the presence or absence of 200 μM tutin was diluted 10-fold by the labeling buffer containing 20 mM Tris-HCl, 1 mM CaCl2, 0.5 mM TCEP, and 150 mM NaCl, in $100\%$ D2O at pD 7.4. After incubation for 30, 90 or 300 seconds at 25 °C, the same volume of ice-cold quench buffer containing 4 M guanidine hydrochloride, 500 mM TECP and 200 mM citric acid in water solution at pH 1.8, $100\%$ H2O, was added to quench deuterium uptake. The sample was digested with pepsin (Promega) on ice for 5 min, and removed by centrifugation. An ACQUITY UPLC BEH C18 column (2.1 μm, 1.0 mm × 50 mm, Waters) equipped with an Ultimate 3000 UPLC system (Thermo Scientific) were used for the obtained peptides separation. A Q Exactive mass spectrometer was used for mass spectrometry analysis of the peptides. Mass spectrometry data were compared with Proteome Discoverer (Thermo Scientific) to match the corresponding peptide in CN. XCALIBUR (Thermo Scientific) was used to inspected peptide peaks. In order to estimate the max deuterium uptake of peptides, a repeated experiment was performed extending incubation in D2O for 24 h. HDExaminer (Sierra Analytics) was used for calculating deuterium uptake levels. Deut % for different peptides were calculated as follows.\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\mathrm{Deut}}}}_i\% = \frac{{\# {{{\mathrm{D}}}}_i/(\# (- {{{\mathrm{CO}}}} - {{{\mathrm{NH}}}} -)_i - \# {{{\mathrm{Pro}}}}_i - 1)}}{{{{{\mathrm{Max}}}}\,{{{\mathrm{D}}}}_i}} \times 100\%$$\end{document}Deuti%=#Di/(#(−CO−NH−)i−#Proi−1)MaxDi×$100\%$# Di: deuterium numbers for peptide i at a certain hydrogen/deuterium exchange time; \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\# (- {{{\mathrm{CO}}}} - - {{{\mathrm{NH}}}} -)_i$$\end{document}#(−CO−−NH−)i: amide bond numbers of each peptide; # Proi: the proline number for peptidei; Mxx Di: maximum deuterium uptake for peptide i.
## Molecular docking
The docking study was performed by using Schrodinger’s soft. The protein coordinates were retrieved from the Protein Data Bank (PDB code: 6NUU). The structures of tutin were generated and energy-minimized. Both the protein and the ligands were prepared by adding polar hydrogen atoms.
## MST assay
The proteins were labeled using the Monolith His-tag Labeling Kit RED-NHS 2nd Generation Kit. Tutin was diluted in a series of concentration. The labeled CN was diluted to working assay buffer (40 nM, $0.05\%$ Tween-20). The mixture was incubated and then loaded into Monolith standard-treated capillaries. The thermophoresis was detected by A Monolith NT.115 instrument (Nano Temper Technologies) and KD values was calculated by NT Analysis software (Nano Temper Technologies).
## Statistic analysis
Statistic analysis was performed using one-way ANOVA for multiple group comparison and unpaired Student’s t-test for two groups. The proportion of animals with stages 0-4 and stage 5 seizures was analyzed by χ2 test. $P \leq 0.05$ was regarded obviously significant.
## Supplementary information
Supplementary Information Supplementary MS Data The online version contains supplementary material available at 10.1038/s41392-023-01312-y.
## References
1. Singh A, Trevick S. **The Epidemiology of Global Epilepsy**. *Neurologic Clin.* (2016) **34** 837-847. DOI: 10.1016/j.ncl.2016.06.015
2. Beghi E. **Epileptic seizures and epilepsy: definitions proposed by the International League Against Epilepsy (ILAE) and the International Bureau for Epilepsy (IBE)**. *Epilepsia* (2010) **46** 470-472
3. Rydzanicz M. **Novel calcineurin A (PPP3CA) variant associated with epilepsy, constitutive enzyme activation and downregulation of protein expression**. *Eur. J. Hum. Genet* (2019) **27** 61-69. DOI: 10.1038/s41431-018-0254-8
4. Wen Y. **Inhibition of Calcineurin A by FK506 Suppresses Seizures and Reduces the Expression of GluN2B in Membrane Fraction**. *Neurochem Res* (2017) **42** 2154-2166. DOI: 10.1007/s11064-017-2221-0
5. Shetty AK, Hattiangady B. **Restoration of calbindin after fetal hippocampal CA3 cell grafting into the injured hippocampus in a rat model of temporal lobe epilepsy**. *Hippocampus* (2007) **17** 943-956. DOI: 10.1002/hipo.20311
6. Dai L. **Target identification and validation of natural products with label-free methodology: A critical review from 2005 to 2020**. *Pharm. Ther.* (2020) **216** 107690. DOI: 10.1016/j.pharmthera.2020.107690
7. Chen X. **Target identification of natural medicine with chemical proteomics approach: probe synthesis, target fishing and protein identification**. *Signal Transduct. Target Ther.* (2020) **5** 72. DOI: 10.1038/s41392-020-0186-y
8. Ma T. **Low-dose metformin targets the lysosomal AMPK pathway through PEN2**. *Nature* (2022) **603** 159-165. DOI: 10.1038/s41586-022-04431-8
9. Fuentealba J. **Inhibitory effects of tutin on glycine receptors in spinal neurons**. *Eur. J. Pharm.* (2007) **559** 61-64. DOI: 10.1016/j.ejphar.2006.12.018
10. He S. **Behavioral and Brain Structural Changes in Kindled Rats Induced by Coriaria Lactone/Pentylenetetrazol**. *Front Behav. Neurosci.* (2021) **15** 727872. DOI: 10.3389/fnbeh.2021.727872
11. Okuda T, Yoshida T. **The correlation of coriamyrtin and tutin, and their absolute configurations**. *Chem. Pharm. Bull. (Tokyo)* (1967) **15** 1955-1965. DOI: 10.1248/cpb.15.1955
12. Craven BM. **Molecular Structure of Tutin**. *Nature* (1963) **197** 1193-1194. DOI: 10.1038/1971193c0
13. Zhou H, Tang YH, Zheng Y. **A new rat model of acute seizures induced by tutin**. *Brain Res* (2006) **1092** 207-213. DOI: 10.1016/j.brainres.2006.03.081
14. Hong Z. **A novel kindling model of temporal lobe epilepsy in rhesus monkeys induced by Coriaria lactone**. *Epilepsy Behav.* (2013) **29** 457-465. DOI: 10.1016/j.yebeh.2013.07.028
15. Fuentealba J. **Potentiation and inhibition of glycine receptors by tutin**. *Neuropharmacology* (2011) **60** 453-459. DOI: 10.1016/j.neuropharm.2010.10.023
16. Nistri A, Constanti A, Quilliam JP. **Central Inhibition, G.A.B.A., and Tutin**. *Lancet* (1974) **303** 996-997. DOI: 10.1016/S0140-6736(74)91311-7
17. Sills GJ, Rogawski MA. **Mechanisms of action of currently used antiseizure drugs**. *Neuropharmacology* (2020) **168** 107966. DOI: 10.1016/j.neuropharm.2020.107966
18. Ueno H. **Alpha-pinene and dizocilpine (MK-801) attenuate kindling development and astrocytosis in an experimental mouse model of epilepsy**. *IBRO Rep.* (2020) **9** 102-114. DOI: 10.1016/j.ibror.2020.07.007
19. Kim RY, Pless SA, Kurata HT. **PIP2 mediates functional coupling and pharmacology of neuronal KCNQ channels**. *Proc. Natl Acad. Sci. USA* (2017) **114** E9702-E9711. DOI: 10.1073/pnas.1705802114
20. Brodie MJ. **Sodium Channel Blockers in the Treatment of Epilepsy**. *CNS Drugs* (2017) **31** 527-534. DOI: 10.1007/s40263-017-0441-0
21. Lin CH. **Effects of anti-epileptic drugs on spreading depolarization-induced epileptiform activity in mouse hippocampal slices**. *Sci. Rep.* (2017) **7** 11884. DOI: 10.1038/s41598-017-12346-y
22. Matthew TJH, Tharakan J, Tai E, Hussein A. **Study of Visual Function in Adult Epileptic Patients on Sodium Valproate or Carbamazepine Monotherapy**. *Cureus* (2019) **11** e4553. PMID: 31275777
23. Liu X. **Selective Blockade of Neuronal BK (alpha + beta4) Channels Preventing Epileptic Seizure**. *J. Med Chem.* (2020) **63** 216-230. DOI: 10.1021/acs.jmedchem.9b01241
24. Yang C. **Homology modeling and molecular docking simulation of martentoxin as a specific inhibitor of the BK channel**. *Ann. Transl. Med* (2022) **10** 71. DOI: 10.21037/atm-21-6967
25. Yang F, Jia G, Guo J, Liu Y, Wang C. **Quantitative Chemoproteomic Profiling with Data-Independent Acquisition-Based Mass Spectrometry**. *J. Am. Chem. Soc.* (2022) **144** 901-911. DOI: 10.1021/jacs.1c11053
26. Zweerink S. **Activity-based protein profiling as a robust method for enzyme identification and screening in extremophilic Archaea**. *Nat. Commun.* (2017) **8** 15352. DOI: 10.1038/ncomms15352
27. Kok BP. **Discovery of small-molecule enzyme activators by activity-based protein profiling**. *Nat. Chem. Biol.* (2020) **16** 997-1005. DOI: 10.1038/s41589-020-0555-4
28. Zhang CJ. **Mechanism-Guided Design and Synthesis of a Mitochondria-Targeting Artemisinin Analogue with Enhanced Anticancer Activity**. *Angew. Chem. Int Ed. Engl.* (2016) **55** 13770-13774. DOI: 10.1002/anie.201607303
29. Jafari R. **The cellular thermal shift assay for evaluating drug target interactions in cells**. *Nat. Protoc.* (2014) **9** 2100-2122. DOI: 10.1038/nprot.2014.138
30. Savitski MM. **Tracking cancer drugs in living cells by thermal profiling of the proteome**. *Science* (2014) **346** 1255784. DOI: 10.1126/science.1255784
31. Kirsch VC. **The Cytotoxic Natural Product Vioprolide A Targets Nucleolar Protein 14, Which Is Essential for Ribosome Biogenesis**. *Angew. Chem. Int Ed. Engl.* (2020) **59** 1595-1600. DOI: 10.1002/anie.201911158
32. Dziekan JM. **Cellular thermal shift assay for the identification of drug-target interactions in the Plasmodium falciparum proteome**. *Nat. Protoc.* (2020) **15** 1881-1921. DOI: 10.1038/s41596-020-0310-z
33. Miletic G. **Changes in calcineurin message, enzyme activity and protein content in the spinal dorsal horn are associated with chronic constriction injury of the rat sciatic nerve**. *Neuroscience* (2011) **188** 142-147. DOI: 10.1016/j.neuroscience.2011.05.013
34. Eckel R, Szulc B, Walker MC, Kittler JT. **Activation of calcineurin underlies altered trafficking of alpha2 subunit containing GABAA receptors during prolonged epileptiform activity**. *Neuropharmacology* (2015) **88** 82-90. DOI: 10.1016/j.neuropharm.2014.09.014
35. Mondragon A. **Overexpression and purification of human calcineurin alpha from Escherichia coli and assessment of catalytic functions of residues surrounding the binuclear metal center**. *Biochemistry* (1997) **36** 4934-4942. DOI: 10.1021/bi9631935
36. Rumi-Masante J. **Structural basis for activation of calcineurin by calmodulin**. *J. Mol. Biol.* (2012) **415** 307-317. DOI: 10.1016/j.jmb.2011.11.008
37. Aramburu J, Rao A, Klee CB. **Calcineurin: from structure to function**. *Curr. Top. Cell Regul.* (2000) **36** 237-295. DOI: 10.1016/S0070-2137(01)80011-X
38. Sun H. **Research advances in hydrogen-deuterium exchange mass spectrometry for protein epitope mapping**. *Anal. Bioanal. Chem.* (2021) **413** 2345-2359. DOI: 10.1007/s00216-020-03091-9
39. Kim EC. **Spontaneous seizure and memory loss in mice expressing an epileptic encephalopathy variant in the calmodulin-binding domain of Kv7.2**. *Proc. Natl Acad. Sci. USA* (2021) **118** e2021265118. DOI: 10.1073/pnas.2021265118
40. Zheng J. **Neuroregenerative gene therapy to treat temporal lobe epilepsy in a rat model**. *Prog. Neurobiol.* (2022) **208** 102198. DOI: 10.1016/j.pneurobio.2021.102198
41. Schmidt-Kastner R, Heim C, Sontag KH. **Damage of substantia nigra pars reticulata during pilocarpine-induced status epilepticus in the rat: immunohistochemical study of neurons, astrocytes and serum-protein extravasation**. *Exp. Brain Res* (1991) **86** 125-140. DOI: 10.1007/BF00231047
42. Dai J. **Chemoproteomics reveals baicalin activates hepatic CPT1 to ameliorate diet-induced obesity and hepatic steatosis**. *Proc. Natl Acad. Sci. USA* (2018) **115** E5896-E5905. DOI: 10.1073/pnas.1801745115
43. Berdan CA. **Parthenolide Covalently Targets and Inhibits Focal Adhesion Kinase in Breast Cancer Cells**. *Cell Chem. Biol.* (2019) **26** 1027-1035.e22. DOI: 10.1016/j.chembiol.2019.03.016
44. Nicholson MW. **Diazepam-induced loss of inhibitory synapses mediated by PLCdelta/Ca**. *Mol. Psychiatry* (2018) **23** 1851-1867. DOI: 10.1038/s41380-018-0100-y
45. Griffith JP. **X-ray structure of calcineurin inhibited by the immunophilin-immunosuppressant FKBP12-FK506 complex**. *Cell* (1995) **82** 507-522. DOI: 10.1016/0092-8674(95)90439-5
46. Mastrangelo M, Cesario S. **Update on the treatment of vitamin B6 dependent epilepsies**. *Expert Rev. Neurother.* (2019) **19** 1135-1147. DOI: 10.1080/14737175.2019.1648212
47. Zhao F. **New sesquiterpenes from the roots of Coriaria nepalensis**. *Tetrahedron* (2012) **68** 6204-6210. DOI: 10.1016/j.tet.2012.05.067
48. Racine RJ. **Modification of seizure activity by electrical stimulation: II. Motor seizure**. *Electroencephalogr. Clin. Neurophysiol.* (1972) **32** 281-294. DOI: 10.1016/0013-4694(72)90177-0
49. Pallen CJ, Wang JH. **Regulation of calcineurin by metal ions. Mechanism of activation by Ni**. *J. Biol. Chem.* (1984) **259** 6134-6141. DOI: 10.1016/S0021-9258(20)82115-2
50. Hendus-Altenburger R. **Molecular basis for the binding and selective dephosphorylation of Na**. *Nat. Commun.* (2019) **10** 3489. DOI: 10.1038/s41467-019-11391-7
|
---
title: Detecting type 2 diabetes mellitus cognitive impairment using whole-brain functional
connectivity
authors:
- Jinjian Wu
- Yuqi Fang
- Xin Tan
- Shangyu Kang
- Xiaomei Yue
- Yawen Rao
- Haoming Huang
- Mingxia Liu
- Shijun Qiu
- Pew-Thian Yap
journal: Scientific Reports
year: 2023
pmcid: PMC9998866
doi: 10.1038/s41598-023-28163-5
license: CC BY 4.0
---
# Detecting type 2 diabetes mellitus cognitive impairment using whole-brain functional connectivity
## Abstract
Type 2 diabetes mellitus (T2DM) is closely linked to cognitive decline and alterations in brain structure and function. Resting-state functional magnetic resonance imaging (rs-fMRI) is used to diagnose neurodegenerative diseases, such as cognitive impairment (CI), Alzheimer’s disease (AD), and vascular dementia (VaD). However, whether the functional connectivity (FC) of patients with T2DM and mild cognitive impairment (T2DM-MCI) is conducive to early diagnosis remains unclear. To answer this question, we analyzed the rs-fMRI data of 37 patients with T2DM and mild cognitive impairment (T2DM-MCI), 93 patients with T2DM but no cognitive impairment (T2DM-NCI), and 69 normal controls (NC). We achieved an accuracy of $87.91\%$ in T2DM-MCI versus T2DM-NCI classification and $80\%$ in T2DM-NCI versus NC classification using the XGBoost model. The thalamus, angular, caudate nucleus, and paracentral lobule contributed most to the classification outcome. Our findings provide valuable knowledge to classify and predict T2DM-related CI, can help with early clinical diagnosis of T2DM-MCI, and provide a basis for future studies.
## Introduction
Type 2 diabetes mellitus (T2DM), accounting for the highest percentage of adults with diabetes, is a series of chronic endocrine and metabolic abnormalities. T2DM is related to clinical complications such as cognitive impairment (CI) and dementia. T2DM patients are at a 1.5 times higher risk for dementia or cognitive decline than individuals without diabetes1,2. Patients with diabetes manifest brain atrophy and microvascular disease in magnetic resonance imaging (MRI) exams3. However, factors contributing to the onset and progression of brain complications in patients with T2DM remain unclear. Therefore, there is an urgent need to identify these factors and early detection as the prevalence of T2DM is rising with population aging.
Increasing evidence indicates that patients with T2DM present structural and functional brain pathological changes4. In China, the prevalence of mild cognitive impairment (MCI), the prodromal stage of Alzheimer’s disease (AD), is $45\%$ (ranges from 21.8 to $67.5\%$) in older patients with T2DM, substantially higher than $14.71\%$ in older populations without T2DM. Prevalence is higher in older women5–7. MCI can gradually develop into moderate or severe CI and even AD5. Since AD cannot be completely cured, early detection and pharmacological and behavioral interventions of MCI are crucial for reducing the risk for AD8.
In the past 20 years, high-precision brain imaging techniques, such as structural, functional, and diffusion MRI as well as positron emission tomography (PET), have been demonstrated to be effective for investigating brain changes in patients with T2DM and MCI (T2DM-MCI)9. For instance, the blood oxygen level-dependent (BOLD) signal in fMRI, which reveals hemodynamic changes associated with neural activities, has been used to detect altered functional connectivity (FC) in patients with T2DM-MCI5,10–12. Diffusion tensor imaging (DTI), which quantifies the diffusion anisotropy of water molecules in white matter (WM), has been used to explore disruptions of structural network connectivity13–16. PET has been used to capture metabolic changes in the brain for early diagnosis17.
FC can be employed to reflect the functional condition of the brain, diagnose neurodegenerative diseases, and provide in-depth insights into pathophysiological mechanisms8,18–20. Region-specific FC provides useful features for T2DM-MCI classification. However, existing T2DM-MCI classification methods based on FC have limited accuracy of less than $70.0\%$21,22. Therefore, further effort is needed to improve the specificity and accuracy. The eXtreme Gradient Boosting (XGBoost)23,24, which improves classification based on iterative learning of weak classifiers. On a single machine, XGBoost is more than tenfold faster than existing popular solutions, with scalability to handle billions of samples. Model learning can be accelerated with parallel and distributed computing23. XGBoost is used in various applications25,26 owing to its high efficiency and accuracy. However, no previous study has used XGBoost for classifying patients with T2DM-MCI based on FC features. Therefore, the aim of the present study was to identify meaningful features to specifically distinguish T2DM-MCI.
Based on previous evidence, this study focused on the following objectives: [1] implement an efficient XGBoost classifier for T2DM-MCI classification; and [2] determine brain regions that distinguish patients with T2DM-MCI, providing a basis for early clinical diagnosis and interventional treatments.
## Clinical and neuropsychological results
A total of 199 participants underwent MRI, clinical blood, and neuropsychological scale tests and fulfilled the inclusion criteria. The mean ages at scanning for the T2DM-MCI, T2DM-NCI, and NC groups were 46.17 ± 8.67, 50.78 ± 8.28, and 46.30 ± 10.40 years, respectively. The demographic, clinical, and neuropsychological characteristics of the 199 participants are summarized in Tables 1,2,3. There were no significant differences between the T2DM-MCI, T2DM-NCI, and NC groups in sex, age, or educational level ($p \leq 0.05$). There were no statistically significant differences between the T2DM-MCI and T2DM-NCI groups in glycated hemoglobin levels, body mass index (BMI), or fasting blood glucose levels($p \leq 0.05$). BMI was significantly different between the T2DM-NCI and NC groups ($$p \leq 0.001$$). Compared with the T2DM-MCI group, the T2DM-NCI and NC groups had higher levels of auditory verbal learning test (AVLT, immediate: $$p \leq 0.002$$ and $$p \leq 0.003$$; 5 min: $$p \leq 0.000$$ and $$p \leq 0.000$$; delay: $$p \leq 0.000$$ and $$p \leq 0.000$$; recall: $$p \leq 0.000$$ and $$p \leq 0.000$$), digit span test (reverse, $$p \leq 0.000$$ and $$p \leq 0.001$$), Montreal Cognitive Assessment-B (MoCA-B, $$p \leq 0.000$$ and $$p \leq 0.000$$), digit symbol substitution (DSST, $$p \leq 0.000$$ and $$p \leq 0.000$$), and lower levels of grooved pegboard test (GPT, L: $$p \leq 0.005$$ and $$p \leq 0.007$$; R: $$p \leq 0.000$$ and $$p \leq 0.000$$). There were no statistically significant differences in the other neuropsychological test outcomes among the three groups ($p \leq 0.05$) (Tables 1,2,3).Table 1Demographic results of T2DM-MCI, T2DM-NCI and NC groups. T2DM-NCIT2DM-MCINCpMeanSDMeanSDMeanSDAge (years)46.178.6750.788.2846.310.4030.058Gender (M/F)$\frac{60}{3323}$/$\frac{1443}{260.969}$#Educational level (years)12.103.4511.273.9012.053.470.762BMI (kg/m2)24.702.9323.433.1223.052.760.003*Data are shown as mean ± standard deviation (SD) and were analyzed using independent sample t-tests. BMI body mass index.#Pearson’s Chi-square test (2-sided).*Statistically significant different ($p \leq 0.05$).Table 2Neuropsychological results of T2DM-MCI, T2DM-NCI and NC groups. T2DM-NCIT2DM-MCINCpMeanSDMeanSDMeanSDAVLT (immediate)23.714.9220.415.5123.845.150.008*AVLT (immediate)23.714.9220.415.5123.845.150.008*AVLT (5 min)9.462.277.502.6710.094.110.000*AVLT (delay)9.142.466.972.969.282.230.000*AVLT (recall)11.072.108.213.9711.031.880.000*TMT-A47.0523.2559.6819.5951.2521.750.000*TMT-B40.1817.0148.7615.8743.2316.720.009*DST (direct)7.851.477.321.6110.089.160.102DST (reverse)4.951.263.891.595.101.640.000*MoCA-B27.511.5524.050.8527.521.540.000*MMSE28.491.5427.831.8428.321.690.170DSST49.7512.8935.0312.7748.2414.830.000*GPT (R)74.1713.5390.7027.5368.4914.170.000*GPT (L)81.6015.2593.8130.8980.1018.220.023*Data are shown as mean ± standard deviation (SD) and were analyzed using independent sample t-tests. AVLT California-Los Angeles auditory verbal learning test, TMT trail-making test, DST digit span test, MoCA montreal cognitive assessment, MMSE mini-mental state examination, DSST digit symbol substitution, GPT grooved pegboard test, L left, R right.*Data was considered significantly different ($p \leq 0.05$).Table 3Clinical results of T2DM-MCI, T2DM-NCI and NC groups. T2DM-NCIT2DM-MCINCpMeanSDMeanSDMeanSDHbAlc (%)9.252.568.752.15NANA0.522DBP (mmHg)126.4216.62127.7919.96123.4819.040.273SBP (mmHg)82.7414.0082.2510.7282.4011.120.783FBG (mmol/L)8.932.967.952.684.500.760.000*FSI (uIU/mL)11.259.7914.0816.05NANA0.507TG (mmol/L)2.672.583.253.83NANA0.775TC (mmol/L)4.761.034.991.06NANA0.694LDL (mmol/L)3.020.993.150.88NANA0.785ACR (mg/g)30.0096.6022.1851.45NANA0.068mALB (mg/L)29.1580.9424.7950.62NANA0.22124 h UPRO (G/24 h)0.190.230.170.14NANA0.436M-TP (mg/L)101.7168.93102.1879.54NANA0.995C-Peptide (ng/mL)2.271.222.311.26NANA0.674Data are shown as mean ± standard deviation, independent sample t-tests.*Data was considered significant different ($p \leq 0.05$).HbA1c hemoglobinA1c, DBP diastolic blood pressure, SBP systolic blood pressure, FBG fasting blood glucose, FSI fasting serum insulin, TG triglyceride, TC total cholesterol, LDL low-density lipoprotein, ACR albumin/creatinine ratio, mALB microalbuminuria, 24 h UPRO 24-h urinary protein, M-TP micro total protein.
## Classification performance
A summary of classification performance using XGBoost is shown in Table 4 in terms of accuracy (ACC), the area under the curve (AUC), sensitivity (SEN), specificity (SPE), precision (PRE), and F127. We constructed the XGBoost model to classify all three groups and pairwise classifications between two groups. The XGBoost model did not perform well in the three classification categories (ACC = $69.39\%$, AUC = $80.07\%$, SEN = $69.39\%$, SPE = $78.14\%$, PRE = $70.90\%$, and F1 = $68.11\%$); however, the model achieved peak performance in discriminating between the two classification categories (T2DM-NCI versus T2DM-MCI: ACC = $87.91\%$, AUC = $81.99\%$, SEN = $61.67\%$, SPE = $98.06\%$, PRE = $93.06\%$, and F1 = $73.95\%$; T2DM-NCI versus NC: ACC = $80.00\%$, AUC = $84.14\%$, SEN = $75.65\%$, SPE = $83.23\%$, PRE = $77.58\%$, and F1 = $76.24\%$).Table 4Classification performance in T2DM-MCI, T2DM-NC and NC differentiation. ACC (%)AUC (%)F1 (%)SEN (%)SPE (%)PRE (%)T2DM-NC vs. T2DM-MCI87.91 ± 1.0481.99 ± 8.3073.95 ± 2.5461.67 ± 4.5698.06 ± 1.7793.06 ± 6.36T2DM-NC vs. NC80.00 ± 0.8384.14 ± 3.4476.24 ± 1.4875.65 ± 7.2883.23 ± 6.2077.58 ± 4.86T2DM-NC vs. T2DM-MCI and NC69.39 ± 1.2780.07 ± 0.7668.11 ± 0.8769.39 ± 1.2778.14 ± 1.3370.90 ± 1.54ACC accuracy, AUC the area under the receiver operating characteristic curve, SEN sensitivity, SPE specificity, PRE precision.
## Connections contributive to classification
Out of 4,005 connections, 511 connections provided useful features for classification (analysis of variance (ANOVA) with Bonferroni correction, $p \leq 0.05$). After using the XGBoost model for classification, we traced the data to further explore and analyze the 511 functional connections. Many functional connections were connected through the same brain region; precisely, we observed the aggregation of connection features. We found that the following eight areas are most discriminative: left caudate nucleus (CAU.L, $14.68\%$), right caudate nucleus (CAU.R, $12.13\%$), left angular gyrus (ANG.L, $10.76\%$), left thalamus (THA.L, $7.83\%$), right paracentral lobule (PCL.R, $7.63\%$), right thalamus (THA.R, $7.44\%$), right angular gyrus (ANG.R, $6.65\%$), and left paracentral lobule (PCL.L, $4.7\%$). There were more than $70\%$ of connections to CAU ($26.81\%$), ANG ($17.42\%$), THA ($15.26\%$), and PCL ($12.33\%$). The FC of these regions contribute most to T2DM-MCI classification ($71.82\%$ total, see Table 5 and Fig. 1).Table 5Brain regions most contributive to classification. Brain regionTotal ratio (%)SubregionRatio (%)CAU26.81CAU.L14.68CAU.R12.13ANG17.42ANG.L10.76ANG.R6.65THA15.26THA.L7.83THA.R7.44PCL12.33PCL.L4.70PCL.R7.63CAU caudate nucleus, ANG angular gyrus, THA thalamus gyrus, PCL paracentral lobule, L left, R right. Figure 1Significant connections rendered on the surface of the automated anatomical labeling atlas in BrainNet viewer. THA thalamus, ANG angular gyrus, CAU caudate nucleus, PCL paracentral lobule, L left, R right.
## Association between classification features and clinical variables
To better understand the relationship between the characteristics of the clinical development of T2DM-MCI, we further analyzed the correlation between imaging data and clinical variables (Bonferroni correction). The correlations between significant cognitive function scores and different brain regions were analyzed using Pearson’s correlation for three groups (Fig. 2). ANG.L was positively correlated with TMT-B ($r = 0.224$, $$p \leq 0.043$$), ANG.R is negatively correlated with BMI (r = –0.215, $$p \leq 0.042$$). PCL.R was positively correlated with AVLT (5 min) ($r = 0.267$, $$p \leq 0.015$$) and AVLT (delay) ($r = 0.233$, $$p \leq 0.037$$). THA.L was positively correlated with educational level ($r = 0.236$, $$p \leq 0.027$$). MoCA was positively correlated with DSST ($r = 0.392$, $$p \leq 0.000$$) and educational level ($r = 0.204$, $$p \leq 0.007$$). There was no significant correlation between other variables and the brain regions. Figure 2Associations between neuropsychological test scores and functional connectivity. Partial correlation was used to determine the relationship between neuropsychological test scores and functional connectivity. ( a) Correlation of functional connectivity with neuropsychological test scores (b) Correlation between neuropsychological test scores. TMT trail-making test, BMI body mass index, AVLT World Health Organization University of California-Los Angeles auditory verbal learning test, MoCA-B montreal cognitive assessment-B, DSST digit symbol substitution test, L left, R right.
## Discussion
T2DM-MCI has a relatively low clinical diagnostic rate owing to its subtle onset and lack of clinical diagnostic approach. This study examined whether FC has discriminative features for accurately detecting T2DM-MCI using XGBoost patterns. The XGBoost algorithm is an accurate and efficient classification algorithm used in data mining with good performance. XGBoost has been applied for early diagnoses of diseases such as tuberculosis, epilepsy, kidney disease, and breast cancer28–32. Notably, this study is the first to apply whole-brain FC for detecting T2DM-MCI using the XGBoost model. Our model yields better classification performance ($87.91\%$ accuracy) than that of previous studies21,22. Using only 23 patients with T2DM and CI, Chen et al.21 used high-order FC for differentiating healthy controls from patients with T2DM and CI ($79.17\%$ accuracy) and patients with T2DM without CI ($59.62\%$ accuracy). With only 16 T2DM-MCI, Shi et al. employed large-scale FC to predict MoCA scores with a connectome-based predictive model and support vector machine, achieving AUC values (T2DM-NCI vs. T2DM-MCI) of 0.65‒0.70, which was significantly lower than that obtained by our method (0.82 in AUC). Moreover, our sample size was larger than those of previous studies21,22,33, including 199 participants in total.
T2DM is typically related to an increased risk of CI and dementia. Patients with T2DM may experience memory, language, attention, concentration, reaction, and executive function decline1,34. Nevertheless, researchers are still unsure of the exact pathophysiology underlying T2DM-related cognitive dysfunction, delaying the development of preventive treatments. We further found that the FC of THA, ANG, CAU, and PCL was highly discriminative in distinguishing T2DM-MCI, T2DM-NCI, and NC.
THA is a relay station or hub transmitting information between subcortical, cortical, and cerebellar areas35. THA declines with normal aging36. There may be no obvious structural damage, however, it develops thalamocortical FC impairment in patients with T2DM21. In our preliminary study37, patients with T2DM without CI already had abnormalities in the dynamic FC of THA, as revealed by a significant decrease in connectivity between the right executive control network and THA.L. Abnormal thalamic connectivity is associated with CI. Thalamic connectivity is likely to be impaired in patients with T2DM and CI, which is consistent with our results21,38,39. When undergoing external working memory tasks, the corresponding working memory brain regions are activated, and the right hippocampal/parahippocampal gyrus and THA are abnormally activated predominantly in the right cerebral hemisphere40. This indicates that THA is involved in processing working memory, and FC is already impaired before the onset of CI in patients with T2DM. ANG is associated with complex language functions and linked to other cognitive domains such as representational and semantic memory41. Patients with T2DM exhibit significantly thinner ANG cortical thickness42, reduced cerebral blood flow43, and less spontaneous neuronal activities44. Moreover, abnormal FC in THA and ANG because of diabetes causes various cognitive dysfunctions, including AD/VaD38,45,46. Compared with NC, bilateral ANG in patients with T2DM exhibit abnormal FC with multiple brain regions, and the FC of ANG with multiple brain regions positively correlated with MoCA, suggesting that the broad functional disconnectedness of ANG may play an essential role in the neuropathology of patients with T2DM-MCI45.
CAU is associated with memory and learning abilities47 as well as executive and cognitive processes48. CAU and the cerebellum function as a network that controls behavior49. FC between the CAU and hippocampus, which is an important anatomic basis for learning and memory, is implicated in altered white matter structure in patients with T2DM50. CAU has extensive connections to cortical and subcortical structures that serve complex regulation of motor function, cognition, and emotion51. In patients with T2DM, the grey matter volume of CAU is significantly reduced42, and the microstructure is abnormal50. The characteristics of the abnormalities are significantly associated with the duration of T2DM. In addition, the activation of the left CAU, hippocampus, and parahippocampal gyrus is weaker in T2DM-MCI than in normal controls under memory task stimulation40. Consistent with the previous studies mentioned above, the abnormality of FC of the CAU in our findings indicates impaired cognitive functions in patients with T2DM. In addition, PCL is associated with motor and sensory innervation of the contralateral lower extremities as well as the regulation of physiological functions. However, FC in PCL is affected in cognition-related diseases, such as vascular cognitive impairment. Sun et al. found that the most obvious regions showing connectivity deficits were between several regions, including PCL, and CAU.R52. They also showed impaired connectivity in the default mode network, and PCL with CAU.R. PCL was also discriminative as a region of interest (ROI) feature in the T2DM classification33. Furthermore, during the analysis of the internal connectivity of the left executive control network, ANG.L and PCL.L had significantly decreased connectivity with other brain regions in the network. In the external network connectivity analysis, significant differences were found between the left executive control network and ANG.R/PCL.L. In addition, significant differences were observed between the right executive control network and PCL.R/ANG.R. Furthermore, significant differences were found between the precuneus network and CAU.R/ANG37. In summary, THA, ANG, CAU, and PCL are highly sensitive to T2DM. They play essential roles in the early diagnosis of T2DM-MCI.
Educational level, age, BMI, blood pressure, and blood glucose levels are key factors influencing MCI in patients with T2DM. Correlation analysis showed that THA.L and MoCA were positively correlated with educational level, suggesting that highly-educated people have a lower risk of developing MCI53,54. Lower FC strength in ANG. R was associated with higher BMI. We also found that higher cognition scores were positively correlated with higher FC in PCL.R and ANG.L. This corresponds with previous findings21,55 that people with higher FC in PCL.R and ANG.L have a smaller risk of developing MCI.
Our study has some advantages. First, this study is the first to apply the XGBoost model to classify T2DM-MCI and achieve a good classification performance. Second, our analysis was based on whole-brain FC, unlike previous studies that were based on brain regions or predefined networks8,21,56,57. Third, we found that THA, ANG, CAU, and PCL demonstrated significant discriminative power in T2DM-MCI detection. However, our study has some limitations. First, the overall T2DM study sample was below 200; the number of T2DM-MCI was small. Therefore, multicenter data collection should be considered to expand the sample size in future studies. Second, this classification study extracted different characteristic connections based on all participants and applied the features to training classification, which has the problem of cross-validation and is slightly limited in the subsequent application. Subsequent studies can consider separating the training and test sets and conducting feature extraction so that the data results can be more objective and random. Third, our research is only a cross-sectional study. We believe that combining follow-up and longitudinal studies will better explain the mechanism of accelerated neurodegenerative changes in T2DM-MCI.
## Conclusion
This study proposes a novel framework to pool the connectivity features extracted from whole brain FC for detecting T2DM-MCI. The current study is the first attempt to use the XGBoost model to detect T2DM-MCI, which significantly enhances the prediction accuracy of the model. We show that the FC within THA, ANG, CAU, and PCL provides major information for detecting T2DM-MCI. Our results affirm that FC contains clinically relevant cognition-related information. Therefore, it is a potential biomarker for assessing the degree of cognitive decline. Overall, our findings provide valuable knowledge for classifying and predicting T2DM-related CI. These results have clinical implications in patients with T2DM. It can help in early clinical diagnosis and provides a basis for future studies.
## Methods
Two hundred and ten individuals were willing to join this study (May 10, 2021, to July 1, 2022). The exclusion criteria for the two groups were as follows: type 1 diabetes mellitus, impaired fasting glucose or impaired glucose tolerance58, hypertension, hypoglycemia (blood sugar levels < 3.9 mmol/L), hyperlipidemia, serious eye diseases (e.g., blindness), symptoms of neurological conditions (e.g., cerebral infarction or hemorrhage), history of neurological abnormality (e.g., Parkinson’s disease), severe head injuries or chronic head discomfort (e.g., migraine), BMI > 31 kg/m2, left- or mixed-handedness, substance (tobacco, alcohol, or psychoactive drug) abuse, taking medications that may affect cognition and memory within 6 months, specific abnormalities detected on conventional MRI scans or any other factors that may influence brain structure or function (e.g., extreme physical weakness, chronic infections, and other endocrine diseases). Patients with T2DM were diagnosed by two experienced endocrinologists following international clinical standards59. MCI was evaluated via Mini-Mental State Examination (MMSE) and MoCA-B (21 ≤ MoCA-B score < 26, and MMSE score > 24 were diagnosed with MCI)60,61.
Participants with brain tumors ($$n = 3$$), neuropsychiatric diseases ($$n = 4$$) (e.g., major depression or schizophrenia), or developmental disorders ($$n = 4$$) were excluded. Finally, 37 patients with T2DM-MCI, 93 patients with T2DM-NCI, and 69 NC were enrolled in this study. The source of patients with T2DM and NC corresponded with our previous study37. This study was approved by the ethics committee of The First Affiliated Hospital of Guangzhou University of Chinese Medicine (ID: NO. JY [2020] 288). Written informed consent was obtained from all participants. In addition, the study was conducted following approved guidelines.
## Demographic, clinical, and neuropsychological assessments
Demographic assessments include age, sex, educational level, past medical history, height, weight and medication history. Clinical assessments include HemoglobinA1c (HbA1c), C-Peptide, systolic blood pressure (SBP), diastolic blood pressure (DBP), total cholesterol (TC), triglyceride (TG), fasting serum insulin (FSI), low-density lipoprotein (LDL), fasting blood glucose (FBG), microalbuminuria (mALB), albumin/creatinine ratio (ACR), micro total protein (M-TP), 24-h urinary protein (24 h UPRO). Neuropsychological assessments include MoCA-B, MMSE, digit span test (DST), AVLT, TMT, GPT, and DSST which can be used to assess cognitive ability.
## MRI data acquisition
A Siemens (Munich, Germany) 3 T Prisma scanner with a standard 64-channel head coil was used to acquire fMRI imaging. All participants were placed in the supine position and tried their best to keep heads as still as possible while acquiring images. The detailed parameters of the multi-slice T2-weighted echo-planar imaging (EPI) sequence were as follows: TR = 2000 ms; TE = 30 ms; FOV = 100 mm; flip angle = 90°; matrix dimensions = 64 × 64; slice thickness = 3.5 mm; and number of slices = 33. Three-dimensional T1-weighted images were acquired with the following parameters: TR = 2,530 ms; TE = 2.98 ms; FOV = 256 × 224 mm2; inversion time = 1,100 ms; flip angle = 7°; matrix size = 224 × 256;, sagittal slices = 192; slice thickness = 1.0 mm; and voxel size = 0.5 × 0.5 × 1 mm3. The BOLD-fMRI gradient EPI sequence acquisition parameters were as follows: TR = 500 ms; TE = 30 ms; matrix dimensions = 64 × 64; FOV = 244 mm × 244 mm; slices thickness = 3.5 mm; voxel size = 3.5 mm × 3.5 mm; number of slices = 960, and scan time = 8 min.
## MRI image pre-processing
For fMRI data, the pre-processing was performed using SPM12 (Wellcome Department of Imaging Neurosciences, University College London, UK, http://www.fil.ion.ucl.ac.uk/spm), and the statistical analyses of imaging data were performed using GRETNA (GRETNA v2.0) in Matlab R2021b. First, the first 10-time point-scanned images were removed owing to the instability of the magnetic field at the beginning of the scan. Second, all functional images were realigned to the first image to correct head movement. All participants met the criteria of < 2 mm translation and < 2° rotation in any direction. Otherwise, their data were excluded. Third, the functional images were normalized to the MNI space using DARTEL and resampled to a 3 × 3 × 3 mm3 voxel size62. Fourth, we used an anisotropic 6-mm full-width half-maximum Gaussian kernel63 for spatial smoothing of the obtained images. Fifth, we detrended and removed linear trends. Sixth, we removed covariates, excluding white matter, grey matter, and cerebrospinal fluid influences. Seventh, 0.01‒0.08 Hz bandpass filtering was used to remove high and low-frequency signals. Eighth, we removed the FD_Threshold > 0.5 mm time points by “scrubbing” 1-time point before and 2-time points after. In summary, the pre-processing procedures included slice timing correction, realignment, normalization, smoothing, detrending, filtering, and scrubbing.
## Statistical analyses
All statistical analyses were performed using the SPSS software package (version 26.0). The measurement data of each group were described by mean ± standard deviation. The demographic, clinical, and neuropsychological assessment scores of the three groups were compared using multiple independent sample ANOVA64. Categorical data were evaluated using Chi-square analysis. Paired-sample t-tests were used for pre-and post-treatment intragroup comparisons. In addition, a partial test was used to examine the relationship between imaging indices, cognitive tests, and clinical data. $p \leq 0.05$ was used as the statistical significance level.
## Feature abstraction based on FC network
The pre-processed fMRI BOLD data has dimensions of 950 × 90, where 950 denotes the number of time points in each fMRI scan, and 90 means the number of ROI derived from the Automated Anatomical Labelling atlas. We calculated the mean BOLD signals for each brain ROI by averaging the time series over all voxels within the ROI. Subsequently, based on the pre-processed fMRI data, we used the Pearson correlation coefficient65 to build an FC network for each participant in a matrix size of 90 × 90. Every node represented a brain ROI, and every edge measured the linear correlation between any pair of ROI.
Subsequently, we flattened the upper triangle elements of FC, thereby deriving a 4,005-dimensional [(90 × 90–90)/2] vector for each participant. However, these features may be redundant for classifying participants into the three experimental groups. Therefore, we applied ANOVA and Bonferroni correction analysis to extract features showing significant differences ($p \leq 0.05$) among the three groups. Finally, we generated 511 discriminative features for further classification.
## Illustration of XGBoost model
The XGBoost model23 is an ensemble machine learning algorithm based on decision trees that used the gradient boosting framework with promising performance in fMRI-related classification tasks29,66,67. The XGBoost model is illustrated in Fig. 3. The XGBoost model was built based on gradient boosting machines which used Gradient Boosting Trees68 as the error predictor. In gradient boosting, we trained a predictor to predict the errors made by the original model and constructed an improved model whose output was fine-tuned based on the original prediction. The improved model is an ensemble of two predictors, i.e., the original and error predictors. We repeated this process until we achieved satisfactory prediction results. Figure 3The illustration of XGBoost model for image classification. A predictor is trained to predict the errors made by the original model, and then construct an improved model whose output is fine-tuned based on the original prediction.
We used the XGBoost model to classify the three experimental groups, and the parameters were defined as follows: the number of gradient-boosted trees was 280; the maximum tree depth of base learners was 2; the minimum sum of sample weight required in a child was 5; the minimum loss reduction needed to make a further partition on a leaf node was 0; the subsample ratio of the training sample was 0.8; the subsample ratio for each tree’s construction was 0.8; the boosting learning rate was 0.1; and the L1 regularization constraint was 0.01. The XGBoost model was implemented based on the XGBoost package in Python (Supplementary Information S1).
## Experimental setting
We randomly partitioned all participants from the three groups into training and testing sets, following a 2:1 ratio, and this procedure was repeated five times. In the data division, we ensured that the three classes in both sets were equally distributed to prevent data imbalance. Five measurement metrics were adopted for model evaluation, including the AUC, ACC, F1, SEN, SPE, and PRE69.
## Supplementary Information
Supplementary Information. The online version contains supplementary material available at 10.1038/s41598-023-28163-5.
## References
1. Cukierman-Yaffe T. **Effect of dulaglutide on cognitive impairment in type 2 diabetes: An exploratory analysis of the REWIND trial**. *Lancet Neurol.* (2020.0) **19** 582-590. DOI: 10.1016/S1474-4422(20)30173-3
2. 2.Srikanth, V., Sinclair, A. J., Hill-Briggs, F., Moran, C. & Biessels, G. J. Type 2 diabetes and cognitive dysfunction—towards effective management of both comorbidities. Lancet Diabetes Endocrinol.8, 535–545 (2020).
3. Biessels GJ, Nobili F, Teunissen CE, Simó R, Scheltens P. **Understanding multifactorial brain changes in type 2 diabetes: A biomarker perspective**. *Lancet Neurol.* (2020.0) **19** 699-710. DOI: 10.1016/S1474-4422(20)30139-3
4. Li W, Wang T, Xiao S. **Type 2 diabetes mellitus might be a risk factor for mild cognitive impairment progressing to Alzheimer’s disease**. *Neuropsychiatr. Dis. Treat.* (2016.0) **12** 2489-2495. DOI: 10.2147/NDT.S111298
5. Qi D. **Default mode network connectivity and related white matter disruption in type 2 diabetes mellitus patients concurrent with amnestic mild cognitive impairment**. *Curr. Alzheimer Res.* (2017.0) **14** 1238-1246. DOI: 10.2174/1567205014666170417113441
6. Xue J, Li J, Liang J, Chen S. **The prevalence of mild cognitive impairment in China: A systematic review**. *Aging Dis.* (2018.0) **9** 706-715. DOI: 10.14336/AD.2017.0928
7. You Y. **The prevalence of mild cognitive impairment in type 2 diabetes mellitus patients: A systematic review and meta-analysis**. *Acta Diabetol.* (2021.0) **58** 671-685. DOI: 10.1007/s00592-020-01648-9
8. Chen X. **Extraction of dynamic functional connectivity from brain grey matter and white matter for MCI classification**. *Hum. Brain Mapp.* (2017.0) **38** 5019-5034. DOI: 10.1002/hbm.23711
9. Lu M, Guo Z, Gao Z, Cao Y, Fu J. **Multiscale brain network models and their applications in neuropsychiatric diseases**. *Electronics* (2022.0) **11** 3468. DOI: 10.3390/electronics11213468
10. Li C. **Type 2 diabetes mellitus may exacerbate gray matter atrophy in patients with early-onset mild cognitive impairment**. *Front. Neurosci.* (2020.0) **14** 856. DOI: 10.3389/fnins.2020.00856
11. Xiong Y. **Altered regional homogeneity and functional brain networks in Type 2 diabetes with and without mild cognitive impairment**. *Sci. Rep.* (2020.0) **10** 1-11. DOI: 10.1038/s41598-020-76495-3
12. Yu Y. **Neurovascular decoupling in type 2 diabetes mellitus without mild cognitive impairment: Potential biomarker for early cognitive impairment**. *Neuroimage* (2019.0) **200** 644-658. DOI: 10.1016/j.neuroimage.2019.06.058
13. Li X. **The role of the Montreal Cognitive Assessment (MoCA) and its memory tasks for detecting mild cognitive impairment**. *Neurol. Sci.* (2018.0) **39** 1029-1034. DOI: 10.1007/s10072-018-3319-0
14. Paschoal AM. **Editorial for “diffusion tensor imaging reveals altered topological efficiency of structural networks in type-2 diabetes patients with and without mild cognitive impairment”**. *J. Magn. Reson. Imaging* (2022.0) **55** 928-929. DOI: 10.1002/jmri.27899
15. Sun Q. **Alterations of white matter integrity and hippocampal functional connectivity in type 2 diabetes without mild cognitive impairment**. *Front. Neuroanat.* (2018.0) **12** 21. DOI: 10.3389/fnana.2018.00021
16. Xiong Y. **Application of neurite orientation dispersion and density imaging to characterize brain microstructural abnormalities in type-2 diabetics with mild cognitive impairment**. *J. Magn. Reson. Imaging* (2019.0) **50** 889-898. DOI: 10.1002/jmri.26687
17. Ciudin A. **Retinal microperimetry: A new tool for identifying patients with type 2 diabetes at risk for developing Alzheimer disease**. *Diabetes* (2017.0) **66** 3098-3104. DOI: 10.2337/db17-0382
18. Chen X. **High-order resting-state functional connectivity network for MCI classification**. *Hum. Brain Mapp.* (2016.0) **37** 3282-3296. DOI: 10.1002/hbm.23240
19. Liang P. **Disruption of cortical integration during midazolam-induced light sedation**. *Hum. Brain Mapp.* (2015.0) **36** 4247-4261. DOI: 10.1002/hbm.22914
20. Qiao L. **Estimating functional brain networks by incorporating a modularity prior**. *Neuroimage* (2016.0) **141** 399-407. DOI: 10.1016/j.neuroimage.2016.07.058
21. Chen Y. **Classification of type 2 diabetes mellitus with or without cognitive impairment from healthy controls using high-order functional connectivity**. *Hum. Brain Mapp.* (2021.0) **42** 4671-4684. DOI: 10.1002/hbm.25575
22. Shi A-P. **Large-scale functional connectivity predicts cognitive impairment related to type 2 diabetes mellitus**. *World J. Diabetes* (2022.0) **13** 110-125. DOI: 10.4239/wjd.v13.i2.110
23. 23.Chen, T. & Guestrin, C. XGBoost: A scalable tree boosting system. in Proceedings of the ACM SIGKDD International Conference on Knowledge Discovery and Data Mining vols 13–17-Augu 785–794 (ACM, 2016).
24. 24.Mackey, L., Bryan, J. & Mo, M. Y. Weighted Classification Cascades for Optimizing Discovery Significance in the HiggsML Challenge. in NIPS 2014 Workshop on High-energy Physics and Machine Learning 129–134 (PMLR, 2014).
25. Devan P, Khare N. **An efficient XGBoost–DNN-based classification model for network intrusion detection system**. *Neural Comput. Appl.* (2020.0) **32** 12499-12514. DOI: 10.1007/s00521-020-04708-x
26. Ramaneswaran S, Srinivasan K, Vincent PMDR, Chang CY. **Hybrid inception v3 XGBoost model for acute lymphoblastic leukemia classification**. *Comput. Math. Methods Med.* (2021.0) **2021** 1-10. DOI: 10.1155/2021/2577375
27. 27.Sokolova, M., Japkowicz, N. & Szpakowicz, S. Beyond accuracy, F-score and ROC: A family of discriminant measures for performance evaluation. in AAAI Workshop - Technical Report vol. WS-06–06 24–29 (Springer, 2006).
28. 28.Ren, X., Guo, H., Li, S., Wang, S. & Li, J. A novel image classification method with CNN-XGBoost model. in Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) vol. 10431 LNCS 378–390 (Springer, 2017).
29. Torlay L, Perrone-Bertolotti M, Thomas E, Baciu M. **Machine learning–XGBoost analysis of language networks to classify patients with epilepsy**. *Brain Inf.* (2017.0) **4** 159-169. DOI: 10.1007/s40708-017-0065-7
30. Ogunleye A, Wang QG. **XGBoost model for chronic kidney disease diagnosis**. *IEEE/ACM Trans. Comput. Biol. Bioinforma.* (2020.0) **17** 2131-2140. DOI: 10.1109/TCBB.2019.2911071
31. Liew XY, Hameed N, Clos J. **An investigation of XGBoost-based algorithm for breast cancer classification**. *Mach. Learn. with Appl.* (2021.0) **6** 100154. DOI: 10.1016/j.mlwa.2021.100154
32. Rahman M, Cao Y, Sun X, Li B, Hao Y. **Deep pre-trained networks as a feature extractor with XGBoost to detect tuberculosis from chest X-ray**. *Comput. Electr. Eng.* (2021.0) **93** 107252. DOI: 10.1016/j.compeleceng.2021.107252
33. Liu Z. **Identification of cognitive dysfunction in patients with T2DM using whole brain functional connectivity**. *Genom. Proteom. Bioinf.* (2019.0) **17** 441-452. DOI: 10.1016/j.gpb.2019.09.002
34. González HM. **Diabetes, cognitive decline, and mild cognitive impairment among diverse Hispanics/Latinos: Study of Latinos-investigation of neurocognitive aging results (HCHS/SOL)**. *Diabetes Care* (2020.0) **43** 1111-1117. DOI: 10.2337/dc19-1676
35. Kim HS, Sasaki JY. **Cultural neuroscience: Biology of the mind in cultural contexts**. *Annu. Rev. Psychol.* (2014.0) **65** 487-514. DOI: 10.1146/annurev-psych-010213-115040
36. Fama R, Sullivan EV. **Thalamic structures and associated cognitive functions: Relations with age and aging**. *Neurosci. Biobehav. Rev.* (2015.0) **54** 29-37. DOI: 10.1016/j.neubiorev.2015.03.008
37. Wu J. **Altered functional network connectivity of precuneus and executive control networks in type 2 diabetes mellitus without cognitive impairment**. *Front. Neurosci.* (2022.0) **16** 887713. DOI: 10.3389/fnins.2022.887713
38. Chen Y-C. **Thalamic resting-state functional connectivity: disruption in patients with type 2 diabetes**. *Metab. Brain Dis.* (2015.0) **30** 1227-1236. DOI: 10.1007/s11011-015-9700-2
39. Tan X. **Micro-structural white matter abnormalities in type 2 diabetic patients: A DTI study using TBSS analysis**. *Neuroradiology* (2016.0) **58** 1209-1216. DOI: 10.1007/s00234-016-1752-4
40. Huang RR. **Spatial working memory impairment in primary onset middle-age type 2 diabetes mellitus: An ethology and BOLD-fMRI study**. *J. Magn. Reson. Imaging* (2016.0) **43** 75-87. DOI: 10.1002/jmri.24967
41. 41.Humphreys, G. F., Lambon Ralph, M. A. & Simons, J. S. A unifying account of angular gyrus contributions to episodic and semantic cognition. Trends Neurosci.44, 452–463 (2021).
42. Peng B, Chen Z, Ma L, Dai Y. **Cerebral alterations of type 2 diabetes mellitus on MRI: A pilot study**. *Neurosci. Lett.* (2015.0) **606** 100-105. DOI: 10.1016/j.neulet.2015.08.030
43. Bangen KJ. **Reduced regional cerebral blood flow relates to poorer cognition in older adults with type 2 diabetes**. *Front. Aging Neurosci.* (2018.0) **10** 270. DOI: 10.3389/fnagi.2018.00270
44. Wang Z-L. **Abnormal spontaneous brain activity in type 2 diabetic retinopathy revealed by amplitude of low-frequency fluctuations: A resting-state fMRI study**. *Clin. Radiol.* (2017.0) **72** 340-e1. DOI: 10.1016/j.crad.2016.11.012
45. Qi F. **Functional disconnection of the angular gyrus related to cognitive impairment in patients with type 2 diabetes mellitus**. *Front. Hum. Neurosci.* (2021.0) **15** 25. DOI: 10.3389/fnhum.2021.621080
46. Zhang Y. **Sensorimotor and pain-related alterations of the gray matter and white matter in Type 2 diabetic patients with peripheral neuropathy**. *Hum. Brain Mapp.* (2020.0) **41** 710-725. DOI: 10.1002/hbm.24834
47. 47.Driscoll, M. E., Bollu, P. C. & Tadi, P. Neuroanatomy, nucleus caudate. in StatPearls [Internet] (StatPearls Publishing, 2021).
48. Habas C. **Distinct cerebellar contributions to intrinsic connectivity networks**. *J. Neurosci.* (2009.0) **29** 8586-8594. DOI: 10.1523/JNEUROSCI.1868-09.2009
49. Graff-Radford J, Williams L, Jones DT, Benarroch EE. **Caudate nucleus as a component of networks controlling behavior**. *Neurology* (2017.0) **89** 2192-2197. DOI: 10.1212/WNL.0000000000004680
50. Hsu J-L. **Microstructural white matter abnormalities in type 2 diabetes mellitus: A diffusion tensor imaging study**. *Neuroimage* (2012.0) **59** 1098-1105. DOI: 10.1016/j.neuroimage.2011.09.041
51. Ghaziri J. **Subcortical structural connectivity of insular subregions**. *Sci. Rep.* (2018.0) **8** 8596. DOI: 10.1038/s41598-018-26995-0
52. Sun Y. **Abnormal functional connectivity in patients with vascular cognitive impairment, no dementia: A resting-state functional magnetic resonance imaging study**. *Behav. Brain Res.* (2011.0) **223** 388-394. DOI: 10.1016/j.bbr.2011.05.006
53. Cesar KG, Yassuda MS, Porto FHG, Brucki SMD, Nitrini R. **MoCA Test: normative and diagnostic accuracy data for seniors with heterogeneous educational levels in Brazil**. *Arq. Neuropsiquiatr.* (2019.0) **77** 775-781. DOI: 10.1590/0004-282x20190130
54. Freitas S, Simões MR, Marôco J, Alves L, Santana I. **Construct validity of the montreal cognitive assessment (MoCA)**. *J. Int. Neuropsychol. Soc.* (2012.0) **18** 242-250. DOI: 10.1017/S1355617711001573
55. Li C. **Alterations of brain structural network connectivity in type 2 diabetes mellitus patients with mild cognitive impairment**. *Front. Aging Neurosci.* (2021.0) **12** 615048. DOI: 10.3389/fnagi.2020.615048
56. 56.Lekha, S. & M., S. Real-time non-invasive detection and classification of diabetes using modified convolution neural network. IEEE J. Biomed. Heal. Inf.22, 1630–1636 (2018).
57. 57.Ndaba, M., Pillay, A. W. & Ezugwu, A. E. An improved generalized regression neural network for type II diabetes classification. in Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) vol. 10963 LNCS 659–671 (Springer, 2018).
58. Ogurtsova K. **IDF Diabetes Atlas: Global estimates for the prevalence of diabetes for 2015 and 2040**. *Diabetes Res. Clin. Pract.* (2017.0) **128** 40-50. DOI: 10.1016/j.diabres.2017.03.024
59. 59.Dobrescu, R. Oxford Textbook of Endocrinology and Diabetes, 3rd Edition. Acta Endocrinologica (Bucharest) vol. 17 (Oxford University Press, 2021).
60. 60.Li, C. et al. Cortical thickness contributes to cognitive heterogeneity in patients with type 2 diabetes mellitus. Medicine (Baltimore).97, e10858 (2018).
61. Schrag A, Siddiqui UF, Anastasiou Z, Weintraub D, Schott JM. **Clinical variables and biomarkers in prediction of cognitive impairment in patients with newly diagnosed Parkinson’s disease: A cohort study**. *Lancet Neurol.* (2017.0) **16** 66-75. DOI: 10.1016/S1474-4422(16)30328-3
62. Berman BD. **Levodopa modulates small-world architecture of functional brain networks in Parkinson’s disease**. *Mov. Disord.* (2016.0) **31** 1676-1684. DOI: 10.1002/mds.26713
63. 63.Thomas Yeo, B. T. et al. The organization of the human cerebral cortex estimated by intrinsic functional connectivity. J. Neurophysiol.106, 1125–1165 (2011).
64. 64.Girden, E. R. ANOVA: Repeated measures. (Sage Publications, Inc., 1992).
65. 65.Benesty, J., Chen, J., Huang, Y. & Cohen, I. Pearson correlation coefficient. in Springer Topics in Signal Processing vol. 2 1–4 (Springer, 2009).
66. Nguyen RD. **A comparison of machine learning classifiers for pediatric epilepsy using resting-state functional MRI latency data**. *Biomed. Rep.* (2021.0) **15** 1-9. DOI: 10.3892/br.2021.1453
67. Xing X. **Modeling essential connections in obsessive–compulsive disorder patients using functional MRI**. *Brain Behav.* (2020.0) **10** e01499. DOI: 10.1002/brb3.1499
68. Friedman JH. **Greedy function approximation: A gradient boosting machine**. *Ann. Stat.* (2001.0) **29** 1189-1232. DOI: 10.1214/aos/1013203451
69. Zou KH, O’Malley AJ, Mauri L. **Receiver-operating characteristic analysis for evaluating diagnostic tests and predictive models**. *Circulation* (2007.0) **115** 654-657. DOI: 10.1161/CIRCULATIONAHA.105.594929
|
---
title: Cre-dependent ACR2-expressing reporter mouse strain for efficient long-lasting
inhibition of neuronal activity
authors:
- Yasutaka Mukai
- Yan Li
- Akiyo Nakamura
- Noriaki Fukatsu
- Daisuke Iijima
- Manabu Abe
- Kenji Sakimura
- Keiichi Itoi
- Akihiro Yamanaka
journal: Scientific Reports
year: 2023
pmcid: PMC9998869
doi: 10.1038/s41598-023-30907-2
license: CC BY 4.0
---
# Cre-dependent ACR2-expressing reporter mouse strain for efficient long-lasting inhibition of neuronal activity
## Abstract
Optogenetics is a powerful tool for manipulating neuronal activity by light illumination with high temporal and spatial resolution. Anion-channelrhodopsins (ACRs) are light-gated anion channels that allow researchers to efficiently inhibit neuronal activity. A blue light-sensitive ACR2 has recently been used in several in vivo studies; however, the reporter mouse strain expressing ACR2 has not yet been reported. Here, we generated a new reporter mouse strain, LSL-ACR2, in which ACR2 is expressed under the control of Cre recombinase. We crossed this strain with a noradrenergic neuron-specific driver mouse (NAT-Cre) to generate NAT-ACR2 mice. We confirmed Cre-dependent expression and function of ACR2 in the targeted neurons by immunohistochemistry and electrophysiological recordings in vitro, and confirmed physiological function using an in vivo behavioral experiment. Our results show that the LSL-ACR2 mouse strain can be applied for optogenetic inhibition of targeted neurons, particularly for long-lasting continuous inhibition, upon crossing with Cre-driver mouse strains. The LSL-ACR2 strain can be used to prepare transgenic mice with homogenous expression of ACR2 in targeted neurons with a high penetration ratio, good reproducibility, and no tissue invasion.
## Introduction
Manipulation of neuronal activity is indispensable for understanding the causal relationships between neuronal activity and outcomes. Optogenetics is a powerful tool for manipulating neuronal activity by light illumination with high temporal and spatial resolution. Opsins are a group of membrane proteins that are activated by light. Channelrhodopsin-2 (ChR2), which is the most well known opsin, is a blue light-gated cation channel that allows for cation inflow to induce depolarization and excitation of neurons1. ChR2-mediated neuronal excitation can often indicate the “sufficiency” of specific neuronal activity for a specific outcome2. On the other hand, the “necessity” of specific neuronal activity is often demonstrated by neuronal inhibition. Archaerhodopsin (Arch) and halorhodopsin (Halo) are commonly used light-driven proton and chloride pumps, respectively, that can induce hyperpolarization and inhibition of neurons3–5. These pump-mediated neuronal inhibitions can demonstrate the “necessity” of specific neuronal activity for a specific outcome6–9. However, as both Arch and Halo are pumps, and not light-driven channels, photocurrents elicited by either of them are lower than those of ion channels.
The discovery of *Guillardia theta* anion-channelrhodopsins (ACRs)10 and the development of chloride-conducting channelrhodopsins11 enabled researchers to overcome the above-mentioned difficulties in optogenetic neuronal inhibition. ACRs are light-gated anion channels that can induce anion transduction, leading to hyperpolarization and inhibition of neurons with a low intracellular chloride ion (Cl−) concentration ([Cl−]). There are two main subtypes of ACRs, namely ACR1 and ACR2, with the major difference between the subtypes being the maximally sensitive light wavelength, which is 515 nm for ACR1 and 470 nm for ACR210.
Compared with the transgene introduction by viral vector injection12, transgenic mouse lines have the advantage of homogeneous gene expression in targeted neurons with a higher penetration ratio, better reproducibility, and no tissue invasion. A reporter mouse strain expressing ACR1 under the control of Cre recombinase was reported recently13, but a reporter mouse strain expressing ACR2 has not been reported yet. In the present study, we generated a new mouse strain in which ACR2 is expressed under the control of Cre recombinase, and confirmed that a selective neuronal population is inhibited by light exposure both in vitro (using slice patch clamp recording) and in vivo (via a behavioral experiment). Since ACR2 is activated by blue light, the present mouse strain can be utilized with the same experimental setup for ChR2.
## NAT-ACR2 mice expressing ACR2 in LC-NA neurons
To express ACR2 exclusively in Cre recombinase-expressing cells, we generated a new mouse strain, namely Rosa26-CAGp-LSL-ACR2-EYFP (LSL-ACR2, Fig. 1a; see also “Materials and methods”). To examine the expression and function of ACR2, we crossed LSL-ACR2 with the noradrenaline-transporter (NAT)-Cre (NAT-Cre) mouse strain14, which expresses Cre recombinase in noradrenergic (NA) neurons, to produce the NAT-Cre;LSL-ACR2 (NAT-ACR2) mouse strain (Fig. 1b). We examined the expression of ACR2 by immunohistochemistry. ACR2-positive (ACR2+) cells were visualized by fused enhanced yellow fluorescent protein (EYFP) and observed in the locus coeruleus (LC), which is a major nucleus of NA neurons (LC-NA neurons) in the forebrain (Fig. 1c,d). ACR2+ cells overlapped with tyrosine-hydroxylase positive (TH+) cells at a high rate, indicating that they are mostly NA neurons in the LC (coverage: ACR2+ in TH+, 86.6 ± $0.8\%$; specificity: TH+ in ACR2+, 93.4 ± $1.8\%$; TH+, 273.3 ± 19.6 cells; ACR2+, 253.0 ± 14.2 cells; TH+ and ACR2+, 236.7 ± 17.4 cells; $$n = 3$$ animals; Fig. 1e). We also observed a characteristic pattern of ACR2 expression in siblings of LSL-ACR2 mice crossed with Vglut2-Cre, Vgat-Cre, and DAT-Cre mice, which express Cre recombinase in glutamatergic, GABAergic, and dopaminergic neurons, respectively (Supplementary Fig. S1). On the other hand, we did not observe any ACR2-EYFP-derived fluorescence in the brains of Cre-negative LSL-ACR2 mice ($$n = 4$$ animals; Supplementary Fig. S2). Taken together, these results suggest that the LSL-ACR2 mouse strain efficiently and exclusively expresses ACR2 in a Cre-dependent manner. Figure 1The LSL-ACR2 mouse strain. ( a) Schematic showing the transgene construction of LSL-ACR2 transgenic mice at Rosa26 locus. R26, Rosa26 locus; CAGp, CAG promoter; loxP, loxP sequence; FRT, flippase recognition target sequence; Neo, neomycin resistant gene; Stop, stop cassette; hGtACR2, human codon-adapted anion channelrhodopsin 2 gene; EYFP, enhanced yellow fluorescent protein; WPRE, woodchuck hepatitis virus posttranscriptional regulatory element; Poly(A), poly-adenylation sequence. ( b) LSL-ACR2 and NAT-Cre mice were crossed to generate NAT-Cre; LSL-ACR2 (NAT-ACR2) mice. ( c) Representative immunostaining image of NAT-ACR2 neurons visualized in the LC. Blue: DAPI, Green: ACR2, Red: TH, scale bar: 500 μm. ( d) Representative immunostaining image of NAT-ACR2 neurons in the right side of the brain visualized in the LC. Insets are magnified images of the dotted area. Blue: DAPI, Green: ACR2, Red: TH, scale bar: 200 μm. ( e) Proportion of positive cells expressing ACR2 and TH in NAT-ACR2 mice ($$n = 3$$ animals). ACR2+/TH+: ACR2+ in TH+ cells, 86.6 ± $0.8\%$. TH+/ACR2+: TH+ in ACR2+ cells, 93.4 ± $1.8\%$.
## Light-induced transient and sustained photocurrents
Next, we examined the function of ACR2 in LC-NA neurons by patch-clamp recording in acute brain slices. Irradiation of ACR2 at a wavelength of 470 nm was previously found to result in rapid inflow of Cl− and inhibition of neuronal activity10. Therefore, we performed patch clamp recording with the whole-cell recording mode using a pipette solution with a low Cl− concentration (8 mM). A brain slice was anchored in a chamber perfused with artificial cerebrospinal fluid (aCSF). Under an epifluorescent microscope, to minimize photoactivation of ACR2, very brief low intensity 505 nm LED illumination (74 µW/mm2, 100 ms) was used to visualize native EYFP fluorescence from ACR2-EYFP as an LC-shaped faint signal, since ACR2-EYFP was expressed on the cellular membrane. We identified LC-NA neurons by a combination of anatomical location, triangular cellular shape, and fluorescence observed in the recording area. In the following experiments, 470 nm light at multiple light intensities (ranging from 1 to $100\%$) were used to activate ACR2 at 0.011 to 3.1 mW/mm2.
Illumination with different intensities of 470 nm light (3 consecutive light exposures with 5-s intervals) induced outward photocurrents, which were recorded by holding the membrane potential at − 60 mV (Fig. 2a). The effect of light intensity on the photocurrent amplitude is shown in the single trace recording in Fig. 2b. To further confirm the correlation between amplitude and light intensity, we defined the baseline, transient current, and sustained current as noted in Fig. 2c. Given that cell size affects the amount of ACR2 expression, we examined current density (i.e., photocurrent amplitude divided by membrane capacity), because absolute value of the current depends on cell size. Indeed, the current density was found to be positively correlated with the light intensity (Fig. 2d,e). Consistent with the results of a previous study10, the transient photocurrent reached the saturation level at $20\%$ (0.93 mW/mm2) light intensity (Fig. 2d), and the sustained photocurrent reached the saturation level at $5\%$ (0.23 mW/mm2) light intensity (Fig. 2e). Furthermore, transient and sustained photocurrents were also induced by long-term (30 s) light illumination (Fig. 2f,g). These results demonstrate that ACR2 in LC-NA neurons was functional, and that it induced an outward current upon illumination with 470 nm light. Figure 2Light-induced photocurrents in ACR2-expressing neurons. ( a) Representative trace of photocurrent recording with illumination at different intensities of 470 nm light. Membrane potential was held at − 60 mV. Light intensity was adjusted from 1 to $100\%$, corresponding to 0.011 to 3.1 mW/mm2, as indicated above the trace. ( b) Single trace of photocurrent recording during illumination at different light intensities. ( c) Definition of baseline, transient, and sustained currents. Baseline: 100 ms before light onset for each illumination. Transient: 100 ms after light onset for each illumination. Sustained: 100 ms before light offset for each illumination. ( d,e) Transient and sustained current densities as a function of light intensity in (d) and (e), respectively. Dots show mean values ± SEM of 9 cells from 2 animals. ( f) Representative trace of photocurrent recording at 470 nm, 30 s light illumination; the membrane potential was held at − 60 mV. Light intensity: $100\%$, corresponding to 3.1 mW/mm2. ( g) Transient and sustained current densities ($$n = 4$$ cells from an animal). Dots show individual data.
## ACR2 inhibits neuronal activity by light-induced chloride ion flow
To further confirm whether ACR2 generates outward photocurrent by inducing the inward flow of Cl−, we investigated the reversal potential of the photocurrent ions by voltage clamp step recording. Light illumination (200 ms, 470 nm) at $50\%$ (2.0 mW/mm2), $5\%$, and $2\%$ (0.067 mW/mm2) intensity induced different degrees of outward or inward photocurrent under a voltage step from − 120 to − 40 mV (Fig. 3a–c). We further analyzed the correlation between sustained photocurrent and holding potential (Fig. 3d) and the reversal potential was calculated by fitting a straight line (Fig. 3e). The holding potential was adjusted with the calculated junction potential (− 14.9 mV). Higher light intensity produced higher sustained induced current density, which was consistent with our previous results (Fig. 2d,e). All reversal potentials of the photocurrent ions during $50\%$ (− 80.7 ± 0.5 mV, $$n = 11$$ cells), $5\%$ (− 82.4 ± 0.7 mV, $$n = 12$$ cells), and $2\%$ (− 83.1 ± 1.0 mV, $$n = 11$$ cells) light illumination were similar to the theoretical reversal potential of Cl− (− 72.6 mV) (Fig. 3e) under the recording conditions. Figure 3ACR2-induced chloride ion inflow and outflow upon illumination. ( a–c) Upper: representative traces of photocurrent recording with illumination at different intensities of 470 nm light held at − 120 mV to − 40 mV. Bottom: schematic of the command voltage. The dotted rectangle indicates the timing of the upper traces. Light intensity: $50\%$, 2.0 mW/mm2 (a); $5\%$, 0.23 mW/mm2 (b); $2\%$, 0.067 mW/mm2 (c). ( d) Sustained current densities as a function of the holding potential. The holding potential was corrected with the junction potential (− 14.9 mV). The dots show mean values ± SEM. Light intensity at $50\%$, $$n = 11$$ cells; $5\%$, $$n = 12$$ cells; $2\%$, $$n = 11$$ cells from 2 animals. The dotted line indicates 0 pA/pF. (e) The reversal potential for different light intensities is based on each recording by linear fitting. Bars show the mean values ± SEM. Dots show individual values from each cell. The dotted line indicates the reversal potential of Cl− (− 72.6 mV).
We also examined membrane potential change by current clamp recording during light illumination (200 ms, 470 nm) at $50\%$, $5\%$, and $2\%$ intensity (Fig. 4a). To analyze the amplitude of the membrane potential change, we defined the baseline potential as the median potential during 15 s before light onset of the first illumination of the session (− 59.5 ± 1.1 mV, $$n = 10$$ cells), and the sustained potential as the mean of 100 ms before light offset of each illumination (Fig. 4b). Similar to the voltage clamp step recordings, the reversal potentials of the photocurrent ions at $50\%$ (− 74.8 ± 1.1 mV, $$n = 10$$ cells), $5\%$ (− 75.9 ± 1.1 mV, $$n = 10$$ cells), and $2\%$ light illumination (− 76.8 ± 1.1 mV, $$n = 10$$ cells) were close to the theoretical reversal potential of Cl− (− 72.6 mV) (Fig. 4c) under the recording conditions. These results suggest that ACR2 inhibited cellular activity by inducing chloride ion inflow upon light illumination in LC-NA neurons. Figure 4ACR2-inhibited neuronal activity by chloride ion inflow upon illumination. ( a) Representative trace of membrane potential recording with illumination at different intensities of 470 nm light. Light intensity was adjusted from 50 to $2\%$, corresponding to 2.0 to 0.067 mW/mm2, as indicated above the trace. Data within the dotted rectangle is also shown in (b). ( b) Definition of baseline and sustained membrane potential. Baseline: 15 s before light onset for the first illumination of the session. Sustained: 100 ms before light offset for each illumination. ( c) The reversal potential for different light intensities was calculated based on current clamp recording ($$n = 10$$ cells from 2 animals). Bars show mean values ± SEM. Dots show individual data from each cell. Dotted line: Reversal potential of Cl− (− 72.6 mV). ( d) Comparison of neuronal firing rate before and after light onset ($$n = 10$$ cells from 2 animals). ** $p \leq 0.01$, paired t-test. The baseline was defined as shown in (b). Bars show mean values ± SEM. Dots show individual cell data including all light intensities.
## ACR2 induced durable and long-lasting inhibition of neuronal activity
To further confirm the influence of ACR2 on neuronal activity, we also examined the change in neuronal firing by current clamp recording. We found that light illumination with $50\%$, $5\%$, and $2\%$ intensity caused hyperpolarization and completely inhibited generation of action potential (paired t-test, $$p \leq 0.004$$, $$n = 10$$ cells, Fig. 4d). In the loose cell-attached mode recording, which does not rupture the membrane and does not alter the natural intracellular environment, ACR2 also inhibited neuronal firing during long-term light illumination (11 µW/mm2, 10 min) and did not significantly affect the firing rate after illumination (Tukey’s test; baseline vs. light on, $$p \leq 0.02$$; light on vs. after light, $$p \leq 0.004$$; baseline vs. after light, $$p \leq 0.45$$; $$n = 5$$ cells, Fig. 5a,b). These results show that ACR2 could inhibit cellular activity in LC-NA neurons, and did not affect neuronal activity after optogenetic inhibition for a long period of time. Figure 5ACR2 continuously inhibited neuronal activity. ( a) Representative trace of loose cell recording at 470 nm, 10 min light illumination. Light intensity was $1\%$, corresponding to 11 µW/mm2. ( b) Comparison of the neuronal firing rate before, during, and after light illumination ($$n = 5$$ cells from an animal). ' Baseline' and 'after light' were defined as 30 s before and after light illumination, respectively. * $p \leq 0.05$, **$p \leq 0.01$, n.s., not significant, Tukey’s test. The bars show mean values ± SEM. The dots show individual cell data.
## ACR2 was functional in free-moving animals in vivo
Finally, to confirm the function of ACR2 in vivo, we performed a real-time place preference test (RT-PPT). We implanted an optical fiber cannula above the LC unilaterally in NAT-ACR2 mice and NAT-Cre; Ai14 mice (NAT-tdTomato mice) (Fig. 6a). NAT-tdTomato mice express the red fluorescent protein tdTomato in NA neurons as a negative control. We allowed the animals to explore a two-chamber apparatus for 10 min without light illumination (‘OFF’ session), and identified a preferred chamber at baseline. Then, we allowed the animals to explore the apparatus for 10 min (‘ON’ session), with illumination of continuous light (470 nm) while the animal was in the un-preferred chamber, i.e., the opposite side of the preferred chamber (opto-paired chamber; Fig. 6b). Then, we examined the shift of the preference induced by light illumination in NAT-ACR2 and NAT-tdTomato mice. We calculated the ‘preference modulation ratio,’ which is the time spent in the opto-paired chamber during the ‘ON’ session divided by that during the ‘OFF’ session. We found that the preference modulation ratio was significantly higher in NAT-ACR2 mice than in NAT-tdTomato mice ($$p \leq 0.026$$, Welch’s t-test; Fig. 6c), suggesting that light illumination of ACR2 in NAT-ACR2 mice significantly increased the animals’ place preference. We also analyzed raw values of the time spent in the opto-paired chamber during ‘OFF’ and ‘ON’ sessions (within-subjects factor ‘light’) of NAT-ACR2 and NAT-tdTomato mice (between-subjects factor ‘gene’) by two-way repeated measures ANOVA. We found a significant interaction between ‘light’ and ‘gene’ factors ($$p \leq 0.021$$). Furthermore, we found a significant difference in the ‘light’ factor in NAT-ACR2 ($$p \leq 0.019$$, post hoc Tukey’s test), but no significant difference in the ‘gene’ factor during the ‘ON’ session ($$p \leq 0.076$$, post hoc Tukey’s test; Fig. 6d). We suggest that the lack of significance could be due to the increased variance of the place preference among animals during the ‘ON’ session. Finally, we examined the preference change throughout a session. We found that the time spent in the opto-paired chamber during the 10 min experiment of NAT-tdTomato mice was unchanged between the ‘OFF’ and ‘ON’ sessions (Fig. 6e), while that of NAT-ACR2 mice was gradually increased during the ‘ON’ session (Fig. 6f). These results suggest that light illumination of ACR2 expressed in NAT-ACR2 successfully modulated animals’ behavior. It was also shown that a unilateral inhibition of LC-NA neuronal activity was sufficient to induce place preference. Figure 6In vivo inhibition of LC-NA neuron-induced place preference. ( a) Representative brain slice image showing the position of the implanted fiber above the LC. f, fiber tract; 4v, fourth ventricle. Green: ACR2, Red: TH, Blue: DAPI, scale bar: 500 μm. ( b) Schematic of real-time place preference test (RT-PPT). After an animal was habituated with an optic fiber for > 1 h in the home cage (Habituation), the place preference at baseline was examined for 10 min in the two-chamber apparatus (Preference check, ‘OFF’). Then, the animal was temporarily placed back into the home cage and preference analysis was performed after ~ 5 min (Home cage). The opto-paired chamber was determined to be the non-preferred chamber, and RT-PPT was performed for 10 min (RT-PPT, ‘ON’). ( c) Preference modulation ratio of control animals expressing tdTomato and animals expressing ACR2. * $p \leq 0.05$ (tdTomato, $$n = 6$$ animals; ACR2, $$n = 6$$ animals, Welch’s t-test). ( d) Time spent in the opto-paired chamber during 10 min, or the preference, at baseline (OFF) and RT-PPT (ON). †$p \leq 0.05$, significant interaction shown in two-way repeated measures (RM) ANOVA between factors ‘light’ and ‘gene’. * $p \leq 0.05$, ‘$$p \leq 0.076$$’, post hoc Tukey’s test. tdTomato, $$n = 6$$ animals; ACR2, $$n = 6$$ animals. ( e and f) The top heat maps show the time spent in the two-chamber cage for a representative animal from each group. P, preferred chamber; O, opto-paired chamber. The line graph shows the cumulative place preference over 10 min. n.s., not significant ($$n = 6$$ animals, Tukey’s test for every 10 s, OFF vs. ON); *$p \leq 0.05$ ($$n = 6$$ animals, Tukey’s test for every 10 s, OFF vs. ON). Line, mean; shadow, standard error of the mean.
## Discussion
In this study, we generated and characterized a new mouse strain, namely LSL-ACR2, for exclusive inhibition of neuronal activity in Cre-expressing targeting neurons (Fig. 1a). The NAT-ACR2 mice were generated by crossing LSL-ACR2 mice with NAT-Cre mice. The function of ACR2 in vitro was confirmed by patch-clamp recording, and it was shown that ACR2 activated by 470 nm light could effectively inhibit neuronal activity by inducing Cl− inflow. This study provides the theoretical basis for the use of the LSL-ACR2 mouse strain in behavioral experiments. To date, ACR2 expression in rodent neurons has been achieved by viral infection15–19. To use viral vectors, researchers need to optimize various indefinite factors of the virus, such as the position, volume, titer, and serotype. When using custom-made viral vectors, researchers also need to consider the promoter and other DNA constructs. On the other hand, using the LSL-ACR2 with Cre-driver mice, researchers can skip all of the above-mentioned processes needed for viral vectors. Thus, the LSL-ACR2 mouse strain will allow researchers to more conveniently express ACR2 in a specific subtype of neurons with homogenous quantity and quality.
We found that neuronal activity was inhibited via ACR2 by 470 nm light at intensities as low as 11 µW/mm2, and that the effects lasted for > 10 min. Remarkably, neuronal activity recovered immediately after the termination of long-term inhibition (Fig. 5). An advantage of ACR2 is the efficiency of phototransduction compared with other inhibitory optogenetic tools, including light-driven pumps such as Arch and Halo, and also gene-engineered chloride-conducting channelrhodopsins, such as slowChloC and iC++10, 18. The characteristic lower intensity and longer duration light-inducibility is advantageous for in vivo applications, since it will cause less heat generation and less phototoxicity20, 21. Furthermore, lower intensity light inducibility might presumably be applicable for optogenetic manipulation without intracranial surgery like ChRmine and OPN4dC22, 23. To inhibit neuronal activity for a prolonged period of time, researchers can also use chemogenetic approaches, such as designer receptors exclusively activated by designer drugs (DREADDs)24. Compared to chemogenetic approaches, which simply require drug administration, using ACR2 requires extra optical equipment and surgery. However, since ACR2 is manipulated by light, researchers can inhibit neuronal activity for any duration with an accuracy of milliseconds, which cannot be achieved by chemogenetic approaches. Regarding subtypes of ACRs, ACR2 exhibits faster closing kinetics than ACR125. This characteristic enables more accurate timing control in the optogenetic manipulation in ACR2 than in ACR1. The LSL-ACR2 mouse strain will provide a new opportunity for researchers to use the efficient inhibitory optogenetic tool.
We found, in the present study, that unilateral continuous inhibition of the activity of LC-NA neurons was sufficient to induce place preference (Fig. 6c). This finding is consistent with a previous study showing that successive photoactivation of LC-NA neurons with ChR2 decreased place preference in RT-PPT26. It is known that stressors can activate LC-NA neurons. On the other hand, the tonic activity of LC-NA neurons can cause anxiety-like and aversion behavior26. Therefore, inhibition of the activity of LC-NA neurons might have suppressed such negative emotions and serve as a positive valence.
A possible limitation of the utility of ACR2 is the dependence on Cl− for its function. Some neurons show higher intracellular [Cl−] than extracellular [Cl−], particularly during development27–30. In these cases, the reversal potential of Cl− can be higher than the resting membrane potential, and ACR2 activation can induce depolarization rather than hyperpolarization. Therefore, one should take into account the reversal potential of Cl− of the targeted neurons for optogenetic manipulation using ACR2. Further, one might utilize this characteristic of ACR2 to examine whether the reversal potential is higher than the resting membrane potential of recorded neurons by loose cell-attached recording—when the reversal potential is higher than the membrane potential, light illumination results in outflow of Cl− and the extracellular potential can be decreased, and vice versa.
It should also be noted that ACR2 used in this study is not the soma-targeted version of ACR2 (stGtACR2) reported by Mahn et al.18. As such, brief illumination of an axon terminal might cause a transient excitation because of the higher axonal reversal potential of Cl−18, 31. We indeed observed axonal signals in the hippocampal dentate gyrus of the NAT-ACR2 mouse brain (Supplementary Fig. S3), which is a region receiving projection from LC-NA neurons32, 33. Therefore, the most appropriate way to utilize the LSL-ACR2 mouse strain is for long-lasting continuous inhibition of neuronal activity, as reported in this study and performed during the RT-PPT. ACR2 expressed in LSL-ACR2 mice can inhibit the activity of targeted neurons for longer periods of time with relatively weak light intensity, which would not generate heat and phototoxicity. Thus, we believe that the LSL-ACR2 mouse strain, by crossing with Cre-driver mice, will be useful for studies aimed at demonstrating “necessity” of specific neuronal activity by investigating the physiological role of targeted neurons in vitro and in vivo.
## Animals
In this study, Rosa26-CAGp-LSL-ACR2-eYFP (Gt(ROSA)26Sortm1(CAG-ACR2/EYFP)Ksak, or LSL-ACR2) was newly generated using a method similar to that used to generate the CAG-floxed STOP tdTomato reporter line (MGI: 6192640) in a previous study34. Briefly, we constructed the targeting vector containing a CAG promoter35, frt flanked pgk-Neo cassette36, STOP cassette consisting of the terminator of the yeast *His3* gene and SV40 poly-adenylation sequence37, cDNA encoding ACR2 tagged with EYFP at the C-terminus from pFUGW-hGtACR2-EYFP (Addgene: Plasmid #67877), woodchuck hepatitis virus posttranscriptional regulatory element38, and rabbit b-globin poly-adenylation sequence39. Two loxP sites were inserted before the frt-Neo cassette and after the STOP cassette37. This vector also exhibits 5’ and 3’ homology arms of 4.7- and 5.2-kb, respectively, which target the Xba1 site of intron 1 at the Rosa26 locus40. The targeting vector (DDBJ: LC744045) was linearized and electroporated into the RENKA C57BL/6 embryonic stem cell line41. G418-resistant ES clones were screened by Southern blot analysis for homologous recombination at the Rosa26 locus. Targeted ES clones were injected into eight-cell stage CD-1, which were cultured to produce blastocysts and later transferred to pseudopregnant CD-1 females. The resulting male chimeric mice were crossed with female C57BL/6 mice to establish the LSL-ACR2 line. The LSL-ACR2 mice used in the present study exhibit the Neo cassette. Previous studies showed that there is no difference in Rosa26 reporter expression with or without removal of the Neo cassette42, 43. Therefore, we did not test removal of the Neo cassette in the present study. The LSL-ACR2 mouse strain was raised in an inbred-manner for 6 to 13 generations after introduction into our animal facility. All progenies of LSL-ACR2 mice crossed with Cre-driver mice showed consistent expression of ACR2 ($$n = 20$$ animals). To express ACR2 in NA neurons, noradrenaline transporter (NAT)-Cre (Tg(Slc6a2-cre)FV319Gsat) mice14 and LSL-ACR2 mice were crossed to generate NAT-Cre;LSL-ACR2 (NAT-ACR2) mice, which were used for optogenetic experiments (total 14 animals) and immunohistochemistry (3 animals). To express tdTomato in NA neurons, Ai14 (B6.Cg-Gt(ROSA)26Sortm14(CAG-tdTomato)Hze/J) mice42 and NAT-Cre mice were crossed to generate NAT-Cre;Ai14 (NAT-tdTomato) mice, which were used for negative control experiments in vivo (total 6 animals). To express ACR2 in Vglut2-positive glutamatergic, Vgat-positive GABAergic, or DAT-positive dopaminergic neurons, Vglut2-Cre (Slc17a6tm2(cre)Lowl)44, Vgat-Cre (Slc32a1tm2(cre)Lowl)44, or DAT-Cre (Slc6a3tm1.1(cre)Bkmn)45 mice were crossed with LSL-ACR2 mice to generate Vglut2-Cre;LSL-ACR2, Vgat-Cre;LSL-ACR2, or DAT-Cre;LSL-ACR2 mice, respectively, which were subsequently used to confirm expression (one animal for each strain). To confirm lack of ACR2 expression in Cre-negative animals, LSL-ACR2 mice (4 animals) and NAT-Cre mice (2 animals) were used. Adult mice (aged > 6 weeks) were used. Animals were housed at 23 ± 2 °C with a 12-h light–dark cycle, and feeding and drinking were available ad libitum. All experiments were carried out following the ARRIVE guidelines 2.046 and the Nagoya University Regulations on Animal Care and Use in Research, and were approved by the Institutional Animal Care and Use Committees of the Research Institute of Environmental Medicine, Nagoya University, Japan (approval R210096 and R210729).
## Immunohistochemistry
Animals were perfused with $4\%$ paraformaldehyde (PFA). The brain was removed and fixed in PFA at 4 °C. After 6 h, PFA was replaced with phosphate-buffered saline (PBS) containing $0.05\%$ NaN3 (PBS + NaN3) and the sample was allowed to sit overnight at 4 °C. On the next day, the brain was embedded in $3\%$ agarose dissolved in PBS + NaN3. Agarose was fixed after 30 min, and 40-µm thick brain slices were sectioned using a vibratome (VT1000S; Leica). Brain slices from NAT-ACR2 mice were collected every 160 µm. Brain slices from Vglut2-Cre;LSL-ACR2, Vgat-Cre;LSL-ACR2, DAT-Cre;LSL-ACR2, LSL-ACR2, and NAT-Cre mice were collected every 320 µm. Brain slices were washed with PBS-BX ($1\%$ bovine serum albumin, $0.25\%$ Triton X-100 in PBS) 3 times every 15 min at room temperature. The slices were then incubated in primary antibodies (rabbit anti-TH (1:1000, AB152, Chemicon) and chicken anti-GFP (1:1000, GFP-1010, Aveslabs)) diluted with PBS-BX overnight at 4 °C. The next day, the slices were washed with PBS-BX 3 times every 15 min at room temperature. The slices were then incubated in secondary antibodies (CF647 conjugated anti-rabbit IgG (1:1000, 20047-1, BTI) and CF488 conjugated anti-chicken IgY (1:1000, 20079-1, BTI)) diluted with PBS-BX at room temperature for 2 h. Next, the slices were washed with PBS-BX 3 times every 10 min at room temperature. Once washed with PBS + NaN3, the slices were incubated with DAPI solution (2 µM, 043-18804, Wako) in PBS + NaN3 for 1 h at room temperature, followed by washing with PBS + NaN3 3 times every 10–15 min.
## Imaging and image analysis
Images were acquired with a Zeiss LSM 710 inverted confocal laser scanning microscope and a Keyence BZ-X710 fluorescence microscope. To count the number of ACR2-expressing cells, a 10× objective lens was used in the Zeiss LSM 710 with 405-, 488-, and 561-nm argon lasers. To verify positions, a 4× objective lens was used in the Keyence BZ-X710 with DAPI, GFP, and Cy5 filter cubes (Keyence). The ImageJ software47 was used for adjustment of brightness and contrast, and quantification of ACR2-expressing cells (Cell Counter plugin). Three brain slices on the right side of 3 different mice were used for cell counting. Cells expressing TH or ACR2, as well as cells expressing both TH and ACR2 simultaneously, were counted.
## Electrophysiology
Animals were anesthetized with isoflurane. After decapitation, the brain was quickly transferred to the frozen cutting solution (containing, in mM, 15 KCl, 3.3 MgCl2, 110 K-gluconate, 0.05 EGTA, 5 HEPES, 25 glucose, 26.2 NaHCO3 and 0.0015 (±)-3-(2-carboxypiperazin-4-yl)propyl-1-phosphonic acid) with carbogen gas ($95\%$ O2 and $5\%$ CO2). The brain was sliced into 250-µm thick sections using a vibratome (VT1200S, Leica), and transferred into artificial cerebrospinal fluid (aCSF, containing, in mM, 124 NaCl, 3 KCl, 2 MgCl2, 2 CaCl2, 1.23 NaH2PO4, 26 NaHCO3, 25 glucose) with carbogen gas ($95\%$ O2 and $5\%$ CO2) at 35 °C for at least 1 h, then at room temperature covered with aluminum foil to avoid light exposure. An amplifier (Multiclamp 700B, Molecular Devices) and a digitizer (Axon Digidata 1550B, Molecular Devices) were used for patch clamp recording. The recording chamber was perfused with aCSF saturated with carbogen gas ($95\%$ O2 and $5\%$ CO2) at room temperature. A glass pipette (GC150-10; Harvard Apparatus) was made with a puller (P-1000, Sutter Instrument) and its resistance was between 2.8 and 7 MΩ. The pipette was loaded with K-gluconate-based pipette solution (in mM, 138 K-gluconate, 8 NaCl, 10 HEPES, 0.2 EGTA-Na3, 2 Mg-ATP, and 0.5 Na2-GTP, pH 7.3 with KOH) for whole-cell recording, or aCSF for loose cell recording. Under an epifluorescent microscope (BX51WI, Olympus), 505 nm LED illumination (74 µW/mm2, 100 ms, Niji, Blue Box Optics) was used to visualize native EYFP fluorescence from ACR2-EYFP. We identified LC-NA neurons by a combination of anatomical location, triangular cellular shape, and fluorescence observed in the recording area. In Fig. 2, the membrane potential was held at − 60 mV for measuring current deflection. In Fig. 3, the membrane potential was held from − 120 to − 40 mV in 20 mV steps with a duration of 700 ms. The voltage deflection was evaluated at a current holding of 0 pA. Clampex 11.0.3 (Molecular Devices) was used to record the data.
## Optogenetic manipulation in the brain slice
Light illumination was delivered through an electronic stimulator (SEN-3301, Nihon Kohden) connected to a light source (470 nm, 3.1 mW/mm2 at maximum, Niji, Blue Box Optics). The light intensity was controlled by our original Python programs48 with a microcontroller (Arduino Uno R3). In Fig. 2a, we set the delay at 0 s, the interval at 10 s, the duration at 5 s, and the train at three times, and the intensity was automatically adjusted to 1, 2, 5, 10, 20, 50, and $100\%$. In Fig. 3a–c, we set the delay at 200 ms, the interval at 0 s, the duration at 200 ms, and the train at 1, and the intensity was adjusted to 50, 5, and $2\%$. In Fig. 4a, we set the delay at 0 ms, the interval at 15 s, the duration at 5 s, and the train at 3 times, and the intensity was automatically adjusted to 50, 5, and $2\%$. In Fig. 2f, we set the delay at 0 ms, the interval at 0 s, the duration at 30 s, the train at 1, and the intensity at $100\%$. In Fig. 5a, we set the delay at 0 ms, the interval at 0 s, the duration at 600 s, the train at 1, and the intensity at $1\%$.
## Real-time place preference test (RT-PPT)
Animals were implanted with an optic cannula (φ400 µm, 0.39 NA; F0618S04B2P, Kyocera) above the LC unilaterally (tip at 5.6 mm posterior and 0.9 mm lateral to the bregma, and 3.0 mm ventral from the brain surface). One day after the surgery, animals were habituated to experimenters’ hands for 30 s twice per day for at least 1 week. After animals were habituated, a fiber cannula of an animal was connected with an optic fiber cable (φ400 µm, 0.39 NA; M98L01, Thorlabs) with an interconnect (ADAL3, Thorlabs), attached to a light source (M470F3, Thorlabs). The animal was placed in one side of a two-chamber cage (11 cm in width, 13 cm in depth, and 14 cm in height/chamber) with different floor textures (metal mesh and smooth) and wall appearance (black–white stripes and white) and allowed to explore both sides of the chamber for 10 min (baseline ‘OFF’ session). Animal behavior was monitored by a USB camera (ELP-USBFHD05MT-KL36IR, Ailipu Technology). The position of the nose was identified by Deeplabcut-live49 with a pre-trained dataset. After the baseline session, the animal was replaced into the home cage, and the duration in each chamber was instantly analyzed. A chamber in which the animal stayed for a longer duration was defined as the preferred chamber. The animal was then placed in the preferred chamber, and allowed to explore the two-chamber cage for 10 min (RT-PPT ‘ON’ session). During the RT-PPT session, continuous light illumination (470 nm, 50–60 µW at the tip of an optic cannula) was delivered while the animal’s nose was in the chamber opposite to the preferred chamber (i.e., non-preferred chamber). Light illumination was controlled via microcontrollers (Arduino Uno R3) and synchronized red LED light, which was not observable from the subject animal, was shown to the camera. After recording, the recorded videos were analyzed with Deeplabcut50, 51 offline and the time spent in each chamber was calculated.
## Statistical analysis
Statistical analysis was performed in OriginPro 2020 (OriginLab Corporation). In Fig. 4d, a paired t-test was used. In Figs. 5b, 6e,f, Tukey’s test was used. In Fig. 6c, Welch’s t-test was used. In Fig. 6d, two-way repeated measures ANOVA followed by post-hoc Tukey’s test was used. Quantitative data are shown as the mean ± standard error of the mean.
## Supplementary Information
Supplementary Figures. The online version contains supplementary material available at 10.1038/s41598-023-30907-2.
## References
1. Boyden ES, Zhang F, Bamberg E, Nagel G, Deisseroth K. **Millisecond-timescale, genetically targeted optical control of neural activity**. *Nat. Neurosci.* (2005.0) **8** 1263-1268. DOI: 10.1038/nn1525
2. Adamantidis AR, Zhang F, Aravanis AM, Deisseroth K, de Lecea L. **Neural substrates of awakening probed with optogenetic control of hypocretin neurons**. *Nature* (2007.0) **450** 420-424. DOI: 10.1038/nature06310
3. Chow BY. **High-performance genetically targetable optical neural silencing by light-driven proton pumps**. *Nature* (2010.0) **463** 98-102. DOI: 10.1038/nature08652
4. Han X, Boyden ES. **Multiple-color optical activation, silencing, and desynchronization of neural activity, with single-spike temporal resolution**. *PLoS ONE* (2007.0) **2** e299. DOI: 10.1371/journal.pone.0000299
5. Zhang F. **Multimodal fast optical interrogation of neural circuitry**. *Nature* (2007.0) **446** 633-639. DOI: 10.1038/nature05744
6. Izawa S. **REM sleep–active MCH neurons are involved in forgetting hippocampus-dependent memories**. *Science* (2019.0) **365** 1308-1313. DOI: 10.1126/science.aax9238
7. Tsunematsu T. **Acute optogenetic silencing of orexin/hypocretin neurons induces slow-wave sleep in mice**. *J. Neurosci.* (2011.0) **31** 10529-10539. DOI: 10.1523/JNEUROSCI.0784-11.2011
8. Tsunematsu T. **Long-lasting silencing of orexin/hypocretin neurons using archaerhodopsin induces slow-wave sleep in mice**. *Behav. Brain Res.* (2013.0) **255** 64-74. DOI: 10.1016/j.bbr.2013.05.021
9. Tsunematsu T. **Optogenetic manipulation of activity and temporally controlled cell-specific ablation reveal a role for MCH neurons in sleep/wake regulation**. *J. Neurosci.* (2014.0) **34** 6896-6909. DOI: 10.1523/JNEUROSCI.5344-13.2014
10. Govorunova EG, Sineshchekov OA, Janz R, Liu X, Spudich JL. **Natural light-gated anion channels: A family of microbial rhodopsins for advanced optogenetics**. *Science* (2015.0) **349** 647-650. DOI: 10.1126/science.aaa7484
11. Berndt A. **Structural foundations of optogenetics: Determinants of channelrhodopsin ion selectivity**. *Proc. Natl. Acad. Sci.* (2016.0) **113** 822-829. DOI: 10.1073/pnas.1523341113
12. Mukai Y, Nagayama A, Itoi K, Yamanaka A. **Identification of substances which regulate activity of corticotropin-releasing factor-producing neurons in the paraventricular nucleus of the hypothalamus**. *Sci. Rep.* (2020.0) **10** 13639. DOI: 10.1038/s41598-020-70481-5
13. Li N. **Spatiotemporal constraints on optogenetic inactivation in cortical circuits**. *Elife* (2019.0) **8** e48622. DOI: 10.7554/eLife.48622
14. Gong S. **Targeting Cre recombinase to specific neuron populations with bacterial artificial chromosome constructs**. *J. Neurosci.* (2007.0) **27** 9817-9823. DOI: 10.1523/JNEUROSCI.2707-07.2007
15. Acharya AR. **Attenuation of hippocampal evoked potentials in vivo by activation of GtACR2, an optogenetic chloride channel**. *Front. Neurosci.* (2021.0) **15** 653844. DOI: 10.3389/fnins.2021.653844
16. Grove JCR. **Dopamine subsystems that track internal states**. *Nature* (2022.0) **608** 374-380. DOI: 10.1038/s41586-022-04954-0
17. Hayat H. **Locus coeruleus norepinephrine activity mediates sensory-evoked awakenings from sleep**. *Sci. Adv.* (2020.0) **6** eaaz4232. DOI: 10.1126/sciadv.aaz4232
18. Mahn M. **High-efficiency optogenetic silencing with soma-targeted anion-conducting channelrhodopsins**. *Nat. Commun.* (2018.0) **9** 4125. DOI: 10.1038/s41467-018-06511-8
19. Vierock J. **BiPOLES is an optogenetic tool developed for bidirectional dual-color control of neurons**. *Nat. Commun.* (2021.0) **12** 4527. DOI: 10.1038/s41467-021-24759-5
20. Owen SF, Liu MH, Kreitzer AC. **Thermal constraints on in vivo optogenetic manipulations**. *Nat. Neurosci.* (2019.0) **22** 1061-1065. DOI: 10.1038/s41593-019-0422-3
21. Stujenske JM, Spellman T, Gordon JA. **Modeling the spatiotemporal dynamics of light and heat propagation for in vivo optogenetics**. *Cell Rep.* (2015.0) **12** 525-534. DOI: 10.1016/j.celrep.2015.06.036
22. Chen R. **Deep brain optogenetics without intracranial surgery**. *Nat. Biotechnol.* (2021.0) **39** 161-164. DOI: 10.1038/s41587-020-0679-9
23. Takahashi TM. **Optogenetic induction of hibernation-like state with modified human Opsin4 in mice**. *Cell Rep. Methods* (2022.0). DOI: 10.1016/j.crmeth.2022.100336
24. Roth BL. **DREADDs for neuroscientists**. *Neuron* (2016.0) **89** 683-694. DOI: 10.1016/j.neuron.2016.01.040
25. Govorunova EG. **The expanding family of natural anion channelrhodopsins reveals large variations in kinetics, conductance, and spectral sensitivity**. *Sci. Rep.* (2017.0) **7** 43358. DOI: 10.1038/srep43358
26. McCall JG. **CRH engagement of the locus coeruleus noradrenergic system mediates stress-induced anxiety**. *Neuron* (2015.0) **87** 605-620. DOI: 10.1016/j.neuron.2015.07.002
27. Ben-Ari Y, Cherubini E, Corradetti R, Gaiarsa JL. **Giant synaptic potentials in immature rat CA3 hippocampal neurones**. *J. Physiol.* (1989.0) **416** 303-325. DOI: 10.1113/jphysiol.1989.sp017762
28. Cherubini E, Gaiarsa JL, Ben-Ari Y. **GABA: An excitatory transmitter in early postnatal life**. *Trends Neurosci.* (1991.0) **14** 515-519. DOI: 10.1016/0166-2236(91)90003-D
29. Glykys J. **Local impermeant anions establish the neuronal chloride concentration**. *Science* (2014.0) **343** 670-675. DOI: 10.1126/science.1245423
30. Rivera C. **The K**. *Nature* (1999.0) **397** 251-255. DOI: 10.1038/16697
31. Turecek R, Trussell LO. **Presynaptic glycine receptors enhance transmitter release at a mammalian central synapse**. *Nature* (2001.0) **411** 587-590. DOI: 10.1038/35079084
32. Takeuchi T. **Locus coeruleus and dopaminergic consolidation of everyday memory**. *Nature* (2016.0) **537** 357-362. DOI: 10.1038/nature19325
33. Seo D. **A locus coeruleus to dentate gyrus noradrenergic circuit modulates aversive contextual processing**. *Neuron* (2021.0). DOI: 10.1016/j.neuron.2021.05.006
34. Nakayama H. **Microglia permit climbing fiber elimination by promoting GABAergic inhibition in the developing cerebellum**. *Nat. Commun.* (2018.0) **9** 2830. DOI: 10.1038/s41467-018-05100-z
35. Niwa H, Yamamura K, Miyazaki J. **Efficient selection for high-expression transfectants with a novel eukaryotic vector**. *Gene* (1991.0) **108** 193-199. DOI: 10.1016/0378-1119(91)90434-D
36. Takeuchi T. **Flp recombinase transgenic mice of C57BL/6 strain for conditional gene targeting**. *Biochem. Biophys. Res. Commun.* (2002.0) **293** 953-957. DOI: 10.1016/S0006-291X(02)00321-2
37. Sauer B. **[53] Manipulation of transgenes by site-specific recombination: Use of cre recombinase**. *Methods Enzymol.* (1993.0) **225** 890-900. DOI: 10.1016/0076-6879(93)25056-8
38. Zufferey R, Donello JE, Trono D, Hope TJ. **Woodchuck hepatitis virus posttranscriptional regulatory element enhances expression of transgenes delivered by retroviral vectors**. *J. Virol.* (1999.0) **73** 2886-2892. DOI: 10.1128/JVI.73.4.2886-2892.1999
39. Gil A, Proudfoot NJ. **Position-dependent sequence elements downstream of AAUAAA are required for efficient rabbit β-globin mRNA 3′ end formation**. *Cell* (1987.0) **49** 399-406. DOI: 10.1016/0092-8674(87)90292-3
40. Zambrowicz BP. **Disruption of overlapping transcripts in the ROSA βgeo 26 gene trap strain leads to widespread expression of β-galactosidase in mouse embryos and hematopoietic cells**. *Proc. Natl. Acad. Sci.* (1997.0) **94** 3789-3794. DOI: 10.1073/pnas.94.8.3789
41. Mishina M, Sakimura K. **Conditional gene targeting on the pure C57BL/6 genetic background**. *Neurosci. Res.* (2007.0) **58** 105-112. DOI: 10.1016/j.neures.2007.01.004
42. Madisen L. **A robust and high-throughput Cre reporting and characterization system for the whole mouse brain**. *Nat. Neurosci.* (2010.0) **13** 133-140. DOI: 10.1038/nn.2467
43. Zariwala HA. **A Cre-dependent GCaMP3 reporter mouse for neuronal imaging in vivo**. *J. Neurosci.* (2012.0) **32** 3131-3141. DOI: 10.1523/JNEUROSCI.4469-11.2012
44. Vong L. **Leptin action on GABAergic neurons prevents obesity and reduces inhibitory tone to POMC neurons**. *Neuron* (2011.0) **71** 142-154. DOI: 10.1016/j.neuron.2011.05.028
45. Bäckman CM. **Characterization of a mouse strain expressing Cre recombinase from the 3′ untranslated region of the dopamine transporter locus**. *Genesis* (2006.0) **44** 383-390. DOI: 10.1002/dvg.20228
46. du Sert NP. **The ARRIVE guidelines 2.0: Updated guidelines for reporting animal research**. *PLoS Biol* (2020.0) **18** e3000410. DOI: 10.1371/journal.pbio.3000410
47. Schindelin J. **Fiji: An open-source platform for biological-image analysis**. *Nat. Methods* (2012.0) **9** 676-682. DOI: 10.1038/nmeth.2019
48. 48.Van Rossum, G. & Drake, FL. Python 3 Reference Manual (CreateSpace, 2009).
49. Kane GA, Lopes G, Saunders JL, Mathis A, Mathis MW. **Real-time, low-latency closed-loop feedback using markerless posture tracking**. *Elife* (2020.0) **9** e61909. DOI: 10.7554/eLife.61909
50. Mathis A. **DeepLabCut: Markerless pose estimation of user-defined body parts with deep learning**. *Nat. Neurosci.* (2018.0) **21** 1281-1289. DOI: 10.1038/s41593-018-0209-y
51. Nath T. **Using DeepLabCut for 3D markerless pose estimation across species and behaviors**. *Nat. Protoc.* (2019.0) **14** 2152-2176. DOI: 10.1038/s41596-019-0176-0
|
---
title: Hormone replacement treatment regimen is associated with a higher risk of hypertensive
disorders of pregnancy in women undergoing frozen-thawed embryo transfer
authors:
- Lijuan Fan
- Na Li
- Xitong Liu
- Xiaofang Li
- He Cai
- Dan Pan
- Ting Wang
- Wenhao Shi
- Pengfei Qu
- Juanzi Shi
journal: Frontiers in Endocrinology
year: 2023
pmcid: PMC9998903
doi: 10.3389/fendo.2023.1133978
license: CC BY 4.0
---
# Hormone replacement treatment regimen is associated with a higher risk of hypertensive disorders of pregnancy in women undergoing frozen-thawed embryo transfer
## Abstract
### Introduction
In frozen-thawed embryo transfer (FET) cycles, hormone replacement treatment (HRT) was associated with a higher risk of hypertensive disorders of pregnancy (HDP) compared with natural cycles (NC). Multiple pregnancy was a risk factor for HDP and several studies did not conduct subgroup analysis of singleton pregnancy and multiple pregnancy.
### Objective
To investigate whether HRT regimen could be a risk factor for HDP in women undergoing FET cycles in singleton and twin pregnancies.
### Methods
A retrospective cohort study at a tertiary hospital, including a total of 9120 women who underwent FET and achieved ongoing pregnancy; 7590 patients underwent HRT-FET and 1530 NC-FET. The main outcome was HDP. HDP were analyzed for singleton and twin pregnancies, respectively.
### Results
In the singleton pregnancy, the risk of HDP in the HRT-FET group was significantly higher than that in the NC-FET group ($6.21\%$ vs. $4.09\%$; $$P \leq 0.003$$). After adjusting for female age oocyte pick up, female age at FET and body mass index (BMI), HRT was found as a risk factor for HDP (adjusted odds ration [aOR]: 1.43; $95\%$ confidence interval [CI]: 1.07 to 1.91; $$P \leq 0.017$$). In the multiple pregnancy, the risk of HDP in the HRT-FET and NC-FET groups was similar.
### Conclusion
HRT was associated with a higher risk of HDP in women who underwent FET and achieved singleton pregnancy.
## Introduction
The application of frozen-thawed embryo transfer (FET) has dramatically increased worldwide over the past decade for its merits. The FET enables the storage of excess embryos, reduces the incidence of ovarian hyperstimulation syndrome (OHSS), provides time required for preimplantation genetic testing (PGT), and facilitates fertility preservation. However, several studies have suggested that compared with fresh embryo transfer, the FET is associated with the reduced incidence rates of preterm birth, low birthweight (LBW), small for gestational age (SGA), and perinatal mortality [1]. However, the safety of FET is challenged by the increased incidence rates of large for gestational age (LGA) and hypertensive disorders of pregnancy (HDP) [2].
The commonly used endometrial preparation regimens for FET include natural cycle (NC) and hormone replacement treatment (HRT) cycle. Given that the HRT cycles rely on exogenous hormone supplement and lack of corpus luteum, this condition might be less ‘physiological’ than a natural ovulatory cycle. Recent studies reported that HRT-FET were associated with a higher risk of HDP compared with NC-FET. However, several studies did not conduct subgroup analysis of singleton pregnancy and multiple pregnancy. Moreover, multiple pregnancy was reported as a risk factor for HDP, indicating its importance. Besides, the effects of the endometrium preparation method on the outcomes of pregnancies conceived with FET have not been fully clarified.
The present study aimed to assess the effects of different FET regimens on the risk of HDP in women who underwent FET cycles, and the risk of HDP was analyzed in singleton pregnancy and multiple pregnancy. This retrospective cohort study was conducted to compare the risk of HDP between NC-FET and HRT-FET groups.
## Study design and patients
This was a retrospective cohort study, in which the in vitro fertilization (IVF) was performed from January 2018 to December 2020 in the Center for Assisted Reproductive Technology of Northwest Women’s and Children’s Hospital (Xi’an, China). The protocol of the study was approved by the institutional review board of the hospital. Data were extracted from electronic medical records. Patients who underwent FET and achieved ongoing pregnancy were enrolled. Ongoing pregnancy was defined as the presence of at least one fetal heart pulsation on ultrasound beyond 20 weeks. All patients were enrolled only once. Women with chronic hypertension before pregnancy were excluded. Written informed consent was obtained from the participants before treatment.
## Controlled ovarian stimulation and vitrified cryopreservation
Ovarian stimulation protocols included gonadotropin-releasing hormone (GnRH) agonist protocol, GnRH antagonist protocol, and progestin-primed ovarian stimulation (PPOS) protocol. Recombinant human chorionic gonadotropin (OVIDREL; Merck Serono, Darmstadt, Germany) or GnRH-a (Decapeptyl; Ferring, Saint-Prex, Switzerland) were administered in patients when two leading follicles reached 18 mm in diameter. Oocyte retrieval was performed at 36 h after recombinant human chorionic gonadotropin or GnRH-a triggered by transvaginal ultrasound-guided aspiration. Insemination method was selected according to the sperm count after sperm preparation. A morphologic score of cleavage-stage embryo was given based on the number of blastomeres, the homogeneous degree of blastomeres, and the degree of cytoplasmic fragmentation, which has been extensively described in our previous study [3]. If a couple has two or more high-quality cleavage-stage embryos on day 3 of embryo culture, the embryos were selected and cultured to blastocyst stage. Blastocyst evaluation was performed according to the Gardner’s grading system [4].
For patients who underwent GnRH agonist protocol and GnRH antagonist protocol, one to two fresh embryos were transferred into the uterus of women free of OHSS, hydrosalpinx, intrauterine adhesion and high progesterone level (> 1.5 ng/ml) on the day of triggering, and then, the spare embryos were cryopreserved for the next FET. Patients who underwent PPOS protocol had to freeze all their embryos. The vitrified cryopreservation was conducted according to standard protocols, as previously described [5].
## Endometrial preparation before FET
The selection of FET regimen is performed based on patients’ conditions, including menstrual regularity, ovulation regularity, doctors’ preference, endometrial development, and the prevalence of endometriosis and adenomyosis. For instance, patients with regular menstrual cycles and ovulation mainly undergo NC-FET. Patients with ovulation disorders or impaired endometrium development often undergo HRT-FET, because these patients have trouble in preparing the endometrium with natural ovulation. Meanwhile, HRT-FET is also selected due to the convenience of scheduling the date of FET. Patients with endometriosis, adenomyosis or recurrent implantation failure mainly undergo combination of GnRH-a and HRT-FET.
In this study, patients in the NC-FET group underwent transvaginal ultrasound on days 8 to 10 of the menstrual cycle. Follicular growth was monitored through transvaginal ultrasound and measurement of serum luteinizing hormone (LH). When the leading follicle had reached a mean diameter of >17 mm and the serum LH level was 20 IU/L, the transvaginal ultrasound was performed every day until ovulation. The day of ovulation was confirmed by transvaginal ultrasound. Cleavage-stage embryo and blastocyst-stage embryo were thawed and transferred on 3 and 5 days after ovulation, respectively.
For patients in the HRT-FET group, endometrial preparation was initiated with oral estradiol valerate (Progynova; Bayer, Berlin, Germany) at a daily dose of 4 mg from day 5 of menstrual cycle. For patients with impaired endometrial development, a daily maximum dose of 6 mg oral estradiol valerate and 3 mg transdermal 17-β estradiol (Besins Healthcare, Paris, France) were given. The serum progesterone level was measured and the transvaginal ultrasound was performed 10-12 days after the usage of exogenous estrogen. When the endometrial thickness reached 7 mm or more and the serum progesterone level was <1.5 ng/mL, exogenous progesterone was added. The FET was scheduled for 5 days for cleavage-stage embryos and for 7 days for blastocyst-stage embryos.
## Luteal support
Three methods of luteal support are implemented in our center. I. Vaginal progesterone gel (90 mg q.d; Crinone, Serono, Hertfordshire, UK); II. Vaginal progesterone soft capsules (0.2 g t.i.d; Utrogestan, Besins, France); III. Intramuscular progesterone (60 mg q.d; Xianju, Zhejiang, China). Patients from both groups could select one of these three luteal support methods and receive oral progesterone (10 mg t.i.d; Dydrogesterone, Abbott Biologicals B.V., Amsterdam, Netherlands) simultaneously. For patients who underwent HRT-FET, exogenous estrogen would be reduced after the confirmation of clinical pregnancy. The luteal support was maintained until week 10 of gestation.
## Definition of outcomes
The primary outcome was the risk of HDP. It was attempted to define HDP as sustained (on at least two occasions 6 h apart) blood pressure ≥ $\frac{140}{90}$ mmHg after 20 weeks, with or without proteinuria and other signs or symptoms of preeclampsia and without a history of hypertension. As secondary outcomes, we analyzed some other perinatal risks and neonatal risks. Other perinatal risks included gestational diabetes mellitus (GDM), placenta previa, premature rupture of membrane, anemia, miscarriage, and stillbirth. Miscarriage was defined as the spontaneous loss of clinical pregnancy before 28 weeks of gestational age. Stillbirth was defined as the absence of signs of life at or after 28 weeks of gestation. Neonatal risks included preterm birth (<37 weeks’ gestation), extremely preterm birth (<32 weeks’ gestation), low birth weight (<2500 g), very low birth weight (<1500 g), and macrosomia (birth weight ≥4000 g) [6].
## Statistical analysis
Categorical variables were presented as count and proportion; normally distributed continuous variables were expressed as the mean and standard deviation, and abnormally distributed continuous variables were presented as the median and interquartile range (IQR). The Chi-squared test or the Fisher’s exact test were utilized to compare the categorical variables. The Student’s t-test was used to compare the continuous variables. The effects of different FET regimens on the risk of HDP and other outcomes were estimated using the generalized linear model adjusted for female age at oocyte pick up (OPU), female age at FET and body mass index (BMI). To assess the influence of potential heterogeneity on the risk of HDP, the effects of different FET regimens were estimated in several subgroups. All data were analyzed using the SPSS 22.0 software (IBM Corp., Armonk, NY, USA). The level of significance was set at $P \leq 0.05.$
## Baseline characteristics of patients
A total of 9120 patients who fulfilled the inclusion and exclusion criteria were included in this study (Figure 1). Of these, 7590 and 1530 patients were in HRT-FET and NC-FET groups, respectively. The baseline characteristics of patients are shown in Table 1. Women who underwent HRT-FET were significantly younger at OPU (30.11 ± 3.90 vs. 30.40 ± 3.77, $P \leq 0.001$) and FET (30.76 ± 3.91 vs. 31.37 ± 3.68, $P \leq 0.001$) compared with those in the NC-FET group. Husbands of women in the HRT-FET group were significantly younger than those of women in the NC-FET group. Besides, a significantly higher BMI, a greater antral follicle count (AFC), a higher number of nulliparous women, and higher incidence rates of adenomyosis and endometriosis were detected in the HRT-FET group compared with the NC-FET group. After ovarian stimulation, the number of oocytes retrieved in the HRT-FET group was significantly greater than that in the NC-FET group. More patients in the HRT-FET group froze all their embryos compared with the NC-FET group. The endometrial thickness in the HRT-FET group was significantly thinner than that in the NC-FET group. In terms of the number of embryos transferred, it was higher in the HRT-FET group compared with that in the NC-FET group. The proportion of D3 cleavage-stage embryo transfer was significantly higher in the HRT-FET group compared with that in the NC-FET group. There were no significant differences in infertility duration, infertility type, proportion of patients with uterine cavity malformation, proportion of patients undergoing PGT, insemination type, and proportion of patients transferred at least one high-quality embryo between the two groups.
**Figure 1:** *Patient selection flowchart.* TABLE_PLACEHOLDER:Table 1
## Different FET regimens and the risk of HDP
The maternal and neonatal outcomes were categorized by the type of FET regimen. The risk of HDP was regarded as the primary outcome. Perinatal outcomes were analyzed in singleton pregnancy (Table 2) and twin pregnancy (Table 3), respectively. In the singleton pregnancy, the risk of HDP in the HRT-FET group was significantly higher than that in the NC-FET group ($6.21\%$ vs. $4.09\%$, $$P \leq 0.003$$). After adjusting for confounders, including female age at OPU, female age at FET and BMI, the HRT-FET group was associated with a higher risk of HDP compared with the NC-FET group in singleton pregnancy (adjusted odds ratio (aOR): 1.43; $95\%$ confidence interval (CI): 1.07 to 1.91; $$P \leq 0.017$$). In the twin pregnancy, the risk of HDP was similar between the HRT-FET and NC-FET groups ($10.48\%$ vs. $6.49\%$, $$P \leq 0.091$$).
Endometrial preparation protocols are commonly categorized into two categories: with and without corpus luteum protocols. HRT protocols combined with or without GnRH-a were classified into artificial preparation without corpus luteum, while with exogenous steroid. For patients with ovulation disorders, monitoring of ovarian follicular development is particularly troublesome. Preparing the endometrium with exogenous hormones is associated with some advantages, such as monitoring and scheduling of the timing of the procedure, making it more convenient and simpler. Therefore, HRT-FET cycle accounted for the majority of our FET cycles.
Hypertensive disorders were reported in $5.9\%$ of assisted reproductive technology (ART) singleton and $12.6\%$ of ART twin pregnancies [7]. Multiple pregnancy was reported as a risk factor for HDP. Previous relevant studies have mainly compared the risk of HDP in NC and HRT groups without separate analysis of singleton pregnancy and multiple pregnancy. Therefore, we analyzed and reported the risk of HDP in singleton pregnancy and multiple pregnancy, respectively.
In our large retrospective cohort study of over 9120 FET cycles, the risk of HDP in the HRT-FET group was risen by $2.12\%$ in the singleton pregnancy compared with that in the NC-FET group. In the multiple pregnancy, the risk of HDP in the HRT-FET group was escalated by $3.99\%$ compared with that in the NC-FET group. This finding is similar to previously reported outcome (2, 8–10). Our results strengthened the pivotal etiopathogenic role of the corpus luteum. In HRT-FET cycles, the absence of corpus luteum was inevitably resulted in the lack of circulating vasoactive factors, such as relaxin [11], prorenin, and renin [12]. These circulating vasoactive factors were required for maternal cardiovascular adaptation during the first trimester of pregnancy. Then, the lack of these circulating vasoactive factors may lead to the disruption of maternal circulatory adaptation. Von Versen-Höynck [11, 13] reported the decline of carotid-femoral pulse-wave velocity (cfPWV) and the increase of femoral pulse-wave transit time (fPWTT) during the first trimester in pregnant women who underwent HRT-FET compared with those who underwent NC-FET and fresh transfer.
Chen et al. [ 14] reported a significantly higher (two times) risk of preeclampsia in HRT-FET cycles compared with that in fresh transfer cycles in women who were diagnosed with polycystic ovary syndrome. The possible reason is that the supra-physiologic number of corpora lutea in fresh transfer cycles could produce the circulating vasoactive factors and protect pregnancies against HDP.
Another possible mechanism underlining the increased risk of HDP after HRT-FET may be attributed to the prematurely elevated level of estradiol. A low estradiol level during implantation allows for extravillous trophoblasts into uterine spiral arteries with vascular remodeling, and elevation of estradiol level later prevents further remodeling [15]. In HRT-FET cycles, estradiol level is elevated more prematurely than that in NC-FET cycles. Trophoblastic invasion of spiral arteries may be suppressed by the prematurely elevated estradiol level [16]. A retrospective study reported a lower risk of HDP in letrozole-induced FET cycles than that in HRT-FET cycles [17]. The administration of letrozole during modified NC-FET cycles might potentially lower estradiol rise and optimal extravillous trophoblasts into uterine spiral arteries with vascular remodeling.
For China the average rate for cesarean section was 54.9 percent in 2014 [18]. Higher rates of cesarean section were reported among women who get pregnant after ART. Meanwhile, patients present multiple pregnancy are prone to have greater need for cesarean section. These may be the reasons for the high rate of cesarean section in this study.
## Different FET regimens and other perinatal outcomes
In the singleton pregnancy (Table 2), a higher cesarean section rate ($78.47\%$ vs. $68.71\%$, $P \leq 0.001$) and a shorter gestational age at birth (271.91 ± 12.58 vs. 272.85 ± 10.00, $$P \leq 0.011$$) were found in the HRT-FET group compared with the NC-FET group. This association remained essentially unchanged after adjusting for female age at oocyte retrieval, female age at FET and BMI (aOR: 1.70; $95\%$ CI: 1.49 to 1.95; $P \leq 0.001$) (aOR: −0.12; $95\%$ CI: −0.23 to −0.02; $$P \leq 0.018$$). A lower rate of placenta previa was detected in the NC-FET group compared with that in the HRT-FET group ($0.63\%$ vs. $0.67\%$), and this association was not significant after adjusting for covariates (aOR: 0.68; $95\%$ CI: 0.32 to 1.47; $$P \leq 0.327$$). Other perinatal outcomes, including the rates of GDM, premature rupture of membranes, anemia, miscarriage, and stillbirth were similar between the two groups in the singleton pregnancy (P≥0.05).
In the multiple pregnancy (Table 3), a lower rate of anemia ($0.71\%$ vs. $3.24\%$, $$P \leq 0.002$$) and a shorter gestational age at birth (251.20 ± 14.69 vs. 253.80 ± 12.62, $$P \leq 0.027$$) were identified in the HRT-FET group compared with those in the NC-FET group. This association remained essentially unchanged after adjusting for female age at oocyte retrieval, female age at FET and BMI (aOR: 0.19; $95\%$ CI: 0.09 to 0.55; $$P \leq 0.002$$) (aOR: −2.44; $95\%$ CI: −4.74 to −0.13; $$P \leq 0.039$$). Other perinatal outcomes, including the rates of GDM, placenta previa, premature rupture of membranes, miscarriage, stillbirth, and method of delivery were similar between the two groups in the multiple pregnancy (P≥0.05).
## Different FET regimens and neonatal outcomes
In the singleton pregnancy (Table 2), the rates of extremely preterm birth ($1.18\%$ vs. $0.38\%$, $$P \leq 0.010$$) and preterm birth ($9.62\%$ vs. $6.51\%$, $P \leq 0.001$) were significantly higher in the HRT-FET group than those in the NC-FET group. This association remained essentially unchanged after adjusting for female age at oocyte retrieval, female age at FET and BMI (aOR: 2.97; $95\%$ CI: 1.19 to 7.39; $$P \leq 0019$$) (aOR: 1.48; $95\%$ CI: 1.17 to 1.88; $$P \leq 0.001$$). Meanwhile, the birth height in the HRT-FET group (47.58 ± 3.06 cm) was significantly shorter than that in the NC-FET group (48.02 ± 2.43 cm, $$P \leq 0.011$$). This association remained essentially unchanged after adjustment (aOR: −0.013; $95\%$ CI: −0.24 to −0.03; $$P \leq 0.013$$). Low birth weight rate was significantly higher in the HRT-FET group compared with that in the NC-FET group ($5.08\%$ vs. $3.68\%$, $$P \leq 0.033$$), and this association was not significant after adjusting for female age at OPU, female age at FET and BMI (aOR: 1.35; $95\%$ CI: 0.99 to 1.84; $$P \leq 0.062$$). Other neonatal outcomes, including birth weight, macrosomia rate, and very low birth weight rate were similar between the two groups in the singleton pregnancy (P≥0.05).
In the multiple pregnancy (Table 3), the birth weight in the HRT-FET group was significantly lower (2.47 ± 0.48 vs. 2.52 ± 0.42 kg, $$P \leq 0.044$$), and birth height (47.58 ± 3.06 vs. 48.02 ± 2.42 cm, $$P \leq 0.011$$) was significantly shorter compared with the NC-FET group. The rate of very low birth weight ($3.97\%$ vs. $1.44\%$ $$P \leq 0.019$$) was significantly higher in the HRT group. These associations remained essentially unchanged after adjustment (aOR: 2.76; $95\%$ CI: 1.11 to 6.87; $$P \leq 0.029$$). There was no macrosomia in both groups. Other neonatal outcomes, including extremely preterm birth rate, preterm birth rate, birth weight, and low birth weight rate were similar between the two groups in the multiple pregnancy (P≥0.05).
## Subgroup analysis
The effects of two different FET regimens on the risk of HDP were analyzed in different subgroups (Table 4; Supplementary Figure 1). It was revealed that HRT-FET was associated with a higher risk of HDP in all the subgroups of women with different infertility durations, whether they attempted to freeze all their embryos or not and the number of embryos transferred, in which these associations were statistically significant ($P \leq 0.05$). For the other subgroups, HRT-FET was associated with a higher risk of HDP compared with NC-FET. Meanwhile, those associations were statistically significant for the subgroups of female age at OPU (≤ 30, between 31 and 35), female age at FET (≤ 30, between 31 and 35), male age (≤ 35), female BMI (< 23 kg/m2), nulliparity, number of oocytes retrieved (< 10), insemination method (IVF), transfer of ≥1 high-quality embryo, and transfer of blastocyst-stage embryo ($P \leq 0.05$).
**Table 4**
| Unnamed: 0 | Subgroup | NC-FET | HRT-FET | OR | 95%CI | P value |
| --- | --- | --- | --- | --- | --- | --- |
| Female age at OPU (y) | ≤30 | 629 (41.11) | 3875 (51.05) | 1.7 | 1.17, 2.47 | 0.005 |
| | 31–35 | 702 (45.88) | 2819 (37.14) | 1.72 | 1.12, 2.65 | 0.013 |
| | ≥36 | 199 (13.01) | 896 (11.81) | 1.16 | 0.58, 2.32 | 0.684 |
| Female age at FET (y) | ≤30 | 806 (52.68) | 4438(58.47) | 1.63 | 1.08, 2.47 | 0.020 |
| | 31–35 | 574 (37.52) | 2431 (32.03) | 1.91 | 1.27, 2.86 | 0.002 |
| | ≥36 | 150 (9.80) | 721 (9.50) | 1.08 | 0.58, 2.01 | 0.802 |
| Male age (y) | <35 | 1251 (81.76) | 6263(82.52) | 1.83 | 1.33, 2.48 | <0.001 |
| | 35–40 | 211(13.79) | 959 (12.64) | 1.1 | 0.64, 1.90 | 0.713 |
| | >40 | 68 (4.44) | 368(4.85) | 1.74 | 0.57, 4.76 | 0.361 |
| BMI (kg/m2) | <23 | 1031(67.39) | 4651 (61.28) | 1.53 | 1.07, 2.20 | 0.021 |
| | 23 –25 | 280 (18.30) | 1398 (18.42) | 1.8 | 0.95, 3.41 | 0.070 |
| | >25 | 219 (14.31) | 1541 (20.30) | 1.3 | 0.81, 2.09 | 0.276 |
| Infertility duration, year | ≤3 | 959 (62.68) | 4614 (60.79) | 1.46 | 1.04, 2.05 | 0.029 |
| | >3 | 571 (37.31) | 2976 (39.21) | 1.9 | 1.25, 2.87 | 0.003 |
| Parity | | 1298 (84.84) | 6626(87.30) | 1.69 | 1.27, 2.25 | <0.001 |
| | High order | 232 (15.16) | 964 (12.70) | 1.5 | 0.76, 2.97 | 0.246 |
| No. of oocytes retrieved | <10 | 673 (44.51) | 2831 (37.67) | 1.89 | 1.25, 2.85 | 0.001 |
| | 10–14 | 446 (29.50) | 2105 (28.01) | 1.83 | 1.08, 3.11 | 0.155 |
| | >14 | 393 (25.99) | 2579 (34.32) | 1.25 | 0.81, 1.94 | 0.218 |
| Insemination type | IVF | 1121 (73.65) | 5595 (74.12) | 1.9 | 1.37, 2.64 | <0.001 |
| | ICSI | 401 (26.35) | 1954 (25.88) | 1.16 | 0.76, 1.79 | 0.489 |
| Freeze all (%) | No | 615 (40.22) | 2109 (27.79) | 2.04 | 1.31, 3.16 | 0.002 |
| | Yes | 914 (59.78) | 5481 (72.21) | 1.44 | 1.04, 1.99 | 0.027 |
| Endometrial thickness (mm) | <9 | 673 (43.99) | 2831 (37.30) | 1.89 | 1.25, 2.85 | <0.001 |
| | 9–12 | 464 (30.33) | 2180 (28.72) | 1.83 | 1.08, 3.11 | 0.155 |
| | >12 | 393 (25.69) | 2579 (33.98) | 1.25 | 0.81, 1.94 | 0.218 |
| No. of embryos transferred, n | 1 | 1012 (66.14) | 4063 (53.53) | 1.53 | 1.11, 2.11 | 0.010 |
| | ≥2 | 518 (33.86) | 3527 (46.47) | 1.76 | 1.13, 2.74 | 0.013 |
| Good quality embryo transfer | | 355 (23.20) | 1858 (24.48) | 1.3 | 0.79, 2.13 | 0.304 |
| | ≥1 high quality embryo | 1175 (76.80) | 5732 (75.52) | 1.75 | 1.29, 2.38 | <0.001 |
| Type of embryo | Cleavage-stage | 279 (18.24) | 1818 (23.95) | 1.74 | 0.92, 3.26 | 0.086 |
| | Blastocyst-stage | 1251 (81.76) | 5772 (76.05) | 1.61 | 1.21, 2.14 | 0.002 |
## Strengths and limitations
The major strength of this study is the large cohort size from a single-center, in which practice consistency can be assured. Controlled ovarian stimulation, IVF protocols, and laboratory conditions remained homogeneous. Additionally, maternal and neonatal outcomes of singleton and multiple pregnancies were analyzed separately, because multiple pregnancy was reported as a risk factor for HDP [19]. Similar to multiple pregnancy, chronic hypertension, BMI>30 kg/m2, and female age were previously found as risk factors for HDP (20–22). We excluded women with chronic hypertension. Meanwhile, we adjusted female age at OPU, female age at FET and BMI in the assessment of the effects of different FET regimens on the risk of HDP. These analyses made our results more reliable.
Our study had several limitations. Firstly, this was a single-center retrospective study, in which inherent bias was inevitable. Regarding this deficiency, we screened patients with strict criteria.
Secondly, endometrial preparation was not randomly assigned in our study population, and imbalance in the number of enrolled patients in the two groups should be noted. However, ovulatory dysfunction was noted as the main indicator to pick HRT-FET rather than NC-FET. Ovulatory dysfunction was identified in approximately $15\%$ of all infertile couples and accounted for up to $40\%$ of female infertility [23]. In our study, the number of patients in the HRT-FET group was 3.96 times greater than that in the NC-FET group. The main advantages of HRT-FET were its convenience, low-cost, and simplicity. Patients who underwent HRT-FET only need to visit the doctor for 2-3 times during their endometrial preparation period. Nevertheless, the imbalance in the number of study subjects could also be found in a previous study [24].
Thirdly, we failed in exact classification of patients into HDP categories after telephone follow-up. It is noteworthy that gestational hypertension and preeclampsia are involved in HDP [25]. We followed up patients by telephone at one month after their expected date of delivery, thus, recall bias might be existed. Besides, some patients could not remember the exact diagnosis of HDP.
## Conclusion
In conclusion, HRT was found to be associated with a higher risk of HDP in women who underwent FET and achieved singleton pregnancy. However, further large-scale, prospective, randomized controlled trials with a longer follow-up are required to verify the increased risk of HDP in women undergoing HRT-FET.
## Data availability statement
The original contributions presented in the study are included in the article/ Supplementary Material. Further inquiries can be directed to the corresponding author.
## Ethics statement
The studies involving human participants were reviewed and approved by Northwest Women’s and Children’s Hospital. The patients/participants provided their written informed consent to participate in this study.
## Author contributions
LF and NL: designed study, drafted the manuscript and reviewed the manuscript. XitL, XiaL, HC, DP, TW, WS and PQ: analyzed data. JS: Study conceptualization. All authors contributed to the article and approved the submitted version.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fendo.2023.1133978/full#supplementary-material
## References
1. Trounson A, Mohr L. **Human pregnancy following cryopreservation, thawing and transfer of an eight-cell embryo**. *Nature* (1983) **305**. DOI: 10.1038/305707a0
2. Saito K, Kuwahara A, Ishikawa T, Saito H, Morisaki N, Miyado M. **Endometrial preparation methods for frozen-thawed embryo transfer are associated with altered risks of hypertensive disorders of pregnancy, placenta accreta, and gestational diabetes mellitus**. *Hum Reprod* (2019) **34**. DOI: 10.1093/humrep/dez079
3. Li M, Wang H, Ma C, Shi J. **Transferring two grades I cleavage-stage embryo might not be a good protocol**. *Gynecol Endocrinol* (2017) **33**. DOI: 10.1080/09513590.2017.1302420
4. Gardner DK, Schoolcraft WB. **Culture and transfer of human blastocysts**. *Curr Opin Obstet Gynecol* (1999) **11**. DOI: 10.1097/00001703-199906000-00013
5. Shi W, Xue X, Zhang S, Zhao W, Liu S, Zhou H. **Perinatal and neonatal outcomes of 494 babies delivered from 972 vitrified embryo transfers**. *Fertil Steril* (2012) **97**. DOI: 10.1016/j.fertnstert.2012.02.051
6. Brown CC, Moore JE, Felix HC, Tilford JM, Bird TM, Lowery CL. **Association of state medicaid expansion status with low birth weight and preterm birth**. *JAMA* (2019) **321**. DOI: 10.1001/jama.2019.3678
7. Opdahl S, Henningsen AA, Tiitinen A, Bergh C, Pinborg A, Romundstad PR. **Risk of hypertensive disorders in pregnancies following assisted reproductive technology: a cohort study from the CoNARTaS group**. *Hum Reprod* (2015) **30**. DOI: 10.1093/humrep/dev090
8. Wang B, Zhang J, Zhu Q, Wang Y. **Effects of different cycle regimens for frozen embryo transfer on perinatal outcomes of singletons**. *Hum Reprod* (2020) **35**. DOI: 10.1093/humrep/deaa093
9. Asserhøj LL, Spangmose AL, Aaris Henningsen AK, Pinborg A. **Adverse obstetric and perinatal outcomes in 1,136 singleton pregnancies conceived after programmed frozen embryo transfer (FET) compared with natural cycle FET**. *Fertil Steril* (2021) **115**. DOI: 10.1016/j.fertnstert.2020.10.039
10. Zaat TR, Brink AJ, de Bruin JP, Mol F. **Increased obstetric and neonatal risks in artificial cycles for frozen embryo transfers**. *Reprod BioMed Online* (2021) **42**. DOI: 10.1016/j.rbmo.2021.01.015
11. von Versen-Ho¨ynck F, Schaub AM, Chi Y-Y, Chiu K-H, Liu J, Lingis M. **Increased preeclampsia risk and reduced aortic compliance with**. *Hypertension* (2019) **73**. DOI: 10.1161/HYPERTENSIONAHA.118.12043
12. Wiegel RE, Jan Danser AH, Steegers-Theunissen RPM, Laven JSE, Willemsen SP, Baker VL. **Determinants of maternal renin-angiotensin-aldosterone-system activation in early pregnancy: insights from 2 cohorts**. *J Clin Endocrinol Metab* (2020) **105**. DOI: 10.1210/clinem/dgaa582
13. von Versen-Höynck F, Narasimhan P, Selamet Tierney ES, Martinez N, Conrad KP, Baker VL. **Absent or excessive corpus luteum number is associated with altered maternal vascular health in early pregnancy**. *Hypertention* (2019) **73**. DOI: 10.1161/HYPERTENSIONAHA.118.12046
14. Chen ZJ, Shi Y, Sun Y, Zhang B, Liang X, Cao Y. **Fresh versus frozen embryos for infertility in the polycystic ovary syndrome**. *New Engl J Med* (2016) **375**. DOI: 10.1056/NEJMoa1513873
15. Bonagura TW, Pepe GJ, Enders AC, Albrecht ED. **Suppression of extravillous trophoblast vascular endothelial growth factor expression and uterine spiral artery invasion by estrogen during early baboon pregnancy**. *Endocrinology* (2008) **149**. DOI: 10.1210/en.2008-0116
16. Bonagura TW, Babischkin JS, Aberdeen GW, Albrecht ED. **Prematurely elevating estradiol in early baboon pregnancy suppresses uterine artery remodeling and expression of extravillous placental vascular endothelial growth factor and a1b1 and a5b1 integrins**. *Endocrinology* (2012) **153**. DOI: 10.1210/en.2012-1141
17. Zhang J, Wei MJ, Bian XJ, Wu L, Zhang S, Mao XY. **Letrozole-induced frozen embryo transfer cycles are associated with a lower risk of hypertensive disorders of pregnancy among women with polycystic ovary syndrome**. *Am J Obstet Gynecol* (2021) **225**. DOI: 10.1016/j.ajog.2021.01.024
18. Liu Y, Li G, Chen Y, Wang X, Ruan Y, Zou L. **A descriptive analysis of the indications for caesarean section in mainland China**. *BMC Pregnancy Childbirth* (2014) **14**. DOI: 10.1186/s12884-014-0410-2
19. Brown MA, Magee LA, Kenny LC, Karumanchi SA, McCarthy FP, Saito S. **The hypertensive disorders of pregnancy: ISSHP classification, diagnosis & management recommendations for international practice**. *Pregnancy Hypertens* (2018) **13** 291-310. DOI: 10.1016/j.preghy.2018.05.004
20. Bartsch E, Medcalf KE, Park AL, Ray JG. **Clinical risk factors for pre-eclampsia determined in early pregnancy: systematic review and meta-analysis of large cohort studies**. *BMJ* (2016) **353**. DOI: 10.1136/bmj.i1753
21. Kawakita T, Downs SK, Franco S, Ghofranian A, Thomas A. **Interpregnancy body mass index change and risk of hypertensive disorders in pregnancy**. *J Matern-Fetal Neo M* (2022) **35**. DOI: 10.1080/14767058.2020.1817371
22. Choi H, Lim JY, Lim NK, Ryu HM, Kwak DW, Chung JH. **Impact of pre-pregnancy body mass index and gestational weight gain on the risk of maternal and infant pregnancy complications in Korean women**. *Int J Obes* (2022) **46** 59-67. DOI: 10.1038/s41366-021-00946-8
23. Mosher WD, Pratt WF. **Fecundity and infertility in the united states: incidence and trends**. *Fertil Steril* (1991) **56**. DOI: 10.1016/S0015-0282(16)54469-0
24. Kaponis A, Chatzopoulos G, Paschopoulos M, Georgiou I, Paraskevaidis V, Zikopoulos K. **Ultralong administration of gonadotropin-releasing hormone agonists before**. *Fertil Steril* (2020) **113**. DOI: 10.1016/j.fertnstert.2019.12.018
25. Behrens I, Basit S, Lykke JA, Ranthe MF, Wohlfahrt J, Boyd HA. **Association between hypertensive disorders of pregnancy and later risk of cardiomyopathy**. *JAMA-J Am Med Assoc* (2016) **315**. DOI: 10.1001/jama.2016.1869
|
---
title: ID3 mediates BMP2-induced downregulation of ICAM1 expression in human endometiral
stromal cells and decidual cells
authors:
- Jin Luo
- Yaqin Wang
- Hsun-Ming Chang
- Hua Zhu
- Jing Yang
- Peter C. K. Leung
journal: Frontiers in Cell and Developmental Biology
year: 2023
pmcid: PMC9998904
doi: 10.3389/fcell.2023.1090593
license: CC BY 4.0
---
# ID3 mediates BMP2-induced downregulation of ICAM1 expression in human endometiral stromal cells and decidual cells
## Abstract
Recurrent pregnancy loss (RPL) remains an unsolved problem in obstetrics and gynecology, and up to $50\%$ of RPL cases are unexplained. Unexplained RPL (uRPL) is widely considered to be related to an aberrant endometrial microenvironment. BMP2 is an important factor involved in endometrial decidualization and embryo implantation, and intercellular adhesion molecule 1 (ICAM1) is a critical inflammatory regulator in the endometrium. In this study, we found that endometrial samples obtained from Unexplained RPL patients have significantly lower BMP2 and higher ICAM1 levels than fertile controls. For further research on the relationship between BMP2 and ICAM1 and the potential molecular mechanisms in Unexplained RPL, immortalized human endometrial stromal cells (HESCs) and primary human decidual stromal cells (HDSCs) were used as study models. Our results showed that BMP2 significantly decreased ICAM1 expression by upregulating DNA-binding protein inhibitor 3 (ID3) in both HESCs and HDSCs. Using kinase receptor inhibitors (dorsomorphin homolog 1 (DMH-1) and dorsomorphin) and siRNA transfection, it has been found that the upregulation of ID3 and the following downregulation of ICAM1 induced by BMP2 is regulated through the ALK3-SMAD4 signaling pathway. This research gives a hint of a novel mechanism by which BMP2 regulates ICAM1 in the human endometrium, which provides insights into potential therapeutics for unexplained RPL.
## Introduction
Recurrent pregnancy loss (RPL), which is defined as the loss of two or more pregnancies before 24 weeks of gestation in the same couple based on the European Society of Human Reproduction and Embryology (ESHRE) guideline (ESHRE Guideline Group on RPL et al., 2018), affects approximately $5\%$ of couples trying to conceive (Garrido-Gimenez and Alijotas-Reig, 2015). Although various etiological factors have been shown to lead to RPL (El Hachem et al., 2017), nearly $50\%$ of RPL cases cannot find identifiable causes or risk factors, which are classified as unexplained RPL (uRPL) (Garrido-Gimenez and Alijotas-Reig, 2015; El Hachem et al., 2017). The profound negative impact of RPL on both physical and mental health combined with the uncertainties in regard to the etiology and management options makes uRPL an ongoing challenge for both the clinical and scientific community. Therefore, exploring the pathogenesis of uRPL is essential for developing early interventions and ensuring a better pregnancy outcome.
As a unique tissue that indirectly contacts with the embryo, the endometrium plays a vital role in the establishment and maintenance of pregnancy. In recent decades, extensive research has been carried out to elucidate the biomolecular mechanisms that promote the acceptance of the embryo by the endometrium and the specific molecules and cellular pathways involved in endometrial receptivity. A number of studies have shown that various factors are dysregulated in the endometrium of women with uRPL (Comba et al., 2015; Dambaeva et al., 2021; Heidari et al., 2021; Zhu et al., 2022), including intercellular adhesion molecule 1 (ICAM1). ICAM1 is a single-chain, 90 kDa inducible cell-surface glycoprotein and a member of the immunoglobulin superfamily (Rothlein et al., 1991). In the human endometrium, ICAM1 is localized to the apical surface of the glandular epithelium, the vascular endothelium, and endometrial stromal cells throughout the menstrual cycle, and its expression in stromal cells is upregulated during menstruation (Thomson et al., 1999). The expression of ICAM1 has also been found in first-trimester human decidual stromal cells (Marzusch et al., 1993). Studies have found that inappropriate expression of ICAM1 may contribute to various gynecological and obstetric disorders, including endometriosis (Pino et al., 2009), gestational diabetes mellitus (Xie et al., 2008), preeclampsia (Austgulen et al., 1997) and RPL (Comba et al., 2015). The wide distribution of ICAM1 and these findings indicate that ICAM1 is involved in the menstrual process, glands, blood vessels and stroma function in the human endometrium and plays a crucial role in a successful pregnancy. However, the regulatory mechanism of ICAM1 during the menstrual cycle and early pregnancy in the human endometrium and decidua is largely unknown.
Bone formation protein 2 (BMP2) is a member of the transforming growth factor β (TGF-β) superfamily which serves as a key regulator of both endometrial degeneration (Li et al., 2007; Stoikos et al., 2008) and trophoblast cell invasion (Zhao et al., 2018a; Zhao et al., 2018b; Zhao et al., 2020). Functionally, Sma- and Mad-related (SMAD) proteins SMAD$\frac{1}{5}$/8 are phosphorylated when BMP2 binds to the TGF-β type II receptor and recruits the TGF-β type I receptors (ALK2, ALK3 and ALK6). After associating with the SMAD4 protein, the activated SMAD$\frac{1}{5}$/8 complex migrates to the nucleus wherein it regulates the target genes expression (Zhao et al., 2018a; Zhang et al., 2020; Luo et al., 2021a; Luo et al., 2021b). BMP2 is significantly elevated during decidualization in immortalized human endometrial stromal cells (HESCs) and primary human endometrial stromal cells (HDSCs) in response to steroid hormones and cyclic adenosine monophosphate (cAMP) (Luo et al., 2020). In addition, exogenous BMP2 treatment promotes the decidual response of these two kinds of cells (Luo et al., 2020). Our previous studies also demonstrated that BMP2 plays an essential role in endometrial stromal remodeling (Luo et al., 2020). However, the expression of BMP2 in the endometrium of patients with RPL has not been reported. We hypothesize that the overexpression of ICAM1 in the endometrium of uRPL patients may be controlled by BMP2 given the spatiotemporal variations in the expression of BMP2 and ICAM1 in the human endometrium throughout the menstrual cycle and pregnancy. To test this hypothesis, we analyzed the endometrial BMP2 and ICAM1 expression levels between uRPL patients and healthy women and explored the underlying molecular mechanisms and signaling pathways using HESCs and primary HDSCs.
## Patient recruitment and tissue collection
The use of endometrial tissue in the research received clearance from the ethics committee of Renmin Hospital, Wuhan University. A total of 16 women diagnosed with uRPL and 12 normal fertile women were recruited. The inclusion and exclusion criteria refer to previously published literature (Comba et al., 2015; Benner et al., 2022). Briefly, uRPL was defined as two or more fetal losses before 24 weeks of gestation without known causes of miscarriages. The control group was made up of normal fertile women with regular periods who had had at least one live birth and no spontaneous miscarriages in the past. The exclusion criteria were the use of immunosuppressive drugs, steroid hormones, antibiotics, diabetes mellitus and smoking. Endometrial biopsies were obtained from women who attended the reproductive center of Renmin Hospital of Wuhan University and received a endometrial biopsy on day 21 or day 22 of menstrual cycle which were identified as mid-secretory phase by pathological examination. In order to isolate primary HDSCs, first-trimester decidual specimens (between the 7th and 12th weeks of gestation) were collected from healthy women who were having an elective abortion as part of the CARE Program at the BC Women’s Hospital and Health Centre. The research was authorized by the University of British Columbia’s Research Ethics Board. All participants in this study were between 20 and 40 years of age and provided written informed consent.
## Cell models
Considering that both BMP2 and ICAM1 are expressed in human endometrial stromal cells throughout the menstrual cycle, and endometrial decidualization is a critical physiological event in the female menstrual cycle, thus immortalized human endometrial stromal cells (HESCs; ATCC® CRL-4003) and primary HDSCs were used as study cell models, representing non-decidual and decidual stromal cells respectively, to systematically study the regulatory effect of BMP2 on ICAM1 in human endometrium at different stages of the menstrual cycle.
## Culture and treatment of immortalized HESCs
HESCs were grown in DMEM/F12 medium without phenol red (Sigma-Aldrich, St. Louis, MO, United States of America), added with $10\%$ charcoal dextran-treated fetal bovine serum (FBS; HyClone Laboratories, Inc., Logan, UT, United States of America), $1\%$ ITS - Premix (BD Biosciences, San Jose, CA, United States of America) and 5 ng/mL puromycin (Thermo Fisher Scientific, Ottawa, ON, CAN). Every culture was kept at 37 °C in an incubator with $5\%$ CO2. HESCs were cultivated for a day after being seeded at a density of 4 × 105 cells per plate in 60-mm tissue culture dishes with full culture media. HESCs were treated with BMP2 (0, 10, 25, or 50 ng/ml) for the concentration-dependent research or with 25 ng/mL BMP2 for the time-course study after serum deprivation in DMEM/F12 media without FBS for 18 h. For the concentration-dependent investigation, cells were taken at 24 h, while for the time-course study, cells were taken at 3, 6, 12, 24 h, and 48 h.
## Isolation and cultivation of primary HDSCs
Primary HDSCs were separated from decidual tissues by means of enzymatic dispersion and mechanical dissociation, as mentioned before (Zhu et al., 2007). Generally, the samples were washed in cold phosphate-buffered saline (Gibco, Life Technologies, Inc., Carlsbad, CA, United States of America) three times, and then minced and treated with $0.1\%$ collagenase (type IV; Sigma‒Aldrich), $0.1\%$ hyaluronidase (type I-S; Sigma‒Aldrich) and 0.5 mg/ml DNase I (Sigma‒Aldrich) and subsequently digested in a shaking water bath for 60 min at 37°C. The supernatant was neutralized by the addition of phenol red-free DMEM/F12 medium supplemented with $10\%$ FBS before the cells were passed through a 40 m nylon filter (BD Biosciences, Bedford, UK). The undigested tissue fragments were left on the filter, and the stromal cell-containing eluate was transferred into a 50 ml tube. The cells were then pelleted by centrifuging them at 1200 g for 3 min at room temperature. Following that, the cell pellets were washed, resuspended, and seeded in phenol red-free DMEM/F12 media with antibiotics (100 U/ml penicillin and 100 μg/ml streptomycin, Life Technologies, Inc.), $10\%$ FBS, 30 nM 17β estradiol (E2; Sigma Aldrich), and 1 μM progesterone (P4; Sigma Aldrich). All decidual stromal cell cultures were afterwards maintained at 37°C in a humid incubator with $5\%$ CO2, unless otherwise stated, in this culture medium. The purity of the HDSCs was determined by immunofluorescent staining for vimentin and cytokeratin-7 as described previously (Zhu et al., 2007). HDSCs were cultivated at a density of 5 × 105 cells per plate in 60-mm tissue culture dishes for the time- and concentration-dependent studies, and BMP2 was added in the same manner as HESCs.
## Antibodies and reagents
Recombinant human BMP2, dorsomorphin homolog 1 (DMH-1) and dorsomorphin dihydrochloride (DM) were obtained from R&D Systems (Minneapolis, MN, United States of America). Monoclonal rabbit anti-DNA-binding protein inhibitor 3 (ID3), monoclonal mouse anti-ICAM1 and polyclonal rabbit anti-SMAD4 antibodies were obtained from Cell Signaling Technology (Beverly, MA, United States of America). Santa Cruz Biotechnology supplied the monoclonal mouse GAPDH antibody sc-47,724 (Santa Cruz, CA, United States of America). Bio-Rad Laboratories, Inc. supplied goat anti-mouse and goat anti-rabbit IgG that had been conjugated with horseradish peroxidase (Hercules, CA, United States of America).
## Reverse transcription-quantitative real-time PCR (RT‒qPCR)
Total RNA was extracted from collected endometrial tissue or cultured cells using TRIzol reagent (Invitrogen, Life Technologies, Inc.) according to the manufacturer’s instructions. Each reverse transcription procedure used 2 μg of RNA to create first-strand complementary DNA (cDNA) utilizing random primers and Moloney murine leukemia virus reverse transcriptase (Promega, Madison, WI, United States of America). Using an Applied Biosystems 7300 Real-Time PCR System, SYBR Green or TaqMan was used for RT-qPCR assays. Each 25 μl qPCR reaction comprised 12.5 μl of SYBR Green PCR Master Mix (Applied Biosystems, Foster City, CA), 100 ng of cDNA, and 7.5 nM of each specific primer. These primers were used in this study: ICAM-1, 5′-CTCCAATGTGCCAGGCTTG-3' (forward) and 5′- 5′-CAGTGGGAAAGTGCCATCCT-3’ (reverse); ID3, 5′- CAGCTTAGCCAGGTGGAAATCC-3' (forward) and 5′-GTCGTTGGAGATGACAAGTTCCG-3’ (reverse); SMAD4, 5′-TGGCCCAGGATCAGT AGGT-3′ (forward) and 5′-CATCAACACCAATTCCAGCA-3′ (reverse); and GAPDH, 5' -GAGTCAACGGATTTGGTCGT-3' (forward) and 5'- GACAAGCTTCCCGTTCTCAG-3' (reverse). Alternatively, TaqMan gene expression assay kits for BMP2 (Hs00154192_m1), ALK2 (Hs00153836_m1), ALK3 (Hs01034913_g1), and GAPDH (Hs02758991_g1) were bought from Applied Biosystems. Each 20 μL TaqMan RT-qPCR reaction comprised 1 × TaqMan Gene Expression Master Mix (Applied Biosystems), 20 ng of cDNA, and 1 × specific TaqMan assay Mix containing primers and probes. Relative quantification of the mRNA levels of target genes was determined based on the comparative cycle threshold (Ct) method, and the 2−ΔΔCt method was used specifically, with the results standardized to endogenous GAPDH.
## Western blot analysis
Total protein extracts were produced using lysis buffer (Cell Signaling Technology) containing protease inhibitor cocktail (Sigma Aldrich) from homogenized endometrial tissues or cultured cells. A DC Protein Assay (Bio-Rad Laboratories, Inc.) was used to measure the protein concentrations. Proteins of same concentration were put onto gels and transferred onto polyvinylidene fluoride (PVDF) membranes (Bio-Rad) after being separated by sodium dodecyl sulfate‒polyacrylamide gel electrophoresis. After blocking the membranes for 1 h at room temperature in Tris-buffered saline containing $0.1\%$ Tween-20 (TBST) and $5\%$ nonfat milk, they were immunoblotted overnight at 4°C with the appropriate primary antibodies. After three washes with TBST, the membranes were incubated with peroxidase-conjugated secondary antibodies for 1 hour. Enhanced chemiluminescent or SuperSignal West Femto chemiluminescent substrates (Thermo Fisher Scientific) were used to identify immunoreactive bands, which were subsequently subjected to X-ray film (Thermo Fisher Scientific). Image-Pro Plus software was used to calculate the band intensities (v4.5; Media Cybernetics, Carlsbad, CA).
## Small interfering RNA transfection
RNA interference was enabled by the transfection of small interfering RNA (siRNA). ON-TARGET plus non-targeting control pool siRNA or an ON-TARGET plus SMART pool targeting ALK2, ALK3, ID3 and SMAD4 were purchased from Dharmacon Inc. A total of 2 × 105 HESCs or primary HDSCs were simultaneously seeded with full culture media 1 day before transfection. Lipofectamine RNA iMAX (Life Technologies) was used to transfect control siRNA or siRNA against ALK2, ALK3, ID3 and SMAD4 into the cells at a dose of 25 nM in accordance with the manufacturer’s instructions. The cells were then cultured for 24 h at 37 °C in a CO2 incubator untill starvation (synchronization of all the cells to the same cell cycle phase and removal of various ligands in serum). RT‒qPCR or Western blot analysis was used to assess the knockdown effectiveness of each target.
## Statistical analysis
All statistical analyses were conducted using PRISM software (GraphPad Software, Inc., San Diego, CA). Using the unpaired Student’s t-test, comparisons were made between two sets of independent samples. Multiple comparisons of means were examined using one-way ANOVA and Newman‒Keuls testing. The results are reported as the mean ± S.E.M. of at least three independent experiments. The significance threshold was established at $p \leq 0.05.$
## Decreased BMP2 and increased ICAM1 expression in the endometrium of patients with uRPL
High expression levels of BMP2 and ICAM1 have been reported in human endometrium in prior studies (Thomson et al., 1999; Bai et al., 2020). Here, the expression levels of these two factors in secretory endometrial tissues were compared between uRPL patients and healthy fertile women by using RT‒qPCR and Western immunoblotting. As shown in Figure 1, compared with normal fertile controls, the expression of BMP2 in the endometrial tissues of uRPL patients was dramatically lower at mRNA level (Figure 1A), whereas the expression of ICAM1 at both mRNA (Figure 1B) and protein (Figure 1C) levels were considerably increased in the endometrium of women with uRPL.
**FIGURE 1:** *Expression levels of BMP2 and ICAM1 in the endometrium obtained from patients with unexplained recurrent pregnancy loss (uRPL) and fertile women (as a control). The endometrial samples were obtained from 12 uRPL patients and 16 fertile women (Control) during their mid-secretory phases. (A,B) The mRNA levels of BMP2 (A) and ICAM1 (B) of the endometrial tissues obtained from uRPL and controls were examined using RT-qPCR. (C) The protein levels of ICAM1 of the endometrial tissues obtained from uRPL and controls were examined using Western blot analysis. The results are expressed as the mean ± S.E.M. Different letters indicate a significant difference (p < 0.05).*
## BMP2 suppresses the expression of ICAM1 in HESCs and primary HDSCs
To investigate if BMP2 affects ICAM1 expression in HESCs, serum-starved cells were initially treated with vehicle control or various concentrations (10, 25, or 50 ng/ml) of recombinant human BMP2 for 24 h. The expression levels of ICAM1 were detected by RT‒qPCR and Western immunoblotting. The results revealed that both the mRNA (Figure 2A) and protein levels (Figure 2B) of ICAM1 were markedly downregulated in a concentration-dependent fashion in response to the BMP2 treatment in HESCs. The time-course study showed that cultivation with 25 ng/mL BMP2 dramatically decreased the mRNA (Figure 2C) and protein levels of ICAM1 at 24 h and 48 h (Figure 2D). We also evaluated the influence of BMP2 on the expression of ICAM1 in primary HDSCs in addition to HESCs. The purity of HDSCs was determined by immunofluorescent staining with markers specific to mesenchymal cells (vimentin) and epithelial cells (cytokeratin-7). HDSCs used in these studies were approximately $98\%$ pure as assessed by vimentin-positive and cytokeratin-negative staining (Supplementary Figure S1). In accordance with the findings in HESCs, treatment with various dosages (10, 25, or 50 ng/mL) of BMP2 for 24 h dramatically decreased the mRNA (Figure 3A) and protein (Figure 3B) levels of ICAM1 in a concentration-dependent way. The time-course analysis revealed that 25 ng/mL BMP2 treatment substantially decreased ICAM1 mRNA (Figure 3C) and protein (Figure 3D) levels at 24 h, and 48 h.
**FIGURE 2:** *BMP2 downregulates the expression of ICAM1 in non-decidualized HESCs. (A,B) HESCs were treated with different concentrations (0, 10, 25, 50 ng/mL) of recombinant human BMP2 (BMP2) for 24 h, and the mRNA (A) and protein (B) levels of ICAM1 were examined using RT-qPCR and Western blot analysis, respectively. (C,D) HESCs were treated with 25 ng/mL of BMP2 for 3, 6, 12, 24 or 48 h, and the mRNA (C) and protein (D) levels of ICAM1 were examined using RT-qPCR and Western blot analysis, respectively. The results are expressed as the mean ± S.E.M. of at least three independent experiments. Different letters indicate significant a difference (p < 0.05). In detail, if the letters on two columns are different (E.g., “a” vs. “b” or “b” vs. “c”), it means that the difference between the two groups is significant, on the other hand, if the letters on the column of two groups are the same (E.g., “a” vs. “a” or “b” vs. “b”), it means there is no significant difference between two groups. C, Ctrl, Control; B2, BMP2.* **FIGURE 3:** *BMP2 downregulates the expression of ICAM1 in primary human endometrial stromal cells (HDSCs). (A,B) HDSCs (n = 3) were treated with different concentrations (0, 10, 25, 50 ng/mL) of BMP2 for 24 h, and the mRNA (A) and protein (B) levels of ICAM1 were examined using RT-qPCR and Western blot analysis, respectively. (C,D) HDSCs (n = 3) were treated with 25 ng/mL of BMP2 for 3, 6, 12, 24 or 48 h, and the mRNA (C) and protein (D) levels of ICAM1 were examined using RT-qPCR and Western blot analysis, respectively. The results are expressed as the mean ± S.E.M. Each experiment was performed in duplicate. Different letters indicate a significant difference (p < 0.05). C, Ctrl, Control; B2, BMP2.*
## BMP2 upregulates the expression of ID3 in HESCs and primary HDSCs
Transcriptional regulator inhibitor of DNA-binding/differentiation-3 (ID3) is a member of the helix-loop-helix (HLH) protein family. Several previous studies have demonstrated that ID3 is a downstream target of BMPs [28–30]. To determine if ID3 is a downstream target of BMP2 in this assay system, the expression levels of ID3 in HESCs and HDSCs were assessed using RT‒qPCR and Western blot analysis following BMP2 treatment. As shown in Figure 4, the mRNA (Figure 4A) and protein (Figure 4B) the expression levels of ID3 were remarkably increased after treatment with gradually increased concentrations of BMP2 in HESCs. In time-response experiments, the mRNA (Figure 4C) and protein (Figure 4D) levels of ID3 markedly rose 3 h after 25 ng/ml BMP2 treatment and persisted for 48 h in HESCs. Similar results were noted in the primary HDSCs. Increasing BMP2 doses resulted in a considerable rise in mRNA (Figure 5A) and protein (Figure 5B) expression levels of ID3, and 25 ng/ml BMP2 treatment markedly enhanced the mRNA (Figure 5C) and protein (Figure 5D) levels of ID3 at 3 h and persisted for 48 h in HDSCs. When cells were treated with BMP2 at a concentration of 10 ng/ml, the expression levels of ID3 and ICAM1 were significantly changed both in HESCs and HDSCs, but the decreased range of ICAM1 expression in protein was mild (0.84-fold change relative to Ctrl in HESCs and 0.85-fold change relative to Ctrl in HDSCs). Therefore, we chose a concentration of 25 ng/ml for the follow-up studies to ensure the stability and reproducibility of the experiment.
**FIGURE 4:** *BMP2 upregulates the expression of ID3 in HESCs. (A and B) HESCs were treated with different concentrations (0, 10, 25, 50 ng/mL) of BMP2 for 24 h, and the mRNA (A) and protein (B) levels of ID3 were examined using RT-qPCR and Western blot analysis, respectively. (C,D) HESCs were treated with 25 ng/mL of BMP2 for 3, 6, 12, 24 or 48 h, and the mRNA (C) and protein (D) levels of ID3 were examined using RT-qPCR and Western blot analysis, respectively. The results are expressed as the mean ± S.E.M. of at least three independent experiments. Different letters indicate a significant difference (p < 0.05).* **FIGURE 5:** *BMP2 upregulates the expression of ID3 by BMP2 in HDSCs. (A and B) HDSCs were treated with different concentrations (0, 10, 25, 50 ng/mL) of recombinant human BMP2 for 24 h, and the mRNA (A) and protein (B) levels of ID3 were examined using RT-qPCR and Western blot analysis, respectively. (C,D) HDSCs were treated with 25 ng/mL of BMP2 for 3, 6, 12, 24 or 48 h, and the mRNA (C) and protein (D) levels of ID3 were examined using RT-qPCR and Western blot analysis, respectively. The results are expressed as the mean ± S.E.M. of at least three independent experiments. Different letters indicate a significant difference (p < 0.05).*
## ID3 mediates the BMP2-induced downregulation of ICAM1 in HESCs and primary HDSCs
By transfecting HESCs and primary HDSCs with ID3-targeting siRNA (siID3), siRNA-based knockdown studies were carried out to investigate the functions of ID3 in BMP2-induced downregulation of ICAM1. The HESCs and primary HDSCs were transfected with siID3 for 24 h, starved in serum-free media for 18 h, and then cocultured with recombinant BMP2 (25 ng/mL) for 24 h. Our results showed that knockdown of ID3 significantly reduced the ID3 expression induced by BMP2 at the mRNA and protein levels (Figure 6A) as well as reversed the mRNA and protein levels of the BMP2-induced downregulation of ICAM1 (Figure 6B) in HESCs. Similar findings were also obtained in primary HDSCs (Figures 6C,D). However, silencing ID3 does not seem to alter basal levels of ICAM1. The same phenomenon has been found in some other studies (Li et al., 2019; Li et al., 2021). We believe that the possible reason is that in the non-stimulated HESCs and HDSCs, the expression of ID3 was very low, and silencing ID3 cannot alter the constitutive expression of ICAM1 but can only affect the inducible ICAM1 expression. These findings imply that ID3 mediates the BMP2-induced downregulation of ICAM1 in HESCs and primary HDSCs.
**FIGURE 6:** *ID3 mediates the BMP2-induced downregulation of ICAM1 expression in HESCs and HDSCs. (A,B) HESCs were transfected with 25 nM control siRNA (siCtrl) or 25 nM siRNA targeting ID3 (siID3) for 48 h, and the cells were then treated with vehicle control (Ctrl) or 25 ng/mL BMP2 for an additional 24 h. The mRNA and protein levels of ID3 (A) and ICAM1 (B) were examined using RT-qPCR and Western blot analysis respectively. (C,D) HDSCs were transfected with 25 nM siCtrl or 25 nM siID3 for 48 h, and the cells were then treated with Ctrl or 25 ng/ml of BMP2 for an additional 24 h. The mRNA and protein levels of ID3 (C) and ICAM1 (D) were examined using RT-qPCRand Western blot analysis, respectively. The results are expressed as the mean ± S.E.M. of at least three independent experiments. Different letters indicate significant a difference (p < 0.05).*
## DMH-1 or dorsomorphin abolishes the BMP2-induced upregulation of ID3 and downregulation of ICAM1 expression in HESCs and primary HDSCs
BMPs transduce signals through transmembrane serine/kinases composed of type I (ALK2, ALK3, and ALK6) and type II receptors (Salazar et al., 2016). To further clarify the role of these type I receptors in the BMP2-induced upregulation of ID3 and downregulation of ICAM1 expression in HESCs and primary HDSCs, these two kinds of cells were pretreated with different TGF-β type I receptor inhibitors, namely, DMH-1 (an inhibitor of ALK2 and ALK3) or dorsomorphin (an inhibitor of ALK2, ALK3, and ALK6) for 1 h, followed by treatment with 25 ng/mL BMP2 for another 24 h. The cells in the control group were treated with equal volumes of DMSO, which has been widely used to formulate compounds for cell administration. Several studies have confirmed that there is no interference with the cell experiment if the final concentration of DMSO is controlled within $0.1\%$ (v/v) (Qi et al., 2008; Dludla et al., 2018). In the present study, the final concentration of DMSO is $0.05\%$ (v/v), which is considered to be a safe vehicle of reagents for almost all cells. The results showed that either DMH-1 or dorsomorphin completely abolished BMP2-induced upregulation of ID3 (Figures 7A, C) and downregulation of ICAM1 (Figures 7B, D) mRNA and protein expression in the HESCs (Figures 7A, B) and primary HDSCs (Figures 7C, D). These results indicate that BMP2 regulates the expression of ID3 and ICAM1 through ALK2 and/or ALK3 receptors.
**FIGURE 7:** *The effects of specific TGF-β type I inhibitors on BMP2-induced upregulation of ID3 expression and downregulation of ICAM1 expression in HESCs and HDSCs. (A and B) HESCs were pretreated with dimethyl sulfoxide (DMSO), DMH-1 (0.25 µM), or dorsomorphin (10 µM) for 1 h, and the cells were then treated with 25 ng/ml BMP2 for an additional 24 h. The mRNA and protein levels of ID3 (A) and ICAM1 (B) were examined using RT-qPCR and Western blot analysis, respectively. (C,D) HDSCs were pretreated with DMSO, DMH-1 (0.25 µM), or dorsomorphin (10 µM) for 1 h, and the cells were then treated with 25 ng/mL BMP2 for an additional 24 h. The mRNA and protein levels of ID3 (C) and ICAM1 (D) were examined using RT-qPCR and Western blot analysis, respectively. The results are expressed as the mean ± S.E.M. of at least three independent experiments. Different letters indicate a significant difference (p < 0.05).*
## ALK3 type I receptor mediates the BMP2-induced upregulation of ID3 and downregulation of ICAM1 expression in HESCs and primary HDSCs
To learn more about the particular type I receptor that BMP2 uses to upregulate ID3 expression and to reduce ICAM1 expression in HESCs and primary HDSCs, the effects of ALK2 or ALK3 were suppressed using a siRNA-based suppression strategy. The cells were pretreated with specific siRNAs targeting ALK2 (siALK2) or ALK3 (siALK3) for 48 h, followed by treatment with 25 ng/mL BMP2 for an additional 24 h. Our results showed that pretreated with siALK2 or siALK3 significantly reduced the target gene expression in HESCs and HDSCs without affecting the expression of the other one, and the expression levels of ALK2 and ALK3 were not affected by BMP2 treatment (Supplementary Figure S2). As shown in Figure 8, knockdown of ALK3 completely abolished the downregulation of ICAM1 (Figure 8A) and upregulation of ID3 (Figure 8B) induced by BMP2 at both the mRNA and protein levels in HESCs. Interestingly, knockdown of ALK2 partially inhibited the BMP2-induced downregulation of ICAM1 (Figure 8C) and upregulation of ID3 expression (Figure 8D) in HDSCs. However, it seemed to have no significant effect on the downregulation of ICAM1 mRNA or protein expression (Figure 8A) induced by BMP2 in HESCs. These results suggest that ALK3 is the major receptor for upregulating ID3 and downregulating ICAM1 expression by BMP2 in HESCs and primary HDSCs.
**FIGURE 8:** *ALK3 type I mediates the BMP2-induced downregulation of ICAM1 and ID3 in HESCs and HDSCs. (A,B) HESCs were transfected with 25 nM siCtrl, siRNA targeting ALK2 (siALK2) or siRNA targeting ALK3 (siALK3) for 48 h, and the cells were then treated with Ctrl or 25 ng/mL of BMP2 for an additional 24 h. The mRNA levels of ICAM1 (A) and ID3 (B) were examined using RT-qPCR and the protein levels of ICAM1 and ID3 were examined using Western blot analysis. (C,D) HDSCs were transfected with 25 nM siCtrl, 25 nM siALK2 or 25 nM siALK3 for 48 h, and the cells were then treated with Ctrl or 25 ng/mL BMP2 for an additional 24 h. The mRNA levels of ICAM1 (C) and ID3 (D) were examined using RT-qPCR, and the protein levels of ICAM1 and ID3 were examined using Western blot analysis. The results are expressed as the mean ± S.E.M. of at least three independent experiments. Different letters indicate significant a difference (p < 0.05).*
## SMAD4 mediates the BMP2-induced upregulation of ID3 and downregulation of ICAM1 expression in HESCs and primary HDSCs
Prior to moving into the nucleus to control target genes, phosphorylated R-SMADs join forces with SMAD4 to create a heterotrimeric transcription factor complex. We transfected HESCs and primary HDSCs with siSMAD4 to inhibit endogenous SMAD4 expression in order to better understand if the SMAD signaling pathway is involved in the BMP2-induced overexpression of ID3 and downregulation of ICAM1. Following a 48-h siSMAD4 transfection, the cells were given 25 ng/mL BMP2 treatment for an additional 24-h period. SMAD4 knockdown effectiveness was confirmed by Western blot and RT-qPCR analysis. Our results showed that the expression of SMAD4 almost eliminated after siSMAD4 transfection and treatment with 25 ng/mL BMP2 did not alter SMAD4 expression in HESCs and HDSCs (Figures 9A, D). The siRNA-mediated depletion of SMAD4 completely abolished the downregulation of ICAM1(Figure 9B) and upregulation of ID3 (Figure 9C) mRNA and protein expression in HESCs. Notably, SMAD4 knockdown also completely abolished BMP2-induced cell activity in HDSCs (Figures 9E, F).
**FIGURE 9:** *SMAD4 is involved in the BMP2-induced upregulation of ID3 expression and downregulation of ICAM1 expression in HESCs and HDSCs. (A–C) HESCs were transfected with 25 nMsiCtrl or 25 nM siRNA targeting SMAD4 (siSMAD4) for 48 h, and the cells were then treated with BMP2 (25 ng/mL) for an additional 24 h. The mRNA and protein levels of SAMD4 (A), ICAM1 (B) and ID3 (C) were examined using RT-qPCR and Western blot analysis, respectively. (D–F) HDSCs were transfected with 25 nM siCtrl or 25 nM siSMAD4 for 48 h, and the cells were then treated with BMP2 (25 ng/mL) for an additional 24 h. The mRNA and protein levels of SAMD4 (D), ICAM1 (E) and ID3 (F) were examined using RT-qPCR and Western blot analysis. Respectively. The results are expressed as the mean ± S.E.M. of at least three independent experiments. Different letters indicate significant a difference (p < 0.05).*
## Discussion
Due to the unclear underlying pathogenesis and the lack of effective interventions, uRPL remains one of the most challenging and frustrating fertility-related diseases. In the present study, we found that mid-secretory endometrium samples obtained from uRPL patients had significantly lower BMP2 and higher ICAM1 levels than fertile controls. Additionally, we demonstrate for the first time that BMP2 suppresses ICAM1 expression through a mechanism reliant on ID3 overexpression both in non-decidual and decidual stromal cells. This information will aid in the development of new pharmacological strategies for unexplained RPL.
There are a large number of cell adhesion molecules in the human endometrium that appear to be necessary to successfully establish the physical interaction between the embryo and the endometrium (van Mourik et al., 2009). One of the best-characterized cell adhesion molecules is ICAM1, which is a ligand for integrin molecule 1 (LFA-1). The expression of ICAM1 has previously been examined in endometrial stromal cells and endometrial epithelial cells in multiple species (Blois et al., 2005; Defrere et al., 2005; Lecce et al., 2011). A previous study demonstrated that ICAM1 was substantially increased in the uterine epithelium and the stroma of high-stress sensing induced abort-prone mice compared to control mice, resulting in more LFA-1-expressing lymphocytes being recruited from the blood into the uterus. Neutralization of the adhesion molecules ICAM1/LFA-1 radically eliminated the effect of stress on the embryonic abortion (Blois et al., 2005). Further research revealed that upregulated ICAM1 in the decidua promoted Th1 polarization via mature dendritic cells, leading to Th1/Th2 imbalance (Blois et al., 2005), which is known to contribute to the pathogenesis of RPL. Notably, evidence from a clinical study showed that elevated ICAM1 levels detected by ELISA in human endometrial tissue correlated with idiopathic RPL (Comba et al., 2015). Similar results were obtained by examining the RNA and protein expression levels of ICAM1 in endometrial tissue obtained from uRPL patients and normal fertile women in our study. At the same time, we also detected that BMP2 expression in the endometrium of these two groups exhibited an opposite phenotype to ICMA1. It has been demonstrated that the upregulation of ICAM1 in HESCs can be induced by various inflammatory mediators such as IL-1 β (Vigano et al., 1994), TNF-α (Thomson et al., 1999), and INF-γ (Mangioni et al., 2005). However, the regulation of ICAM1 by BMP2 has not yet been reported. Definitive evidence shows that BMP2 is essential for pregnancy establishment and maintenance by regulating blastocyst implantation, uterine decidualization and placental/fetal development (Lee et al., 2007; Yi et al., 2021). Transcriptome analysis revealed that BMP2 targets primarily play a key role in regulating cell adhesion and extracellular matrix (ECM) transformation in human endometrium (Yi et al., 2021). By controlling the expression of IGFBP3, our prior work also showed that BMP2 contributes to endometrial remodeling in human non-decidual and decidual stromal cells (Luo et al., 2020). In this study, our in vitro analysis showed that the expression of ICAM1 in HESCs and primary HDSCs can also be regulated by BMP2 via ID3.
The inhibitors of DNA binding proteins (ID) are dominant negative antagonists of basic helixloop-helix (bHLH) transcription factors. To date, four ID family proteins have been identified in mammalian cells and have been demonstrated to be expressed in the uterine endometrium as well as the maternal-fetal interface (Han et al., 2018). Our studies showed that ID1 (Luo et al., 2020), ID2 (Supplementary Figure S3), and ID3 (Figure 4; Figure 5) are among the most significantly upregulated genes upon stimulation of BMP2. Further studies based on the siRNA-based knockdown experiments found that it was ID3 but not ID1 or ID2 that mediated BMP2-induced downregulation of ICAM1 in HESCs and primary HDSCs (Figure 6 and Supplementary Figure S4). ID3 expression can be differentially regulated by members of the TGF-β superfamily in various cell types; for example, TGF-β1 represses ID3 expression in adult neural stem/precursor cells (Bohrer et al., 2015), while TGF-β1 increases ID3 mRNA and nuclear ID3 protein levels in immortalized human granulosa cells (Li et al., 2019). In addition, other TGF-β superfamily members, such as BMP4 and BMP6, upregulate ID3 expression in a range of different cell lines, including embryonic stem cells, human B progenitor cells, intestinal stem cells, and neuronal stem cells (Hollnagel et al., 1999; Kersten et al., 2006; Hu et al., 2021). Importantly, a previous study revealed that ID3 was dramatically upregulated by BMP2 in adult neural stem/precursor cells and was essential for BMP2- induced differentiation of neural stem/precursor cells into astrocytes (Bohrer et al., 2015). In this study, our findings add to growing evidence that the overexpression of ID3 induced by BMP2 was required for BMP2-suppressed ICAM1 expression in HESCs and primary HDSCs. Furthermore, we explored the underlying molecular mechanism by which BMP2 regulates the ID3 and ICAM1 expression by pretreating HESCs and primary HDSCs with different TGF-β type I receptor inhibitors, including DMH-1 (an inhibitor of ALK$\frac{2}{3}$) and dorsomorphin (an inhibitor of ALK$\frac{2}{3}$/6) prior to BMP2 treatment. Our results show that the inhibitors DMH-1 and dorsomorphin can significantly eliminate the upregulation of ID3 and downregulation of ICAM1 induced by BMP2, but the siRNA-mediated gene downregulation provided more accurate evidence that it is ALK3 rather than ALK2 mainly responsible for the downstream pathway of BMP2 induction. We suppose that although ALK3 and ALK2 are paralogous genes and both are downstream receptors of BMPs, their protein primary structures are about $10\%$ different, which may lead to some functional divergence. This will be interesting research content in the future. Ligand‒receptor complexes induce downstream signaling in a SMAD-dependent manner following BMP2 binding to specific receptors (Shi and Massague, 2003). SMAD$\frac{1}{5}$ is thought to be the main downstream signaling pathway that mediates BMP2 signaling pathway in HESCs and primary HDSCs (Zhang et al., 2020; Zhang et al., 2022). Upon phosphorylation of type I receptors in the majority of tissues, phosphorylated SMAD$\frac{1}{5}$ bind to a common SMAD (SMAD4) to create a heterotrimer complex, which translocates into the nucleus to control the expression of target genes. Here, knocking down SMAD4 totally reversed the effects of BMP2 on ID3 and ICAM1 expression, suggesting that SMAD4 is required for BMP2-induced intracellular signaling in HESCs and primary HDSCs.
In conclusion, our data reveal that downregulation of BMP2 in the endometrium may contribute to the pathogenesis of uRPL by increasing ICAM1 expression via the ALK3-SMAD4-ID3 signaling pathway (Figure 10). These findings not only deepen the understanding of the molecular regulatory mechanisms of ICAM1 expression in the human endometrium but also suggest that it may be possible to improve the pregnancy outcomes in patients with uRPL by regulating the local expression of BMP2 or ICAM1 in the endometrium.
**FIGURE 10:** *Schematic diagram of the proposed molecular mechanisms by which BMP2 downregulates the expression of ICAM1 in human endometrial stromal cells and decidual cells. BMP2 binds to a pair of ALK3 type 1 receptor and BMP type II receptors, leading to the activation of canonical R-SMADs, which are associated with a common SMAD (SMAD4) and further increases the transcription of ID3. The increase in ID3 suppresses the expression of ICAM1.*
## Data availability statement
The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding authors.
## Ethics statement
The studies involving human participants were reviewed and approved by Ethics Committee of Renmin Hospital, Wuhan University. The patients/participants provided their written informed consent to participate in this study.
## Author contributions
JL, JY, and PL conceived and designed the research; JL and YW performed experiments; JL and YW analyzed and interpreted the results of the experiments; JL and HZ drafted the manuscript; H.-MC and PL edited and revised the manuscript. All the authors approved the final version of the manuscript.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fcell.2023.1090593/full#supplementary-material
## References
1. Austgulen R., Lien E., Vince G., Redman C. W.. **Increased maternal plasma levels of soluble adhesion molecules (ICAM-1, VCAM-1, E-selectin) in preeclampsia**. *Eur. J. Obstet. Gynecol. Reprod. Biol.* (1997) **71** 53-58. DOI: 10.1016/s0301-2115(96)02647-4
2. Bai L., Chang H. M., Zhang L., Zhu Y. M., Leung P. C. K.. **BMP2 increases the production of BDNF through the upregulation of proBDNF and furin expression in human granulosa-lutein cells**. *FASEB J.* (2020) **34** 16129-16143. DOI: 10.1096/fj.202000940R
3. Benner M., Feyaerts D., Lopez-Rincon A., van der Heijden O. W. H., van der Hoorn M. L., Joosten I.. **A combination of immune cell types identified through ensemble machine learning strategy detects altered profile in recurrent pregnancy loss: a pilot study**. *F. S Sci.* (2022) **3** 166-173. DOI: 10.1016/j.xfss.2022.02.002
4. Blois S., Tometten M., Kandil J., Hagen E., Klapp B. F., Margni R. A.. **Intercellular adhesion molecule-1/LFA-1 cross talk is a proximate mediator capable of disrupting immune integration and tolerance mechanism at the feto-maternal interface in murine pregnancies**. *J. Immunol.* (2005) **174** 1820-1829. DOI: 10.4049/jimmunol.174.4.1820
5. Bohrer C., Pfurr S., Mammadzada K., Schildge S., Plappert L., Hils M.. **The balance of Id3 and E47 determines neural stem/precursor cell differentiation into astrocytes**. *EMBO J.* (2015) **34** 2804-2819. DOI: 10.15252/embj.201591118
6. Comba C., Bastu E., Dural O., Yasa C., Keskin G., Ozsurmeli M.. **Role of inflammatory mediators in patients with recurrent pregnancy loss**. *Fertil. Steril.* (2015) **104** 1467-1474.e1. DOI: 10.1016/j.fertnstert.2015.08.011
7. Dambaeva S., Bilal M., Schneiderman S., Germain A., Fernandez E., Kwak-Kim J.. **Decidualization score identifies an endometrial dysregulation in samples from women with recurrent pregnancy losses and unexplained infertility**. *F. S Rep.* (2021) **2** 95-103. DOI: 10.1016/j.xfre.2020.12.004
8. Defrere S., Van Langendonckt A., Moulin P., Befahy P., Gonzalez D., Martinez-Madrid B.. **Human endometrial epithelial cells (EEC) constitutively express more intercellular adhesion molecule (ICAM)-1 than endometrial stromal cells (ESC) in culture**. *Am. J. Reprod. Immunol.* (2005) **54** 5-12. DOI: 10.1111/j.1600-0897.2005.00272.x
9. Dludla P. V., Jack B., Viraragavan A., Pheiffer C., Johnson R., Louw J.. **A dose-dependent effect of dimethyl sulfoxide on lipid content, cell viability and oxidative stress in 3T3-L1 adipocytes**. *Toxicol. Rep.* (2018) **5** 1014-1020. DOI: 10.1016/j.toxrep.2018.10.002
10. El Hachem H., Crepaux V., May-Panloup P., Descamps P., Legendre G., Bouet P. E.. **Recurrent pregnancy loss: current perspectives**. *Int. J. Womens Health* (2017) **9** 331-345. DOI: 10.2147/IJWH.S100817
11. Bender Atik R., Christiansen O. B., Elson J., Kolte A. M., Lewis S.. **ESHRE guideline: recurrent pregnancy loss**. *Hum. Reprod. Open* (2018) **2018** hoy004. DOI: 10.1093/hropen/hoy004
12. Garrido-Gimenez C., Alijotas-Reig J.. **Recurrent miscarriage: causes, evaluation and management**. *Postgrad. Med. J.* (2015) **91** 151-162. DOI: 10.1136/postgradmedj-2014-132672
13. Han J., Seo H., Choi Y., Lee C., Kim M. I., Jeon Y.. **Expression and regulation of inhibitor of DNA binding proteins ID1, ID2, ID3, and ID4 at the maternal-conceptus interface in pigs**. *Theriogenology* (2018) **108** 46-55. DOI: 10.1016/j.theriogenology.2017.11.029
14. Heidari Z., Moudi B., Sheibak N., Asemi-Rad A., Keikha N., Mahmoudzadeh-Sagheb H.. **Interleukin 22 expression during the implantation window in the endometrium of women with unexplained recurrent pregnancy loss and unexplained infertility compared to healthy parturient individuals**. *J. Interferon Cytokine Res.* (2021) **41** 461-468. DOI: 10.1089/jir.2021.0160
15. Hollnagel A., Oehlmann V., Heymer J., Ruther U., Nordheim A.. **Id genes are direct targets of bone morphogenetic protein induction in embryonic stem cells**. *J. Biol. Chem.* (1999) **274** 19838-19845. DOI: 10.1074/jbc.274.28.19838
16. Hu L., Xu J., Wang X., Feng L., Zhang C., Wang J.. **Bone morphogenetic protein 4 alleviates DSS-induced ulcerative colitis through activating intestinal stem cell by target ID3**. *Front. Cell Dev. Biol.* (2021) **9** 700864. DOI: 10.3389/fcell.2021.700864
17. Kersten C., Dosen G., Myklebust J. H., Sivertsen E. A., Hystad M. E., Smeland E. B.. **BMP-6 inhibits human bone marrow B lymphopoiesis-upregulation of Id1 and Id3**. *Exp. Hematol.* (2006) **34** 72-81. DOI: 10.1016/j.exphem.2005.09.010
18. Lecce L., Kaneko Y., Madawala R. J., Murphy C. R.. **ICAM1 and fibrinogen-gamma are increased in uterine epithelial cells at the time of implantation in rats**. *Mol. Reprod. Dev.* (2011) **78** 318-327. DOI: 10.1002/mrd.21307
19. Lee K. Y., Jeong J. W., Wang J., Ma L., Martin J. F., Tsai S. Y.. **Bmp2 is critical for the murine uterine decidual response**. *Mol. Cell Biol.* (2007) **27** 5468-5478. DOI: 10.1128/MCB.00342-07
20. Li Q., Kannan A., Wang W., Demayo F. J., Taylor R. N., Bagchi M. K.. **Bone morphogenetic protein 2 functions via a conserved signaling pathway involving Wnt4 to regulate uterine decidualization in the mouse and the human**. *J. Biol. Chem.* (2007) **282** 31725-31732. DOI: 10.1074/jbc.M704723200
21. Li H., Chang H. M., Shi Z., Leung P. C. K.. **ID3 mediates the TGF-β1-induced suppression of matrix metalloproteinase-1 in human granulosa cells**. *FEBS J.* (2019) **286** 4310-4327. DOI: 10.1111/febs.14964
22. Li H., Chang H. M., Lin Y. M., Shi Z., Leung P. C. K.. **TGF-β1 inhibits microvascular-like formation by decreasing VCAM1 and ICAM1 via the upregulation of SNAIL in human granulosa cells**. *Mol. Cell Endocrinol.* (2021) **111395** 111395. DOI: 10.1016/j.mce.2021.111395
23. Luo J., Zhu H., Chang H. M., Lin Y. M., Yang J., Leung P. C. K.. **The regulation of IGFBP3 by BMP2 has a role in human endometrial remodeling**. *FASEB J.* (2020) **34** 15462-15479. DOI: 10.1096/fj.202000508R
24. Luo X., Chang H. M., Yi Y., Leung P. C. K., Sun Y.. **Bone morphogenetic protein 2 upregulates SERPINE2 expression through noncanonical SMAD2/3 and p38 MAPK signaling pathways in human granulosa-lutein cells**. *FASEB J.* (2021a) **35** e21845. DOI: 10.1096/fj.202100670RR
25. Luo X., Chang H. M., Yi Y., Sun Y., Leung P. C. K.. **Bone morphogenetic protein 2 inhibits growth differentiation factor 8-induced cell signaling via upregulation of gremlin2 expression in human granulosa-lutein cells**. *Reprod. Biol. Endocrinol.* (2021b) **19** 173. DOI: 10.1186/s12958-021-00854-6
26. Mangioni S., Vigano P., Florio P., Borghi O., Vignali M., Petraglia F.. **Effect of activin A on tumor necrosis factor-alpha/intercellular adhesion molecule-1 pathway in endometrial stromal cells**. *Eur. J. Obstet. Gynecol. Reprod. Biol.* (2005) **123** 218-223. DOI: 10.1016/j.ejogrb.2005.03.017
27. Marzusch K., Ruck P., Geiselhart A., Handgretinger R., Dietl J. A., Kaiserling E.. **Distribution of cell adhesion molecules on CD56++, CD3-CD16-large granular lymphocytes and endothelial cells in first-trimester human decidua**. *Hum. Reprod.* (1993) **8** 1203-1208. DOI: 10.1093/oxfordjournals.humrep.a138229
28. Pino M., Galleguillos C., Torres M., Sovino H., Fuentes A., Boric M. A.. **Association between MMP1 and MMP9 activities and ICAM1 cleavage induced by tumor necrosis factor in stromal cell cultures from eutopic endometria of women with endometriosis**. *Reproduction* (2009) **138** 837-847. DOI: 10.1530/REP-09-0196
29. Qi W., Ding D., Salvi R. J.. **Cytotoxic effects of dimethyl sulphoxide (DMSO) on cochlear organotypic cultures**. *Hear Res.* (2008) **236** 52-60. DOI: 10.1016/j.heares.2007.12.002
30. Rothlein R., Czajkowski M., Kishimoto T. K.. **Intercellular adhesion molecule-1 in the inflammatory response**. *Chem. Immunol.* (1991) **50** 135-142. PMID: 1686175
31. Salazar V. S., Gamer L. W., Rosen V.. **BMP signalling in skeletal development, disease and repair**. *Nat. Rev. Endocrinol.* (2016) **12** 203-221. DOI: 10.1038/nrendo.2016.12
32. Shi Y., Massague J.. **Mechanisms of TGF-beta signaling from cell membrane to the nucleus**. *Cell* (2003) **113** 685-700. DOI: 10.1016/s0092-8674(03)00432-x
33. Stoikos C. J., Harrison C. A., Salamonsen L. A., Dimitriadis E.. **A distinct cohort of the TGFbeta superfamily members expressed in human endometrium regulate decidualization**. *Hum. Reprod.* (2008) **23** 1447-1456. DOI: 10.1093/humrep/den110
34. Thomson A. J., Greer M. R., Young A., Boswell F., Telfer J. F., Cameron I. T.. **Expression of intercellular adhesion molecules ICAM-1 and ICAM-2 in human endometrium: regulation by interferon-gamma**. *Mol. Hum. Reprod.* (1999) **5** 64-70. DOI: 10.1093/molehr/5.1.64
35. van Mourik M. S., Macklon N. S., Heijnen C. J.. **Embryonic implantation: cytokines, adhesion molecules, and immune cells in establishing an implantation environment**. *J. Leukoc. Biol.* (2009) **85** 4-19. DOI: 10.1189/jlb.0708395
36. Vigano P., Pardi R., Magri B., Busacca M., Di Blasio A. M., Vignali M.. **Expression of intercellular adhesion molecule-1 (ICAM-1) on cultured human endometrial stromal cells and its role in the interaction with natural killers**. *Am. J. Reprod. Immunol.* (1994) **32** 139-145. DOI: 10.1111/j.1600-0897.1994.tb01104.x
37. Xie L., Galettis A., Morris J., Jackson C., Twigg S. M., Gallery E. D.. **Intercellular adhesion molecule-1 (ICAM-1) expression is necessary for monocyte adhesion to the placental bed endothelium and is increased in type 1 diabetic human pregnancy**. *Diabetes Metab. Res. Rev.* (2008) **24** 294-300. DOI: 10.1002/dmrr.793
38. Yi Y., Zhu H., Klausen C., Chang H. M., Inkster A. M., Terry J.. **Dysregulated BMP2 in the placenta may contribute to early-onset preeclampsia by regulating human trophoblast expression of extracellular matrix and adhesion molecules**. *Front. Cell Dev. Biol.* (2021) **9** 768669. DOI: 10.3389/fcell.2021.768669
39. Zhang Y., Zhu H., Chang H. M., Leung P. C. K.. **ALK3-SMAD1/5 signaling mediates the BMP2-induced decrease in PGE2 production in human endometrial stromal cells and decidual stromal cells**. *Front. Cell Dev. Biol.* (2020) **8** 573028. DOI: 10.3389/fcell.2020.573028
40. Zhang Y., Chang H. M., Zhu H., Leung P. C. K.. **BMP2 suppresses the production of pentraxin 3 in human endometrial stromal and decidual stromal cells**. *FASEB J.* (2022) **36** e22319. DOI: 10.1096/fj.202200081RR
41. Zhao H. J., Chang H. M., Zhu H., Klausen C., Li Y., Leung P. C. K.. **Bone morphogenetic protein 2 promotes human trophoblast cell invasion by inducing activin A production**. *Endocrinology* (2018a) **159** 2815-2825. DOI: 10.1210/en.2018-00301
42. Zhao H. J., Klausen C., Li Y., Zhu H., Wang Y. L., Leung P. C. K.. **Bone morphogenetic protein 2 promotes human trophoblast cell invasion by upregulating N-cadherin via non-canonical SMAD2/3 signaling**. *Cell Death Dis.* (2018b) **9** 174. DOI: 10.1038/s41419-017-0230-1
43. Zhao H. J., Klausen C., Zhu H., Chang H. M., Li Y., Leung P. C. K.. **Bone morphogenetic protein 2 promotes human trophoblast cell invasion and endothelial-like tube formation through ID1-mediated upregulation of IGF binding protein-3**. *FASEB J.* (2020) **34** 3151-3164. DOI: 10.1096/fj.201902168RR
44. Zhu H., Leung P. C., MacCalman C. D.. **Expression of ADAMTS-5/implantin in human decidual stromal cells: regulatory effects of cytokines**. *Hum. Reprod.* (2007) **22** 63-74. DOI: 10.1093/humrep/del356
45. Zhu R. H., Dai F. F., Yang D. Y., Liu S. Y., Zheng Y. J., Wu M. L.. **The mechanism of insulin-like growth factor II mRNA-binging protein 3 induce decidualization and maternal-fetal interface cross talk by TGF-β1 in recurrent spontaneous abortion**. *Front. Cell Dev. Biol.* (2022) **10** 862180. DOI: 10.3389/fcell.2022.862180
|
---
title: Development and validation of the diabetic self-management scale based on information-motivation-behavioral
skills theory
authors:
- Zhenwei Dai
- Shu Jing
- Xiaoyang Liu
- Haoran Zhang
- Yijin Wu
- Hao Wang
- Weijun Xiao
- Yiman Huang
- Jiaqi Fu
- Xu Chen
- Lei Gao
- Xiaoyou Su
journal: Frontiers in Public Health
year: 2023
pmcid: PMC9998917
doi: 10.3389/fpubh.2023.1109158
license: CC BY 4.0
---
# Development and validation of the diabetic self-management scale based on information-motivation-behavioral skills theory
## Abstract
### Background
Self-management is important for the blood sugar control of middle-aged and elderly Type 2 diabetes mellitus (T2DM) patients, of which diet, exercise, and drug compliance are the most common components. The Information-Motivation-Behavioral Skills Model (IMB) has been widely used in health behavior management and intervention.
### Objective
The purpose of this study is to develop and validate the Diabetic Self-Management Scale (DSMS) based on the IMB model.
### Methods
Self-report survey data was collected from middle-aged and elderly T2DM patients in Zhongmu City, Henan Province, China in November 2021 using convenience sampling. The original DSMS was developed through a literature review and summary of previous similar scales using an inductive approach. Item modification was finished by a panel of specialists. Exploratory factor analysis and confirmatory factor analysis were used to evaluate the reliability, convergent validity, discriminant validity, and criterion validity of DSMS.
### Results
Four hundred and sixty nine T2DM patients completed the questionnaire survey. The final DSMS consists of 22 items with three dimensions, including information (five items), motivation (eight items), and behavior skills (nine items). The results of simple factor analysis showed that the KMO value was 0.839, Bartlett spherical test 2 = 3254.872, $P \leq 0.001.$ The results of confirmatory factor analysis showed that 2/df = 2.261, RMSEA = 0.073, CFI = 0.937, TLI = 0.930, and SRMR = 0.096. The standardized factor loadings of 22 DSMS items were all above 0.6, and the CR values of 3 dimensions were all higher than 0.9. In addition, DSMS also showed good discriminant and criterion validity.
### Conclusion
The 22-item DSMS has good reliability and validity, and can be used to make diabetic self-management assessment regarding diet, physical activity, and medication among middle-aged and elderly Chinese T2DM patients. DSMS is of moderate length and easy to understand. It can be promoted in China in the future to understand the self-management status of middle-aged and elderly T2DM patients in China.
## 1. Introduction
With the increasing rates of aging and obesity worldwide, diabetes has become one of the most serious and common chronic diseases currently [1]. According to International Diabetes Federation (IDF), there were about 537 million people suffering from diabetes in 2021 worldwide, with a global prevalence estimated to be more than $10\%$, and the number is expected to reach 783 million by 2045 [2]. Due to the large aging population, China has now become the country with the highest prevalence of diabetes in the world and has the largest number of diabetic patients [3]. According to the 2021 Global Diabetes Atlas released by International Diabetes Federation (IDF), the total number of diabetic patients in mainland China was estimated to be 140.9 million in 2021 [2]. The incidence of diabetes continues to increase with age, especially after the age of 50, and among the middle-aged and elderly populations in China, the prevalence of diabetes and pre-diabetes was more than 10 and $40\%$ respectively in 2018 (4–6). Worse still, the high incidence of diabetes in middle-and low-income countries has brought huge costs and burdens to the global health economy [7]. Empirical data showed that in 2021, the direct health expenditure caused by diabetes has reached nearly 1 trillion dollars, which has increased about $316\%$ over the past 15 years globally [2]. Moreover, the socio-economic inequities in diabetes are also not conducive to the prevention and control of diabetes worldwide, especially in developing countries [8]. Type 2 diabetes mellitus (T2DM) is the most common type of diabetes mellitus, accounting for $90\%$ of all diabetic patients [9]. T2DM patients may develop microvascular and macrovascular complications such as cardiovascular disease, diabetes nephropathy, and diabetes ophthalmopathy without effective control of blood sugar [10]. In addition, T2DM patients have a $15\%$ increased risk of all-cause death compared with healthy people [11]. This will lead to both the compromise of life quality of T2DM patients and huge financial burden to their families [12].
The main contributing factors for T2DM were obesity and unhealthy lifestyles like sedentariness [13]. Early control of T2DM and patient-centered self-management can reduce blood glucose levels and minimize complications [9]. A series of randomized controlled trials indicated that lifestyle interventions, such as increasing physical activity and having a healthy diet, are simple and effective ways to control the progression of T2DM (14–18). Apart from the lifestyle changes recommended by their family, friends or doctors, most of the T2DM patients are benefiting from the diabetes medication, typically given to the control of the blood glucose and further occurrence of complications of T2DM. Hence, adhering to a doctor's prescription for hypoglycemic medications and suggestions is crucial for managing the condition and preventing the emergence of T2DM complications [4, 19, 20]. However, the self-management practice of T2DM patients need further optimization and refinement, since it involves multiple aspects, such as eating habits, physical activities, and medication adherence, and might be complicated for T2DM patients to follow it strictly [21]. Recent studies have found that most T2DM patients only adhere to their treatment to a moderate degree (22–24). In this case, an instrument to systematically assess eating habits, physical activities, and medication adherence is necessary for T2DM patients to evaluate their capacity for self-management. Currently, most self-management scales employed among Chinese T2DM patients were introduced from other countries, such as the Diabetes Self-Care Scale (DSCS) which evaluates the self-management ability from the perspective of diet, blood sugar detection, feet care, physical activity, and medication; Summary of Diabetes Self-Care Activities Scale (SDSCA) that contains the assessment of diet, physical activity, blood sugar detection, feet care, and smoking; The Personal Diabetes Questionnaire (PDQ) that was developed under the structure of knowledge, self-decision, self-management, and psychology; and Diabetes Care Profiles (DCP) that evaluate the mental and social health of diabetic patients from multiple dimensions (25–28). However, few studies focused on the development and validation of a self-management scale for T2DM patients in China considering the culture of the Chinese context. Additionally, the existing self-management evaluation tools for diabetes often involve multiple dimensions such as diet, physical activity, blood glucose monitoring, feet care, and drug compliance. Among them, diet, physical activity and drug treatment are of utmost concerns by majority of the diabetes and prediabetes patients, and it is also the focus of medical staff [6]. Therefore, developing a tool focusing on the above mentioned three aspects to evaluate the self-management status of diabetes patients might be expected to provide a measurement applicable to a broader T2DM population.
The Information-Motivation-Behavioral Skills Model (IMB) developed by Fisher et al. in the 1990's was originally used to evaluate the risk of HIV infection and promote the prevention of HIV/AIDS [29]. According to the IMB theory, the performance of behavior requires behavior-specific information, motivation, and behavior skills. Individuals with higher levels of information, motivation and behavioral skills are more likely to adopt healthy behaviors. Therefore, measuring the level of the above three dimensions can well-predict and reflect individual's behavior. Information is a factor directly related to health-related behavior, and motivation is an additional determinant of health-related behavior. Adequate information and motivation can promote individuals to develop appropriate behavioral skills and ultimately lead to health-promoting behaviors (Figure 1) [30, 31]. At present, the IMB model has demonstrated a satisfactory predictive ability to improve the compliance of self-management behavior and ameliorate the health outcome of T2DM patients, showing good practicability and maturity [32]. For example, Qin compiled the IMB-SMBG questionnaire based on the IMB model to investigate the self-monitoring of blood glucose in adult type I diabetes patients [33]. However, the diabetic self-management scales employed in these studies were mostly self-designed without strict validation, and few studies focused on middle-aged and elderly T2DM patients in China [34]. Therefore, the purpose of the current study is to develop a diabetic self-management scale (DSMS) in Chinese middle-aged and elderly T2DM patients based on IMB theory. And it is intended to evaluate the self-management of T2DM patients from the perspective of diet, physical activity, and medication, respectively, and to provide a tool for systematically understanding of their relevant health-related behaviors.
**Figure 1:** *Conceptual framework of the role of IMB model on health behavior.*
## 2.1. Scale development
The Diabetic Self-Management Scale (DSMS) developed in this study was based on the three dimensions of IMB theory, namely information, motivation, and behavioral skills. And each dimension covered items on physical activity, diet, and medication of T2DM. The information dimension was developed based on the items from the diabetes knowledge questionnaires of the Diabetes Education Project in China of Project Hope, while items of motivation and behavioral skill dimensions were developed based on comprehensive literature reading and other scales that our research team had previously developed and employed based on IMB or similar models [35, 36]. Initially, a 55-item pool (15 items of information, 20 items of motivation, and 20 items of behavioral skill) was generated via literature review and group discussion. Then, similar items were collapsed into 1 item to avoid redundancy. For example, “Appropriate physical activities can lower my blood sugar level,” “Physical activities can make my blood sugar well controlled”, and “Blood sugar level is difficult to control if I do not exercise.” were collapsed into “Appropriate physical activities can lower my blood sugar level.” Meanwhile, some items were also removed by considering the practicability and applicability in the study population by investigators and field workers who were familiar with the study population and the local culture of the study site, such as “Do you know what your ideal weight is?.” *After this* phase, the 55-item pool was reduced to 43 items (13 items of information, 15 items of motivation, and 15 items of behavioral skill). Later, a panel of specialists in epidemiology, psychology, and behavioral science were invited to further review the 43-item pool, evaluating the face validity of the items, and to make the final modification suggestion. According to the suggestions of specialists, some unnecessary items were removed, such as “High-fat food will increase the risk of complications of diabetes,” and “I will feel anxious if my blood sugar level cannot be well-controlled,” and a few items were slightly reworded to improve their linguistic clarity. Finally, a total of 25 items (eight items of information, eight items of motivation, and nine items of behavioral skill) remained for subsequent analyses.
## 2.2. Study design and participants
The sample size was planned to be at least 120 in this study, with the set of α = 0.05; β = 0.2; degree of freedom (df) = 120; RMSEA = 0.05 in the null hypothesis; RMSEA = 0.08 in the test of close fit, and RMSEA = 0.01 in the test of non-close fit [37]. The calculation was completed in R 4.2.2.
A descriptive cross-sectional questionnaire survey was conducted in the present study. Participants were recruited from Zhongmu, Henan province, China, and they were invited to fill out a questionnaire including demographics and DSMS by convenience sampling from November 2nd, 2021 to November 12nd, 2021. The inclusion criteria were: [1] Registered clinically diagnosed diabetic patients aged from 45 to 65 years old; [2] Fasting blood glucose level is not lower than 7.0 mmol/L or HbA1c is not lower than $6.5\%$; [3] Can independently finish questionnaires; [4] Can sign the informed consent form and cooperate to complete all the research contents. The exclusion criteria were: [1] Patients with serious diseases (such as malignant tumors), immunodeficiency or immunosuppressants, or those with severe neurological or mental disorders; [2] Patients who are deaf-mute, unable to move, etc. Investigators who are familiar with the local dialect were recruited and trained. Unified instructions were set for each item in the questionnaire for the investigators to ask questions, and they would fill out the questionnaire according to the answers of the participants. After the investigators and proofreaders sign at the end of each questionnaire, it is deemed that the investigation of this sample is completed. Epidata software was used for data entry and double check to ensure the accuracy of the data. In this study, 484 participants completed the questionnaires, and 469 out of them met the eligibility criteria of the study, which were employed for subsequent analysis, with an effective recovery rate of $97\%$. The study protocol was approved by the Ethics Committee of the Institute of Pathogen Biology, Chinese Academy of Medical Sciences (Beijing, China) (IPB-2021-09).
## 2.3.1. Demographic information
Demographic information included age, gender, educational level, marital status, annual household income in 2020, whether drank in the past year, whether smoked in the last 6 months, whether have high blood pressure, blood glucose (GLU), and glycosylated hemoglobin or glycated hemoglobin (GHB). ( GLU and GHB are the current diagnostic criteria for diabetes in China) [38]. GLU and GHB were measured by researchers during investigation.
## 2.3.2. Preliminary version of diabetic self-management scale
The DSMS developed in this study was based on IMB theory, which included three dimensions: information, motivation, and behavioral skill. The information dimension consisted of eight items on knowledge of diet, physical activity, and medication of T2DM. Each item was of dichotomous response on “Yes” and “No” (“Yes” equals 1 while “No” equals 0), participants would receive 1-point for each correct response, and the higher total score of this dimension indicated a higher level of knowledge on T2DM. The Cronbach's α of this dimension was 0.630 in this study. The motivation dimension consisted of eight items and each item was 5-point Likert scaled from 1–5, and higher total scores indicated a higher level of motivation on diabetic self-management. The Cronbach's α of this dimension was 0.938 in this study. The behavioral skill dimension consisted of nine items and each item was a question that need an answer from “Yes,” “No,” or “Not clear.” Participants would receive 1-point for each correct response and get the total score after finishing all items. Higher total scores indicated a higher level of behavioral skill. The Cronbach's α of this dimension was 0.898 in this study.
## 2.4. Statistical analysis
Descriptive analysis was used to describe the demographic characteristics. Pearson correlation analysis was employed to examine the correlation among the 3 dimensions of DSMS. When assessing the psychometric properties of DSMS, the sample was randomly divided into two parts via a random number generator, to perform exploratory factor analysis (EFA) and confirmatory factor analysis (CFA), respectively. The final sample size of the EFA sample (sample 1) was 234 and the CFA sample (sample 2) was 235. Kaiser–Meyer–Olkin (KMO) and Bartlett's test of sphericity was used to test whether our data were suitable for factor analysis. In EFA, principal component factor analysis with varimax rotation was conducted to assess the underlying structure for the 25 items of the DSMS. Items that had a factor loading of more than 0.50 and did not load on multiple factors were obtained for further CFA [39]. After EFA, three-factor CFA with oblique rotation was employed to evaluate the reliability and validity of the DSMS. Since the scales of indicators of “information” and “behavioral skill” were binary, mean and variance-adjust weight least squares (WLSMV) was used to estimate the parameters of the CFA model [40]. The structural validity of the DSMS was evaluated by model fit indices, which include χ2, df, Root Mean Square Error of Approximation (RMSEA), Comparative Fit Index (CFI), Tucker-Lewis Index (TLI), and Standardized Root Mean Square Residual (SRMR). The reliability and convergent validity of the DSMS were assessed by standardized factor loadings, composite reliability (CR), and average variance extracted (AVE) [41]. The discriminant validity of the DSMS was assessed by the AVE method [41]. The criterion validity was assessed by the correlation among the three dimensions of DSMS and GLU and GHB [42]. All the analyses were completed with SAS 9.4 and Mplus 8.3.
## 3.1. Participant characteristics
A total of 469 T2DM patients were recruited in this study. $28.4\%$ of the participants were above 60 years old, $59.9\%$ were female, and $78.7\%$ had high blood pressure (see Table 1).
**Table 1**
| Variables | N (%) |
| --- | --- |
| Age (years) | |
| 45–60 | 336 (71.6) |
| >60 | 133 (28.4) |
| Gender | |
| Male | 188 (40.1) |
| Female | 281 (59.9) |
| Educational level | |
| Primary school or below | 265 (56.5) |
| Above primary school | 204 (43.5) |
| Marital status | |
| unmarried/divorced/widowed | 32 (6.8) |
| Married | 437 (93.2) |
| Annual household income in 2020 (RMB) | |
| ≤ 30,000 | 294 (62.7) |
| >30,000 | 175 (37.3) |
| Drink in the past year | |
| No | 363 (77.4) |
| Yes | 106 (22.6) |
| Smoke in the last 6 months | |
| No | 395 (84.2) |
| Yes | 74 (15.8) |
| High blood pressure | |
| No | 100 (21.3) |
| Yes | 369 (78.7) |
| GLU (mmol/L) | 9.93 ± 4.36 |
| GHB (%) | 10.27 ± 2.62 |
## 3.2. Exploratory factor analysis
The KMO measure of the 25-item original DSMS was 0.839, indicating enough items are predicted by each factor in the current study. The result of Bartlett's test of sphericity was statistically significant (χ2 = 3254.872, $P \leq 0.001$), suggesting that the items are correlated highly enough to provide a reasonable basis for factor analysis. In EFA, 3 major factors were expected, based on the fact that the items were designed to index 3 constructs: information, motivation, and behavioral skill. Table 2 displays the items and factor loadings for the rotated factors, with loadings < 0.50 omitted in the further analysis [39]. Three items in the information dimension were deleted: “Eating too much sugar or sweet food is a cause of diabetes,” “People with diabetes cannot eat fruits and vegetables,” “If I forget to take the hypoglycemic drugs in the morning, then I can take the two drugs together at noon to make up for the morning's vacancy.”
**Table 2**
| Items and expected dimensions | Dimension | Dimension.1 | Dimension.2 |
| --- | --- | --- | --- |
| | Information | Motivation | Behavioral skill |
| Information | Information | Information | Information |
| Eating too much sugar or sweet food is a cause of diabetes. | 0.279 | −0.148 | 0.277 |
| People with diabetes cannot eat fruits and vegetables. | 0.122 | −0.139 | −0.077 |
| Appropriate physical activities can lower my blood sugar level | 0.546 | 0.198 | 0.046 |
| Diabetic patients should start exercising 1/2 to 1 h after a meal | 0.703 | 0.157 | 0.167 |
| Diabetic patients should take sweets with them when exercising | 0.696 | −0.068 | 0.148 |
| If I forget to take the hypoglycemic drugs in the morning, then I can take the two drugs together at noon to make up for the morning's vacancy | 0.411 | 0.127 | −0.253 |
| All hypoglycemic agents may cause hypoglycemia | 0.563 | 0.051 | 0.082 |
| Paying attention to diet and strengthening physical activities are as important as taking hypoglycemic agents | 0.591 | 0.294 | −0.061 |
| Motivation | Motivation | Motivation | Motivation |
| I attach great importance to my health | 0.177 | 0.616 | −0.198 |
| It will be difficult to control my blood sugar if I do not control my diet | −0.038 | 0.741 | 0.119 |
| Poor control of blood sugar can easily lead to diabetic complications (such as nephropathy, etc.) | 0.168 | 0.784 | −0.098 |
| My relatives and friends around me think I should stick to the diabetes diet | 0.081 | 0.826 | 0.161 |
| Moderate physical activities can control my blood sugar well | 0.078 | 0.817 | 0.128 |
| My relatives and friends around me think I should keep moderate physical activity to control my blood sugar | 0.079 | 0.849 | 0.184 |
| Medication according to the doctor's advice can control my blood sugar well | 0.149 | 0.824 | −0.062 |
| My relatives and friends around me think that I should stick to the doctor's advice | 0.160 | 0.845 | 0.157 |
| Behavioral skill | Behavioral skill | Behavioral skill | Behavioral skill |
| When you are busy, can you still follow the dietary principles suggested by your doctor? | 0.020 | 0.039 | 0.756 |
| If you make up your mind, can you stick to the diabetes diet? | 0.021 | −0.011 | 0.800 |
| Do you know how to eat and drink to help control blood sugar? | 0.117 | 0.159 | 0.667 |
| When you are busy, can you still keep exercising? | 0.059 | 0.090 | 0.820 |
| If you make up your mind, can you keep exercising? | 0.108 | 0.105 | 0.795 |
| Do you know how to exercise to help control blood sugar? | 0.163 | 0.110 | 0.755 |
| Can you take the medicine according to the doctor's advice during the period of taking the medicine recommended by the doctor? | −0.135 | −0.043 | 0.659 |
| If you make up your mind, can you follow the doctor's advice? | −0.075 | 0.048 | 0.588 |
| Do you know how you should take medicine to control blood sugar? | 0.085 | 0.047 | 0.796 |
## 3.3.1. Structural validity
The CFA model with 22 items extracted from the EFA showed χ2/df = 2.261, RMSEA = 0.073, CFI = 0.937, TLI = 0.930, and SRMR = 0.096, suggesting an acceptable model fit and structural validity of the DSMS. The model fit indices of the CFA are illustrated in Table 3.
**Table 3**
| Unnamed: 0 | χ2 | df | χ2/df | CFI | TLI | RMSEA | SRMR |
| --- | --- | --- | --- | --- | --- | --- | --- |
| Model | 465.681 | 206 | 2.261 | 0.937 | 0.93 | 0.073 | 0.096 |
## 3.3.2. Reliability and convergent validity
The standardized factor loadings of 22-item DSMS were all above 0.6 and statistically significant. The values of CRs for the three dimensions were all above 0.9. The AVEs of the dimensions were above 0.6, suggesting good reliability and convergent validity of the DSMS (see Table 4, Figure 2).
## 3.3.3. Discriminant and criterion validity
To evaluate discriminant and criterion validity, the correlations of 3 dimensions were examined. When the square root of each factor's AVE is greater than the absolute value of the correlation between this dimension and the other two dimensions, the model demonstrates discriminant validity. As is shown in Table 5, the diagonal elements in the correlation matrix of DSMS factors were the square root of AVE. All the diagonal elements were greater than corresponding off-diagonal elements, indicating the DSMS showed good discriminant validity. Additionally, criterion validity was evaluated by the correlation between 3 dimensions of DSMS and GLU, and GHB. As shown in Table 5, three dimensions of DSMS were all negatively associated with GLU and GHB, except the correlation between behavioral skill and GHB, but this correlation was not statistically significant. Additionally, despite the non-significant correlation between motivation and GLU, the correlation between motivation and GHB were statistically significant, indicating that DSMS developed in this study had good criterion validity.
**Table 5**
| Unnamed: 0 | Information | Motivation | Behavioral skill |
| --- | --- | --- | --- |
| Information | 0.815 | | |
| Motivation | 0.365* | 0.838 | |
| Behavioral skill | 0.177* | 0.130* | 0.909 |
| GLU | −0.196* | −0.068 | −0.148* |
| GHB | −0.091* | −0.171* | 0.015 |
## 3.4. Characteristics of dimensions in DSMS
The average scores of information, motivation, and behavioral skill for all participants were (2.697 ± 1.607), (4.073 ± 0.517), and (6.058 ± 3.033) (see Table 6).
**Table 6**
| Unnamed: 0 | Min | Max | Mean | SD |
| --- | --- | --- | --- | --- |
| Information | 0.0 | 5.0 | 2.6972 | 1.60728 |
| Motivation | 1.0 | 5.0 | 4.0728 | 0.5172 |
| Behavioral skill | 0.0 | 9.0 | 6.0576 | 3.0331 |
## 4. Discussion
T2DM is a chronic disease requiring patients' lifelong self-management to avoid the occurrence of complications and to ensure the achievement of the best clinical outcomes [43, 44]. Recognizing the weak links and misconceptions of self-management of T2DM patients in *China is* of positive significance for the health authorities and medical care workers to implement targeted intervention among this population. Therefore, it is necessary to develop a scale to comprehensively assess the self-management status of T2DM patients. The primary aim of this study was to develop a scale to evaluate the self-management ability, including diet, physical activity, and medication, of middle-aged and elderly T2DM patients in China based on IMB theory. The final DSMS is a 22-item scale with a systematic yet simple 3-factor structure encapsulating information, motivation, and behavioral skill, which is consistent with the established theoretical perspectives of IMB [34]. The standardized factor loadings of 22 items in CFA were all above the recommended value of 0.5 and statistically significant, indicating good communalities of items [39]. The value of Cronbach's α and CR for the three dimensions were all above 0.7, indicating good reliability and convergent validity of the CFA model [41, 45]. In addition, the discriminant validity of the CFA model was acceptable according to the comparison between correlation coefficients and AVEs. These findings provide reasonable evidence that DSMS has satisfied psychometric properties to meet the requirement for a self-report measure of self-management ability among Chinese middle-aged and elderly T2DM patients.
Blood sugar is an important indicator to evaluate the self-management of T2DM patients [46, 47]. The three dimensions of DSMS were all negatively associated with hyperglycemia markers, such as higher levels of GHB and GLU. The first dimension is information, which is the precondition for healthy behavior in the IMB model. Researchers have pointed out that information can directly affect diabetes patients' self-management behavior, and a high level of diabetes-related knowledge is beneficial for patients' glycemic control [34, 48]. In addition, sufficient diabetes-related information can improve drug compliance and regular glycemic monitoring of T2DM patients, which will also help to avoid the deterioration of the disease [32, 49]. The second dimension is motivation, including personal motivation and social motivation, which are independent and direct predictors of T2DM self-management behavior [50]. Researches indicated that positive motivation can promote physical activity and a healthy diet in T2DM patients [49, 51]. However, the relationship between motivation and blood glucose monitoring and drug compliance has not been identified, since these behaviors may be more strictly limited by information such as doctor's advice, while the diet and physical activity are more flexible and can be easily adjusted according to the patient's motivation [52]. The last part of the IMB model is behavior skills. Behavior skills are composed of personal objective skills and self-efficacy, which are also positively associated with diabetes self-management behavior [53]. Previous studies have shown that information and motivation are positively associated with behavioral skills, and behavioral skills are also positively associated with T2DM self-care behaviors [54]. Although information and motivation are essential for self-management behaviors in T2DM patients, it might be difficult to adopt correct healthy behaviors without solid practical skills [50]. In addition, HIV-related intervention studies suggested that behavioral skills mediate the relationship among information, motivation, and health behaviors [55, 56]. Therefore, DSMS developed in this study might give a direct evaluation of the health outcome of T2DM self-management. Based on the DSMS, health authorities and medical care workers can understand the factors of poor self-management behavior of T2DM patients and then take targeted health education and intervention.
In this study, all the items in the dimensions of motivation and behavioral skill of the preliminary DSMS have been completed and retained, indicating that the items of these two dimensions could well-assess the self-management of T2DM patients. However, in EFA, we found three redundant items and deleted them in the “Information” dimension from the original version of DSMS due to their low factor loadings and poor interpretability to the whole DSMS. Since the launch of the new round of health system reform in China in 2009, community-based diabetes management and care has become one of the key contents of the country's basic public health services, including regular blood glucose testing, the guidance of medication, diet control, and physical activity, which has helped to improve the knowledge and self-care awareness level of T2DM patients (57–59). In this case, the deleted three items in “Information” dimension were more like common sense for T2DM patients, which might be too simple for them, thus demonstrating low consistency with other items.
This study has several strengths. First, the 22-item scale is relatively appropriate for the middle-aged and elderly and is comparable in length to other widely used measures, such as DSCS which has 26 items [26]. Second, researchers in this study had background in diabetic epidemiology and were familiar with psychology and behavioral science, which could ensure the face and content validity of the scale. Third, during the scale-development process, we employed the IMB model as our theoretical framework and considered the items from aspects of diet, physical activity, and medication, which provided a systematic and comprehensive assessment for the T2DM. Moreover, all 469 T2DM patients fully completed the questionnaires, indicating the acceptability of the scale, and the sample size was large enough for the psychometric testing.
This study also has several limitations. First, this study is limited by the convenience sampling method and the fact that it consisted only of middle-aged and elderly T2DM patients in Zhongmu, Henan, China. Further studies including T2DM patients of different ages and regions are necessary to examine the validity of DSMS in the Chinese context. Second, the CR values of the three subscales were all above 0.9, despite displaying good reliability of the scale, it suggested that further scale reduction might be applicable, however, further deletion may reduce the face and content validity of the scale. Third, this study may also be limited by the fact that the DSMS is a self-report instrument. Participants may give a subjective or socially appropriate answer that does not reflect their true thoughts. Further comprehensive investigation with the additional use of qualitative methods may be valuable.
## 5. Conclusion
Currently, no instrument can be used to systematically assess the self-management readiness of middle-aged and elderly T2DM patients in China. The 22-item DSMS developed in our study is an important step toward closing this gap, and can be used to make comprehensive assessment of diabetic self-management regarding diet, physical activity, and medication, based on IMB theory. The DSMS is validated with good reliability and validity, with moderate length and understandable content for middle-aged and elderly T2DM patients in China. Thus, the DSMS can be applied in China to identify levels of self-management among middle-aged and elderly T2DM patients in China.
## Data availability statement
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.
## Ethics statement
The studies involving human participants were reviewed and approved by the Ethics Committee of the Institute of Pathogen Biology, Chinese Academy of Medical Sciences (Beijing, China) (IPB-2021-09). The patients/participants provided their written informed consent to participate in this study.
## Author contributions
XS, ZD, SJ, and XL prepared the first draft. XS provided overall guidance and managed the overall project. ZD, SJ, XL, HZ, YW, HW, WX, YH, JF, XC, and LG were responsible for the questionnaire survey, intervention implementation, and data analysis. All authors contributed to the article and approved the submitted version.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher's note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## References
1. Tinajero MG, Malik VS. **An update on the epidemiology of type 2 diabetes: a global perspective**. *Endocrinol Metab Clin North Am.* (2021) **50** 337-55. DOI: 10.1016/j.ecl.2021.05.013
2. Sun H, Saeedi P, Karuranga S, Pinkepank M, Ogurtsova K, Duncan BB. **IDF Diabetes Atlas: global, regional, and country-level diabetes prevalence estimates for 2021 and projections for 2045**. *Diabetes Res Clin Pract.* (2022) **183** 109119. DOI: 10.1016/j.diabres.2021.109119
3. Tu WJ, Xue Y, Nie D. **The prevalence and treatment of diabetes in China from 2013 to 2018**. *JAMA.* (2022) **327** 1706. DOI: 10.1001/jama.2022.3014
4. Wang L, Peng W, Zhao Z, Zhang M, Shi Z, Song Z. **Prevalence and treatment of diabetes in China, 2013–2018**. *JAMA.* (2021) **326** 2498-506. DOI: 10.1001/jama.2021.22208
5. Saeedi P, Petersohn I, Salpea P, Malanda B, Karuranga S, Unwin N. **Global and regional diabetes prevalence estimates for 2019 and projections for 2030 and 2045: results from the international diabetes federation diabetes atlas, 9(th) edition**. *Diabetes Res Clin Pract.* (2019) **157** 107843. DOI: 10.1016/j.diabres.2019.107843
6. Li YZ, Teng D, Shi XG, Qin GJ, Qin YF, Quan HB. **Prevalence of diabetes recorded in mainland China using 2018 diagnostic criteria from the American diabetes association: national cross sectional study**. *Bmj-Brit Med J* (2020) **369** 997. DOI: 10.1136/bmj.m997
7. Liu C, Wang B, Liu S, Li S, Zhang K, Luo B. **Type 2 diabetes attributable to PM2**. *Environ Int.* (2021) **156** 106725. DOI: 10.1016/j.envint.2021.106725
8. Cao G, Cui Z, Ma Q, Wang C, Xu Y, Sun H. **Changes in health inequalities for patients with diabetes among middle-aged and elderly in China from 2011 to 2015**. *BMC Health Serv Res.* (2020) **20** 719. DOI: 10.1186/s12913-020-05609-4
9. Wu Y, Ding Y, Tanaka Y, Zhang W. **Risk factors contributing to type 2 diabetes and recent advances in the treatment and prevention**. *Int J Med Sci.* (2014) **11** 1185-200. DOI: 10.7150/ijms.10001
10. Kautzky-Willer A, Harreiter J, Pacini G. **Sex and gender differences in risk, pathophysiology and complications of type 2 diabetes mellitus**. *Endocr Rev.* (2016) **37** 278-316. DOI: 10.1210/er.2015-1137
11. Tancredi M, Rosengren A, Svensson AM, Kosiborod M, Pivodic A, Gudbjornsdottir S. **Excess mortality among persons with type 2 diabetes**. *N Engl J Med.* (2015) **373** 1720-32. DOI: 10.1056/NEJMoa1504347
12. Chatterjee S, Khunti K, Davies MJ. **Type 2 diabetes**. *Lancet.* (2017) **389** 2239-51. DOI: 10.1016/S0140-6736(17)30058-2
13. Zheng Y, Ley SH, Hu FB. **Global aetiology and epidemiology of type 2 diabetes mellitus and its complications**. *Nat Rev Endocrinol.* (2018) **14** 88-98. DOI: 10.1038/nrendo.2017.151
14. Li G, Zhang P, Wang J, Gregg EW, Yang W, Gong Q. **The long-term effect of lifestyle interventions to prevent diabetes in the China Da Qing diabetes prevention study: a 20-year follow-up study**. *Lancet.* (2008) **371** 1783-9. DOI: 10.1016/S0140-6736(08)60766-7
15. Gong Q, Gregg EW, Wang J, An Y, Zhang P, Yang W. **Long-term effects of a randomised trial of a 6-year lifestyle intervention in impaired glucose tolerance on diabetes-related microvascular complications: the China Da Qing diabetes prevention outcome study**. *Diabetologia.* (2011) **54** 300-7. DOI: 10.1007/s00125-010-1948-9
16. Gong Q, Zhang P, Wang J, Ma J, An Y, Chen Y. **Morbidity and mortality after lifestyle intervention for people with impaired glucose tolerance: 30-year results of the Da Qing diabetes prevention outcome study**. *Lancet Diabetes Endocrinol.* (2019) **7** 452-61. DOI: 10.1016/S2213-8587(19)30093-2
17. Knowler WC, Barrett-Connor E, Fowler SE, Hamman RF, Lachin JM, Walker EA. **Reduction in the incidence of type 2 diabetes with lifestyle intervention or metformin**. *N Engl J Med.* (2002) **346** 393-403. DOI: 10.1056/NEJMoa012512
18. Ojo O. **Dietary intake and type 2 diabetes**. *Nutrients* (2019) **11** 2177. DOI: 10.3390/nu11092177
19. Yu CG, Fu Y, Fang Y, Zhang N, Sun RX, Zhao D. **Fighting type-2 diabetes: present and future perspectives**. *Curr Med Chem.* (2019) **26** 1891-907. DOI: 10.2174/0929867324666171009115356
20. Gloyn AL, Drucker DJ. **Precision medicine in the management of type 2 diabetes**. *Lancet Diabetes Endocrinol.* (2018) **6** 891-900. DOI: 10.1016/S2213-8587(18)30052-4
21. Yuksel M, Bektas H. **Compliance with treatment and fear of hypoglycaemia in patients with type 2 diabetes**. *J Clin Nurs.* (2021) **30** 1773-86. DOI: 10.1111/jocn.15736
22. Zkaptan BB, Kapucu S, Demirci N. **Relationship between adherence to treatment and acceptance of illness in patients with type 2 diabetes**. *Cukurova Med J.* (2019) **3** 25-8
23. American Diabetes A. **Standards of medical care in diabetes-2018 abridged for primary care providers**. *Clin Diabetes.* (2018) **36** 14-37. DOI: 10.2337/cd17-0119
24. Bussell JK, Cha E, Grant YE, Schwartz DD, Young LA. **Ways health care providers can promote better medication adherence**. *Clin Diabetes.* (2017) **35** 171-7. DOI: 10.2337/cd016-0029
25. Fitzgerald JT, Davis WK, Connell CM, Hess GE, Funnell MM, Hiss RG. **Development and validation of the diabetes care profile**. *Eval Health Prof* (1996) **19** 208-30. DOI: 10.1177/016327879601900205
26. Lee NP, Fisher WP Jr. **Evaluation of the diabetes self-care scale**. *J Appl Meas.* (2005) **6** 366-81. PMID: 16192661
27. Stetson B, Schlundt D, Rothschild C, Floyd JE, Rogers W, Mokshagundam SP. **Development and validation of the personal diabetes questionnaire (PDQ): a measure of diabetes self-care behaviors, perceptions and barriers**. *Diabetes Res Clin Pract.* (2011) **91** 321-32. DOI: 10.1016/j.diabres.2010.12.002
28. Toobert DJ, Hampson SE, Glasgow RE. **The summary of diabetes self-care activities measure: results from seven studies and a revised scale**. *Diabetes Care.* (2000) **23** 943-50. DOI: 10.2337/diacare.23.7.943
29. Fisher WA, Fisher JD, Harman J, Suls J, Wallston KA. **The information-motivation-behavioral skills model: A general social psychological approach to understanding and promoting health behavior**. *Social Psychological Foundations of Health and Illness* (2003) 82-106. DOI: 10.1002/9780470753552.ch4
30. Jeon E, Park HA. **Development of the IMB model and an evidence-based diabetes self-management mobile application**. *Healthc Inform Res.* (2018) **24** 125-38. DOI: 10.4258/hir.2018.24.2.125
31. Motahari-Tabari NS, Nasiri-Amiri F, Faramarzi M, Shirvani MA, Bakhtiari A, Omidvar S. **The effectiveness of information-motivation-behavioral skills model on self-care practices in early pregnancy to prevent gestational diabetes mellitus in iranian overweight and obese women: a randomized controlled trial**. *Int Q Community Health Educ* (2021) **3** 272684X211020300. DOI: 10.21203/rs.2.24420/v2
32. Nelson LA, Wallston KA, Kripalani S, LeStourgeon LM, Williamson SE, Mayberry LS. **Assessing barriers to diabetes medication adherence using the information-motivation-behavioral skills model**. *Diabetes Res Clin Pract.* (2018) **142** 374-84. DOI: 10.1016/j.diabres.2018.05.046
33. Qin ZY, Yan JH, Yang DZ, Deng HR, Yao B, Weng JP. **Behavioral analysis of chinese adult patients with type 1 diabetes on self-monitoring of blood glucose**. *Chinese Med J-Peking.* (2017) **130** 39-44. DOI: 10.4103/0366-6999.196574
34. Chen Q, Wang H, Wang Y, Wang Z, Zhao D, Cai Y. **Exploring effects of self-management on glycemic control using a modified information-motivation-behavioral skills model in type 2 diabetes mellitus patients in Shanghai, China: A cross-sectional study**. *J Diabetes.* (2018) **10** 734-43. DOI: 10.1111/1753-0407.12655
35. Si M, Jiang Y, Su X, Wang W, Zhang X, Gu X. **Willingness to accept human papillomavirus vaccination and its influencing factors using information-motivation-behavior skills model: a cross-sectional study of female college freshmen in mainland China**. *Cancer Control.* (2021) **28** 10732748211032899. DOI: 10.1177/10732748211032899
36. Wang H, Huang YM, Su XY, Xiao WJ, Si MY, Wang WJ. **Acceptance of the COVID-19 vaccine based on the health belief model: a multicenter national survey among medical care workers in China**. *Hum Vaccin Immunother* (2022) **5** 2076523. DOI: 10.1080/21645515.2022.2076523
37. MacCallum RC, Browne MW, Sugawara HM. *Power Analysis and Determination of Sample Size for Covariance Structure Modeling* (1996). DOI: 10.1037/1082-989X.1.2.130
38. **Clinical guidelines for prevention and treatment of type 2 diabetes mellitus in the elderly in China (2022 edition)**. *Zhonghua Nei Ke Za Zhi* (2022) **61** 12-50. DOI: 10.3760/cma.j.cn112138-20211027-00751
39. Hair JF, Black WC, Babin BJ, Anderson RE. *Multivariate Data Analysis* (2019)
40. Flora DB, Curran PJ. **An empirical evaluation of alternative methods of estimation for confirmatory factor analysis with ordinal data**. *Psychol Methods.* (2004) **9** 466-91. DOI: 10.1037/1082-989X.9.4.466
41. Fornell C, Larcker DF. **Evaluating structural equation models with unobservable variables and measurement error**. *J Mark Res* (1981) **18** 39-50. DOI: 10.1177/002224378101800104
42. Li C. **Yu-ning W, Xiao-mei L. The reliability and validity of Chinese version of the personal diabetes questionnaire**. *Chin J Nurs.* (2014) **49** 734-8. DOI: 10.3761/j.issn.0254-1769.2014.06.024
43. Ausili D, Bulgheroni M, Ballatore P, Specchia C, Ajdini A, Bezze S. **Self-care, quality of life and clinical outcomes of type 2 diabetes patients: an observational cross-sectional study**. *Acta Diabetol.* (2017) **54** 1001-8. DOI: 10.1007/s00592-017-1035-5
44. Weisman A, Fazli GS, Johns A, Booth GL. **Evolving trends in the epidemiology, risk factors, and prevention of type 2 diabetes: a review**. *Can J Cardiol.* (2018) **34** 552-64. DOI: 10.1016/j.cjca.2018.03.002
45. Kline RB. *Principles and Practice of Structural Equation Modeling* (2016)
46. Peng X, Guo X, Li H, Wang D, Liu C, Du Y. **Qualitative exploration of self-management behaviors and influencing factors in patients with type 2 diabetes**. *Front Endocrinol.* (2022) **13** 771293. DOI: 10.3389/fendo.2022.771293
47. Chen L, Chuang LM, Chang CH, Wang CS, Wang IC, Chung Y. **Evaluating self-management behaviors of diabetic patients in a telehealthcare program: longitudinal study over 18 months**. *J Med Internet Res.* (2013) **15** e266. DOI: 10.2196/jmir.2699
48. Al-Qazaz HK, Sulaiman SA, Hassali MA, Shafie AA, Sundram S, Al-Nuri R. **Diabetes knowledge, medication adherence and glycemic control among patients with type 2 diabetes**. *Int J Clin Pharm-Net.* (2011) **33** 1028-35. DOI: 10.1007/s11096-011-9582-2
49. Meunier S, Coulombe S, Beaulieu MD, Cote J, Lesperance F, Chiasson JL. **Longitudinal testing of the information-motivation-behavioral skills model of self-care among adults with type 2 diabetes**. *Patient Educ Couns.* (2016) **99** 1830-6. DOI: 10.1016/j.pec.2016.06.011
50. Gao JL, Wang JL, Zhu YC, Yu JM. **Validation of an information-motivation-behavioral skills model of self-care among Chinese adults with type 2 diabetes**. *Bmc Public Health* (2013) **13** 100. DOI: 10.1186/1471-2458-13-100
51. Austin S, Guay F, Senecal C, Fernet C, Nouwen A. **Longitudinal testing of a dietary self-care motivational model in adolescents with diabetes**. *J Psychosom Res.* (2013) **75** 153-9. DOI: 10.1016/j.jpsychores.2013.04.013
52. Ferrari M, Speight J, Beath A, Browne JL, Mosely K. **The information-motivation-behavioral skills model explains physical activity levels for adults with type 2 diabetes across all weight classes**. *Psychol Health Med.* (2021) **26** 381-94. DOI: 10.1080/13548506.2020.1749292
53. Osborn CY, Egede LE. **Validation of an Information-Motivation-Behavioral Skills model of diabetes self-care (IMB-DSC)**. *Patient Educ Couns.* (2010) **79** 49-54. DOI: 10.1016/j.pec.2009.07.016
54. Yao JJ, Wang HP, Yin X, Yin J, Guo XL, Sun Q. **The association between self-efficacy and self-management behaviors among Chinese patients with type 2 diabetes**. *PLoS ONE* (2019) **14** 4869. DOI: 10.1371/journal.pone.0224869
55. Fisher JD, Fisher WA, Amico KR, Harman JJ. **An information-motivation-behavioral skills model of adherence to antiretroviral therapy**. *Health Psychol.* (2006) **25** 462-73. DOI: 10.1037/0278-6133.25.4.462
56. Qiu TY, Huang J, Wang WX. **Association between Diabetes knowledge and self-efficacy in patients with type 2 diabetes mellitus in china: a cross-sectional study**. *Int J Endocrinol* (2020) **3** 2020. DOI: 10.1155/2020/2393150
57. Jia H, Wang X, Cheng J. **Knowledge, attitudes, and practices associated with diabetic foot prevention among rural adults with diabetes in North China**. *Front Public Health.* (2022) **10** 876105. DOI: 10.3389/fpubh.2022.876105
58. Wang Q, Zhang X, Fang L, Guan Q, Guan L, Li Q. **Prevalence, awareness, treatment and control of diabetes mellitus among middle-aged and elderly people in a rural Chinese population: a cross-sectional study**. *PLoS ONE.* (2018) **13** e0198343. DOI: 10.1371/journal.pone.0198343
59. Yao J, Wang H, Yin J, Shao D, Guo X, Sun Q. **Factors associated with the utilization of community-based diabetes management care: a cross-sectional study in Shandong Province, China**. *BMC Health Serv Res.* (2020) **20** 407. DOI: 10.1186/s12913-020-05292-5
|
---
title: Characteristics of the fecal microbiome and metabolome in older patients with
heart failure and sarcopenia
authors:
- Jieting Peng
- Hui Gong
- Xing Lyu
- Yang Liu
- Shizhen Li
- Shengyu Tan
- Lini Dong
- Xiangyu Zhang
journal: Frontiers in Cellular and Infection Microbiology
year: 2023
pmcid: PMC9998919
doi: 10.3389/fcimb.2023.1127041
license: CC BY 4.0
---
# Characteristics of the fecal microbiome and metabolome in older patients with heart failure and sarcopenia
## Abstract
### Background
Increasing evidence supports that gut microbiota plays an important role in the development of cardiovascular diseases. The prevalence of sarcopenia is increasing in patients with heart failure. Muscle wasting is an independent predictor of death in heart failure patients.
### Aims
In this study, we aimed to explore the characteristics of gut microbiota and metabolites in heart failure patients with or without sarcopenia.
### Methods
Fecal samples of 33 heart failure patients without sarcopenia, 29 heart failure patients with sarcopenia, and 15 controls were collected. The intestinal microbiota was analyzed using 16S rRNA sequencing and the metabolites were detected using the gas chromatography-mass spectrometry method.
### Results
There were significant differences in the overall microbial community structure and diversity between control and heart failure patients with or without sarcopenia. However, no clear clustering of samples was observed in heart failure with and without sarcopenia patients. Several bacterial, particularly Nocardiaceae, Pseudonocardiaceae, Alphaproteobacteria, and Slackia were significantly enriched in the heart failure patients without sarcopenia, while Synergistetes was more abundant in the heart failure patients with sarcopenia. Isobutyric acid, isovaleric acid, and valeric acid were lower in heart failure patients with sarcopenia than that without sarcopenia but lacked significance.
### Conclusions
This study demonstrates that there are differences in the gut microbiota between control individuals and heart failure patients with or without sarcopenia. Modulating the gut microbiota may be a new target for the prevention and treatment of sarcopenia in heart failure patients.
## Introduction
Along with population aging, geriatric syndromes have become a major healthcare problem worldwide. Sarcopenia is one of the most challenging geriatric syndromes with multiple contributing factors. It has been described as an age-related decline in skeletal muscle mass as well as muscle function, which could lead to a decrease in physical capability, falls, disability, and higher mortality in older people (Cruz-Jentoft et al., 2010). On the other hand, chronic heart failure (CHF) remains one of the most important healthcare problems, with high costs and poor outcomes. The prevalence of heart failure (HF) is more than $10\%$ among people aged ≥70 years. Sarcopenia is a common complication of CHF patients and the prevalence of sarcopenia in CHF patients was nearly $20\%$ (Fulster et al., 2013). The prognosis of HF with sarcopenia is worse than that of patients without sarcopenia (Narumi et al., 2015). A prospective cohort study revealed that sarcopenia was significantly linked with a cardiac event in the New York Heart Association Class II–IV group (Zhang et al., 2021). In addition, a multicenter prospective cohort study indicated that sarcopenia was an independent predictor of 1-year mortality in both ejection fraction preserved and reduced HF patients (Konishi et al., 2021).
More than 1000 different bacterial species are observed in the gut. The diversity of the bacterial species plays a major role in maintaining homeostasis (Przewlocka et al., 2020). The composition of gut microbiota is proven to be dynamic due to dietary changes, age, disease, and so on (Marti et al., 2017). The gut microbiome is being increasingly recognized as a modulator of atherosclerosis, hypertension, atrial fibrillation, and myocardial fibrosis and contributes to the development of these diseases (Peng et al., 2018). Accumulating evidences suggest that gut microbiota has a significant impact on skeletal muscle metabolism and is related to the development of HF (Branchereau et al., 2019; Lahiri et al., 2019). Gut microbiota-derived micronutrients and metabolites such as lipopolysaccharide, trimethylamine N-oxide, short-chain fatty acids (SCFAs), and secondary bile acids can reach and act on muscle (Liu et al., 2021). Gut microbiome affects HF by generating bioactive metabolites that can impact host physiology (Tang et al., 2019). Numerous studies have demonstrated the existence of the gut microbiome-muscle axis and heart-gut microbiome axis (Liao et al., 2020; Madan and Mehra, 2020). Germ-free mice had less skeletal muscle mass than pathogen-free mice when fed the same food (Lahiri et al., 2019). Besides, gut microbiota influenced HF mainly by affecting immunity and inflammation (Yuzefpolskaya et al., 2020). Modulating the microbiota through dietary intervention, probiotics or prebiotics supplements, may be a novel strategy against muscle aging and HF.
SCFAs including acetate, propionate, butyrate, isovalerate, valerate, and caproate, are produced by microbial fermentation of fiber and prebiotics in the colon and mediate the interaction among diet, microbiota, and the host (Kimura et al., 2014). A study found that treating germ-free mice with SCFAs partially reversed muscle atrophy induced by dexamethasone (Lahiri et al., 2019). Professional athletes had relatively increased SCFAs compared with more sedentary subjects (Barton et al., 2018). SCFAs influence HF by regulating T-cell proliferation and exerting anti-inflammatory effects (Mamic et al., 2021) and proved to be an efficient energy source during pathological stress in the failing heart (Carley et al., 2021). Meanwhile, SCFAs were reported to have a positive impact on reducing cardiac hypertrophy and pressure overload (Pakhomov and A. Baugh, 2021).
Emerging evidences have shown the association between gut microbiota and sarcopenia. Sarcopenia may accelerate the progression of HF and increase the mortality of patients with HF. However, there has no standard treatment for slowing muscle loss in patients with HF. It is not clear whether alterations in gut microbiota and metabolites are related to sarcopenia in HF patients. This study is the first to explore the changes of gut microbiota composition and SCFAs levels in HF patients with or without sarcopenia.
## Study participants
Seventy-seven patients aged ≥ 65 years old with similar diet and environmental conditions in the Second Xiangya Hospital of Central South University were enrolled in this study. Patients were classified into the following 3 categories: 33 HF patients without sarcopenia (HF group), 29 HF patients with sarcopenia (SHF group), and 15 control individuals (Control group). Sarcopenia was diagnosed according to the Asian Working Group for Sarcopenia 2019 Guidelines (Chen et al., 2020). Low skeletal muscle mass was defined as muscle mass < 7.0 kg/m2 (male) or < 5.7 kg/m2 (female) by bioelectrical impedance analysis using the InBodyS10 body composition analyzer (Chen et al., 2014). Low muscle strength was defined as handgrip strength <28 kg for male and <18 kg for female. Criteria for low physical performance is a 6-m walk speed < 1 m/s. Sarcopenia was defined as low muscle mass plus either diminished muscle strength or physical performance. Exclude subjects included recurrent diarrhea or constipation, unusual dietary habits (vegetarians), edema, those with tumors, diabetes, intestinal inflammation, irritable bowel syndrome, history of intestinal surgery, being treated with antibiotics or probiotics within 1 month. Demographic characteristics and clinical laboratory examinations were documented for all patients. The study was approved by the local Ethics Committee of the Second Xiangya Hospital of Central South University. Written informed consent was obtained from all participants. This study was conducted under the Declaration of Helsinki.
## Fecal samples collection
Fecal samples of the patients were collected within one hour of excretion in the morning during hospitalization and stored at -80°C in preservation tubes before being delivered to the detected center.
## 16S rRNA gene sequencing
Microbial genomic DNA of fecal samples was extracted by the Quant-iT PicoGreen dsDNA Assay Kit (Invitrogen) according to the manufacturer’s instructions. The purity and integrity of total DNA were detected using the NanoDrop NC-2000 spectrophotometer (Thermo Fisher Scientific, United States) and $1.2\%$ agarose gel electrophoresis. Hypervariable regions V3–V4 of the 16S rRNA gene were amplified by polymerase chain reaction (PCR) using the primer ACTCCTACGGGAGGCAGCA and GGACTACHVGGGTWTCTAAT and sequenced by the NovaSeqPE250 according to the manufacturer’s manuals.
## Fecal short-chain fatty acid analysis
SCFAs quantification was evaluated by the gas chromatography-mass spectrometry method (GC-MS). An appropriate stool sample was taken and 50 μL $15\%$ phosphoric acid, 100 μL 125 μg/mL isocaproic acid, and 400 μL ether were added; then placed in a high-throughput tissue grinder and grind at 55 Hz for 60 s twice. The samples were centrifuged and the supernatant was collected. The mixture was briefly vortexed before GC-MS analysis. A Thermo TRACE, 1310-ISQ LT GC-MS system (Thermo, USA) was used to perform the analysis.
## Bioinformatics analysis
The DADA2 plugin is used for quality filtering, denoising, merging and removing chimeras of sequences. Data were filtered by removing low-quality reads. Briefly, the entire sequence was removed if the average quality < 20 and the read length after truncation was $75\%$ lower than the original read length. Primer/Adaptor- containing reads, N-containing reads were also removed to obtain high-quality clean data. After quality control, the raw data were filed, and the optimized sequences were identified, and then UPARSE (version7.1) were used to assign the raw sequences to operational taxonomic units (OTUs) based on OTUs having≥$97\%$ similarity. Each OTU is considered to represent each taxonomic level, i.e., kingdom, phylum, class, order, family, and genus. By comparing the RDP Classifier algorithm (http://rdp.cme.msu.edu/) with the Silva (SSU123) 16S rRNA database, the taxonomy of each 16S rRNA gene sequence was analyzed based on a $70\%$ confidence threshold. Bioinformatics analyses were conducted using QIIME (version 1.9.1). The diversity within the gut community was assessed by alpha diversity and beta diversity. Alpha diversity indices include the Chao1 richness estimator, Shannon, and Inverse Simpson (Chen et al., 2021). Beta diversity includes principal coordinate analysis (PCoA) plots and nonmetric multidimensional scaling (NMDS) which were created to visualize the structural variation of microbial communities between different groups. Additionally, to further detect differentially abundant taxa in the community structure (phylum and genus) between the groups of samples, the linear discriminant analysis effect size (LEfSe) method was used to compare the differences in the taxonomic levels.
## Statistical analyses
Statistical analyses were performed using SPSS 25.0 software and R (3.0.2). Normal distribution data, non-normal distribution data and categorical data were expressed as mean ± standard deviation (SD), median with interquartile range and number (%) respectively. Normal distribution data were analyzed using Student’s t-test between two groups, non-normal distribution data were analyzed using Mann Whitney U-test. Statistical analysis of beta diversity among different groups was conducted using the adonis test. Kruskal-Wallis tests were applied to determine the statistical differences in the relative abundance of OTUs and alpha diversity indexes between different groups ($P \leq 0.05$). Correlations between SCFAs and different microbiota were evaluated using Spearman rank correlation, and presented as a heatmap.
## Demographic and clinical characteristics of the subjects
Clinical characteristics including sex, age, history of smoking and drinking, blood lipid parameters, and history of the disease and medication were obtained from medical records. Body mass index (BMI) in the SHF group was significantly lower than that in the control and HF groups, while ALT and Cr were significantly higher in the HF group than that in the control group. The percentages of diuretic and angiotensin-converting enzyme inhibitor/angiotensin receptor blocker use in the control group were significantly lower than that in the HF and SHF groups. There were no significant differences in sex, age, WBC, RBC, AST, percentages of smoking and drinking, percentages of β-receptor blocker, statin, and calcium channel blocker use among the three groups (Table 1).
**Table 1**
| Unnamed: 0 | HF (n=33) | SHF (n=29) | Control (n=15) | P |
| --- | --- | --- | --- | --- |
| Sex (Male, %) | 24 (72.70%) | 13 (44.80%) | 8 (53.30%) | 0.076 |
| Age (years) | 71.76 ± 7.93 | 75.14 ± 8.18 | 67.67 ± 9.76 | 0.059 |
| BMI (kg/m2) | 24.24 ± 2.81### | 20.27 ± 3.75 | 23.52 ± 3.12# | 0.0 |
| WBC (×10 9/L) | 5.56 (4.83, 6.29) | 5.16 (4.35, 7.25) | 4.90 (4.14, 5.99) | 0.437 |
| RBC (×10 9/L) | 4.05 ± 0.84 | 4.00 ± 0.54 | 4.28 ± 0.42 | 0.3 |
| ALT (IU/L) | 19.80 (16.90, 23.30)* | 15.10 (11.00, 30.70) | 13.10 (11.10, 15.60) | 0.002 |
| AST (IU/L) | 17.60 (11.80, 23.20) | 20.90 (14.90, 36.00) | 15.70 (14.30, 17.65) | 0.055 |
| Cr (μmol/L) | 97.00 (77.20, 112.30)* | 83.00 (63.30, 122.50) | 73.00 (58.00, 89.00) | 0.018 |
| Smoking, n (%) | 14 (42.40%) | 11 (37.90%) | 5 (33.30%) | 0.827 |
| Drinking, n (%) | 11 (32.30%) | 9 (31.00%) | 2 (13.30%) | 0.34 |
| Diuretic, n (%) | 20 (60.60%) | 11 (37.90%) | | |
| ACEI/ARB, n (%) | 25 (75.80%)*** | 22 (75.90%)*** | 2 (13.30%) | 0.0 |
| β-receptor blocker, n (%) | 23 (69.70%) | 25 (86.20%) | 8 (46.70%) | 0.059 |
| Statin, n (%) | 27 (81.80%) | 25 (86.20%) | 11 (73.30%) | 0.577 |
| CCB, n (%) | 6 (18.18%) | 9 (31.03%) | 5 (33.33%) | 0.396 |
The body composition parameters and echocardiographic data of the three groups were compared. There were no significant differences among the control, HF, and SHF groups in the visceral adipose area, body fat, left ventricular end-diastolic diameter (LVEDd), right atrial diameter (RAD), and right ventricular end-diastolic diameter (RVEDd). The HF and SHF groups had significantly higher N-terminal pro-brain natriuretic peptide (NT-proBNP) levels, left atrial diameter (LAD), and lower left ventricular ejection fraction (LVEF) compared to the control group (Table 2). Basic metabolic rate, bone mineral content, upper arm circumference, and SMI were significantly lower in the SHF group when compared to the control and HF groups ($p \leq 0.05$) (Table 2). The results are consistent with the characteristics of HF and sarcopenia patients.
**Table 2**
| Unnamed: 0 | HF (n=33) | SHF (n=29) | Control (n=15) | P |
| --- | --- | --- | --- | --- |
| Visceral adipose area | 84.00 (68.55, 102.80) | 67.30 (55.05, 93.25) | 70.10(51.60, 89.95) | 0.096 |
| Basic metabolic rate | 1397.00 (1268.00, 1515.00) ### | 1151.00 (1073.00, 1298.50) | 1280.00 (1230.00, 1582.00) ## | 0.004 |
| Body fat (%) | 26.33 ± 7.91 | 25.61 ± 9.78 | 25.11 ± 6.85 | 0.758 |
| Bone mineral content | 2.65 ± 0.35### | 2.29 ± 0.40 | 2.66 ± 0.46# | 0.0 |
| Upper arm circumference | 30.13 ± 2.98### | 26.47 ± 2.49 | 29.83 ± 2.96## | 0.0 |
| SMI | 7.42 ± 0.93### | 5.73 ± 0.77 | 7.14 ± 1.02## | 0.0 |
| NT-pro BNP | 1084.00 (372.00,3200.00)*** | 1424.00 (514.00, 4830.00)*** | 73.60 (27.10, 120.00) | 0.0 |
| LVEDd | 49.76 ± 7.72 | 47.66 ± 8.50 | 44.87 ± 3.52 | 0.123 |
| LAD | 39.27 ± 7.25** | 39.14 ± 6.47** | 31.73 ± 6.81 | 0.001 |
| RVEDd | 30.30 ± 3.44 | 30.48 ± 4.56 | 28.13 ± 1.51 | 0.073 |
| RAD | 30.00 (28.00, 36.00) | 31.00 (26.50, 36.50) | 28.00 (27.00, 29.50) | 0.081 |
| LVEF (%) | 57.00 (39.50, 61.50)* | 55.00 (38.00, 60.00)* | 63.00 (60.00, 65.00) | 0.01 |
## Alpha and beta diversity of gut microbiota
The gut microbial composition was compared in three groups. There were significant differences among the control, HF group, and SHF group in diversity (Observed-species index and Chao1) and abundance (Simpson and Shannon indexes) (Figure 1A). These indexes were similar between the HF group and the SHF group. Beta diversity reflects the between-habitat diversity in microbial community structure assessed using Bray and Curtis distances. Axis.1 and Axis.2 explained $7.9\%$ and $6.2\%$ of the variation in microbiota, respectively. Significant separations were found between the control group with HF or SHF group ($$p \leq 0.002$$). No significant differences were observed in microbial community composition between HF and SHF group ($$p \leq 0.641$$) (Figure 1B). These findings suggested gut microbial dysbiosis in HF and SHF groups compared to the control group.
**Figure 1:** *Gut microbial diversity in three groups. (A) Alpha diversity index measured by Observed, Chao1, Shannon, and Simpson methods. The boxplots are showing interquartile (IQR) ranges with the median and whiskers extending up to the most extreme point within 1.5 folds IQR. Figures under the diversity index label are p-values from the Kruskal-Wallis test. (B) Beta diversity index measured by Bray-Curtis method using PCoA based on OTUs relative abundance profile. The two-variance explained by Axis.1 and Axis.2 are 7.9% and 6.2%, respectively. *p <0 .05, **p < 0.01, ***p < 0.001 compared to control group.*
## Gut microbiota composition at phylum and genus levels
As shown in Figure 2A, each ellipse represents a group, the overlapping area between the ellipses indicates the shared OTU between the groups and the number indicates the number of OTUs. There were, 1755 shared OTUs in the three groups, while 6302, 11852, and, 10230 OTUs were unique to the control, HF, and SHF groups respectively. Similar to previous studies, the microbial communities in the human gut mainly belong to Firmicutes, Bacteroidetes, Proteobacteria, Actinobacteria, and Verrucomicrobia (Figure 2B). At the phylum level, Firmicutes and Bacteroidetes were more abundant in the control group than in the HF and SHF group. In the HF and SHF groups, the relative abundance of Proteobacteria and Actinobacteria was higher than that in the control group (Figure 2B). At the genus level, Faecalibacterium, Blautia, and Prevotella were more abundant in the control group than in the HF and SHF groups (Figure 2C). Compared with the SHF group, the HF group displayed a higher abundance of Shigella, Bacteroides, Streptococcus, Gemmiger, and Akkermansia, while lesser in Lactobacillus, Bifidobacterium, Enterococcus (Figure 2C).
**Figure 2:** *Gut microbiota compositional shifts at phylum and genus level. (A) Venn diagram of OTUs shared by and exclusive to the three groups. Each ellipse represents a group, the overlapping area between the ellipses indicates the shared OTU between the groups and the number indicates the number of OTUs. Corresponding percentages are noted for relevant overlaps. The overall bacterial structures of the three groups at (B) phylum and (C) genus levels are expressed as the relative abundance of OTUs in each group (The top 20 phylum and genus with the average OTU frequency of the sample).*
## Different abundance of gut microbiota in the three groups
LEfSe analysis based on the Linear Discriminant Analysis (LDA) score was used to measure the unique microbial characteristics that distinguished each group. The results showed that the Nocardiaceae, Pseudonocardiaceae, Leuconostocaceae, and Alphaproteobacteria, Slackia were significantly more enriched in the HF group when compared to the control group (Figures 3A, B). Meanwhile, the Synergistetes was significantly enriched in the SHF versus the HF group (Figures 4A, B). Furthermore, compared with the HF and SHF groups, the control group had a higher abundance of Clostridia, Faecalibacterium, Peptostreptococcaceae, and Prevotellaceae (Figures 3A, B).
**Figure 3:** *The different abundance of gut microbiota between Control and HF group. Gut microbial markers were measured by LEfSe analysis (cut-off value LDA > 2.5) between the Control and HF group. Histograms (A); Cladogram (B).* **Figure 4:** *The different abundance of gut microbiota between the HF and SHF group. Gut microbial markers were measured by LEfSe analysis (cut-off value LDA > 2.0) between the HF and SHF group. Histograms (A); Cladogram (B).*
## Fecal levels of SCFAs in Control, HF, and SHF groups
The concentrations of SCFAs were assessed in feces. Fecal levels of acetic acid, propionate acid, butyric acid, and caproic acid were similar in the three groups (Table 3). The levels of isobutyric acid, isovaleric acid, and valeric acid were lower in the SHF group compared with that in the HF and control group but has no significant differences (Table 3).
**Table 3**
| Unnamed: 0 | HF (n=33) | SHF (n=29) | Control (n=15) | P |
| --- | --- | --- | --- | --- |
| Acetic acid | 1286.28 ± 661.61 | 1303.31 ± 653.97 | 1301.65 ± 556.98 | 0.97 |
| Propanoic acid | 763.30 ± 494.90 | 646.18 ± 395.50 | 637.59 ± 306.94 | 0.715 |
| Isobutyric acid | 92.05 (53.80, 158.67) | 71.57 (27.85, 130.96) | 104.43 (63.78, 132.84) | 0.453 |
| Butyric acid | 764.32 ± 689.10 | 624.64 ± 451.59 | 502.65 ± 328.88 | 0.689 |
| Isovaleric acid | 108.56 (45.30, 187.76) | 68.41 (26.25, 121.89) | 102.50 (59.15, 133.87) | 0.336 |
| Valeric acid | 116.89 (43.19, 198.58) | 88.64 (14.02, 193.55) | 101.00 (19.04, 154.88) | 0.516 |
| Caproic acid | 4.04 (1.62, 24.29) | 2.86 (0.66, 7.70) | 2.94 (1.40, 5.37) | 0.244 |
## Relationship between the gut microbiota and clinical parameters with SCFAs
An association analysis was conducted to examine whether the microbiota is associated with SCFA production in all subjects. The results showed Tenericutes and Bacteroidetes were positively associated with most of SCFAs production. Verrucomicrobia was positively correlated with Valeric acid and Isovaleric acid. The abundance of Chlamydiae was positively associated with caproic acid (Figure 5). In addition, a positive correlation was found between SMI and bone mineral content (Figure 6) ($r = 0.709$, $p \leq 0.001$).
**Figure 5:** *The correlation between the gut microbiota and SCFAs. The intensity of the colors represents the degree of association between the variables measured and asterisks indicate significant associations. *p < 0.05; **p < 0.01, ***p < 0.001.* **Figure 6:** *The correlation between SMI and bone mineral content. SMI, skeletal muscle index.*
## Discussion
In this study, we first presented the composition of gut microbiota, SCFA levels, and their relationship in HF patients with and without sarcopenia and controls. We found that overall microbial diversity (alpha‐diversity) and community structure (beta‐diversity) were significantly different between the control individuals and HF patients with or without sarcopenia, while no difference between the HF patients with sarcopenia and HF patients without sarcopenia. The Nocardiaceae, Pseudonocardiaceae, Leuconostocaceae, Alphaproteobacteria, and Slackia were found to be more abundant in HF individuals without sarcopenia. Conversely, the Synergistetes showed a higher abundance in HF patients with sarcopenia. In addition, the levels of isobutyric acid, isovaleric acid, and valeric acid were lower in HF patients with sarcopenia compared with that in the HF patients without sarcopenia and control group but has no significant differences.
Microbiota diversity has been proposed as a health biomarker. Diversity analyses indicated a difference in gut microbiota composition between the control group and HF patients with or without sarcopenia in the current study. Loss of gut flora biodiversity is associated with HF patients with and without sarcopenia. Based on a metagenomics study of patients with sarcopenia (Ticinesi et al., 2020), we expected that alpha-diversity and beta-diversity would be reduced in HF patients with sarcopenia compared to HF patients without sarcopenia, but the results exhibited that there were no significant differences between these two groups in our study population.
The main bacterial phyla in the gut are Firmicutes, Bacteroidetes, Actinobacteria, and Proteobacteria, which constitute the vast majority of the dominant human intestinal flora (Arumugam et al., 2011). The results were the same in our study. Some studies have examined gut microbiome differences in HF patients compared to control individuals (Sun et al., 2021). Likewise, we found significant differences in the composition of gut microbiome between control and HF patients with or without sarcopenia. In line with the study by Elena et al (Gutierrez-Calabres et al., 2020), Firmicutes and Bacteroidetes were the two most abundant phyla in control individuals. Proteobacteria are gram-negative bacteria, with the outer membrane mainly composed of lipopolysaccharide (LPS). The majority of the *Proteobacteria phylum* are pathogenic bacteria, it is thought that they serve as a microbial signature of dysbiosis in gut microbiota (Shin et al., 2015). An elevated level of Actinobacteria in cardiovascular diseases has been reported (Khan et al., 2022). Consistent with these findings, the relative abundance of Proteobacteria and Actinobacteria were higher in HF patients with and without sarcopenia than that in control individuals.
Differential abundances of certain bacterial taxa were observed in HF patients without sarcopenia compared with the control. Specifically, in line with previous studies, the abundance of Nocardiaceae which is associated with steroid hormone synthesis (Sherman et al., 2018), was found to be increased in HF patients without sarcopenia in our study. Pseudonocardiaceae, Slackia, and Alphaproteobacteria belong to the phylum Actinobacteria and Proteobacteria, respectively. Proteobacteria and Actinobacteria have an impact on the prognosis of patients with HF by influencing the production of trimethylamine-N-oxide (Bin-Jumah et al., 2021). The increased abundance of Proteobacteria and Actinobacteria was also discovered in HF patients without sarcopenia in our study. Based on LEfSe analyses, Synergistetes were found abundant in the HF patients with sarcopenia. Synergistetes was correlated with infection and enriched in HF patients with preserved ejection fraction (Huang et al., 2021; Zhu et al., 2022). Increasing evidence demonstrated that *Synergistetes is* associated with periodontal diseases, which is prevalent in sarcopenia patients (Belibasakis et al., 2013; Hatta and Ikebe, 2021). It may be a biomarker of HF patients with sarcopenia that influences the progression of sarcopenia in HF patients. Clostridia, Faecalibacterium, Peptostreptococcaceae, and Prevotellaceae were believed to be probiotic species that enriched the fecal microbiomes of control.
Changes in the microbiota of those suffering from various diseases have been associated with reduced diversity of bacteria and the amount of SCFAs in the feces (Machiels et al., 2014; Wang et al., 2014). SCFAs are produced by the gut microbiota metabolizing dietary fiber and undigested carbohydrates. Alterations in fecal SCFAs may lead to dysbiosis of the intestinal microbiota and inflammatory changes. Muscle mass and strength were increased in germ-free mice fed acetate, propionate, and butyrate (Lahiri et al., 2019). SCFAs, particularly butyric acid, can improve muscle atrophy during the aging process (E Walsh et al. [ 2015]). Our data suggested that there were no significant differences in SCFAs in fecal samples among the three groups. This was in contrast with our hypothesis. Treatment-related confounders in the HF patients with and without sarcopenia cannot be entirely ruled out.
Correlation analysis showed a positive correlation between SCFAs and SCFA-producing bacteria. Different SCFAs are known to be produced by various bacteria. For example, Bacteroides spp. produced acetate and propionate (Koh et al., 2016). The abundances of the Verrucomicrobia were increased in mice feeding with the butyrate and SCFA mix water (Lee et al., 2022). Metabolism and function of skeletal muscle are susceptible to changes in the gut microbiota, this association appears to be partially mediated by the generation of SCFAs. SCFAs have been proven to influence carbohydrate, lipid, and protein metabolism in skeletal muscle tissues both in vitro and in vivo (Frampton et al., 2020). Meanwhile, SCFAs inhibit systemic inflammation by binding to G-protein receptors on the surface of the cells and enhance myocardial energy metabolism, which plays a cardioprotective role (Borchers and Pieler, 2010). Moreover, a study on older Koreans found a positive correlation between sarcopenia and bone mineral content (Kim et al., 2014). Lower bone mass may have effects on falls and fracture risk in sarcopenia patients (Gandham et al., 2021). In our study, bone mineral content was positively correlated with SMI. HF patients with sarcopenia had also significantly lower bone mineral content compared with HF patients without sarcopenia.
There are some limitations in our study, including the small sample size, and the lack of detailed information about dietary intake and physical activity. Although all participants were from the same region and had similar dietary habits, we were still unable to fully rule out the influence of diet on gut microbiota and SCFAs. Furthermore, this is a single-center clinical study and may have a selection bias. In addition, a cross-sectional research is difficult to determine the causal relationships. Larger cohort studies are needed to explore the effect of gut flora on sarcopenia in HF patients and whether modulating the gut microbiota and metabolites can alter the natural course of sarcopenia in HF patients.
## Conclusion
In summary, the preliminary study suggests that HF patients with and without sarcopenia differ from control subjects in intestinal microbial composition. Nocardiaceae, Pseudonocardiaceae, and the Alphaproteobacteria, Slackia were significantly enriched in HF patients without sarcopenia. Synergistetes may be biomarkers of HF patients with sarcopenia. Modulating the gut microbiota may be a new target for the prevention and treatment of sarcopenia in heart failure patients.
## Data availability statement
The original contributions presented in the study are included in the article/supplementary material. Further inquiries can contact the corresponding author. The data presented in the study are deposited in the NCBI repository, accession number PRJNA930609.
## Ethics statement
The study involving human participants was reviewed and approved by Ethics Committee of the Second Xiangya Hospital of Central South University. The patients/participants provided their written informed consent to participate in this study.
## Author contributions
JP: Writing-Original Draft, Investigation, Methodology. HG: Visualization, Methodology. XL: Project administration, Funding acquisition. YL: Software, Date Curation. SL: Conceptualization, Validation. ST: Funding acquisition, Software. LD: Resources, Supervision. XZ: Conceptualization, Funding acquisition, Writing-Reviewing, and Editing. All authors contributed to the article and approved the submitted version.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## References
1. Arumugam M., Raes J., Pelletier E., Le Paslier D., Yamada T., Mende D. R.. **Enterotypes of the human gut microbiome**. *Nature* (2011) **473** 174-180. DOI: 10.1038/nature09944
2. Barton W., C Penney N., Cronin O., Garcia-Perez I., G Molloy M., Holmes E.. **The microbiome of professional athletes differs from that of more sedentary subjects in composition and particularly at the functional metabolic level**. *Gut* (2018) **67** 625-633. DOI: 10.1136/gutjnl-2016-313627
3. Belibasakis G. N., Ozturk V. O., Emingil G., Bostanci N.. **Synergistetes cluster a in saliva is associated with periodontitis**. *J. Periodontal Res.* (2013) **48** 727-732. DOI: 10.1111/jre.12061
4. Bin-Jumah M. N., Gilani S. J., Hosawi S., Al-Abbasi F. A., Zeyadi M., Imam S. S.. **Pathobiological relationship of excessive dietary intake of Choline/L-carnitine: A TMAO precursor-associated aggravation in heart failure in sarcopenic patients**. *Nutrients* (2021) **13** 3453. DOI: 10.3390/nu13103453
5. Borchers A., Pieler T.. **Programming pluripotent precursor cells derived from xenopus embryos to generate specific tissues and organs**. *Genes (Basel)* (2010) **1** 413-426. DOI: 10.3390/genes1030413
6. Branchereau M., Burcelin R., Heymes C.. **The gut microbiome and heart failure: A better gut for a better heart**. *Rev. Endocr. Metab. Disord.* (2019) **20** 407-414. DOI: 10.1007/s11154-019-09519-7
7. Carley A. N., Maurya S. K., Fasano M., Wang Y., Selzman C. H., Drakos S. G.. **Short-chain fatty acids outpace ketone oxidation in the failing heart**. *Circulation* (2021) **143** 1797-1808. DOI: 10.1161/CIRCULATIONAHA.120.052671
8. Chen L. K., Liu L. K., Woo J., Assantachai P., Auyeung T. W., Bahyah K. S.. **Sarcopenia in Asia: consensus report of the Asian working group for sarcopenia**. *J. Am. Med. Dir Assoc.* (2014) **15** 95-101. DOI: 10.1016/j.jamda.2013.11.025
9. Chen Z., Radjabzadeh D., Chen L., Kurilshikov A., Kavousi M., Ahmadizar F.. **Association of insulin resistance and type 2 diabetes with gut microbial diversity: A microbiome-wide analysis from population studies**. *JAMA Netw. Open* (2021) **4**. DOI: 10.1001/jamanetworkopen.2021.18811
10. Chen L. K., Woo J., Assantachai P., Auyeung T. W., Chou M. Y., Iijima K.. **Asian Working group for sarcopenia: 2019 consensus update on sarcopenia diagnosis and treatment**. *J. Am. Med. Dir Assoc.* (2020) **21** 300-307 e302. DOI: 10.1016/j.jamda.2019.12.012
11. Cruz-Jentoft A. J., Landi F., Topinková E., Michel J.-P.. **Understanding sarcopenia as a geriatric syndrome**. *Curr. Opin. Clin. Nutr. Metab. Care* (2010) **13** 1-7. DOI: 10.1097/MCO.0b013e328333c1c1
12. E Walsh M., Bhattacharya A., Sataranatarajan K., Qaisar R., Sloane L., M Rahman Md. **The histone deacetylase inhibitor butyrate improves metabolism and reduces muscle atrophy during aging**. *Aging Cell* (2015) **14** 957-970. DOI: 10.1111/acel.12387
13. Frampton J., Murphy K. G., Frost G., Chambers E. S.. **Short-chain fatty acids as potential regulators of skeletal muscle metabolism and function**. *Nat. Metab.* (2020) **2** 840-848. DOI: 10.1038/s42255-020-0188-7
14. Fulster S., Tacke M., Sandek A., Ebner N., Tschope C., Doehner W.. **Muscle wasting in patients with chronic heart failure: results from the studies investigating co-morbidities aggravating heart failure (SICA-HF)**. *Eur. Heart J.* (2013) **34** 512-519. DOI: 10.1093/eurheartj/ehs381
15. Gandham A., Mesinovic J., Jansons P., Zengin A., Bonham M. P., Ebeling P. R.. **Falls, fractures, and areal bone mineral density in older adults with sarcopenic obesity: A systematic review and meta-analysis**. *Obes. Rev.* (2021) **22**. DOI: 10.1111/obr.13187
16. Gutierrez-Calabres E., Ortega-Hernandez A., Modrego J., Gomez-Gordo R., Caro-Vadillo A., Rodriguez-Bobada C.. **Gut microbiota profile identifies transition from compensated cardiac hypertrophy to heart failure in hypertensive rats**. *Hypertension* (2020) **76** 1545-1554. DOI: 10.1161/HYPERTENSIONAHA.120.15123
17. Hatta K., Ikebe K.. **Association between oral health and sarcopenia: A literature review**. *J. Prosthodont Res.* (2021) **65** 131-136. DOI: 10.2186/jpr.JPOR_2019_567
18. Huang Z., Mei X., Jiang Y., Chen T., Zhou Y.. **Gut microbiota in heart failure patients with preserved ejection fraction (GUMPTION study)**. *Front. Cardiovasc. Med.* (2021) **8**. DOI: 10.3389/fcvm.2021.803744
19. Khan I., Khan I., Kakakhel M. A., Xiaowei Z., Ting M., Ali I.. **Comparison of microbial populations in the blood of patients with myocardial infarction and healthy individuals**. *Front. Microbiol.* (2022) **13**. DOI: 10.3389/fmicb.2022.845038
20. Kimura I., Inoue D., Hirano K., Tsujimoto G.. **The SCFA receptor GPR43 and energy metabolism**. *Front. Endocrinol. (Lausanne)* (2014) **5**. DOI: 10.3389/fendo.2014.00085
21. Kim S., Won C. W., Kim B. S., Choi H. R., Moon M. Y.. **The association between the low muscle mass and osteoporosis in elderly Korean people**. *J. Korean Med. Sci.* (2014) **29** 995-1000. DOI: 10.3346/jkms.2014.29.7.995
22. Koh A., De Vadder F., Kovatcheva-Datchary P., Backhed F.. **From dietary fiber to host physiology: Short-chain fatty acids as key bacterial metabolites**. *Cell* (2016) **165** 1332-1345. DOI: 10.1016/j.cell.2016.05.041
23. Konishi M., Kagiyama N., Kamiya K., Saito H., Saito K., Ogasahara Y.. **Impact of sarcopenia on prognosis in patients with heart failure with reduced and preserved ejection fraction**. *Eur. J. Prev. Cardiol.* (2021) **28** 1022-1029. DOI: 10.1093/eurjpc/zwaa117
24. Lahiri S., Kim H., Garcia-Perez I., Reza M. M., Martin K. A., Kundu P.. **The gut microbiota influences skeletal muscle mass and function in mice**. *Sci. Transl. Med.* (2019) **11**. DOI: 10.1126/scitranslmed.aan5662
25. Lee J. G., Lee J., Lee A. R., Jo S. V., Park C. H., Han D. S.. **Impact of short-chain fatty acid supplementation on gut inflammation and microbiota composition in a murine colitis model**. *J. Nutr. Biochem.* (2022) **101**. DOI: 10.1016/j.jnutbio.2021.108926
26. Liao X., Wu M., Hao Y., Deng H.. **Exploring the preventive effect and mechanism of senile sarcopenia based on "Gut-muscle axis"**. *Front. Bioeng Biotechnol.* (2020) **8**. DOI: 10.3389/fbioe.2020.590869
27. Liu C., Cheung W. H., Li J., Chow S. K., Yu J., Wong S. H.. **Understanding the gut microbiota and sarcopenia: a systematic review**. *J. Cachexia Sarcopenia Muscle* (2021) **12** 1393-1407. DOI: 10.1002/jcsm.12784
28. Machiels K., Joossens M., Sabino J., De Preter V., Arijs I., Eeckhaut V.. **A decrease of the butyrate-producing species roseburia hominis and faecalibacterium prausnitzii defines dysbiosis in patients with ulcerative colitis**. *Gut* (2014) **63** 1275-1283. DOI: 10.1136/gutjnl-2013-304833
29. Madan S., Mehra M. R.. **The heart-gut microbiome axis in advanced heart failure**. *J. Heart Lung Transplant* (2020) **39** 891-893. DOI: 10.1016/j.healun.2020.04.003
30. Mamic P., Chaikijurajai T., Tang W. H. W.. **Gut microbiome - a potential mediator of pathogenesis in heart failure and its comorbidities: State-of-the-art review**. *J. Mol. Cell Cardiol.* (2021) **152** 105-117. DOI: 10.1016/j.yjmcc.2020.12.001
31. Marti J. M., Martinez-Martinez D., Rubio T., Gracia C., Pena M., Latorre A.. **Health and disease imprinted in the time variability of the human microbiome**. *mSystems* (2017) **2**. DOI: 10.1128/mSystems.00144-16
32. Narumi T., Watanabe T., Kadowaki S., Takahashi T., Yokoyama M., Kinoshita D.. **Sarcopenia evaluated by fat-free mass index is an important prognostic factor in patients with chronic heart failure**. *Eur. J. Intern. Med.* (2015) **26** 118-122. DOI: 10.1016/j.ejim.2015.01.008
33. Pakhomov N., A. Baugh J.. **The role of diet-derived short chain fatty acids in regulating cardiac pressure overload**. *Am. J. Physiol. Heart Circ. Physiol.* (2021) **320** H475-H486. DOI: 10.1152/ajpheart.00573.2020
34. Peng J., Xiao X., Hu M., Zhang X.. **Interaction between gut microbiome and cardiovascular disease**. *Life Sci.* (2018) **214** 153-157. DOI: 10.1016/j.lfs.2018.10.063
35. Przewlocka K., Folwarski M., Kazmierczak-Siedlecka K., Skonieczna-Zydecka K., Kaczor J. J.. **Gut-muscle AxisExists and may affect skeletal muscle adaptation to training**. *Nutrients* (2020) **12** 1451. DOI: 10.3390/nu12051451
36. Sherman S. B., Sarsour N., Salehi M., Schroering A., Mell B., Joe B.. **Prenatal androgen exposure causes hypertension and gut microbiota dysbiosis**. *Gut Microbes* (2018) **9** 400-421. DOI: 10.1080/19490976.2018.1441664
37. Shin N. R., Whon T. W., Bae J. W.. **Proteobacteria: microbial signature of dysbiosis in gut microbiota**. *Trends Biotechnol.* (2015) **33** 496-503. DOI: 10.1016/j.tibtech.2015.06.011
38. Sun W., Du D., Fu T., Han Y., Li P., Ju H.. **Alterations of the gut microbiota in patients with severe chronic heart failure**. *Front. Microbiol.* (2021) **12**. DOI: 10.3389/fmicb.2021.813289
39. Tang W. H. W., Li D. Y., Hazen S. L.. **Dietary metabolism, the gut microbiome, and heart failure**. *Nat. Rev. Cardiol.* (2019) **16** 137-154. DOI: 10.1038/s41569-018-0108-7
40. Ticinesi A., Mancabelli L., Tagliaferri S., Nouvenne A., Milani C., Del Rio D.. **The gut-muscle axis in older subjects with low muscle mass and performance: A proof of concept study exploring fecal microbiota composition and function with shotgun metagenomics sequencing**. *Int. J. Mol. Sci.* (2020) **21** 8946. DOI: 10.3390/ijms21238946
41. Wang W., Chen L. P., Zhou R., Wang X. B., Song L., Huang S.. **Increased proportions of bifidobacterium and the lactobacillus group and loss of butyrate-producing bacteria in inflammatory bowel disease**. *J. Clin. Microbiol.* (2014) **52** 398-406. DOI: 10.1128/JCM.01500-13
42. Yuzefpolskaya M., Bohn B., Nasiri M., Zuver A. M., Onat D. D., Royzman E. A.. **Gut microbiota, endotoxemia, inflammation, and oxidative stress in patients with heart failure, left ventricular assist device, and transplant**. *J. Heart Lung Transplant* (2020) **39** 880-890. DOI: 10.1016/j.healun.2020.02.004
43. Zhang Y., Zhang J., Ni W., Yuan X., Zhang H., Li P.. **Sarcopenia in heart failure: a systematic review and meta-analysis**. *ESC Heart Fail.* (2021) **8** 1007-1017. DOI: 10.1002/ehf2.13255
44. Zhu Y. Y., Thakur K., Feng J. Y., Zhang J. G., Hu F., Cespedes-Acuna C. L.. **Riboflavin bioenriched soymilk alleviates oxidative stress mediated liver injury, intestinal inflammation, and gut microbiota modification in B2 depletion-repletion mice**. *J. Agric. Food Chem.* (2022) **70** 3818-3831. DOI: 10.1021/acs.jafc.2c00117
|
---
title: Produce D-allulose from non-food biomass by integrating corn stalk hydrolysis
with whole-cell catalysis
authors:
- Qing Jia
- Hui Zhang
- Anqi Zhao
- Lingbo Qu
- Wenlong Xiong
- Md. Asraful Alam
- Jixing Miao
- Weigao Wang
- Feihu Li
- Jingliang Xu
- Yongkun Lv
journal: Frontiers in Bioengineering and Biotechnology
year: 2023
pmcid: PMC9998921
doi: 10.3389/fbioe.2023.1156953
license: CC BY 4.0
---
# Produce D-allulose from non-food biomass by integrating corn stalk hydrolysis with whole-cell catalysis
## Abstract
D-allulose is a high-value rare sugar with many health benefits. D-allulose market demand increased dramatically after approved as generally recognized as safe (GRAS). The current studies are predominantly focusing on producing D-allulose from either D-glucose or D-fructose, which may compete foods against human. The corn stalk (CS) is one of the main agricultural waste biomass in the worldwide. Bioconversion is one of the promising approach to CS valorization, which is of significance for both food safety and reducing carbon emission. In this study, we tried to explore a non-food based route by integrating CS hydrolysis with D-allulose production. Firstly we developed an efficient *Escherichia coli* whole-cell catalyst to produce D-allulose from D-glucose. Next we hydrolyzed CS and achieved D-allulose production from the CS hydrolysate. Finally we immobilized the whole-cell catalyst by designing a microfluidic device. Process optimization improved D-allulose titer by 8.61 times, reaching 8.78 g/L from CS hydrolysate. With this method, 1 kg CS was finally converted to 48.87 g D-allulose. This study validated the feasibility of valorizing corn stalk by converting it to D-allulose.
## 1 Introduction
The excessive intake of high energy sugars has caused many health problems, such as diabetes, hypertension, hyperlipidemia, and other diseases (Xia et al., 2021). D-allulose (D-psicose or D-ribo-2-hexulose) is a good sugar substitute, because it is an low calorie sweeter ($70\%$ sweetness of sucrose), and inert in energy metabolism (Iida et al., 2010). Besides, D-allulose also has many other health benefits, such as anti-oxidative, anti-diabetic (Hossain et al., 2015), anti-obesity (Nagata et al., 2015), and neuroprotective effects, and others (Juneja et al., 2019). The approval of D-allulose as GRAS by FDA has stimulated its market demand as food ingredient and dietary supplement (Zhang W. L. et al., 2017). However, as a rare sugar, D-allulose is hardly found in nature (Mu et al., 2011). Very small quantities of D-allulose can be found in the stems or leaves of Itea and wheat and some bacteria (Zhang et al., 2016b).
In the past decades, both chemical (Doner, 1979) and biological (Mu et al., 2011; Zhu et al., 2012) methods have been developed to synthesize D-allulose. Among them, biological methods are more attractive, because it is greener, lower cost, and its product is easier to purify (Izumori, 2002). The currently two most studied bioproduction approaches are the aldol condensation pathway and the Izumoring strategy (Zhang W. L. et al., 2017). In the aldol condensation pathway, D-allulose is produced by L-rhamnulose-1-phosphate aldolase-catalyzed condensation of dihydroxyacetone phosphate and D-glyceraldehyde and subsequent dephosphorylation (Li et al., 2017). This pathway involves the cofactor (ATP and NAD(P)+) regeneration and at least 5 functional genes (from glycerol to D-allulose) (Brovetto et al., 2011; Li et al., 2011; Wei et al., 2015). In comparison, the Izumoring strategy is simpler and straightforward. D-allulose production from D-glucose utilizes only 2 enzymes: D-glucose isomerase (interconversion of D-glucose and D-fructose) and D-psicose 3-epimerase (interconversion of D-fructose and D-allulose), and involves no cofactor regeneration (Izumori, 2006; Men et al., 2014). Consequently, the Izumoring strategy is the predominantly explored approach (Li et al., 2015; Zhang W. et al., 2017; Chen et al., 2017; Zhang et al., 2020). Based on this strategy, either D-fructose (Zhang et al., 2013; Zhang et al., 2015; Zhang et al., 2016a; He et al., 2016) or D-glucose (Men et al., 2014) was used as the substrate for D-allulose production.
With the continues increase of world population, food and resource insufficiency is becoming a great challenge to human society. The large scale commercial production of D-allulose from D-glucose or D-fructose (D-fructose is also produced from D-glucose) may “struggle” for food against human (Juneja et al., 2019). Consequently, developing non-food based and sustainable D-allulose producing process is necessary (Song et al., 2017). Corn stalk (CS) is a very promising alternative to the current D-glucose feedstock, because it is continuously produced as one of the major agricultural wastes and composed of $30\%$–$40\%$ cellulose, $20\%$–$30\%$ hemicellulose, and $10\%$–$20\%$ lignin (Yang et al., 2021). Cellulose is a polymer of glucose, and can be hydrolyzed into monomer (D-glucose). Among the annual lignocellulose output (about 170 billion tons), only $3\%$ has been efficiently utilized (Liu et al., 2021; Shen and Sun, 2021). Compared with the free-enzyme reactions, whole-cell catalysis has the advantages of without tedious and costly enzyme purification process, protecting enzymes from harsh reaction conditions, enhancing reactions by colocalizing multiple enzymes within the cell, and preventing intermediates from diffusion (Chen et al., 2022). In this study, we will try to develop an efficient whole-cell catalyst to produce D-allulose from CS hydrolysate. By integration the whole-cell catalysis with CS hydrolysis, this study will provide a sustainable process for D-allulose production from the non-food biomass, as well as the valorization of agricultural waste CS.
## 2.1 Genes, plasmids, and strains
The encoding genes of glucose isomerase from *Acidothermus cellulolyticus* 11B (AcceGI, NCBI access number: WP_011720899) (Mu et al., 2012) and D-psicose 3-epimerase (CcDPEase, NCBI access number: 3VNI_A) (Mu et al., 2011) were codon optimized and synthesized by Sangon Biotech (Shanghai, China). Other putative glucose isomerase, xylose isomerase, and D-psicose 3-epimerase genes were obtained by bioinformatic analysis and amplified from corresponding genome DNA (Table 1). The ePathBrick plasmid pET-28a (PB) was used for gene expression, fusion gene construction, and gene copy number optimization (Xu et al., 2012; Lv et al., 2017). E. coli JM109 was used for plasmid construction, maintenance, and propagation. E. coli BL21 (DE3) was used for protein expression, whole-cell catalyst development, and cell immobilization.
**TABLE 1**
| Gene | Enzyme | Original strain | Length (bp) |
| --- | --- | --- | --- |
| YlXI | Xylose isomerase | Yarrowia lipolytica Po1f | 1179 |
| YlGPI | Glucose-6-phosphate isomerase | Yarrowia lipolytica Po1f | 1668 |
| GoDPEase | D-psicose 3-epimerase or xylose isomerase( a ) | Gluconobacter oxydans 621H | 852 |
| GoXI_02 | Xylose isomerase | Gluconobacter oxydans 621H | 744 |
| GoGPI | Glucose-6-phosphate isomerase | Gluconobacter oxydans 621H | 1068 |
| PpDPEase_01 | D-psicose 3-epimerase | Pseudomonas putida KT2440 | 1248 |
| PpDPEase_02 | D-psicose 3-epimerase | Pseudomonas putida KT2440 | 783 |
| PpXI | Xylose isomerase | Pseudomonas putida KT2440 | 816 |
| BsDPEase_01 | D-psicose 3-epimerase | Bacillus subtilis subsp. subtilis str. 168 | 915 |
| BsDPEase_02 | D-psicose 3-epimerase | Bacillus subtilis subsp. subtilis str. 168 | 894 |
| BsXI | Xylose isomerase | Bacillus subtilis subsp. subtilis str. 168 | 1338 |
| BsGPI | Glucose-6-phosphate isomerase | Bacillus subtilis subsp. subtilis str. 168 | 1353 |
| PaXI | Xylose isomerase | Pseudomonas aeruginosa PAO1 | 816 |
| PaDPEase_01 | D-psicose 3-epimerase | Pseudomonas aeruginosa PAO1 | 783 |
| PaDPEase_02 | D-psicose 3-epimerase | Pseudomonas aeruginosa PAO1 | 798 |
| BtDPEase | D-psicose 3-epimerase or xylose isomerase( a ) | Bacillus thuringiensis ATCC 10792 | 843 |
## 2.2 Bioinformatic analysis
The genome sequences of *Yarrowia lipolytica* strain CLIB89(W29), *Bacillus subtilis* subsp. subtilis str. 168 (NC_000964.3), *Bacillus thuringiensis* strain ATCC 10792 (NZ_CP021061.1), Gluconobacter oxydans 621H (NC_006677.1), *Pseudomonas aeruginosa* PAO1 (NC_002516.2), and *Pseudomonas putida* KT2440 (NC_002947.4) were downloaded from NCBI. The amino acid sequences of all glucose isomerase, xylose isomerase, and D-psicose 3-epimerase were downloaded from NCBI (update 5 July 2020) and used to query the genome sequences using TBLASTN. The predicted genes were double checked by querying the amino acid sequences of the corresponding enzymes.
## 2.3 Molecular biology
AcceGI and CcDPEase were subcloned into pET-28a (PB) between BamHI and HindIII sites to yield pET28a (PB)-AcceGI and pET28a (PB)-CcDPEase, respectively. The putative genes (Table 1) were subcloned into pET-28a (PB) by homologous one-step cloning, resulting in corresponding recombinant plasmids (Table 2) (Xu et al., 2012; Lv et al., 2017). The primers were flanked with homologous sequence of pET-28a (PB) at 5′-terminal (Supplementary Table S1). The homologous one-step cloning was carried out using ClonExpress II One Step Cloning Kit (Vazyme, Nanjing, China) (Zhang et al., 2014).
**TABLE 2**
| Plasmid | Genetic characteristics | Reference or source |
| --- | --- | --- |
| pET-28a (PB) | An ePathBrick vector | Xu et al. (2012), Lv et al. (2017) |
| pET28a (PB)-AcceGI | pET-28a (PB) carrying a glucose isomerase gene AcceGI from Acidothermus cellulolyticus 11B | This study |
| pET28a (PB)-CcDPEase | pET-28a (PB) carrying a D-psicose 3-epimerase gene CcDPEase from Clostridium cellulolyticum H10 | This study |
| pET28a (PB)-YlXI | pET-28a (PB) carrying a putative xylose isomerase gene from Yarrowia lipolytica Po1f | This study |
| pET28a (PB)-YlGPI | pET-28a (PB) carrying a putative glucose-6-phosphate isomerase gene from Yarrowia lipolytica Po1f | This study |
| pET28a (PB)-GoDPEase | pET-28a (PB) carrying a putative D-psicose 3-epimerase or xylose isomerase( a ) gene from Gluconobacter oxydans 621H | This study |
| pET28a (PB)-GoXI_02 | pET-28a (PB) carrying a putative xylose isomerase gene from Gluconobacter oxydans 621H | This study |
| pET28a (PB)-GoGPI | pET-28a (PB) carrying a putative glucose-6-phosphate isomerase gene from Gluconobacter oxydans 621H | This study |
| pET28a (PB)-PpDPEase_01 | pET-28a (PB) carrying a putative D-psicose 3-epimerase gene from Pseudomonas putida KT2440 | This study |
| pET28a (PB)-PpDPEase_02 | pET-28a (PB) carrying a putative D-psicose 3-epimerase gene from Pseudomonas putida KT2440 | This study |
| pET28a (PB)-PpXI | pET-28a (PB) carrying a putative xylose isomerase gene from Pseudomonas putida KT2440 | This study |
| pET28a (PB)-BsDPEase_01 | pET-28a (PB) carrying a putative D-psicose 3-epimerase gene from Bacillus subtilis subsp. subtilis str. 168 | This study |
| pET28a (PB)-BsDPEase_02 | pET-28a (PB) carrying a putative D-psicose 3-epimerase gene from Bacillus subtilis subsp. subtilis str. 168 | This study |
| pET28a (PB)-BsXI | pET-28a (PB) carrying a putative xylose isomerase gene from Bacillus subtilis subsp. subtilis str. 168 | This study |
| pET28a (PB)-BsGPI | pET-28a (PB) carrying a putative glucose-6-phosphate isomerase gene from Bacillus subtilis subsp. subtilis str. 168 | This study |
| pET28a (PB)-PaXI | pET-28a (PB) carrying a putative xylose isomerase gene from Pseudomonas aeruginosa PAO1 | This study |
| pET28a (PB)-PaDPEase_01 | pET-28a (PB) carrying a putative D-psicose 3-epimerase gene from Pseudomonas aeruginosa PAO1 | This study |
| pET28a (PB)-PaDPEase_02 | pET-28a (PB) carrying a putative D-psicose 3-epimerase gene from Pseudomonas aeruginosa PAO1 | This study |
| pET28a (PB)-BtDPEase | pET-28a (PB) carrying a putative D-psicose 3-epimerase or xylose isomerase( a ) gene from Bacillus thuringiensis ATCC 10792 | This study |
| pET28a (PB)-GP | pET-28a (PB) carrying AcceGI and CcDPEase in fusion form | This study |
| pET28a (PB)-GSP | pET-28a (PB) carrying AcceGI and CcDPEase linked with “GGGGS” encoding sequence | This study |
| pET28a (PB)-GS2P | pET-28a (PB) carrying AcceGI and CcDPEase linked with “GGGGSGGGGS” encoding sequence | This study |
| pET28a (PB)-GS3P | pET-28a (PB) carrying AcceGI and CcDPEase linked with “GGGGSGGGGSGGGGS” encoding sequence | This study |
| pET28a (PB)-GEP | pET-28a (PB) carrying AcceGI and CcDPEase linked with “EAAAK” encoding sequence | This study |
| pET28a (PB)-GE2P | pET-28a (PB) carrying AcceGI and CcDPEase linked with “EAAAKEAAAK” encoding sequence | This study |
| pET28a (PB)-GE3P | pET-28a (PB) carrying AcceGI and CcDPEase linked with “EAAAKEAAAKEAAAK” encoding sequence | This study |
| pET28a (PB)-AcceGI-CcDPEase | pET-28a (PB) carrying AcceGI and CcDPEase in monocistronic form | This study |
| pET28a (PB)-AcceGI×2-CcDPEase | pET-28a (PB) carrying 2-copy AcceGI and CcDPEase in monocistronic form | This study |
| pET28a (PB)-AcceGI-CcDPEase×2 | pET-28a (PB) carrying AcceGI and 2-copy CcDPEase in monocistronic form | This study |
| pET28a (PB)-AcceGI×3-CcDPEase | pET-28a (PB) carrying 3-copy AcceGI and CcDPEase in monocistronic form | This study |
| pET28a (PB)-AcceGI×4-CcDPEase | pET-28a (PB) carrying 4-copy AcceGI and CcDPEase in monocistronic form | This study |
| pET28a (PB)-AcceGI×5-CcDPEase | pET-28a (PB) carrying 5-copy AcceGI and CcDPEase in monocistronic form | This study |
| pET28a (PB)-AcceGI×6-CcDPEase | pET-28a (PB) carrying 6-copy AcceGI and CcDPEase in monocistronic form | This study |
| pET28a (PB)-AcceGI×7-CcDPEase | pET-28a (PB) carrying 7-copy AcceGI and CcDPEase in monocistronic form | This study |
AcceGI and CcDPEase were fused together directedly or linked with flexible or rigid linkers. The detailed processes were as follows. AcceGI_GP was amplified with primer pair GI_GP_Fusion F/GI_GP_Fusion R, and subcloned into pET28a (PB)-CcDPEase at BamHI site. The resulting plasmid pET28a (PB)-GP will produce a fusion protein GP, which links the C-terminal of AcceGI to the N-terminal of CcDPEase directly. AcceGI_GS1P was amplified with primer pairs GI_GP_Fusion F/GI_GS1P_Fusion R, and subcloned into pET28a (PB)-CcDPEase at BamHI site. The resulting plasmid pET28a (PB)-GS1P will produce a fusion protein GS1P, which links AcceGI and CcDPEase with linker GGGGS. AcceGI_GS2P and CcDPEase_GS2P were amplified with primer pairs GI_GP_Fusion F/GS2P R1 and GS2P F2/DPEase_GP_Fusion R2, and subcloned into pET-28a (PB) between BamHI and HindIII sites. The resulting plasmid pET28a (PB)-GS2P will produce a fusion proteins GS2P, which links AcceGI and CcDPEase with linkers GGGGSGGGGS. AcceGI_GS3P and CcDPEase_GS3P were amplified with primer pairs GI_GP_Fusion F/GS3P R1 and GS3P F2/DPEase_GP_Fusion R2, and subcloned into pET-28a (PB) between BamHI and HindIII sites. The resulting plasmid pET28a (PB)-GS3P will produce a fusion proteins GS3P, which links AcceGI and CcDPEase with linkers GGGGSGGGGSGGGGS. Plasmids pET28a (PB)-GE1P, pET28a (PB)-GE2P, and pET28a (PB)-GE3P were constructed following the same process, except that AcceGI_GE1P, AcceGI_GE2P, CcDPEase_GE2P, AcceGI_GE3P, and CcDPEase_GE3P, were amplified with primer pairs GI_GP_Fusion F/GI_GE1P_Fusion R, GI_GP_Fusion F/GE2P R1, GE2P F2/DPEase_GP_Fusion R2, GI_GP_Fusion F/GE3P R1, and GE3P F2/DPEase_GP_Fusion R2 (Supplementary Table S1; Table 2). All the above subcloning were carried out using ClonExpress II One Step Cloning Kit (Vazyme, Nanjing, China). All the encoding genes of flexible linker ((GGGGS)n ($$n = 1$$–3)) and rigid linker ((EAAAK)n ($$n = 1$$–3)) were codon optimized with GenSmart™ Codon Optimization online tool (https://www.genscript.com/gensmart-free-gene-codon-optimization.html) (Supplementary Table S2) (Guo et al., 2013).
To co-overexpress glucose isomerase and D-psicose epimerase in single cell, AcceGI and CcDPEase were assembled into monocistronic form by using isocaudomers (AvrII, NheI, and SalI) and subsequent T4 ligation (Xu et al., 2012). *The* gene copy number optimization was also carried out by using this method.
## 2.4 Protein expression and enzymatic activity analysis
The recombinant protein expression was carried out following previous methods with moderate modification (Lv et al., 2017). A E. coli BL21 (DE3) colony containing corresponding recombinant plasmid was inoculated into a sterile incubating tube containing 2 mL LB medium. After overnight incubation at 37°C and 200 rpm, 1 mL culture was inoculated into 250 mL shaking flask containing 25 mL LB medium. The strain was cultured at 37°C and 200 rpm until OD600 reach 0.6–0.8. Cool down the culture to 25°C, and add 0.5 mM isopropyl-β-D-thiogalactopyranoside (IPTG) for induction. After another 8 h incubation at 25°C and 200 rpm, centrifuge at 4°C and 8,000 rpm to collect cell pellet. After wash with pure water, the cells were resuspended in water, and used for SDS-PAGE assay, enzymatic activity analysis, process optimization, or immobilization.
The biomass was measured by recording the optical density at 600 nm (OD600) with a Tecan Infinite® M Plex microplate reader (Tecan, Männedorf, Switzerland). SDS-PAGE (Sangon Biotech, Shanghai, China) was used to validate the recombinant protein expression. The enzymatic activities of AcceGI and CcDPEase were measured following previous methods (Mu et al., 2011; Mu et al., 2012).
## 2.5 Analytical methods
D-glucose, D-fructose, and D-allulose were analyzed with a Thermo Scientific™ Dionex™ ICS-6000 ion chromatography system (HPIC) equipped with a Dionex CarboPac™ PA20 BioLC™ 3 × 150 mm analytical column. The mobile phase A was pure water, and B was water with 200 mM NaOH. The gradient (B%) was as follows: 0–15 min $10\%$, 15.1–25 min $100\%$, 25.1–35 min $10\%$. The flow rate was 0.5 mL/min. The oven and detector temperature was maintained at 30°C. The inject volume was 25 μL.
## 2.6 Escherichia coli cell immobilization
The recombinant cells were cultured and collected as the above description, and mixed with sodium alginate solutions of different concentrations ($0.1\%$, $0.5\%$, $1.0\%$, $1.5\%$, $2.0\%$, w/v). The mixture was then pumped into calcium chloride solutions of different concentrations ($1.0\%$, $1.5\%$, $2.0\%$, $2.5\%$, $3.0\%$, w/v). The flow rate was 60 mL/h. Incubate at room temperature for different period (20 min, 40 min, 60 min, 80 min, 100 min) to obtain calcium alginate fibers containing recombinant cells.
## 2.7 Corn stalk pretreatment
The CS was collected from local farm (Yuanyang county, Henan province, China, 113.9o E, 35.1o N), and then totally dried, ground and screened with a 60 mesh sieve. The pulping was carried out by cooking with $2\%$ (w/v) NaOH solution at 80°C and atmospheric pressure for 2 h. The ratio of CS to NaOH was 1:20 (w/w). Wash the pulp with pure water until pH maintain stable, then totally dry the pulp in oven at 105°C. The hydrolysis was carried out with cellulases (Qingdao Vland Biotech Inc., Qingdao, China) in deionized water (adjust to pH5.0 with acetic acid). It should be noted that the hydrolysis buffer should not contain any sodium ion, which will cause the calcium alginate fiber instable in the next step. A final concentration of $10\%$ (w/v) substrate and (10 FPU cellulase)/(g substrate) was added. Incubate the mixture at 50°C and 200 rpm for 3 days for sufficient hydrolysis.
## 2.8 Process optimization
The process optimization was carried out by single factor optimization. Recombinant protein expression, induction time (4–20 h) and IPTG concentration (0.001–0.5 mM) were optimized stepwise following the previous descriptions (Lv et al., 2017). For cell immobilization, calcium chloride concentration ($1.0\%$–$3.0\%$), sodium alginate concentration ($0.1\%$–$2.0\%$), cell dosage (OD600 = 10–35), and immobilization time (20–100 min) were optimized stepwise. For the D-allulose production, D-glucose concentration (20–60 g/L), reaction temperature (55–75°C), and reaction time (2–12 h) were optimized stepwise. All the bioconversions were carried out in pure water at natural pH (around 7.5). For each bioconversion, triplicated biological repeats were carried out in 250-mL shaking flasks. The statical analysis and graphing were performed using Origin Lab software (OriginLab Corporation, Northampton, MA).
## 3.1 Producing D-allulose from glucose by developing whole-cell catalysts
We firstly explored the potential glucose isomerase and D-psicose 3-epimerase in *Yarrowia lipolytica* strain CLIB89(W29), *Bacillus subtilis* subsp. subtilis str. 168, *Bacillus thuringiensis* strain ATCC 10792, Gluconobacter oxydans 621H, *Pseudomonas aeruginosa* PAO1, and *Pseudomonas putida* KT2440 by using local TBLASTN method (Altschul et al., 1997). We chose these microbes because they are readily available in our laboratory. Previous studies showed that many xylose isomerases also show glucose isomerase activities, consequently we also analyzed the xylose isomerases in these microbes (Mu et al., 2012; Karaoglu et al., 2013). The bioinformatic analysis predicted 3 glucose-6-phosphate isomerases, 7 xylose isomerases, and 8 D-psicose 3-epimerases, among which 2 were annotated as both D-psicose 3-epimerase and xylose isomerase (Table 1). To validate the activities of these putative genes, we overexpressed them in E. coli BL21 (DE3). Besides, the known glucose isomerase from *Acidothermus cellulolyticus* 11B (AcceGI) and D-psicose 3-epimerase from *Clostridium cellulolyticum* H10 (CcDPEase) were also overexpressed (Mu et al., 2011; Mu et al., 2012). SDS-PAGE results clearly showed that almost all these genes were successfully expressed, except GoXI, PaXI, GoGPI, and PpDPEase01 (Figures 1A, B). Unfortunately, in the subsequent enzymatic analysis, only AcceGI and CcDPEase showed obvious glucose isomerase and D-psicose 3-epimerase activities, respectively (Figures 1C, D).
**FIGURE 1:** *Expression and activity analysis of predicted glucose isomerase, xylose isomerase, and D-psicose 3-epimerase. (A) Expression of predicted glucose isomerase and glucose isomerase (B) Expression of predicted D-psicose 3-epimerase. (C) Enzyme activity analysis of predicted glucose isomerase and glucose isomerase. The analysis was carried out by converting D-glucose to D-fructose (D) Enzyme activity analysis of predicted D-psicose 3-epimerase. The analysis was carried out by converting D-fructose to D-allulose. In all panels, control refers to E. coli BL21 (DE3) containing empty pET-28a (PB) plasmid. In SDS-PAGE results, red arrows indicate the recombinant proteins. In HPIC results, standards refer to mixture of D-glucose, D-fructose, and D-allulose standards.*
To develop the one-step D-allulose producing whole-cell catalyst, we co-overexpressed AcceGI and CcDPEase in monocistronic form (G-P) in 1 cell by using the ePathBrick method (Xu et al., 2012). When incubating the whole-cell catalyst (G-P) with 50 g/L D-glucose, 3.94 g/L D-allulose was produced with a yield of $7.88\%$ (Figures 2B, C).
**FIGURE 2:** *Optimization of the whole-cell catalyst. (A) Expression of the fusion enzymes. The calculated molecular weights were as follows: GP 81.49 kDa, GS1P 81.80 kDa, GS2P 82.11 kDa, GS3P 82.42 kDa, GE1P 81.96 kDa, GE2P 82.42 kDa, GE3P 82.89 kDa. M refers to marker. C refers to the control, which contain the empty pET-28a (PB) plasmid. (B) HPIC analysis of the enzymatic cascades (C) Efficiency of the enzymatic cascades in converting D-glucose to D-allulose. GP refers to the fusion protein linked AcceGI and CcDPEase directly. Other fusion protein are those linked AcceGI and CcDPEase through corresponding flexible or rigid linkers. G-P refers to the cascade composed of free AcceGI and CcDPEase. Standards refer to the D-glucose, D-fructose, and D-allulose mixture. (D) Analysis of rate limiting step of the enzymatic cascade by stepwise increasing gene copy number (E) Improving the activity of the whole-cell catalyst by optimizing the enzyme expression levels. (F) Expression level analysis of AcceGI and CcDPEase. M refers to marker. C refers to the control, which contain the empty pET-28a (PB) plasmid. G indicates the bands of AcceGI; P indicates the bands of CcDPEase.*
## 3.2 Improving the activity of whole-cell catalyst by balancing enzyme expression level
Spatially confining enzymes is a commonly used approach to preventing intermediates diffusion and thus improving enzymatic cascade efficiency (Zhong et al., 2022). Consequently, we designed a panel of fusion proteins, which linked AcceGI and CcDPEase directly or through flexible or rigid linkers. The expression of fusion proteins was validated by SDS-PAGE analysis (Figure 2A). The enzymatic activity analysis showed that all these fusion enzymes maintained their native activities. Moreover, longer linkers conferred enzymes with higher activities, which should be a result of less steric hindrance (Supplementary Figure S1). Consequently, we used the two best performing fusion proteins (GS3P and GE3P) to validate the enzymatic cascade. HPIC results showed that the fusion enzymes (GS3P and GE3P) directly converted D-glucose to D-allulose (Figure 2B). However, neither of the fusion enzymes performed better than that of the free enzyme cascade (Figure 2C). We deem this to be resulted from the steric hindrance effect, because the enzymes with longer linkers showed higher activities and the rigid linker performed better than the flexible one (Supplementary Figure S1) (Chen et al., 2013). These results indicated that the intermediate diffusion should not to be the predominant rate limiting factor in this whole-cell catalyst.
We next tried to improve the activity of the whole-cell catalyst by balancing the expression levels of the enzymes. This was achieved by optimizing gene copy numbers by using the ePathBrick method (Xu et al., 2012). The results showed that increasing AcceGI copy number improved the whole-cell activity by $24.38\%$, while increasing CcDPEase copy number had no effect on the whole-cell catalyst (Figure 2D). This result indicated that AcceGI is the rate limiting step of the enzymatic cascade. Subsequently we used this straightforward method to further stepwise increase the gene copy number to 7. The results showed that the activities of the while-cell catalyst improved along with the increasing of the AcceGI copy number until 5, and thereafter activities decreased (Figure 2E). The improvement of AcceGI expression level was validated with SDS-PAGE analysis (Figure 2F). Moreover, improving AcceGI copy number did not have obvious affect on the biomass (Supplementary Figure S2). Consequently, we used the optimal whole-cell catalyst containing 5 AcceGI (AcceGI×5-CcDPEase) in the subsequent research.
## 3.3 Producing D-allulose from corn stalk
The purpose of this study is to produce D-allulose from CS. So we next developed a process to convert CS into monosaccharide by subsequential grind, alkaline pretreatment, and digestion. The resulting hydrolysate was a light yellow solution, containing 70.82 g/L D-glucose, 19.13 g/L D-xylose, 1.69 g/L L-arabinose, and other minor components (D-mannose, D-galactose, lignans, and furfural) (Figures 3A, B). The D-glucose yield from dry CS was $27.83\%$.
**FIGURE 3:** *Producing D-allulose from CS. (A) CS hydrolysate by subsequential grind, alkaline pretreatment, and digestion (B) Major monosaccharide components of the CS hydrolysate. (C) D-allulose titers from pure D-glucose and CS hydrolysate (D) Relative activities of free AcceGI to D-glucose and D-xylose mixture and CS hydrolysate. The activity of free AcceGI to pure D-glucose was defined as 100%. (E) D-xylose can be used as the substrate of AcceGI, and converted to D-xylulose (F) Relative activities of free CcDPEase to mixtures of monosaccharides. The activity of free CcDPEase to pure D-fructose was defined as 100%. D-Glu refers to D-glucose. D-Xyl refers to D-xylose. L-Ara refers to L-arabinose. Hydro refers to CS hydrolysate. When using CS hydrolysate as substrate, the performance conditions were identical to those using D-glucose as substrate. The detailed performance conditions were described in the materials and methods section. Red stars indicate the effects of CS hydrolysate to enzymes.*
When using the best performing whole-cell catalyst (AcceGI×5-CcDPEase) to convert this hydrolysate, 2.82 g/L D-allulose was produced from 50 g/L D-glucose (from CS hydrolysate) with a yield of $5.64\%$ (Figure 3C). This yield equals to $61.82\%$ of that from pure D-glucose (Figure 3C). We deem the loss of catalytic activity to two reasons, the inhibition effect of inhibitory factors (such as furfural) and the competition effect of other potential substrates. For instance, the glucose isomerase has been shown being capable of utilizing both D-glucose and D-xylose (Patra and Bera, 2014). To validate our hypothesis, we mimicked the hydrolysate by adding D-xylose to D-glucose to the same final concentrations. The activities of free AcceGI and CcDPEase to pure D-glucose and D-fructose were defined as $100\%$, respectively. The results showed that adding D-xylose decreased AcceGI activity by $24.14\%$, which is similar to that of the CS hydrolysate (Figure 3D). Moreover, the HPIC results showed that free AcceGI converted D-xylose into D-xylulose, which indicated the competitive effect between D-xylose and D-glucose to AcceGI (Figure 3E). On the other hand, potential substrates did not have effect on CcDPEase activity (Figure 3F). These results indicated that the activity decrease when using CS hydrolysate as substate is predominantly resulted from competition between D-glucose and D-xylose.
## 3.4 Enhancing D-allulose production by developing and optimizing a recyclable catalytic fiber
To improve the whole-cell catalyst reusability and process economical efficiency, we designed a simple microfluidic system to immobilize the whole-cell catalyst in calcium alginate fiber. As shown in Figure 4A, a mixture of whole-cell catalyst and sodium alginate solution was in the syringe, which was driven by a motor. Calcium chloride solution in the beaker reacts with the fiber from the syringe, resulting in stable calcium alginate fibers (Figure 4B). The fiber’s diameter was controlled by the needle, and the flow rate was controlled by the controller. To validate the cell immobilization in the fiber, we substituted the enzymes (AcceGI and CcDPEase) with an enhanced green fluorescent protein (EGFP) to track the recombinant cells. Fluorescent microscope results showed that these cells were immobilized in the fibers, whose diameter was 500 μm (Figure 4C).
**FIGURE 4:** *Whole-cell catalyst immobilization and optimization. (A) Architecture of the microfluidic systems (B) Calcium alginate fiber made by the microfluidic system. (C) Calcium alginate fiber containing fluorescent cells. This photo was taken with a fluorescent microscope. The cells were tracked by overexpressing EGFP. The red bar indicates 100 μm (D) Induction time and (E) IPTG concentration optimization for protein expression. (F) CaCl2 concentration, (G) sodium alginate concentration, (H) cell dosage, and (I) immobilization time optimization for the immobilization process. (J) Substrate concentration, (K) reaction temperature, and (L) reaction time optimization for the D-allulose production process (M) Reusability analysis of the calcium alginate fiber in D-allulose production. The detailed performance conditions were described in the materials and methods section.*
We next tried to improve the fiber’s activity by process optimization. The results showed that for the recombinant protein induction, the optimal induction time was 12 h, and optimal IPTG concentration was 0.05 mM (Figures 4D, E). For the immobilization process, the optimal concentrations of calcium chloride and sodium alginate were $2.5\%$ (w/w) and $0.5\%$ (w/w) respectively (Figures 4F, G), and the optimal immobilization time was 80 min (Figure 4I). Although higher D-allulose production was obtained along with higher cell dosage (OD600 = 10–35), however it increased very slow after OD600 reaching 15 (Figure 4H). As a result, we used OD600 = 15 in the subsequent optimization. For the D-allulose production process, the optimal substrate (D-glucose in CS hydrolysate) concentration and temperature were 50 g/L and 65°C (Figures 4J, K) respectively. Although higher D-allulose production was obtained along with longer reaction time during 2–12 h, however it increased very slow after 8 h (Figure 4L). As a result, we used 8 h as the reaction time in the subsequent reactions. Taking together, the process optimization improved D-allulose production by 8.61 times from 1.02 g/L to 8.78 g/L. To validate the fiber’s reusability, we reused one single fiber in 10 catalytic reactions. The results showed that the fiber maintained $84.53\%$ relative activity after 10-cycle reuse (Figure 4M). It should be noted that calcium alginate fiber is unstable in the presence of some ions like sodium. Consequently, we did not optimize the reaction pH (avoiding the involvement of sodium buffer) and carried out all the bioconversions in pure water at natural pH (around 7.5). The results showed that the calcium alginate fiber was stable in pure water (Figure 4M). This not only stabilized the fiber, but also could be helpful for reducing the cost of future large scale applications (eliminating the involvement of any buffer).
## 4 Discussion
D-allulose is a GRAS sugar substitute, and characterized with many health benefits (Iida et al., 2010). As a rare sugar, D-allulose is hardly found in nature. The currently predominant production method is based on the Izumoring strategy, which use either D-glucose or D-fructose as substrate (Izumori, 2006; Juneja et al., 2019; Zhang et al., 2020). With the increasing concerns about insufficient food supply, exploring non-food based D-allulose producing method is becoming necessary (Hobbs, 2020). CS is one of the main agricultural wastes continuously produced worldwide. Dry CS contains more than $30\%$ cellulose, which is a polymer of D-glucose (Zhang et al., 2021). Consequently, CS biomass could be a promising alternative feedstock to D-glucose or D-fructose.
The Izumoring strategy is limited by thermodynamic equilibrium. When the reaction reach equilibrium in the D-allulose production from D-glucose, the D-allulose yield was $18.18\%$ (the ratio of D-glucose:D-fructose:D-allulose was 6.5:7:3) (Zhang W. et al., 2017). In the present study, 8.78 g/L D-allulose was produced from 50 g/L D-glucose (obtained by CS hydrolysis). The D-glucose yield from dry CS was $27.83\%$. Taking together, 1 kg CS was finally converted to 48.87 g D-allulose. This yield seemed very low. However, the D-allulose yield from D-glucose was $17.56\%$ (8.78 g/L D-allulose from 50 g/L D-glucose in CS hydrolysate), which is close to the equilibrium state ($18.18\%$).
The dry CS is mainly composed of cellulose, hemicellulose, and lignin. In this study, the hydrolysis yield three main monosaccharides (70.82 g/L D-glucose, 19.13 g/L D-xylose, and 1.69 g/L L-arabinose). Both D-glucose and D-xylose can be utilized by D-glucose isomerase, and the products are D-fructose and D-xylulose respectively (Mu et al., 2012). The substrate competition between D-xylose and D-glucose decreased D-fructose production by $24.14\%$ (Figure 3D). This issue can be resolved by removing D-xylose from the solution or improving D-glucose isomerase’s substrate specificity. However, neither of them is easy to achieve. In the following research, we are going to simultaneously produce D-allulose and D-ribose, which is a important pentose with many valuable physiological functions (Mahoney et al., 2018; Li et al., 2021). D-ribose is currently produced from D-glucose through the pentose phosphate pathway (Cheng et al., 2017; Park et al., 2017). However, using this method 1 carbon is lost by releasing 1 CO2 for each D-ribose production, which is neither economical efficient or environmental friendly (Srivastava et al., 2012). By converting D-xylose to D-ribose (both of them are pentose) will not only explore novel D-ribose producing method, but also further improve the full utilization of CS.
## 5 Conclusion
In the present study, we produced D-allulose from CS by integrating the CS hydrolysis with a D-allulose producing whole-cell catalyst. *By* gene screening, co-overexpressing, copy number optimization, we obtained a high-efficient whole-cell catalyst, and used it to produce D-allulose from CS hydrolysate. We next designed a microfluidic system and used it to immobilize the whole-cell catalyst. After process optimization, the D-allulose titer improved by 8.61 times, reaching 8.78 g/L from CS hydrolysate. In this process, 1 kg CS was finally converted to 48.87 g D-allulose. The bioconversions were carried out in pure water. This study validated the feasibility of producing D-allulose from CS, a non-food feedstock.
## Data availability statement
The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding authors.
## Author contributions
YL and JX conceived the topic. QJ, HZ, FL, and YL performed the experiments. WW, QJ, and YL drafted and revised the manuscript. FL, AZ, LQ, WX, JM, WW, and MA gave suggestions.
## 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/fbioe.2023.1156953/full#supplementary-material
## References
1. Altschul S. F., Madden T. L., Schaffer A. A., Zhang J., Zhang Z., Miller W.. **Gapped blast and PSI-blast: A new generation of protein database search programs**. *Nucleic Acids Res.* (1997) **25** 3389-3402. DOI: 10.1093/nar/25.17.3389
2. Brovetto M., Gamenara D., Mendez P. S., Seoane G. A.. **C-C bond-forming lyases in organic synthesis**. *Chem. Rev.* (2011) **111** 4346-4403. DOI: 10.1021/cr100299p
3. Chen J., Jin Z., Gai Y., Sun J., Zhang D.. **A food-grade expression system for D-psicose 3-epimerase production in**. *Bioresour. Bioprocess.* (2017) **4** 9. DOI: 10.1186/s40643-017-0139-7
4. Chen X., Zaro J. L., Shen W. C.. **Fusion protein linkers: Property, design and functionality**. *Adv. Drug Deliv. Rev.* (2013) **65** 1357-1369. DOI: 10.1016/j.addr.2012.09.039
5. Chen Z., Gao X. D., Li Z.. **Recent advances regarding the physiological functions and biosynthesis of D-allulose**. *Front. Microbiol.* (2022) **13** 881037. DOI: 10.3389/fmicb.2022.881037
6. Cheng J., Zhuang W., Li N. N., Tang C. L., Ying H. J.. **Efficient biosynthesis of D-ribose using a novel co-feeding strategy in**. *Lett. Appl. Microbiol.* (2017) **64** 73-78. DOI: 10.1111/lam.12685
7. Doner L. W.. **Isomerization of D-fructose by base: Liquid-chromatographic evaluation and the isolation of D-psicose**. *Carbohyd. Res.* (1979) **70** 209-216. DOI: 10.1016/S0008-6215(00)87101-3
8. Guo N., Zheng J., Wu L. S., Jian T. A., Zhou H. B.. **Engineered bifunctional enzymes of endo-1,4-beta-xylanase/endo-1,4-beta-mannanase were constructed for synergistically hydrolyzing hemicellulose**. *J. Mol. Catal. B Enzym.* (2013) **97** 311-318. DOI: 10.1016/j.molcatb.2013.06.019
9. He W., Jiang B., Mu W., Zhang T.. **Production of D-allulose with D-psicose 3-epimerase expressed and displayed on the surface of**. *J. Agric. Food Chem.* (2016) **64** 7201-7207. DOI: 10.1021/acs.jafc.6b03347
10. Hobbs J. E.. **Food supply chains during the COVID-19 pandemic**. *Can. J. Agr. Econ.* (2020) **68** 171-176. DOI: 10.1111/cjag.12237
11. Hossain A., Yamaguchi F., Hirose K., Matsunaga T., Sui L., Hirata Y.. **Rare sugar D-psicose prevents progression and development of diabetes in T2DM model Otsuka Long-Evans Tokushima Fatty rats**. *Drug Des. Dev. Ther.* (2015) **9** 525-535. DOI: 10.2147/DDDT.S71289
12. Iida T., Hayashi N., Yamada T., Yoshikawa Y., Miyazato S., Kishimoto Y.. **Failure of D-psicose absorbed in the small intestine to metabolize into energy and its low large intestinal fermentability in humans**. *Metab. Clin. Exp.* (2010) **59** 206-214. DOI: 10.1016/j.metabol.2009.07.018
13. Izumori K.. **Bioproduction strategies for rare hexose sugars**. *Naturwissenschaften* (2002) **89** 120-124. DOI: 10.1007/s00114-002-0297-z
14. Izumori K.. **Izumoring: A strategy for bioproduction of all hexoses**. *J. Biotechnol.* (2006) **124** 717-722. DOI: 10.1016/j.jbiotec.2006.04.016
15. Juneja A., Zhang G., Jin Y.-S., Singh V.. **Bioprocessing and technoeconomic feasibility analysis of simultaneous production of D-psicose and ethanol using engineered yeast strain KAM-2GD**. *Bioresour. Technol.* (2019) **275** 27-34. DOI: 10.1016/j.biortech.2018.12.025
16. Karaoglu H., Yanmis D., Sal F. A., Celik A., Canakci S., Belduz A. O.. **Biochemical characterization of a novel glucose isomerase from**. *J. Mol. Catal. B Enzym.* (2013) **97** 215-224. DOI: 10.1016/j.molcatb.2013.08.019
17. Li A. M., Cai L., Chen Z., Wang M. Y., Wang N., Nakanishi H.. **Recent advances in the synthesis of rare sugars using DHAP-dependent aldolases**. *Carbohyd. Res.* (2017) **452** 108-115. DOI: 10.1016/j.carres.2017.10.009
18. Li S., Wang J., Xiao Y., Zhang L., Fang J., Yang N.. **D-ribose: Potential clinical applications in congestive heart failure and diabetes, and its complications (Review)**. *Exp. Ther. Med.* (2021) **21** 496. DOI: 10.3892/etm.2021.9927
19. Li Z., Cai L., Qi Q., Styslinger T. J., Zhao G., Wang P. G.. **Synthesis of rare sugars with L-fuculose-1-phosphate aldolase (FucA) from**. *Bioorg. Med. Chem. Lett.* (2011) **21** 5084-5087. DOI: 10.1016/j.bmcl.2011.03.072
20. Li Z., Li Y., Duan S., Liu J., Yuan P., Nakanishi H.. **Bioconversion of D-glucose to D-psicose with immobilized D-xylose isomerase and D-psicose 3-epimerase on**. *J. Ind. Microbiol. Biotechnol.* (2015) **42** 1117-1128. DOI: 10.1007/s10295-015-1631-8
21. Liu Y., Tang Y., Gao H., Zhang W., Jiang Y., Xin F.. **Challenges and future perspectives of promising biotechnologies for lignocellulosic biorefinery**. *Molecules* (2021) **26** 5411. DOI: 10.3390/molecules26175411
22. Lv Y., Cheng X., Du G., Zhou J., Chen J.. **Engineering of an H**. *Biotechnol. Bioeng.* (2017) **114** 2066-2074. DOI: 10.1002/bit.26319
23. Mahoney D. E., Hiebert J. B., Thimmesch A., Pierce J. T., Vacek J. L., Clancy R. L.. **Understanding D-ribose and mitochondrial function**. *Adv. Biosci. Clin. Med.* (2018) **6** 1-5. DOI: 10.7575/aiac.abcmed.v.6n.1p.1
24. Men Y., Zhu Y. M., Zeng Y., Izumori K., Sun Y. X., Ma Y. H.. **Co-expression of D-glucose isomerase and D-psicose 3-epimerase: Development of an efficient one-step production of D-psicose**. *Enzyme Microb. Tech.* (2014) **64-65** 1-5. DOI: 10.1016/j.enzmictec.2014.06.001
25. Mu W., Chu F., Xing Q., Yu S., Zhou L., Jiang B.. **Cloning, expression, and characterization of a D-psicose 3-epimerase from**. *J. Agric. Food Chem.* (2011) **59** 7785-7792. DOI: 10.1021/jf201356q
26. Mu W. M., Wang X. L., Xue Q. H., Jiang B., Zhang T., Miao M.. **Characterization of a thermostable glucose isomerase with an acidic pH optimum from**. *Food Res. Int.* (2012) **47** 364-367. DOI: 10.1016/j.foodres.2011.09.006
27. Nagata Y., Kanasaki A., Tamaru S., Tanaka K.. **D-psicose, an epimer of D-fructose, favorably alters lipid metabolism in Sprague-Dawley rats**. *J. Agric. Food Chem.* (2015) **63** 3168-3176. DOI: 10.1021/jf502535p
28. Park H.-C., Kim Y.-J., Lee C.-W., Rho Y.-T., Kang J., Lee D.-H.. **Production of D-ribose by metabolically engineered**. *Process Biochem.* (2017) **52** 73-77. DOI: 10.1016/j.procbio.2016.10.001
29. Patra A., Bera M.. **Spectroscopic investigation of new water soluble Mn**. *Carbohyd. Res.* (2014) **384** 87-98. DOI: 10.1016/j.carres.2013.12.002
30. Shen X., Sun R.. **Recent advances in lignocellulose prior-fractionation for biomaterials, biochemicals, and bioenergy**. *Carbohydr. Polym.* (2021) **261** 117884. DOI: 10.1016/j.carbpol.2021.117884
31. Song Y., Nguyen Q. A., Wi S. G., Yang J., Bae H.-J.. **Strategy for dual production of bioethanol and D-psicose as value-added products from cruciferous vegetable residue**. *Bioresour. Technol.* (2017) **223** 34-39. DOI: 10.1016/j.biortech.2016.10.021
32. Srivastava R. K., Maiti S. K., Das D., Bapat P. M., Batta K., Bhushan M.. **Metabolic flexibility of D-ribose producer strain of**. *J. Ind. Microbiol. Biotechnol.* (2012) **39** 1227-1243. DOI: 10.1007/s10295-012-1115-z
33. Wei M., Li Z., Li T., Wu B., Liu Y., Qu J.. **Transforming flask reaction into cell-based synthesis: Production of polyhydroxylated molecules via engineered**. *ACS Catal.* (2015) **5** 4060-4065. DOI: 10.1021/acscatal.5b00953
34. Xia Y., Cheng Q., Mu W., Hu X., Sun Z., Qiu Y.. **Research advances of D-allulose: An overview of physiological functions, enzymatic biotransformation technologies, and production processes**. *Foods* (2021) **10** 2186. DOI: 10.3390/foods10092186
35. Xu P., Vansiri A., Bhan N., Koffas M. a. G.. **ePathBrick: A synthetic biology platform for engineering metabolic pathways in**. *ACS Synth. Biol.* (2012) **1** 256-266. DOI: 10.1021/sb300016b
36. Yang S., Peng L., Liu E., He L., Guan Q., Zhang J.. **Development of a general kinetic model for organic acid-catalyzed hydrolysis of corn stalk**. *Cellulose* (2021) **28** 6935-6952. DOI: 10.1007/s10570-021-03977-6
37. Zhang J., Xu C., Chen X., Ruan X., Zhang Y., Xu H.. **Engineered**. *Enzyme Microb. Tech.* (2020) **136** 109531. DOI: 10.1016/j.enzmictec.2020.109531
38. Zhang W., Li H., Jiang B., Zhang T., Mu W.. **Production of D-allulose from D-glucose by**. *J. Sci. Food Agric.* (2017) **97** 3420-3426. DOI: 10.1002/jsfa.8193
39. Zhang W., Fang D., Xing Q., Zhou L., Jiang B., Mu W.. **Characterization of a novel metal-dependent D-psicose 3-epimerase from**. *Plos One* (2013) **8** e62987. DOI: 10.1371/journal.pone.0062987
40. Zhang W., Jia M., Yu S., Zhang T., Zhou L., Jiang B.. **Improving the thermostability and catalytic efficiency of the D-psicose 3-epimerase from**. *J. Agric. Food Chem.* (2016) **64** 3386-3393. DOI: 10.1021/acs.jafc.6b01058
41. Zhang W. L., Li H., Zhang T., Jiang B., Zhou L., Mu W. M.. **Characterization of a D-psicose 3-epimerase from**. *J. Mol. Catal. B Enzym.* (2015) **120** 68-74. DOI: 10.1016/j.molcatb.2015.05.018
42. Zhang W. L., Zhang T., Jiang B., Mu W. M.. **Enzymatic approaches to rare sugar production**. *Biotechnol. Adv.* (2017) **35** 267-274. DOI: 10.1016/j.biotechadv.2017.01.004
43. Zhang W., Yu S., Zhang T., Jiang B., Mu W.. **Recent advances in D-allulose: Physiological functionalities, applications, and biological production**. *Trends Food Sci. Technol.* (2016b) **54** 127-137. DOI: 10.1016/j.tifs.2016.06.004
44. Zhang X. O., Wang H. B., Zhang Y., Lu X., Chen L. L., Yang L.. **Complementary sequence-mediated exon circularization**. *Cell* (2014) **159** 134-147. DOI: 10.1016/j.cell.2014.09.001
45. Zhang Y., Wang H., Sun X., Wang Y., Liu Z.. **Separation and characterization of biomass components (cellulose, hemicellulose, and lignin) from corn stalk**. *Bioresources* (2021) **16** 7205-7219. DOI: 10.15376/biores.16.4.7205-7219
46. Zhong X., Ma Y., Zhang X., Zhang J., Wu B., Hollmann F.. **More efficient enzymatic cascade reactions by spatially confining enzymes via the SpyTag/SpyCatcher technology**. *Mol. Catal.* (2022) **521** 112188. DOI: 10.1016/j.mcat.2022.112188
47. Zhu Y., Men Y., Bai W., Li X., Zhang L., Sun Y.. **Overexpression of D-psicose 3-epimerase from**. *Biotechnol. Lett.* (2012) **34** 1901-1906. DOI: 10.1007/s10529-012-0986-4
|
---
title: A controlled open clinical trial of the positive effect of a physical intervention
on quality of life in schizophrenia
authors:
- Viviane Batista Cristiano
- Michele Fonseca Szortyka
- Paulo Belmonte-de-Abreu
journal: Frontiers in Psychiatry
year: 2023
pmcid: PMC9998925
doi: 10.3389/fpsyt.2023.1066541
license: CC BY 4.0
---
# A controlled open clinical trial of the positive effect of a physical intervention on quality of life in schizophrenia
## Abstract
### Justification
Schizophrenia is a severe mental disorder associated with important physical (obesity and low motor functional capacity) and metabolic (diabetes and cardiovascular diseases) changes that contribute to a more sedentary lifestyle and a low quality of life.
### Objective
The study aimed to measure the effect of two different protocols of physical exercise [aerobic intervention (AI) versus functional intervention ([FI)] on lifestyle in schizophrenia compared with healthy sedentary subjects.
### Methodology
A controlled clinical trial involving patients diagnosed with schizophrenia from two different locations [Hospital de Clinicas de Porto Alegre (HCPA) and Centro de Atenção Psicosocial (CAPS) in the city of Camaquã] was carried out. The patients undertook two different exercise protocols (IA: 5-min warm-up of comfortable intensity; 45 min of aerobic exercise of increasing intensity using any of the three modalities—a stationary bicycle, a treadmill, or an elliptical trainer; and 10 min of global stretching of large muscle groups; and FI: a 5 min warm-up with a stationary walk; 15 min of muscle and joint mobility exercises; 25 min of global muscle resistance exercises; and 15 min of breathing body awareness work) twice a week for 12 weeks and were compared with physically inactive healthy controls. Clinical symptoms (BPRS), life quality (SF-36), and physical activity levels (SIMPAQ) were evaluated. The significance level was p ≤ 0.05.
### Results
The trial involved 38 individuals, of which 24 from each group performed the AI, and 14 from each group underwent the FI. This division of interventions was not randomized but was instead decided upon for convenience. The cases showed significant improvements in quality of life and lifestyle, but these differences were greater in the healthy controls. Both interventions were very beneficial, with the functional intervention tending to be more effective in the cases and the aerobic intervention more effective in the controls.
### Conclusion
Supervised physical activity improved life quality and reduced sedentary lifestyle in adults with schizophrenia.
## Introduction
Severe mental disorders, such as schizophrenia, are associated with various personal impairments, including cognitive, physical, and metabolic changes [1, 2]. These, in turn, are associated with other diseases (cardiovascular, obesity, and diabetes) (1–3) that are the consequences of an unhealthy lifestyle, mental illness manifestations, and the side effects of drug treatment.
Furthermore, this population undertakes little physical activity, which contributes to these additional metabolic and cardiovascular pathologies, such as diabetes, acute myocardial infarction, and stroke (1–4). Additionally, the diet of these individuals is a significant factor to be considered as obesity and a sedentary lifestyle will lead to the development of metabolic and cardiovascular diseases, particularly considering that many drugs used to control the disease also contribute to weight gain [5].
All these scenarios of a sedentary lifestyle, obesity, and metabolic and/or cardiovascular diseases directly influence the quality of life of these individuals. Therefore, several studies have evaluated the effect of physical activity on people with schizophrenia and demonstrated that increased physical activity may induce changes in functional capacity, social interaction, and pain tolerance, even when compared to sedentary individuals without mental disorders [6, 7]. These studies suggest that regular exercise may positively affect individuals with schizophrenia, especially those who are sedentary and overweight. This may occur because of the convergence of increased cellular mitosis, increased metabolism, increased production of endorphins and neurotrophic factors, and muscle and neuronal plasticity (8–12). With this hypothesis in mind, many recent studies have focused on physical activity in this population. Almost all the physical activities mentioned were related to aerobic exercise because it has well-known oxidative effects and results can be achieved in a relatively short period of time (on average, 8 weeks are sufficient for systemic responses [13]. Another format of physical activity that can be very beneficial for this population and that has not yet been studied is functional training as it uses everyday movements, such as sitting and standing, pulling and pushing, and spinning, thus favoring and stimulating the individual’s autonomy.
The main results of the studies focused on cognitive issues, functional capacity, weight, and biomarkers, revealing positive effects in this population. However, most failed to study the impact on lifestyle and compare it with other physical activities, such as anaerobic exercise. This points to the need for additional data comparing the effect of different types of physical exercise on different health outcomes. Therefore, this study aims to measure the effect of two different physical activity protocols, aerobic intervention (AI) versus functional intervention (FI), on lifestyle in individuals with schizophrenia compared with healthy sedentary individuals.
## Trial design
In this section, we describe the clinical trial of physical intervention [aerobic physical intervention (AI) and functional physical intervention (FI)] in two groups of stable outpatients with a diagnosis of schizophrenia (SCZ) and one group of healthy sedentary controls. The AI group received regular care at a public health facility [Psychosocial Attention Center (CAPS)]. Patients under continued outpatient care at CAPS-Camaquã in the surrounding cities of Metropolitan Porto Alegre in southern Brazil received AI, and patients under regular care at a university-based hospital [schizophrenia outpatient clinic (Prodesq) of Hospital de Clínicas de Porto Alegre (HCPA)] received FI.
## Participants
Stable outpatients under regular treatment received prior psychiatric diagnosis after a three-step procedure consisting of the following: (a) careful clinical observation with at least three evaluations; (b) a family interview; and (c) a review of their medical records performed by a trained psychiatrist. All met the following inclusion criteria: Diagnostic and Statistical Manual of Mental Disorders, DSM-5; [14] diagnosis of schizophrenia; aged between 18 and 65 years; under stable drug treatment adjusted to their clinical state for at least 3 months; and not involved in any other physical activity programs during the intervention. Patients were recruited from the services where they were being clinically followed (HCPA or CAPES) and were allocated to the intervention group (individuals from HCPA underwent IF, and individuals from CAPES underwent IA) (Figure 1).
**FIGURE 1:** *Study flowchart: Aerobic intervention (AI) and functional intervention (FI) in patients with schizophrenia.*
The exclusion criteria were as follows: alcohol or other drug abuse in the previous month; major systemic or neurological diseases; physical disability contraindicating physical activity or any physical condition that makes physical activity unsafe; suicide risk confirmed by direct contact with the patient and family; pregnancy or women of reproductive age that did not use a contraception method; and not agreeing to participate in the study after full explanation of the program.
Controls were recruited through specific social networks (Figure 2). Then, they were paired by sex, age (3 years older or younger), and social class [we followed the classification criteria by classes of the Brazilian Institute of Geography and Statistics (IBGE), which uses the monthly income of all the residents of the same house to list from the richest to the poorest]. Thus, they were divided into the following classes: A (monthly income above R$ 20,900), B (monthly income between R$ 10,450.01 and R$ 20,900), C (monthly income above R$ 4,180 but up to R$ 10,450), D (monthly income between R$ 2,090.01 and R$ 4,180), and E (monthly income of no more than R$ 2,090).
**FIGURE 2:** *Study flowchart: Aerobic intervention (AI) and functional intervention (FI) in controls.*
The absence of any major mental illness was defined by a direct interview in which questions about life experiences of memory loss, psychosis (delusions and/or hallucinations), depression, mania, generalized anxiety disorder, and obsessive-compulsive symptoms were asked. Additionally, the subjects were asked about regular physical activity as they were supposed to be sedentary. Exclusion criteria were the same as those applied to patients with SCZ.
## Ethical standards
The authors assert that all procedures contributing to this work complied with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008.
The study was registered in the Brazilian Research Registry (43408615.7.0000.5327) and the Brazilian Registry of Clinical Trials (ReBEC, 1 No. RBR-2h2hjy, registration date $\frac{09}{29}$/2020, study start date $\frac{07}{22}$/2017) and approved [150066] by the Research Ethics Committee of Hospital de Clínicas de Porto Alegre (HCPA). Patients and their legal guardians provided written informed consent after reading and understanding the intervention program and their rights. Registration took place later because we thought it was not a randomized clinical trial, so there would be no need.
## Clinical assessment
After patient recruitment, previously trained professionals performed a standardized clinical and physical assessment of the study participants before physical intervention and after 3 months of treatment.
## Sedentary lifestyle
The Simple Physical Activity Questionnaire (SIMPAQ) [15] is a 5-item clinical tool designed to assess physical activity among populations at high risk of sedentary behavior. The questionnaire evaluates the last 7 days, including time in bed, sedentary time, time spent walking, type and time spent exercising, and time spent in other activities, including leisure, domestic, work, and transportation activities. It was simultaneously developed and validated in several languages, including English, Spanish, and Portuguese. Following its validation, our group pioneered its use in clinical research.
## Disease severity
The Brief Psychiatric Rating Scale (BPRS) is one of the most used instruments to assess the presence and severity of various psychiatric symptoms and has been in the public domain since 1965 [16]. It is valid in several languages, including Portuguese [17], and in Brazil, it is used by the Brazilian Unified Health System (SUS) for monitoring patients. This tool assesses 18 symptom domains, such as somatic worry, anxiety, emotional withdrawal, conceptual disorganization, feelings of guilt, tension, mannerisms and posture, grandiosity, depressed mood, hostility, distrust, hallucinatory behavior, motor retardation, lack of cooperation, unusual thought content, blunted affect, excitement, and disorientation. The assessment takes approximately 5–10 min after an interview with the patient. The clinician then rates each item on a scale ranging from 0 (absent) to 6 (extremely severe) through observation and questioning, depending on the item assessed.
## Quality of life
The Medical Outcomes Study 36-Item Short Form (SF-36) is a commonly used validated questionnaire that is provided in several languages, including Portuguese [18], and has high sensitivity in detecting functional status, among other aspects of quality of life. The SF-36 questionnaire includes eight multiple-item subscales that evaluate functional capacity, physical limitation, pain, general health, vitality, social aspects, emotional limitations, and mental health. The total score on each SF-36 subscale ranges between 0 and 100; the higher the score, the better the patient is.
## Physical intervention
The physical intervention for cases and controls followed an initial assessment that took place after the consent form was read and signed and measured disease severity (BPRS) (cases only), quality of life (SF-36), and physical activity level (SIMPAQ). The aerobic or functional physical intervention program lasted 12 weeks in healthy cases and controls. Patients continued with regular clinical treatment in addition to standardized activity, and after completion of the intervention program, revaluation was performed using all the tests and questionnaires mentioned above.
The aerobic protocol was as follows: 24 patients diagnosed with SCZ were paired with 24 sedentary controls without mental illness. The program lasted 12 weeks and consisted of 1-h aerobic exercise sessions twice a week. The sessions were carried out individually or at most in pairs and monitored by a physiotherapist blinded to the evaluations. The participants were monitored using a Polar FT1® frequency meter with results adjusted for age, sex, weight, and height. Measurements ranged from 70 to $80\%$ of the maximum heart rate calculated using Karvonen’s formula.
A standard aerobic session consisted of the following: a 5-min warm-up at a comfortable intensity followed by aerobic exercise of increasing intensity with one of three modalities: (a) a bicycle ergometer (Embreex 367C, Brazil), (b) a treadmill (Embreex 566BX, Brazil), or (c) an elliptical trainer (Embreex 219, Brazil). This strategy was consistent with public health recommendations that suggest tailoring the program to individual preferences, which has been proven to be feasible in patients diagnosed with SCZ. A trained professional coordinated the intervention sessions with guidance and equipment adjustments and encouraged each participant to perform the exercises in the best way possible. After completing the aerobic exercise, participants globally stretched the major muscle groups.
The functional protocol was as follows: 14 patients diagnosed with SCZ were paired with 14 sedentary controls without mental illness. The program lasted 12 weeks and consisted of 1-h physical function training sessions twice a week. The participants carried out the program in trios or quartets and were trained by a physical therapist blinded to the evaluations.
A standard session consisted of the following: a 5-min warm-up with stationary walking, followed by 15 min of muscle and joint mobility exercises. Then, 25 min of global muscle endurance exercises (paravertebrae, abdominals, extensors, flexors, adductors, hip abductors, flexors and extensors of the shoulders, knees, and elbows) based on the basic movements of functional training (sit and stand, pull and push, and rotate and advance) were performed, followed by 15 min of respiratory body awareness work. A maximum number of repetitions were performed in 30 s (only once per exercise) and accessories such as balls, elastic bands, and dumbbells were used according to the level of resistance required.
## Statistical analyses
The sample size calculation for this study was performed using the WinPepi program (with the evaluation of five co-variables), using a previous study [7] as a baseline. This calculation estimated a minimum number of 30 patients in each group (group 1 patients with SCZ and group 2 controls, for 60 subjects). The normality of the data distribution was assessed using the Kolmogorov–Smirnov test. Quantitative variables with a normal distribution were presented as mean ± standard deviations, while variables with an asymmetric distribution were presented as median and interquartile ranges. Student’s paired t-test/independent t-test or the Wilcoxon test/Mann–Whitney test were used to compare normal and asymmetric variables, respectively. ANOVA followed by Bonferroni correction was used for comparisons between three or more groups. Relationships between two variables were assessed through Spearman correlation coefficients. Categorical variables were presented as frequencies and analyzed using the Pearson chi-square test, Fisher exact test, or McNemar test. The main outcome measure was assessed using generalized linear model (GLM) analysis (gamma distribution), and confounding factors were determined based on statistical criteria (association with either study factor and the outcome with a p ≤ 0.2). The significance level was p ≤ 0.05. The analysis of effect size was used to evaluate the magnitude of the difference derived from GLM. SPSS Statistics 22.0 was used to process and analyze data.
## Results
Five of 43 individuals with SCZ who started the exercise protocol were excluded for not having the minimum required frequency ($80\%$, 5 absences over 24 appointments were allowed), whereas in the control group, 49 started, and 11 were excluded for the same reason. Thus, the final sample numbers in both groups was 38 individuals, of which 24 from each group performed the aerobic intervention, and 14 from each group performed the functional intervention. This division of interventions was not randomized but was instead decided upon for convenience. The sociodemographic and clinical data of the samples are shown in Table 1, where we can observe that the groups were homogeneous for sex, weight, and BMI, showing statistical differences only in age and height.
**TABLE 1**
| Variables/Group | Cases | Controls | P-value |
| --- | --- | --- | --- |
| Gender n (%) | Gender n (%) | Gender n (%) | Gender n (%) |
| Male | 32 (84.2) | 32 (84.2) | – |
| Female | 6 (15.8) | 6 (15.8) | |
| Schooling n (%) | Schooling n (%) | Schooling n (%) | Schooling n (%) |
| Basic education (up to 12 years of schooling) | 38 (100) | 27 (71.1) | – |
| Higher education (more than 12 years of schooling) | – | 11 (28.9) | – |
| Marital status n (%) | Marital status n (%) | Marital status n (%) | Marital status n (%) |
| Single | 37 (97.4) | 11 (28.9) | – |
| Married | 1 (2.6) | 27 (71.1) | |
| Smoking n (%) | Smoking n (%) | Smoking n (%) | Smoking n (%) |
| Yes | 14 (36.8) | – | – |
| No | 24 (63.2) | 38 (100) | – |
| Years of illness n (%) | Years of illness n (%) | Years of illness n (%) | Years of illness n (%) |
| <7 years | 4 (10.5) | – | – |
| >7 years | 34 (89.5) | – | – |
| Antipsychotic medication n (%) | Antipsychotic medication n (%) | Antipsychotic medication n (%) | Antipsychotic medication n (%) |
| Typical antipsychotic | 13 (34.2) | – | – |
| Atypical antipsychotic | 17 (44.7) | – | – |
| Combination of typical and atypical | 8 (21.1) | – | – |
| Age (years, mean ± SD) | 40.95 ± 11.37 | 41.68 ± 11.22 | 0.039* |
| Weight (kg, mean ± SD) | 83.77 ± 23.56 | 88.66 ± 18.51 | 0.274 |
| Height (m, mean ± SD) | 1.69 ± 0.080 | 1.73 ± 0.070 | 0.011* |
| BMI (mean ± SD) | 29.23 ± 7.96 | 29.55 ± 5.88 | 0.829 |
| Psychiatric hospitalizations median (p. 25–75) | 2.00 (0.75–4.00) | – | – |
## Brief Psychiatric Rating Scale
Table 2 shows the results of the BPRS clinical scale in the different protocols for both the pre- and post-moments in the case group. There was a worsening of hostility symptoms ($$p \leq 0.02$$) in the AI group.
**TABLE 2**
| Unnamed: 0 | Total cases (mean ± SD) | Total cases (mean ± SD).1 | Total cases (mean ± SD).2 | Total cases (mean ± SD).3 |
| --- | --- | --- | --- | --- |
| Intervention | Aerobic (n = 24) | Aerobic (n = 24) | Functional (n = 14) | Functional (n = 14) |
| Time | Before | After | Before | After |
| Variable | Variable | Variable | Variable | Variable |
| Anxiety and depression 5-2-9-1 | 5.71 ± 4.83 | 5.75 ± 5.11 | 5.29 ± 5.34 | 7.36 ± 4.88 |
| Retardation 13-16 3-18-14-7 | 5.67 ± 6.58 | 6.38 ± 6.86 | 5.21 ± 5.18 | 5.86 ± 4.47 |
| Thinking disorder 11-15-12-4-10-8 | 6.33 ± 5.30 | 7.04 ± 5.93 | 7.36 ± 7.68 | 8.64 ± 8.19 |
| Activation 7-6-17-8 | 1.00 ± 2.25 | 1.96 ± 2.97 | 1.29 ± 2.61 | 1.79 ± 2.49 |
| Hostility 10-11-14-8 | 2.17 ± 2.16 | 3.83 ± 4.29 | 2.57 ± 3.74 | 3.93 ± 4.25 |
## 36-Item short form quality of life scale
Patients with SCZ showed an increase of approximately $20\%$ in almost all domains of the SF-36, except for the pain domain, which decreased in the AI group. Even with this clear improvement, only two domains in each intervention (AI, functional capacity, $p \leq 0.001$, and limitations by emotional aspects, $$p \leq 0.014$$; FI, pain, $$p \leq 0.002$$, and limitations by emotional aspects, $$p \leq 0.039$$) changed significantly (Table 3).
**TABLE 3**
| Unnamed: 0 | Patients (mean ± SD) | Patients (mean ± SD).1 | Patients (mean ± SD).2 | Patients (mean ± SD).3 | Controls (mean ± SD) | Controls (mean ± SD).1 | Controls (mean ± SD).2 | Controls (mean ± SD).3 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Intervention | Aerobic (n = 24) | Aerobic (n = 24) | Functional (n = 14) | Functional (n = 14) | Aerobic (n = 24) | Aerobic (n = 24) | Functional (n = 14) | Functional (n = 14) |
| Time | Before | After | Before | After | Before | After | Before | After |
| Variable | Variable | Variable | Variable | Variable | Variable | Variable | Variable | Variable |
| Functional capacity | 64.38 ± 31.98 | 80.65 ± 20.36 | 69.29 ± 26.23 | 73.21 ± 22.41 | 75.83 ± 16.66 | 85.25 ± 12.24 | 78.93 ± 20.86 | 84.29 ± 17.74 |
| Physical limitations | 45.83 ± 45.25 | 57.61 ± 37.26 | 39.29 ± 47.75 | 66.07 ± 38.44 | 68.83 ± 31.82 | 79.79 ± 27.99 | 61.79 ± 29.52 | 80.36 ± 24.53 |
| Pain | 68.79 ± 25.99 | 61.65 ± 34.43 | 57.36 ± 32.37 | 80.36 ± 18.25 | 59.67 ± 27.15 | 74.67 ± 23.87 | 59.57 ± 11.57 | 80.36 ± 12.48 |
| General health | 46.79 ± 23.17 | 54.82 ± 23.28 | 55.00 ± 17.72 | 59.29 ± 21.38 | 53.42 ± 17.57 | 61.92 ± 18.39 | 62.79 ± 14.90 | 77.14 ± 16.02 |
| Vitality | 56.67 ± 26.89 | 66.14 ± 29.40 | 56.79 ± 18.57 | 62.14 ± 26.73 | 62.08 ± 18.82 | 71.04 ± 17.51 | 65.36 ± 15.12 | 74.64 ± 17.27 |
| Social aspects | 56.33 ± 27.00 | 64.22 ± 24.83 | 56.79 ± 29.64 | 69.64 ± 31.28 | 73.75 ± 28.06 | 83.21 ± 22.12 | 74.11 ± 21.63 | 82.14 ± 21.21 |
| Emotional limitations | 38.79 ± 44.66 | 60.50 ± 44.43 | 38.09 ± 48.67 | 66.66 ± 43.39 | 70.50 ± 30.57 | 84.33 ± 14.79 | 73.81 ± 26.76 | 78.61 ± 30.95 |
| Mental health | 64.00 ± 28.48 | 66.00 ± 26.84 | 62.86 ± 28.13 | 68.86 ± 20.29 | 73.96 ± 20.79 | 81.63 ± 19.50 | 72.00 ± 11.09 | 78.57 ± 17.37 |
By contrast, the control group improved significantly in seven of the SF-36 domains in the AI (functional capacity $p \leq 0.001$; limitations by physical aspects, $$p \leq 0.013$$; pain, $$p \leq 0.001$$; vitality, $$p \leq 0.002$$; social aspects, $$p \leq 0.0271$$; limitations by emotional aspects, $$p \leq 0.014$$; and mental health, $$p \leq 0.024$$) and in four domains in the FI (limitations by physical aspects, $$p \leq 0.013$$; pain, $p \leq 0.001$; general health status, $p \leq 0.001$; and vitality, $$p \leq 0.049$$) (Table 3).
## Simple Physical Activity Questionnaire
The AI case group showed a significant change in exercise time from 16 to 126 min per week ($p \leq 0.001$) but not in the other SIMPAQ items. The FI cases, on the other hand, also showed significant improvement in exercise time from 0 to 110 min per week ($p \leq 0.001$) and in weekly walking time (from 108 to 177 min/week; $p \leq 0.001$) (Table 4).
**TABLE 4**
| Unnamed: 0 | Patients (mean ± SD) | Patients (mean ± SD).1 | Patients (mean ± SD).2 | Patients (mean ± SD).3 | Controls (mean ± SD) | Controls (mean ± SD).1 | Controls (mean ± SD).2 | Controls (mean ± SD).3 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Intervention | Aerobic (n = 24) | Aerobic (n = 24) | Functional (n = 14) | Functional (n = 14) | Aerobic (n = 24) | Aerobic (n = 24) | Functional (n = 14) | Functional (n = 14) |
| Time | Before | After | Before | After | Before | After | Before | After |
| Variable | Variable | Variable | Variable | Variable | Variable | Variable | Variable | Variable |
| Time in bed (min) | 607.5 ± 114.07 | 586.25 ± 84.84 | 660.36 ± 133.14 | 615.71 ± 75.72 | 466.25 ± 82.51 | 422.92 ± 67.02 | 437.14 ± 57.27 | 420.71 ± 73.85 |
| Sedentary time (min) | 365.0 ± 144.73 | 368.75 ± 163.0 | 268.57 ± 250.60 | 225.0 ± 179.95 | 427.71 ± 107.52 | 337.29 ± 111.79 | 387.76 ± 236.65 | 465.0 ± 285.52 |
| Time spent walking (min) | 161.25 ± 211.87 | 195.42 ± 271.50 | 108.57 ± 206.24 | 177.14 ± 191.60 | 132.92 ± 150.52 | 309.88 ± 309.40 | 269.29 ± 247.87 | 338.57 ± 258.57 |
| Type and time spent for exercises (min) | 16.67 ± 44.40 | 126.67 ± 140.76 | 0 | 110,0 ± 47.72 | 20.0 ± 70.03 | 202.5 ± 175.46 | 137.86 ± 116.57 | 207.57 ± 107.42 |
| Time spent for other physical activities (min) | 34.58 ± 76.44 | 42.92 ± 75.84 | 4.29 ± 16.04 | 8.57 ± 24.76 | 69.38 ± 107.42 | 158.54 ± 180.42 | 15.00 ± 27.10 | 38.93 ± 51.60 |
The AI control group showed significant improvement in three measures: (a) sedentary time (from 427 to 337 min/day; $$p \leq 0.007$$), (b) walking time (from 132 to 309 min/week; $$p \leq 0.004$$), and (c) exercise time (from 20 to 202 min/week; $p \leq 0.001$). The FI control, on the other hand, had only showed a significant improvement in exercise time (from 137 to 207 min/week; $p \leq 0.001$) (Table 4).
## Discussion
The results of the study on quality of life and levels of physical inactivity are promising and unprecedented, as the cases were compared with healthy controls and submitted to two different interventions. Of the eight domains of quality of life assessed by the SF-36, only two showed statistically significant differences in the AI cases (functional capacity and limitations by emotional aspects) in the comparison between before and after. Additionally, only two domains were statistically significantly different in the FI cases (pain and limitations by emotional aspects). The healthy AI controls showed significant improvement in seven domains of quality of life (pain, limitations by physical aspects, limitations by emotional aspects, general health status, social aspects, functional capacity, and vitality domain), whereas the FI controls improved in four domains in the before and after comparison (limitations by physical aspects, pain, general health status, and vitality). These results allow us to conclude that the benefits of physical activity, independent of the protocol, were superior in healthy controls, and this should be better investigated in future studies. Could the disease be responsible for this difference in response?
The sedentary lifestyle levels assessed by SIMPAQ were different from those of quality of life, with the healthy control group AI and the case group FI showing more significant results in the before and after comparison. This demonstrates that the groups respond better to a specific type of physical exercise, even though they are homogeneous and have the same previous levels of physical inactivity.
Most of the studies available in the literature with this population (SCZ) proposed an aerobic-type activity and their results, in the great majority, were positive in the symptoms of the disease, however our findings were opposite to them, highlighting the hostility variable of the BPRS, where individuals who performed aerobic exercise showed a significant worsening of symptoms (9–11). What we can safely say is that both interventions produced positive results in cases and in controls in terms of quality of life, favoring the change in the lifestyle of these individuals, taking them out of the “comfort zone” of their sedentary lifestyle, which had made them inactive, and leading them to a more healthy life. After all, physical inactivity is responsible for numerous pathologies, such as obesity and cardiovascular diseases, which are both very prevalent with SCZ. Data from the WHO indicate that the world population is experiencing a sedentary lifestyle; therefore, any physical activity should be encouraged and should last at least 75–150 min per week [19].
When we look at each variable individually, we can see that the quality of life pain domain improved scores in cases and in controls. However, although this difference occurred for both the AI and FI in the controls, it only occurred for the FI in the cases. The same occurred for the variable time spent in other activities of SIMPAQ. These two differences suggest that the functional protocol tends to produce better responses in SCZ patients. Few studies [20, 21] have evaluated this in terms of functional and/or postural activities in this population; therefore, it is necessary to explore new lines of research with this type of physical activity. Thus, we will feel more confident when creating protocols. Furthermore, our results emphasize the correlation between the quality of life and the physical activity levels in this population, a hypothesis that has already been discussed previously with psychoses [22]. Another critical variable to be considered in studies on physical activities in this population is sleep, as recent studies have shown its correlation with sports performance, the development of chronic and inflammatory diseases (diabetes and cardiovascular diseases), and mental disorders (depression and psychoses) [23, 24].
## Limitations
Our study has significant limitations, such as the small sample size. Additionally, our sample was mostly male and was already in the chronic course of the disease. Some of these factors occurred due to a selection bias, in which there was a higher male prevalence in the centers of origin (CAPES and HCPA) of individuals SCZ; to avoid interference in the results, statistical adjustments were made. Furthermore, because we depended on a specific structure and specific equipment to carry out the physical exercise protocols, we divided the groups by convenience and not by randomization. It would have been interesting to divide the group of individuals with SCZ into sedentary and non-sedentary groups for a better comparison with healthy controls. Still, we were again restricted to the profile of the patients in the centers studied, where the majority were sedentary. We suggest further research should include a nutritional control associated with physical activity, as well as additional clinical questionnaires about the disease.
## Conclusion
The results of our study are significant and unprecedented for this population. We were pioneers in comparing different physical protocols and assessing lifestyle, as well as the importance of quality of life. Our findings, although preliminary, prove the effectiveness of the practice of guided physical activity among patients with a severe mental disorder, such as SCZ, and serve to guide further studies.
## 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 Research Ethics Committee of Hospital de Clínicas de Porto Alegre (HCPA). The patients/participants provided their written informed consent to participate in this study.
## Author contributions
All authors listed have made a substantial, direct, and intellectual contribution to the work, and approved it for publication.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## References
1. Vancampfort D, Correll C, Galling B, Probst M, De Hert M, Ward P. **Diabetes mellitus in people with schizophrenia, bipolar disorder and major depressive disorder: a systematic review and large scale meta-analysis.**. (2016) **15** 166-74. DOI: 10.1002/wps.20309
2. Moreno-Küstner B, Martín C, Pastor L. **Prevalence of psychotic disorders and its association with methodological issues: a systematic review and meta-analyses.**. (2018) **13**. DOI: 10.1371/journal.pone.0195687
3. Nemani K, Li C, Olfson M, Blessing E, Razavian N, Chen J. **Association of psychiatric disorders with mortality among patients with covid-19.**. (2021) **78** 380-6. DOI: 10.1001/jamapsychiatry.2020.4442
4. Soares D, Carvalho D, Ribeiro M, Diniz E, Rêgo AN. **Prevalence and predictors of treatment-resistant schizophrenia in a tertiary hospital in Northeast Brazil.**. (2021) **43** 270-7. DOI: 10.47626/2237-6089-2020-0151
5. Alvarez-Herrera S, Escamilla R, Medina-Contreras O, Saracco R, Flores Y, Hurtado-Alvarado G. **Immunoendocrine peripheral effects induced by atypical antipsychotics.**. (2020) **21**. DOI: 10.3389/fendo.2020.00195
6. Vancampfort D, Stubbs B, Mitchell A, De Hert M, Wampers M, Ward P. **Risk of metabolic syndrome and its components in people with schizophrenia and related psychotic disorders, bipolar disorder and major depressive disorder: a systematic review and meta-analysis.**. (2015) **14** 339-47. DOI: 10.1002/wps.20252
7. Cristiano V, Vieira Szortyka M, Lobato M, Ceresér K, Belmonte-de-Abreu P. **Postural changes in different stages of schizophrenia is associated with inflammation and pain: a cross-sectional observational study.**. (2016) **21** 104-11. DOI: 10.1080/13651501.2016.1249892
8. Vancampfort D, Probst M, Sweers K, Maurissen K, Knapen J, De Hert M. **Relationships between obesity, functional exercise capacity, physical activity participation and physical self-perception in people with schizophrenia.**. (2011) **123** 423-30. PMID: 21219266
9. Firth J, Stubbs B, Vancampfort D, Schuch F, Lagopoulos J, Rosenbaum S. **Effect of aerobic exercise on hippocampal volume in humans: a systematic review and meta-analysis.**. (2018) **166** 230-8. DOI: 10.1016/j.neuroimage.2017.11.007
10. Abdul Rashid N, Nurjono M, Lee J. **Clinical determinants of physical activity and sedentary behavior in individuals with schizophrenia.**. (2019) **46** 62-7. DOI: 10.1016/j.ajp.2019.10.004
11. Ospina L, Wall M, Jarskog L, Ballon J, McEvoy J, Bartels M. **Improving cognition via exercise (ICE): study protocol for a multi-site, parallel-group, single-blind, randomized clinical trial examining the efficacy of aerobic exercise to improve neurocognition, daily functioning, and biomarkers of cognitive change in individuals with schizophrenia.**. (2019) **4**. DOI: 10.20900/jpbs.20190020
12. De Sousa R, Improta-Caria A, Aras-Júnior R, de Oliveira E, Soci Ú, Cassilhas R. **Physical exercise effects on the brain during COVID-19 pandemic: links between mental and cardiovascular health.**. (2021) **42** 1325-34. DOI: 10.1007/s10072-021-05082-9
13. Neufer P. **The effect of detraining and reduced training on the physiological adaptations to aerobic exercise training.**. (1989) **8** 302-20. DOI: 10.2165/00007256-198908050-00004
14. 14.American Psychiatric Association [APA].
Diagnostic and Statistical Manual of Mental Disorders, (DSM-5).
5th ed. Arlington, VA: American Psychiatric Association (2013).. (2013)
15. Rosenbaum S, Morell R, Abdel-Baki A, Ahmadpanah M, Anilkumar T, Baie L. **Assessing physical activity in people with mental illness: 23-country reliability and validity of the simple physical activity questionnaire (SIMPAQ).**. (2020) **20**. DOI: 10.1186/s12888-020-2473-0
16. Overall J, Gorham D. **The brief psychiatric rating scale.**. (1962) **10** 799-812
17. Crippa J, Sanches R, Hallak J, Loureiro S, Zuardi A. **Factor structure of Bech’s version of the Brief psychiatric rating scale in Brazilian patients.**. (2002) **35** 1209-13. DOI: 10.1590/s0100-879x2002001000014
18. Ciconelli R, Ferraz M, Santos W, Meinão I, Quaresma M. **Tradução para a língua portuguesa e validação do questionário genérico de avaliação de qualidade de vida SF-36 (Brasil SF-36)/Brazilian-Portuguese version of the SF-36. A reliable and valid quality of life outcome measure.**. (1999) **39** 143-50
19. 19.World Health Organization [WHO].
Physical Activity.
Geneva: World Health Organization (2020).. (2020)
20. Akbas E, Erdem E, Gunes Yalcin E, Özkan T, Kinikli G. **Effects of pilates-based exercises on functional capacity and mental health in individuals with schizophrenia: a pilot study.**. (2021) **24** 1-9. DOI: 10.1080/09593985.2021.1929613
21. Tréhout M, Leroux E, Bigot L, Jego S, Leconte P, Reboursière E. **A web-based adapted physical activity program (e-APA) versus health education program (e-HE) in patients with schizophrenia and healthy volunteers: study protocol for a randomized controlled trial (PEPSY V@Si).**. (2021) **271** 325-37. DOI: 10.1007/s00406-020-01140-z
22. Gouveia R, Ferreira-Junior A, Schuch F, Zanetti G, da Silva A, Del-Bem C. **Physical activity, quality of life and global functioning an early stages of psychosis.**. (2020) **32** 373-9. DOI: 10.24869/psyd.2020.373
23. Adams M, Katz D, Shenson D. **A healthy lifestyle composite measure: significance and potential uses.**. (2016) **84** 41-7. DOI: 10.1016/j.ypmed.2015.12.005
24. Werneck A, Vancampfort D, Stubbs B, Silva D, Cucato G, Christofaro D. **Prospective associations between multiple lifestyle behaviors and depressive symptoms.**. (2022) **301** 233-9. DOI: 10.1016/j.jad.2021.12.131
|
---
title: 'Effect of the chronic medication use on outcome measures of hospitalized COVID-19
patients: Evidence from big data'
authors:
- Mohammad-Reza Malekpour
- Mohsen Abbasi-Kangevari
- Ali Shojaee
- Sahar Saeedi Moghaddam
- Seyyed-Hadi Ghamari
- Mohammad-Mahdi Rashidi
- Alireza Namazi Shabestari
- Mohammad Effatpanah
- Mohammadmehdi Nasehi
- Mehdi Rezaei
- Farshad Farzadfar
journal: Frontiers in Public Health
year: 2023
pmcid: PMC9998941
doi: 10.3389/fpubh.2023.1061307
license: CC BY 4.0
---
# Effect of the chronic medication use on outcome measures of hospitalized COVID-19 patients: Evidence from big data
## Abstract
### Background
Concerns about the role of chronically used medications in the clinical outcomes of the coronavirus disease 2019 (COVID-19) have remarkable potential for the breakdown of non-communicable diseases (NCDs) management by imposing ambivalence toward medication continuation. This study aimed to investigate the association of single or combinations of chronically used medications in NCDs with clinical outcomes of COVID-19.
### Methods
This retrospective study was conducted on the intersection of two databases, the Iranian COVID-19 registry and Iran Health Insurance Organization. The primary outcome was death due to COVID-19 hospitalization, and secondary outcomes included length of hospital stay, Intensive Care Unit (ICU) admission, and ventilation therapy. The Anatomical Therapeutic Chemical (ATC) classification system was used for medication grouping. The frequent pattern growth algorithm was utilized to investigate the effect of medication combinations on COVID-19 outcomes.
### Findings
Aspirin with chronic use in $10.8\%$ of hospitalized COVID-19 patients was the most frequently used medication, followed by Atorvastatin ($9.2\%$) and Losartan ($8.0\%$). Adrenergics in combination with corticosteroids inhalants (ACIs) with an odds ratio (OR) of 0.79 ($95\%$ confidence interval: 0.68–0.92) were the most associated medications with less chance of ventilation therapy. Oxicams had the least OR of 0.80 (0.73–0.87) for COVID-19 death, followed by ACIs [0.85 (0.77–0.95)] and Biguanides [0.86 (0.82–0.91)].
### Conclusion
The chronic use of most frequently used medications for NCDs management was not associated with poor COVID-19 outcomes. Thus, when indicated, physicians need to discourage patients with NCDs from discontinuing their medications for fear of possible adverse effects on COVID-19 prognosis.
## 1. Introduction
Non-communicable diseases (NCDs), accounting for about three-quarters of global deaths and one-third of disability-adjusted life years (DALYs) in 2019, place the most significant burden of disease on public health [1]. However, due to their unknown nature and speed of spread, emerging infectious diseases can pose tremendous challenges to the health system. The coronavirus disease 2019 (COVID-19), the latest and most widespread pandemic in the twenty-first century, has resulted in more than 600 million cases and nearly 6.4 million deaths until September 2022 [2].
Several studies have shown the interaction between NCDs and COVID-19 (3–5). One of the most controversial topics with contradictory results is the effect of chronic use of medications on the clinical outcomes of COVID-19. For instance, some research suggested that angiotensin-converting enzyme (ACE) inhibitors have been associated with an increased risk of death from COVID-19, while others indicated an association between desirable clinical outcomes and the chronic use of ACE inhibitors [6, 7].
According to a World Health Organization (WHO) survey of 155 countries, $53\%$, $49\%$, and $45\%$ of the countries had partially disrupted services for hypertension, diabetes, and cancer treatment, respectively, during the COVID-19 pandemic [8]. Causes such as quarantine, self-isolation, and unprecedented load on the health systems, have been shown to have a detrimental effect on the quality of NCDs management, especially in the early days of the COVID-19 pandemic (9–12). However, uncertainty about the role of chronic medication use in the clinical outcome of COVID-19 has an abiding potential for the breakdown of NCDs management by imposing ambivalence toward medication continuation.
Despite the interest in the effect of individual medications with chronic use on COVID-19 clinical outcomes, the role of medication combinations has not been explored in depth. This study aimed to investigate whether single or combinations of widely used medications in treating chronic diseases was associated with clinical outcomes of COVID-19.
## 2.1. Design and data acquisition
This retrospective cohort study was conducted on hospitalized COVID-19 patients in Iran. Data of hospitalized COVID-19 patients were obtained from the Iranian COVID-19 registry, which included about 1.2 million patients from February 1st, 2020, to June 8th, 2021. The diagnosis of COVID-19 was made based on the results of Reverse Transcription Polymerase Chain Reactions (RT-PCR) for SARS-CoV-2, or lung Computed Tomography (CT) scan. All patients received supportive and antiviral treatment according to the national guidelines. Extracted variables from the Iranian COVID-19 registry consisted of age, sex, admission date, chronic comorbidities, length of hospital stay, Intensive Care Unit (ICU) admission, ventilation therapy, and death. Cardiovascular diseases (CVDs), chronic respiratory diseases (CRDs), diabetes mellitus (DM), and malignancies were inspected as chronic comorbidities. Claimed prescriptions data were retrieved from Iran Health Insurance Organization (IHIO), including medication names and quantities of prescriptions for the years 2018 and 2019. To investigate the burden on both COVID-19 patients and the healthcare system, outcome measures with the most prominent health and economic impacts were included in this study. The primary outcome of this study was death, and secondary outcomes included length of hospital stay, ICU admission, and ventilation therapy.
## 2.2. Data preprocessing
The national identification number was used for the record linkage of the two databases. COVID-19 patients under treatment were excluded from the study due to an unknown outcome. Admission dates were divided into 3-month intervals to be included in the statistical models due to the different virulence of SARS-CoV-2 variants and vital equipment shortage during the pandemic surges. The Anatomical Therapeutic Chemical (ATC) classification system was applied to medication names for obtaining standard coding. Among five levels of granularity in the ATC classification system, level four was used for medication grouping (ATC groups) in this study. The level five ATC codes claimed at least four times during the 2018–2019 period were considered chronically used medications for each patient. Chronically used medications claimed by less than one percent of COVID-19 patients were excluded from the investigation (Figure 1).
**Figure 1:** *The flowchart of data preprocessing and merging stages of the Iranian COVID-19 registry and IHIO.*
## 2.3. Association rule mining
To investigate the effect of medication combinations on COVID-19 outcomes, the frequent pattern growth algorithm, an efficient association rule mining algorithm, was utilized for item set extraction [13]. For elimination of uncommon combinations, the support hyperparameter was used, defined as below: Among all combinations, two- and three-medication sets with support of at least 0.01 were included in the study.
## 2.4. Statistical analysis
The accelerated failure time (AFT) model was utilized with the Weibull survival function to assess medications' association with length of hospital stay. The logistic regression model was used to evaluate the association of medications with binary outcomes, including ICU admission, ventilation therapy, and death. The models were adjusted for age, gender, admission date, chronic comorbidities, and receiving ventilation therapy. Moreover, the use of each drug of the set alone was also considered a covariate. The 0.05 alpha level was used for significance inference and confidence interval (CI) computation of coefficients and odds ratio (OR) in all statistical analyses. Apache Spark, version 3.2.0, a unified big data processing engine, was utilized for data preprocessing and association rule mining [14]. Statistical analyses were performed using Python libraries Statsmodels, version 0.13, and Lifelines, version 0.27 [15, 16].
## 2.5. Ethical considerations
This study was conducted according to the guidelines of the Declaration of Helsinki, and approved by Research Ethics Committees of Endocrine and Metabolism Research Institute, Tehran University of Medical Sciences (IR.TUMS.EMRI.REC.1400.046).
## 3.1. Overview
A total of 258,917 patients were identified by joining the IHIO database and the Iranian COVID-19 registry following the exclusion of patients under treatment. Overall, $53.0\%$ of patients were female, and the median age was 63 [interquartile range (IQR): 48–75] years. The last and first quarters of 2020 had the highest and lowest number of COVID-19 patients, respectively (Table 1). CVDs with affecting $30.3\%$ of patients, were the leading chronic comorbidities, followed by DM ($19.7\%$) and CRDs ($5.6\%$). The median length of hospital stay was 5 days (IQR: 3–8) and 7 days (IQR: 3–13) in recovered and deceased patients, respectively. Among the study population, 21,363 ($8.3\%$) were admitted to ICU, 18,564 ($7.2\%$) underwent ventilation therapy, and 41,618 ($16.1\%$) were deceased due to COVID-19.
**Table 1**
| Unnamed: 0 | Patients count (percent) |
| --- | --- |
| Admission date | Admission date |
| 2020Q4 | 71,494 (27.6%) |
| 2020Q3 | 51,456 (19.9%) |
| 2021Q2 | 42,582 (16.4%) |
| 2020Q2 | 39,369 (15.2%) |
| 2021Q1 | 37,440 (14.5%) |
| 2020Q1 | 16,576 (6.4%) |
| Chronic comorbidity | Chronic comorbidity |
| CVDs | 78,446 (30.3%) |
| DM | 50,917 (19.7%) |
| CRDs | 14,628 (5.6%) |
| DM (insulin therapy) | 11,441 (4.4%) |
| Malignancies | 4,924 (1.9%) |
| Outcomes | Outcomes |
| Death | 41,618 (16.1%) |
| ICU admission | 21,363 (8.3%) |
| Ventilation therapy | 18,564 (7.2%) |
## 3.2. Drug utilization
Of the total 607 chronically used medications, 42 ATC groups and 42 level five ATC codes were claimed by at least $1\%$ of the study population. Platelet aggregation inhibitors excluding Heparin (B01AC), plain Angiotensin II receptor blockers (ARBs; C09CA), and HMG CoA reductase inhibitors (C10AA) were the most frequent ATC groups with chronic use in $10.8\%$, $9.4\%$, and $9.4\%$ of patients, respectively (Supplementary Data Sheet 1). Besides, Aspirin ($10.8\%$), Atorvastatin ($9.2\%$), and Losartan ($8.0\%$) were the leading medications in chronic use (Supplementary Data Sheet 2). Aspirin-Atorvastatin ($5.7\%$) and Aspirin-Atorvastatin-Losartan ($2.5\%$) had the highest frequency of chronic use among two- and three-medication sets, respectively (Supplementary Data Sheet 3). There was no significant difference between chronic use of a single medication vs. two or more than two medications in ICU admission, need for ventilation therapy, and death (Figure 2).
**Figure 2:** *Change in length of hospital stay and ORs of ICU admission, ventilation therapy, and death by the number of chronically used medications.*
## 3.3.1. Medication groups
The most associated ATC groups with the prolongation of hospital stay were preparations inhibiting uric acid production (M04AA) and fatty acid derivatives antiepileptics (N03AG) with $11.6\%$ ($95\%$ CI: 7.2–16.1) and $11.3\%$ (7.7–15.0) increase, respectively (Supplementary Table 1). Among ATC groups, chronic use of acetic acid derivatives and related substances (M01AB) with an OR of 0.85 (0.76–0.96) had the least association with ICU admission (Figure 3; Supplementary Table 1). Adrenergics in combination with corticosteroids inhalants (ACIs) had the lowest OR for undergoing ventilation therapy [0.79 (0.68–0.92)], followed by plain Thiazides [C03AA; 0.84 (0.72–0.98)] and Biguanides [A10BA; 0.87 (0.80–0.94)]. Regarding the association of chronic use of ATC groups with death, Oxicams (M01AC) had a significantly low OR of 0.80 (0.73–0.87), followed by ACIs [R03AK; 0.85 (0.77–0.95)] and Biguanides [A10BA; 0.86 (0.82–0.91)].
**Figure 3:** *The ORs of ICU admission vs. death for medication groups, colored by the OR of ventilation therapy.*
## 3.3.2. Single medications
Valproic acid with a $14.4\%$ (10.8–18.1) increase in hospital stay, along with Prednisolone [$14.2\%$ (11.2–17.3)] and Allopurinol [$12.1\%$ (7.6–16.6)] were the most associated medications with the lengthening of hospital stay (Figure 4; Supplementary Table 2). Salmeterol/Fluticasone inhaler with an OR of 0.84 (0.73–0.97) was the leading medication in negative association with ICU admission, followed by Cholecalciferol [0.84 (0.72–0.96)], and Diclofenac [0.87 (0.76–0.98)]. Regarding the need for ventilation therapy, Salmeterol/Fluticasone inhaler [OR: 0.71 (0.61–0.83)], Valsartan [OR: 0.82 (0.72–0.93)], and Hydrochlorothiazide [OR: 0.85 (0.73–0.99)] had the lowest association. Similarly, Salmeterol/Fluticasone inhaler [OR: 0.84 (0.76–0.92)] and Metformin [OR: 0.87 (0.83–0.92)], respectively, were the second and third least associated medications with COVID-19 death. While chronic use of Betamethasone was negatively associated with death [OR: 0.81 (0.72–0.91)] more than any other medication, Dexamethasone [OR: 1.58 (1.44–1.73)] was among the leading medications in positive association with death.
**Figure 4:** *Change in length of hospital stay along with ORs of ICU admission, ventilation therapy, and death by medications.*
## 3.3.3. Two-medication sets
The most associated two-medication sets with the decrease in hospital stay were Amlodipine-Hydrochlorothiazide [$16.4\%$ (9.3–22.8)], Carvedilol-Spironolactone [$15.7\%$ (8.7–22.1)], and Aspirin-Valproic acid [$15.3\%$ (8.6–21.4); Supplementary Table 3]. Hydrochlorothiazide-Metformin with an OR of 0.64 (0.46–0.90) had the least association with ICU admission, followed by Salbutamol-Salmeterol/Fluticasone [0.66 (0.47–0.94)], and Alprazolam-Metformin [0.72 (0.52–0.98)]. In contrast, Folic acid-Losartan and Losartan-Spironolactone were the most related two-medication sets to ICU admission, with ORs of 1.35 (1.05–1.74) and 1.30 (1.01–1.68), respectively. While Glibenclamide-Losartan with an OR of 0.79 (0.63–0.99) was the least associated two-medication set with the need for ventilation therapy, Gliclazide-Losartan [1.63 (1.22–2.17)] was the most associated set. Among all two-medication sets, Folic acid-Prednisolone with a significant OR of 0.70 (0.59–0.84) had the least association with death, followed by Carvedilol-Spironolactone [0.73 (0.61–0.88)].
## 3.3.4. Three-medication sets
Aspirin-Carvedilol-Spironolactone with a $17.8\%$ (10.7–24.4) decrease in length of hospital stay was the most negatively associated three-medication set, followed by Aspirin-Glyceryl Trinitrate-Spironolactone [$16.8\%$ (9.9–23.2); Supplementary Table 4]. No three-medication set was significantly associated with reducing ICU admission or the need for ventilation therapy. The least associated three-medication sets with death were Aspirin-Furosemide-Spironolactone and Amlodipine-Atorvastatin-Metformin, with ORs of 0.72 (0.61–0.85) and 0.73 (0.63–0.84), respectively.
## 4. Discussion
Aspirin, Atorvastatin, and Losartan were the leading chronically used medications among patients hospitalized due to COVID-19. ACIs were the most associated medications with less chance of ventilation therapy, followed by plain Thiazides and Biguanides. Oxicams had the most inverse association with death, followed by ACIs and Biguanides.
Hypertension is associated with severe COVID-19 [17, 18]. There has been a growing body of literature investigating the potential role of various antihypertensive agents in COVID-19 prognosis. The results of this study suggested that Amlodipine-Hydrochlorothiazide was associated with a $16\%$ decrease in hospital stay. Among ARBs, Valsartan was associated with $18\%$ less likelihood of ventilation therapy. Moreover, the chronic use of plain Thiazides was associated with a $15\%$ lower risk of ventilation therapy during COVID-19 hospitalization. There is evidence that ACE inhibitors/ARBs use was associated with a significantly decreased risk of COVID-19 mortality [19]. Moreover, they were not associated with significantly increased chances of receiving ICU care [20]. They may even have superior beneficial effects on treating hypertension during the COVID-19 pandemic [21]. Thus, continuing ACE inhibitors/ARBs during the COVID-19 pandemic can be a suitable strategy [22], as current guidance suggests [23].
The chronic use of Biguanides was associated with about $13\%$ lower mortality due to COVID-19. Similarly, a meta-analysis reported that Metformin was associated with $34\%$ (22–44) lower COVID-19 mortality [24]. While the chronic use of Metformin was associated with a $13\%$ lower probability of receiving ventilation therapy, a meta-analysis did not identify a statistically significant association between Metformin and intubation [24]. The underlying cause of the association between chronic Metformin use and death due to COVID-19 is yet to be understood. Nevertheless, there is evidence that Metformin could alter the expression and stability of ACE-2, a target of SARS-CoV-2 [25, 26], while also reducing tumor necrosis factor alpha (TNF-α) [27, 28], a proinflammatory cytokine involved in the prognosis of COVID-19 [29].
Due to evidence suggestive of possible links between non-steroidal anti-inflammatory drugs (NSAIDs) and respiratory or cardiovascular adverse effects in several settings, regular NSAIDs use was not recommended as the first-line option for managing the symptoms of COVID-19 [30]. However, further evidence suggested that NSAIDs use was not associated with poor COVID-19 outcomes [31]. In this study, chronic NSAIDs use was associated with a $14\%$ lower ICU admission due to COVID-19. Notably, chronic Diclofenac users had a $13\%$ lower chance of ICU admission due to COVID-19. Furthermore, Piroxicam was associated with a $14\%$ less risk of ventilation therapy and a $12\%$ less mortality due to COVID-19. Currently, there seems to be no evidence suggesting that clinicians should refrain from or discontinue NSAIDs in patients with COVID-19 upon proper indication [32].
Reports on the impact of inhaled corticosteroids (ICS) on COVID-19 clinical outcomes have been discrepant [33]. In this study, Salmeterol/Fluticasone inhaler was associated with $16\%$, $29\%$, and $16\%$ reductions in ICU admission, ventilation therapy, and death due to COVID-19, respectively. A study among more than 800,000 patients with asthma reported a non-significant increase in COVID-19-related mortality associated with ICS use [34]. While the potential beneficial effects of chronic ICS use, alone or in combination with adrenergics, are yet to be confidently established, there seems to be sufficient evidence for the continuation of these medications when indicated [35].
While the administration of Dexamethasone in hospitalized patients with COVID-19 resulted in lower all-cause mortality [36, 37], the chronic use of Dexamethasone in this study was associated with a $58\%$ higher mortality due to COVID-19. Furthermore, chronic Prednisolone users tended to stay in hospital $14\%$ longer. Similarly, a study reported that the use of systemic glucocorticoids prior to COVID-19 hospitalization was associated with higher mortality [38]. Thus, the administration of systemic glucocorticoids needs to be with extra caution during the COVID-19 pandemic. In the meantime, patients who need to take such medications for extended periods must beware of their short-, mid-, and long-term adverse effects, especially during the pandemic. In contrast to Dexamethasone, the chronic use of Betamethasone was associated with a $19\%$ lower mortality due to COVID-19, which needs to be validated and investigated in future studies.
The NCDs pandemic has existed prior to the COVID-19 era, and CVDs, cancers, CRDs, and DM are the four groups of diseases responsible for more than $80\%$ of all NCDs mortality [39, 40]. Similarly, in this study, $30\%$ of participants had CVDs, nearly $20\%$ had DM, and some $6\%$ had CRDs. NCDs have resulted in a double burden due to the COVID-19 pandemic, as patients with HTN, DM, CRDs, and CVDs have been reportedly more likely to have poor COVID-19 outcomes [41, 42]. COVID-19 pandemic has inevitably resulted in rapid exploitation of healthcare system resources and leaving health systems tremendously overstretched [43]. The significant disruptions to conventional NCDs care could lead to increased morbidity and mortality in the long term [44, 45]. In the meantime, poor adherence to prescribed medications for NCDs, also fueled by fear of potential adverse effects on COVID-19 infection, could complicate the management of NCDs. In this sense, the present study provides evidence for continuing the most used medications for NCDs when indicated.
## 4.1. Strengths and limitations
This study is the first nationwide study to investigate whether single or combinations of widely used medications in treating chronic diseases affect clinical outcomes of COVID-19, based on the real-world data of more than 250,000 hospitalized patients. The strength of this study lies in a large sample and data gathering since the early days of the outbreak in Iran. Moreover, chronic use of medications was investigated based on the patients' prescriptions from 2 years before the admission date due to COVID-19. Despite the interest in the effect of individual medications with chronic use on COVID-19 clinical outcomes, the role of medication combinations has not been explored in depth.
We also acknowledge the limitations of this study. The focus of this study was the chronic use of medications; hence, the findings have limited implications in the context of the acute use of medications. Moreover, the results of this study do not reflect the potential biochemical pathways through which chronic use of medications may interact with SARS-CoV-2; thus, further in-vitro/in-vivo studies are required in this sense. Another noteworthy limitation of this study was neglecting the role of disease control due to the lack of an integrated electronic healthcare system. Similarly, the role of CVDs, DM, and CRDs could be adjusted while performing the statistical analysis. The effects of other comorbidities could not be investigated due to the lack of data. Thus, the results of this study need to be interpreted with caution.
## 4.2. Conclusions
The chronic use of most frequently used medications for NCDs management was not associated with poor COVID-19 outcomes. Thus, based on the available evidence and when indicated, physicians need to discourage patients with NCDs, particularly those with CVDs, DM, or CRDs, from discontinuing their medications for fear of possible adverse effects on COVID-19 prognosis. In the meantime, future studies need to investigate the clinical/biochemical role of acute/chronic medication use in the context of COVID-19.
## Data availability statement
The data used in this study is owned by IHIO and Iranian COVID-19 registry. Therefore, authors are not allowed to share the data publicly or privately. However, any researcher with written permission from IHIO and Iranian COVID-19 registry can request to obtain the anonymized data. Requests to access these datasets should be directed to IHIO.gov.ir.
## Ethics statement
This study was conducted according to the guidelines of the Declaration of Helsinki and approved by Research Ethics Committees of Endocrine and Metabolism Research Institute, Tehran University of Medical Sciences (IR.TUMS.EMRI.REC.1400.046).
## Author contributions
Conceptualization: FF and M-RM. Methodology, formal analysis, and visualization: M-RM. Software: SS and M-RM. Validation: MA-K, S-HG, and SS. Investigation and data curation: M-RM, MA-K, S-HG, and M-MR. Resources: AS, AN, ME, MN, and MR. Writing—original draft preparation: M-RM, MA-K, and S-HG. Writing—review and editing: FF, SS, M-RM, and MA-K. Supervision: FF. Project administration: AS. All authors contributed to the article and approved the submitted version.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher's note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpubh.2023.1061307/full#supplementary-material
## References
1. **Global burden of 369 diseases and injuries in 204 countries and territories, 1990–2019: a systematic analysis for the Global Burden of Disease Study 2019**. *Lancet.* (2020.0) **396** 1204-1222. DOI: 10.1016/S0140-6736(20)30925-9
2. Dong E, Du H, Gardner L. **An interactive web-based dashboard to track COVID-19 in real time**. *Lancet Infect Dis.* (2020.0) **20** 533-4. DOI: 10.1016/S1473-3099(20)30120-1
3. de Almeida-Pititto B, Dualib PM, Zajdenverg L, Dantas JR, de Souza FD, Rodacki M. **Severity and mortality of COVID 19 in patients with diabetes, hypertension and cardiovascular disease: a meta-analysis**. *Diabetol Metab Synd* (2020.0) **12** 1-12. DOI: 10.1186/s13098-020-00586-4
4. Rezaei N, Montazeri F, Malekpour M-R, Ghanbari A, Azadnajafabad S, Mohammadi E. **COVID-19 in patients with diabetes: factors associated with worse outcomes**. *J Diabetes Metab Disord.* (2021.0) **20** 1605-14. DOI: 10.1007/s40200-021-00910-3
5. Kumar A, Arora A, Sharma P, Anikhindi SA, Bansal N, Singla V. **Is diabetes mellitus associated with mortality and severity of COVID-19? A meta-analysis**. *Metab Syndrome.* (2020.0) **15** 535-45. DOI: 10.1016/j.dsx.2020.04.044
6. Selçuk M, Çinar T, Keskin M, Çiçek V, Kiliç S, Kenan B. **Is the use of ACE inb/ARBs associated with higher in-hospital mortality in COVID-19 pneumonia patients?**. *Clin Exp Hypertens.* (2020.0) **42** 738-42. DOI: 10.1080/10641963.2020.1783549
7. Zhang P, Zhu L, Cai J, Lei F, Qin J-J, Xie J. **Association of inpatient use of angiotensin-converting enzyme inhibitors and angiotensin II receptor blockers with mortality among patients with hypertension hospitalized with COVID-19**. *Circ Res.* (2020.0) **126** 1671-81. DOI: 10.1161/CIRCRESAHA.120.317134
8. 8.Rapid Assessment of Service Delivery for NCDs During the COVID-19 Pandemic. https://www.who.int/publications/m/item/rapid-assessment-of-service-delivery-for-ncds-during-the-covid-19-pandemic (accessed May 2, 2022).. *Rapid Assessment of Service Delivery for NCDs During the COVID-19 Pandemic*
9. Akrami F, Peykari N, Payab M, Moghisi A, Yousefi E, Ranjbar M. **The effect of COVID-19 pandemic on non-communicable disease prevention and management services in the primary health care system of Iran**. *SJKU.* (2021.0) **26** 33-49. DOI: 10.52547/sjku.26.5.33
10. Dyer O. **COVID-19: pandemic is having “severe” impact on non-communicable disease care, WHO survey finds**. *BMJ.* (2020.0) **369** m2210. DOI: 10.1136/bmj.m2210
11. Yadav UN, Rayamajhee B, Mistry SK, Parsekar SS, Mishra SK. **A syndemic perspective on the management of non-communicable diseases amid the COVID-19 pandemic in low- and middle-income countries**. *Front. Public Health.* (2020.0) **8** e508. DOI: 10.3389/fpubh.2020.00508
12. Basu S. **Non-communicable disease management in vulnerable patients during COVID-19**. *Indian J Med Ethics.* (2020.0) 103-105. DOI: 10.20529/IJME.2020.041
13. Han J, Pei J, Yin Y. **Mining frequent patterns without candidate generation**. *SIGMOD Rec.* (2000.0) **29** 1-12. DOI: 10.1145/335191.335372
14. Zaharia M, Xin RS, Wendell P, Das T, Armbrust M, Dave A. **Apache Spark: a unified engine for big data processing**. *Commun ACM.* (2016.0) **59** 56-65. DOI: 10.1145/2934664
15. Seabold S, Perktold J. **Statsmodels: econometric and statistical modeling with python**. *Proceedings of the 9th Python in Science Conference* (2010.0) 61
16. Davidson-Pilon C. **Lifelines: survival analysis in Python**. *JOSS.* (2019.0) **4** 1317. DOI: 10.21105/joss.01317
17. Fang L, Karakiulakis G, Roth M. **Are patients with hypertension and diabetes mellitus at increased risk for COVID-19 infection?**. *Lancet Respir Med.* (2020.0) **8** e21. DOI: 10.1016/S2213-2600(20)30116-8
18. Qian Z, Li Z, Peng J, Gao Q, Cai S, Xu X. **Association between hypertension and prognosis of patients with COVID-19: a systematic review and meta-analysis**. *Clin Exp Hypertens.* (2022.0) **44** 451-8. DOI: 10.1080/10641963.2022.2071914
19. Lee HW, Yoon C-H, Jang EJ, Lee C-H. **Renin-angiotensin system blocker and outcomes of COVID-19: a systematic review and meta-analysis**. *Thorax.* (2021.0) **76** 479-86. DOI: 10.1136/thoraxjnl-2020-215322
20. Hippisley-Cox J, Young D, Coupland C, Channon KM, Tan PS, Harrison DA. **Risk of severe COVID-19 disease with ACE inhibitors and angiotensin receptor blockers: cohort study including 83 million people**. *Heart.* (2020.0) **106** 1503-11. DOI: 10.1136/heartjnl-2020-317393
21. Nozari F, Hamidizadeh N. **The effects of different classes of antihypertensive drugs on patients with COVID-19 and hypertension: a mini-review**. *Int J Hypertens.* (2022.0) **2022** 5937802. DOI: 10.1155/2022/5937802
22. Azad GN, Kumar A. **ACEi/ARB and deaths of COVID-19 patients**. *Curr Hypertens Rev.* (2022.0) **18** 158-62. DOI: 10.2174/1573402118666220407093332
23. JAMA Network. **Coronavirus Disease 2019 (COVID-19) infection and renin angiotensin system blockers | clinical pharmacy and pharmacology**. *JAMA Cardiol*
24. Li Y, Yang X, Yan P, Sun T, Zeng Z, Li S. **Metformin in patients with COVID-19: a systematic review and meta-analysis**. *Front Med.* (2021.0) **8** 704666. DOI: 10.3389/fmed.2021.704666
25. Zhang J, Dong J, Martin M, He M, Gongol B, Marin TL. **AMP-activated protein kinase phosphorylation of angiotensin-converting enzyme 2 in endothelium mitigates pulmonary hypertension**. *Am J Respir Crit Care Med.* (2018.0) **198** 509-20. DOI: 10.1164/rccm.201712-2570OC
26. Ursini F, Ciaffi J, Landini MP, Meliconi R. **COVID-19 and diabetes: is metformin a friend or foe?**. *Diabetes Res Clin Pract.* (2020.0) **164** 108167. DOI: 10.1016/j.diabres.2020.108167
27. Park JW, Lee JH, Park YH, Park SJ, Cheon JH, Kim WH. **Sex-dependent difference in the effect of metformin on colorectal cancer-specific mortality of diabetic colorectal cancer patients**. *World J Gastroenterol.* (2017.0) **23** 5196-205. DOI: 10.3748/wjg.v23.i28.5196
28. Matsiukevich D, Piraino G, Lahni P, Hake PW, Wolfe V, O'Connor M. **Metformin ameliorates gender-and age-dependent hemodynamic instability and myocardial injury in murine hemorrhagic shock**. *Biochim Biophys Acta Mol Basis Dis.* (2017.0) **1863** 2680-91. DOI: 10.1016/j.bbadis.2017.05.027
29. Costela-Ruiz VJ, Illescas-Montes R, Puerta-Puerta JM, Ruiz C, Melguizo-Rodríguez L. **SARS-CoV-2 infection: the role of cytokines in COVID-19 disease**. *Cytokine Growth Factor Rev.* (2020.0) **54** 62-75. DOI: 10.1016/j.cytogfr.2020.06.001
30. Little P. **Non-steroidal anti-inflammatory drugs and COVID-19**. *BMJ.* (2020.0) **368** m1185. DOI: 10.1136/bmj.m1185
31. Drake TM, Fairfield CJ, Pius R, Knight SR, Norman L, Girvan M. **Non-steroidal anti-inflammatory drug use and outcomes of COVID-19 in the ISARIC Clinical Characterisation Protocol UK cohort: a matched, prospective cohort study**. *Lancet Rheumatol.* (2021.0) **3** e498-506. DOI: 10.1016/S2665-9913(21)00104-1
32. Zhao H, Huang S, Huang S, Liu F, Shao W, Mei K. **Prevalence of NSAID use among people with COVID-19 and the association with COVID-19-related outcomes: systematic review and meta-analysis**. *Br J Clin Pharmacol.* (2022.0) **88** 5113-27. DOI: 10.1111/bcp.15512
33. Luque-Paz D, Tattevin P, Loubet P, Bénézit F, Thibault V, Lainé F. **Chronic use of inhaled corticosteroids in patients admitted for respiratory virus infections: a 6-year prospective multicenter study**. *Sci Rep.* (2022.0) **12** 4199. DOI: 10.1038/s41598-022-08089-0
34. Schultze A, Walker AJ, MacKenna B, Morton CE, Bhaskaran K, Brown JP. **Inhaled corticosteroid use and risk COVID-19 related death among 966,461 patients with COPD or asthma: an OpenSAFELY analysis**. *medRxiv.* (2020.0)
35. Kow CS, Hasan SS. **Preadmission use of inhaled corticosteroids and risk of fatal or severe COVID-19: a meta-analysis**. *J Asthma.* (2022.0) **59** 787-90. DOI: 10.1080/02770903.2021.1878531
36. **Dexamethasone in hospitalized patients with COVID-19**. *New Engl J Med.* (2021.0) **384** 693-704. DOI: 10.1056/NEJMoa2021436
37. Zeng Y, Zeng W, Yang B, Liu Z. **Effectiveness of corticosteroids to treat coronavirus disease 2019 symptoms: a meta-analysis**. *Med Clin (Barc).* (2022.0) **159** 575-83. DOI: 10.1016/j.medcli.2022.03.013
38. Calderón-Parra J, Cuervas-Mons V, Moreno-Torres V, Rubio-Rivas M, Blas PA, Pinilla-Llorente B. **Influence of chronic use of corticosteroids and calcineurin inhibitors on COVID-19 clinical outcomes: analysis of a nationwide registry**. *Int J Infect Dis.* (2022.0) **116** 51-8. DOI: 10.1016/j.ijid.2021.12.327
39. 39.Non-communicable Diseases. Available online at: https://www.who.int/news-room/fact-sheets/detail/noncommunicable-diseases (accessed April 21, 2022).. *Non-communicable Diseases*
40. 40.Non-communicable Disease Facility-Based Monitoring Guidance. Available online at: https://www.who.int/publications-detail-redirect/9789240057067 (accessed December 29, 2022).. *Non-communicable Disease Facility-Based Monitoring Guidance*
41. 41.BMC Public Health. COVID-19 and Non-Communicable Diseases: Evidence From a Systematic Literature Review. Available online at: https://bmcpublichealth.biomedcentral.com/articles/10.1186/s12889-021-11116-w (accessed June 7, 2022). *COVID-19 and Non-Communicable Diseases: Evidence From a Systematic Literature Review*
42. Azadnajafabad S, Ghasemi E, Saeedi Moghaddam S, Rezaei N, Farzadfar F. **Non-communicable diseases' contribution to the COVID-19 mortality: a global warning on the emerging syndemics**. *Arch Iran Med.* (2021.0) **24** 445-6. DOI: 10.34172/aim.2021.63
43. **NCD Countdown 2030: efficient pathways and strategic investments to accelerate progress towards the Sustainable Development Goal target 3.4 in low-income and middle-income countries**. *Lancet.* (2022.0) **399** 1266-78. DOI: 10.1016/S0140-6736(21)02347-3
44. Ahy CRH. **Preventing a post-pandemic double burden of disease in the COVID-19 pandemic**. *Glob Adv Health Med.* (2021.0) **10** 21649561211010137. DOI: 10.1177/21649561211010137
45. GBD 2019 Iran Collaborators. **Health system performance in Iran: a systematic analysis for the Global Burden of Disease Study 2019**. *Lancet.* (2022.0) **399** 1625-45. DOI: 10.1016/S0140-6736(21)02751-3
|
---
title: 'Changes in the health literacy of residents aged 15–69 years in central China:
A three-round cross-sectional study'
authors:
- Xin Mei
- Gong Chen
- Yuting Zuo
- Qian Wu
- Junlin Li
- Yilin Li
journal: Frontiers in Public Health
year: 2023
pmcid: PMC9998942
doi: 10.3389/fpubh.2023.1092892
license: CC BY 4.0
---
# Changes in the health literacy of residents aged 15–69 years in central China: A three-round cross-sectional study
## Abstract
### Introduction
Health literacy is an effective strategy to promote more cost-effective use of health care services and a crucial tool for preventing the spread of infectious diseases. The main objective of this study was to analyze changes in health literacy from 2019 to 2021.
### Methods
Data were from the latest three-round cross-sectional studies with the same study design.
### Results
Although the prevalence of adequate health literacy rose significantly over time, increasing from $26.9\%$ ($95\%$ CI 20.1–33.7) in 2019 to $34.1\%$ (32.9–35.3) in 2021, it was still at a relatively low level. The most significant decrease was observed for health skills among the three aspects and health information literacy among the six dimensions. Working as medical staff was a protective factor for adequate health literacy, health skills literacy, and health information literacy. Risk factors for adequate health literacy and health information literacy were older age and lower education level. Furthermore, adequate health information literacy was positively related to annual family income.
### Discussion
More practical and effective policies targeting health literacy for critical aspects and groups in Central China, are urgently needed, especially during the epidemic.
## Introduction
As defined by the U.S. National Library of Medicine and the World Health Organization (WHO), health literacy refers to the ability to obtain, understand, and process basic health information and services and use them to make sound health-related decisions to maintain and promote health [1, 2]. Health literacy is an essential factor affecting health and a strong predictor of the population's health status [3]. Studies have shown that limited health literacy is not only related to adverse health behaviors such as smoking, alcoholism, low physical activity, difficulty communicating with doctors, and poor adherence to medicines prescribed by doctors but also closely related to adverse health outcomes such as hypertension, diabetes, stroke, and high mortality (4–8). Limited health literacy will also lead to increased medical expenses and waste of healthcare services [9]. Thus, improving health literacy can be an effective strategy to promote more cost-effective use of healthcare services, contributing to the ultimate goal of primary healthcare and improving the population's health (10–14).
Therefore, the international emphasis on health literacy is increasing (15–17). Health literacy has become a research hotspot in clinical medicine, health education, and health promotion [18, 19]. Research on health literacy is mainly based on two perspectives: the clinical perspective [20] and the public health perspective [21].
Research on health literacy in North American and European countries mainly focuses on the clinical perspective [20, 21]. With the development of the health literacy evaluation system, many countries have successively launched health literacy surveys. The setting for these studies has been expanded from patients to the general public, and the measurement content has expanded from the clinical context to disease prevention, healthcare, and health promotion [22].
From the perspective of public health, the impact of health literacy on disease prevention, healthy lifestyle and behavior, and maintenance and promotion of health was studied in China [23]. The National Health Commission of China released the educational book “Health Literacy of Chinese Citizens-Basic Knowledge and Skills (Trial)” and organized the first national health literacy survey in 2008 [24]. In 2010, the China Health Education Center studied the evaluation system of health literacy, with the educational book as the evaluation content, and compiled the Chinese Health Literacy Survey Questionnaire (CHLSQ) [25]. Since 2012, China has carried out scientific and continuous health literacy monitoring. Focusing on basic knowledge and concepts, healthy lifestyle and behavior, and health skills, a health literacy monitoring system for permanent residents aged 15–69 has gradually been established in China [23].
Although the definitions, measurement instruments, evaluation perspectives, and survey methods of health literacy are different in different countries or regions, many surveys have come to the same conclusion: Globally, health literacy needs to be improved [26, 27]. The National Assessment of Adult Literacy (NAAL) found that $88\%$ of adults do not have sufficient health literacy in the USA [26]. A systematic review indicated that the prevalence of low health literacy ranged from 27 to $48\%$ in Europe [27]. The Chinese national health literacy survey showed that the prevalence of adequate health literacy among residents aged 15–69 was only $19\%$ [28].
On 31 January 2020, the WHO declared the COVID-19 outbreak a Public Health Emergency of International Concern [17]. The pandemic posed a considerable threat to human health [29, 30]. Health literacy is a crucial determinant of health at both the social and individual levels, in healthy populations and with diverse infectious diseases [31], which is also a crucial tool for preventing the spread of infectious diseases [32].
Previous studies have highlighted the significance of health literacy for the outcomes of infectious diseases and the role that health literacy plays regarding infectious diseases [33, 34]. People with low health literacy may not obtain adequate health knowledge on time and cannot implement protective behaviors, such as the adoption of immunization, to prevent infectious diseases [33]. Therefore, it is significant to study the changes in the health literacy level and its determinants during this time. However, we found few studies describing the changes in health literacy during the pandemic.
Wuhan is located in central China, with a permanent population of 13.6 million [35]. The main objective of this study was to analyze, based on three-wave city-level representative data among 15- to 69-year-old permanent residents in Wuhan, China, levels and changes in health literacy from 2019 to 2021 and the relationship between health literacy and related factors.
## Study population
The China Health Literacy Survey (CHLS) is a nationally representative household survey of the permanent population aged 15–69 [36]. In conjunction with the CHLS, the Wuhan Health Literacy Survey (WHLS) aimed to provide data necessary to estimate health literacy since 2016 at the 1-year interval. The WHLS is a cross-sectional survey using the CHLS standardized protocol and questionnaire. We based our study on the latest three rounds (conducted from August to November 2019, 2020, and 2021) of the WHLS. The processes and sampling design of the survey were reviewed and approved by the Institutional Review Board (IRB) of Wuhan CDC (WHCDCIRB-K-2019016). All study participants provided electronic informed consent. All collected data were anonymous and self-administered.
## Sampling method
The sample size was calculated by the formula N=μα2×p(1-p)δ2×deff, where α was the significance level, μα was the α-quantile of the standard normal distribution, p was the health literacy level, δ was the maximum permissible error, and deff was the design effect of complex sampling. Considering the rate of invalid questionnaires and rejections, the final sample size is expected to be calculated. The sampling procedure involved five stages to ensure the representativeness of the selected study population. First, the simple random sampling (SRS) method was used to select several administrative districts (six in 2019 and 2020 and five in 2021) from the 15 districts in Wuhan. Second, the SRS method was used in each administrative district to select several streets (four in 2019 and 2021 three in 2020). Third, the SRS method was used in each street to select several neighborhood committees (three in 2019 and 2021 and two in 2020). Fourth, the SRS method was used in each neighborhood committee to select several households (55 in 2019, 85 in 2020, and 80 in 2021). Fifth, one resident was selected from each household using the KISH method, and a certain number of questionnaires were completed in each neighborhood committee (40 in 2019, 70 in 2020, and 52 in 2021).
## Measurement instrument
The CHLSQ, as compiled by the China Health Education Center [36], was used to measure health literacy. The questionnaire has strong internal consistency and split-half reliability [23], which consists of two parts: sociodemographic characteristics and health literacy content (a total of 50 items). The 50 items include eight true-or-false questions, 23 single-choice questions, 15 multiple-choice questions, and four situational questions (including three single and one multiple-choice questions). The 50-item health literacy is further categorized into three aspects and six dimensions. Based on the knowledge, attitude, practice (KAP) theory, the three aspects of literacy are basic knowledge and concept literacy, healthy lifestyles and behavior literacy, and health skill literacy [25]. Guided by public health problems, the six dimensions of literacy are scientific views of health, infectious disease literacy, chronic disease literacy, safety and first aid literacy, medical care literacy, and health information literacy [24].
The total score of 50 items ranged from 0 to 66 points, with one point for every true-or-false and every single-choice question and two points for every multiple-choice question. Moreover, every wrong or missing choice received 0 points. The total scores of the three aspects were 28 (basic knowledge and concepts literacy, 22 items), 22 (healthy lifestyles and behavior literacy, 16 items), and 16 (health skill literacy, 12 items) points. The maximum total scores for the six dimensions of literacy were 11 points (scientific views of health, eight items), seven points (infectious disease literacy, six items), 12 points (chronic disease literacy, nine items), 14 points (safety and first aid literacy, ten items), 14 points (medical care literacy, 11 items), and eight points (health information literacy, six items).
Adequate health literacy is defined as when participants achieve more than $80\%$ of the total score (53–66 points), and limited health literacy is defined as when participants score <$80\%$ of the total score (0–52 points) [24, 25]. The judgment criterion for adequate health literacy in each aspect or dimension was ≥$80\%$ of the total score for the aspect or dimension. Health literacy level was defined as the proportion of participants who had adequate health literacy out of the total number of participants, as was the health literacy level of the three aspects and six dimensions [37].
## Survey method
Before the fieldwork, the neighborhood committee issued an investigation announcement about the purpose of the study to encourage residents to participate. In the investigation, face-to-face interviews were conducted at each participant's home or in other public places at their convenience. A portable tablet was used to complete electronic questionnaires. If participants could not complete the questionnaire, the investigators would neutrally interview them as an alternative to completing the questionnaire on behalf of the participants. In addition, participants were sent small gifts as an incentive for participating. If the individuals were already participants, they could withdraw at any time without penalty or loss of benefits. Strict quality control was applied to the whole investigative process. Two training sessions were held, and all staff participated and passed the on-site exams. The investigator complied with the investigation guidelines during all processes.
## Statistical analysis
We used the following independent variables drawn from the literature in our analysis: [21, 24, 25, 38] sociodemographic characteristics (i.e., gender, age, marital status, education level, occupation, and annual family income) and self-reported health status (Supplementary Table 1).
Data cleansing rules were created to ensure accuracy and eliminate internal inconsistencies. The sampling weight was considered since the survey adopted a multi-stage sampling procedure. The three waves of data were weighted: calculation of the sampling weight, non-response adjustment, and poststratification calibration adjustment of the sample totals to the known population totals. All of the analyses were based on a complex survey design. Rao-Scott chi-square tests were used to compare the differences in health literacy among subgroups in bivariate analyses. Cochran–Armitage trend tests were used to measure trends in health literacy over time. Multivariable logistic regression analysis was conducted to identify the risk factors related to adequate health literacy. A two-sided $5\%$ significance level assessed statistical inferences. Data cleaning, weighting, and analysis were conducted using SAS software version 9.4 (SAS Institute Inc. Cary, NC).
## Participant characteristics
Table 1 shows descriptive statistics of the study population over time. A total of 2,880 individuals in 2019, 2,520 individuals in 2020, and 3,120 individuals in 2021 were invited to participate in the survey, with effective response rates of $94.7\%$ in 2019 (2,544 individuals), $95.3\%$ in 2020 (2,295 individuals), and $99.0\%$ in 2021 (3,088 individuals).
**Table 1**
| Characteristic | Survey year 2019 (n = 2,544) | Survey year 2019 (n = 2,544).1 | Survey year 2020 (n = 2,295) | Survey year 2020 (n = 2,295).1 | Survey year 2021 (n = 3,088) | Survey year 2021 (n = 3,088).1 | Pc |
| --- | --- | --- | --- | --- | --- | --- | --- |
| Characteristic | Pc | | | | | | |
| | n a | % (95%CI)b | n a | % (95%CI)b | n a | % (95%CI)b | |
| Gender | Gender | Gender | Gender | Gender | Gender | Gender | Gender |
| Male | 1238 | 51.3 (45.5–57.1) | 1110 | 51.3 (44.3–58.3) | 1492 | 51.3 (47.2–55.4) | |
| Female | 1306 | 48.7 (42.9–54.5) | 1185 | 48.7 (41.7–55.7) | 1596 | 48.7 (44.6–52.8) | |
| Age, years | Age, years | Age, years | Age, years | Age, years | Age, years | Age, years | Age, years |
| 15–29 | 281 | 38.0 (32.7–43.4) | 202 | 38.0 (28.8–47.2) | 352 | 38.0 (30.6–45.4) | |
| 30–44 | 830 | 28.2 (21.9–34.5) | 629 | 28.2 (21.6–34.8) | 889 | 28.2 (24.8–31.6) | |
| 45–59 | 841 | 25.1 (19.2–31.1) | 794 | 25.1 (16.8–33.4) | 1082 | 25.1 (20.8–29.5) | |
| 60–69 | 592 | 8.7 (5.4–11.9) | 670 | 8.7 (4.5–12.8) | 765 | 8.7 (4.8–12.6) | |
| Marital status | | | | | | | <0.001 |
| Unmarried | 249 | 26.2 (18.2–34.3) | 203 | 24.3 (14.5–34.0) | 401 | 30.1 (22.8–37.5) | |
| Married | 2160 | 70.8 (61.7–80.0) | 1953 | 73.3 (63.6–83.0) | 2479 | 66.4 (59.4–73.5) | |
| Divorced/Widowed | 135 | 2.9 (1.0–4.9) | 139 | 2.5 (1.0–3.9) | 208 | 3.4 (2.8–4.1) | |
| Education level | | | | | | | <0.001 |
| College or above | 790 | 43.5 (22.2–64.9) | 688 | 48.0 (29.5–66.5) | 1011 | 46.8 (24.7–68.8) | |
| Senior high school and below | 1754 | 56.5 (35.1–77.8) | 1607 | 52.0 (33.5–70.5) | 2077 | 53.2 (31.2–75.3) | |
| Occupation | | | | | | | <0.001 |
| Medical staff | 47 | 2.3 (0.8–3.8) | 53 | 3.0 (0.4–5.5) | 69 | 2.7 (0.8–4.6) | |
| Civil servant/teacher | 136 | 4.7 (1.4–7.9) | 62 | 2.9 (1.1–4.7) | 61 | 2.2 (1.2–3.1) | |
| Farmer/worker | 788 | 18.5 (4.9–32.0) | 829 | 23.4 (0.0–47.4) | 943 | 24.1 (0.0–50.5) | |
| Others | 1573 | 74.6 (63.5–85.6) | 1351 | 70.7 (46.4–95.0) | 2015 | 71.0 (46.3–95.8) | |
| Annual family income (CNY) d | | | | | | | <0.001 |
| ≥100,000 | 1084 | 48.6 (25.4–71.8) | 779 | 49.5 (26.2–72.8) | 1184 | 45.0 (24.8–65.2) | |
| < 100,000 | 1460 | 51.4 (28.2–74.6) | 1516 | 50.5 (27.2–73.8) | 1904 | 55.0 (34.8–75.2) | |
| Self-reported health status | | | | | | | <0.001 |
| Good | 1816 | 76.5 (67.6–85.3) | 1798 | 86.0 (81.0–91.0) | 2478 | 86.2 (81.8–90.7) | |
| Medium | 658 | 21.4 (13.7–29.1) | 450 | 12.7 (8.3–17.2) | 518 | 11.7 (8.6–14.8) | |
| Poor | 70 | 2.2 (0.8–3.5) | 47 | 1.3 (0.2–2.4) | 92 | 2.0 (0.3–3.7) | |
| Health literacy | | | | | | | <0.001 |
| Limited (40–52 points) | 1948 | 73.1 (66.3–79.9) | 1648 | 66.4 (59.5–73.3) | 2175 | 65.9 (64.7–67.1) | |
| Adequate (53–66 points) | 596 | 26.9 (20.1–33.7) | 647 | 33.6 (26.7–40.5) | 913 | 34.1 (32.9–35.3) | |
The unweighted average ages in 2019, 2020, and 2021 were 46.9 ± 13.4, 49.5 ± 13.7, and 47.8 ± 13.9, respectively. The male:female ratios in 2019, 2020, and 2021 were 0.95:1, 0.94:1, and 0.93:1, respectively. No statistically significant difference was found in the gender or age composition of the participants among the different years.
## Bivariate analysis of health literacy level with variables of sociodemographic characteristics
As shown in Table 2, there were significant differences in health literacy level by age, education level, and occupation but not by gender or self-reported health status in 2019, 2020, and 2021.
**Table 2**
| Characteristic | AHL (Survey year 2019) | AHL (Survey year 2019).1 | AHL (Survey year2020) | AHL (Survey year2020).1 | AHL (Survey year 2021) | AHL (Survey year 2021).1 | Z | Pbfor Trend |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Characteristic | Z | Pbfor Trend | | | | | | |
| | %(95%CI) | P a | %(95%CI) | P a | %(95%CI) | P a | | |
| Gender | | 0.914 | | 0.225 | | 0.725 | | |
| Male | 26.8 (19.3–34.3) | | 35.3 (26.0–44.6) | | 33.6 (28.7–38.5) | | 212.172 | <0.001 |
| Female | 27.0 (20.1–34.0) | | 31.8 (25.7–37.8) | | 34.7 (30.6–38.7) | | 235.174 | <0.001 |
| Age, years | | <0.001 | | <0.001 | | <0.001 | | |
| 15–29 | 31.8 (18.5–45.0) | | 36.7 (26.4–47.0) | | 38.4 (34.0–42.7) | | 172.675 | <0.001 |
| 30–44 | 29.6 (25.1–34.1) | | 41.5 (30.5–52.6) | | 36.9 (30.2–43.5) | | 164.221 | <0.001 |
| 45–59 | 21.3 (15.7–27.0) | | 23.7 (16.6–30.9) | | 30.8 (23.1–38.4) | | 221.557 | <0.001 |
| 60–69 | 13.2 (7.8–18.6) | | 22.7 (13.3–32.2) | | 16.4 (11.9–21.0) | | 51.455 | <0.001 |
| Marital status | | 0.625 | | 0.022 | | 0.041 | | |
| Unmarried | 28.5 (16.4–40.7) | | 32.0 (22.0–42.0) | | 38.7 (32.2–45.2) | | 233.494 | <0.001 |
| Married | 26.5 (21.4–31.6) | | 34.6 (28.3–40.9) | | 32.5 (28.1–37.0) | | 223.990 | <0.001 |
| Divorced/Widowed | 23.2 (7.6–38.8) | | 18.8 (6.1–31.5) | | 25.6 (20.3–30.9) | | 22.414 | <0.001 |
| Education level | | <0.001 | | 0.042 | | 0.001 | | |
| College or above | 37.2 (24.4–50.0) | | 40.3 (35.8–44.8) | | 44.5 (38.6–50.5) | | 204.254 | <0.001 |
| Senior high school and below | 19.0 (12.3–25.7) | | 27.4 (14.0–40.8) | | 25.0 (15.8–34.2) | | 216.982 | <0.001 |
| Occupation | | <0.001 | | <0.001 | | <0.001 | | |
| Medical staff | 63.0 (34.7–91.2) | | 63.5 (54.5–72.4) | | 72.0 (46.8–97.2) | | 63.337 | <0.001 |
| Civil servant/teacher | 39.4 (20.2–58.5) | | 60.2 (27.0–93.4) | | 40.4 (11.4–69.4) | | 42.213 | <0.001 |
| Farmer/worker | 19.1 (12.8–25.4) | | 22.9 (13.3–32.5) | | 29.1 (23.7–34.5) | | 220.361 | <0.001 |
| Others | 26.9 (21.0–32.9) | | 34.8 (26.6–43.0) | | 34.2 (30.9–37.5) | | 272.719 | <0.001 |
| Annual family income (CNY)c | | 0.003 | | 0.001 | | 0.178 | | |
| ≥100,000 | 32.7 (30.3–35.1) | | 42.2 (31.3–53.1) | | 36.8 (33.4–40.1) | | 123.490 | <0.001 |
| < 100,000 | 21.5 (10.8–32.1) | | 25.2 (17.0–33.3) | | 32.0 (25.9–38.0) | | 354.688 | <0.001 |
| Self-reported health status | | 0.699 | | 0.636 | | 0.067 | | |
| Good | 27.6 (19.9–35.4) | | 33.3 (26.0–40.6) | | 35.3 (34.9–35.7) | | 297.991 | <0.001 |
| Medium | 24.4 (17.2–31.6) | | 36.5 (21.8–51.2) | | 27.5 (17.5–37.6) | | 88.458 | <0.001 |
| Poor | 27.0 (0.0–58.7) | | 24.8 (5.8–43.8) | | 22.6 (0.0–45.5) | | −29.843 | <0.001 |
## Trend analysis of health literacy, three aspects, and six dimensions of literacy over time
Table 2 shows the level and trend in health literacy for subgroups of sociodemographic characteristics. The prevalence of adequate health literacy in most subgroups showed a significant upward trend, but the subgroup of poor self-reported health status showed a significant downward trend from 2019 to 2021.
The level and trend of health literacy, the three aspects, and the six dimensions of literacy over time are presented in Table 3. The prevalence of adequate health literacy rose significantly over time, increasing from $26.9\%$ ($95\%$ CI 20.1–33.7) in 2019 to $34.1\%$ (32.9–35.3) in 2021.
**Table 3**
| Unnamed: 0 | Survey year 2021 | Survey year 2020 | Survey year 2019 | Z | Pa for Trend |
| --- | --- | --- | --- | --- | --- |
| Health literacy | 34.1 (32.9–35.3) | 33.6 (26.7–40.5) | 26.9 (20.1–33.7) | 316.001 | <0.001 |
| Three aspects | Three aspects | Three aspects | Three aspects | Three aspects | Three aspects |
| Basic knowledge and concepts | 46.3 (42.1–50.5) | 39.6 (27.2–52.0) | 45.8 (32.4–59.1) | 23.539 | <0.001 |
| Healthy lifestyles and behavior | 41.1 (36.4–45.7) | 38.7 (30.5–46.8) | 27.8 (19.7–36.0) | 561.653 | <0.001 |
| Health skills | 26.5 (23.4–29.6) | 29.6 (21.5–37.7) | 31.1 (20.1–42.1) | −207.328 | <0.001 |
| Six dimensions | Six dimensions | Six dimensions | Six dimensions | Six dimensions | Six dimensions |
| Scientific views of health | 57.3 (48.5–66.1) | 55.9 (44.0–67.9) | 59.7 (49.8–69.6) | −97.339 | <0.001 |
| Infectious disease literacy | 39.4 (28.1–50.7) | 39.9 (31.6–48.2) | 19.8 (12.9–26.7) | 848.035 | <0.001 |
| Chronic disease literacy | 37.5 (32.5–42.5) | 32.7 (25.1–40.2) | 37.2 (26.8–47.5) | 13.836 | <0.001 |
| Safety and first aid literacy | 61.4 (50.2–72.7) | 61.3 (50.2–72.4) | 64.6 (53.1–76.2) | −134.418 | <0.001 |
| Medical care literacy | 34.2 (26.3–42.1) | 34.4 (24.6–44.3) | 27.8 (17.1–38.5) | 277.968 | <0.001 |
| Health information literacy | 39.8 (29.5–50.2) | 44.2 (35.5–53.0) | 45.9 (30.5–61.3) | −249.012 | <0.001 |
In 2021, the lowest prevalence of adequate health literacy of the three aspects was for health skills; the lowest prevalence of the six dimensions was for medical care literacy.
In the trend analysis, the most significant increase was observed for healthy lifestyles and behavior (increased $39\%$ in 2020 and $48\%$ in 2021) among the three aspects and infectious disease literacy (increased $101\%$ in 2020 and $99\%$ in 2021) among the six dimensions; the most significant decrease was observed for health skills (decreased $15\%$ in 2021) among the three aspects and health information literacy (decreased $13\%$ in 2021) among the six dimensions.
## Multivariable logistic regression analysis of health literacy, health skill literacy, and health information literacy
As the most significant decrease was observed for health skills among the three aspects and health information literacy of the six dimensions, they were also included in the multivariable logistic regression analysis along with health literacy (Table 4).
**Table 4**
| Variables | Health literacy | Health literacy.1 | Health skills literacy | Health skills literacy.1 | Health information literacy | Health information literacy.1 |
| --- | --- | --- | --- | --- | --- | --- |
| Variables | | | | | | |
| | OR (95% CI) a | P | OR (95% CI) a | P | OR (95% CI) a | P |
| Year | Year | Year | Year | Year | Year | Year |
| 2021 | Ref | | ref | | ref | |
| 2020 | 0.9 (0.7–1.3) | 0.691 | 1.1 (0.8–1.7) | 0.528 | 1.2 (0.7–1.9) | 0.510 |
| 2019 | 0.7 (0.5–0.9) | 0.005 | 1.3 (0.9–1.9) | 0.242 | 1.3 (0.8–2.1) | 0.285 |
| Gender | Gender | Gender | Gender | Gender | Gender | Gender |
| Male | Ref | | ref | | ref | |
| Female | 0.9 (0.8–1.1) | 0.245 | 0.9 (0.7–1.1) | 0.309 | 1.0 (0.9–1.1) | 0.930 |
| Age, years | Age, years | Age, years | Age, years | Age, years | Age, years | Age, years |
| 15–29 | Ref | | ref | | ref | |
| 30–44 | 0.9 (0.7–1.2) | 0.575 | 1.0 (0.8–1.3) | 0.905 | 0.9 (0.7–1.0) | 0.131 |
| 45–59 | 0.7 (0.5–1.0) | 0.024 | 0.7 (0.5–1.0) | 0.083 | 0.7 (0.6–0.8) | <0.001 |
| 60–69 | 0.5 (0.3–0.7) | <0.001 | 0.5 (0.3–0.8) | 0.003 | 0.5 (0.4–0.7) | <0.001 |
| Marital status | Marital status | Marital status | Marital status | Marital status | Marital status | Marital status |
| Unmarried | Ref | | ref | | ref | |
| Married | 1.2 (0.9–1.7) | 0.158 | 1.1 (0.8–1.5) | 0.724 | 1.2 (0.9–1.5) | 0.146 |
| Divorced/Widowed | 1.1 (0.7–1.8) | 0.704 | 0.8 (0.6–1.2) | 0.344 | 1.1 (0.8–1.5) | 0.546 |
| Education level | Education level | Education level | Education level | Education level | Education level | Education level |
| College or above | Ref | | ref | | ref | |
| Senior high school and below | 0.6 (0.4–0.9) | 0.011 | 0.7 (0.5–1.1) | 0.099 | 0.6 (0.5–0.8) | <0.001 |
| Occupation | Occupation | Occupation | Occupation | Occupation | Occupation | Occupation |
| Medical staff | Ref | | ref | | ref | |
| Civil servant/teacher | 0.4 (0.2–0.7) | 0.001 | 0.4 (0.2–0.6) | <0.001 | 0.5 (0.3–0.7) | <0.001 |
| Farmer/worker | 0.3 (0.2–0.4) | <0.001 | 0.3 (0.2–0.4) | <0.001 | 0.4 (0.2–0.6) | <0.001 |
| Others | 0.3 (0.2–0.4) | <0.001 | 0.3 (0.2–0.4) | <0.001 | 0.4 (0.2–0.6) | <0.001 |
| Annual family income (CNY) b | Annual family income (CNY) b | Annual family income (CNY) b | Annual family income (CNY) b | Annual family income (CNY) b | Annual family income (CNY) b | Annual family income (CNY) b |
| ≥100,000 | Ref | | ref | | ref | |
| < 100,000 | 0.8 (0.5–1.0) | 0.078 | 0.8 (0.6–1.2) | 0.308 | 0.8 (0.6–1.0) | 0.021 |
| Self-reported health status | Self-reported health status | Self-reported health status | Self-reported health status | Self-reported health status | Self-reported health status | Self-reported health status |
| Good | Ref | | ref | | ref | |
| Medium | 1.1 (0.9–1.4) | 0.297 | 1.0 (0.8–1.2) | 0.991 | 1.0 (0.8–1.1) | 0.582 |
| Poor | 1.0 (0.6–1.7) | 0.984 | 1.1 (0.7–1.7) | 0.705 | 0.7 (0.4–1.3) | 0.295 |
Compared to 2021, the odds of adequate health literacy were significantly lower in 2019. Working as medical staff was a protective factor for adequate health literacy, health skill literacy, and health information literacy compared with other occupations. Risk factors for adequate health literacy and health information literacy were older age (45–69) and lower education level (senior high school and below). Risk factors for adequate health skill literacy were older age (60–69). Furthermore, adequate health information literacy was positively related to annual family income.
## Discussion
This is the first study describing the changes over time in health literacy in Wuhan, central China, based on representative three-time-series survey data. We observed that the prevalence of adequate health literacy rose significantly over time, increasing from $26.9\%$ ($95\%$ CI 20.1–33.7) in 2019 to $34.1\%$ (32.9–35.3) in 2021. Although the prevalence showed the same upward trend as a previous study [37] and is slightly higher than that of the Chinese national level ($25.4\%$) [39], it is still at a relatively low level, similar to American and European countries [26, 27]. The significant rise may be mainly related to economic and social development, the in-depth development of health education and health promotion, and the people's close attention to and urgency regarding health during the COVID-19 epidemic [24, 37, 40, 41].
In 2021, the highest prevalence of adequate health literacy among the three aspects was for basic knowledge and concepts, and the lowest was for health skills. The prevalence of adequate health literacy for healthy lifestyles and behaviors has risen rapidly, and health skills have shown a significant downward trend. In recent years, healthy lifestyle actions have been vigorously carried out, and knowledge of infectious diseases has been spread, effectively promoting healthy behavior [42]. Health education should focus on behavioral intervention and health skill training in the future.
In 2021, the lowest prevalence among the six dimensions was medical care literacy. Residents who lack medical care literacy may not be able to access and understand basic health information and services and cannot effectively utilize the complex healthcare system when they seek treatment [2, 10]. From the perspective of trend changes, the most significant increase was observed for infectious disease literacy among the six dimensions. It may be that the government and health departments paid more attention to educating the public about infectious disease prevention and control due to the COVID-19 epidemic [30]. Against this background, people not only knew about virus transmission routes but also knew how to engage in effective preventive behaviors such as hand washing, mask-wearing, household ventilation and disinfection, and reduced interpersonal contact by avoiding visiting crowded spaces [42, 43]. In addition, the prevalence of adequate literacy of the six dimensions for scientific views of health, safety and first aid, and health information showed a downward trend from 2019 to 2021, and health information literacy declined the most. Therefore, health education in Wuhan should focus on the aforementioned dimensions of literacy.
In multivariable logistic regression analysis, working as the medical staff was a protective factor for adequate health literacy, health skill literacy, and health information literacy compared with other occupations, which is in line with the characteristics of an occupation engaged in the medical and healthcare industries [23, 24, 44]. The education level, knowledge reserve, and information acquisition channels of medical staff are better than those of other occupations. This study also showed that the prevalence of health literacy of residents who reported poor health status showed a significant downward trend from 2019 to 2021, indicating that medical staff can be used to carry out health education of residents with poor health status seeking treatment, to improve their health literacy in a targeted manner.
Risk factors for adequate health literacy and health information literacy were older age and lower education level, consistent with previous studies [24, 25, 37, 45]. This may be due to the following reasons: the cognitive ability, learning ability, and memory of elderly people decline, and their ability to accept new knowledge is relatively poor, directly leading to the poor acquisition of health knowledge and skills and limited health literacy; well-educated individuals are more likely to seek beneficial information and medical care and can communicate effectively with healthcare workers [46]. In addition, adequate health information literacy was positively related to annual family income, consistent with previous studies [21, 38, 47]. This may be because a good economic situation positively affects the acquisition of health information and the utilization of healthcare resources. This indicates that targeted health education and health promotion should be strengthened, focusing on residents with older ages, lower education levels, and lower annual family incomes.
Our study has several limitations that can be improved in further research. First, the study design was cross-sectional, and no causal relationships could be made. Second, some factors, such as health behaviors, and health service quality were not assessed. Third, we obtained data from self-reported items, which are prone to bias. Finally, our research population consisted of permanent residents aged 15–69, and some groups were not included, which should be further studied.
This is the first study to characterize the levels, changes, and factors related to health literacy among residents aged 15–69 from 2019 to 2021 in central China. Overall, although the prevalence of adequate health literacy rose significantly, increasing from $26.9\%$ ($95\%$ CI 20.1–33.7) in 2019 to $34.1\%$ (32.9–35.3) in 2021, it was still at a relatively low level. In the context of the COVID-19 epidemic, the prevalence of adequate infectious disease literacy rose rapidly, but health skills and health information literacy declined. The protective factor for adequate health literacy, health skill literacy, and health information literacy was working as medical staff, and the risk factors were older age, lower education level, and lower annual family income. Tailored health education and promotion strategies are needed for different subgroups of residents to improve health literacy, especially for health skills and health information literacy. At the same time, medical staff with adequate health literacy can effectively be used by providing health education for people who seek treatment with a poor health status to improve the health literacy of this population.
## Data availability statement
The data analyzed in this study is subject to the following licenses/restrictions: The datasets generated during and/or analyzed during the current study are not publicly available due to restrictions applied to the availability of these data but are available from the corresponding author on reasonable request. Requests to access these datasets should be directed to [email protected].
## Ethics statement
The processes and sampling design of the survey were reviewed and approved by the Institutional Review Board (IRB) of Wuhan CDC (WHCDCIRB-K-2019016). Written informed consent to participate in this study was provided by the participants' legal guardian/next of kin.
## Author contributions
XM: conceptualization, formal analysis, investigation, data curation, and writing—reviewing and editing. GC, YZ, and QW: formal analysis, investigation, and writing—reviewing and editing. JL: conceptualization, methodology, writing—reviewing and editing, supervision, and project administration. YL: conceptualization, methodology, investigation, writing—reviewing and editing, supervision, and project administration. All authors contributed to the article and approved the submitted version.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher's note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpubh.2023.1092892/full#supplementary-material
## References
1. 1.National Library of Medicine. An Introduction to Health Literacy. (2022). https://nnlm.gov/guides/intro-health-literacy (accessed August 1, 2022).. *An Introduction to Health Literacy* (2022)
2. 2.World Health Organization. Health Literacy Toolkit for Low- and Middle-Income Countries. (2014). Available online at: https://www.who.int/publications/i/item/9789290224754 (accessed August 1, 2022).. *Health Literacy Toolkit for Low- and Middle-Income Countries* (2014)
3. 3.U.S. National Library of Medicine. Health Literacy. (2022). Available online at: https://medlineplus.gov/healthliteracy.html#summary (accessed August 1, 2022).. *Health Literacy* (2022)
4. Schaffler J, Leung K, Tremblay S, Belzile E, Lambrou A, Lambert SD. **The effectiveness of self-management interventions for individuals with low health literacy and/or low income: a descriptive systematic review**. *J Gen Intern Med.* (2018) **33** 510-23. DOI: 10.1007/s11606-017-4265-x
5. Marciano L, Camerini AL, Schulz PJ. **The role of health literacy in diabetes knowledge, self-care, and glycemic control: a meta-analysis**. *J Gen Intern Med.* (2019) **34** 1007-17. DOI: 10.1007/s11606-019-04832-y
6. Waters EA, Biddle C, Kaphingst KA, Schofield E, Kiviniemi MT, Orom H. **Examining the interrelations among objective and subjective health literacy and numeracy and their associations with health knowledge**. *J Gen Intern Med.* (2018) **33** 1945-53. DOI: 10.1007/s11606-018-4624-2
7. Muscat DM, Shepherd HL, Nutbeam D, Trevena L, Mccaffery KJ. **Health literacy and shared decision-making: exploring the relationship to enable meaningful patient engagement in healthcare**. *J Gen Intern Med.* (2021) **36** 521-4. DOI: 10.1007/s11606-020-05912-0
8. Chin J, Wang HP, Awwad AW, Graumlich JF, Wolf MS, Morrow DG. **Health literacy, processing capacity, illness knowledge, and actionable memory for medication taking in type 2 diabetes: cross-sectional analysis**. *J Gen Intern Med.* (2021) **36** 1921-7. DOI: 10.1007/s11606-020-06472-z
9. Liu LF, Huang J, Li GX, Chen Z, He TF. **The economic costs of limited health literacy in China: evidence from China's national health literacy surveillance data**. *BMC Health Serv Res* (2022) **22** 521. DOI: 10.1186/s12913-022-07795-9
10. Vandenbosch J, Van den Broucke S, Vancorenland S, Avalosse H, Verniest R, Callens M. **Health literacy and the use of healthcare services in Belgium**. *J Epidemiol Commun H.* (2016) **70** 1032-8. DOI: 10.1136/jech-2015-206910
11. Dewalt DA, Broucksou KA, Broucksou KA, Hawk V, Baker DW, Schillinger D. **Relationship between literacy, knowledge, self-care behaviors, and heart failure-related quality of life among patients with heart failure**. *J Gen Intern Med.* (2011) **26** 979-86. DOI: 10.1007/s11606-011-1668-y
12. Lindquist LA, Go L, Fleisher J, Jain N, Friesema E, Baker DW. **Relationship of health literacy to intentional and unintentional non-adherence of hospital discharge medications**. *J Gen Intern Med.* (2012) **27** 173-8. DOI: 10.1007/s11606-011-1886-3
13. Wu JR, Holmes GM, Dewalt DA, Macabasco-O'Connell A, Bibbins-Domingo K, Ruo B. **Low literacy is associated with increased risk of hospitalization and death among individuals with heart failure**. *J Gen Intern Med.* (2013) **28** 1174-80. DOI: 10.1007/s11606-013-2394-4
14. Sanders K, Schnepel L, Smotherman C, Livingood WC, Dodani S, Antonios N. **Assessing the impact of health literacy on education retention of stroke patients**. *Prev Chronic Dis* (2014) **11** E55. DOI: 10.5888/pcd11.130259
15. 15.World Health Organization. New Bangkok Charter for Health Promotion Adopted to Address Rapidly Changing Global Health Issues. (2005). Available online at: https://www.who.int/news/item/11-08-2005-new-bangkok-charter-for-health-promotion-adopted-to-address-rapidly-changing-global-health-issues (accessed August 1, 2022).. *New Bangkok Charter for Health Promotion Adopted to Address Rapidly Changing Global Health Issues* (2005)
16. 16.World Health Organization. 9th Global Conference on Health Promotion: Global Leaders Agree to Promote Health in Order to Achieve Sustainable Development Goals. (2016). Available online at: https://www.who.int/news/item/21-11-2016-9th-global-conference-on-health-promotion-global-leaders-agree-to-promote-health-in-order-to-achieve-sustainable-development-goals (accessed August 1, 2022).. *9th Global Conference on Health Promotion: Global Leaders Agree to Promote Health in Order to Achieve Sustainable Development Goals* (2016)
17. 17.World Health Organization. WHO Director-General's Statement on IHR Emergency Committee on Novel Coronavirus (2019-nCoV). (2020). Available online at: https://www.who.int/director-general/speeches/detail/who-director-general-s-statement-on-ihr-emergency-committee-on-novel-coronavirus-(2019-ncov) (accessed August 1, 2022).. *WHO Director-General's Statement on IHR Emergency Committee on Novel Coronavirus* (2019-nCoV)
18. Brach C, Harris LM. **Healthy people 2030 health literacy definition tells organizations: make information and services easy to find, understand, and use**. *J Gen Intern Med.* (2021) **36** 1084-5. DOI: 10.1007/s11606-020-06384-y
19. Curtis LM, Kwasny MJ, Opsasnick L, O'Conor RM, Yoshino-Benavente J, Eifler M. **Change in health literacy over a decade in a prospective cohort of community-dwelling older adults**. *J Gen Intern Med.* (2021) **36** 916-22. DOI: 10.1007/s11606-020-06423-8
20. Visscher BB, Steunenberg B, Heijmans M, Hofstede JM, Devillé W, van der Heide I. **Evidence on the effectiveness of health literacy interventions in the EU: a systematic review**. *BMC Public Health.* (2018) **18** 1414. DOI: 10.1186/s12889-018-6331-7
21. Huang YQ, Qi F, Wang R, Jia X, Wang Y, Lin P. **The effect of health literacy on health status among residents in Qingdao, China: a path analysis**. *Environ Health Prev* (2021) **26** 78. DOI: 10.1186/s12199-021-01001-8
22. Rui L, Shan H. **Foreign studies on the assessment tools for health literacy: a systematic review**. *J Modern Inform* (2020) **40** 154-66, 77. DOI: 10.3969/j.issn.1008-0821.2020.11.016
23. Nie XQ, Li YH, Li CN, 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) **60** 569-78. DOI: 10.1016/j.amepre.2020.05.032
24. Mei X, Zhong Q, Chen G, Huang YX, Li JL. **Exploring health literacy in Wuhan, China: a cross-sectional analysis**. *BMC Public Health* (2020) **20** 1417. DOI: 10.1186/s12889-020-09520-9
25. Li ZH, Tian YQ, Gong ZC, Qian L. **Health literacy and regional heterogeneities in china: a population-based study**. *Front Public Health* (2021) **9** 603325. DOI: 10.3389/fpubh.2021.603325
26. Kutner M, Greenberg E, Jin Y, Paulsen C. *The Health Literacy of America's Adults: Results from the 2003 National Assessment of Adult Literacy.* (2006)
27. Baccolini V, Rosso A, Di Paolo C, Isonne C, Salerno C, Migliara G. **What is the prevalence of low health literacy in European Union member states? a systematic review and meta-analysis**. *J Gen Intern Med.* (2021) **36** 753-61. DOI: 10.1007/s11606-020-06407-8
28. 28.Publicity Department of the National Health Commission of the People's Republic of China. In 2019, the Health Literacy Level of Chinese Residents Rose to 19.17%. (2020). Available online at: http://www.nhc.gov.cn/xcs/s3582/202004/df8d7c746e664ad783d1c1cf5ce849d5.shtml (accessed July 29, 2022).. (2020)
29. Jia YH, Ma SL, Bai LJ, Xiao Q, Wu Y, Gao Y. **Health literacy and disparities in knowledge, attitude and practice regarding COVID-19 among college students during the COVID-19 outbreak in China: a cross-sectional study**. *Risk Manag Healthc Policy.* (2021) **14** 4477-88. DOI: 10.2147/RMHP.S319331
30. Lai CC, Shih TP, Ko WC, Tang HJ, Hsueh PR. **Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) and coronavirus disease-2019 (COVID-19): the epidemic and the challenges**. *Int J Antimicrob Ag* (2020) **55** 105924. DOI: 10.1016/j.ijantimicag.2020.105924
31. Albus C. **Health literacy: do we have effective interventions to enhance it in socially disadvantaged people?**. *Eur J Prev Cardiol.* (2019) **26** 1760-1. DOI: 10.1177/2047487319861221
32. Paakkari L, Okan O. **COVID-19: health literacy is an underestimated problem**. *Lancet Public Health.* (2020) **5** E249-50. DOI: 10.1016/S2468-2667(20)30086-4
33. Castro-Sanchez E, Chang P, Vila-Candel R, Escobedo AA, Holmes AH. **Health literacy and infectious diseases: why does it matter?**. *Int J Infect Dis.* (2016) **43** 103-10. DOI: 10.1016/j.ijid.2015.12.019
34. Yang P, Dunzhu C, Widdowson MA, Wu S, Ciren P, Duoji D. **Infectious disease-specific health literacy in Tibet, China**. *Health Promot Int.* (2018) **33** 84-91. DOI: 10.1093/heapro/daw054
35. 35.Statistical Bulletin of Wuhan's National Economic Social Development. Population Distribution. (2022). Available online at: http://www.wuhan.gov.cn/zjwh/whgk/202003/t20200316_976479.shtml (accessed July 29, 2022).. (2022)
36. 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) **24** 100464. DOI: 10.1016/j.lanwpc.2022.100464
37. Yang PY, Ou YR, Yang HL, Pei X, Li J, Wang Y. **Research on influencing factors and dimensions of health literacy in different age groups: before and after the COVID-19 era in Chongqing, China**. *Front Public Health* (2021) **9** 690525. DOI: 10.3389/fpubh.2021.690525
38. Wu Y, Wang L, Cai ZY, Bao LQ, Ai P, Ai ZS. **Prevalence and risk factors of low health literacy: a community-based study in Shanghai, China**. *Int J Env Res Pub Health* (2017) **14** 628. DOI: 10.3390/ijerph14060628
39. 39.Publicity Department of the National Health Commission of the People's Republic of China. In 2021, the National Health Literacy Level of Residents Will Reach 25.40% in China. (2022). Available online at: http://www.nhc.gov.cn/xcs/s3582/202206/5dc1de46b9a04e52951b21690d74cdb9.shtml (accessed August 1, 2022).. (2022)
40. Li SJ, Cui GH, Kaminga AC, Cheng SX, Xu HL. **Associations between health literacy, eHealth literacy, and COVID-19-related health behaviors among Chinese college students: cross-sectional online study**. *J Med Internet Res* (2021) **23** e25600. DOI: 10.2196/25600
41. Naveed MA, Shaukat R. **Health literacy predicts Covid-19 awareness and protective behaviours of university students**. *Health Info Libr J.* (2022) **39** 46-58. DOI: 10.1111/hir.12404
42. Zhong B-L, Luo W, Li H-M, Zhang Q-Q, Liu X-G, Li W-T. **Knowledge, attitudes, and practices towards COVID-19 among Chinese residents during the rapid rise period of the COVID-19 outbreak: a quick online cross-sectional survey**. *Int J Biol Sci.* (2020) **16** 1745-52. DOI: 10.7150/ijbs.45221
43. Chen Y, Zhou R, Chen BY, Chen H, Li Y, Chen Z. **Knowledge, perceived beliefs, and preventive behaviors related to COVID-19 among chinese older adults: cross-sectional web-based survey**. *J Med Internet Res* (2020) **22** e23729. DOI: 10.2196/23729
44. Wallington S, Cummings C, Greaney ML, Rampa S, Lees FD. **Exploring healthcare. employees' perceptions of health literacy to increase health literacy among underserved populations**. *Ann Behav Med.* (2020) **54** S317
45. Mcclintock H, Schrauben S, Andrews A, Wiebe D. **Measurement of health literacy to advance global health research: a study based on demographic and health surveys in 14 sub-Saharan countries**. *Lancet Glob Health.* (2017) **5** 18-18. DOI: 10.1016/S2214-109X(17)30125-0
46. Nutbeam D, Lloyd JE. *Understanding and Responding to Health Literacy as a Social Determinant of Health* (2021) p. 159-73. DOI: 10.1146/annurev-publhealth-090419-102529
47. Tang CX, Wu XJ, Chen XF, Pan BY, Yang XC. **Examining income-related inequality in health literacy and health-information seeking among urban population in China**. *BMC Public Health* (2019) **19** 221. DOI: 10.1186/s12889-019-6538-2
|
---
title: Saccharomyces cerevisiae fermentation product improves robustness of equine
gut microbiome upon stress
authors:
- Erika Ganda
- Anirikh Chakrabarti
- Maria I. Sardi
- Melissa Tench
- Briana K. Kozlowicz
- Sharon A. Norton
- Lori K. Warren
- Ehsan Khafipour
journal: Frontiers in Veterinary Science
year: 2023
pmcid: PMC9998945
doi: 10.3389/fvets.2023.1134092
license: CC BY 4.0
---
# Saccharomyces cerevisiae fermentation product improves robustness of equine gut microbiome upon stress
## Abstract
### Introduction
Nutritional and environmental stressors can disturb the gut microbiome of horses which may ultimately decrease their health and performance. We hypothesized that supplementation with a yeast-derived postbiotic (*Saccharomyces cerevisiae* fermentation product-SCFP) would benefit horses undergoing an established model of stress due to prolonged transportation.
### Methods
Quarter horses ($$n = 20$$) were blocked based on sex, age (22 ± 3 mo) and body weight (439 ± 3 kg) and randomized to receive either a basal diet of $60\%$ hay and $40\%$ concentrate (CON) or the basal diet supplemented with 21 g/d Diamond V TruEquine C (SCFP; Diamond V, Cedar Rapids, IA) for 60 days. On day 57, horses were tethered with their heads elevated 35cm above wither height for 12 h to induce mild upper respiratory tract inflammation. Fecal samples were collected at days 0, 28, and 56 before induction of stress, and at 0, 12, 24, and 72 h post-stress and subjected to DNA extraction and Nanopore shotgun metagenomics. Within sample (alpha) diversity was evaluated by fitting a linear model and between sample (beta) diversity was tested with permutational ANOVA.
### Results
The SCFP stabilized alpha diversity across all time points, whereas CON horses had more fluctuation ($P \leq 0.05$) at 12, 24, and 72 h post-challenge compared to d 56. A significant difference between CON and SCFP was observed at 0 and 12 h. There was no difference in beta-diversity between SCFP and CON on d 56.
### Discussion
Taken together, these observations led us to conclude that treatment with SCFP resulted in more robust and stable microbial profiles in horses after stress challenge.
## Introduction
The role of the microbiome and its importance has been well-established in several systems over the past two decades, including in environmental [1, 2], biomedical (3–5), and agricultural (6–8) contexts. Herbivores are particularly impacted by the gastrointestinal microbiome, given their interdependency with metabolic pathways only present in microbes that are necessary for digestion of complex carbohydrates present in forages that typify the equine daily diet.
In horses relatively fewer research studies investigating the microbiome are available, however the number of such studies is increasing rapidly [8, 9]. Among the many factors that can impact the equine microbiome, stress is one of the most preeminent ones. Both diet- [10] and exercise- [11, 12] induced stress have been associated with microbiome changes in horses. Unstable microbiomes represent an open niche for opportunistic pathogen establishment and are associated with worse health outcomes. In fact, colonization resistance is one of the biggest roles played by the microbiome in maintaining host health (13–15). Thus, maintaining a robust microbiome upon stressful events would be beneficial for horse health.
Several techniques can be applied to intentionally manipulate the diversity and composition of the gut microbiome in the quest to maintain an optimal microbial community. Diet modification, pre-, pro-, and postbiotic administration, and more drastic therapeutics such as antibiotic therapy and fecal microbiota transplantation are also used for microbiome modulation [16]. Postbiotics are defined as a “preparation of inanimate microorganisms and/or their components that confers a health benefit on the host” [17, 18]. These preparations do not necessarily originate from probiotic microorganisms and must contain an unpurified mixture of inanimate organisms and their metabolites. Because the mode of action of postbiotics does not rely on presence of live organisms in the final product, they represent an attractive alternative for feed supplementation given their better stability during feed processing [19]. Several studies evaluating the efficacy of postbiotic supplementation of *Saccharomyces cerevisiae* fermentation products (SCFP) in bovine [20, 21], avian [22], and equine [23] species have been performed. While the mechanisms by which postbiotics confer benefits to the host have not yet been completely elucidated, much of the literature indicates that postbiotic supplementation is associated with microbiome optimization [24] and improvement of immune function [20, 22, 25]. However, less is known about the effects of SCFP on horses. A few recent studies have indicated improvement in immune parameters in a vaccine challenge model [23, 26] while no difference was observed in the microbiota of racehorses fed a yeast supplement [27]. Taking the wealth of evidence of the beneficial effects of postbiotic administration in many species, it is reasonable to hypothesize that postbiotic administration would benefit horses under stress.
Horses are exposed to stressful situations daily, including transportation, exercise, and diet changes. Although several studies have demonstrated the impact of stressful events on the equine fecal microbiome [10, 28] little evidence is available on how postbiotic administration can impact the robustness of microbiome in horses under stress. Thus, the objective of this study was to determine if supplementation with SCFP would result in more robust microbiome in an established equine model to simulate stress due to prolonged transportation. We hypothesized that SCFP supplementation would result in a more robust microbiome that would be less impacted by experimental stress.
## Experimental design, animals, and sample collection
The animal experiment for present microbiome study was described by Tench et al. [ 29]. The protocol for the use of experimental animals was approved by the Institutional Animal Care and Use Committee at the University of Florida in Gainesville, FL (#201810324) under the Guide for the Care and Use of Agricultural Animals in Research and Teaching [30].
Briefly, 20 young and clinically healthy horses in training (mean ± SEM; initial age 22 ± 0.3 mo and BW 439 ± 3 kg) were paired by age and sex and randomly assigned to one of the two experimental treatments for 60 days. Treatments included supplementation with 0 g/d (Control; no treatment Control) or 21 g/d Diamond V TruEquine C (SCFP; Diamond V, Cedar Rapids, IA). A basal diet of $60\%$ Coastal bermudagrass hay and $40\%$ concentrate formulated to meet the nutrient requirements of horses at a moderate rate of growth [31] was offered to all horses. Treatment administration was done by top dressing SCFP on the concentrate ration. Horses were exercised 4 days per week for 30–45 min/d at light to moderate intensity. On day 57, horses were placed in individual stalls and tethered with their heads elevated 35 cm above wither height for 12 h to induce mild upper respiratory tract inflammation according to a previously established protocol to mimic long-distance transport stress [32, 33]. Induction of inflammation was confirmed by significantly elevated serum cortisol and blood leukocyte measurements performed after stress induction compared to pre-stress [34, 35]. The stress period was relieved after the 12 h timepoint by untethering of the horse heads. Fecal samples were collected into sterile containers at seven time points: days 0, 28, and 56 before induction of stress, and at 0, 12, 24, and 72 h post-stress, where 0 h is the time at which the horses were untethered. Samples were immediately placed on ice and transported to the laboratory where they were kept in a −80°C freezer until DNA extraction. A schematic of the experimental design and sample collection is given in Figure 1.
**Figure 1:** *Study overview. Young and clinically healthy horses were paired by age and sex and randomly assigned into Control (n = 10) or Saccharomyces cerevisiae fermentation product (SCFP; n = 10). Horses received diets for 60 days. On day 57, horses were subjected to a previously established stress protocol to induce mild upper respiratory tract inflammation that mimicked long-distance transportation. Samples were collected on days 0, 28, and 56 before stress and at 0, 12, 24, and 72 h post-stress.*
## DNA extraction
Fecal samples were removed from the −80°C freezer 1 day prior to DNA extraction and thawed in a 4°C refrigerator overnight. The ZymoBIOMICS 96 MagBead DNA kit (Zymo Research Corporation, Irvine, CA) was used in a Biomek i7 (Beckman Coulter, Indianapolis, IN) workstation for DNA extraction according to manufacturer's instructions. Four extraction blanks were included in each 96 well plate to confirm that cross contamination did not occur.
## Nanopore sequencing
Libraries were constructed using the SQK-RPB004 Rapid PCR Barcoding kit (ONT, Oxford, UK). Library preparation included DNA extraction blanks for quality control. Shotgun metagenomic sequencing was performed using R9.4.1 FLO-MIN 106 flow cells on the GridION platform (ONT, Oxford, UK), multiplexing 12 samples in each flow cell. Each sequencing run lasted 70 h. The MinKNOW ONT software (v 3.6.5) with Guppy basecaller was used for sequencing using the high-accuracy basecalling setting, followed by de-multiplexing, adapter trimming, and quality control using default settings.
## Taxonomic assignment and microbial diversity
Fastq files obtained from the MinKNOW ONT workflow were used for microbial taxonomic classification. First, host DNA was removed by mapping fastq files to the horse genome (assembly EquCab3.0) using Minimap2 [36] followed by the removal of any reads matching the horse genome using SAMtools [37]. The remaining reads were assumed to be from microbial origin and used for taxonomic assignment. To improve microbial classification, a custom database was made, which contained high quality genomes from the RefSeq database [38] and published metagenome-assembled genomes (38–40). The Kraken2 pipeline [41] was used for species identification and Bracken was used to estimate species abundances [42]. Diversity metrics were calculated in R [43] using the Phyloseq package [44, 45] with the rarefied species count table from Bracken as input. Species tables were center-log transformed using the microbiome package [46] after imputation of zeros using a Bayesian multiplicative replacement method from the zCompositions package [47]. Species with non-zero presence in at least $75\%$ of samples and relative abundance >$0.001\%$ were identified separately in the pre-stress and post-stress periods and the superset containing all species was used for differential abundance analysis.
## Functional potential
To have a better understanding of the microbiome functional potential, the Carbohydrate-Active enZymes (CAZy) [48] present in the microbiome communities for each sample were identified. First, genomes from microbial species identified with Kraken2 were annotated using PROKKA [45], followed by additional assessment of gene function using EggNOG-mapper v2 [49]. After the annotation process was completed, a custom python script was used to compile the CAZy for each genome, generating a table with the accumulated CAZy potential for all the microbes identified for each sample. Results were compiled into a final table containing numbers of annotated features for each sample.
## Diversity metrics
Within sample (alpha) diversity was evaluated by fitting a linear model with the lmer function of the lme4 package [50] in R. The model included Shannon diversity index as the dependent variable, horse as a random effect, treatment, timepoint, and their interactions as independent variables. Because stress is nested within timepoint, the effect of stress only is evaluated in a separate model. Between sample (beta) diversity was tested with permutational ANOVA using the adonis function in the vegan package [51] in R. The model included Aitchison distances [52] calculated based on CLR transformed values as the dependent variable, and treatment, timepoint, and their interactions as independent variables. Data was visualized with PCA using the Phyloseq package [44, 45].
## Differential abundance
A modified version of the linear discriminant analysis from the LinDA package [53] was used to fit linear models that included relative abundances as the dependent variable, treatment, timepoint, and their interaction as independent variables, and horse as a random effect. The output from each model was then analyzed with the emmeans package [54] to calculate fold changes of centralized log ratio (CLR) transformed data of each measurement (species) for each animal with respect to their initial sample collected at day 0. False discovery rate (FDR) correction [55, 56] was used to identify species within each timepoint that significantly differed between treatment and Control.
## Correlation networks
An adaptation of the CoNet framework [57], which includes generation of a combination of diverse measures of correlation (including Pearson's, Spearman's, and Kendall's correlation coefficients) using CLR transformed data was used for correlation network analyses. Distributions of all pair-wise scores between the nodes were computed for each timepoint. Only edges (correlations) with p-values < 0.05 after FDR correction [55, 56] were taken into further consideration, and edges not supported by at least two measures were discarded.
## Clustering
Identification of the optimal number of clusters and clustering was calculated and performed using gap statistics [58] in MATLAB R2019b [59] using the spearman correlation for species and CAZy identifiers, and Aitchison distance for samples. The difference of the CLR transformed values at any time point and its corresponding value at day 0 were used as the input. The data was sorted based on experimental variables or clusters and visualized.
## Sequencing parameters
A total of 140 samples were sequenced. On average, 389,680 reads were obtained per sample (mean 389,680, median 377,834, SD 118,596). Read N50 lengths averaged 4,043 bp (mean 4,043 median 4,052, SD 318). Reads had an average quality score of 12 (mean 12, median 12, SD 0.6). On average, 1,429,212,272 total bases were obtained per sample, with a standard deviation of 413,891,226 bases per sample. Four samples had low sequencing throughput and were removed from further analysis.
## Taxonomic assignment
On average, $67\%$ of reads were assigned at the species level (mean $66.9\%$, median $67.4\%$, SD $4.4\%$). A total of 119 taxa were identified (Supplementary Table 5). Of those, 27 taxa fit the criteria of being present in at least $75\%$ of samples and relative abundance >$0.001\%$ in the pre-stress period and 18 taxa fit the criteria in the post-stress period. The final superset that was used for differential abundance analysis contained 27 taxa.
## Stress significantly impacts microbial diversity and SCFP treatment leads to a more robust microbiome after stress
Alpha diversity was similar between Control and SCFP groups in the pre-stress period (Figure 2), indicating treatment with SCFP did not significantly alter Shannon microbial diversity index values. Stress impacted ($P \leq 0.0001$) diversity levels both in the Control and SCFP groups. However, stress had a lower impact in changing the SCFP group's diversity levels when compared to Controls. Overall, horses treated with SCFP exhibited robust microbial diversity after stress, with less variation and overall lower stress-induced drop in diversity when compared to the Control group (Supplementary Table 1). When within-group comparisons were made, statistical differences were observed in the Control group between several timepoints (Figure 2, gray dotted lines; Supplementary Table 2). On the other hand, fewer timepoints were significantly different from one another when within-group comparisons were made in the SCFP group (Figure 2, blue dotted lines; Supplementary Table 3), indicating that SCFP treatment might have contributed to more stable diversity levels post-stress.
**Figure 2:** *Alpha diversity comparisons. Shannon diversity metrics were analyzed with a linear model which included horse as a random effect, treatment, timepoint, and their interactions. Raw means and standard deviations are shown in the plot. Blue diamonds represent SCFP, and gray circles represent Control. Transparent lines represent the hypothetical trajectory of diversity. Dashed horizontal lines represent within-group pairwise significant differences at the 0.05 level after Bonferroni multiple comparison adjustment. Asterisks indicate timepoints in which SCFP significantly differs from Control **P < 0.01, *P < 0.05.*
Beta diversity was variable in the pre-stress period (Figure 3). At time 0 h (time at which the horses were untethered), horses assigned to the SCFP treatment formed two subclusters, whereas horses assigned to the Control treatment clustered in the same overall region (Figure 3, panel 1). On day 28, treated and untreated horses clustered in two overlapping groups (Figure 3, panel 2), and became homogeneous over time, with no clear difference between Control and SCFP-treated horses on day 56 (Figure 3, panel 3). However, Control and SCFP-treated horses had two completely different clustering trajectories after stress, with SCFP and Control horses clustering separately at 0 and 12 h post-stress (Figure 3, panels 4 and 5) and culminating again in a homogenous group at 72 h post-stress (Figure 3, panel 7).
**Figure 3:** *Beta diversity comparisons. Aitchison distances were analyzed with permutational ANOVA (PERMANOVA) model which included treatment, timepoint, and their interactions. Plots depict principal component analysis of Aitchison distances with CLR transformed data.*
## The stress impact was greater for Control horses
Stress challenge and SCFP treatment significantly influenced microbial composition at the species level (PERMANOVA of Aitchison distances: Treatment, $$P \leq 0.01$$; Timepoint, $$P \leq 0.01$$; Treatment × Timepoint, $$P \leq 0.01$$). Two species clusters were identified (Figure 4A, vertical clusters A and B). The larger cluster (cluster A−18 species) comprised mainly species that increased in abundance after stress challenge. The smaller cluster (cluster B—nine species) comprised species that decreased in abundance after stress challenge (Figure 4A, vertical clusters A and B, Supplementary Table 5). Notably, Control horses had a much more marked reduction in species belonging to cluster B after stress when compared to those treated with SCFP. When total microbial composition was used as a basis for clustering analysis of the samples, five major sample clusters were identified (Figure 4A, clusters I, II, III, IV, and V). Very different trajectories were observed between the SCFP and Control treatments after stress (Figure 4B), with microbial composition of Control horses mostly belonging to cluster V, while SCFP treated horses exhibited microbiome compositions representatives of all clusters.
**Figure 4:** *Stress challenge results in different microbial profiles in Control and SCFP treated horses. (A) Heatmap of centered log-ratio relative abundance of bacterial species detected in fecal samples of horses. (B) Pie charts represent the presence of samples across different clusters for Control and SCFP groups across time. Species cluster A contains species that drastically increase in relative abundance upon stress challenge, whereas species cluster B is comprised of species which decrease after stress challenge.*
## Stress challenge resulted in significant differential abundances in a time-dependent manner
Treatment with SCFP significantly increased the abundances of Erysipelotrichaceae before stress challenge. In fact, this was the only significantly different taxa between Control and SCFP in the pre-challenge period (Figure 5A, panels 1 and 2), and SCFP treated animals had an overall positive log ratios throughout the entire study (Supplementary Figures).
**Figure 5:** *Differential abundances. The effect size (log2-fold change) is shown for each species, and only significantly different species are shown in each plot with their correspondent confidence interval. Statistical models included relative abundance as the dependent variable, horse as a random effect, treatment, timepoint, and their interactions as independent variables. Multiple hypothesis testing correction was performed with Benjamini Hochberg False Discovery Rate method. A negative fold change indicates an increase in relative abundance in SCFP compared to Control, and a positive fold indicates a decrease in relative abundance in SCFP compared to Control. (A) Differentially abundant species before stress. (B) Differentially abundant species after stress.*
Many more species were significantly differentially abundant after stress challenge particularly at times 0 and 12 h post-stress (Figure 5B). At time 0 h, eight species were significantly increased in the SCFP group compared to Control, and three species were significantly decreased (Figure 5B, panel 1). Statistically different species were observed between groups up to 24 h after stress (Figure 5B, panels 2 and 3), with no significantly different species observed at 72 h after stress (Figure 5B, panel 4).
## SCFP treated horses demonstrated more robust microbial functionality post-stress as compared to Control horses
Clustering analysis of the functional potential of the samples, measured by CAZy families, identified two major functional sample clusters (Figure 6A, clusters I and II). The CAZy families identified were Auxiliary Activity Family (AA), Carbohydrate-Binding Module Family (CBM), Carbohydrate Esterase Family (CE), Glycoside Hydrolase Family (GH), Glycosyl Transferase Family (GT), and Polysaccharide Lyase Family (PL). A more pronounced increase in CAZy families was observed after stress in Control horses compared to SCFP horses (Figure 6A). Similar to compositional clustering outcomes, Control and SCFP groups exhibited markedly different functional profiles following imposition of the stressor (Figure 6B), with Control horses demonstrating a switch to cluster I immediately after stress challenge, and again completely switching to cluster II from 12 to 72 h post-stress. Conversely SCFP treated horses displayed microbiome functional potential representatives of both clusters throughout the entire study period.
**Figure 6:** *Functional potential is differentially impacted by treatment and stress in SCFP and Control horses. (A) Heatmap of centered log-ratio relative abundance of carbohydrate active enzymes (CAZy) families detected in fecal samples of horses. (B) Pie charts represent the presence of samples across different clusters for Control and SCFP groups across time. AA, Auxiliary Activity Family; CBM, Carbohydrate-Binding Module Family; CE, Carbohydrate Esterase Family; GH, Glycoside Hydrolase Family; GT, Glycosyl Transferase Family; PL, Polysaccharide Lyase Family.*
## Correlation networks reveal that post-stress microbial communities are more stable in SCFP treated horses
Overall, a smaller number of significant interactions were observed in the SCFP group compared to Control, particularly following stress (SCFP = 198 positive interactions and 30 negative interactions: Control = 304 positive interactions and 210 negative interactions; Supplementary Table 4). Treatment with SCFP resulted in a smaller number of significant species interactions overall (maximum of 310 interactions before challenge) while the Control group had a total of 520 interactions before challenge. Horses that received SCFP had fewer interactions in total compared to Control, both pre- and post-stress. While no difference was observed in the percentage of positive interactions before stress, SCFP treated horses had a substantially higher number of positive interactions after stress when compared to untreated Control horses (87 vs. $59\%$).
## Discussion
To evaluate the potential effect of supplementing horses under stressful conditions with a postbiotic, we sequenced the fecal metagenomes of 20 horses undergoing a previously established stress model that mimics prolonged transportation. The rationale that SCFP supplementation could lead to improved microbiome stability is based on recent reports of SCFP having a positive impact in other species undergoing stressful conditions (22, 24, 60–62). Here, we observed that untreated Control horses and treated (SCFP) horses presented very different microbiome trajectories upon stress, both in within- and between-sample diversity measurements. Moreover, a lower magnitude of changes was observed in the functional potential and microbial profile of SCFP horses vs. Control. Taken together, these observations led us to conclude that treatment with SCFP resulted in more robust and stable microbial profiles in horses after stress challenge.
Less variation in microbial and functional profiles were observed for SCFP compared to Control horses. This was noted in several of our analyses including Shannon diversity index, total number of microbial network interactions, percentage of positive network interactions, and microbial and functional clustering profiles, which was illustrated in heatmaps with Control horses having a higher degree of change than SCFP treated horses. These data led us to conclude that dietary SCFP supplementation results in a more stable and robust community that is less impacted by stress. Our findings are in agreement with Tun et al. [ 24] who observed that postbiotic treatment tends to stabilize the microbiota of cows in a subclinical acidosis challenge. In that study authors concluded that SCFP supplementation attenuated the impacts of subacute ruminal acidosis on the composition and functionality of the rumen microbiome. Taken together, our results and those from others leads us to hypothesize that SCFP treatment results in a microbial community that is more robust (defined as resistance against change) in responding to stress. This hypothesis is corroborated by observations in our correlation network analyses, where the number of connections generally decreased with SCFP, but the percentage of positive interactions increased, indicating that a leaner (defined as a community with fewer connections among members but with a larger number of positive connections) and more connected microbial community in SCFP horses.
Our results are also in agreement with other studies that demonstrated that postbiotic supplementation is associated with improved microbiome balance which has been translated into host health in many species [23, 60, 62, 63]. In horses, postbiotic supplementation resulted in increased relative abundances of fibrolytic bacteria [64] and attenuated exercise-induced stress markers [28]. Additionally, Lucassen et al. showed that postbiotic-treated horses have more efficient response to vaccination [23]. In contrast, a recent study by the same group evaluating SCFP supplementation did not identify significant alterations in the fecal microbiota of thoroughbred racehorses [27]. Possible explanations for this disparity are the use of a lower resolution technique (16S rRNA gene sequencing) and the smaller dataset (11 horses) in that study when compared to our study, which used shotgun metagenomics to analyze the fecal microbiome of 20 horses. Additionally, those authors reported a high degree of horse-dependent effects of treatment, which can be attributed to high horse-to-horse variability.
The potential effects of stress and SCFP treatment on CAZy families was evaluated due to the importance and dependence of the horse on the degradation of structural carbohydrates of forages by gut microbiome for health and wellbeing. We observed that SCFP stabilized the composition and functionality of the hindgut microbial community. This was observed particularly immediately after stress relief (0 h) where lower CAZy abundance was observed in Control horses (light green in most cases) while abundances remained relatively unchanged or increased in SCFP (with the exception of one horse). At 12 h post-stress, Control horses displayed a dramatic switch in functional profile, with most showing increased relative abundances as illustrated in the heat maps in Figure 6. The small sample size of this study precludes us from making further statements regarding the functional potential, but what is evident from this study is that larger swings in relative abundances of CAZy families were associated with Control horses when compared to treated horses throughout the entire post-stress period. Additionally, a strong horse-to-horse effect was observed, indicating that treatment effect is highly dependent on the animal. These findings are similar to those of Lucassen et al. [ 27] who observed a high degree of horse-to-horse variability in their study of the equine microbiome of horses fed a postbiotic.
The bacteria identified in our study are in agreement with previous reports of healthy equine gut microbiomes, with a composition that is mainly dominated by fibrolytic bacteria [8, 9, 65]. It is important to highlight that we chose a very strict threshold for taxa selection for statistical comparisons between treatment groups. Specifically, to be included in the statistical analysis, a microbial species had to be present in at least $75\%$ of all samples. This was a deliberate choice to decrease the chances for spurious findings due to multiple hypothesis testing.
Here, we identified that SCFP treatment significantly impacted the relative abundance of Erysipelotrichaceae before stress, with a small, but significant increase in SCFP treated horses compared to Control horses. Biddle and colleagues also observed significant temporal changes in Erysipelotrichaceae in obese horses [66], and this family had previously been identified as part of the core microbial community of horse feces [67]. However, little is known about the role of this species in the horse gut and diverging evidence has been presented about the role of Erysipelotrichaceae in other organisms, with varying levels of Erysipelotrichaceae reported in murine and human studies of disease [12, 68].
Immediately after stress 11 bacteria were identified to be significantly different albeit with very small effect sizes. From those, eight were increased in the SCFP group and three were increased in the Control group, with relatively higher effect sizes when compared to species increased in SCFP horses. Microorganisms significantly increased in Control horses included one uncultured Butyrivibrio species, Pseudobutirivibrio ruminis, and Ryzophagus irregularis. We identified Ryzophagus irregularis, an arbuscular mycorrhizal fungus that is common in plants, and which [69] has not been previously reported in the horse gastrointestinal tract. Given the presence of this organism in many plant species, and the plant-based diet of horses, this finding is not completely surprising.
A larger number of significantly different species were observed at 12 h post stress. Out of 18 significantly different species, six had relatively high effect sizes and were increased in SCFP horses. These included three Butyrivibrio species in addition to Blautia, Acetivibrio, and Methanobrevibacter, which were found to have significantly higher relative abundances in SCFP treated horses 12 h after stress. Butyrivibrio are very versatile bacteria and encode a variety of enzymes to hydrolyze complex carbohydrates [70, 71]. They have been reported to carry many genes encoding glycoside hydrolases (GH) that are involved in carbohydrate fermentation and butyrate production. Likewise, Blautia and Acetivibrio are also fiber fermenters [8]. In agreement with increases with Butyrivibrio species, in our functional annotation analyses, 36 out of 62 enzymes found to be significantly different in the present study encode for glycoside hydrolases. Lastly, Methanobrevibacter was also identified to be increased in the SCFP group at 12 and 24 h post stress. The presence of methanogenic archaea in the horse gut has been previously reported, and the diversity of methane producers in the horse gut is believed to be high [8, 72].
Despite the wealth of data collected as part of this project we acknowledge that a sample size of 20 horses is relatively small. It was further substantiated that horse intrinsic factors impacted the response to stress or treatment, as it could be observed by single animals behaving differently than the remainder of the group at a given timepoint. In fact, horse-to-horse variability has been well-documented in immune parameters in horses subjected to this model of stress [33]. Raidal et al. observed varying degrees of change in white cell count, neutrophil count, and total bacterial numbers in six horses subjected to prolonged head elevation [33]. This added variability might have confounded our analyses and precluded us from identifying strong signals. Nevertheless, animals are different and thus further research should account for animal-to-animal changes and perhaps quantify the effect in terms of microbiome changes within an animal. Even with the relatively high heterogeneity of this dataset we were able to identify a clear overall signal that treatment with SCFP tends to promote microbiome robustness and stability after stress, as we observed in measures of alpha and beta diversity, as well as bacterial and functional profiles, and bacteria interaction dynamics.
This study adds to the body of knowledge regarding the beneficial impacts of postbiotic administration to horses undergoing stressful situations. While the specific mechanisms by which this robustness and stability are imparted in an equine's gut microbiome by postbiotic administration are not fully elucidated, studies in other species suggested that potential underlying mechanisms by which postbiotic supplementation led to improved health include effects in immunomodulatory pathways [73] and improved microbiome composition and functionality [24]. From an immune perspective, animals receiving SCFP seem to be primed to respond with elevated (magnitude of response) and accelerated (speed of response) cytokine production when a threat is detected [61, 62]. Additionally, at the site of challenge, increased phagocytic activity and killing ability of white blood cells and reduced activation of inflammatory system leads to a reduction in localized inflammation, and potentially immunopathology in SCFP supplemented animals [62, 74]. From the microbiome perspective, ruminant studies have shown that SCFP supplementation boosts the abundances of influential members of the microbiome which promote richness and diversity, and hence, functionality of the microbiome resulting in increased VFA production and improved energetic efficiency of rumen fermentation [24, 75]. Therefore, it can be speculated that the dual action of SCFP postbiotic via immunomodulatory pathways and optimized microbiome functionality increases robustness of animals against a wide range of infectious and metabolic stressors.
Our results indicate that prophylactic supplementation with a yeast-derived postbiotic might be a beneficial strategy for horses prior to exposure to stress. This exploratory study is limited in the ability to draw mechanistic conclusions on the effects of SCFP in horses subjected to a stress model. We observed a lower degree of change both in microbial diversity and functional profile of horses fed SCFP when compared to Control. Mechanistically, having a more robust and stable microbiome plausibly results in less opportunity for pathogen colonization and better health maintenance. Postbiotics have been demonstrated to have positive impacts in several species, and further research into the mechanisms by which these beneficial effects occur is warranted.
## Data availability statement
The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found at: https://www.ncbi.nlm.nih.gov/, bioproject PRJNA788958.
## Ethics statement
The animal study was reviewed and approved by Institutional Animal Care and Use Committee at the University of Florida in Gainesville, FL (#201810324).
## Author contributions
LW designed the animal experiment. EK and BK designed the microbiome study. MT performed the animal experiment and collected samples. MS performed sequencing and bioinformatics. AC performed statistical analyses. EG, EK, and SN interpreted the data. EG and EK prepared the first draft. All authors read and approved the final manuscript.
## Conflict of interest
AC, MS, BK, SN, and EK were employed by the company Cargill Inc. The remaining 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/fvets.2023.1134092/full#supplementary-material
## References
1. Dini-Andreote F, van Elsas JD, Olff H, Salles JF. **Dispersal-competition tradeoff in microbiomes in the quest for land colonization**. *Sci Rep* (2018) **8** 6. DOI: 10.1038/s41598-018-27783-6
2. Forsberg KJ, Reyes A, Wang B, Selleck EM, Sommer MOA, Dantas G. **The shared antibiotic resistome of soil bacteria and human pathogens**. *Science.* (2012) **337** 1107-11. DOI: 10.1126/science.1220761
3. Leung H, Long X, Ni Y, Qian L, Nychas E, Siliceo SL. **Risk assessment with gut microbiome and metabolite markers in NAFLD development**. *Sci Transl Med.* (2022) **14** 855. DOI: 10.1126/scitranslmed.abk0855
4. Bäckhed F, Ding H, Wang T, Hooper L, Gou YK, Nagy A. **The gut microbiota as an environmental factor that regulates fat storage**. *Proc Natl Acad Sci USA.* (2004) **101** 15718-23. DOI: 10.1073/pnas.0407076101
5. Jacobson A, Lam L, Rajendram M, Tamburini F, Honeycutt J, Pham T. **A gut commensal-produced metabolite mediates colonization resistance to salmonella infection**. *Cell Host Microbe* (2018) **24** 296-307. DOI: 10.1016/j.chom.2018.07.002
6. Borda-Molina D, Vital M, Sommerfeld V, Rodehutscord M, Camarinha-Silva A. **Insights into broilers' gut microbiota fed with phosphorus, calcium, and phytase supplemented diets**. *Front Microbiol.* (2016) **2016** 2033. DOI: 10.3389/fmicb.2016.02033
7. Cangiano LR, Yohe TT, Steele MA, Renaud DL. **Invited review: Strategic use of microbial-based probiotics and prebiotics in dairy calf rearing**. *Appl Anim Sci* (2020) **36** 630-51. DOI: 10.15232/aas.2020-02049
8. Kauter A, Epping L, Semmler T, Antao EM, Kannapin D, Stoeckle SD. **The gut microbiome of horses: Current research on equine enteral microbiota and future perspectives**. *Anim Microbiome.* (2019) **1** 14. DOI: 10.1186/s42523-019-0013-3
9. Morrison PK, Newbold CJ, Jones E, Worgan HJ, Grove-White DH, Dugdale AH. **The equine gastrointestinal microbiome: Impacts of weight-loss**. *BMC Vet Res.* (2020) **16** 1-18. DOI: 10.1186/s12917-020-02295-6
10. Destrez A, Grimm P, Julliand V. **Dietary-induced modulation of the hindgut microbiota is related to behavioral responses during stressful events in horses**. *Physiol Behav.* (2019) **202** 94-100. DOI: 10.1016/j.physbeh.2019.02.003
11. Plancade S, Clark A, Philippe C, Helbling JC, Moisan MP, Esquerré D. **Unraveling the effects of the gut microbiota composition and function on horse endurance physiology**. *Sci Rep* (2019) **9** 7. DOI: 10.1038/s41598-019-46118-7
12. Mach N, Ruet A, Clark A, Bars-Cortina D, Ramayo-Caldas Y, Crisci E. **Priming for welfare: Gut microbiota is associated with equitation conditions and behavior in horse athletes**. *Sci Rep.* (2020) **10** 1-19. DOI: 10.1038/s41598-020-65444-9
13. Litvak Y, Bäumler AJ. **The founder hypothesis: A basis for microbiota resistance, diversity in taxa carriage, and colonization resistance against pathogens**. *PLoS Pathog.* (2019) **15** e1007563. DOI: 10.1371/journal.ppat.1007563
14. Ranallo RT, Mcdonald LC, Halpin AL, Hiltke T, Young VB. **The state of microbiome science at the intersection of infectious diseases and antimicrobial resistance**. *J Infect Dis* (2021) **223** S187-93. DOI: 10.1093/infdis/jiab020
15. McLaren MR, Callahan BJ. **Pathogen resistance may be the principal evolutionary advantage provided by the microbiome**. *Philos Trans Royal Soc B* (2020) **375** 592. DOI: 10.1098/rstb.2019.0592
16. Lozupone CA, Stombaugh JI, Gordon JI, Jansson JK, Knight R. **Diversity, stability and resilience of the human gut microbiota**. *Nature.* (2012) **489** 220. DOI: 10.1038/nature11550
17. Salminen S, Collado MC, Endo A, Hill C, Lebeer S, Quigley EMM. **The International Scientific Association of Probiotics and Prebiotics (ISAPP) consensus statement on the definition and scope of postbiotics**. *Nat Rev Gastroenterol Hepatol.* (2021) **18** 649-67. DOI: 10.1038/s41575-021-00440-6
18. Gibson GR, Hutkins R, Sanders ME, Prescott SL, Reimer RA, Salminen SJ. **Expert consensus document: The International Scientific Association for Probiotics and Prebiotics (ISAPP) consensus statement on the definition and scope of prebiotics**. *Nat Rev Gastroenterol Hepatol.* (2017) **14** 491-502. DOI: 10.1038/nrgastro.2017.75
19. Zółkiewicz J, Marzec A, Ruszczyński M, Feleszko W. **Postbiotics-A step beyond pre- and probiotics**. *Nutrients.* (2020) **12** 1-17. DOI: 10.3390/nu12082189
20. Magalhães VJA, Susca F, Lima FS, Branco AF, Yoon I, Santos JEP. **Effect of feeding yeast culture on performance, health, and immunocompetence of dairy calves**. *J Dairy Sci.* (2008) **91** 1497-509. DOI: 10.3168/jds.2007-0582
21. Guo J, Xu L, Khalouei H, Fehr K, Senaratne V, Ghia JE. **Saccharomyces cerevisiae fermentation products reduce bacterial endotoxin concentrations and inflammation during grain-based subacute ruminal acidosis in lactating dairy cows**. *J Dairy Sci.* (2022) **105** 2354-68. DOI: 10.3168/jds.2021-20572
22. Chou WK, Park J, Carey JB, McIntyre DR, Berghman LR. **Immunomodulatory effects of**. *Front Vet Sci* (2017) **4** 37. DOI: 10.3389/fvets.2017.00037
23. Lucassen A, Finkler-Schade C, Schuberth HJ. **A**. *Animals* (2021) **11** 92726. DOI: 10.3390/ani11092726
24. Tun HM, Li S, Yoon I, Meale SJ, Azevedo PA, Khafipour E. *BMC Vet Res.* (2020) **16** 1-17. DOI: 10.1186/s12917-020-02437-w
25. Moyad MA, Robinson LE, Kittelsrud JM, Reeves SG, Weaver SE, Guzman AI. **Immunogenic yeast-based fermentation product reduces allergic rhinitis-induced nasal congestion: A randomized, double-blind, placebo-controlled trial**. *Adv Ther.* (2009) **26** 795-804. DOI: 10.1007/s12325-009-0057-y
26. van Dorland HA, Zanoni R, Gerber V, Jeannerat E, Wiederkehr D, Burger D. **Antibody response to influenza booster vaccination in franches-montagnes stallions supplemented with equi-strath**. *Vet Med Sci.* (2018) **4** 133-9. DOI: 10.1002/vms3.95
27. Lucassen A, Hankel J, Finkler-Schade C, Osbelt L, Strowig T, Visscher C. **Feeding a**. *Animals.* (2022) **12** 1496. DOI: 10.3390/ani12121496
28. Valigura HC, Leatherwood JL, Martinez RE, Norton SA, White-Springer SH. **Dietary supplementation of a**. *J Anim Sci* (2021) **99** skab199. DOI: 10.1093/jas/skab199
29. 29.Tench. Effect of Saccharomyces Cerevisiae Fermentate on Mucosal Immunity in Young Stress-Challenged Horse. UF Digital Collections. (2022). Available online at: https://ufdc.ufl.edu/UFE0056169/00001/pdf (accessed December 24, 2022).. *Effect of Saccharomyces Cerevisiae Fermentate on Mucosal Immunity in Young Stress-Challenged Horse* (2022)
30. 30.ADSA. Guide for the Care and Use of Agricultural Animals in Research and Teaching. 4th ed. (2020). Available online at: www.adsa.org (accessed December 28, 2022).. *Guide for the Care and Use of Agricultural Animals in Research and Teaching* (2020)
31. 31.National Research Council. Nutrient Requirements of Horses: Sixth Revised Edition.
Washington, DC: The National Academies Press (2007). 10.17226/11653. *Nutrient Requirements of Horses: Sixth Revised Edition.* (2007). DOI: 10.17226/11653
32. Bobel JM, Long MT, Warren LK. **119 Use of a head elevation and nasopharyngeal flush model for short-term immune distress in horses**. *J Equine Vet Sci.* (2015) **35** 433-4. DOI: 10.1016/j.jevs.2015.03.130
33. Raidal SL, Love DN, Bailey GD. **Effects of posture and accumulated airway secretions on tracheal mucociliary transport in the horse**. *Aust Vet J.* (1996) **73** 45-9. DOI: 10.1111/j.1751-0813.1996.tb09963.x
34. Johnson AC, Biddle AS, Tench M, Bobel JM, Bazurto C, Hansen TL. **198 effect of**. *J Anim Sci* (2021) **99** 106-7. DOI: 10.1093/jas/skab235.193
35. Tench M, Bobel JM, Bazurto C, Dolly J, Hansen TL, Kirk N. **40 dietary**. *J Equine Vet Sci.* (2021) **100** 103503. DOI: 10.1016/j.jevs.2021.103503
36. Li H. **Minimap2: Pairwise alignment for nucleotide sequences**. *Bioinformatics.* (2018) **34** 3094-100. DOI: 10.1093/bioinformatics/bty191
37. Li H, Handsaker B, Wysoker A, Fennell T, Ruan J, Homer N. **The sequence alignment/map format and SAMtools**. *Bioinformatics.* (2009) **25** 2078-9. DOI: 10.1093/bioinformatics/btp352
38. Tatusova T, Ciufo S, Fedorov B, O'Neill K, Tolstoy I. **RefSeq microbial genomes database: New representation and annotation strategy**. *Nucleic Acids Res.* (2015) **43** 3872. DOI: 10.1093/nar/gkv278
39. Almeida A, Nayfach S, Boland M, Strozzi F, Beracochea M, Shi ZJ. **A unified catalog of 204,938 reference genomes from the human gut microbiome**. *Nat Biotechnol.* (2021) **39** 105-14. DOI: 10.1038/s41587-020-0603-3
40. Stewart RD, Auffret MD, Warr A, Walker AW, Roehe R, Watson M. **Compendium of 4,941 rumen metagenome-assembled genomes for rumen microbiome biology and enzyme discovery**. *Nat Biotechnol.* (2019) **37** 953-61. DOI: 10.1038/s41587-019-0202-3
41. Wood DE, Lu J, Langmead B. **Improved metagenomic analysis with Kraken 2**. *Genome Biol* (2019) **20** 1891. DOI: 10.1186/s13059-019-1891-0
42. Lu J, Breitwieser FP, Thielen P, Salzberg SL. **Bracken: Estimating species abundance in metagenomics data**. *PeerJ Comput Sci.* (2017). DOI: 10.1101/051813
43. 43.R Core Development Team. R: A Language Environment for Statistical Computing, 3.2.1. (2015). Available online at: http://www.r-project.org (accessed November 22, 2021).. *R: A Language Environment for Statistical Computing, 3.2.1* (2015)
44. McMurdie PJ, Holmes S. **phyloseq: An R package for reproducible interactive analysis and graphics of microbiome census data**. *PLoS ONE.* (2013) **8** e61217. DOI: 10.1371/journal.pone.0061217
45. Seemann T. **Prokka: Rapid prokaryotic genome annotation**. *Bioinformatics.* (2014) **30** 2068-9. DOI: 10.1093/bioinformatics/btu153
46. Lahti L. *Tools for Microbiome Analysis in R* (2023)
47. Palarea-Albaladejo J, Martín-Fernández JA. **ZCompositions—R package for multivariate imputation of left-censored data under a compositional approach**. *Chemometr Intell Lab Syst.* (2015) **143** 85-96. DOI: 10.1016/j.chemolab.2015.02.019
48. Lombard V, Golaconda Ramulu H, Drula E, Coutinho PM, Henrissat B. **The carbohydrate-active enzymes database (CAZy) in 2013**. *Nucleic Acids Res* (2014) **42** gkt1178. DOI: 10.1093/nar/gkt1178
49. Cantalapiedra CP, Hernández-Plaza A, Letunic I, Bork P, Huerta-Cepas J. **eggNOG-mapper v2: Functional annotation, orthology assignments, and domain prediction at the metagenomic scale**. *Mol Biol Evol* (2021) **6** 3. DOI: 10.1101/2021.06.03.446934
50. Bates D, Mächler M, Bolker BM, Walker SC. **Fitting linear mixed-effects models using lme4**. *J Stat Softw.* (2015) **67** 1-48. DOI: 10.18637/jss.v067.i01
51. Oksanen J, Kindt R, Legendre P, O'Hara B, Simpson GL, Solymos P. *vegan: Community Ecology Package* (2008)
52. Aitchison J, Barceló-Vidal C, Martín-Fernández JA, Pawlowsky-Glahn V. **Logratio analysis and compositional distance**. *Math Geol.* (2000) **32** 271-5. DOI: 10.1023/A:1007529726302
53. Zhou H, He K, Chen J, Zhang X. **LinDA: Linear models for differential abundance analysis of microbiome compositional data**. *Genome Biol.* (2022) **23** 1-23. DOI: 10.1186/s13059-022-02655-5
54. Russell L, Singmann H, Love J, Buerkner PMH. *Emmeans: Estimated Marginal Means* (2020)
55. Benjamini Y, Yekutieli D. **The control of the false discovery rate in multiple testing under dependency**. *Ann Statist* (2001) **29** 1165-88. DOI: 10.1214/aos/1013699998
56. Benjamini Y, Hochberg Y. **Controlling the false discovery rate: A practical and powerful approach to multiple testing**. *J Roy Stat Soc B.* (1995) **57** 289-300. DOI: 10.1111/j.2517-6161.1995.tb02031.x
57. Faust K, Raes J. **CoNet app: Inference of biological association networks using Cytoscape**. *F1000Res* (2016) **5** 1519. DOI: 10.12688/f1000research.9050.2
58. Tibshirani R, Walther G, Hastie T. **Estimating the number of clusters in a data set**. *J R Stat Soc Series B Stat Methodol.* (2001) **63** 411-23. DOI: 10.1111/1467-9868.00293
59. 59.MATLAB. Version R2019B. Natick, MA: The MathWorks Inc. (2019).. *Version R2019B* (2019)
60. Al-Qaisi M, Horst EA, Mayorga EJ, Goetz BM, Abeyta MA, Yoon I. **Effects of a**. *J Dairy Sci.* (2020) **103** 9634-45. DOI: 10.3168/jds.2020-18721
61. Klopp RN, Yoon I, Eicher S, Boerman JP. **Effects of feeding Saccharomyces cerevisiae fermentation products on the health of Holstein dairy calves following a lipopolysaccharide challenge**. *J Dairy Sci.* (2022) **105** 1469-79. DOI: 10.3168/jds.2021-20341
62. Mahmoud AHA, Slate JR, Hong S, Yoon I, McGill JL. **Supplementing a**. *J Anim Sci* (2020) **98** skaa252. DOI: 10.1093/jas/skaa252
63. Gingerich E, Frana T, Logue CM, Smith DP, Pavlidis HO, Chaney WE. **Effect of feeding a postbiotic derived from saccharomyces cerevisiae fermentation as a preharvest food safety hurdle for reducing**. *J Food Prot.* (2021) **84** 275-80. DOI: 10.4315/JFP-20-330
64. Garber A, Hastie PM, Farci V, McGuinness D, Bulmer L, Alzahal O. **The effect of supplementing pony diets with yeast on 2. The faecal microbiome**. *Animal.* (2020) **14** 2493-502. DOI: 10.1017/S1751731120001512
65. Gilroy R, Leng J, Ravi A, Adriaenssens EM, Oren A, Baker D. **Metagenomic investigation of the equine faecal microbiome reveals extensive taxonomic diversity**. *PeerJ* (2022) **10** 13084. DOI: 10.7717/peerj.13084
66. Biddle AS, Tomb JF, Fan Z. **Microbiome and blood analyte differences point to community and metabolic signatures in lean and obese horses**. *Front Vet Sci* (2018) **5** 225. DOI: 10.3389/fvets.2018.00225
67. Dougal K, de la Fuente G, Harris PA, Girdwood SE, Pinloche E, Newbold CJ. **Identification of a core bacterial community within the large intestine of the horse**. *PLoS ONE.* (2013) **8** e77660. DOI: 10.1371/journal.pone.0077660
68. Kaakoush NO. **Insights into the role of**. *Front Cell Infect Microbiol.* (2015) **5** 84. DOI: 10.3389/fcimb.2015.00084
69. Krüger M, Krüger C, Walker C, Stockinger H, Schüßler A. **Phylogenetic reference data for systematics and phylotaxonomy of arbuscular mycorrhizal fungi from phylum to species level**. *New Phytol.* (2012) **193** 970-84. DOI: 10.1111/j.1469-8137.2011.03962.x
70. Kopečný J, Zorec M, Mrázek J, Kobayashi Y, Marinšek-Logar R. *Int J Syst Evol Microbiol.* (2003) **53** 201-9. DOI: 10.1099/ijs.0.02345-0
71. Bryant MP, Small N. **The anaerobic monotrichous butyric acid-producing curved rod-shaped bacteria of the rumen**. *J Bacteriol.* (1956) **72** 16-21. DOI: 10.1128/jb.72.1.16-21.1956
72. Elghandour MMMY, Adegbeye MJ, Barbabosa-Pilego A, Perez NR, Hernández SR, Zaragoza-Bastida A. **Equine contribution in methane emission and its mitigation strategies**. *J Equine Vet Sci.* (2019) **72** 56-63. DOI: 10.1016/j.jevs.2018.10.020
73. Jastrzab R, Graczyk D, Siedlecki P. **Molecular and cellular mechanisms influenced by postbiotics**. *Int J Mol Sci* (2021) **22** 2413475. DOI: 10.3390/ijms222413475
74. Vailati-Riboni M, Coleman DN, Lopreiato V, Alharthi A, Bucktrout RE, Abdel-Hamied E. **Feeding a**. *J Anim Sci Biotechnol.* (2021) **12** 1-19. DOI: 10.1186/s40104-021-00560-8
75. Guo J, Zhang T, Yoon I, Khafipour E, Plaizier JC. **236 effects of**. *J Anim Sci* (2021) **99** 175-6. DOI: 10.1093/jas/skab235.321
|
---
title: 'RETRACTED: High glucose induced HIF-1α/TREK1 expression and myometrium relaxation
during pregnancy'
authors:
- Tengteng Li
- Jiajia Fei
- Huihui Yu
- Xingxing Wang
- Jingjing Bai
- Fucai Chen
- Dan Li
- Zongzhi Yin
journal: Frontiers in Endocrinology
year: 2023
pmcid: PMC9998977
doi: 10.3389/fendo.2023.1115619
license: CC BY 4.0
---
# RETRACTED: High glucose induced HIF-1α/TREK1 expression and myometrium relaxation during pregnancy
## Abstract
### Background
The incidence of gestational diabetes mellitus (GDM) is increasing worldwide. GDM patients have a significantly higher rate of cesarean section and postpartum hemorrhage, suggesting changes in uterine contractility. TWIK-1-related potassium channel (TREK1) expressed in the pregnant uterus and its role in uterine contraction. In this study, we examined the expression of HIF-1α and TREK1 proteins in GDM uterine and investigated whether high glucose levels are involved in the regulation of human uterine smooth muscle cells (HUSMCs) contraction through TREK1, and verified the role of HIF-1α in this process.
### Methods
Compared the uterine contractility between GDM and normal patients undergoing elective lower segment cesarean section. The HUSMCs were divided into normal glucose group, high glucose group, normal glucose with CoCl2 group, CoCl2 with echinomycin/L-Methionine group, and high glucose with echinomycin/L-Methionine group; Compare the cell contractility of each group. Compared the expression of hypoxia-inducible factor-1α (HIF-1α) and TREK1 protein in each group.
### Results
The contractility of human uterine strips induced by both KCl and oxytocin was significantly lower in patients with GDM compared with that in normal individuals, with increased TREK1 and HIF-1α protein expression. The contractility of cultured HUSMCs was significantly decreased under high glucose levels, which was consistent with increased expression of HIF-1α and TREK1 proteins. The contractility of HUSMCs was decreased when hypoxia was induced by CoCl2 and increased when hypoxia was inhibited by echinomycin. The TREK1 inhibitor L-methionine also recovered the decreased contractility of HUSMCs under high glucose levels or hypoxia.
### Discussion
The high glucose levels decreased the contractility of the myometrium, and increased expression of HIF-1a and TREK1 proteins play a role in changes in uterus contractility.
## Introduction
Gestational diabetes mellitus (GDM) is defined as diabetes diagnosed in the second or third trimester of pregnancy, which is not overt diabetes prior to gestation [1]. GDM is one of the most common pregnancy complications, and the prevalence of GDM is increasing globally [2].
Several studies have shown that the rate of cesarean section and postpartum hemorrhage in patients with GDM is significantly higher than that in women without GDM (3–6), and more than $30\%$ of cesarean sections in patients with GDM are due to failed vaginal birth [7]. Failed vaginal delivery trial and postpartum hemorrhage are associated with uterine atony [8, 9], suggesting that GDM affects uterine contractility. An in vitro trial of isolated uteri in patients with diabetes confirmed that high glucose levels caused poor uterine contractility [10]. In a non-pregnant murine model of non-obese type-1 diabetes mellitus, isolated uteri isometric contraction showed a significant reduction in spontaneous motility and hypo-contractility compared with controls [11]. Telma et al. [ 12] found that the myometrium had a reduced number and irregular arrangement of myogenic fibers and decreased contractility in pregnant female rats with diabetes compared with controls. These results indicate that hyperglycemia may affect myometrial contractility of pregnant and nonpregnant uteri. Although these studies indicate that diabetes reduces uterine contractility, the underlying mechanisms remain unclear.
Uterine smooth muscle cells are stimulated by specific signals to cause membrane depolarization and generate action potentials, which trigger electro-mechanical coupling, leading to cell contraction. The TWIK-1-related potassium channel (TREK1) is a double-pore potassium channel protein expressed in human uterine tissue, which plays an important role in potassium efflux and maintaining the resting potential of smooth muscle cells; TREK1 is regulated by temperature, PH, stretching, arachidonic acid, L-methionine, progesterone, and other factors [13, 14]. We previously demonstrated that the incubation of human pregnant myometrial strips with the TREK1 activator arachidonic acid significantly reduced myometrial strip contractility, whereas incubation with the TREK1 inhibitor L-methionine significantly enhanced myometrial strip contractility, confirming the importance of TREK1 in regulating uterine contractility (15–17). Several studies [15, 16] have shown that TREK1 expression is altered in the pregnant uterus and that this change is associated with changes in the contractility of pregnant uterine tissue [18]. However, it is unknown whether the expression of TREK1 protein changes in GDM uterine tissues and whether the glucose-related changes in myometrium contractility are also associated with TREK1.
Studies comparing patients with diabetes and individuals without diabetes have reported significantly lower contractility of the myocardium, vascular smooth muscle, and gastrointestinal smooth muscle among patients with diabetes, with an increased incidence of heart failure [19], decreased ventricular systolic function [20], impaired vascular smooth muscle contraction [21, 22], and gastroparesis [23]. Tissue or cell injury caused by diabetes was found to be associated with an increased expression of hypoxia-inducible factor (HIF) triggered by glycemia (24–27). HIFs are nuclear factors that reflect the degree of tissue or cell hypoxia and consist of alpha and beta subunits, which include mainly HIF-1 and HIF-2 [28, 29]. HIF-1α plays a central role in hypoxia and regulates many targets, promoting erythropoietin, cell proliferation, and angiogenesis for tissue or cells to respond to hypoxia (29–32). HIF-1α is rapidly degraded by hydroxylation under proline hydroxylase, and CoCl2 can prevent this hydroxylation [28, 33]. Lim et al. [ 34] showed that upregulated expression of HIF-1α in vascular smooth muscle leads to reduced contractility of vascular smooth muscle and that vascular smooth muscle strips incubated with the HIF-1α inhibitors echinocandin and U0126 restored vascular smooth muscle contractility. In uterine tissue, Alotaibi reported significantly decreased contractility in hypoxic myometrium [35]. Osman et al. [ 36] found that the uterine contractility was significantly reduced after incubation with cyanide to induce hypoxia, and the contractility was restored when the uterine was removed from the cyanide environment. Studies have shown that cyanide induced the nuclear accumulation of HIF-1α [37]. Eun et al. [ 38] found that HIF-1α mRNA and protein overexpression significantly increased TREK1 mRNA and protein expression in primary astrocytes using CoCl2. Astrocyte elevated gene-1 (AEG-1) is a major mediator of hypoxia-regulated TREK1 expression in astrocytes, and HIF-1α binds directly to the AEG-1 promoter. AEG-1 knockdown dramatically decreased the mRNA and protein levels of TREK1, suggesting that TERK1 is regulated by HIF-1α. However, the expression of HIF-1α protein in uterine tissues of GDM and the association between HIF-1α protein changes and changes in uterine contractility in patients with GDM are unclear.
In this study, we examined the expression changes of HIF-1α and TREK1 proteins in uterine tissues of patients with GDM, investigated whether high glucose levels are involved in the regulation of uterine smooth muscle contraction during pregnancy through TREK1, and verified the role of HIF-1α in this process.
## Ethics statement
The study was reviewed and approved by the First Affiliated Hospital of Anhui Medical University Ethics Committee for the Protection of Human Subjects in Research and Tissue Collection (PJ2020-06-12). Uterine tissues were collected from pregnant women undergoing elective lower segment cesarean section at the First Affiliated Hospital of Anhui Medical University, Hefei, China. All participating women provided written informed consent to participate in the study.
The studies involving human participants were reviewed and approved by First Affiliated Hospital of Anhui Medical University Ethics Committee for the Protection of Human Subjects in Research and Tissue Collection (PJ2020-06-12). The patients/participants provided their written informed consent to participate in this study.
## Tissue collection
Diagnosis of GDM was made according to the Chinese Society of Obstetrics and Gynecology and the Chinese Medical *Association consensus* [39]. A 75-g oral glucose tolerance test was performed between the 24th and 28th weeks of gestation. The values meeting the diagnostic criteria for GDM were as follows: fasting plasma glucose ≥5.1 mmol/L (92 mg/dL), 1-h plasma glucose ≥10.0 mmol/L (180 mg/dL), and 2-h plasma glucose ≥8.5 mmol/L (153 mg/dL). As previously described [40], patients aged 18 years or older with a singleton pregnancy, vertex presentation, and GDM diagnosis were included. The following exclusion criteria were applied: gestational age at birth of less than 35 weeks, patients with GDM with uncontrolled blood glucose levels, multiple pregnancies, placenta previa, scarred uterus, and medical/surgical comorbidity as an indication for cesarean section. All gestational ages were verified using the last menstrual period and confirmed using the first-trimester sonographic measurement of crown-rump length.
After safe delivery of the fetus and placenta by elective lower segment cesarean section, tissue specimens from the lower uterine margin were resected and immediately placed in a cryopreservation incubator in a refrigerated Krebs solution for transport to the laboratory. The uterine tissue was trimmed to a 7 mm × 3 mm muscle strip, and contraction of the uterine muscle strip was measured. Since oxytocin is synthesized in the decidual tissue immediately adjacent to the myometrium, it was necessary to separate the decidual tissue from the surface of the myometrial strip before measuring the contractility of the myometrium. The uterine tissue was trimmed to 3 mm × 3 mm × 3 mm segments for western blot analysis.
## Measurement of uterine contraction
The method for the measurement of uterine contraction has been previously described [15]. Briefly, the uterine muscle strips are fixed to the constant temperature bath and multi-channel physiological signal acquisition and processing system while continuously ventilated with $95\%$ O2 and $5\%$ CO2 at 37.0 ± 0.5°C. Our previous study has confirmed that uterus strip contractility performance is most appropriate at a 2-g stretch [15]. After the appearance of a stable and regular contraction curve, the strips were stimulated with 96 mM KCl and different concentrations of oxytocin (from 10-11 to 10-6 mM), and the contraction response was recorded. Finally, the uterine muscle strips were weighed at the end of the experiment for calibration.
Quantitative analysis of the contractility of the uterine muscle strips was performed by a multichannel physiological signal acquisition system (RM6240E, Chengdu, China) by calculating the area under the curve (AUC) of the contraction curve presented as AUC/g tissue weight. The AUC was measured at time 0 and was subtracted from the AUC measured after 5 min of application of KCl or each oxytocin concentration.
## Cell culture
Isolation of primary human gestational uterine smooth muscle cells was achieved by the enzymatic dispersion method. Myometrium was obtained from women after elective cesarean delivery in late pregnancy. The endometrium and epithelium were slightly scraped off the surface of the myometrium with a sterile blade, and the myometrium was then cut up with tissue scissors. The cut uterine tissue was digested using 15 ml of digestion solution, followed by shaking for 1 h at 37°C on a shaker. Then, the digested solution was filtered through a 100-μm filter to remove the tissue fragments, and the filtrate was transferred to a sterile centrifuge tube and centrifuged at 1000 rpm/min for 5 min before discarding the upper layer. The cell precipitate was resuspended in normal glucose medium, and the cell suspension was placed in a 25 cm2 cell culture flask (Corning) and incubated at 37°C, $5\%$ CO2 incubator until fusion.
## Cell contraction assay
According to the protocol (Cell Contraction Assay Kit, Cell Biolabs, San Diego), the collagen solution, 5× phosphate-buffered saline (PBS), and neutral solution were mixed and diluted in proportion and then placed on ice. The number of uterine smooth muscle cells was adjusted to 800,000 cells/100 µl cell suspension. The cell suspension and diluted collagen solution were mixed (1:4) to configure the gel, and the gel was added to a 24-well plate at 500 µl/well and then incubated for 1 h in an incubator at 37°C to allow the gel to solidify. The gel was divided into 7 groups: normal glucose group, high glucose group, normal with CoCl2 group, CoCl2 with echinomycin group, high glucose with echinomycin group, CoCl2 with L-methionine group, and high glucose with L-methionine group. After collagen solidification of the first three groups, they were cultured in the corresponding medium. The gels of the latter four groups were first incubated in a CoCl2 medium or high glucose medium for 4 h, then changed to the corresponding medium containing echinomycin or L-methionine, and incubated for 4 h. KCl and oxytocin were added to every group at 4 h after gel solidification. The gel areas of each group were observed and recorded with a camera at the time of solidification (0 h) and the addition of KCl/Oxytocin at 4 h; the areas were measured using Image J.
## Western blot analysis
Total protein was extracted with RIPA lysis buffer containing benzoyl fluoride and phosphatase inhibitors (Beyotime Biotechnology, China) from uterine muscle strips. The supernatant was collected by centrifugation at 4°C and 12,000 rpm/min for 10 min. The BCA protein assay was used to determine the total protein concentration. The supernatant and loading buffer were mixed (1:4) and heated at 100°C for 10 min. Protein homogenates were electrophoresed on $10\%$ SDS (sodium dodecyl sulfate) polyacrylamide gels and then electrophoretically transferred to polyvinylidene difluoride (PVDF) membranes. Membranes were incubated in PBS-Tween buffer containing $5\%$ skim milk for 2 h to block non-specific sites and then incubated overnight at 4°C in primary antibody solution containing the monoclonal mouse antibody to GAPDH (1:5000, Abcam, England) or the monoclonal rabbit antibody to TREK1 (1:200, Sigma-Aldrich, American) or HIF-1α (1:500, Abcam, England). The PVDF membranes were then washed 3 times in PBS-Tween for 15 min each and then incubated with horseradish peroxidase-coupled goat anti-rabbit secondary antibody (1:10,000, Abcam, England) or goat anti-mouse secondary antibody (1:10,000, Abcam, England) for 2 h. The membrane blots were washed with PBS-Tween and visualized by enhanced chemiluminescence (ECL, Biosharp, China). A band of 37 kDa for GAPDH, 110 kDa for HIF-1α, and 47 kDa for TREK1 was detected according to the corresponding protocols of the antibody products and were confirmed by the Marker. GAPDH was used as the internal control. Reaction bands corresponding to GAPDH, TREK1, and HIF-1α were analyzed with Image J.
## Solutions and drugs
The composition of all solutions in this study is summarized as follows. Normal Krebs solution contained the following (in mM): 120 NaCl, 5.9 KCl, 25 NaHCO3, 1.2 NaH2PO4, 11.5 dextrose, 2.5 CaCl2, 1.2 MgCl2 (Biosharp, China). High KCl solution (96 mM) was prepared as normal Krebs but with equimolar substitution of NaCl with KCl. Oxytocin (MedChemExpress, China) was dissolved in deionized water to prepare 10-11 to 10-6 M concentration for isometric contractions. Digestion solution was prepared: 2 mg/ml type II collagenase, 1 mg/ml BSA, and 0.5 mg/ml deoxyribonuclease I was dissolved in Dulbecco’s Modified Eagle Medium (DMEM) (Sigma-Aldrich, American). The normal glucose medium consisted of DMEM with 5.5 mmol/L glucose, $10\%$ fetal bovine serum (FBS) (Sigma-Aldrich, American), and $1\%$ penicillin/streptomycin (Gibco, American) solution. The high glucose medium consisted of DMEM with 25 mmol/L glucose, $10\%$ FBS, and $1\%$ penicillin/streptomycin solution. CoCl2 medium was prepared in normal glucose containing 200 µmol/L CoCl2 (Sigma-Aldrich, American). High glucose with echinomycin medium was prepared in high glucose containing 10 nmol/L echinomycin (MedChemExpress, China). CoCl2 with echinomycin medium was prepared in CoCl2 medium containing 10 nmol/L echinomycin. CoCl2 with L-methionine medium was 10 µmol/L L-methionine (Sigma-Aldrich, American) dissolved in CoCl2 medium. High glucose with L-methionine medium was 10 µmol/L of L-methionine dissolved in high glucose medium. High KCl (96 mM) and oxytocin (10-7 M) were dissolved in the respective medium for cell contractions.
## Statistical analysis
All data were analyzed and presented as mean ± standard error of the mean (SEM) using Prism (v.8.01; GraphPad Software, San Diego, CA), with the “n” value representing the number of subjects. For uterine contraction experiments, individual concentration-contraction curves were constructed, and sigmoidal curves were fitted to the data using the least squares method. Data were first analyzed using the analysis of variance with multiple classification criteria between the normal and high glucose groups. When a statistical difference was observed, the data were further analyzed using Bonferroni’s post-hoc test for multiple comparisons. Unpaired Student’s t-test was used for the comparison of two means. Differences were considered significant if $P \leq 0.05.$
## Uterine contractility decreased in patients with GDM
High concentrations of 96 mM KCl cause depolarization of cell membranes and stimulate Ca2+ influx through voltage-gated Ca2+ channels [41]. Normal human pregnancy uterine strips respond rapidly to KCl stimulation, with contractility peaking rapidly and then decreasing but remaining at a high level. Uterine strips in patients with GDM also respond rapidly to KCl, although the peak contractility is lower than normal. The sum of contractility produced by KCl stimulation of the uterine muscle strips at 5 min (AUC, 5 min) was also significantly lower ($P \leq 0.05$) in patients with GDM than in normal individuals (Figures 1A, C).
**Figure 1:** *Decreased uterine contractility in patients with GDM. Uterine strips from normal (Nor) individuals and patients with GDM were stimulated with 96 mM KCl (A) followed by washing 3 times in Krebs solution and then stimulated with increasing concentrations (10-11 to 10-6 M) of oxytocin (B). The contraction to KCl (C) and oxytocin (D, E) was measured as AUC/g tissue. Data are presented as means ± SEM, n = 6 to 7. * to ***** mean GDM vs. Nor and P<0.05. AUC, area under the curve; GDM, gestational diabetes mellitus; SEM, standard error of the mean.*
The contractility of uterine muscle strips in normal individuals and patients with GDM responds to oxytocin in a concentration-dependent manner, reaching a maximum at 10-7 M. The contraction curve of the muscle strips induced by oxytocin shows a cyclic oscillation, with an increase in contraction frequency and peak contractility with increasing oxytocin concentration. The peak and sum of uterine strip contractility (AUC, 5 min) induced by all concentration subgroups showed that GDM was weaker than normal (Figures 1B, D, E) ($P \leq 0.05$).
## Increased expression of HIF-1α and TREK1 protein in the uterus of pregnant women with GDM
We extracted total proteins from pregnant uterine tissues for western blot analysis and detected a band corresponding to HIF-1α at the 110 kDa and TREK1 at the 47 kDa position. Western blot analysis showed that HIF-1α ($P \leq 0.05$) and TREK1 ($P \leq 0.05$) protein levels were significantly higher in the uterine tissues of patients with GDM than in those of normal individuals (Figures 2A, B).
**Figure 2:** *Expression of HIF-1α and TREK1 protein increased in the uterine tissues of patients with GDM. Uterine tissues from normal (Nor) individuals and patients with GDM. The total protein of each group was extracted, and HIF-1α (A) and TREK1 (B) protein expression was measured by western blot analysis. Data are presented as means ± SEM, n = 6 to 7. * means GDM vs, Nor and P<0.05. GDM, gestational diabetes mellitus; HIF-1α, hypoxia-inducible factor-1 alpha; SEM, standard error of the mean; TREK1, TWIK-1-related potassium channel.*
## High glucose levels decreased cell contraction and increased protein expression of HIF-1α and TREK1 in human uterine smooth muscle cells
To further confirm the modulation of contractility of human uterine smooth muscle cells (HUSMCs) with increased glucose levels, we performed cell-collagen contraction experiments and divided the gels into the normal glucose group and the high glucose group. The total proteins of HUSMCs in the normal glucose and high glucose groups were extracted and analyzed for HIF-1α and TREK1 by western blot.
We observed the gels of both normal glucose and high glucose groups after stimulation with mM KCl or 10-7 M oxytocin for 4 h; the gel area in the normal glucose group was significantly smaller ($P \leq 0.05$) than that in the high glucose group (Figures 3A, B), indicating that the normal glucose group has greater cell contractility. Western blot analysis of HUSMCs showed that high glucose levels significantly increased ($P \leq 0.05$) HIF-1α (Figure 3C) and TREK1 (Figure 3D) protein expression.
**Figure 3:** *High glucose level decreased cell contraction and increased protein expression of HIF-1α and TREK1 in HUSMCs. HUSMCs mixed with collagen were cultured in normal glucose (5.5 mmol/L) (NG) medium and high glucose (25 mmol/L) (HG) medium in a 24-well culture plate. The medium was added after collagen polymerization (0 h). The images of collagen stimulated by KCl (A) or oxytocin (B) for 4 h. The contraction of HUSMCs was assessed by measuring the mean gel area change ([the area of KCl 4 h – the area of 0 h]/the area of 0 h and [the area of oxytocin 4 h – the area of 0 h]/the area of 0 h). The HUSMCs of each group were cultured in NG medium at the beginning, and the HG group was replaced with HG medium for 24 h when the cell density reached 60%–70%. Then, the total protein of each group was extracted, and HIF-1α (C) and TREK1 (D) protein expression was measured by western blot analysis. Three or four independent experiments were conducted, and the average of each group was calculated. Data are presented as means ± SEM, n = 4 for each group. * means HG vs. NG and P<0.05. HIF-1α, hypoxia-inducible factor-1 alpha; HUSMCs, human uterine smooth muscle cells; SEM, standard error of the mean; TREK1, TWIK-1-related potassium channel.*
## Induced cell hypoxia decreased the contraction and increased TREK1 protein expression in HUSMCs
High glucose levels cause increased HIF-1α/TREK1 expression and decreased cell contractility. CoCl2 was used to induce cell hypoxia to determine whether the decrease in cell contractility was related to hypoxia. We divided the HUSMCs into the normal glucose group and the normal glucose with CoCl2 group. In collagen contraction, following incubation with KCl or oxytocin for 4 h, the gel area of the normal glucose group was significantly smaller ($P \leq 0.05$) than that of the normal glucose with CoCl2 group (Figures 4A, B). Western blot analysis showed that normal glucose levels with CoCl2 induced cell hypoxia, which was detected as increased HIF-1α protein expression ($P \leq 0.05$) (Figure 4C). This induced hypoxia also significantly increased ($P \leq 0.05$) TREK1 protein expression (Figure 4D).
**Figure 4:** *Induced cell hypoxia decreased the contraction and increased TREK1 protein expression in HUSMCs. HUSMCs mixed with collagen were cultured in normal glucose (5.5 mmol/L) (NG) medium and normal glucose with CoCl2 medium (NG+CoCl2), which induced hypoxia in a 24-well plate. The medium was added after collagen polymerization (0 h). The images of collagen stimulated by KCl (A) or oxytocin (B) for 4 h. The contraction of HUSMCs was assessed by measuring the mean gel area change. The HUSMCs of each group were first cultured in NG medium, and the NG + CoCl2 group was replaced with NG + CoCl2 medium for 24 h when the cell density reached 60%–70%. Then, the total protein of each group was extracted, and HIF-1α (C) and TREK1 (D) protein expression was measured by western blot analysis. Data are presented as means ± SEM, n = 4 for each group. * means NG+CoCl2 vs. NG and P<0.05. HIF-1α, hypoxia-inducible factor-1 alpha; HUSMCs, human uterine smooth muscle cells; SEM, standard error of the mean; TREK1, TWIK-1-related potassium channel.*
## Hypoxia inhibition recovered the decreased cell contractility under high glucose conditions
Echinomycin was used to inhibit HIF-1α. High glucose levels induced cell hypoxia and increased HIF-1α. We used echinomycin to inhibit HIF-1α and detected KCl- or oxytocin-induced cell contraction for 4 h. We divided the HUSMCs into the normal glucose group, the high glucose group, and the high glucose with echinomycin group. The gel area was significantly smaller ($P \leq 0.05$) when echinomycin was used in the high glucose group compared with that obtained without echinomycin (Figures 5A, B), indicating a recovery of cell contractility that had decreased with high glucose levels. Echinomycin inhibited cell hypoxia, decreasing HIF-1α and TREK1 expression ($P \leq 0.05$) (Figures 5C, D).
**Figure 5:** *Hypoxia inhibition recovered the decreased cell contractility under high glucose levels. We divided HUSMCs into the normal glucose (NG, 5.5 mmol/L) group, the high glucose (HG, 25 mmol/L) group, and the high glucose with echinomycin (HG+Ech) group to inhibit hypoxia in a 24-well culture plate. The medium was added after collagen polymerization (0 h). The images of collagen stimulated by KCl (A) or oxytocin (B) for 4 h. The contraction of HUSMCs was assessed by measuring the mean gel area change. The HUSMCs of each group were cultured in NG medium at the beginning. Then, the NG medium in the HG and HG + Ech groups was replaced with HG medium for 24 h when the cell density reached 60%–70%, after which echinomycin was added to the HG + Ech group for 4 h. The total protein of each group was then extracted, and HIF-1α (C) and TREK1 (D) protein expression was measured by western blot analysis. Data are presented as means ± SEM, n = 4 for each group. * means HG vs. NG and P<0.05. # means HG+Ech vs. HG and P<0.05. HIF-1α, hypoxia-inducible factor-1 alpha; HUSMCs, human uterine smooth muscle cells; SEM, standard error of the mean; TREK1, TWIK-1-related potassium channel.*
## Hypoxia inhibition recovered the CoCl2-decreased cell contractility in HUSMCs
CoCl2-induced hypoxia decreased cell contractility and increased ($P \leq 0.05$) the HIF-1α and TREK1 expression. Echinomycin was then used to inhibit HIF-1α expression, and gels were incubated with KCl or oxytocin for 4 h. We divided the HUSMCs into the normal glucose group, the normal glucose with CoCl2 group, and the normal glucose + CoCl2 with echinomycin group. We observed that the gel area in the echinomycin group was significantly smaller ($P \leq 0.05$) than without echinomycin (Figures 6A, B), indicating that the contractility of HUSMCs was significantly recovered by echinomycin after hypoxia. The HIF-1α and TREK1 protein expression variations were consistent (Figures 6C, D), which increased ($P \leq 0.05$) under hypoxia and decreased when hypoxia was inhibited.
**Figure 6:** *Hypoxia inhibition recovered the CoCl2-decreased cell contractility in HUSMCs. We divided HUSMCs into the normal glucose (NG, 5.5 mmol/L) group, the normal glucose with CoCl2 (NG+CoCl2) group, and the normal glucose + CoCl2 with echinomycin (NG+CoCl2+Ech) group in a 24-well culture plate. The medium was added after collagen polymerization (0 h). The images of collagen stimulated by KCl (A) or oxytocin (B) for 4 h. The contraction of HUSMCs was assessed by measuring the mean gel area change. The HUSMCs of each group were cultured in NG medium at the beginning. Then, the NG medium in NG+CoCl2 and NG+CoCl2+Ech groups was replaced with NG+CoCl2 medium for 24 h when the cell density reached 60%–70%, after which echinomycin was added to the NG+CoCl2+Ech group for 4 h. The total protein of each group was then extracted, and HIF-1α (C) and TREK1 (D) protein expression was measured by western blot analysis. Data are presented as means ± SEM, n = 4 for each group. * means NG+CoCl2 vs. NG and P<0.05, # means NG+CoCl2+Ech vs. NG+CoCl2 and P<0.05. HIF-1α, hypoxia-inducible factor-1 alpha; HUSMCs, human uterine smooth muscle cells; SEM, standard error of the mean; TREK1, TWIK-1-related potassium channel.*
## TREK1 inhibitor L-methionine increased HUSMC contractility that was decreased by high glucose levels or hypoxia
High glucose levels and hypoxia induced TREK1 protein expression of HUSMCs and decreased cell contractility. To check whether the TREK1 variation contributed to contractility change, we used L-methionine to inhibit the TREK1 function. We divided the gels into the normal glucose group, the high glucose group, and the high glucose with L-methionine to check the function of TREK1 in a high glucose environment. We then divided the gels into the normal glucose group, the normal glucose with CoCl2 group, and the normal glucose group + CoCl2 with L-methionine group to check the function of TREK1 in the CoCl2 environment. After KCl or oxytocin was added to the gel for 4 h, the gel area of the high glucose with L-methionine group (Figures 7A, B) and the CoCl2 with L-methionine group (Figures 7C, D) was significantly smaller ($P \leq 0.05$) than that of the high glucose group and the normal glucose with CoCl2 group.
**Figure 7:** *TREK1 inhibitor L-methionine increased the HUSMC contractility decreased by high glucose levels or hypoxia. We divided HUSMCs into the normal glucose (NG, 5.5 mmol/L) group, high glucose (HG, 25 mmol/L) group, high glucose with L-methionine (HG+L-M) group, normal glucose with CoCl2 (NG+CoCl2) group, and normal glucose + CoCl2 with L-methionine (NG+CoCl2+L-M) group in a 24-well culture plate. The medium was added after collagen polymerization (0 h). The images of collagen stimulated by KCl (A) or oxytocin (B) for 4 h. The contraction of HUSMCs was assessed by measuring the mean gel area change. The medium was added after collagen polymerization (0 h). The images of collagen stimulated by KCl (C) or oxytocin (D) for 4 h. The contraction of HUSMCs was assessed by measuring the mean gel area change. Data are presented as means ± SEM, n = 4 for each group. (A, B) * means HG vs. NG and P<0.05. (A, B) # means P<0.05 HG+L-M vs. HG and P<0.05. (C, D) * means P<0.05 NG+CoCl2 vs. NG and P<0.05. (C, D) # means P<0.05 NG+CoCl2+L-M vs. NG+CoCl2 and P<0.05. HUSMCs, human uterine smooth muscle cells; SEM, standard error of the mean.*
## Discussion
Our study showed that 1) the hyperglycemic environment in patients with GDM causes a significant decrease in uterine contractility in late pregnancy and increases HIF-1α/TREK1 protein expression and that 2) hyperglycemia promotes hypoxia in HUSMCs, causing increased TREK1 expression and decreased HUSMCs contractility.
Patients with GDM have a significantly higher incidence of prolonged labor, cesarean section, and postpartum hemorrhage than women without GDM, and this may be associated with abnormal uterine contractility caused by hyperglycemia (3–6). However, previous studies (11, 42–46) on the regulation of uterine contractility by diabetes have reported inconsistent findings. In diabetic animal models, uterine tissues of non-pregnant diabetic rats or mice induced by KCl or oxytocin produced significantly weaker contractility than normal rats or mice [11, 42, 43]. The differences in contractility between diabetic pregnant and normal rats vary considerably during different gestation periods. The contractility of uterine tissue in diabetic rats did not differ from normal rats at 22 days of gestation [44, 45], significantly increased at 15 days of gestation [45], and significantly decreased at 10 days of gestation [46]. The differences in uterine contractility changes in diabetic animal models have been observed in isolated human uterine tissues by in vitro experiments. Sarioglu et al. [ 47] found no difference in spontaneous uterine contractility between patients with GDM and normal individuals, but the study did not compare uterotonic-induced uterine contractions. Al-Qahtani et al. [ 10] found that spontaneous, KCl-induced, and oxytocin-induced uterine contractility were significantly weaker in patients with GDM and concluded that reduced uterine contractility was associated with reduced calcium channel expression, intracellular calcium signaling, and decreased muscle mass. These results are consistent with our present data. In 96 mM KCl- and 10-7 M oxytocin-induced uterine contractions, the contractility was significantly weaker in patients with GDM than in normal individuals. We also found that HUSMCs contractility decreased in an environment with high glucose levels, showing consistent outcomes with those observed in GDM uterine tissues.
Previous studies [15, 48] have shown that uterine tissue contractility is regulated by the potassium channel TREK1, and high TREK1 expression in pregnant uterine tissues can cause uterine tissue diastole. However, it is not clear whether decreased uterine contractility in patients with GDM is caused by TREK1 and hyperglycemia. This study found that TREK1 protein expression in uterine smooth muscle was significantly higher in patients with GDM than in normal individuals, while TREK1 protein expression was also significantly higher in HUSMCs in the high glucose group. When TREK1 function was inhibited with L-methionine, HUSMC contractility was restored significantly. These results indicate that high glucose levels promote the expression of TREK1 proteins in uterine smooth muscle, which leads to reduced contractility of uterine smooth muscle.
Hyperglycemia can lead to hypoxia in various tissues or cells, and HIF-1α is considered a marker closely associated with hypoxia [28, 29]. There is inconsistent evidence on the effects of hyperglycemia on HIF-1α in different tissues or cells, with some studies showing that hyperglycemia promotes HIF-1α expression and others reporting its role in HIF-1α degradation in some tissues (49–53). We found that HIF-1α protein expression was significantly increased in uterine tissues of patients with GDM compared with those of normal individuals. Moreover, HIF-1α protein expression in HUSMCs was also significantly increased in the high glucose group. When CoCl2 was used to simulate HUSMC hypoxia, cell contractility was also reduced, as in the high glucose group. This finding is consistent with the observation by Alotaibi et al. [ 35] of a significant decrease in uterine strip contractility during hypoxia. However, whether the TREK1 protein, which regulates cell contractility, also functions during hypoxia-induced reduction of contractility in uterine smooth muscle cells remains unclear.
TREK1 expression was found to be significantly elevated in astrocytes after ischemia and hypoxia [38]. However, this finding did not establish whether hyperglycemia-induced HIF-1α expression in the uteri of patients with GDM also regulates uterine tissue contractility by modulating TREK1 expression. To determine whether TREK1 is regulated by HIF-1α in uterine tissue, we inhibited hypoxia and investigated TREK1 expression. Our results showed that both HIF-1α and TREK1 protein expression in HUSMCs were significantly decreased by echinomycin. This demonstrates that TREK1 is regulated by HIF-1α in HUSMCs and that HIF-1α modulating TREK1 protein expression may contribute to changes in uterine contractility.
The collagen gel contraction assay was used to verify cell contractility; the data indicated that hypoxia decreased cell contractility and that the decrease was recovered by the TREK1 inhibitor L-methionine. The induced hypoxia was associated with the changes in contractility observed with high glucose levels. The cell contraction data confirmed that high glucose induced myometrium hypoxia and decreased contractility through TREK1. Interestingly, a recent study [54] reported that hypoxia and increased HIF-1α expression promote the contraction of myometrial and smooth muscle cells. These findings contradict our experimental results; however, we found some inconsistencies in the supplement Figure 4 of their paper and the conclusion about the cell contraction. In a study related to HIF and venous contraction, Lim et al. [ 34] found that induced HIF-1α overexpression was associated with reduced venous contraction. Our study still has some limitations. We did not use TREK1 shRNA/siRNA treatment in HUSMCs to specifically inhibit endogenous TREK1 to examine the changes in cell contractility under high glucose or CoCl2 conditions. We also did not propose a specific mechanism for HIF-1α regulation of TREK1. Mammalian microRNAs (miRNA/miR) play especially powerful roles in smooth muscle cells [55]. Hypoxia induces further downregulation of miR‐124 [56], and Bucharest et al. [ 57] revealed an inverse correlation between miR-124 and TREK1 expression in neurons. Whether miR-124 is involved in the regulation of TREK1 by HIF-1α in the uterus may be a future research direction.
In conclusion, our study confirms that the HIF-1α and TREK1 protein expression is significantly increased in the gestational diabetic uterus and HUSMCs cultured under high glucose and immediate hypoxia conditions. Hypoxia is involved in the regulation of uterine smooth muscle contraction through TREK1, which is an important pathway allowing hyperglycemia to regulate the process of uterine smooth muscle contraction in patients with GDM. Intervention with hypoxia and TREK1 restores the contractility of uterine smooth muscle. Therefore, hypoxia and TREK1 may be potential targets for future modulation of uterine contractility and reduction in the complications of GDM.
## Data availability statement
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.
## Author contributions
ZY, DL, and TL wrote the manuscript and researched data. JF, HY, XW, JB, and FC researched data and contributed to the discussion. TL and JF contributed equally to this work. ZY and DL contributed equally to this work. All authors contributed to the article and approved the submitted version.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fendo.2023.1115619/full#supplementary-material
## References
1. **2. classification and diagnosis of diabetes: standards of medical care in diabetes-2021**. *Diabetes Care* (2021) **44**. DOI: 10.2337/dc21-S002
2. Buchanan TA, Xiang AH, Page KA. **Gestational diabetes mellitus: risks and management during and after pregnancy**. *Nat Rev Endocrinol* (2012) **8**. DOI: 10.1038/nrendo.2012.96
3. McIntyre HD, Catalano P, Zhang C, Desoye G, Mathiesen ER, Damm P. **Gestational diabetes mellitus**. *Nat Rev Dis Primers* (2019) **5** 47. DOI: 10.1038/s41572-019-0098-8
4. Lucas IM, Barr ELM, Barzi F, Longmore DK, Lee IL, Kirkwood M. **Gestational diabetes is associated with postpartum hemorrhage in indigenous Australian women in the Pandora study: A prospective cohort**. *Int J Gynaecol Obstet* (2021) **155** 296-304. DOI: 10.1002/ijgo.13846
5. Cauldwell M, Chmielewska B, Kaur K, van-de-l’Isle Y, Sherry A, Watt Coote I. **Screening for late-onset gestational diabetes: Are there any clinical benefits**. *BJOG* (2022) **129**. DOI: 10.1111/1471-0528.17154
6. Li Y, Wang X, Jiang F, Chen W, Li J, Chen X. **Serum lipid levels in relation to clinical outcomes in pregnant women with gestational diabetes mellitus: An observational cohort study**. *Lipids Health Dis* (2021) **20** 125. DOI: 10.1186/s12944-021-01565-y
7. Grabowska K, Stapińska-Syniec A, Saletra A, Jarmużek P, Bomba-Opoń D. **Labour in women with gestational diabetes mellitus**. *Ginekol Pol* (2017) **88**. DOI: 10.5603/GP.a2017.0016
8. Kruit H, Mertsalmi S, Rahkonen L. **Planned vaginal and planned cesarean delivery outcomes in pregnancies complicated with pregestational type 1 diabetes – a three-year academic tertiary hospital cohort study**. *BMC Pregnancy Childbirth* (2022) **22** 173. DOI: 10.1186/s12884-022-04510-8
9. Evensen A, Anderson JM, Fontaine P. **Postpartum hemorrhage: Prevention and treatment**. *Am Fam Physician* (2017) **95**
10. Al-Qahtani S, Heath A, Quenby S, Dawood F, Floyd R, Burdyga T. **Diabetes is associated with impairment of uterine contractility and high caesarean section rate**. *Diabetologia* (2012) **55**. DOI: 10.1007/s00125-011-2371-6
11. Mitidieri E, Vanacore D, Turnaturi C, Sorrentino R, d’Emmanuele di Villa Bianca R. **Uterine dysfunction in diabetic mice: The role of hydrogen sulfide**. *Antioxidants (Basel)* (2020) **9** 917. DOI: 10.3390/antiox9100917
12. Favaro RR, Salgado RM, Raspantini PR, Fortes ZB, Zorn TM. **Effects of long-term diabetes on the structure and cell proliferation of the myometrium in the early pregnancy of mice**. *Int J Exp Pathol* (2010) **91**. DOI: 10.1111/j.1365-2613.2010.00718.x
13. Parkington HC, Stevenson J, Tonta MA, Paul J, Butler T, Maiti K. **Diminished hERG k+ channel activity facilitates strong human labour contractions but is dysregulated in obese women**. *Nat Commun* (2014) **5** 4108. DOI: 10.1038/ncomms5108
14. Bai X, Bugg GJ, Greenwood SL, Glazier JD, Sibley CP, Baker PN. **Expression of TASK and TREK, two-pore domain k+ channels, in human myometrium**. *Reproduction* (2005) **129**. DOI: 10.1530/rep.1.00442
15. Yin Z, He W, Li Y, Li D, Li H, Yang Y. **Adaptive reduction of human myometrium contractile activity in response to prolonged uterine stretch during term and twin pregnancy**. *Role TREK-1 channel. Biochem Pharmacol* (2018) **152**. DOI: 10.1016/j.bcp.2018.03.021
16. Buxton IL, Singer CA, Tichenor JN. **Expression of stretch-activated two-pore potassium channels in human myometrium in pregnancy and labor**. *PLos One* (2010) **5**. DOI: 10.1371/journal.pone.0012372
17. Monaghan K, Baker SA, Dwyer L, Hatton WC, Sik Park K, Sanders KM. **The stretch-dependent potassium channel TREK-1 and its function in murine myometrium**. *J Physiol* (2011) **589**. DOI: 10.1113/jphysiol.2010.203869
18. Alotaibi MF. **The response of rat and human uterus to oxytocin from different gestational stages**. *Gen Physiol Biophys* (2017) **36** 75-82. DOI: 10.4149/gpb_2016022
19. Dillmann WH. **Diabetic cardiomyopathy**. *Circ Res* (2019) **124**. DOI: 10.1161/CIRCRESAHA.118.314665
20. Ballo P, Cameli M, Mondillo S, Giacomin E, Lisi M, Padeletti M. **Impact of diabetes and hypertension on left ventricular longitudinal systolic function**. *Diabetes Res Clin Pract* (2010) **90**. DOI: 10.1016/j.diabres.2010.08.004
21. Adingupu DD, Göpel SO, Grönros J, Behrendt M, Sotak M, Miliotis T. **SGLT2 inhibition with empagliflozin improves coronary microvascular function and cardiac contractility in prediabetic ob/ob-/- mice**. *Cardiovasc Diabetol* (2019) **18** 16. DOI: 10.1186/s12933-019-0820-6
22. Zhang Z, Tremblay J, Raelson J, Sofer T, Du L, Fang Q. **EPHA4 regulates vascular smooth muscle cell contractility and is a sex-specific hypertension risk gene in individuals with type 2 diabetes**. *J Hypertens* (2019) **37**. DOI: 10.1097/HJH.0000000000001948
23. Wang H, Zhao K, Shi N, Niu Q, Liu C, Chen Y. **Electroacupuncture regularizes gastric contraction and reduces apoptosis of interstitial cells of cajal in diabetic rats**. *Front Physiol* (2021) **12**. DOI: 10.3389/fphys.2021.560738
24. Akat F, Fıçıcılar H, Durak A, Tuncay E, Dursun AD, Topal Çelikkan F. **Intermittent hypoxia induces beneficial cardiovascular remodeling in left ventricular function of type 1 diabetic rat**. *Anatol J Cardiol* (2018) **19**. DOI: 10.14744/AnatolJCardiol.2018.00236
25. Kang MK, Kim SI, Oh SY, Na W, Kang YH. **Tangeretin ameliorates glucose-induced podocyte injury through blocking epithelial to mesenchymal transition caused by oxidative stress and hypoxia**. *Int J Mol Sci* (2020) **21** 8577. DOI: 10.3390/ijms21228577
26. Zhu Y, Ma WQ, Han XQ, Wang Y, Wang X, Liu NF. **Advanced glycation end products accelerate calcification in VSMCs through HIF-1alpha/PDK4 activation and suppress glucose metabolism**. *Sci Rep* (2018) **8** 13730. DOI: 10.1038/s41598-018-31877-6
27. Miura Y, Hayakawa A, Kikuchi S, Tsumoto H, Umezawa K, Chiba Y. **Fumarate accumulation involved in renal diabetic fibrosis in goto-kakizaki rats**. *Arch Biochem Biophys* (2019) **678**. DOI: 10.1016/j.abb.2019.108167
28. Gunton JE. **Hypoxia-inducible factors and diabetes**. *J Clin Invest* (2020) **130**. DOI: 10.1172/JCI137556
29. Catrina SB, Zheng X. **Hypoxia and hypoxia-inducible factors in diabetes and its complications**. *Diabetologia* (2021) **64**. DOI: 10.1007/s00125-021-05380-z
30. Lee JW, Bae SH, Jeong JW, Kim SH, Kim KW. **Hypoxia-inducible factor (HIF-1)alpha: its protein stability and biological functions**. *Exp Mol Med* (2004) **36** 1-12. DOI: 10.1038/emm.2004.1
31. Ke Q, Costa M. **Hypoxia-inducible factor-1 (HIF-1)**. *Mol Pharmacol* (2006) **70**. DOI: 10.1124/mol.106.027029
32. Prabhakar NR, Peng YJ, Nanduri J. **Hypoxia-inducible factors and obstructive sleep apnea**. *J Clin Invest* (2020) **130**. DOI: 10.1172/JCI137560
33. Packer M. **Mechanisms leading to differential hypoxia-inducible factor signaling in the diabetic kidney: Modulation by SGLT2 inhibitors and hypoxia mimetics**. *Am J Kidney Dis* (2021) **77**. DOI: 10.1053/j.ajkd.2020.04.016
34. Lim CS, Qiao X, Reslan OM, Xia Y, Raffetto JD, Paleolog E. **Prolonged mechanical stretch is associated with upregulation of hypoxia-inducible factors and reduced contraction in rat inferior vena cava**. *J Vasc Surg* (2011) **53**. DOI: 10.1016/j.jvs.2010.09.018
35. Alotaibi M. **Hypoxic preconditioning ameliorates endometrial and myometrial damage and improves uterine function following prolonged hypoxia in nonpregnant rats**. *Gen Physiol Biophys* (2019) **38** 497-503. DOI: 10.4149/gpb_2019031
36. Wray S, Duggins K, Iles R, Nyman L, Osman V. **The effects of metabolic inhibition and acidification on force production in the rat uterus**. *Exp Physiol* (1992) **77**. DOI: 10.1113/expphysiol.1992.sp003590
37. Zhang L, Li L, Liu H, Prabhakaran K, Zhang X, Borowitz JL. **HIF-1alpha activation by a redox-sensitive pathway mediates cyanide-induced BNIP3 upregulation and mitochondrial-dependent cell death**. *Free Radic Biol Med* (2007) **43**. DOI: 10.1016/j.freeradbiomed.2007.04.005
38. Kim A, Jung HG, Kim SC, Choi M, Park JY, Lee SG. **Astrocytic AEG-1 regulates expression of TREK-1 under acute hypoxia**. *Cell Biochem Funct* (2020) **38**. DOI: 10.1002/cbf.3469
39. **Diabetes in pregnancy: Diagnosis and treatment**. *Chin J Obstet Gynecol (Published Chinese)* (2014) **49**. DOI: 10.3760/cma.j.issn.0529-567X.2014.08.001
40. Yin Z, Li T, Zhou L, Fei J, Su J, Li D. **Optimal delivery time for patients with diet-controlled gestational diabetes mellitus: A single-center real-world study**. *BMC Pregnancy Childbirth* (2022) **22** 356. DOI: 10.1186/s12884-022-04683-2
41. Crews JK, Novak J, Granger JP, Khalil RA. **Stimulated mechanisms of Ca2+ entry into vascular smooth muscle during NO synthesis inhibition in pregnant rats**. *Am J Physiol* (1999) **276**. DOI: 10.1152/ajpregu.1999.276.2.R530
42. Samir SM, Mostafa AF. **Abscisic acid: A novel uterine stimulator in normal and diabetic rats**. *Can J Physiol Pharmacol* (2018) **96**. DOI: 10.1139/cjpp-2018-0040
43. McMurtrie EM, Ginsberg GG, Frederick GT, Kirkland JL, Stancel GM, Gardner RM. **Effect of a diabetic state on myometrial ultrastructure and isolated uterine contractions in the rat**. *Proc Soc Exp Biol Med* (1985) **180** 497-504. DOI: 10.3181/00379727-180-42208
44. Jawerbaum A, Catafau JR, Gonzalez ET, Novaro V, Gómez G, Gelpi E. **Eicosanoid production, metabolism and contractile activity in the isolated uterus from non-insulin-dependent diabetic rats during late pregnancy**. *Prostaglandins* (1996) **51**. DOI: 10.1016/0090-6980(96)00023-8
45. Spiegl G, Zupkó I, Minorics R, Csík G, Csonka D, Falkay G. **Effects of experimentally induced diabetes mellitus on pharmacologically and electrically elicited myometrial contractility**. *Clin Exp Pharmacol Physiol* (2009) **36**. DOI: 10.1111/j.1440-1681.2009.05162.x
46. Jawerbaum A, Roselló Catafau J, Gonzalez ET, Novaro V, Gomez G, Gelpi E. **Glucose metabolism, triglyceride and glycogen levels, as well as eicosanoid production in isolated uterine strips and in embryos in a rat model of non-insulin-dependent diabetes mellitus during pregnancy**. *Prostaglandins* (1994) **47** 81-96. DOI: 10.1016/0090-6980(94)90079-5
47. Kaya T, Cetin A, Cetin M, Sarioglu Y. **Effects of endothelin-1 and calcium channel blockers on contractions in human myometrium. A study on myometrial strips from normal and diabetic pregnant women**. *J Reprod Med* (1999) **44**
48. Cowles CL, Wu YY, Barnett SD, Lee MT, Burkin HR, Buxton IL. **Alternatively spliced human TREK-1 variants alter TREK-1 channel function and localization**. *Biol Reprod* (2015) **93** 122. DOI: 10.1095/biolreprod.115.129791
49. Kapustin RV, Kopteeva EV, Alekseenkova EN, Tral TG, Tolibova GK, Arzhanova ON. **Placental expression of endoglin, placental growth factor, leptin, and hypoxia-inducible factor-1 in diabetic pregnancy and pre-eclampsia**. *Gynecol Endocrinol* (2021) **37**. DOI: 10.1080/09513590.2021.2006513
50. Isoe T, Makino Y, Mizumoto K, Sakagami H, Fujita Y, Honjo J. **High glucose activates HIF-1-mediated signal transduction in glomerular mesangial cells through a carbohydrate response element binding protein**. *Kidney Int* (2010) **78** 48-59. DOI: 10.1038/ki.2010.99
51. Guo S, Bragina O, Xu Y, Cao Z, Chen H, Zhou B. **Glucose up-regulates HIF-1 alpha expression in primary cortical neurons in response to hypoxia through maintaining cellular redox status**. *J Neurochem* (2008) **105**. DOI: 10.1111/j.1471-4159.2008.05287.x
52. Lee HJ, Ryu JM, Jung YH, Lee SJ, Kim JY, Lee SH. **High glucose upregulates BACE1-mediated abeta production through ROS-dependent HIF-1alpha and LXRα/ABCA1-regulated lipid raft reorganization in SK-N-MC cells**. *Sci Rep* (2016) **6**. DOI: 10.1038/srep36746
53. Yang X, Bao M, Fang Y, Yu X, Ji J, Ding X. **STAT3/HIF-1alpha signaling activation mediates peritoneal fibrosis induced by high glucose**. *J Transl Med* (2021) **19** 283. DOI: 10.1186/s12967-021-02946-8
54. Wen B, Zheng Z, Wang L, Qian X, Wang X, Chen Y. **HIF-1alpha is essential for the augmentation of myometrial contractility during labor**. *Biol Reprod* (2022) **107** 1540–50. DOI: 10.1093/biolre/ioac174
55. Liu Y, Gao L. **Preterm labor, a syndrome attributed to the combination of external and internal factors**. *MFM* (2022) **4** 61-71. DOI: 10.1097/FM9.0000000000000136
56. Gu H, Liu M, Ding C, Wang X, Wang R, Wu X. **Hypoxia-responsive miR-124 and miR-144 reduce hypoxia-induced autophagy and enhance radiosensitivity of prostate cancer cells**. *Cancer Med* (2016) **5**. DOI: 10.1002/cam4.664
57. Paschou M, Maier L, Papazafiri P, Selescu T, Dedos SG, Babes A. **Neuronal microRNAs modulate TREK two-pore domain k**. *RNA Biol* (2020) **17**. DOI: 10.1080/15476286.2020.1722450
|
---
title: Gut microbiota composition alteration analysis and functional categorization
in children with growth hormone deficiency
authors:
- Congfu Huang
- Dongming Meng
- Yinhu Li
- Shiyang Lu
- Wei Yang
- Bin Wu
- Shufen Chen
- Zhenyu Yang
- Haiying Liu
journal: Frontiers in Pediatrics
year: 2023
pmcid: PMC9998986
doi: 10.3389/fped.2023.1133258
license: CC BY 4.0
---
# Gut microbiota composition alteration analysis and functional categorization in children with growth hormone deficiency
## Abstract
### Objective
To study changes in the composition and functions of the gut microbiota (GM) in children with growth hormone deficiency (GHD) using high-throughput sequencing.
### Methods
Thirty-three children with GHD diagnosed in Longgang District Maternity and Child Health Hospital were included in the disease group and 24 healthy children of the same age comprised the control group. Total DNA was extracted and amplified from stool samples obtained from all subjects. High-throughput sequencing was used to analyze the GM composition and functions.
### Results
The GM from the two groups of children showed significant differences in α-diversity ($P \leq 0.05$). In comparison with the control group, the abundance of the phylum Bacteroidetes was significantly higher ($45.96\%$ vs. $65.71\%$) while the Firmicutes count was significantly lower ($47.09\%$ vs. $25.20\%$). At the genus level, the abundance of Prevotella in the disease group was significantly higher ($3.16\%$ vs. $20.67\%$) and that of *Lachnospiracea incertae* sedis, Clostridium XlVa, and Megamonas was lower ($6.576\%$ vs. $1.75\%$; $4.51\%$ vs. $0.80\%$; $5.08\%$ vs. $2.02\%$, respectively). GM functions, including those involved in membrane_transport, energy_metabolism, poorly_characterized, metabolism_of_cofactors_and_vitamins, glycan_biosynthesis_and_metabolism, transcription, folding,_sorting,_and_degradation, were significantly altered in the disease group. The abundance of various GM components was correlated with endocrine hormone levels.
### Conclusion
Significant alterations in the GM are seen in children with growth hormone deficiency, which may affect both energy metabolism and the levels of endocrine hormones, potentially leading to growth restriction.
## Introduction
Short stature is defined as a height of less than two standard deviations or less than the third percentile among children of the same sex, age, or race. Growth hormone deficiency (GHD) is a growth disorder caused by reduced or absent production of growth hormone (GH). It is one of the most common causes of short stature in children, accounting for $38.6\%$ of all causes [1]. The worldwide incidence of GHD in children varies between $\frac{1}{4}$,000 and $\frac{1}{10}$,000 and most children show idiopathic GHD [2].
The stability of the gut microbiota (GM) is an important factor influencing the growth and development of children [3]. Intestinal microorganisms and metabolites such as short-chain fatty acids (SCFAs) can regulate the production of hormones related to bone health, including sex steroids, vitamin D, and serotonin [4, 5]. In addition, they mediate signal transduction via the intestinal–brain axis and affect the secretion of GH-releasing peptide, somatostatin, and leptin, all of which regulate the GH/insulin-like growth factor-1 (IGF-1) axis and modulate processes such as GH secretion, appetite regulation, and bone growth (1, 6–10). Growth hormone can not only directly promote the growth of all organs but also stimulate the production of IGF-1. The latter is an effective growth factor that plays a synergistic role with growth hormone to maintain overall growth and metabolism [11, 12]. Conversely, GH or IGF-1 can also affect the composition and functions of the GM in different ways [1]. Li et al. [ 13] reported significant changes in the GM of children with idiopathic short stature where intestinal *Clostridium and* Eubacterium were significantly and positively correlated with their height standard deviation score (SDS) and IGF-1 SDS. The authors believed that the decrease in IGF-1 synthesis by *Clostridium and* Eubacterium through SCFAs might be one of the underlying causes.
The hypothalamus–pituitary–IGF-1 axis is the main hormonal regulator of growth and development, of which GH and IGF-1 are key components [14]. GHD children have reduced levels of GH and IGF-1. Imbalances in the GM can lead to endocrine hormone disorders. We speculate that children with GHD may also have GM imbalances. In this study, the intestinal composition and function of GHD children and healthy children of the same age were compared, and correlations between their GM and several hormones were analyzed to explore the characteristics of the GM of GHD children and the possible mechanism of action.
## Sample screening
We selected 33 children with GHD diagnosed at Longgang District Maternity and Child Health Hospital as the disease group, and 24 healthy children of the same age as the control group. The ages of children in the two groups ranged between 5 and 14 years, with no statistical difference seen in the comparative analysis ($P \leq 0.05$) (Table 1). All the children with GHD were diagnosed at the Department of Growth and Development, Shenzhen Longgang District Maternity and Child Health Hospital. The disease group met the diagnostic criteria for GHD in Chinese children [15]: ① Below the third percentile of the height of normal healthy children of the same age and sex (−1.88 standard deviations [−1.88 SD] or minus 2 standard deviations [−2 SD]); ② Annual growth rate <5 cm/year; ③ Symmetrical dwarfism and childish face; ④ Normal intelligence development; ⑤ Bone age lagging behind actual age; ⑥ Peak values of two GH drug provocation tests of <10 µg/L; ⑦ Lower than normal level of serum IGF-1. The exclusion criteria for children in the two groups included: ① Severe liver or gastrointestinal disorders; ② Severe infection; ③ Treatment with antibiotics or probiotic preparations within one month before the test. All children provided informed consent from their guardians before enrollment.
**Table 1**
| Group | Age (year) | Gender (male/female) | Weight (kg) | Height (cm) | IGF-1 |
| --- | --- | --- | --- | --- | --- |
| Disease group (n = 33) | 8.73 ± 2.40 | 21/12 | 22.53 ± 1.02 | 120.72 ± 1.87 | 179.71 ± 75.73 |
| Control group (n = 24) | 8.78 ± 2.04 | 14/10 | 27.45 ± 1.34 | 128.55 ± 2.49 | 235.55 ± 70.89 |
| F/t Value | 0.057 | 0.165 | 0.037 | 2.566 | 2.171 |
| P-value | 0.955 | 0.685 | 0.154 | 0.013 | 0.038 |
## Collection of fecal samples for DNA extraction and sequencing from two groups of children
Approximately 5 g of the middle section of the feces was collected and immediately frozen and stored at ‒80°C. The samples were transported on dry ice to Shenzhen Micro Health Gene Technology Co., Ltd. for high-throughput sequencing. MoBio's PowerSoil® DNA Isolation Kit was used to extract bacterial DNA from fecal samples. Amplification of the V3 – V4 region of the 16S rRNA gene in DNA was performed by polymerase chain reaction (PCR). Amplified samples were sequenced using the Illumina MiSeq high-throughput sequencing platform.
## Sequencing data analysis
Low-quality reads were filtered from the sequencing data using self-programming bioinformatics tools, and the data were spliced using FLASH software (v12.11, http://ccb.jhu.edu/software/FLASH/index.shtml). The splicing sequences were aggregated into OTUs (sortable elements) with USEARCH, which were compared with the bacterial library (Greengene V201305) to obtain the GM compositions of all samples. The bacterial abundance in the samples of both groups was analyzed only at the phylum and genus levels.
## Statistical methods
The ade4 package in R (v3.3.3) software was used to perform principal component analysis (PCA) based on the composition and relative abundance of bacteria in all samples at the genus level. The overall distribution of the microbiota compositions in the two groups was plotted. Bacteria were classified to the phylum and genus levels, and different species between the two groups were investigated by the Wilcoxon method where $P \leq 0.05$ indicated a significant difference. The 16S rDNA sequencing data were used to evaluate differences in bacterial functions between the two groups of children based on the functional analysis performed by the Kyoto Encyclopedia of Genes and Genomes (KEGG) database. SPSS 22.0 software was used for general data analysis. The age, weight, height, and IGF-1 values were compared by χ2 tests or two-group independent sample t-tests.
## Comparison of differences in the composition of the GM
The GM of two groups of children showed significant differences in α-diversity ($$P \leq 0.033$$) (Figure 1). We used PCA to reduce the dimensionality of the GM data of the two groups, finding that there were marked differences in the GM between the two groups. *The* genera that contributed most to this difference included Prevotella ($P \leq 0.001$), Megamonas ($$P \leq 0.01$$), Bacteroides ($$P \leq 0.765$$), Bifidobacterium ($$P \leq 0.011$$), and Faecalibacterium ($$P \leq 0.094$$) (Figure 2).
**Figure 1:** *Chart showing comparison of microbiota diversity between the two groups.* **Figure 2:** *Principal component analysis.*
## Comparison of the dominant bacterial phyla between the two groups of children
The top five dominant bacterial phyla differed between the groups with a significant increase in the abundance of Bacteroides in the disease group ($$P \leq 0.000$$) together with a significant reduction in the abundance of Firmicutes ($$P \leq 0.000$$). In addition, there was also a significant difference between the two groups in the abundance of Fusobacteria and Actinomycetes ($P \leq 0.05$) (Table 2 and Figure 3).
**Figure 3:** *Comparison of the GM levels in the two groups.* TABLE_PLACEHOLDER:Table 2
## Comparison of the dominant bacterial genera between the two groups of children
We selected the top 15 dominant bacterial genera in the two groups for comparison. The results showed that the abundance of Prevotella, Fusobacterium, Klebsiella, and Alistipes was significantly increased in the disease group ($P \leq 0.05$) while that of *Lachnospiracea incertae* sedis, Megamonas, Blautia, Clostridium XlVa, Bifidobacterium, and Eubacterium was significantly decreased ($P \leq 0.05$) (Table 3 and Figure 4).
**Figure 4:** *Comparison of the abundance of bacterial genera between the two groups. Remarks: * indicates $P \leq 0.05$, ** indicates $P \leq 0.01$, and *** indicates $P \leq 0.001$ statistically significant differences between the two groups. The higher the number of asterisks, more significant the difference.* TABLE_PLACEHOLDER:Table 3
## Alterations of GM functions in the GHD children
In comparison with the healthy children, the GHD patients showed significant changes in GM functions, including the decreased “Membrane transport” ($P \leq 0.001$, FDR < 0.001), “Lipid metabolism” ($$P \leq 0.025$$, FDR = 0.042), and “Transcription” ($P \leq 0.001$, FDR < 0.001, Figure 5), which indicated the. In contrast, the functional categories, such as “Energy metabolism” ($P \leq 0.001$, FDR < 0.001), “Metabolism of cofactors and vitamins” ($P \leq 0.001$, FDR < 0.001), “Nucleotide metabolism” ($$P \leq 0.008$$, FDR = 0.016), “*Glycan biosynthesis* and metabolism” ($P \leq 0.001$, FDR < 0.001), and “Folding sorting and degradation” ($P \leq 0.001$, FDR < 0.001) were enriched in the GHD patients (Figure 5). These elevated GM metabolic activities in the GHD patients, especially the “*Glycan biosynthesis* and metabolism” function, affect the neuro-regulations in hosts and is probably related to the occurrence of GHD.
**Figure 5:** *Comparison of GM functions between the two groups of children.*
## GM and clinical phenotypes
Spearman's correlation analysis was used to investigate associations between the GM of children with GHD and eight endocrine hormones. Our results showed that Bacteroides were positively correlated and Prevotella was negatively correlated with insulin, while Alistipes and Haemophilus showed a negative correlation with GH. A positive correlation was also reported between Fusicatenibacter, Fusobacterium, and Sutterella, whereas Veillonella was negatively correlated with prolactin. Faecaliterium and FSH were positively correlated (Figure 6).
**Figure 6:** *Correlations between GM compositions and endocrine hormones in children with GHD. Legend description: A correlation analysis was performed with eight clinical phenotypes and genera with a relative abundance of ≥0.1%. Results are shown as above where significance was expressed as *P < 0.05 and **P < 0.01.*
## The GM composition differed markedly between the disease and control groups
Compared with the healthy controls, children in the disease group showed reduced α-diversity in the GM, consistent with results reported in malnourished children [16]. The phylum Bacteroides was more abundant in children from the disease group than in those from the control group while the opposite trend was observed for Firmicutes, in contrast to findings on obese and diabetic patients [17]. The abundance of Prevotella in the disease group was also significantly higher than that in the control group. Prevotella can degrade broad-spectrum plant polysaccharides [18], and carbohydrate-based diets tend to form a Prevotella-dominated “gut type.” Increased abundance of Prevotella abundance has been shown to reduce blood sugar and insulin levels, thus affecting energy absorption and promoting weight loss [19]. In the disease group, the abundance of Fusobacterium, Klebsiella, Alistipes, and other genera was found to be significantly increased. Fusobacterium is present in the normal oral flora and can inhibit the immune response as well as promote the transformation of inflammation to malignancy [20]. An increase in the abundance of both Klebsiella and Alistipes has been shown to be associated with intestinal inflammation [21, 22]; therefore, the increase in the population of these genera can promote chronic inflammation in the intestine and disrupt the function of the intestinal barrier. This can lead to a cellular biochemical imbalance, reduced absorption capacity, and increased susceptibility to enteric pathogen infections, and consequently affect energy metabolism and nutrient absorption [23]. In addition, Klebsiella and Alistipes are both associated with neurological diseases [24] and can produce neurotransmitter-related metabolites such as serotonin, dopamine, and histamine [25]. These neurotransmitters enter the brain through the gut-brain axis to regulate the energy balance and function of the hypothalamus [26, 27]. The hypothalamus is the highest regulatory center of thehypothalamic–pituitary–growth axis (HPA) and can reduce appetite and cause weight loss [28, 29]. Lachnospiracea incertae sedis, Megamonas, Blautia, Clostridium XlVa, and Bifidobacterium were found to be significantly reduced in the intestines of the disease group, which could lead to reduced concentrations of SCFAs such as butyric acid produced by these beneficial bacteria [30]. Decreased abundance of *Lachnospiracea incertae* sedis might also affect protein synthesis [31], disturb the intestinal energy supply, and retard growth and development. Jensen et al. [ 5] reported that increased Prevotella abundance together with reduced numbers of Bifidobacterium can reduce the levels of growth hormone-releasing peptide (GHRP) and leptin, thus reducing the release of GH.
## Significant differences in GM function between the two groups
The enriched functional categories in the GHD group included “Replication and repair, Energy metabolism, Poorly characterized, Metabolism of cofactors and vitamins, Nucleotide metabolism, Cellular processes and signaling, Nucleotide metabolism, *Glycan biosynthesis* and metabolism, Transcription, Folding sorting and degradation”. Children with GHD showed dysregulation in energy metabolism, vitamin and related factor metabolism, and polysaccharide metabolism and biosynthesis. Considering that Prevotella significantly increases the catabolism of carbohydrates and that the abundance of butyric acid and other bacteria such as *Lachnospiracea incertae* sedis is significantly reduced in children with GHD, the GM imbalance in this population may affect the functions of the flora. This phenomenon may lead to chronic inflammation of the intestine and poor intake and absorption of nutrients such as fats and proteins, affecting both growth and development.
## GM is closely related to the clinical phenotype
We conducted a correlation analysis of the GM and endocrine hormones and found that Prevotella abundance was negatively correlated with insulin. Significantly higher abundance of Prevotella can affect insulin secretion, which can not only regulate food intake [19, 32] but also modulate blood glucose levels through signaling pathways essential for maintaining energy storage, glucose metabolism, sugar production, adipogenesis, cell growth, survival, and reproduction [33]. We speculate that this significant increase in Prevotella abundance may be detrimental to growth and development. We also found a variety of other intestinal bacteria related to endocrine hormones and confirmed the interaction between the GM and endocrine hormones. Maintaining the stability of the GM is conducive to the promotion of growth and development.
## Conclusion
There was a significant reduction in the α-diversity of the intestinal microbial composition in GHD children, together with an increased abundance of Bacteroides and reduced numbers of Firmicutes. Fusobacterium, Klebsiella, Alistipes, and other genera were significantly enriched in children with GHD while the numbers of *Lachnospiracea incertae* sedis, Megamonas, Blautia, Clostridium XlVa, and Bifidobacterium were significantly reduced. These imbalances in the GM were predicted to affect pathways involved in energy metabolism and biosynthesis, as well as induce abnormal secretion of insulin and other endocrine hormones, which may promote the occurrence and development of GHD.
## Deficiencies and next steps
There are many factors that cause insufficiency in GH secretion in children with GHD, and GM imbalance may be one of the major factors. On the one hand, GM imbalance leads to the abnormal secretion of endocrine hormones as well as an abnormal production of microbial metabolites, especially neurotransmitters, that can influence the HPA through the gut–brain axis. The sample size in the present study was small, consisting of only 33 children with GHD; hence, large-sample, multi-center research is needed to verify the associations between the GM and GHD. Studies combined with metabolomics could better clarify the mechanism of action of the GM and its metabolites in growth and development.
## Data availability statement
The data presented in the study are deposited in the NCBI sequence Archive (SRA) database, accession number: PRJNA899674.*The data* can be found at the following link: https://dataview.ncbi.nlm.nih. gov/object/PRJNA899674?reviewer=tu7mnej3p04c61u31c6f4v5hgo.
## Ethics statement
The studies involving human participants were reviewed and approved by The Ethics Committee of Shenzhen Longgang District Maternity and Child Health Care Hospital approved the study, with the approval number of LGFYYXLL-024. Written informed consent to participate in this study was provided by the participants’ legal guardian/next of kin. Written informed consent was obtained from the minor(s)’ legal guardian/next of kin for the publication of any potentially identifiable images or data included in this article.
## Author contributions
CH and HL: managed the project. DM, YL, SL, and WY: were responsible for the registration of the clinical information of the enrolled children and the collection of stool samples according to the standard configuration. CH, YL, and ZY: were responsible for DNA extraction and biological information analysis, etc. HL and ZY: were responsible for tabulation and statistical analyses. CH and DM: were responsible for interpreting various data and writing papers. All authors contributed to the article and approved the submitted version.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher's note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## References
1. Han XW, Dong ZY, Zhang WY, Ma XY, An JG, Xiao Y. **Analysis of the etiology and clinical characteristics of short stature**. *J Clin Pediatr* (2019.0) **37** 39-42. DOI: 10.3969/j.issn.1000-3606.2019.01.010
2. Chinoy A, Murray PG. **Diagnosis of growth hormone deficiency in the paediatric and transitional age**. *Best Pract Res Clin Endocrinol Metab* (2016.0) **30** 737-47. DOI: 10.1016/j.beem.2016.11.002
3. Qi XY, Yun CY, Pang YL, Qiao J. **The impact of the GM on the reproductive and metabolic endocrine system**. *Gut Microbes* (2021.0) **13** 1-21. DOI: 10.1080/19490976.2021.1894070
4. Hsu E, Pacififici R. **From osteoimmunology to osteomicrobiology: how the microbiota and the immune system regulate bone**. *Calcif Tissue Int* (2018.0) **102** 512-21. DOI: 10.1007/s00223-017-0321-0
5. Jensen EA, Young JA, Mathes SC, List EO, Carroll RK, Kuhn J. **Crosstalk between the growth hormone/insulin-like growth factor-1 axis and the gut microbiome: a new frontier for microbial endocrinology**. *Growth Horm IGF Res* (2020.0) **53-54** 101333. DOI: 10.1016/j.ghir.2020.101333
6. 6.Martel-Pelletier J, Di Battista JA, Lajeunesse D, Pelletier JP. GF/IGFBP axis in cartilage and bone in osteoarthritis pathogenesis. Inflamm Res. (1998) 47(3):90–100. 10.1007/s000110050288. DOI: 10.1007/s000110050288
7. Philbrick KA, Wong CP, Branscum AJ, Turner RT, Iwaniec UT. **Leptin stimulates bone formation in ob/ob mice at doses having minimal impact on energy metabolism**. *J Endocrinol* (2017.0) **232** 461-74. DOI: 10.1530/JOE-16-0484
8. Villa-Osaba A, Gahete MD, Cordoba-Chacon J, De Lecea L, Castaño JP, Luque RM. **Fasting modulates GH/IGF-I axis and its regulatory systems in the mammary gland of female mice: influence of endogenous cortistatin**. *Mol Cell Endocrinol* (2016.0) **434** 14-24. DOI: 10.1016/j.mce.2016.06.014
9. Shamsi BH, Chatoo M, Xu XK, Xu X, Chen XQ. **Versatile functions of somatostatin and somatostatin receptors in the gastrointestinal system**. *Front Endocrinol* (2021.0) **12** 652363. DOI: 10.3389/fendo.2021.652363
10. Schubert ML. **Physiologic, pathophysiologic, and pharmacologic regulation of gastric acid secretion**. *Curr Opin Gastroenterol* (2017.0) **33** 430-8. DOI: 10.1097/MOG.0000000000000392
11. Hjelholt AJ, Lee KY, Arlien-Søborg MC, Pedersen SB, Kopchick JJ, Puri V. **Temporal patterns of lipolytic regulators in adipose tissue after acute growth hormone exposure in human subjects: a randomized controlled crossover trial**. *Mol Metab* (2019.0) **29** 65-75. DOI: 10.1016/j.molmet.2019.08.013
12. Dixit M, Poudel SB, Yakar S. **Effects of GH/IGF axis on bone and cartilage**. *Mol Cell Endocrinol* (2021.0) **519** 111052. DOI: 10.1016/j.mce.2020.111052
13. Li L, An JJ, Wang JQ, Wang XQ, Dong ZY. **The structure of gut microbiome in idiopathic short stature profiled by 16S rRNA second generation sequencing sequencing**. *J Diagn Concepts Pract* (2021.0) **20** 149-54. DOI: 10.16150/j.1671-2870.2021.02.006
14. Al-Samerria S, Radovick S. **The role of Insulin-like Growth Factor-1 (IGF-1) in the control of neuroendocrine regulation of growth**. *Cells* (2021.0) **10** 2664. DOI: 10.3390/cells10102664
15. **Guidelines for diagnosis and treatment of children with short stature**. *Chin J Pediatr* (2008.0) **46** 20-1
16. Guerrant RL, DeBoer MD, Moore SR. **The impoverished gut–a triple burden of diarrhoea, stunting and chronic disease**. *Nat Rev Gastroenterol Hepatol* (2013.0) **10** 220-9. DOI: 10.1038/nrgastro.2012.239
17. Ley RE, Turnbaugh PJ, Klein S, Gordon JI. **Microbial ecology: human gut microbes associated with obesity**. *Nature* (2006.0) **444** 1022-3. DOI: 10.1038/4441022a
18. Wu GD, Chen J, Hoffmann C, Bittinger K, Chen YY, Keilbaugh SA. **Linking long-term dietary patterns with gut microbial enterotypes**. *Science* (2011.0) **333** 105-8. DOI: 10.1126/science.1208344
19. Hjorth MF, Blædel T, Bendtsen LQ, Lorenzen JK, Holm JB, Kiilerich P. **Prevotella-to-bacteroides ratio predicts body weight and fat loss success on 24-week diets varying in macronutrient composition and dietary fiber: results from a post-hoc analysis**. *PLoS One* (2016.0) **11** e0161211. DOI: 10.1038/s41366-018-0093-2
20. Zhou Z, Chen J, Yao H, Hu H. **Fusobacterium and colorectal cancer**. *Front Oncol* (2018.0) **8** 371. DOI: 10.3389/fonc.2018.00371
21. Atarashi K, Suda W, Luo C, Kawaguchi T, Motoo I, Narushima S. **Ectopic colonization of oral bacteria in the intestine drives TH1 cell induction and inflammation**. *Science* (2017.0) **358** 359-65. DOI: 10.1126/science.aan4526
22. Saulnier DM, Riehle K, Mistretta TA, Diaz MA, Mandal D, Raza S. **Gastrointestinal microbiome signatures of pediatric patients with irritable bowel syndrome**. *Gastroenterology* (2011.0) **141** 1782-91. DOI: 10.1053/j.gastro.2011.06.072
23. Sharon G, Cruz NJ, Kang DW, Gandal MJ, Wang B, Kim YM. **Human GM from autism spectrum disorder promote behavioral symptoms in mice**. *Cell* (2019.0) **177** 1600-1618 e17. DOI: 10.1016/j.cell.2019.05.004
24. Kau AL, Ahern PP, Griffin NW, Goodman AL, Gordon JI. **Human nutrition, the gut microbiome and the immune system**. *Nature* (2011.0) **474** 327-36. DOI: 10.1038/nature10213
25. Strandwitz P. **Neurotransmitter modulation by the GM**. *Brain Res* (2018.0) **1693** 128-33. DOI: 10.1016/j.brainres.2018.03.015
26. Louwies T, Johnson AC, Orock A, Yuan T, Meerveld BGV. **The microbiota-gut-brain axis: an emerging role for the epigenome**. *Exp Biol Med* (2020.0) **245** 138-45. DOI: 10.1177/1535370219891690
27. Heiss CN, Olofsson LE. **GM-dependent modulation of energy metabolism**. *J Innate Immun* (2018.0) **10** 163-71. DOI: 10.1159/000481519
28. Georgescu T, Lyons D, Heisler LK. **Role of serotonin in body weight, insulin secretion and glycaemic control**. *J Neuroendocrinol* (2021.0) **33** e12960. DOI: 10.1111/jne.12960
29. Li H, Li Y, Zhang X, Ren G, Ren G, Wang LF. **The combination of**. *Front Pharmacol* (2021.0) **11** 614413. DOI: 10.3389/fphar.2020.614413
30. Zhang JD, Song LJ, Wang YJ, Liu C, Zhang L, Zhu SW. **Benefificial effect of butyrate-producing Lachnospiraceae on stress-induced visceral hypersensitivity in rats**. *J Gastroenterol Hepatol* (2019.0) **34** 1368-76. DOI: 10.1111/jgh.14536
31. Chen WG, Liu FL, Ling ZX, Tong XJ, Xiang CL. **Human intestinal lumen and mucosa-associated microbiota in patients with colorectal cancer**. *PLoS One* (2012.0) **7** e39743. DOI: 10.1371/journal.pone.0039743
32. Mitchell CS, Begg DP. **The regulation of food intake by insulin in the central nervous system**. *J Neuroendocrinol* (2021.0) **33** e12952. DOI: 10.1111/jne.12952
33. Edgerton DS, Moore MC, Gregory JM, Kraft G, Cherrington AD. **Importance of the route of insulin delivery to its control of glucose metabolism**. *Am J Physiol Endocrinol Metab* (2021.0) **320** E891-7. DOI: 10.1152/ajpendo.00628.2020
|
---
title: 'Constipation and cardiovascular disease: A two-sample Mendelian randomization
analysis'
authors:
- Qichao Dong
- Delong Chen
- Yuxuan Zhang
- Yi Xu
- Longhui Yan
- Jun Jiang
journal: Frontiers in Cardiovascular Medicine
year: 2023
pmcid: PMC9998987
doi: 10.3389/fcvm.2023.1080982
license: CC BY 4.0
---
# Constipation and cardiovascular disease: A two-sample Mendelian randomization analysis
## Abstract
### Background
Although several observational studies have suggested positive associations between constipation and cardiovascular disease (CVD), a solid causal association has not been demonstrated. Therefore, a two-sample Mendelian randomization (MR) study was performed to investigate the causal associations between constipation and CVD.
### Methods
*Independent* genetic variants strongly associated with constipation were obtained from the FinnGen consortium. Summary-level data for CVD, including coronary artery disease (CAD), myocardial infarction (MI), heart failure (HF), atrial fibrillation (AF), stroke, and its subtypes, were collected from a few extensive genome-wide association studies (GWASs). The inverse-variance weighted methods, weighted median, and MR-Egger were used for the MR estimates. The Cochran’s Q test, MR-Egger intercept tests, MR-PRESSO, MR Steiger test, leave-one-out analyses, and funnel plot were used in the sensitivity analysis.
### Results
Genetically determined constipation was suggestively associated with AF risk (odds ratio (OR), 1.07; $95\%$ confidence interval (CI), 1.01, 1.14; $$p \leq 0.016$$). Constipation and other CVD do not appear to be causally related. It was demonstrated that the results were robust through sensitivity analyses.
### Conclusion
This MR study demonstrated suggestive causal associations of constipation on AF, despite no associations achieving a significance value after multiple testing corrections. There was no evidence of an association between constipation and the risk of CAD, MI, HF, stroke, or stroke subtypes.
## Introduction
As the primary cause of mortality and disability globally, cardiovascular disease (CVD) poses an increasingly healthy and social burden with the world’s population aging [1]. In view of the severe social and clinical consequences, prompt action was required to identify risk factors of CVD for early prevention and intervention [2]. Genetic or environmental factors may lead to the occurrence and progression of CVD. In addition, some studies have indicated that constipation is probably associated with CVD (3–5).
Constipation is a prevalent worldwide health issue reported daily in clinical practice [6]. The prevalence of constipation among patients hospitalized for cardiovascular disease is approximately $50\%$ [7]. Several epidemiological studies have reported associations between constipation and CVD. A cohort study including 73,047 postmenopausal women showed that patients with severe constipation experienced a $23\%$ higher risk of coronary artery disease (CAD) at a median follow-up of 6.9 years [4]. In another cohort of 3,359,653 US veterans, patients with constipation experienced an $11\%$ higher risk of CAD and a $19\%$ higher risk of ischemic stroke [5]. Meanwhile, Honkura and colleagues robustly demonstrated that constipation is significantly related to overall cardiovascular disease mortality in the general population, which was mainly related to the risk of stroke [8]. Additionally, constipation also increases the risk of atrial fibrillation (AF) as well as heart failure (HF) [3].
However, it is still not fully elucidated whether the effects of constipation on CVD risk were merely biased by shared pleiotropic factors or reverse causation due to the inherent defects of conventional observational studies [9, 10]. Besides, randomized controlled studies (RCTs) are time and labor-consuming to implement this topic. Recently, Mendelian randomization (MR) has been increasingly used to assess credible causal relationships between exposures and outcomes [11]. Founded on the principle of the random assortment of genetic variants through meiosis, MR used genetic variations related to exposure as instrumental variables (IVs) to infer the association between risk factors (e.g., constipation) and disease outcomes (e.g., CVD) [12]. *Because* genetic variants are randomly allocated at conception before disease onset, MR analysis could avoid confounding factors and reverse causality, further identifying causal determinants of a particular outcome [13]. In the present study, a two-sample MR study was implemented to investigate the potential causality between constipation and CVD outcomes using large-scale genome-wide association study (GWAS) data.
## Study design
We conducted a two-sample MR study using data from the publicly available FinnGen (https://www.finngen.fi/en) and the GWAS summary data (https://gwas.mrcieu.ac.uk/). Informed consent and ethical approval were provided in the original publications and these publicly available databases. This MR analysis was founded on three critical assumptions as follows: [1] IVs must be strongly associated with constipation, [2] IVs must not be associated with confounders, and [3] IVs cannot lead to CVD unless through their effects on constipation (Figure 1) [14, 15].
**Figure 1:** *Study flow diagram. The dashed lines represent possible pleiotropic or direct causal effects between variables that might violate MR assumptions. CAD, coronary artery disease; MI, myocardial infarction; HF, heart failure; AF, atrial fibrillation; IS, ischemic stroke; CES, cardioembolic stroke; LAS, large-artery atherosclerotic stroke; SVS, small-vessel stroke; IV, instrumental variable; IVW, inverse-variance weighted; WM, weighted-median; MR, Mendelian randomization; MR-PRESSO, MR pleiotropy residual sum and outlier.*
## Data sources
Summary statistics for constipation were obtained from FinnGen with a sample size of 309,154 European individuals comprising 26,919 cases and 282,235 controls [16]. Genetic associations with CAD were obtained from a GWAS meta-analysis comprising 122,733 CAD cases and 424,528 controls from the CARDIoGRAMplusC4D consortium and UK Biobank [17]. Summary data for myocardial infarction (MI) were also derived from the CARDIoGRAMplusC4D consortium, which comprised 171,875 participants ($77\%$ for European ancestry; 43,676 MI cases and 128,199 controls) [18]. Summary-level data for HF were extracted from the HERMES consortium, including 977,323 subjects of European ancestry (47,309 HF cases and 930,014 controls) [19]. Summary statistics for AF came from a large-scale GWAS meta-analysis, including 60,620 AF cases and 970,216 controls of European ancestry [20]. Summary statistics for stroke were obtained from the MEGASTROKE consortium, which comprised 446,696 participants of European ancestry (40,585 stroke cases and 406,111 controls) [21]. 34,217 subjects were defined as having an ischemic stroke (IS) among these stroke cases. Further, ischemic stroke was divided into three subtypes, including large-artery atherosclerotic stroke (LAS, 4373 cases), small-vessel stroke (SVS, 5386 cases), and cardioembolic stroke (CES, 7193 cases). The overlapping populations did not exist between the exposures and outcomes GWASs.
## Selection of genetic instrumental variables and statistical power
First, we identified three single-nucleotide polymorphisms (SNPs) robustly associated with constipation ($p \leq 5$ × 10−8). Then, a more relaxed threshold ($p \leq 5$ × 10−6) was used to identify SNPs since the number of SNPs meeting genome-wide significance was limited. Second, to obtain independent SNPs, we collected SNPs at linkage disequilibrium (LD) r2 threshold at r2 < 0.001 and kb > 10,000 based on European ancestry reference data, which come from the 1,000 Genomes Project [22]. Third, we calculated the F-statistics to test the strength of each instrument with the following formula: F = R2 × (N−2)/(1−R2) [23], where N represents the sample size of constipation and R2 represents the proportion of variance in constipation explained by each selected SNP (calculated by the method described previously) [24, 25]. Then we selected SNPs with an F-statistic of more than 10 to prevent potential weak instrument bias. Fourth, we searched SNPs in PhenoScanner V21 to assess whether these SNPs were associated ($p \leq 1$ × 10−5) with possible horizontal pleiotropic effects or risk factors for CVD [26]. Next, we removed those SNPs with confounding traits that may influence the results. Subsequently, we extracted the remaining SNPs from the summary statistics of CAD, MI, HF, AF, stroke, and stroke subtypes. To meet the third assumption, SNPs that were significantly ($p \leq 5$ × 10−6) associated with the outcomes directly were dropped. When the specified SNP for constipation was unavailable in these outcomes data, a highly relevant SNP (r2 > 0.8) on SniPA2 was selected as a proxy. Without appropriate proxies available for those absent in these outcomes data, we then excluded them. Then, We excluded SNPs being palindromic based on the allele frequency after harmonizing the SNPs-exposure and SNPs-outcome. Finally, we performed MR-pleiotropy residual sum and outlier (MR-PRESSO) before MR analysis to discard any outliers with potential pleiotropy to guarantee the liability of MR estimates [27]. The remaining SNPs were finally utilized as genetic instruments following the abovementioned steps. Statistical power was calculated with an online tool available at https://shiny.cnsgenomics.com/mRnd/ [28].
## Mendelian randomization analyses
Three MR analytical methods were conducted to assess the causal effects of constipation on CVD in this study to avoid the influence of potential pleiotropic effects of genetic variants. The primary MR analysis was conducted by the random-effects inverse-variance weighted (IVW) method, which combines the Wald ratio estimates of each SNP on the outcome to gain a pooled causal estimate and provides the highest statistical power. For random-effect IVW, it permits that all the instruments are ineffective on the condition that overall horizontal pleiotropy is balanced [29]. Furthermore, another two MR analyses, weighted median (WM) and MR-Egger, were implemented as complements to detect the causality. The weighted median method can generate unbiased causal estimates on the condition that at least $50\%$ of the weight comes from valid instrumental variables [30]. The MR-Egger method provides consistent estimates accounting for pleiotropy on the condition that all the instruments are invalid, although with the lowest power [31]. Our MR estimates of the risk of CVD were presented as odds ratio (OR, $95\%$ confidence interval [CI]) per-1-log unit increase in the risk of constipation. A two-sided value of $p \leq 0.05$ were deemed as suggestive significance and associations with p-values <0.0056 (Bonferroni correction $$p \leq 0.05$$/9 outcomes) were considered statistically significant.
## Sensitivity analysis
Sensitivity analysis was conducted to detect the existence of horizontal pleiotropy, which violated the main MR assumptions. Thus, we perform Cochran’s Q test, MR-Egger intercept tests, MR-PRESSO, MR Steiger test, leave-one-out (LOO) analyses, and funnel plot to examine the presence of pleiotropy to evaluate the robustness of the results. Specifically, the Cochran Q test was applied to evaluate the heterogeneity, which was detected if the p value was less than 0.05. Horizontal pleiotropy was appraised by estimating the intercept term derived from MR-Egger regression, indicating potential bias with the intercept term difference from 0 [31]. MR Steiger test was applied to estimate the potential reverse causal association between CVD and constipation [32]. The LOO analysis was performed to detect any pleiotropy driven by a single SNP.
All these MR analyses were performed using the TwoSampleMR package (version 0.5.6) in R Version 4.2.1.
## Genetic instruments selected in Mendelian randomization
In this study, we obtained 20 SNPs associated with constipation, which met the universally accepted genomewide significance threshold ($p \leq 5$ × 10−6, r2 < 0.001, kb = 10,000) for exposure (Supplementary Table S1). One SNP (rs2130630) in constipation was removed to eliminate confounding factors associated with body mass index. Furthermore, estimates of the F-statistic suggested that no weak instrument was employed in our MR analysis (all F-statistics >10) (Supplementary Table S1). No relevant proxy SNPs were identified to replace the small number of SNPs (2–4) absent in different CVD GWASs data. After removing outliers identified by MR-PRESSO and excluding ambiguous and palindromic SNPs through harmonizing processes, the remaining SNPs were selected as instrumental variables. Details of instrumental variables of each CVD were exhibited in Supplementary Tables S2–S10. In the present study, given a type I error of $5\%$ and a statistical power of 0.80, the minimum detectable ORs for the 9 CVDs ranged from 1.31 to 2.41.
## Causal effects of constipation on CVD
IVW analysis showed that constipation was suggestively associated with the risk of AF (OR = 1.07, $95\%$ CI 1.01–1.14; $$p \leq 0.016$$). The results from WM and MR-Egger indicated a nonsignificant but consistent direction (Figure 2). There was no significant or suggestive association between genetic liability to constipation and the risk of CAD, MI, HF, stroke, or stroke subtypes (all $p \leq 0.05$). The results were consistent with IVW, WM, and MR-Egger (Figure 2). However, we may not have reached sufficient statistical power to detect such weak associations.
**Figure 2:** *Causal effects for constipation on CVD. MR-Egger, weighted median (WM), and inverse-variance weighted (IVW) estimates of Mendelian randomization (MR) are summarised. CI, confidence interval; nSNP, number of single nucleotide polymorphism; OR, odds ratio. See Figure 1 for other abbreviations.*
## Sensitivity analyses
To evaluate the robustness of the results, several sensitivity analyses, consisting of Cochran’s Q test, MR-PRESSO global test, MR Steiger test, and MR Egger intercept test, were conducted (Table 1). All p values were > 0.05 in the MR-PRESSO global tests and the MR-Egger intercept tests, manifesting that no horizontal pleiotropy existed across the analyses. MR Steiger test identified no evidence of reverse causality, and the causal direction was reliable. Nevertheless, heterogeneity was detected in Cochran’s Q test analysis between constipation and HF ($Q = 24.63$, $$p \leq 0.04$$), constipation and SVS ($Q = 27.47$, $$p \leq 0.04$$). However, the detected heterogeneity in certain results did not invalidate the MR estimates because the random-effect IVW used in this study could balance the pooled heterogeneity. Aside from that, the MR-Egger intercepts did not reveal any pleiotropy, which suggests that MR estimates were not biased by heterogeneity (Supplementary Figures S1–S9). Other analyses did not find any heterogeneity. Furthermore, after deleting 1 SNP at a time from the LOO analysis, the risk estimates did not change much, proving that no specific SNP was critical for the causal association (Supplementary Figures S10–S18). Moreover, as shown by the funnel plot, the effect size variation around the point estimate was symmetrical, meaning that horizontal pleiotropy was not apparent (Supplementary Figures S1–S9).
**Table 1**
| Outcome | Cochran Q test | Cochran Q test.1 | MR-PRESSO | MR-Egger | MR-Egger.1 | MR steiger test |
| --- | --- | --- | --- | --- | --- | --- |
| Outcome | Q_value | p_value | p_value | Intercept | p_value | p_value |
| CAD | 8.15 | 0.88 | 0.76 | −0.003 | 0.50 | 9.05E-40 |
| MI | 10.94 | 0.90 | 0.81 | −0.005 | 0.42 | 1.93E-41 |
| HF | 24.63 | 0.04 | 0.07 | −0.0003 | 0.96 | 2.92E-45 |
| AF | 14.48 | 0.56 | 0.60 | 0.003 | 0.48 | 1.96E-55 |
| Stroke | 19.93 | 0.22 | 0.25 | 0.003 | 0.63 | 3.40E-35 |
| IS | 15.19 | 0.44 | 0.47 | 0.003 | 0.61 | 4.02E-34 |
| LAS | 16.08 | 0.45 | 0.51 | 0.010 | 0.56 | 1.02E-15 |
| SVS | 27.47 | 0.04 | 0.06 | 0.005 | 0.83 | 1.01E-16 |
| CES | 20.58 | 0.15 | 0.19 | −0.003 | 0.83 | 1.57E-18 |
## Discussion
Constipation incidence varies from 3 to $79\%$ in diverse adult groups, depending on age, gender, and definition of constipation [33, 34]. Though constipation has imposed an immense social and economic burden, limited attention was paid to it in the medical field, which causes its genesis and physiopathology to be poorly elucidated [35]. Consequently, there is little information about the potential relationship between constipation and cardiovascular risk. Though a few researchers have examined the relationship between constipation and CVD, the majority of them only concentrate on stroke and CAD [4, 5, 36, 37]. Meanwhile, the results from previous works of literature are confined to observational correlations, and reverse causality may be unavoidable.
Utilizing the available large-scale GWAS data, we adopted MR which is a time- and labor-saving way to examine the association between constipation and the risk of nine CVDs. Intriguingly, genetically determined constipation is suggestively associated with enhanced AF risk was revealed. No clear pattern of associations of genetically determined constipation with the risk of CAD, MI, HF, stroke, or stroke subtypes was found. Five sensitivity tests revealed that causal effects were not caused by outliers, horizontal pleiotropy, or reverse causality. To the best of our knowledge, this is the first MR study to estimate the causal association between constipation and CVDs.
Our finding that there is a suggestive causal link between constipation and the risk of AF concurs with a Danish population-based matched cohort study that found a $27\%$ higher risk of AF in those with constipation [3]. However, the mechanisms underlying this association are unidentified. Gut microbiota imbalance caused by constipation may be one of the possible explanations for the causality. Studies have revealed that the gut microbiota of those with constipation differs from those of healthy individuals [38, 39]. Unbalanced gut microbiota may cause the intestinal mucosal barrier to disrupt, resulting in inflammation, cytokine release, and immune suppression (40–42). On the flip side, gut microbiota dysbiosis may be made worse by higher levels of inflammation, leading to aberrant bowel function and the ensuing major chronic illnesses, such as AF [43]. Interestingly, Zhang et al. provided solid evidence that gut microbiota dysbiosis directly contributes to the pathophysiology of AF by raising the levels of circulating LPS and glucose and activating the atrial NLRP3 inflammasome [44]. Besides, increased blood pressure linked to gut microbiota dysbiosis may also lead to AF [45]. Additionally, oxidative stress and constipation-induced anxiety may be another link between constipation and AF [7, 46]. Nevertheless, the causal role of constipation in AF needs to be interpreted cautiously and future study is warranted to investigate the potential mechanisms.
In the last decade, several observational studies that explored the relationship between chronic constipation and CAD, MI, HF, stroke, and its subtypes yielded contradictory results (3–5, 8, 36, 37). Elena et al. conducted a cohort study in postmenopausal women, revealing that only the severe constipation group was associated with an increased risk of cardiovascular events, including CAD, MI, and stroke and its subtypes [4]. In patients from US veterans, Keiichi et al. demonstrated that patients with constipation and patients using laxatives experienced a similarly higher risk of CAD and ischemic stroke [5]. However, Yasuhiko et al. reported that the risk of constipation on CAD and stroke would no longer be statistically significant after adjusting for potential confounding variables [36]. Our MR analysis does not provide evidence of the causal effects of constipation on CAD, MI, HF, stroke, or stroke subtypes, suggesting that associations observed clinically are likely to be biased. Thus, further studies are needed to clarify whether there are driving factors that account for bias or confounding in previous observational studies.
The IVW method generally has significantly greater statistical power than the other MR approaches, particularly MR-Egger [47]. Therefore, in most cases, IVW was used as the primary method for identifying potentially significant outcomes. Other MR methods and sensitivity analyses were conducted to ensure that IVW estimates were robust. Our study also used the consistent beta direction requirement in all MR approaches, as is the case for most MR analyses [48, 49].
This study possesses several strengths. The major strength is the MR design we used for evaluating independent causal effects of constipation on multiple CVD outcomes without interference from reverse causality or residual confounding. Additionally, we used the most significant GWAS to reduce the “winner’s curse,” even though some causal estimates were relatively small. Another advantage is the magnitude of the sample size, which allowed us to perform an adequately powered MR analysis.
The current study also has several drawbacks. Firstly, even though we used the biggest GWAS on constipation, only a small number of SNPs conform to genome-wide significance, which can result in the use of weak genetic instruments. To remedy this, we eased the statistical threshold ($p \leq 5$ × 10−6) to provide additional SNPs whose F-statistics are all above 10. When bigger GWAS numbers become available, further study will be needed to corroborate our findings. Secondly, it is challenging to completely rule out pleiotropy since the biological functions of the chosen SNPs are still unclear. However, given that we cannot discover any horizontal pleiotropy in our research, it is reassuring that the causal effect estimates were robust through various MR models and sensitivity analyses depending on different assumptions. Thirdly, the statistical power of this study may be insufficient since only $0.1\%$ of the variance in constipation was explained by IVs. Therefore, we should be cautious with interpreting the negative results; the null association might be due to a lack of power. Fourthly, since the GWAS used in our study derives from participants of European ancestry, the findings cannot be generalized to other ethnic groups. Due to these limitations, future studies are needed to confirm the causality and investigate potential mechanisms, which is compulsory for making pertinent clinical suggestions.
## Conclusion
This MR study demonstrated suggestive causal associations of constipation on AF, despite no associations achieving a significance value after multiple testing corrections. There was no evidence of an association between constipation and the risk of CAD, MI, HF, stroke, or stroke subtypes.
## Data availability statement
The original contributions presented in the study are included in the article/Supplementary material, further inquiries can be directed to the corresponding author.
## Ethics statement
Ethical approval and written informed consent were provided in the original publications and these publicly available databases.
## Author contributions
QD and JJ designed this study and drafted the manuscript. QD, DC, YZ, YX, LY, and JJ contributed to the data acquisition and data analysis. All authors contributed to the article and approved the submitted version.
## Funding
This study was supported by the National Natural Science Foundation of China (grant 82170332).
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fcvm.2023.1080982/full#supplementary-material
## References
1. Evans MA, Sano S, Walsh K. **Cardiovascular disease, aging, and clonal hematopoiesis**. *Annu Rev Pathol* (2020) **15** 419-38. DOI: 10.1146/annurev-pathmechdis-012419-032544
2. Roger VL, Sidney S, Fairchild AL, Howard VJ, Labarthe DR, Shay CM. **Recommendations for cardiovascular health and disease surveillance for 2030 and beyond: a policy statement from the American Heart Association**. *Circulation* (2020) **141** e104-19. DOI: 10.1161/CIR.0000000000000756
3. Sundbøll J, Szépligeti SK, Adelborg K, Szentkúti P, Gregersen H, Sørensen HT. **Constipation and risk of cardiovascular diseases: a Danish population-based matched cohort study**. *BMJ Open* (2020) **10** e037080. DOI: 10.1136/bmjopen-2020-037080
4. Salmoirago-Blotcher E, Crawford S, Jackson E, Ockene J, Ockene I. **Constipation and risk of cardiovascular disease among postmenopausal women**. *Am J Med* (2011) **124** 714-23. DOI: 10.1016/j.amjmed.2011.03.026
5. Sumida K, Molnar MZ, Potukuchi PK, Thomas F, Lu JL, Yamagata K. **Constipation and risk of death and cardiovascular events**. *Atherosclerosis* (2019) **281** 114-20. DOI: 10.1016/j.atherosclerosis.2018.12.021
6. Bassotti G. **Understanding constipation treatment: do we need to strain to obtain better results?**. *Expert Opin Drug Metab Toxicol* (2013) **9** 387-9. DOI: 10.1517/17425255.2013.773974
7. Ishiyama Y, Hoshide S, Mizuno H, Kario K. **Constipation-induced pressor effects as triggers for cardiovascular events**. *J Clin Hypertens* (2019) **21** 421-5. DOI: 10.1111/jch.13489
8. Honkura K, Tomata Y, Sugiyama K, Kaiho Y, Watanabe T, Zhang S. **Defecation frequency and cardiovascular disease mortality in Japan: the Ohsaki cohort study**. *Atherosclerosis* (2016) **246** 251-6. DOI: 10.1016/j.atherosclerosis.2016.01.007
9. Smith GD, Ebrahim S. **‘Mendelian randomization’: can genetic epidemiology contribute to understanding environmental determinants of disease?**. *Int J Epidemiol* (2003) **32** 1-22. DOI: 10.1093/ije/dyg070
10. Sekula P, Del Greco MF, Pattaro C, Köttgen A. **Mendelian randomization as an approach to assess causality using observational data**. *J Am Soc Nephrol* (2016) **27** 3253-65. DOI: 10.1681/ASN.2016010098
11. Richmond RC, Davey SG. **Mendelian randomization: concepts and scope**. *Cold Spring Harb Perspect Med* (2022) **12** a040501. DOI: 10.1101/cshperspect.a040501
12. Burgess S, Thompson SG. **Multivariable Mendelian randomization: the use of pleiotropic genetic variants to estimate causal effects**. *Am J Epidemiol* (2015) **181** 251-60. DOI: 10.1093/aje/kwu283
13. Yarmolinsky J, Relton CL, Lophatananon A, Muir K, Menon U, Gentry-Maharaj A. **Appraising the role of previously reported risk factors in epithelial ovarian cancer risk: a Mendelian randomization analysis**. *PLoS Med* (2019) **16** e1002893. DOI: 10.1371/journal.pmed.1002893
14. Burgess S, Scott RA, Timpson NJ, Davey Smith G, Thompson SG. **Using published data in Mendelian randomization: a blueprint for efficient identification of causal risk factors**. *Eur J Epidemiol* (2015) **30** 543-52. DOI: 10.1007/s10654-015-0011-z
15. Evans DM, Davey SG. **Mendelian randomization: new applications in the coming age of hypothesis-free causality**. *Annu Rev Genomics Hum Genet* (2015) **16** 327-50. DOI: 10.1146/annurev-genom-090314-050016
16. Mitja I, Kurki JK, Palta P, Sipilä TP, Kristiansson K, Donner K. **Finngen: unique genetic insights from combining isolated population and national health register data**. *MedRxiv* (2022). DOI: 10.1101/2022.03.03.22271360
17. van der Harst P, Verweij N. **Identification of 64 novel genetic loci provides an expanded view on the genetic architecture of coronary artery disease**. *Circ Res* (2018) **122** 433-43. DOI: 10.1161/CIRCRESAHA.117.312086
18. Nikpay M, Goel A, Won H-H, Hall LM, Willenborg C, Kanoni S. **A comprehensive 1,000 genomes-based genome-wide association meta-analysis of coronary artery disease**. *Nat Genet* (2015) **47** 1121-30. DOI: 10.1038/ng.3396
19. Shah S, Henry A, Roselli C, Lin H, Sveinbjörnsson G, Fatemifar G. **Genome-wide association and Mendelian randomisation analysis provide insights into the pathogenesis of heart failure**. *Nat Commun* (2020) **11** 163. DOI: 10.1038/s41467-019-13690-5
20. Nielsen JB, Thorolfsdottir RB, Fritsche LG, Zhou W, Skov MW, Graham SE. **Biobank-driven genomic discovery yields new insight into atrial fibrillation biology**. *Nat Genet* (2018) **50** 1234-9. DOI: 10.1038/s41588-018-0171-3
21. Malik R, Chauhan G, Traylor M, Sargurupremraj M, Okada Y, Mishra A. **Multiancestry genome-wide association study of 520,000 subjects identifies 32 loci associated with stroke and stroke subtypes**. *Nat Genet* (2018) **50** 524-37. DOI: 10.1038/s41588-018-0058-3
22. Abecasis GR, Altshuler D, Auton A, Brooks LD, Durbin RM, Gibbs RA. **A map of human genome variation from population-scale sequencing**. *Nature* (2010) **467** 1061-73. DOI: 10.1038/nature09534
23. Burgess S, Thompson SG. **Avoiding bias from weak instruments in Mendelian randomization studies**. *Int J Epidemiol* (2011) **40** 755-64. DOI: 10.1093/ije/dyr036
24. Shim H, Chasman DI, Smith JD, Mora S, Ridker PM, Nickerson DA. **A multivariate genome-wide association analysis of 10 Ldl subfractions, and their response to statin treatment, in 1868 Caucasians**. *PLoS One* (2015) **10** e0120758. DOI: 10.1371/journal.pone.0120758
25. Chen D, Zhang Y, Yidilisi A, Xu Y, Dong Q, Jiang J. **Causal associations between circulating Adipokines and cardiovascular disease: a Mendelian randomization study**. *J Clin Endocrinol Metab* (2022) **107** e2572-80. DOI: 10.1210/clinem/dgac048
26. Kamat MA, Blackshaw JA, Young R, Surendran P, Burgess S, Danesh J. **Phenoscanner V2: an expanded tool for searching human genotype-phenotype associations**. *Bioinformatics* (2019) **35** 4851-3. DOI: 10.1093/bioinformatics/btz469
27. Verbanck M, Chen C-Y, Neale B, Do R. **Detection of widespread horizontal pleiotropy in causal relationships inferred from Mendelian randomization between complex traits and diseases**. *Nat Genet* (2018) **50** 693-8. DOI: 10.1038/s41588-018-0099-7
28. Brion M-JA, Shakhbazov K, Visscher PM. **Calculating statistical power in Mendelian randomization studies**. *Int J Epidemiol* (2013) **42** 1497-501. DOI: 10.1093/ije/dyt179
29. Pierce BL, Burgess S. **Efficient Design for Mendelian Randomization Studies: subsample and 2-sample instrumental variable estimators**. *Am J Epidemiol* (2013) **178** 1177-84. DOI: 10.1093/aje/kwt084
30. Bowden J, Davey Smith G, Haycock PC, Burgess S. **Consistent estimation in Mendelian randomization with some invalid instruments using a weighted median estimator**. *Genet Epidemiol* (2016) **40** 304-14. DOI: 10.1002/gepi.21965
31. Bowden J, Davey Smith G, Burgess S. **Mendelian randomization with invalid instruments: effect estimation and bias detection through egger regression**. *Int J Epidemiol* (2015) **44** 512-25. DOI: 10.1093/ije/dyv080
32. Hemani G, Tilling K, Davey SG. **Orienting the causal relationship between imprecisely measured traits using Gwas summary data**. *PLoS Genet* (2017) **13** e1007081. DOI: 10.1371/journal.pgen.1007081
33. Pare P, Ferrazzi S, Thompson WG, Irvine EJ, Rance L. **An epidemiological survey of constipation in Canada: definitions, rates, demographics, and predictors of health care seeking**. *Am J Gastroenterol* (2001) **96** 3130-7. DOI: 10.1111/j.1572-0241.2001.05259.x
34. Peppas G, Alexiou VG, Mourtzoukou E, Falagas ME. **Epidemiology of constipation in Europe and Oceania: a systematic review**. *BMC Gastroenterol* (2008) **8** 5. DOI: 10.1186/1471-230X-8-5
35. Leung FW. **Etiologic factors of chronic constipation: review of the scientific evidence**. *Dig Dis Sci* (2007) **52** 313-6. DOI: 10.1007/s10620-006-9298-7
36. Kubota Y, Iso H, Tamakoshi A. **Bowel movement frequency, laxative use, and mortality from coronary heart disease and stroke among Japanese men and women: the Japan collaborative cohort (Jacc) study**. *J Epidemiol* (2016) **26** 242-8. DOI: 10.2188/jea.JE20150123
37. Ma W, Li Y, Heianza Y, Staller KD, Chan AT, Rimm EB. **Associations of bowel movement frequency with risk of cardiovascular disease and mortality among us women**. *Sci Rep* (2016) **6** 33005. DOI: 10.1038/srep33005
38. Ohkusa T, Koido S, Nishikawa Y, Sato N. **Gut microbiota and chronic constipation: a review and update**. *Front Med* (2019) **6** 19. DOI: 10.3389/fmed.2019.00019
39. Zhao Y, Yu Y-B. **Intestinal microbiota and chronic constipation**. *Springerplus* (2016) **5** 1130. DOI: 10.1186/s40064-016-2821-1
40. Kasahara K, Rey FE. **The emerging role of gut microbial metabolism on cardiovascular disease**. *Curr Opin Microbiol* (2019) **50** 64-70. DOI: 10.1016/j.mib.2019.09.007
41. Tremaroli V, Bäckhed F. **Functional interactions between the gut microbiota and host metabolism**. *Nature* (2012) **489** 242-9. DOI: 10.1038/nature11552
42. Alagiakrishnan K, Halverson T. **Holistic perspective of the role of gut microbes in diabetes mellitus and its management**. *World J Diabetes* (2021) **12** 1463-78. DOI: 10.4239/wjd.v12.i9.1463
43. Zheng J, Hoffman KL, Chen J-S, Shivappa N, Sood A, Browman GJ. **Dietary inflammatory potential in relation to the gut microbiome: results from a cross-sectional study**. *Br J Nutr* (2020) **124** 931-42. DOI: 10.1017/S0007114520001853
44. Zhang Y, Zhang S, Li B, Luo Y, Gong Y, Jin X. **Gut microbiota Dysbiosis promotes age-related atrial fibrillation by lipopolysaccharide and glucose-induced activation of Nlrp3-inflammasome**. *Cardiovasc Res* (2022) **118** 785-97. DOI: 10.1093/cvr/cvab114
45. Yang T, Santisteban MM, Rodriguez V, Li E, Ahmari N, Carvajal JM. **Gut Dysbiosis is linked to hypertension**. *Hypertension* (2015) **65** 1331-40. DOI: 10.1161/HYPERTENSIONAHA.115.05315
46. Vermorken A, Andrès E, Cui Y. **Letter: chronic constipation - a warning sign for oxidative stress?**. *Aliment Pharmacol Ther* (2015) **42** 385-6. DOI: 10.1111/apt.13257
47. Lin Z, Deng Y, Pan W. **Combining the strengths of inverse-variance weighting and egger regression in Mendelian randomization using a mixture of regressions model**. *PLoS Genet* (2021) **17** e1009922. DOI: 10.1371/journal.pgen.1009922
48. Venkatesh SS, Ferreira T, Benonisdottir S, Rahmioglu N, Becker CM, Granne I. **Obesity and risk of female reproductive conditions: a Mendelian randomisation study**. *PLoS Med* (2022) **19** e1003679. DOI: 10.1371/journal.pmed.1003679
49. Chen X, Hong X, Gao W, Luo S, Cai J, Liu G. **Causal relationship between physical activity, leisure sedentary behaviors and Covid-19 risk: a Mendelian randomization study**. *J Transl Med* (2022) **20** 216. DOI: 10.1186/s12967-022-03407-6
|
---
title: 'Establishment and validation of a predictive nomogram for gestational diabetes
mellitus during early pregnancy term: A retrospective study'
authors:
- Luman Li
- Quan Zhu
- Zihan Wang
- Yun Tao
- Huanyu Liu
- Fei Tang
- Song-Mei Liu
- Yuanzhen Zhang
journal: Frontiers in Endocrinology
year: 2023
pmcid: PMC9998988
doi: 10.3389/fendo.2023.1087994
license: CC BY 4.0
---
# Establishment and validation of a predictive nomogram for gestational diabetes mellitus during early pregnancy term: A retrospective study
## Abstract
### Objective
This study aims to develop and evaluate a predictive nomogram for early assessment risk factors of gestational diabetes mellitus (GDM) during early pregnancy term, so as to help early clinical management and intervention.
### Methods
A total of 824 pregnant women at Zhongnan Hospital of Wuhan University and Maternal and Child Health Hospital of Hubei Province from 1 February 2020 to 30 April 2020 were enrolled in a retrospective observational study and comprised the training dataset. Routine clinical and laboratory information was collected; we applied least absolute shrinkage and selection operator (LASSO) logistic regression and multivariate ROC risk analysis to determine significant predictors and establish the nomogram, and the early pregnancy files (gestational weeks 12–16, $$n = 392$$) at the same hospital were collected as a validation dataset. We evaluated the nomogram via the receiver operating characteristic (ROC) curve, C-index, calibration curve, and decision curve analysis (DCA).
### Results
We conducted LASSO analysis and multivariate regression to establish a GDM nomogram during the early pregnancy term; the five selected risk predictors are as follows: age, blood urea nitrogen (BUN), fibrinogen-to-albumin ratio (FAR), blood urea nitrogen-to-creatinine ratio (BUN/Cr), and blood urea nitrogen-to-albumin ratio (BUN/ALB). The calibration curve and DCA present optimal predictive power. DCA demonstrates that the nomogram could be applied clinically.
### Conclusion
An effective nomogram that predicts GDM should be established in order to help clinical management and intervention at the early gestational stage.
## Introduction
Gestational diabetes mellitus (GDM) is a universal metabolic disturbance syndrome with a complicated etiology during pregnancy; insulin resistance and pancreatic β cell failure were significant factors for the pathogenesis of the disease, which gradually leads to hyperglycemia (1–4). Hyperglycemia exposure contributes to both maternal and fetal adverse complications. The degree of dysregulation of blood glucose is highly related to the risks of obstetrical and neonatal outcomes, which include cesarean section, hypertension, preeclampsia, polyhydramnios, preterm delivery, fetal growth restriction, birth injury, and respiratory distress. In the long term, there is an increased risk of developing obesity, cardiovascular disease, and type 2 diabetes mellitus in both the mother and the offspring [5]. Multiple variables have been reported in previous research, such as age, gestational week, ethnicity, obesity, lifestyle, environment, and metabolism [6, 7]. Since the GDM etiology is complicated, the short-term and long-term outcomes are not optimistic and have profound influences, and the demand for early prediction and intervention is increasing.
Two acceptable diagnosis methods that are acknowledged by expert professional organizations such as the International Association of the Diabetes and Pregnancy Study Group (IADPSG) are one-step screening approach (currently preferred by the American Diabetes Association) and the two-step Carpenter–Coustan screening approach (recommended by the American College of Obstetricians and Gynecologists). The one-step screening method can diagnose more patients than the two-step screening method in a large randomized trial, and there is no statistical difference regarding maternal and neonatal adverse outcomes between these two methods [8]. Both methods have their own pros and cons, and each has its own cutoff threshold [9]. Due to the varying diagnostic criteria, the incidence of GDM varies from $3\%$ to $21.2\%$ in Asia and from $0.31\%$ to $18\%$ globally, and the prevalence continues to rise (10–12). The WHO recommended a 75-g anhydrous glucose load screening test for diagnosis after 8–14 h overnight fasting at 24–28 gestational weeks [13]. Because pregnant women undergo the oral glucose tolerance test (OGTT) at the second stage of the trimester, early warning signs for dysglycemia may be missed.
Our study aims at establishing a nomogram to predict the risk factors of GDM during early pregnancy term and to apply early intervention. Early management and intervention of GDM improves maternal and perinatal outcomes [14, 15]. Prediction models can correctly identify GDM at early gestational weeks and could mostly benefit women with targeted risk factors, which helps them focus on precision lifestyle changes. These models can be used as tools to identify risk factors and stratify diseases, which can be largely applied in clinical management and treatment [16]. Using statistical modeling combined with clinical variables and laboratory information, we developed prediction tools for GDM, which can be applied in early gestational weeks.
## Data collection
This study is a retrospective study that recruited 1,216 pregnant women at Zhongnan Hospital of Wuhan University and Maternal and Child Health Hospital of Hubei Province from 1 February to 30 April 2020. A total of 824 pregnant women in the second and third trimesters were enrolled in the training dataset, and their clinical and laboratory data during their 12th–16th gestational weeks were retrospectively collected. We also recruited 392 pregnant women during early pregnancy as a validation dataset. We collected the following maternal clinical and laboratory information: age, gestational week, gravidity and parity history, white blood cell (WBC), red blood cell (RBC), platelet (PLT), prothrombin time (PT), activated partial thromboplastin time (APTT), thrombin time (TT), fibrinogen (FIB), total protein (TP), albumin (ALB), alkaline phosphatase (ALP), blood urea nitrogen (BUN), creatinine (Cr), uric acid (UA), fibrinogen-to-albumin ratio (FAR), blood urea nitrogen-to-creatinine ratio (BUN/Cr), and blood urea nitrogen-to-albumin (BUN/ALB). All blood samples were collected by skilled nurses, and the blood tests were taken in the laboratory of Zhongnan Hospital of Wuhan University and Maternal and Child Health Hospital of Hubei Province. The levels of these factors were measured by commercial diagnostic kits: RBC, PLT, and PLT (DXH800, UniCel automated hematology analyzer, USA); PT, APTT, and FIB (CA1500, Sysmex coagulation analyzer, USA); and TT, TP, ALP, ALB, BUN, Crea, and UA (AU5800, Beckman biochemical analyzer, USA). Inclusion criteria were as follows: [1] GDM patients with confirmed diagnosis of GDM based on the 75-g OGTT test (2010 IADPSG criteria [17]; cutoff threshold values: 0 h fasting plasma glucose ≥ 5.1 mmol/L, 1 h plasma glucose ≥ 10.0 mmol/L, and 2 h plasma glucose ≥ 8.5 mmol/L) and normal pregnant women with no coexisting diseases and complications; [2] singleton pregnancy; and [3] age between 18 and 45 years. Exclusion criteria were as follows: [1] history of obstetric abnormality history, tumor, coinfection, and blood diseases; [2] presence of inflammation, cardiovascular, metabolic, immune, and endocrine diseases; and [3] type 1 diabetes mellitus or type 2 diabetes mellitus which were diagnosed before pregnancy. The details of our study process are depicted in the flowchart in Figure 1. LASSO logistic regression and multivariate ROC risk analysis were applied to establish significant factors and establish the nomogram. The early pregnancy files (gestational weeks 12–16, $$n = 392$$) were collected as a validation dataset. AUC, C-index, calibration curve, and DCA were used to evaluate the nomogram. The risk factor “age”, which acts as a continuous variable, has a poor predictive value; according to multivariate logistic regression and clinical meaning, we select the cutoff value of “30” to divide “age” as a categorical variable.
**Figure 1:** *Flowchart of this study. A total of 1,216 pregnant women enrolled in our study were selected by inclusion criteria. The training dataset (n = 824) was used to estimate a GDM predictive model, and our study applied LASSO logistic regression and multivariate ROC risk analysis to determine significant predictors and establish the nomogram. A total of 392 early pregnant women were used as a validation dataset. We evaluate the nomogram by AUC, C-index, calibration curve, and decision curve analysis (DCA).*
## Statistical methods
Statistical analysis was performed with SPSS 24.0 and R 4.0.0 software (R Statistical Computing Foundation, Vienna, Austria). Continuous data were expressed as mean ± standard deviation. Clinical characteristics were compared using t-test (continuous variables) and χ 2 test (categorical variables). LASSO regression was used to select the best predictive factors [18]. The nomogram was established as a result of the binary logistic regression model with fivefold cross-validation. Selected factors applied in the nomogram fit the following: selected by multivariable analysis and clinically relevant. The calibration curve was applied to assess the accuracy of the predictive model (the Hosmer–Lemeshow test was used to access goodness of fit). The ROC curve evaluates discriminative ability by the area under the ROC curve (AUC). The DCA curve was conducted to determine the clinical utility and benefit of the nomogram. All cutoff values were determined by the total risk scores in the training cohort. Differences with p-value < 0.05 were considered statistically significant.
Based on the 10EPP rule [19], the sample size of our predictor model should be at least 170; our study sample consists of 824 women in the training dataset and 392 women in the validation dataset, and based on the sample size of our study, the power (1 − β) calculation is equal to 1.0.
## Patients’ clinical characteristics
We included 824 pregnant women in the training cohort and 392 pregnant women in the validation dataset. All the p-values of these factors are greater than 0.001, which indicates that there were no statistically significant differences between the training dataset and the validation dataset as shown in Table 1. The baseline characteristics of each dataset are presented in Table 2, in which data on non-GDM and GDM pregnant women from both datasets are shown separately. We selected the following predictive factors by logistic regression analysis: age, WBC, PLT, APTT, BUN, UA, FAR, BUN/Cr, and BUN/ALB. Then, we selected the statistically significant factors in multivariate logistic regression and clinical correlated factors to establish a predictive model, including the following factors: age, FAR, BUN, BUN/Cr, and BUN/ALB (shown in Table 3).
## Development and validation of the nomogram
Based on the factors selected from the training cohort, LASSO regression analysis was conducted to select the predictive factors from Table 1 and establish the model with factors shown in Table 2: Five of the eighteen variables were enrolled to build the predictive model (Figure 2). These selected factors showed significant statistical differences, and they were independent of each other. The “Rms” package was used to build a nomogram to establish a GDM diagnosis model; the nomogram was constructed to predict the risk of GDM during early pregnancy (Figure 3). These five variables are given in Table 3. The AUC aimed to evaluate the discrimination of the nomogram in Figure 4; the AUC value of the training dataset is 0.808, $95\%$ CI: 0.770–0.842 ($p \leq 0.05$, Figure 4A), and the AUC value of the validation dataset is 0.769, $95\%$ CI: 0.722–0.815 ($p \leq 0.05$, Figure 4B). The calibration curve was used to evaluate the predictive power shown in Figure 5. The predictive model and the validation set showed the optimal predictive degree of the fitting. The DCA demonstrated the threshold probability of the prediction model nomogram in the training and validation datasets, respectively, and it was used to evaluate the clinical effects of the nomogram more visually, which indicated that the nomogram has optimal predictive power. DCA demonstrated that the nomogram could be applied clinically (Figure 6).
**Figure 2:** *Variable selection by the LASSO binary logistic regression model. (A) eighteen variables with nonzero coefficients were selected by deriving the optimal lambda. (B) Following verification of the optimal parameter (λ) in the LASSO model, the mean squared error changes with respect to the Log (λ) value, and the vertical dotted line near Log (λ) = −4 is drawn based on 1 standard error criteria.* **Figure 3:** *Nomogram to estimate the probability of GDM. A nomogram used basic pregnancy file information to predict GDM. Find the predictor points on the uppermost point scale that correspond to each variable of the pregnant woman and add them up; the total points projected to the bottom scale indicate the probability of GDM.* **Figure 4:** *Receiver operating characteristic (ROC) curves of nomograms in the training dataset and validation dataset, respectively. (A) The AUC value of the training dataset is 0.808, 95% CI: 0.770–0.842 (p < 0.05). (B) The AUC value of the validation dataset is 0.769, 95% CI: 0.722–0.815 (p < 0.05).* **Figure 5:** *The calibration curve of the nomogram for predicting GDM in the training dataset and validation dataset, respectively. Calibration focused on the accuracy of the probability between the predictive model and the actually observed value. The y-axis represents the actual diagnosed cases of GDM, the x-axis represents the predicted risk of GDM, and the solid line represents the prediction of the training dataset (A) and the validation dataset (B).* **Figure 6:** *Decision curve analysis for the GDM risk nomogram. The y-axis estimates the net benefit, the transverse solid line represents the probability of risk that pregnant Asian women have no GDM, and the oblique solid line represents the probability of risk that pregnant Asian women have GDM. (A) Training dataset. (B) Validation dataset.*
## Discussion
GDM is defined as dysglycemia with onset or first recognition during pregnancy [20]; insulin resistance and pancreatic β cell failure have been reported to be significant factors in GDM aside from the other main causes of GDM such as maternal age, obesity, inflammation, and inadequate physical exercise [21]. GDM increases maternal and neonatal adverse effects in both short-term and long-term periods. In addition, it is necessary to identify and address the risk factors of GDM early and accurately. Tools that could accurately target these GDM predictors in early pregnant women will most likely benefit these women [22].
On the other hand, early warning and intervention during early pregnancy may prevent the adverse outcomes of GDM by controlling glucose level. The first-line treatment for GDM is medical nutrition therapy, weight management, and physical activities (23–25). to "$70\%$ to $85\%$ of women diagnosed with GDM could modify their glucose condition through targeted lifestyle changes [26]. *In* general, an early prediction model of GDM should be established, which could positively affect prevention, treatment, and prognosis.
Prior studies indicated that BUN was dose-response related with GDM during the first trimester [27]. Diabetes mellitus drives the occurrence of kidney diseases [28]. Meanwhile, kidney metabolites such as urea or other uremic components may increase the risk of diabetes [29]. BUN was considered as a kidney function marker; a high level of urea increases insulin resistance and suppresses insulin secretion, which is associated with an increased risk of incident diabetes mellitus [30]. The underlying mechanism is as follows: urea induced the production of reactive oxygen species and restrains insulin signaling by suppressing insulin receptor substrate–serine phosphorylation [31]; on the other hand, uremic metabolite accumulation impaired β-cell normal function and negatively affected glucose homeostasis [32].
Meanwhile, fibrinogen is a long-acting plasma acute-phase reactant [33], and the change in albumin level has been attributed to the changes in nutritional status; furthermore, hypoalbuminemia represents a chronic inflammatory state caused by malnutrition [33, 34]. Likewise, FAR has been proven to be a more powerful inflammatory-based prognostic predictor of overall survival than other single prognostic markers (35–37); compared with healthy pregnancies, FAR was considered to be an independent risk factor for predicting spontaneous abortion, and increased FAR levels were considered to be related to the thrombotic process in recurrent abortion [38, 39]. BUN/*Cr is* an important indicator to evaluate acute renal injury and gastrointestinal hemorrhage, and a low BUN/Cr level is associated with higher risks of total and ischemic stroke (40–42). BUN/ALB is a novel prognostic marker that has a higher predictive ability than single urea nitrogen and albumin in pneumonia and acute pulmonary embolism (43–46). Given the low cost and the abundance of laboratory offerings, and the fact that these markers provided poor clinical outcomes in previous studies, we generated a predictive nomogram of GDM through serial measures.
From an economics perspective, our study takes advantage of early pregnancy files and validates the nomogram that was set up for 824 enrolled pregnant women. In this multicenter study, we have identified five predictors, namely, age, BUN, FAR, BUN/ALB, and BUN/Cr, which were significantly associated with GDM. These five predictors are independent of each other, and research about their relationship has rarely been reported. We also developed a nomogram that could predict the incidence of GDM during early pregnancy.
Our study has strengths and limitations; this is a multicenter retrospective study with a large sample size of pregnant women, and we used an early pregnant stage dataset verified by the nomogram. The GDM predictive nomogram focused on several clinical factors which could be readily available at low cost via routine blood tests in clinical practice, and the nomogram can be performed with optimal predictive power with better combined clinical characteristics with laboratory results. This model can be widely used in less-developed and developing countries where the incidence of GDM is rapidly increasing. It provides risk assessment based on first pregnancy profiles for early detection and intervention and to control glucose level. Thus, it should be widely carried out in more basic-level hospitals. However, many factors should be considered first. We should expand the sample size via dynamic monitoring of different gestational weeks and detect more variables and risk factors during pregnancy before the model can be widely used in clinical practice.
In summary, by analyzing basic information from pregnancy files, we found five independent risk factors of GDM: age, BUN, FAR, BUN/Cr, and BUN/ALB. According to the GDM nomogram predictive model validated by the early pregnancy dataset, we could help patients’ clinical management at the early gestational stage.
## Data availability statement
The datasets presented in this article are not readily available due to the nature of this research, participants of this study did not agree for their data to be shared publicly, so supporting data is not available. Requests to access the datasets should be directed to [email protected].
## Ethics statement
The study was approved by the Institutional Ethics Committee of the Zhongnan Hospital of Wuhan University and the Maternal and Child Health Hospital of Hubei Province, the informed consent number was No. 2020072K.
## Author contributions
LL wrote the first draft of the manuscript. LL, S-ML, and YZ contributed to the conception and design of the study. YZ is the first corresponding author. QZ, ZW, and FT collected the data. LL, QZ, and HL performed the data processing and analysis. YT contributed to the critical revision of the manuscript. All authors contributed to manuscript revision, and read and approved the submitted version.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## References
1. Hod M, Kapur A, Sacks DA, Hadar E, Agarwal M, Di Renzo GC. **The international federation of gynecology and obstetrics (Figo) initiative on gestational diabetes mellitus: A pragmatic guide for diagnosis, management, and care**. *Int J Gynaecol Obstet* (2015) **131 Suppl 3**. DOI: 10.1016/s0020-7292(15)30033-3
2. Lavery JA, Friedman AM, Keyes KM, Wright JD, Ananth CV. **Gestational diabetes in the united states: Temporal changes in prevalence rates between 1979 and 2010**. *Bjog* (2017) **124**. DOI: 10.1111/1471-0528.14236
3. McIntyre HD, Catalano P, Zhang C, Desoye G, Mathiesen ER, Damm P. **Gestational diabetes mellitus**. *Nat Rev Dis Primers* (2019) **5** 47. DOI: 10.1038/s41572-019-0098-8
4. Ying W, Fu W, Lee YS, Olefsky JM. **The role of macrophages in obesity-associated islet inflammation and β-cell abnormalities**. *Nat Rev Endocrinol* (2020) **16** 81-90. DOI: 10.1038/s41574-019-0286-3
5. Kurtzhals LL, Nørgaard SK, Secher AL, Nichum VL, Ronneby H, Tabor A. **The impact of restricted gestational weight gain by dietary intervention on fetal growth in women with gestational diabetes mellitus**. *Diabetologia* (2018) **61**. DOI: 10.1007/s00125-018-4736-6
6. Wei X, Song H, Yin L, Rizzo MG, Sidhu R, Covey DF. **Fatty acid synthesis configures the plasma membrane for inflammation in diabetes**. *Nature* (2016) **539**. DOI: 10.1038/nature20117
7. Hillier TA, Pedula KL, Schmidt MM, Mullen JA, Charles MA, Pettitt DJ. **Childhood obesity and metabolic imprinting: The ongoing effects of maternal hyperglycemia**. *Diabetes Care* (2007) **30**. DOI: 10.2337/dc06-2361
8. Hillier TA, Pedula KL, Ogasawara KK, Vesco KK, Oshiro CES, Lubarsky SL. **A pragmatic, randomized clinical trial of gestational diabetes screening**. *N Engl J Med* (2021) **384** 895-904. DOI: 10.1056/NEJMoa2026028
9. **2. classification and diagnosis of diabetes: Standards of medical care in diabetes-2021**. *Diabetes Care* (2021) **44** S15-s33. DOI: 10.2337/dc21-S002
10. Yuen L, Wong VW. **Gestational diabetes mellitus: Challenges for different ethnic groups**. *World J Diabetes* (2015) **6**. DOI: 10.4239/wjd.v6.i8.1024
11. Chamberlain JJ, Doyle-Delgado K, Peterson L, Skolnik N. **Diabetes technology: Review of the 2019 american diabetes association standards of medical care in diabetes**. *Ann Intern Med* (2019) **171**. DOI: 10.7326/m19-1638
12. Lende M, Rijhsinghani A. **Gestational diabetes: Overview with emphasis on medical management**. *Int J Environ Res Public Health* (2020) **17**. DOI: 10.3390/ijerph17249573
13. Koning SH, van Zanden JJ, Hoogenberg K, Lutgers HL, Klomp AW, Korteweg FJ. **New diagnostic criteria for gestational diabetes mellitus and their impact on the number of diagnoses and pregnancy outcomes**. *Diabetologia* (2018) **61**. DOI: 10.1007/s00125-017-4506-x
14. Landon MB, Spong CY, Thom E, Carpenter MW, Ramin SM, Casey B. **A multicenter, randomized trial of treatment for mild gestational diabetes**. *N Engl J Med* (2009) **361**. DOI: 10.1056/NEJMoa0902430
15. Harrison J, Melov S, Kirby AC, Athayde N, Boghossian A, Cheung W. **Pregnancy outcomes in women with gestational diabetes mellitus by models of care: A retrospective cohort study**. *BMJ Open* (2022) **12** e065063. DOI: 10.1136/bmjopen-2022-065063
16. **Practice bulletin no. 163 summary: Screening for fetal aneuploidy**. *Obstet Gynecol* (2016) **127**. DOI: 10.1097/aog.0000000000001439
17. Metzger BE, Gabbe SG, Persson B, Buchanan TA, Catalano PA, Damm P. **International association of diabetes and pregnancy study groups recommendations on the diagnosis and classification of hyperglycemia in pregnancy**. *Diabetes Care* (2010) **33**. DOI: 10.2337/dc09-1848
18. Friedman J, Hastie T, Tibshirani R. **Regularization paths for generalized linear models**. *J Stat Softw* (2010) **33** 1-22. PMID: 20808728
19. Riley RD, Ensor J, Snell KIE, Harrell FE, Martin GP, Reitsma JB. **Calculating the sample size required for developing a clinical prediction model**. *Bmj* (2020) **368**. DOI: 10.1136/bmj.m441
20. **Classification and diagnosis of diabetes mellitus and other categories of glucose intolerance**. *Diabetes* (1979) **28**. DOI: 10.2337/diab.28.12.1039
21. Lekva T, Norwitz ER, Aukrust P, Ueland T. **Impact of systemic inflammation on the progression of gestational diabetes mellitus**. *Curr Diabetes Rep* (2016) **16** 26. DOI: 10.1007/s11892-016-0715-9
22. White SL, Lawlor DA, Briley AL, Godfrey KM, Nelson SM, Oteng-Ntim E. **Early antenatal prediction of gestational diabetes in obese women: Development of prediction tools for targeted intervention**. *PloS One* (2016) **11** e0167846. DOI: 10.1371/journal.pone.0167846
23. Rasmussen L, Poulsen CW, Kampmann U, Smedegaard SB, Ovesen PG, Fuglsang J. **Diet and healthy lifestyle in the management of gestational diabetes mellitus**. *Nutrients* (2020) **12**. DOI: 10.3390/nu12103050
24. Yamamoto JM, Kellett JE, Balsells M, García-Patterson A, Hadar E, Solà I. **Gestational diabetes mellitus and diet: A systematic review and meta-analysis of randomized controlled trials examining the impact of modified dietary interventions on maternal glucose control and neonatal birth weight**. *Diabetes Care* (2018) **41**. DOI: 10.2337/dc18-0102
25. **14. management of diabetes in pregnancy: Standards of medical care in diabetes-2020**. *Diabetes Care* (2020) **43**. DOI: 10.2337/dc20-S014
26. **13. management of diabetes in pregnancy**. *Diabetes Care* (2017) **40**. DOI: 10.2337/dc17-S016
27. Feng P, Wang G, Yu Q, Zhu W, Zhong C. **First-trimester blood urea nitrogen and risk of gestational diabetes mellitus**. *J Cell Mol Med* (2020) **24**. DOI: 10.1111/jcmm.14924
28. Lai S. **Chronic kidney disease and diabetes-a potential causal link**. *EBioMedicine* (2016) **6**. DOI: 10.1016/j.ebiom.2016.03.025
29. Allison SJ. **Diabetes: Urea inhibits insulin secretion in ckd**. *Nat Rev Nephrol* (2016) **12** 581. DOI: 10.1038/nrneph.2016.131
30. Xie Y, Bowe B, Li T, Xian H, Yan Y, Al-Aly Z. **Higher blood urea nitrogen is associated with increased risk of incident diabetes mellitus**. *Kidney Int* (2018) **93**. DOI: 10.1016/j.kint.2017.08.033
31. Jha JC, Banal C, Chow BS, Cooper ME, Jandeleit-Dahm K. **Diabetes and kidney disease: Role of oxidative stress**. *Antioxid Redox Signal* (2016) **25**. DOI: 10.1089/ars.2016.6664
32. Koppe L, Nyam E, Vivot K, Manning Fox JE, Dai XQ, Nguyen BN. **Urea impairs β cell glycolysis and insulin secretion in chronic kidney disease**. *J Clin Invest* (2016) **126**. DOI: 10.1172/jci86181
33. Amaro E, Moore-Lotridge SN, Wessinger B, Benvenuti MA, An TJ, Oelsner WK. **Albumin and the fibrinogen-to-Albumin ratio: Biomarkers for the acute phase response following total knee arthroplasty**. *PloS One* (2021) **16**. DOI: 10.1371/journal.pone.0247070
34. Rothschild MA, Oratz M, Mongelli J, Schreiber SS. **Effects of a short-term fast on albumin synthesis studied in vivo, in the perfused liver, and on amino acid incorporation by hepatic microsomes**. *J Clin Invest* (1968) **47**. DOI: 10.1172/jci105941
35. Wang Y, Xu W, Wang Y. **Prognostic role of preoperative fibrinogen to albumin ratio in breast cancer**. *Clin Chim Acta* (2020) **510**. DOI: 10.1016/j.cca.2020.07.055
36. Zhang J, Ding Y, Wang W, Lu Y, Wang H, Wang H. **Combining the Fibrinogen/Albumin ratio and systemic inflammation response index predicts survival in resectable gastric cancer**. *Gastroenterol Res Pract* (2020) **2020**. DOI: 10.1155/2020/3207345
37. Xu Q, Yan Y, Gu S, Mao K, Zhang J, Huang P. **A novel inflammation-based prognostic score: The Fibrinogen/Albumin ratio predicts prognoses of patients after curative resection for hepatocellular carcinoma**. *J Immunol Res* (2018) **2018**. DOI: 10.1155/2018/4925498
38. Usta CS, Atik TK, Ozcaglayan R, Bulbul CB, Camili FE, Adali E. **Does the Fibrinogen/Albumin ratio predict the prognosis of pregnancies with abortus imminens**. *Saudi Med J* (2021) **42**. DOI: 10.15537/smj.2021.42.3.20200695
39. Cimsir MT, Yildiz MS. **Could fibrinogen to albumin ratio be a predictive marker for recurrent pregnancy loss**. *Int J Clin Pract* (2021) **75**. DOI: 10.1111/ijcp.14520
40. Lameire N, Van Biesen W, Vanholder R. **Acute renal failure**. *Lancet* (2005) **365**. DOI: 10.1016/s0140-6736(05)17831-3
41. Deng L, Wang C, Qiu S, Bian H, Wang L, Li Y. **Association between blood urea nitrogen-to-Creatinine ratio and three-month outcome in patients with acute ischemic stroke**. *Curr Neurovasc Res* (2019) **16**. DOI: 10.2174/1567202616666190412123705
42. Peng R, Liu K, Li W, Yuan Y, Niu R, Zhou L. **Blood urea nitrogen, blood urea nitrogen to creatinine ratio and incident stroke: The dongfeng-tongji cohort**. *Atherosclerosis* (2021) **333** 1-8. DOI: 10.1016/j.atherosclerosis.2021.08.011
43. Ryu S, Oh SK, Cho SU, You Y, Park JS, Min JH. **Utility of the blood urea nitrogen to serum albumin ratio as a prognostic factor of mortality in aspiration pneumonia patients**. *Am J Emerg Med* (2021) **43**. DOI: 10.1016/j.ajem.2020.02.045
44. Fang J, Xu B. **Blood urea nitrogen to serum albumin ratio independently predicts mortality in critically ill patients with acute pulmonary embolism**. *Clin Appl Thromb Hemost* (2021) **27**. DOI: 10.1177/10760296211010241
45. Küçükceran K, Ayrancı MK, Girişgin AS, Koçak S, Dündar ZD. **The role of the Bun/Albumin ratio in predicting mortality in covid-19 patients in the emergency department**. *Am J Emerg Med* (2021) **48**. DOI: 10.1016/j.ajem.2021.03.090
46. Lin Z, Zhao Y, Xiao L, Qi C, Chen Q, Li Y. **Blood urea nitrogen to serum albumin ratio as a new prognostic indicator in critical patients with chronic heart failure**. *ESC Heart Fail* (2022) **9**. DOI: 10.1002/ehf2.13825
|
---
title: 'Association between aberrant amino acid metabolism and nonchromosomal modifications
fetal structural anomalies: A cohort study'
authors:
- Huizhen Yuan
- Chang Liu
- Xinrong Wang
- Tingting Huang
- Danping Liu
- Shuhui Huang
- Zeming Wu
- Yanqiu Liu
- Peiyuan Yin
- Bicheng Yang
journal: Frontiers in Endocrinology
year: 2023
pmcid: PMC9998993
doi: 10.3389/fendo.2023.1072461
license: CC BY 4.0
---
# Association between aberrant amino acid metabolism and nonchromosomal modifications fetal structural anomalies: A cohort study
## Abstract
### Background
More than half of the cases of fetal structural anomalies have no known cause with standard investigations like karyotype testing and chromosomal microarray. The differential metabolic profiles of amniotic fluid (AF) and maternal blood may reveal valuable information about the physiological processes of fetal development, which may provide valuable biomarkers for fetal health diagnostics.
### Methods
This cohort study of singleton-pregnant women had indications for amniocentesis, including structural anomalies and a positive result from maternal serum screening or non-invasive prenatal testing, but did not have any positive abnormal karyotype or chromosomal microarray analysis results. A total of 1580 participants were enrolled between June 2021 and March 2022. Of the 1580 pregnant women who underwent amniocentesis, 294 were included in the analysis. There were 137 pregnant women in the discovery cohort and 157 in the validation cohort.
### Results
High-coverage untargeted metabolomic analysis of AF revealed distinct metabolic signatures with 321 of the 602 metabolites measured ($53\%$) (false discovery rate, q < 0.005), among which amino acids predominantly changed in structural anomalies. Targeted metabolomics identified glutamate and glutamine as novel predictive markers for structural anomalies, their vital role was also confirmed in the validation cohort with great predictive ability, and the area under the receiver operating characteristic curves (AUCs) were 0.862 and 0.894 respectively. And AUCs for glutamine/glutamate were 0.913 and 0.903 among the two cohorts.
### Conclusions
Our results suggested that the aberrant glutamine/glutamate metabolism in AF is associated with nonchromosomal modificantions fetal structural anomalies. Based on our findings, a novel screening method could be established for the nonchromosomal modificantions fetal structural anomalies. And the results also indicate that monitoring fetal metabolic conditions (especially glutamine and glutamine metabolism) may be helpful for antenatal diagnosis and therapy.
## Graphical Abstract
General view of aberrant glutamate and glutamine metabolism in fetal ultrasound anomalies.
## Introduction
Fetal structural anomalies, which can range from minor deficiencies in a single organ to severe multi-organ system malformations, have a considerable impact on fetal morbidity and mortality [1]. Prenatal ultrasound is now regarded as a routine analysis in obstetrical care, and with increasingly high resolution, fetal structural anomalies are identified in approximately $3\%$ of pregnancies. Fetal structural anomalies have various genetic causes, including chromosomal aneuploidy, copy number variations (CNVs), and pathogenic sequence variants in developmental genes [2]. Genetic investigations are essential for the assessment and clinical triage of fetal structural anomalies. Clinically, when fetal anomalies are identified, further prospective evaluations included karyotype testing and chromosomal microarray analysis (CMA) to detect aneuploidies and CNVs [3, 4]. Overall, approximately $32\%$ of fetuses with a structural anomaly identified by ultrasound have a clinically relevant abnormal karyotype, and $6.5\%$ of them have a causative CNV (1, 3–5). Additionally, where karyotype testing and CMA failed to determine the underlying cause, whole-exome sequencing was reported to identify a well-described genetic cause in 8.5-$10\%$ of fetuses with structural anomalies [2, 6]. However, more than $50\%$ of fetal structural anomalies are left without a prospectively screening or identification method.
Pregnancy is related to the onset of many adaptation processes that change throughout gestation [7]. Maternal blood constantly exchanges with the fetus’s blood through the placenta to provide the nutrients needed for fetal growth and development. Amniotic fluid (AF) can also be considered a pool of metabolites reflecting the biological process of anabolism and catabolism [8, 9]. The biochemical nature of AF and maternal blood makes them extremely valuable materials for fetal health diagnostics.
Spurred by tremendous technological advancements, the metabolome has become widely acknowledged as the dynamic and sensitive expression of biological phenotypes at the molecular level, placing metabolomics at the forefront of biomarker and mechanistic discoveries associated with pathophysiological processes [10]. Untargeted metabolomics is applied to measure the most comprehensive range of compounds or putative metabolites present in an extracted sample without prior knowledge of the metabolome [11]. In contrast, targeted metabolomics focuses on a small group (50–500) of compounds of interest; here, methods are generated and optimized for the investigation of specific metabolites and metabolic pathways with higher sensitivity and selectivity than untargeted metabolomics [12]. The targeted analysis is also outstanding for hypothesis validation and expanding upon the results of untargeted analysis [13].
Liquid chromatography-tandem mass spectrometry (LC-MS/MS) is a current, routine, highly accurate application in newborn screening [14, 15]. Similarly, metabolomics can be applied to fetal malformations by exploring the AF metabolome, and several studies have reported promising results [16, 17], revealing the possibility of using this technology in clinical practice. Since AF can reflect both maternal and fetal health, linking AF metabolic profiles with structural anomalies is conducive to biomarker discovery, and will better guide clinical practice.
The present study aimed to characterize the metabolic signature of AF in fetal structural anomalies. Also, we tried to investigate whether metabolic changes reflect maternal or fetal conditions. In this study, we measured AF metabolites in the structural anomalies and control groups from two independent cohorts using both untargeted and targeted metabolomics. First, a high-coverage untargeted metabolomic assay based on ultra-high performance liquid chromatography-tandem mass spectrometry (UHPLC-MS/MS) was applied to 137 participants (the discovery cohort). To assay the changes in metabolites more quantitatively, we performed targeted metabolomic analysis using the UHPLC-MS/MS system and isotope-labeled internal standards. The findings in the discovery cohort were confirmed by targeted metabolomic analysis of a validation cohort of 157 participants. At the same time, we analyzed maternal serum metabolites using targeted metabolomics, which reflected the amino acid metabolism of the mothers.
## Study design and participant enrollment
This study was approved by the medical ethics committee of Jiangxi Maternal and Child Health Hospital (Approval number: EC-KT-202210). All the participants provided written informed consent. All participants were recruited from the prenatal diagnosis center of Jiangxi Maternal and Child Health Hospital from June 2021 to March 2022. Inclusion criterion: Pregnant women who had an indication for amniocentesis, including structural anomalies and a positive result from maternal serum screening or non-invasive prenatal testing. Exclusion criteria: [1] abnormal karyotype or chromosomal microarray analysis results; gestational age beyond 140-154 days; [3] multiple pregnancies; [4] other risk factors for prenatal diagnoses. Finally, 294 participants were included and separated into the discovery ($$n = 137$$, from June 2021 to October 2021) and validation ($$n = 157$$, from November 2021 to March 2022) cohorts. Fetuses with structural anomalies were categorized into three phenotypic groups based on abnormalities in different organ systems detected by ultrasound, including cardiac, central nervous systems, and renal anomalies. The control group in this study included women with singleton pregnancies whose fetuses had no structural malformations, but who had indications for amniocentesis, including a positive result from maternal serum screening or non-invasive prenatal testing.
## Collection and processing of samples
20-25 mL of AF and 3-5 mL of blood were obtained from the pregnant women at the time of amniocentesis. The AF was centrifuged at 1200 rpm for 10 min at 4°C, and the supernatant was collected. Blood was placed at 4°C for 1 h and centrifuged at 3000 rpm at 4°C for 10 min, and serum was collected from the upper layer. All samples were stored at -80°C before analysis, and their use for research was approved by the ethical committee. In the validation cohort, AF and blood samples were obtained from the same pregnant woman.
## Untargeted LC-MS metabolomics profiling
Broad-based metabolomic profiling was performed using UHPLC-MS/MS platform. Further details are provided in the Supplementary Material.
## Targeted LC-MS metabolomics data collection and processing
Fifty-four amino acids and their derivatives were quantified using a Shimadzu LC-20ADXR (Shimadzu, Kyoto, Japan) coupled with a Sciex 5500+ triple quadrupole mass spectrometer (AB Sciex, Singapore). Further details are provided in the Supplementary Material.
## Statistical analysis
The metabolites included in the statistical analyses were those which were consistently detected in at least $80\%$ of the samples. The metabolome data derived from different methods were normalized. Data scaling was assessed using Pareto scaling. Multivariate statistical analyses, partial least squares discrimination (PLS-DA), functional enrichment, metabolic pathway analysis of metabolites and lipids, and receiver operating characteristic (ROC) analysis, and the respective area under the ROC curve (AUC) were performed using an online data analysis platform- MetaboAnalyst 5.0 (https://www.metaboanalyst.ca). Unit statistical analyses, such as t-tests, were performed using SPSS software (version 26.0; IBM, USA). Bar and line plots were drawn by GraphPad Prism 8.0 (GraphPad Software Inc., USA). Chemical similarity enrichment analysis was conducted using ChemRICH R package [18], and significant metabolites alterations were visualized in an enhanced heat map in gplots package using the in R (version 3.6). All p-values involved in this study were two-tailed probabilities and were adjusted by false discovery rate (FDR). Differences were considered statistically significant at FDR <0.05.
## High-coverage untargeted metabolomics analysis revealed distinct metabolic signatures with amino acids predominantly changed in the structural anomalies group
To comprehensively detect the metabolic profiles of structural anomalies, we implemented a high-coverage untargeted metabolomic analysis of AF samples by integrating five different analytical methods that could cover both hydrophobic and hydrophilic metabolomes. Between June 2021 and March 2022, 1580 pregnant women whose fetuses were diagnosed with structural anomalies were screened for their eligibility for inclusion in our study (Figure 1). Finally, 137 and 157 participants prospectively enrolled in the study as described in Table 1. The untargeted metabolomic analysis enabled the detection and relative quantification of 602 metabolites in all AF samples. As shown in the PLS-DA score plot, the structural anomalies group was separated from the control group in the direction of the first principal component (Figure 2A). Running 10-fold cross-validation showed that the accuracy of one component was 0.87 (0.54 for R2 and 0.50 for Q2) (Supplementary Table 1). Moreover, 321 metabolites were identified as differential metabolites between the structural anomalies and control groups (FDR, q <0.05) (Supplementary Table 2, Supplementary Figure 1A). Among these, differential amino acids were the most abundant (Figure 2B). The KEGG pathway enrichment analysis of these differential metabolites also showed that amino acid metabolic pathways, such as glutamine (Gln) and glutamate (Glu) metabolism; alanine, aspartate and Glu metabolism; and phenylalanine, tyrosine and tryptophan biosynthesis, were the most significant changed (Figure 2C). Among all amino acids, Gln ($32\%$ increase, FDR, q<1×10−13) and Glu ($84\%$ decrease, FDR, q<1×10−11) were the significantly the significantly altered metabolite in structural anomalies group (Figure 2D, Supplementary Table 2).
**Figure 1:** *Study outline of workflow.* TABLE_PLACEHOLDER:Table 1 **Figure 2:** *Amniotic fluid metabolic landscape for fetal ultrasound anomalies. (A) PLS-DA score plot for untargeted metabolomics data. (B) Classed enrichment analysis for differential metabolites between ultrasound anomalies group and the control group. (C) Pathway enrichment analysis for significantly different metabolites between ultrasound anomalies group and the control group. (D) Volcano plot for all metabolites from untargeted metabolomics. (E) Venn plot of differential metabolites from three kinds of ultrasound anomalies compared with the control group. (F) Chemical similarity enrichment analysis for differential metabolites from three kinds of ultrasound anomalies. CA, cardiac anomalies; CNA, central nervous system anomalies; RA, renal anomalies.*
Based on the above results, we focused on the amino acid changes among different structural anomalies, including cardiac, central nervous system, and renal system anomalies. Compared to the control group, each type of structural anomaly demonstrated a distinct metabolic profile, with 172 overlapping differential metabolites (Figure 2E). There were 14 amino acids in the 172 overlapping metabolites. Surprisingly, Glu levels were dramatically lower while Gln levels were significantly higher in the cardiac, central nervous system, and renal anomalies (Supplementary Table 2). Gln-Glu exchange is important in placental amino acid transport, and Gln and Glu are the most utilized amino acids in fetuses during late gestation. Therefore, we hypothesized that Gln and Glu are vital for the early diagnosis of fetal structural anomalies.
In addition to the significant changes in Glu metabolism in the three types of structural anomalies, it is worth noting that fetuses with renal anomalies uniquely showed significantly inhibited urea cycle (arginine and proline metabolism), and that creatine metabolism was positively regulated in fetuses with central nervous system anomalies subjects (Figure 2F). These metabolic pathway changes may be typical responses to different structural anomalies.
## Amniotic fluid-targeted metabolomics of identified glutamate and glutamine as novel predictive markers for structural anomalies
We performed a targeted metabolomic assay of 54 amino acids and their derivatives to quantify the metabolite changes in the structural anomalies and control groups more precisely. We first quantified AF amino acids obtained from 137 participants in the discovery cohort, confirming that aberrant amino acid metabolism occurred in the structural anomalies group (Supplementary Table 3). Gln and Glu were significantly altered in targeted metabolomics.
To further validated these results, we applied targeted metabolomics to the validation cohort. Based on the concentrations presented in the different groups, 33 amino acids, including Gln and Glu, showed significant differences between the structural anomalies and control groups (Supplementary Table 4). Twenty amino acids were shared by the three types of structural anomalies (cardiac, central nervous system and renal anomalies), and significant differences existed between the structural anomalies and control groups (Supplementary Figures 1B, C). We found that Glu levels in the AF were significantly lower (Figure 3A), while Gln levels were significantly higher in the structural anomalies group than in the control group (Figure 3B). Using Gln/Glu as a metric indicating Gln-Glu conversion, we found that this ratio fell approximately 14-fold on an average among participants in the structural anomalies group (Figure 3C). Notably, regardless of the types of anomaly present, the Gln/Glu ratio was significantly reduced in the structural anomalies group than in the control group (Supplementary Figure 1D). These results were consistent with our findings in the discovery cohort (Supplementary Figure 1E). In addition, Glu (AUC=0.862, $95\%$CI: 0.800-0.925), Gln (AUC=0.894, $95\%$CI: 0.838-0.950) and Gln/Glu (AUC=0.903, $95\%$CI: 0.851-0.954) had great prediction ability in distinguishing structural anomalies from the control group in the validation cohort (Figure 3D–F).
**Figure 3:** *Glutamine and Glutamate were novel predictive markers for ultrasound anomalies. (A–C) Expression of glutamine (A), glutamate (B) and glutamine/glutamate (C) in amniotic fluid of validation cohort. (D, E) ROC curves of glutamine (D), glutamate (E) and glutamine/glutamate (F) in discovery cohort (red line) and validation cohort (blue line). ****, P<0.0001.*
We then investigated whether the Gln/Glu in AF correlated with maternal metabolic conditions. Serum samples were collected from women in the validation cohort and analyzed using the same amino acid-targeted metabolomic assay. Notably, maternal serum Glu (Supplementary Figure 1F) and Gln levels (Supplementary Figure 1G) did not differ significantly between the structural anomalies and control groups. Gln/Glu ratio also did not differ between the two groups (Supplementary Figure 1H). In addition, almost all the quantified amino acids demonstrated no big differences between the structural anomalies group and control group (Supplementary Table 5), except for threonine (Supplementary Figure 1I) and leucyl-leucine (Supplementary Figure 1J). Taken together, these results suggest that changes in Gln/Glu ratio in the AF of the structural anomalies group are associated with the fetal condition rather than the maternal condition (Figure 4).
**Figure 4:** *General view of aberrant glutamate and glutamine metabolism in fetal ultrasound anomalies. NS, no significance; ****, P<0.0001.*
## Discussion
Despite the use of karyotype testing and chromosomal microarray as routine investigations in obstetric care, a large proportion of fetal structural anomalies still have no proven cause. Herein, we explored the underlying causes of fetal malformations using AF metabolomics study. First, we performed an untargeted metabolomic assay on AF samples, starting with the discovery cohort. The results demonstrated that AF metabolic signatures were remarkably altered in the structural anomalies group compared to the control group. The most apparent alterations were observed in amino acids and their derivatives. These amino acid changes were further confirmed using targeted metabolomics, and we found 23 amino acids that were differentially expressed in the three types of structural anomalies (cardiac, central nervous system, and renal anomalies). Among these amino acids, Glu and Gln were the most significantly altered metabolites. The structural anomalies group was characterized by a significantly lower Gln/Glu ratio than the control group. To strengthen this finding, the results were validated using samples from an independent validation cohort. The results of the validation cohort were consistent with those of the discovery cohort; aberrant Glu and Gln metabolism was found in fetal structural anomalies. In addition, analysis of maternal blood samples through targeted metabolomics demonstrated no significant difference in Gln/Glu ratio between the fetal structural anomalies and the control groups, suggesting that the contributors to these Glu-Gln changes in AF were closely related to fetal metabolic conditions rather than maternal metabolic status. It is also worth noting that most amino acids in maternal blood did not show significant changes in the structural anomalies and the control groups.
During pregnancy, amino acids serve as important precursors for the biosynthesis of macromolecules, including proteins and nucleotides, which are involved in fetal development and growth (19–21). Glu and Gln are among the most abundant and most utilized amino acids in the fetus during late pregnancy [19]. The human placenta mediates the net transfer of amino acids to the fetus, with amino acid concentrations being generally higher in the fetus than in the mother, indicating an active transfer process across the placenta [22, 23]. One notable exception to this process is Glu, which is the net placental uptake from the fetus [23]. To meet the acquisitive demand for nutrients, Gln, a non-essential amino acid, is essential when fetal demand for amino acids exceeds maternal supply during pregnancy [24, 25]. This demand is met through the interorgan recycling of Gln and Glu. In the fetal liver, the deamination of Gln produces Glu. Glu is transported across the syncytiotrophoblast microvillous membrane and basal membranes by high-affinity excitatory amino acid transporters and is converted to Gln in the placenta [26, 27]. Glu is also an important nitrogen resource and a precursor of γ-aminobutyric acid, a key inhibitory neurotransmitter [28, 29]. Therefore, the Glu-Gln cycle and exchange in the placenta-fetus unit likely play important roles in fetal growth and development.
In our study, the significant increase in Gln/Glu ratio in the AF observed in the fetal structural anomalies group suggested a disturbing Glu-Gln cycle in the fetus rather than in the mother, since no obvious changes were detected for either Glu or Gln in maternal blood. Decreased levels of Glu and increased levels of Gln in AF have also been reported in the studies of fetal malformations, prediagnostic gestational diabetes, preterm delivery and early rupture of membranes [16, 30]. The underlying cause may be the dysfunction of transporters utilized by Glu and Gln. The amino acids that the fetus requires for metabolic processes and biosynthesis pathways can only be obtained from the placenta and delivered by different amino acid transporters [23, 31]. For example, in fetal growth restriction, the initial rate of uptake of Gln and Glu into placental villous fragments is reportedly reduced but increases with the expression of their transporter proteins (Gln: LAT1, LAT2, SNAT5, Glu: EAAT1) [32, 33]. Transporter activity is not simply determined by the protein expression levels; it is also influenced by factors that regulate substrate levels on both sides of the membrane. Interestingly, a study demonstrated that Glu efflux down its transmembrane gradient drives placental uptake via OAT4 and OATP2B1 from the fetal circulation and that the reuptake of Glu maintains this driving gradient, although OAT4 and OATP2B1 are not currently understood Glu transporters [26].
In the group with renal anomalies, we also found inhibited urea cycle metabolism. Arginine is the precursor for the synthesis of ornithine, proline, and nitric oxide [34], detecting the levels of arginine and its metabolites may provide insight into discriminating fetal renal anomalies and monitoring fetal urinary development.
However, there are some limitations to our study. First, this study was limited to one center: the Prenatal Diagnosis Center of Jiangxi Maternal and Child Health Hospital. UHPLC-MS/MS analysis is simple and sensitive, and it uses only a small amount of AF for metabolic analysis, AF acquisition is still invasive. Additionally, details of clinical examination results were not available in our study, so the study did not reveal the correlations between changed metabolites and the clinical data.
## Data availability statement
The original contributions presented in the study are included in the article/ Supplementary Material. Further inquiries can be directed to the corresponding authors.
## Ethics statement
The studies involving human participants were reviewed and approved by The medical ethics committee of Jiangxi Maternal and Child Health Hospital. The patients/participants provided their written informed consent to participate in this study. Written informed consent was obtained from the individual(s) for the publication of any potentially identifiable images or data included in this article.
## Author contributions
BY, YL, PY conceived and designed the experiment. HY, XW, TH,DL and SH collected the samples and clinical information. HY, CL, ZW performed the experiments and analyzed the data. CL, ZW drafted the manuscript, BY, YL, PY reviewed the manuscript, and BY, PY edited the final manuscript. All authors contributed to the article and approved the submitted version.
## Conflict of interest
ZW and PY are co-founders of iphenome Yun Pu Kang biotechnology Inc. Author ZW is employed by PY.
The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fendo.2023.1072461/full#supplementary-material
## References
1. Kilby MD. **The role of next-generation sequencing in the investigation of ultrasound-identified fetal structural anomalies**. *BJOG an Int J Obstetrics Gynaecol* (2021.0) **128**. DOI: 10.1111/1471-0528.16533
2. Lord J, McMullan DJ, Eberhardt RY, Rinck G, Hamilton SJ, Quinlan-Jones E. **Prenatal exome sequencing analysis in fetal structural anomalies detected by ultrasonography (Page): A cohort study**. *Lancet* (2019.0) **393**. DOI: 10.1016/S0140-6736(18)31940-8
3. Wapner RJ, Martin CL, Levy B, Ballif BC, Eng CM, Zachary JM. **Chromosomal microarray versus karyotyping for prenatal diagnosis**. *N Engl J Med* (2012.0) **367**. DOI: 10.1056/NEJMoa1203382
4. Robson SC, Chitty LS, Morris S, Verhoef T, Ambler G, Wellesley DG. *Efficacy and mechanism evaluation. evaluation of array comparative genomic hybridisation in prenatal diagnosis of fetal anomalies: A multicentre cohort study with cost analysis and assessment of patient, health professional and commissioner preferences for array comparative genomic hybridisation*
5. Hillman SC, Pretlove S, Coomarasamy A, McMullan DJ, Davison EV, Maher ER. **Additional information from array comparative genomic hybridization technology over conventional karyotyping in prenatal diagnosis: A systematic review and meta-analysis**. *Ultrasound Obstet Gynecol* (2011.0) **37** 6-14. DOI: 10.1002/uog.7754
6. Petrovski S, Aggarwal V, Giordano JL, Stosic M, Wou K, Bier L. **Whole-exome sequencing in the evaluation of fetal structural anomalies: A prospective cohort study**. *Lancet* (2019.0) **393**. DOI: 10.1016/S0140-6736(18)32042-7
7. Lain KY, Catalano PM. **Metabolic changes in pregnancy**. *Clin Obstet Gynecol* (2007.0) **50**. DOI: 10.1097/GRF.0b013e31815a5494
8. **Amniotic fluid and its clinical significance**. *Arch Dis Child* (1981.0) **56**. DOI: 10.1136/adc.56.12.966-c
9. Underwood MA, Gilbert WM, Sherman MP. **Amniotic fluid: Not just fetal urine anymore**. *J Perinatol* (2005.0) **25**. DOI: 10.1038/sj.jp.7211290
10. Johnson CH, Ivanisevic J, Siuzdak G. **Metabolomics: Beyond biomarkers and towards mechanisms**. *Nat Rev Mol Cell Biol* (2016.0) **17**. DOI: 10.1038/nrm.2016.25
11. Wishart DS. **Metabolomics for investigating physiological and pathophysiological processes**. *Physiol Rev* (2019.0) **99**. DOI: 10.1152/physrev.00035.2018
12. Roberts LD, Souza AL, Gerszten RE, Clish CB. **Targeted metabolomics**. *Curr Protoc Mol Biol* (2012.0) 1-24. DOI: 10.1002/0471142727.mb3002s98
13. Ivanisevic J, Elias D, Deguchi H, Averell PM, Kurczy M, Johnson CH. **Arteriovenous blood metabolomics: A readout of intra-tissue metabostasis**. *Sci Rep* (2015.0) **5**. DOI: 10.1038/srep12757
14. Seger C, Salzmann L. **After another decade: Lc-Ms/Ms became routine in clinical diagnostics**. *Clin Biochem* (2020.0) **82** 2-11. DOI: 10.1016/j.clinbiochem.2020.03.004
15. Lehotay DC, Hall P, Lepage J, Eichhorst JC, Etter ML, Greenberg CR. **Lc-Ms/Ms progress in newborn screening**. *Clin Biochem* (2011.0) **44** 21-31. DOI: 10.1016/j.clinbiochem.2010.08.007
16. Graca G, Goodfellow BJ, Barros AS, Diaz S, Duarte IF, Spagou K. **Uplc-Ms metabolic profiling of second trimester amniotic fluid and maternal urine and comparison with nmr spectral profiling for the identification of pregnancy disorder biomarkers**. *Mol Biosyst* (2012.0) **8**. DOI: 10.1039/c2mb05424h
17. Fedou C, Camus M, Lescat O, Feuillet G, Mueller I, Ross B. **Mapping of the amniotic fluid proteome of fetuses with congenital anomalies of the kidney and urinary tract identifies plastin 3 as a protein involved in glomerular integrity**. *J Pathol* (2021.0) **254**. DOI: 10.1002/path.5703
18. Barupal DK, Fiehn O. **Chemical similarity enrichment analysis (Chemrich) as alternative to biochemical pathway mapping for metabolomic datasets**. *Sci Rep* (2017.0) **7** 14567. DOI: 10.1038/s41598-017-15231-w
19. Wu X, Xie C, Zhang Y, Fan Z, Yin Y, Blachier F. **Glutamate-glutamatetamine cycle and exchange in the placenta-fetus unit during late pregnancy**. *Amino Acids* (2015.0) **47** 45-53. DOI: 10.1007/s00726-014-1861-5
20. Parimi PS, Kalhan SC. **Glutamatetamine supplementation in the newborn infant**. *Semin Fetal Neonatal Med* (2007.0) **12** 19-25. DOI: 10.1016/j.siny.2006.10.003
21. Pochini L, Scalise M, Galluccio M, Indiveri C. **Membrane transporters for the special amino acid glutamatetamine: Structure/Function relationships and relevance to human health**. *Front Chem* (2014.0) **2**. DOI: 10.3389/fchem.2014.00061
22. Cetin I, Ronzoni S, Marconi AM, Perugino G, Corbetta C, Battaglia FC. **Maternal concentrations and fetal-maternal concentration differences of plasma amino acids in normal and intrauterine growth-restricted pregnancies**. *Am J Obstet Gynecol* (1996.0) **174**. DOI: 10.1016/s0002-9378(96)70609-9
23. Cleal JK, Lofthouse EM, Sengers BG, Lewis RM. **A systems perspective on placental amino acid transport**. *J Physiol* (2018.0) **596**. DOI: 10.1113/JP274883
24. Neu J. **Glutamatetamine in the fetus and critically ill low birth weight neonate: Metabolism and mechanism of action**. *J Nutr* (2001.0) **131**. DOI: 10.1093/jn/131.9.2585S
25. Tapiero H, Mathe G, Couvreur P, Tew KD. **II. glutamatetamine and glutamate**. *BioMed Pharmacother* (2002.0) **56**. DOI: 10.1016/s0753-3322(02)00285-8
26. Lofthouse EM, Brooks S, Cleal JK, Hanson MA, Poore KR, O'Kelly IM. **Glutamate cycling may drive organic anion transport on the basal membrane of human placental syncytiotrophoblast**. *J Physiol* (2015.0) **593**. DOI: 10.1113/jp270743
27. Day PE, Cleal JK, Lofthouse EM, Goss V, Koster G, Postle A. **Partitioning of glutamatetamine synthesised by the isolated perfused human placenta between the maternal and fetal circulations**. *Placenta* (2013.0) **34**. DOI: 10.1016/j.placenta.2013.10.003
28. Moores RR, Vaughn PR, Battaglia FC, Fennessey PV, Wilkening RB, Meschia G. **Glutamate metabolism in fetus and placenta of late-gestation sheep**. *Am J Physiol* (1994.0) **267**. DOI: 10.1152/ajpregu.1994.267.1.R89
29. Vaughn PR, Lobo C, Battaglia FC, Fennessey PV, Wilkening RB, Meschia G. **Glutamatetamine-glutamate exchange between placenta and fetal liver**. *Am J Physiol* (1995.0) **268**. DOI: 10.1152/ajpendo.1995.268.4.E705
30. Graca G, Duarte IF, Barros AS, Goodfellow BJ, Diaz SO, Pinto J. **Impact of prenatal disorders on the metabolic profile of second trimester amniotic fluid: A nuclear magnetic resonance metabonomic study**. *J Proteome Res* (2010.0) **9**. DOI: 10.1021/pr100815q
31. Cleal JK, Glazier JD, Ntani G, Crozier SR, Day PE, Harvey NC. **Facilitated transporters mediate net efflux of amino acids to the fetus across the basal membrane of the placental syncytiotrophoblast**. *J Physiol* (2011.0) **589**. DOI: 10.1113/jphysiol.2010.198549
32. Persson M, Cnattingius S, Villamor E, Soderling J, Pasternak B, Stephansson O. **Risk of major congenital malformations in relation to maternal overweight and obesity severity: Cohort study of 1.2 million singletons**. *BMJ* (2017.0) **357**. DOI: 10.1136/bmj.j2563
33. McIntyre KR, Vincent KMM, Hayward CE, Li X, Sibley CP, Desforges M. **Human placental uptake of glutamatetamine and glutamate is reduced in fetal growth restriction**. *Sci Rep* (2020.0) **10** 16197. DOI: 10.1038/s41598-020-72930-7
34. Wu G, Morris SM. **Arginine metabolism: Nitric oxide and beyond**. *Biochem J* (1998.0) **336** 1-17. DOI: 10.1042/bj3360001
|
---
title: 'Aspartate aminotransferase/alanine aminotransferase ratio was associated with
type 2 diabetic peripheral neuropathy in a Chinese population: A cross-sectional
study'
authors:
- Pijun Yan
- Yuru Wu
- Xiaofang Dan
- Xian Wu
- Qian Tang
- Xiping Chen
- Yong Xu
- Jianhua Zhu
- Ying Miao
- Qin Wan
journal: Frontiers in Endocrinology
year: 2023
pmcid: PMC9998996
doi: 10.3389/fendo.2023.1064125
license: CC BY 4.0
---
# Aspartate aminotransferase/alanine aminotransferase ratio was associated with type 2 diabetic peripheral neuropathy in a Chinese population: A cross-sectional study
## Abstract
### Objective
Despite previous research that focused on aspartate aminotransferase/alanine aminotransferase ratio (AAR) as predictors of type 2 diabetes mellitus (T2DM) and cardiovascular disease, there has been limited research evaluating the association between AAR and diabetic microvascular complications. This study aimed to investigate the association of AAR with diabetic peripheral neuropathy (DPN).
### Methods
A total of 1562 hospitalized patients with T2DM were divided into four groups according to AAR quartiles. The relationship between AAR and DPN and related parameters was explored by the Spearman correlation coefficients, multivariable logistic regression analysis, and receiver operating characteristic (ROC) curves.
### Results
Patients with higher AAR quartiles had higher levels of vibration perception threshold (VPT) and presence of DPN, and AAR was positively associated with VPT and presence of DPN independent of sex, age, body mass index, and diabetic duration ($P \leq 0.01$ or $P \leq 0.05$). Moreover, AAR remained significantly associated with a higher odds ratio (OR) of DPN (OR 2.413, $95\%$ confidence interval [CI] 1.081-5.386, $P \leq 0.05$) after multivariate adjustment. Additionally, the risk of presence of DPN increased progressively as AAR quartiles increased (all P for trend <0.01) in both male and female subjects, and the highest quartile of AAR of male and female subjects was respectively associated with $107.3\%$ ($95\%$ CI: 1.386-3.101; $P \leq 0.01$) and $136.8\%$ ($95\%$ CI: 1.550-3.618; $P \leq 0.01$) increased odds of DPN compared with the lower quartiles. Last, the analysis of receiver operating characteristic curves revealed that the best cutoff values for AAR to predict the presence of DPN were 0.906 (sensitivity: $70.3\%$; specificity: $49.2\%$; and area under the curve [AUC]: 0.618) and 1.402 (sensitivity: $38\%$; specificity: $81.9\%$; and AUC: 0.600) in male and female subjects, respectively.
### Conclusions
These findings suggest that the high AAR may be associated with the presence of DPN in Chinese patients with T2DM, and may be used as an additional indicator of risk of DPN.
## Introduction
Diabetic peripheral neuropathy (DPN) is the most common but usually underestimated chronic microvascular complication that first present in the distal extremities and can result in either numbness or chronic pain; and is a major risk factor for Charcot joints, diabetic foot ulcers (DFU), and limb amputation in diabetic patients [1, 2]. DPN has now been considered an increasing public health problem, owing to its close association with considerable morbidity and mortality, heavy economic burden, and compromised quality of life [1, 3]. However, the current treatment for DPN involves only symptomatic relief, and often the results are disappointing. Therefore, it is urgent to find an indicator for screening the high-risk population of DPN, resulting in early identification and, consequently, early intervention.
A number of studies have shown that type 2 diabetes mellitus (T2DM) is an independent risk factor for the development of nonalcoholic fatty liver disease (NAFLD) and progression to liver fibrosis and cirrhosis [4, 5]. Also, NAFLD and liver fibrosis have been reported to play an important role in the presence and progression of DPN (6–8). Alanine aminotransferase (ALT) and aspartate aminotransferase (AST) were the two most common liver enzymes that reflect hepatocellular injury and death, and liver function. The concept of AST/ALT ratio (AAR) that represents the simultaneous alteration of AST and ALT levels was first put forwarded by De Ritis in 1957 [9]. Since then, AAR has been reported to be a widely used liver fibrosis marker, and in addition, an established predictive marker of liver fibrosis severity in patients with liver disease and other non-hepatic diseases [10, 11]. Besides, AAR was correlated with oxidative stress, systemic inflammation, and insulin resistance (IR) [12, 13], and implicated in the incidence and development of a wider range of cardiometabolic diseases, including metabolic syndrome (MetS) and its components including obesity, hyperglycemia or T2DM, hypertension, and hyperlipidemia, NAFLD as a hepatic manifestation of MetS, peripheral artery disease (PAD), arteriosclerosis, arterial stiffness, stroke, and cardiovascular diseases (CVD) (14–19), all of which have been proved to be closely associated with diabetic microvascular complications (20–22). Considering the strong interrelationship between diabetic microvascular complications and above-mentioned cardiometabolic diseases and the important role of liver fibrosis in diabetic microvascular complications [7, 23, 24], it is reasonable to hypothesize that T2DM individuals with high AAR would have a high risk for diabetic microvascular complications. Indeed, two clinical studies suggested that high AAR was an independent risk factor for diabetic nephropathy (DN), and was associated with more severe renal pathologic lesions and worse renal function [12, 25]. As far as we are aware, the relationship of the AAR with DPN, however, has never been determined, and the underpinning mechanisms are less well understood.
Therefore, this cross-sectional study was conducted to investigate the relationship between AAR and risk of presence of DPN in Chinese adults with T2DM. Moreover, the possible mechanisms were explored by analyzing the potential relationships among AAR and metabolic and vascular parameters, and inflammation and oxidative stress markers.
## Study population
A total of 3514 confirmed or newly diagnosed T2DM inpatients aged 18–89 years between August 2012 and September 2015, who were admitted to the Department of Endocrinology at the Affiliated Hospital of Southwest Medical University for screening of diabetic chronic complications and to optimize their anti-diabetic regimen, were initially recruited. T2DM was diagnosed based on the 1999 World Health *Organization criteria* [26]. Subjects were excluded if they had any of the following criteria: 1) other types of diabetes other than T2DM, severe DFU (grades III-V according to the Wagner classification) or previous amputation, recent acute complications of diabetes, including diabetic ketoacidosis, hyperglycemic hyperosmolar state, hyperosmolar coma and hypoglycemia; 2) endocrine diseases other than T2DM, such as thyroid disease, parathyroid disease, adrenal diseases, pituitary diseases; 3) presence of non-diabetes-related neuropathy such as chronic inflammatory demyelinating polyneuropathy, cervical and lumbar diseases, and severe cerebrovascular disease (ischemic and haemorrhagic stroke); 4) severe respiratory disease, congestive heart failure (New York Heart Association functional class IV), severe renal failure (estimated glomerular filtration rate (eGFR) <30 mL/min/1.73 m2), hematological diseases, thromboembolic disease; 5) autoimmune or viral hepatitis, alcohol-induced or drug-induced liver disease, cholestatic or metabolic/genetic liver disease, liver cirrhosis, and other chronic liver disease, gall bladder and biliary tract diseases; 6) connective tissue, inflammatory and recent active infectious disease, and stress conditions, autoimmune diseases; 7) history of malignancies and mental illness; 8) alcoholism; 9) pregnancy and lactation; 10) use of immunosuppressive agents, antioxidants, anti-inflammatory, antibiotics, analgesics, systemic corticosteroids, multivitamins or vitamin B12 supplements; 11) use of possible or known drugs affecting peripheral nerve function and sympathetic system; 12) missing or incomplete demographic or clinical characteristic data. After applying the exclusion criteria, 1562 participants aged 18-89 years were eligible and finally enrolled in the cross-sectional study.
The study was reviewed and approved by the human research ethics committee of the Affiliated Hospital of Southwest Medical University, and was performed in accordance with the Helsinki Declaration. All patients gave informed consent before participating in this study.
## Data collection and measurements
During face-to-face interviews, trained interviewers administered a detailed standardized questionnaire, which consisted of information on their demographic characteristics (sex, age), lifestyle characteristics (physical activity, smoking and drinking status, etc.), personal medical history (hypertension, coronary heart disease (CHD), DFU, PAD, diabetic retinopathy (DR), DN, NAFLD, and other diseases), disease duration, family history, as detailed elsewhere. Then, all patients with T2DM received anthropometric examination, physical examination, laboratory tests, and evaluation of diabetes-related complications.
Body weight and height were measured by trained interviewers under standardized conditions following a standardized protocol, and body mass index (BMI) was calculated as body weight (kg) divided by the square of the height (m). Systolic blood pressures (SBP) and diastolic blood pressures (DBP) were measured in all subjects on the right arm using a standard mercury sphygmomanometer [27].
Venous blood samples were gathered from each participant in the morning after an overnight fast (at least 8 h) for measurement of fasting blood glucose (FBG), glycated hemoglobin A1C (HbA1c), total cholesterol (TC), triglyceride (TG), high density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), apolipoprotein A (apoA), apolipoprotein B (apoB), AST, ALT, total bilirubin (TBIL), glutamyl transpeptidase (GGT), serum albumin (ALB), creatinine (Cr), uric acid (UA), white blood cell (WBC), neutrophil, and lymphocyte counts, red blood cell distribution width (RDW), and fibrinogen according to relevant protocols and guidelines at the registered central laboratory located at the Affiliated Hospital of Southwestern Medical University, which is accredited in line with the international organization for standardization (ISO) 15189 standard for quality management specific to medical laboratories.
Triglyceride-glucose (TyG) index was calculated using the following equation: ln (fasting TG [mg/dL] × FBG [mg/dL]/2) [28]. The atherogenic index of plasma (AIP) was calculated as ln (TG/HDL-C) and the atherogenic coefficient (AC) was calculated as (TC-HDL-C/HDL-C) [29]. Hepatic steatosis index (HSI) was defined as follows: HSI = 8 × ALT/AST ratio + BMI (+2, if diabetes; +2, if female) [30]. The AAR was calculated as AST/ALT ratio. Neutrophil to lymphocyte ratio (NLR) was calculated by dividing the neutrophil count by lymphocyte count. The eGFR was evaluated with the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equations modified by a Japanese coefficient [31, 32]. Urinary albumin and Cr were measured from three fresh morning spot urine sample on three separate occasions within 6 months. Urinary albumin was measured with immunoturbidimetric tests. Urinary Cr was measured enzymatically. The urinary albumin-to- Cr ratio (ACR; mg/g creatinine) was calculated by dividing urinary albumin by urinary Cr [33, 34]. Patients were then classified as having DN if they had an eGFR < 60 mL/min/1.73m2 and/or an ACR > 30 mg/g in two out of three random voided urine samples (32–34).
## Foot examination and definition of DPN, PAD, and DFU
All patients with T2DM were asked whether they had numbness, pain (prickling or stabbing, shooting, burning or aching pain), and paresthesia (abnormal cold or heat sensation, allodynia and hyperalgesia) in the toes, feet, legs or upper-limb. Then, an experienced physician performed the neurologic examination which included vibration, light touch, and achilles tendon reflexes on both sides in the knee standing position (as being either presence or weakening or loss). Vibration perception threshold (VPT) was assessed at the metatarsophalangeal joint dig I using a neurothesiometer (Bio- Thesiometer; Bio-Medical Instrument Co., Newbury, OH, USA). First, the patients were informed how to know the vibration sensation is felt by gradually turning the amplitude from zero to maximum, then the test began again from zero and they were asked to say the moment that they first felt it. Measurements were made on the planter aspect of the big toe bilaterally, three times consecutively for each big toe. The median of three readings is accepted as the VPT value of that measurement [35]. Sensitivity to touch was also tested using a $\frac{5.07}{10}$-g Semmes-Weinstein monofilament (SWM) at four points on each foot: three on the plantar and one on the dorsal side. The 10-g SWM was placed perpendicular to the skin and pressure was applied until the filament just buckled with a contact time of 2 s. Inability to perceive the sensation at any one site was considered abnormal [36, 37]. DPN was defined as VPT ≥25 V and/or inability to feel the monofilament [35], and then participants were divided into DPN group and no DPN group.
Ankle brachial index (ABI) was measured noninvasively by a continuous-wave Doppler ultrasound probe (Vista AVS, Summit Co., USA) with participants in the supine position after at least 5 min of rest. Leg-specific ABI was calculated by dividing the higher SBP in the posterior tibial or dorsalis pedis by the higher of the right or left brachial SBP [33, 38]. Patients were diagnosed as having PAD if an ABI value <0.9 on either limb [33, 38].
DFU was defined as ulceration of the foot (distally from the ankle and including the ankle) associated with neuropathy and different grades of ischemia and infection [39].
## Other classifications and definitions
A Canon CR-2 Digital Retinal Camera was performed to obtain two-field fundus photography of patient’s eyes (Canon Inc., Kanagawa, Japan). The presence of DR was assessed by high-quality fundus photographs and an ophthalmologist. NAFLD diagnosis was based on the detection of hepatic steatosis by abdominal ultrasound while excluding drugs, viruses, or alcohol as the cause [40]. MetS was defined according to Chinese Diabetes Society (CDS) criteria [41] if they have three or more of the following risk factors [1]: overweight or obese (BMI ≥ 25.0 kg/m2) [2]; hyperglycemia (FBG ≥ 6.1 mmol/l and/or 2-hour postprandial plasma glucose ≥ 7.8 mmol/l, or under treatment for diabetes) [3]; hypertension (SBP ≥ 140 mmHg, DBP ≥ 90 mmHg, or on antihypertensive medication); and [4] dyslipidemia, defined as TG ≥ 1.7 mmol/l and/or HDL-C < 0.9 mmol/l (men) or <1.0 mmol/l (women). CHD was defined as a positive history of myocardial infarction, bypass operation, a diagnostic finding in angiography or positive exercise test [42].
## Statistical analysis
Statistical analyses were conducted using the Statistical Package for Social Sciences (SPSS) (version 20.0; IBM, Chicago, IL). All data were first analyzed for normality of distribution using the Kolmogorov–Smirnov test of normality, and homogeneity of variance using the Levene homogeneity of variance test. Continuous data are presented as mean ± standard deviation (SD), and categorical data are presented as absolute and relative frequencies (n, %).
All patients with T2DM were placed into four groups according to AAR quartiles: quartile (Q) 1 group, 0.32–0.80; Q2 group, 0.81–1.00; Q3 group, 1.01–1.27; and Q4 group, 1.28-5.26. Meanwhile, male and female patients were divided into four quartile groups by AAR level, respectively: Q1 group (male: 0.32-0.75; female: 0.36–0.88), Q2 group (male: 0.76–0.94; female: 0.89–1.08), Q3 group (male: 0.95–1.19; female: 1.09–1.34), and Q4 group (male: 1.20– 3.98; female: 1.35– 5.26). Continuous variables were compared by Student’s t test and one-way analysis of variance (ANOVA), whereas skewed distribution variables were compared by Mann-Whitney U and Kruskal-Wallis tests. Categorical variables were compared across groups using χ2 tests. As AAR was non-normally distributed, Spearman correlation coefficients were performed to assess whether there was an association between AAR and other variables, and the partial correlation coefficient was also used to control for the effects of age, sex, BMI, and diabetic duration. The collinearity diagnostics analysis in linear regression models was also performed to assess whether multiple collinearity exists in these independent variables. The associations of AAR and other variables with the risk of presence of DPN in all T2DM patients were explored by the univariable logistic regression analysis, and then determined using a multivariable logistic regression analysis with those variables achieving P ≤ 0.20 in our univariable analysis entered into this model. Further, binary logistic regression analyses were conducted to investigate the association of AAR quartiles with the risk of presence of DPN in all subjects, male subjects, and female subjects, and odd ratio (OR) and $95\%$ confidence interval (CI) were estimated. Possible dose-response relationships between AAR and DPN were examined by the trend test. Last, the predictive validity of AAR for the presence of DPN was determined using receiver operating characteristic (ROC) curves and area under the curve (AUC) in all subjects, male subjects, and female subjects.
Results were considered to be statistically significant at a P value <0.05.
## Clinical and laboratory characteristics of study participants
The clinical and laboratory characteristics of 1562 patients with T2DM (774 male, $49.55\%$, and 788 female, $50.45\%$) according to AAR quartiles were summarized in Table 1. Overall, mean age was 59.74 years, BMI was 24.19 kg/m2, diabetic duration was 7.55 years, and AAR was 1.10. Patients with higher AAR quartiles tended to be female and relatively older, less user of smoking, and have longer diabetic duration, higher levels of SBP, HDL-C, apoA, AST, AAR, RDW, serum Cr, urinary ACR, VPT, presence of DPN, DN, hypertension, DFU, PAD, and lower BMI, DBP, TG, LDL-C, apoB, TyG, FBG, HbA1c, ALT, TBIL, GGT, serum ALB, lymphocyte counts, eGFR, ABI, HSL, prevalence of dyslipidemia, NAFLD, and MetS compared to those with lower quartiles ($P \leq 0.01$ or $P \leq 0.05$). Supplementary Table 1 reported characteristics of all T2DM patients by DPN. Patients with DPN had significantly older age, longer diabetic duration, higher SBP, FBG, HbA1c, AAR, WBC, neutrophil counts, NLR, fibrinogen, serum Cr, urinary ACR, VPT, prevalence of DN, DR, hypertension, CHD, DFU, PAD, and lower BMI, DBP, TC, TG, apoA, ALT, AST, TBIL, GGT, serum ALB, lymphocyte counts, eGFR, ABI, HSL, and prevalence of NAFLD than those without DPN ($P \leq 0.01$ or $P \leq 0.05$).
**Table 1**
| Variable | Total | Q1 | Q2 | Q3 | Q4 | P |
| --- | --- | --- | --- | --- | --- | --- |
| | (n=1562) | (n=394) | (n=387) | (n=386) | (n=395) | P |
| | | 0.32–0.80 | 0.81–1.00 | 1.01–1.27 | 1.28– 5.26 | value |
| Male (n, %) | 774 (49.55%) | 252 (63.96%) | 195 (50.39%) | 182 (47.15%) | 145 (36.71%) | 0.000 |
| Age (years) | 59.74 ± 11.32 | 54.79 ± 11.32 | 59.33 ± 10.46 | 60.89 ± 10.72 | 63.94 ± 10.80 | 0.000 |
| BMI (kg/m2) | 24.19 ± 3.66 | 24.78 ± 3.66 | 24.69 ± 3.49 | 24.21 ± 3.60 | 23.08 ± 3.64 | 0.000 |
| Diabetic duration (years) | 7.55 ± 6.45 | 5.10 ± 5.09 | 7.76 ± 6.18 | 8.09 ± 6.42 | 9.27 ± 7.22 | 0.000 |
| Smoking (n, %) | 333 (21.32%) | 120 (30.46%) | 72 (18.60%) | 73 (18.91%) | 68 (17.22%) | 0.000 |
| SBP (mmHg) | 132.48 ± 20.82 | 129.16 ± 18.94 | 132.97 ± 20.94 | 133.19 ± 22.03 | 134.60 ± 20.98 | 0.001 |
| DBP (mmHg) | 72.06 ± 12.17 | 74.50 ± 12.31 | 72.81 ± 11.05 | 71.34 ± 12.72 | 69.61 ± 12.05 | 0.000 |
| TC(mmol/L) | 4.86 ± 1.35 | 4.88 ± 1.28 | 4.83 ± 1.27 | 5.00 ± 1.46 | 4.75 ± 1.37 | 0.086 |
| TG (mmol/L) | 2.36 ± 2.60 | 2.67 ± 2.57 | 2.42 ± 2.00 | 2.42 ± 3.59 | 1.92 ± 1.82 | 0.000 |
| HDL-C (mmol/L) | 1.18 ± 0.37 | 1.08 ± 0.30 | 1.14 ± 0.33 | 1.20 ± 0.40 | 1.29 ± 0.40 | 0.000 |
| LDL-C (mmol/L) | 2.78 ± 1.00 | 2.78 ± 0.94 | 2.75 ± 0.96 | 2.91 ± 1.09 | 2.69 ± 1.02 | 0.025 |
| ApoA (g/L) | 1.33 ± 0.30 | 1.28 ± 0.26 | 1.33 ± 0.28 | 1.35 ± 0.32 | 1.36 ± 0.35 | 0.001 |
| ApoB (g/L) | 0.91 ± 0.29 | 0.94 ± 0.26 | 0.90 ± 0.26 | 0.93 ± 0.30 | 0.87 ± 0.33 | 0.000 |
| TyG | 9.48 ± 1.08 | 9.67 ± 1.11 | 9.61 ± 0.92 | 9.46 ± 1.04 | 9.17 ± 1.15 | 0.000 |
| AIP | 0.48 ± 0.02 | 0.67 ± 0.04 | 0.57 ± 0.04 | 0.46 ± 0.05 | 0.20 ± 0.04 | 0.000 |
| AC | 3.52 ± 2.42 | 3.88 ± 2.46 | 3.61 ± 2.26 | 3.66 ± 3.09 | 2.95 ± 1.53 | 0.000 |
| FBG (mmol/L) | 10.87 ± 5.20 | 11.68 ± 5.20 | 11.00 ± 4.88 | 10.52 ± 5.08 | 10.28 ± 5.50 | 0.000 |
| HbA1c (%) | 9.52 ± 2.50 | 9.96 ± 2.47 | 9.55 ± 2.30 | 9.38 ± 2.50 | 9.20 ± 2.68 | 0.000 |
| ALT (U/L) | 23.15 ± 17.53 | 38.09 ± 22.46 | 23.11 ± 10.58 | 17.92 ± 12.20 | 13.40 ± 10.66 | 0.000 |
| AST (U/L) | 21.86 ± 15.81 | 24.20 ± 12.88 | 20.98 ± 9.63 | 20.15 ± 13.48 | 22.08 ± 23.39 | 0.000 |
| AAR | 1.10 ± 0.46 | 0.66 ± 0.10 | 0.91 ± 0.06 | 1.13 ± 0.08 | 1.68 ± 0.50 | 0.000 |
| TBIL (μmol/L) | 12.27 ± 5.62 | 13.11 ± 6.52 | 12.61 ± 5.37 | 12.10 ± 5.39 | 11.24 ± 4.94 | 0.000 |
| GGT (U/L) | 43.74 ± 2.53 | 55.57 ± 7.57 | 41.92 ± 2.66 | 40.50 ± 4.84 | 36.86 ± 3.70 | 0.048 |
| Serum ALB (g/L) | 40.97 ± 4.91 | 41.79 ± 4.48 | 41.76 ± 4.42 | 41.08 ± 4.77 | 39.32 ± 5.48 | 0.000 |
| WBC (*109/L) | 6.81 ± 2.41 | 7.01 ± 2.40 | 6.64 ± 1.85 | 6.64 ± 2.12 | 6.94 ± 3.07 | 0.119 |
| Neutrophil (*109/L) | 4.60 ± 2.27 | 4.73 ± 2.22 | 4.40 ± 1.66 | 4.43 ± 1.92 | 4.82 ± 3.02 | 0.264 |
| Lymphocyte (*109/L) | 1.65 ± 0.62 | 1.73 ± 0.64 | 1.66 ± 0.61 | 1.63 ± 0.59 | 1.57 ± 0.64 | 0.006 |
| NLR | 3.36 ± 0.08 | 3.22 ± 0.14 | 3.11 ± 0.11 | 3.16 ± 0.12 | 3.94 ± 0.22 | 0.192 |
| RDW (%) | 13.17 ± 1.28 | 13.17 ± 1.37 | 13.13 ± 1.21 | 13.01 ± 1.19 | 13.35 ± 1.32 | 0.013 |
| Fibrinogen(g/L) | 3.68 ± 1.35 | 3.49 ± 1.25 | 3.62 ± 1.34 | 3.71 ± 1.29 | 3.86 ± 1.47 | 0.060 |
| Serum UA (μmol/L) | 316.52 ± 108.40 | 321.33 ± 113.45 | 313.86 ± 95.81 | 316.39 ± 108.38 | 314.45 ± 115.00 | 0.908 |
| Serum Cr (μmol/L) | 73.74 ± 47.41 | 69.31 ± 40.69 | 71.05 ± 39.74 | 74.87 ± 49.83 | 79.69 ± 56.72 | 0.008 |
| eGFR(mL/min/1.73 m2) | 91.91 ± 26.12 | 100.40 ± 24.34 | 93.29 ± 25.12 | 89.90 ± 25.32 | 84.07 ± 27.00 | 0.000 |
| Urinary ACR (mg/g) | 238.78 ± 22.07 | 140.82 ± 35.57 | 177.12 ± 33.57 | 209.69 ± 32.92 | 427.31 ± 64.97 | 0.000 |
| ABI | 1.02 ± 0.15 | 1.04 ± 0.12 | 1.03 ± 0.15 | 1.01 ± 0.16 | 1.00 ± 0.18 | 0.003 |
| VPT (V) | 16.40 ± 9.90 | 14.65 ± 9.47 | 15.42 ± 8.96 | 16.49 ± 9.17 | 19.03 ± 11.28 | 0.000 |
| HSL | 33.46 ± 5.45 | 37.92 ± 4.65 | 34.45 ± 3.82 | 32.23 ± 4.25 | 29.14 ± 4.82 | 0.000 |
| Dyslipidemia (n, %) | 844 (54.03%) | 245 (62.18%) | 219 (56.59%) | 211 (54.66%) | 169 (42.78%) | 0.000 |
| NAFLD (n, %) | 713 (45.65%) | 229 (58.12%) | 204 (52.71%) | 159 (41.19%) | 121 (30.63%) | 0.000 |
| MetS (n, %) | 740 (47.38%) | 191 (48.48%) | 193 (49.87%) | 196 (50.78%) | 160 (40.51%) | 0.015 |
| Microvascular complications | Microvascular complications | Microvascular complications | Microvascular complications | Microvascular complications | Microvascular complications | Microvascular complications |
| DN (n, %) | 657 (42.06%) | 121 (30.71%) | 150 (38.76%) | 169 (43.78%) | 217 (54.94%) | 0.000 |
| DR (n, %) | 201 (12.87%) | 46 (11.68%) | 43 (11.11%) | 58 (15.03%) | 54 (13.67%) | 0.335 |
| DPN (n, %) | 236 (15.11%) | 43 (10.91%) | 51 (13.18%) | 55 (14.25%) | 87 (22.03%) | 0.000 |
| Macrovascular complications | Macrovascular complications | Macrovascular complications | Macrovascular complications | Macrovascular complications | Macrovascular complications | Macrovascular complications |
| Hypertension (n, %) | 828 (53.01%) | 180 (45.69%) | 206 (53.23%) | 213 (55.18%) | 229 (57.97%) | 0.004 |
| CHD (n, %) | 140 (8.96%) | 24 (6.09%) | 39 (10.08%) | 40 (10.36%) | 37 (9.37%) | 0.134 |
| DFU (n, %) | 115 (7.36%) | 19 (4.82%) | 24 (6.20%) | 31 (8.03%) | 41 (10.38%) | 0.018 |
| PAD (n, %) | 154 (9.86%) | 26 (6.60%) | 28 (7.24%) | 40 (10.36%) | 60 (15.19%) | 0.000 |
## Association of AAR with clinical and laboratory characteristics in study subjects
Table 2 showed the association of AAR with clinical and laboratory characteristics in all patients with T2DM performed by Spearman and partial correlation coefficient. The results revealed that AAR was positively associated with age, sex distribution, diabetic duration, SBP, HDL-C, apoA, fibrinogen, serum Cr, urinary ACR, VPT and prevalence of DPN, DN, hypertension, DFU, PAD, and negatively with BMI, smoking, DBP, TG, apoB, TyG, FBG, HbA1c, ALT, AST, TBIL, GGT, serum ALB, WBC, lymphocyte counts, eGFR, ABI, HSL, and prevalence of dyslipidemia, NAFLD and MetS ($P \leq 0.01$ or $P \leq 0.05$). After adjustments for sex, age, BMI, and diabetic duration, the associations among HbA1c, ALT, serum ALB, eGFR, urinary ACR, VPT, HSL, presence of dyslipidemia, NAFLD, DPN, DN, PAD and AAR were attenuated but remained statistically significant ($P \leq 0.01$ or $P \leq 0.05$).
**Table 2**
| Unnamed: 0 | r | P-value | Adjusted r | Adjusted P-value |
| --- | --- | --- | --- | --- |
| Age | 0.305 | 0.01 | – | – |
| Sex (female vs male) | 0.197 | 0.0 | – | – |
| BMI | -0.185 | 0.0 | – | – |
| Diabetic duration | 0.226 | 0.0 | – | – |
| Smoking | -0.105 | 0.0 | 0.027 | 0.295 |
| SBP | 0.099 | 0.0 | 0.029 | 0.496 |
| DBP | -0.166 | 0.0 | -0.054 | 0.214 |
| TC | -0.034 | 0.177 | -0.017 | 0.690 |
| TG | -0.17 | 0.0 | -0.034 | 0.425 |
| HDL-C | 0.207 | 0.0 | 0.060 | 0.162 |
| LDL-C | -0.029 | 0.249 | -0.004 | 0.929 |
| ApoA | 0.091 | 0.0 | -0.008 | 0.846 |
| ApoB | -0.1 | 0.0 | 0.041 | 0.344 |
| TyG | -0.201 | 0.0 | -0.084 | 0.053 |
| AIP | -0.208 | 0.0 | -0.051 | 0.235 |
| AC | -0.199 | 0.0 | -0.032 | 0.461 |
| FBG | -0.154 | 0.0 | -0.060 | 0.161 |
| HbA1c | -0.145 | 0.0 | -0.086 | 0.047 |
| ALT | -0.703 | 0.0 | -0.380 | 0.000 |
| AST | -0.138 | 0.0 | 0.038 | 0.384 |
| TBIL | -0.131 | 0.0 | -0.060 | 0.161 |
| GGT | -0.324 | 0.0 | 0.053 | 0.220 |
| Serum ALB | -0.181 | 0.0 | -0.130 | 0.003 |
| WBC | -0.052 | 0.039 | 0.035 | 0.423 |
| Neutrophil | -0.037 | 0.148 | 0.044 | 0.303 |
| Lymphocyte | -0.085 | 0.001 | -0.066 | 0.128 |
| NLR | 0.044 | 0.086 | 0.082 | 0.057 |
| RDW | 0.047 | 0.064 | 0.081 | 0.060 |
| Fibrinogen | 0.086 | 0.017 | 0.055 | 0.205 |
| Serum UA | -0.02 | 0.44 | 0.039 | 0.371 |
| Serum Cr | 0.087 | 0.001 | 0.068 | 0.114 |
| eGFR | -0.258 | 0.0 | -0.112 | 0.010 |
| Urinary ACR | 0.192 | 0.0 | 0.105 | 0.014 |
| ABI | -0.098 | 0.0 | -0.050 | 0.251 |
| VPT | 0.209 | 0.0 | 0.110 | 0.011 |
| HSL | -0.665 | 0.0 | -0.807 | 0.000 |
| Dyslipidemia | -0.141 | 0.0 | -0.068 | 0.009 |
| NAFLD | -0.215 | 0.0 | -0.131 | 0.000 |
| MetS | -0.06 | 0.017 | 0.016 | 0.530 |
| DPN | 0.124 | 0.0 | 0.068 | 0.008 |
| DN | 0.186 | 0.0 | 0.118 | 0.000 |
| DR | 0.03 | 0.238 | 0.006 | 0.817 |
| Hypertension | 0.087 | 0.001 | 0.007 | 0.785 |
| CHD | 0.046 | 0.072 | -0.002 | 0.926 |
| DFU | 0.084 | 0.001 | 0.043 | 0.104 |
| PAD | 0.129 | 0.0 | 0.052 | 0.048 |
## Univariate and multivariate analysis of determinants of DPN in study subjects
Table 3 displayed the associations of AAR and other variables with the risk of presence of DPN. The univariate logistic regression analysis revealed that age, BMI, diabetic duration, SBP, DBP, TC, TG, apoA, FBG, HbA1c, ALT, AST, AAR, TBIL, serum ALB, WBC, neutrophil and lymphocyte counts, NLR, fibrinogen, serum Cr, eGFR, urinary ACR, ABI, HSL, and prevalence of NAFLD, DN, DR, hypertension, CHD, DFU, PAD were significantly associated with the presence of DPN ($P \leq 0.01$ or $P \leq 0.05$).Multivariable logistic regression analysis showed that age, TyG, AAR, serum ALB, and DFU were significantly and independently associated with the presence of DPN ($P \leq 0.01$ or $P \leq 0.05$). Notably, each SD increase in AAR was associated with a significant 2.413-fold increased odds of DPN ($95\%$ CI, 1.081-5.386, $P \leq 0.05$).
**Table 3**
| Variables | Univariate analysis | Univariate analysis.1 | Univariate analysis.2 | Multivariate analysis | Multivariate analysis.1 | Multivariate analysis.2 |
| --- | --- | --- | --- | --- | --- | --- |
| | B | OR (95%CI) | P-value | B | OR (95%CI) | P-value |
| Sex (female vs male) | -0.221 | 0.802 (0.607-1.058) | 0.119 | | | |
| Age | 0.072 | 1.075 (1.059-1.090) | 0.000 | 0.040 | 1.041 (1.004-1.080) | 0.029 |
| BMI | -0.059 | 0.943 (0.905-0.982) | 0.005 | | | |
| Diabetic duration | 0.068 | 1.071 (1.050-1.092) | 0.000 | | | |
| Smoking | -0.101 | 0.904 (0.640-1.277) | 0.568 | | | |
| SBP | 0.007 | 1.007 (1.000-1.014) | 0.038 | | | |
| DBP | -0.017 | 0.983 (0.971-0.994) | 0.004 | | | |
| TC | -0.149 | 0.862 (0.769-0.966) | 0.011 | | | |
| TG | -0.175 | 0.839 (0.758-0.930) | 0.001 | | | |
| HDL-C | 0.040 | 1.041 (0.710-1.525) | 0.838 | | | |
| LDL-C | -0.040 | 0.961 (0.834-1.107) | 0.579 | | | |
| ApoA | -0.822 | 0.440 (0.269-0.718) | 0.001 | | | |
| ApoB | -0.030 | 0.970 (0.596-1.579) | 0.903 | | | |
| TyG | -0.119 | 0.888 (0.786-1.002) | 0.054 | 2.156 | 8.639 (1.036-72.058) | 0.046 |
| AC | -0.129 | 0.879 (0.803-0.962) | 0.005 | | | |
| FBG | 0.027 | 1.028 (1.002-1.053) | 0.032 | | | |
| HbA1c | 0.103 | 1.109 (1.051-1.169) | 0.000 | | | |
| ALT | -0.031 | 0.970 (0.957-0.982) | 0.000 | | | |
| AST | -0.030 | 0.971 (0.954-0.987) | 0.001 | | | |
| AAR | 0.667 | 1.949 (1.497-2.537) | 0.000 | 0.881 | 2.413 (1.081-5.386) | 0.032 |
| TBIL | -0.058 | 0.944 (0.916-0.973) | 0.000 | | | |
| GGT | -0.001 | 0.999 (0.997-1.001) | 0.381 | | | |
| Serum ALB | -0.151 | 0.860 (0.835-0.886) | 0.000 | -0.102 | 0.903 (0.843-0.967) | 0.004 |
| WBC | 0.101 | 1.106 (1.052-1.162) | 0.000 | | | |
| Lymphocyte | -0.767 | 0.464 (0.355-0.608) | 0.000 | | | |
| NLR | 0.111 | 1.117 (1.074-1.163) | 0.000 | | | |
| RDW | 0.060 | 1.062 (0.958-1.177) | 0.256 | | | |
| Fibrinogen | 0.409 | 1.505 (1.328-1.704) | 0.000 | | | |
| Serum UA | 0.001 | 1.001 (1.000-1.002) | 0.127 | | | |
| Serum Cr | 0.006 | 1.006 (1.004-1.008) | 0.000 | | | |
| eGFR | -0.023 | 0.978 (0.973-0.983) | 0.000 | | | |
| Urinary ACR | 0.678 | 1.969 (1.613-2.404) | 0.000 | | | |
| ABI | -3.087 | 0.046 (0.021-0.098) | 0.000 | | | |
| HSL | -0.060 | 0.941 (0.917-0.966) | 0.000 | | | |
| Dyslipidemia | -0.191 | 0.826 (0.626-1.090) | 0.178 | | | |
| NAFLD | -0.426 | 0.653 (0.491-0.869) | 0.003 | | | |
| MetS | 0.144 | 1.155 (0.875-1.523) | 0.309 | | | |
| DN | 1.035 | 2.814 (2.111-3.751) | 0.000 | | | |
| DR | 0.738 | 2.092 (1.466-2.984) | 0.000 | | | |
| Hypertension | 0.490 | 1.632 (1.227-2.171) | 0.001 | | | |
| CHD | 0.958 | 2.606 (1.757-3.865) | 0.000 | | | |
| DFU | 1.817 | 6.153 (4.133-9.159) | 0.000 | 1.337 | 3.807 (1.753-8.267) | 0.001 |
| PAD | 1.632 | 5.117 (3.575-7.324) | 0.000 | | | |
## Association of AAR quartiles with the risk of presence of DPN in study subjects
As shown in Table 4, the risk of presence of DPN also increased progressively as AAR quartiles increased in all subjects, male subjects, and female subjects, respectively (all P for trend <0.01). When compared to the lower quartiles (Q1, Q2, and Q3), the highest quartile of AAR (Q4) of all subjects, male subjects, and female subjects were significantly associated with increased odds for DPN (OR = 1.930, 2.073, and 2.368, respectively). Even per SD increase in AAR of all subjects, male subjects, and female subjects were respectively associated with were more likely to have DPN (OR = 1.358, 1.416, and 1.348, respectively).
**Table 4**
| AAR | DPN | DPN.1 |
| --- | --- | --- |
| | Odds ratio (95% CI) | P |
| All subjects | All subjects | All subjects |
| Per SD increase | 1.358 (1.204-1.533) | 0.000 |
| Quartiles of AAR | Quartiles of AAR | Quartiles of AAR |
| Q1 (0.32–0.80) | 1 (reference) | |
| Q2 (0.81–1.00) | 1.239 (0.804-1.909) | 0.331 |
| Q3 (1.01–1.27) | 1.356 (0.886-2.077) | 0.161 |
| Q4 (1.28– 5.26) | 2.306 (1.552-3.426) | 0.000 |
| P for trend | 0.000 | |
| Q4 versus. Q1, Q2, Q3 | 1.930 (1.439-2.588) | 0.000 |
| | Male subjects | |
| Per SD increase | 1.416 (1.198-1.674) | 0.000 |
| Quartiles of AAR | Quartiles of AAR | Quartiles of AAR |
| Q1 (0.32–0.75) | 1 (reference) | |
| Q2 (0.76–0.94) | 1.450 (0.774-2.717) | 0.246 |
| Q3 (0.95–1.19) | 2.115 (1.163-3.849) | 0.014 |
| Q4 (1.20– 3.98) | 3.101 (1.746-5.510) | 0.000 |
| P for trend | 0.000 | |
| Q4 versus. Q1, Q2, Q3 | 2.073 (1.386-3.101) | 0.000 |
| | Female subjects | |
| Per SD increase | 1.348 (1.133-1.603) | 0.001 |
| Quartiles of AAR | Quartiles of AAR | Quartiles of AAR |
| Q1 (0.36–0.88) | 1 (reference) | |
| Q2 (0.89–1.08) | 1.071 (0.572-2.004) | 0.831 |
| Q3 (1.09–1.34) | 0.945 (0.498-1.793) | 0.864 |
| Q4 (1.35– 5.26) | 2.379 (1.366-4.145) | 0.002 |
| P for trend | 0.002 | |
| Q4 versus. Q1, Q2, Q3 | 2.368 (1.550-3.618) | 0.000 |
## Predictive value of AAR in screening for the presence of DPN in T2DM patients
To explore the predictive value of AAR for DPN, we analyzed the ROC curves of AAR. The results revealed that the best cutoff value for AAR to predict the presence of DPN was 1.40 (sensitivity: $30.90\%$; specificity: $85.50\%$; and AUC: 0.600; Figure 1A) in all subjects, and the best cutoff values for AAR to predict the presence of DPN were 0.906 (sensitivity: $70.3\%$; specificity: $49.2\%$; and AUC: 0.618; Figure 1B) and 1.402 (sensitivity: $38\%$; specificity: $81.9\%$; and AUC: 0.600; Figure 1C) in male and female subjects, respectively.
**Figure 1:** *Receiver operating characteristics (ROC) curve analysis of aminotransferase to alanine aminotransferase ratio (AAR) to inidicate DPN. (A). all subjects; (B). male subjects; (C). female subjects.*
## Discussion
To our knowledge, this was the first study to investigate the relationship between AAR and risk of presence of DPN. We found that patients with higher AAR quartiles had higher presence of DPN, and AAR was an independent determinant of presence of DPN after multivariate adjustment. Additionally, the risk of presence of DPN increased progressively as AAR quartiles increased in both sexes. Last, the analysis of ROC curves revealed that AAR could predict the presence of DPN in both sexes. These findings suggest that high AAR may be associated with the presence of DPN in hospitalized Chinese T2DM patients, and may be used as an additional indicator of risk of DPN.
As mentioned earlier, AAR, an emerging indicator of liver function, has been reported to effectively predict the severity of liver fibrosis in patients with various liver disease including NAFLD [10, 11]. There is now growing evidence that NAFLD is more common and often advanced in patients with T2DM, easily progressing to nonalcoholic steatohepatitis and advanced liver fibrosis, than in the general population (6, 43–46). Considering a certain intrinsic correlation among AAR, NAFLD and liver fibrosis, and diabetic vascular complication (6, 10, 11, 15–19, 43–46), it is plausible that AAR may be associated with the presence of DPN, and high AAR may be an early signal for being at risk for DPN. In the present study, we found that patients with higher AAR quartiles tended to have higher VPT, a widely recommended indicator of the presence and severity of confirmed clinical neuropathy [47], and similarly, patients with DPN had significantly higher AAR than those without. Moreover, AAR was positively associated with VPT and presence of DPN. Altogether, these data preliminarily argue that there was a potential relationship between AAR and the presence and severity of DPN. Besides, AAR was significantly and independently associated with the presence of DPN after multivariate adjustment. Additionally, the risk of presence of DPN increased progressively as AAR quartiles increased in both sexes. More importantly, AAR could predict the presence of DPN in both sexes. These data were broadly similar to the findings of previous studies showing that noninvasive biomarkers of liver fibrosis, such as NAFLD fibrosis score and fibrosis-4 score were independently associated with DPN [6, 8, 48, 49], further suggesting that higher AAR, another novel liver fibrosis marker, could be linked to an increased risk of the presence and severity of DPN, and AAR may be a novel and reliable marker for identifying subjects at high risk for DPN in patients with T2DM, however, the underlying mechanisms potentially responsible for the association remain unclear.
Growing evidence suggests that NAFLD is closely associated with the presence of DPN (50–52), while IR has been suggested to play a central role in the development and progression of NAFLD [53]. In the present study, we found that patients with DPN had significantly lower HSL, which is a accurate proxy of NAFLD that can assess liver steatosis in predominantly Asian populations [30], and prevalence of NAFLD than those without DPN. Moreover, the logistic regression analysis revealed that HSL, TyG, a biochemical marker of IR [28], and prevalence of NAFLD were significantly associated with the presence of DPN. Our findings are largely in line with results from prior studies [50, 51, 54, 55]. Yan et al. reported that patients with NAFLD diagnosed earlier than T2DM had a lower prevalence of DPN compared with those with T2DM diagnosed earlier than NAFLD or those with T2DM only [51]. Another cross-sectional study demonstrated that the prevalence of NAFLD in Chinese T2DM patients with DPN was significantly lower than those without DPN, and NAFLD was negatively correlated with the prevalence of DPN [50]. Recently, Zhao and colleagues revealed that a higher level of AUC of C-peptide was inversely associated with prevalence of diabetic neuropathy, and positively associated with homeostasis model assessment of IR index and NAFLD in 885 patients with T2DM [54]. Similar results were also obtained by Guo et al. in T2DM patients [55]. Together, these lines of evidence, combined with our results, suggest that NAFLD and its key component IR may protect against the development and progression of DPN in T2DM patients. Moreover, we demonstrated that patients with higher AAR quartiles tended to have longer diabetic duration and lower TyG, HSL, and prevalence of NAFLD compared to those with lower quartiles. Additionally, the Spearman correlation analysis revealed that AAR was negatively associated with HSL, TyG, and prevalence of NAFLD. Qiao et al. found that C-peptide and insulin levels progressively decreased (inadequate insulin secretion) and IR was relatively low because of weakened or even deterioration of pancreatic islet β cell function induced by long-term hyperglycemia along with increased diabetic duration, leading to increased prevalence of DPN [56]. Combined, these data suggest that there might be a negative correlation between AAR and IR and NAFLD, and higher AAR might contribute to the development of DPN through a complex mechanism associated with IR and NAFLD; however, the mechanism of action needs to be further investigated.
Numerous studies have demonstrated that low-grade inflammation and oxidative stress are also contributing factors in the development and progression of DPN [7, 21, 49]. Serum ALB is the most abundant circulating protein in blood synthesized and secreted from liver cells. It has been reported that serum ALB is the major source of extracellular reduced sulfhydryl groups, which act as potent scavengers of reactive oxygen and nitrogen species, thus constituting the dominant antioxidant in the circulatory system [57, 58]. In addition, some substances such as nitric oxide and bilirubin are carried by serum albumin and provide additional protection against oxidative stress [57]. Also, serum ALB can bind various inflammatory mediators and inhibit the secretion of pro-inflammatory cytokines, thus involving in regulating the inflammatory immune response and endothelial stabilization [58, 59]. It has been suggested that elderly people are susceptible to oxidative stress due to a decline in the inefficiency of their endogenous antioxidant systems [60], and oxidative stress and inflammatory mediators increase with aging [61]. In the present study, we found that patients with DPN had significantly older age and lower serum ALB than those without DPN. The logistic regression analysis revealed that age and serum ALB were significantly and independently associated with the presence of DPN after multivariate adjustment. Our findings were in agreement with previous studies (62–65) showing that serum ALB has neuroprotective effects via its antioxidant/anti-inflammatory activity, and its lower level was related to abnormal peripheral nerve function and a significantly increased risk of DPN, and older age is a risk factor for DPN, providing further evidence that inflammation and oxidative stress induced by lower serum ALB and older age may be closely associated with the presence of DPN. Moreover, patients with higher AAR quartiles tended to be relatively older and had significantly lower serum ALB compared to those with lower quartiles. Additionally, AAR was positively associated with age, and negatively with serum ALB. Our findings were in agreement with most previous studies (15, 66–69). Liu et al. reported that Chinese hypertensive patients with higher AAR had significantly lower levels of serum ALB and other endogenous antioxidant substances compared with those with lower AAR [15]. Evidence from an animal study has also suggested that mice with elevated AAR had a reduced ability to carry oxygen, which was accompanied by significantly elevated levels of markers of oxidative stress [66]. Several studies also have announced that hepatic steatosis assessed by AAR is associated with increased production of interleukin-6 and other pro-inflammatory cytokines by hepatozytes and nonperynchymal cells (67–69). Together, these lines of evidence, combined with our results, suggest that higher AAR may be closely associated with increased inflammation and oxidative stress, and inflammation and oxidative stress induced by lower serum ALB and older age might at least partially mediate the potential relationship between AAR and DPN, but larger, well-characterized, prospective research is still needed to validate these findings.
Experimental and epidemiological studies have shown that atherosclerotic vascular disease plays a critical role in the development and progression of DPN [23, 64]. In the present study, we found that patients with DPN had significantly higher prevalence of DFU and PAD, two major diabetic macrovascular complications associated with atherosclerosis, than those without DPN. Moreover, the logistic regression analysis revealed that the prevalence of PAD was significantly positively associated with the presence of DPN, while DFU was an independent risk factor for DPN. Our findings further provided evidence that atherosclerotic vascular disease, especially DFU, and DPN are closely interconnected, and nerve ischemia associated with vascular dysfunction may contribute to nerve damage, eventually leading to the development of DPN. Moreover, patients with higher AAR quartiles tended to have higher prevalence of DFU and PAD and lower ABI, and AAR was significantly positively associated with prevalence of PAD and DFU, which was in general agreement with two previous studies [15, 70]. A cross-sectional study that included 10900 Chinese adults with hypertension discovered that a high AAR was independently and positively associated with associated with PAD risk [15]. Similarly, another cross-sectional study conducted by Rief and his colleagues reported that an elevation in AAR is significantly associated with an increased risk of occurrence of critical limb ischemia, independently of well-established risk factors, in patients with peripheral arterial occlusive disease [70]. Together, these results indicate that high AAR might be linked to PAD and critical limb ischemia, an important risk factor for DFU, and vascular damage, especially DFU, might be associated with the presence of DPN. It is well-known that AST is abundantly present in many different types of tissue in addition to the liver, such as skeletal, cardiac, smooth muscle, kidney, and brain, whereas ALT is low in cells other than hepatocytes [71]. Thus, an increased vulnerability of the liver and several other tissue associated with AST distribution to ischemia due to vascular damage caused by DFU would lead to an higher AAR in T2DM patients with DPN [6, 71, 72]. However, the exact mechanism responsible for the relationship between AAR and DPN is still obscure and required further investigation.
Some potential limitations of our study should be noted. First, the causality of the relationship between AAR and DPN could not be established because this design of the present study is cross-sectional. Second, individuals with T2DM are at high risk for both microvascular complications and macrovascular complications, and thus may usually needs to take multiple medications at the same time, of which might affect liver transaminase due to potential drug-drug interactions. However, detailed medication history, such as statins, for these subjects was unavailable. Third, the present study population consisted of T2DM inpatients of Chinese Han ancestry, who generally had more serious illness than diabetic outpatients, and thus, our findings cannot to be extrapolated to diabetic outpatients and other types of diabetes with different ethnic back grounds. Finally, it has been reported that a sedentary lifestyle and unhealthy dietary habits are associated with elevated liver enzyme levels [73], however, insufficient data were available for the information about their diet and lifestyle, which might have influenced the results. Despite these limitations, this study has several strengths such as a relatively large sample size, use of a standardized method at a single center, and thorough adjustment for possible confounding variables.
In conclusion, the present study demonstrated that AAR was significantly increased in T2DM patients with DPN, and was independently associated with increased risk of presence of DPN in Chinese patients with T2DM, thereby suggesting that AAR may serve as an useful and reliable biomarker of DPN, and highlighting that it is crucial to pay more attention to T2DM patients with high AAR to further prevent and reduce the development of DPN and related unfavorable health outcomes.
## Data availability statement
The original contributions presented in the study are included in the article/ Supplementary Material. Further inquiries can be directed to the corresponding author.
## Ethics statement
The studies involving human participants were reviewed and approved by the human research ethics committee of the Affiliated Hospital of Southwest Medical University. The patients/participants provided their written informed consent to participate in this study. Written informed consent was obtained from the individual(s) for the publication of any potentially identifiable images or data included in this article.
## Author contributions
All the authors contributed significantly to the manuscript. PY conducted the population study, analysed and interpreted the data, and drafted the manuscript. QW significantly revised the draft, interpreted the data, and involved in data analyses. YW, XD, XW, and QT conducted the study, collected the information and participated in data interpretation. XC, YX, JZ, and YM involved in the sample test, data management and draft revision. QW is the PI of project, who designed the study and critically revised the manuscript. All authors read and approved the final manuscript. All authors contributed to the article and approved the submitted version.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fendo.2023.1064125/full#supplementary-material
## References
1. Tesfaye S, Sloan G. **Diabetic polyneuropathy - advances in diagnosis and intervention strategies**. *Eur Endocrinol* (2020) **16** 15-20. DOI: 10.17925/EE.2020.16.1.15
2. Dalla Paola L, Faglia E. **Treatment of diabetic foot ulcer: An overview strategies for clinical approach**. *Curr Diabetes Rev* (2006) **2**. DOI: 10.2174/1573399810602040431
3. Argoff CE, Cole BE, Fishbain DA, Irving GA. **Diabetic peripheral neuropathic pain: Clinical and quality-of-life issues**. *Mayo Clin Proc* (2006) **81** S3-11. DOI: 10.1016/s0025-6196(11)61474-2
4. Tuong TTK, Tran DK, Phu PQT, Hong TND, Dinh TC, Chu DT. **Non-alcoholic fatty liver disease in patients with type 2 diabetes: Evaluation of hepatic fibrosis and steatosis using fibroscan**. *Diagnostics (Basel)* (2020) **10**. DOI: 10.3390/diagnostics10030159
5. Chen K, Sng WK, Quah JH, Liu J, Chong BY, Lee HK. **Clinical spectrum of non-alcoholic fatty liver disease in patients with diabetes mellitus**. *PloS One* (2020) **15**. DOI: 10.1371/journal.pone.0236977
6. Kim K, Oh TJ, Cho HC, Lee YK, Ahn CH, Koo BK. **Liver fibrosis indices are related to diabetic peripheral neuropathy in individuals with type 2 diabetes**. *Sci Rep* (2021) **11** 24372. DOI: 10.1038/s41598-021-03870-z
7. Huang J, Li R, Liu N, Yi N, Zheng H, Zhang Q. **Liver fibrosis is independently associated with diabetic peripheral neuropathy in type 2 diabetes mellitus**. *J Diabetes Investig* (2021) **12**. DOI: 10.1111/jdi.13562
8. Williams KH, Burns K, Constantino M, Shackel NA, Prakoso E, Wong J. **An association of large-fibre peripheral nerve dysfunction with non-invasive measures of liver fibrosis secondary to non-alcoholic fatty liver disease in diabetes**. *J Diabetes Complications* (2015) **29**. DOI: 10.1016/j.jdiacomp.2015.06.015
9. De Ritis F, Coltorti M, Giusti G. **An enzymic test for the diagnosis of viral hepatitis; the transaminase serum activities**. *Clin Chim Acta* (1957) **2**. DOI: 10.1016/0009-8981(57)90027-x
10. Åberg F, Danford CJ, Thiele M, Talbäck M, Rasmussen DN, Jiang ZG. **A dynamic aspartate-to-Alanine aminotransferase ratio provides valid predictions of incident severe liver disease**. *Hepatol Commun* (2021) **5**. DOI: 10.1002/hep4.1700
11. Long MT, Pedley A, Massaro JM, Hoffmann U, Fox CS. **The association between non-invasive hepatic fibrosis markers and cardiometabolic risk factors in the framingham heart study**. *PloS One* (2016) **11**. DOI: 10.1371/journal
12. Xu J, Shi X, Pan Y. **The association of aspartate Aminotransferase/Alanine aminotransferase ratio with diabetic nephropathy in patients with type 2 diabetes**. *Diabetes Metab Syndr Obes* (2021) **14**. DOI: 10.2147/DMSO.S330741
13. Simental-Mendía LE, Rodríguez-Morán M, Gómez-Díaz R, Wacher NH, Rodríguez-Hernández H, Guerrero-Romero F. **Insulin resistance is associated with elevated transaminases and low aspartate aminotransferase/alanine aminotransferase ratio in young adults with normal weight**. *Eur J Gastroenterol Hepatol* (2017) **29**. DOI: 10.1097/MEG.0000000000000811
14. Hanley AJ, Williams K, Festa A, Wagenknecht LE, D'Agostino RB, Haffner SM. **Liver markers and development of the metabolic syndrome: The insulin resistance atherosclerosis study**. *Diabetes* (2005) **54**. DOI: 10.2337/diabetes.54.11.3140
15. Liu H, Zha X, Ding C, Hu L, Li M, Yu Y. **AST/ALT ratio and peripheral artery disease in a Chinese hypertensive population: A cross-sectional study**. *Angiology* (2021) **72**. DOI: 10.1177/00033197211004410
16. Mori T, Yoshioka K, Tanno Y. **Non-alcoholic fatty liver disease frequency and associated factors at admission of acute stroke**. *Hepatol Int* (2022) **16**. DOI: 10.1007/s12072-021-10253-z
17. Alexander KS, Zakai NA, Lidofsky SD, Callas PW, Judd SE, Tracy RP. **Non- alcoholic fatty liver disease, liver biomarkers and stroke risk: The reasons for geographic and racial differences in stroke cohort**. *PloS One* (2018) **13**. DOI: 10.1371/journal.pone.0194153
18. Weng SF, Kai J, Guha IN, Qureshi N. **The value of aspartate aminotransferase and alanine aminotransferase in cardiovascular disease risk assessment**. *Open Heart* (2015) **2**. DOI: 10.1136/openhrt-2015-000272
19. Liu Y, Zhao P, Cheng M, Yu L, Cheng Z, Fan L. **AST to ALT ratio and arterial stiffness in non-fatty liver Japanese population: A secondary analysis based on a cross-sectional study**. *Lipids Health Dis* (2018) **17** 275. DOI: 10.1186/s12944-018-0920-4
20. Selvarajah D, Kar D, Khunti K, Davies MJ, Scott AR, Walker J. **Diabetic peripheral neuropathy: Advances in diagnosis and strategies for screening and early intervention**. *Lancet Diabetes Endocrinol* (2019) **7**. DOI: 10.1016/S2213-8587(19)30081-6
21. Su JB, Zhao LH, Zhang XL, Cai HL, Huang HY, Xu F. **HbA1c variability and diabetic peripheral neuropathy in type 2 diabetic patients**. *Cardiovasc Diabetol* (2018) **17** 47. DOI: 10.1186/s12933-018-0693-0
22. Hägg S, Thorn LM, Putaala J, Liebkind R, Harjutsalo V, Forsblom CM. **Incidence of stroke according to presence of diabetic nephropathy and severe diabetic retinopathy in patients with type 1 diabetes**. *Diabetes Care* (2013) **36**. DOI: 10.2337/dc13-0669
23. Tesfaye S, Chaturvedi N, Eaton SE, Ward JD, Manes C, Ionescu-Tirgoviste C. **Vascular risk factors and diabetic neuropathy**. *N Engl J Med* (2005) **352**. DOI: 10.1056/NEJMoa032782
24. Lombardi R, Airaghi L, Targher G, Serviddio G, Maffi G, Mantovani A. **Liver fibrosis by FibroScan® independently of established cardiovascular risk parameters associates with macrovascular and microvascular complications in patients with type 2 diabetes**. *Liver Int* (2020) **40**. DOI: 10.1111/liv.14274
25. Wu Y, Liu F. **Aspartate aminotransferase to alanine aminotransferase ratio and the risk of diabetic nephropathy progression in patients with type 2 diabetes mellitus: A biopsy-based study**. *J Diabetes Complications* (2022) **36** 108235. DOI: 10.1016/j
26. Mostafa SA, Davies MJ, Webb D, Gray LJ, Srinivasan BT, Jarvis J. **The potential impact of using glycated haemoglobin as the preferred diagnostic tool for detecting type 2 diabetes mellitus**. *Diabetes Med* (2010) **27**. DOI: 10.1111/j
27. Wan H, Wang Y, Xiang Q, Fang S, Chen Y, Chen C. **Associations between abdominal obesity indices and diabetic complications: Chinese visceral adiposity index and neck circumference**. *Cardiovasc Diabetol* (2020) **19**. DOI: 10.1186/s12933-020-01095-4
28. Liu L, Xia R, Song X, Zhang B, He W, Zhou X. **Association between the triglyceride-glucose index and diabetic nephropathy in patients with type 2 diabetes: A cross-sectional study**. *J Diabetes Investig* (2021) **12**. DOI: 10.1111/jdi
29. Yan PJ, Xu Y, Wan Q, Feng J, Li H, Gao CL. **Decreased plasma neuregulin 4 concentration is associated with increased high-sensitivity c-reactive protein in newly diagnosed type 2 diabetes mellitus patients: A cross-sectional study**. *Acta Diabetol* (2017) **54**. DOI: 10.1007/s00592-017-1044-4
30. Damba T, Bourgonje AR, Abdulle AE, Pasch A, Sydor S, van den Berg EH. **Oxidative stress is associated with suspected non-alcoholic fatty liver disease and all-cause mortality in the general population**. *Liver Int* (2020) **40**. DOI: 10.1111/liv.14562
31. Levey AS, Inker LA, Coresh J. **GFR estimation: From physiology to public health**. *Am J Kidney Dis* (2014) **63**. DOI: 10.1053/j.ajkd.2013.12.006
32. Levey AS, Stevens LA, Schmid CH, Zhang YL, Castro AF, Feldman HI. **A new equation to estimate glomerular filtration rate**. *Ann Intern Med* (2009) **150**. DOI: 10.7326/0003-4819-150-9-200905050-00006
33. Yan P, Zhang Z, Wan Q, Zhu J, Li H, Gao C. **Association of serum uric acid with bone mineral density and clinical fractures in Chinese type 2 diabetes mellitus patients: A cross-sectional study**. *Clin Chim Acta* (2018) **486** 76-85. DOI: 10.1016/j.cca.2018.07.033
34. Amatruda JG, Katz R, Peralta CA, Estrella MM, Sarathy H, Fried LF. **Association of non-steroidal anti-inflammatory drugs with kidney health in ambulatory older adults**. *J Am Geriatr Soc* (2021) **69**. DOI: 10.1111/jgs.16961
35. Lu B, Yang Z, Wang M, Yang Z, Gong W, Yang Y. **High prevalence of diabetic neuropathy in population-based patients diagnosed with type 2 diabetes in the shanghai downtown**. *Diabetes Res Clin Pract* (2010) **88**. DOI: 10.1016/j.diabres.2010.02.002
36. Bansal D, Gudala K, Muthyala H, Esam HP, Nayakallu R, Bhansali A. **Prevalence and risk factors of development of peripheral diabetic neuropathy in type 2 diabetes mellitus in a tertiary care setting**. *J Diabetes Investig* (2014) **5**. DOI: 10.1111/jdi.12223
37. Boulton AJ, Armstrong DG, Albert SF, Frykberg RG, Hellman R, Kirkman MS. **Comprehensive foot examination and risk assessment: A report of the task force of the foot care interest group of the American diabetes association, with endorsement by the American association of clinical endocrinologists**. *Diabetes Care* (2008) **31**. DOI: 10.2337/dc08-9021
38. Chen J, Mohler ER, Garimella PS, Hamm LL, Xie D, Kimmel S. **Ankle brachial index and subsequent cardiovascular disease risk in patients with chronic kidney disease**. *J Am Heart Assoc* (2016) **5**. DOI: 10.1161/JAHA.116.003339
39. Jeffcoate WJ, Macfarlane RM, Fletcher EM. **The description and classification of diabetic foot lesions**. *Diabetes Med* (1993) **10**. DOI: 10.1111/j.1464-5491
40. **American Gastroenterological association medical position statement: Nonalcoholic fatty liver disease**. *Gastroenterology* (2002) **123**. DOI: 10.1053/gast.2002.36569
41. Wei D, Chen T, Li J, Gao Y, Ren Y, Zhang X. **Association of serum gamma- glutamyl transferase and ferritin with the metabolic syndrome**. *J Diabetes Res* (2015) **2015**. DOI: 10.1155/2015/741731
42. Gordin D, Hiilesmaa V, Fagerudd J, Rönnback M, Forsblom C, Kaaja R. **Pre-eclampsia but not pregnancy-induced hypertension is a risk factor for diabetic nephropathy in type 1 diabetic women**. *Diabetologia* (2007) **50**. DOI: 10.1007/s00125-006-0544-5
43. Hazlehurst JM, Woods C, Marjot T, Cobbold JF, Tomlinson JW. **Non-alcoholic fatty liver disease and diabetes**. *Metabolism* (2016) **65**. DOI: 10.1016/j.metabol.2016.01.001
44. Khan RS, Bril F, Cusi K, Newsome PN. **Modulation of insulin resistance in nonalcoholic fatty liver disease**. *Hepatology* (2019) **70**. DOI: 10.1002/hep.30429
45. Perumpail BJ, Khan MA, Yoo ER, Cholankeril G, Kim D, Ahmed A. **Clinical epidemiology and disease burden of nonalcoholic fatty liver disease**. *World J Gastroenterol* (2017) **23**. DOI: 10.3748/wjg.v23.i47.8263
46. Jarvis H, Craig D, Barker R, Spiers G, Stow D, Anstee QM. **Metabolic risk factors and incident advanced liver disease in non-alcoholic fatty liver disease (NAFLD): A systematic review and meta-analysis of population-based observational studies**. *PloS Med* (2020) **17** e1003100. DOI: 10.1371/journal.pmed.1003100
47. Martin CL, Waberski BH, Pop-Busui R, Cleary PA, Catton S, Albers JW. **Vibration perception threshold as a measure of distal symmetrical peripheral neuropathy in type 1 diabetes: Results from the DCCT/EDIC study**. *Diabetes Care* (2010) **33**. DOI: 10.2337/dc10-0616
48. Williams KH, Burns K, Twigg SM. **Differing clinical phenotype for higher alanine-aminotransferase (ALT) compared with high-risk NAFLD fibrosis score in type 2 diabetes mellitus**. *J Diabetes Complications* (2018) **32**. DOI: 10.1016/j
49. Lombardi R, Airaghi L, Targher G, Serviddio G, Maffi G, Mantovani A. **NAFLD fibrosis score (NFS) can be used in outpatient services to identify chronic vascular complications besides advanced liver fibrosis in type 2 diabetes**. *J Diabetes Complications* (2020) **34**. DOI: 10.1016/j.jdiacomp.2020.107684
50. Lv WS, Sun RX, Gao YY, Wen JP, Pan RF, Li L. **Nonalcoholic fatty liver disease and microvascular complications in type 2 diabetes**. *World J Gastroenterol* (2013) **19**. DOI: 10.3748/wjg.v19.i20.3134
51. Yan LH, Mu B, Guan Y, Liu X, Zhao N, Pan D. **Assessment of the relationship between non-alcoholic fatty liver disease and diabetic complications**. *J Diabetes Investig* (2016) **7**. DOI: 10.1111/jdi.12518
52. Mantovani A, Rigolon R, Mingolla L, Pichiri I, Cavalieri V, Salvotelli L. **Nonalcoholic fatty liver disease is associated with an increased prevalence of distal symmetric polyneuropathy in adult patients with type 1 diabetes**. *J Diabetes Complications* (2017) **31**. DOI: 10.1016/j.jdiacomp.2017.01.024
53. Ikegami T, Hyogo H, Honda A, Miyazaki T, Tokushige K, Hashimoto E. **Increased serum liver X receptor ligand oxysterols in patients with non-alcoholic fatty liver disease**. *J Gastroenterol* (2012) **47**. DOI: 10.1007/s00535-012-0585-0
54. Zhao L, Ma J, Wang S, Xie Y. **Relationship between β-cell function, metabolic control, and microvascular complications in type 2 diabetes mellitus**. *Diabetes Technol Ther* (2015) **17** 29-34. DOI: 10.1089/dia.2014.0214
55. Guo Y, Zhao X, Liu CQ, Huang ZP, Zou DJ. **A novel refined classification system for type 2 diabetes in adults: A Chinese retrospective cohort study**. *Diabetes Metab Res Rev* (2022) **38** e3577. DOI: 10.1002/dmrr.3577
56. Qiao X, Zheng H, Zhang S, Liu S, Xiong Q, Mao F. **C-peptide is independent associated with diabetic peripheral neuropathy: A community-based study**. *Diabetol Metab Syndr* (2017) **9** 12. DOI: 10.1186/s13098-017-0208-2
57. Arques S. **Human serum albumin in cardiovascular diseases**. *Eur J Intern Med* (2018) **52** 8-12. DOI: 10.1016/j.ejim.2018.04.014
58. Caraceni P, Domenicali M, Tovoli A, Napoli L, Ricci CS, Tufoni M. **Clinical indications for the albumin use: Still a controversial issue**. *Eur J Intern Med* (2013) **24**. DOI: 10.1016/j.ejim.2013.05.015
59. Sun L, Yin H, Liu M, Xu G, Zhou X, Ge P. **Impaired albumin function: a novel potential indicator for liver function damage**. *Ann Med* (2019) **51**. DOI: 10.1080/07853890.2019.1693056
60. Tan BL, Norhaizan ME, Liew WP, Sulaiman Rahman H. **Antioxidant and oxidative stress: A mutual interplay in age-related diseases**. *Front Pharmacol* (2018) **9**. DOI: 10.3389/fphar.2018.01162
61. Uribarri J, Cai W, Peppa M, Goodman S, Ferrucci L, Striker G. **Circulating glycotoxins and dietary advanced glycation endproducts: Two links to inflammatory response, oxidative stress, and aging**. *J Gerontol A Biol Sci Med Sci* (2007) **62**. DOI: 10.1093/gerona/62.4.427
62. Li L, Liu B, Lu J, Jiang L, Zhang Y, Shen Y. **Serum albumin is associated with peripheral nerve function in patients with type 2 diabetes**. *Endocrine* (2015) **50** 397-404. DOI: 10.1007/s12020-015-0588-8
63. Hu Y, Wang J, Zeng S, Chen M, Zou G, Li Y. **Association between serum albumin levels and diabetic peripheral neuropathy among patients with type 2 diabetes: Effect modification of body mass index**. *Diabetes Metab Syndr Obes* (2022) **15**. DOI: 10.2147/DMSO.S347349
64. Yan P, Tang Q, Wu Y, Wan Q, Zhang Z, Xu Y. **Serum albumin was negatively associated with diabetic peripheral neuropathy in Chinese population: A cross- sectional study**. *Diabetol Metab Syndr* (2021) **13** 100. DOI: 10.1186/s13098-021-00718-4
65. Aleidan FAS, Ahmad BA, Alotaibi FA, Aleesa DH, Alhefdhi NA, Badri M. **Prevalence and risk factors for diabetic peripheral neuropathy among Saudi hospitalized diabetic patients: A nested case-control study**. *Int J Gen Med* (2020) **13**. DOI: 10.2147/IJGM.S273807
66. Liu H, Ding C, Hu L, Li M, Zhou W, Wang T. **The association between AST/ ALT ratio and all-cause and cardiovascular mortality in patients with hypertension**. *Med (Baltimore)* (2021) **100** e26693. DOI: 10.1097/MD.0000000000026693
67. Day CP. **From fat to inflammation**. *Gastroenterology* (2006) **130**. DOI: 10.1053/j.gastro.2005.11.017
68. Stefan N, Kantartzis K, Häring HU. **Causes and metabolic consequences of fatty liver**. *Endocr Rev* (2008) **29**. DOI: 10.1210/er.2008-0009
69. Tilg H, Moschen AR. **Insulin resistance, inflammation, and non-alcoholic fatty liver disease**. *Trends Endocrinol Metab* (2008) **19**. DOI: 10.1016/j.tem.2008.08.005
70. Rief P, Pichler M, Raggam R, Hafner F, Gerger A, Eller P. **The AST/ALT (De- ritis) ratio: A novel marker for critical limb ischemia in peripheral arterial occlusive disease patients**. *Med (Baltimore)* (2016) **95**. DOI: 10.1097/MD.0000000000003843
71. Goessling W, Friedman LS. **Increased liver chemistry in an asymptomatic patient**. *Clin Gastroenterol Hepatol* (2005) **3**. DOI: 10.1016/S1542-3565(05)00416-7
72. Chen SL, Li JP, Li LF, Zeng T, He X. **Elevated preoperative serum alanine Aminotransferase/Aspartate aminotransferase (ALT/AST) ratio is associated with better prognosis in patients undergoing curative treatment for gastric adenocarcinoma**. *Int J Mol Sci* (2016) **17**. DOI: 10.3390/ijms17060911
73. Labayen I, Ruiz JR, Huybrechts I, Ortega FB, Castillo M, Sjöstrom M. **Ideal cardiovascular health and liver enzyme levels in European adolescents; the HELENA study**. *J Physiol Biochem* (2017) **73**. DOI: 10.1007/s13105-016-0546-9
|
---
title: Tryptophan metabolism as bridge between gut microbiota and brain in chronic
social defeat stress-induced depression mice
authors:
- Jing Xie
- Wen-tao Wu
- Jian-jun Chen
- Qi Zhong
- Dandong Wu
- Lingchuan Niu
- Sanrong Wang
- Yan Zeng
- Ying Wang
journal: Frontiers in Cellular and Infection Microbiology
year: 2023
pmcid: PMC9999000
doi: 10.3389/fcimb.2023.1121445
license: CC BY 4.0
---
# Tryptophan metabolism as bridge between gut microbiota and brain in chronic social defeat stress-induced depression mice
## Abstract
### Backgrounds
Gut microbiota plays a critical role in the onset and development of depression, but the underlying molecular mechanisms are unclear. This study was conducted to explore the relationships between gut microbiota and host’s metabolism in depression.
### Methods
Chronic social defeat stress (CSDS) model of depression was established using C57BL/6 male mice. Fecal samples were collected from CSDS group and control group to measure gut microbiota and microbial metabolites. Meanwhile, tryptophan metabolism-related metabolites in hippocampus were also analyzed.
### Results
CSDS successfully induced depressive-like behaviors in CSDS group. The 24 differential bacterial taxa between the two groups were identified, and 14 ($60.87\%$) differential bacterial taxa belonged to phylum Firmicutes. Functional analysis showed that tryptophan metabolism was significantly affected in CSDS mice. Meanwhile, 120 differential microbial metabolites were identified, and two key tryptophan metabolism-related metabolites (tryptophan and 5-hydroxytryptophan (5-HTP)) were significantly decreased in feces of CSDS mice. The correlation analysis found the significant relationships between tryptophan and differential bacterial taxa under Firmicutes, especially genus Lactobacillus ($r = 0.801$, $$p \leq 0.0002$$). In addition, the significantly decreased 5-hydroxytryptamine (5-HT) in hippocampus of depressed mice was also observed.
### Conclusions
Our results showed that tryptophan metabolism might have an important role in the crosstalk between gut microbioa and brain in depression, and phylum Firmicutes, especially genus Lactobacillus, might be involved in the onset of depression through regulating tryptophan metabolism.
## Introduction
Depression is a common mental disorder that severely affects quality of life and psychosocial functioning of patients (Affatato et al., 2021). During a depressive episode, people will experience different symptoms, such as empty, low self-worth, and a loss of pleasure, even thoughts about suicide (Grover and Adarsh, 2022; Ning et al., 2022). According to the reports of World Health Organization, depression affects about $3.8\%$ of the population and causes huge economic burdens on individual, family and society. However, the pathogenesis of depression is yet unclear (He et al., 2022; Zhong et al., 2022), which results in two serious problems: i) only about $70\%$ patients response to the first-line antidepressants; and ii) no objective laboratory methods are developed to diagnose depression (Cai et al., 2022; Mojtabavi et al., 2022; Tian et al., 2023). Therefore, it is urgently needed to further explore the pathogenesis of depression.
Gut microbiota recently has been considered to be participated in the onset and development of depression (Jiang et al., 2021; Liu et al., 2021; Pu et al., 2022; Song et al., 2022). Previous study reported that gut microbiota might play a key role in the physiopathology of depression by regulating brain neurotransmitters (Huang and Wu, 2021). Carlessi et al. suggested that the disordered gut microbiota would cause a systemic inflammation, which eventually influenced responses to depression treatment (Carlessi et al., 2021). In our previous studies, we found that there were significant differences on gut microbiota compositions between healthy controls and depression patients, and gut microbiota might participate in the onset of depressive-like behaviors by affecting host’s metabolism (Zheng et al., 2016; Chen et al., 2020; Bai et al., 2021). Using depression mice model, we found that gut microbiota might influence the levels of neurotransmitters in hypothalamus through its metabolic products (Wu et al., 2020).
Although much work has been done, the detailed mechanism of gut microbitoa in depression has still not been completely understood. Thus, in this study, we established depression mice model using chronic social defeat stress (CSDS) method to further investigate the role of gut microbiota in the pathogenesis of depression. The 16S rRNA gene sequencing analysis was used to identify the differential gut microbiota, and then the functions of these differerntial bacterial taxa were predicted. Meanwhile, microbial metabolites in feces were also detected using liquid chromatography-mass spectrometry (LC-MS). Considering the important role tryptophan metabolism in the pathogenesis of depression (Chojnacki et al., 2022; Lu et al., 2022; Pu et al., 2022; Xiao et al., 2022), tryptophan metabolism-related metabolites in hippocampus were also analyzed. Integrating these data, we sought to find out the potential pathways in the crosstalk of gut and brain in depressed mice.
## Depression model
This study was performed according to the National Institutes of Health’s Animal Research Guide, and Ethics Committee of Chongqing Medical University reviewed and approved this study. C57BL/6 male mice were provided by Laboratory Animal Center of Chongqing Medical University (Chongqing, China) and housed in groups of five. CD1 male breeders were used as aggressors for this study and singly housed. After one week adaptation, we randomly assigned the C57BL/6 male mice into control group and CSDS group. Sucrose preference and body weight (BW) were matched in the two groups. In the next 10 days, mice in CSDS group were subjected to social defeated stress from CD1, and mice in control group were not disturbed (Wang et al., 2016). The mice after the CSDS models were prepared were single-housed. Subsequently, mice in two groups underwent behavioral tests to assess whether the depression model was successfully built or not.
## Behavioral tests
The mice were subjected to behavioral tests one day after CSDS. The following behavioral tests were successively conducted here: social interaction test (SIT), sucrose preference test, open field test (OFT), and forced swim test (FST). The behavioral tests were carried out in another same room to rule out the potential effects CD1 male. The procedures of these tests were the same as those in our previous studies (Gong et al., 2021; Xie et al., 2022) (Supplementary File 1). During these tests, the operators were blinded to the group allocation of mice. Each behavioral test continued for six minutes, and we recorded the activities of each mouse in the last five minutes. Three indicators (center area time (CT) (%) in OFT, sucrose preference (SPT) in sucrose preference test and immobility time (IT) in FST) were calculated to assess whether the mice in CSDS group showed depressive-like behaviors or not. Meanwhile, BW of each mouse in the two groups was collected.
## Data collection
After completing these behavioral tests, the mice in the two groups were sacrificed. Samples were rapidly collected and then stored at -80°C. The fecal samples were used for both 16s rRNA analysis and metabolism analysis. Meanwhile, the hippocampus is used for metabolism analysis. In this study, the mice in CSDS group with a social interaction (SI) ratio>=1 were considered stress-resilient and excluded from subsequent experiments ($$n = 4$$). During the development of CSDS, three mice in CSDS group incurred bite marks: i) two mice (SI ratio <1) were still included in this study; and ii) one mouse (SI ratio >=1) was excluded from this study). The procedure of 16S rRNA gene sequencing analysis was exactly conducted in accordance to our previous studies (Wu et al., 2020; Tian et al., 2022) (Supplementary File 1), and the procedures of detecting microbial metabolites in feces using LC-MS were exactly completed according to our previous studies (Wu et al., 2020; Tian et al., 2022) (Supplementary File 1).
## Statistical analysis
Student’s t-test, nonparametric Mann-Whitney, or Pearson correlation analysis was used when appropriate. Here, four parameters (ace, chao, shannon and simpson) were calculated to evaluate the alpha diversity of gut microbiota. The principal coordinate analysis (PCoA) was used to assess the beta diversity of gut microbiota. The linear discriminant-analysis (LDA) effect size (LEfSe) was conducted to identify the differential gut microbiota between the two groups, and the phylogenetic investigation of communities by reconstruction of unobserved states (PICRUSt) analysis based on Kyoto Encyclopedia of Genes and Genomes (KEGG) database was performed to predict the potential functions of the differential gut microbiota. To identify the differential microbial metabolites (variable importance in projection (VIP) > 1.0 and p-value<0.05) between the two groups, the orthogonal partial least squares (OPLS) model was built using microbial metabolites. All the analyses was carried out using SPSS 19.0, R software 4.0 and Cytoscape 5.0, and $p \leq 0.05$ was considered to be statistically significant.
## CSDS-induced depressive-like behaviors
As shown in Figure 1A, CSDS group showed a significantly lower SI ratio of social time in SIT compared to control group ($$p \leq 0.00005$$), indicating that the mice in CSDS group had the social interaction deficiency behavior. In OFT, the total distance was similar between the two groups (Figure 1B), indicating the comparable motor functions between control mice and CSDS mice; but the CT(%) was significantly lower in CSDS group than in control group (Figure 1C, $$p \leq 0.031$$). In FST, CSDS group showed the significantly higher IT compared to control group (Figure 1D, $$p \leq 0.0036$$). During the whole CSDS, the food intake between the two groups was not statistically different. After the CSDS procedure, significant differences in both SPT (Figure 1E, $$p \leq 0.032$$) and BW (Figure 1F, $$p \leq 0.0082$$) were found between the two groups. These results showed that the mice in CSDS group had the depressive-like behaviors, demonstrating that CSDS-induced depression model was successfully established.
**Figure 1:** *Depressive-like behaviors in CSDS-induced depressed mice. (A) CSDS mice had the significantly lower SI ratio compared to control mice; (B, C) in open field test, CSDS mice had the similar total distance (B) and significantly lower center time (%) (C) compared to control mice; (D) the immobility time was significantly higher in CSDS mice than in control mice; (E, F) after CSDS procedure, both sucrose preference (%) (E) and body weight (F) were significantly lower in CSDS mice than in control mice. SI, social interaction; Con, control; CSDS, chronic social defeat stress; FST, force swimming test.*
## Differential gut microbiota
The results of within-sample (α) phylogenetic diversity analysis showed no significant differences on alpha diversity between the two groups (ace, $$p \leq 0.64$$; chao, $$p \leq 0.64$$; Shannon, $$p \leq 0.11$$; simpson, $$p \leq 0.09$$). But the results of PCoA showed that there were significant differences on gut microbiota compositions between the two group (Figure 2A, $$p \leq 0.0010$$). The relative abundance on the phylum level was described in Figure 2B, and the dominant bacteria taxa on phylum level in both groups were Firmicutes and Bacteroidota. To identify the bacterial taxa responsible for discriminating CSDS mice from control mice, the LEfSe was used here. The bacterial taxa with LDA>2.0 was identified as the differential bacterial taxa. The results showed that there were 24 differential bacterial taxa between the two groups were found (Figure 2C). The 14 ($60.87\%$) of these differential bacterial taxa belonged to phylum Firmicutes. The detailed information of these differential bacterial taxa was described in Supplementary Table S1.
**Figure 2:** *Differential gut microbiota composition between the two groups. (A) principal coordinate analysis (PCoA) showed an obvious difference in gut microbiota composition between the two groups; (B) the dominant bacteria taxa on phylum level in both groups were Firmicutes and Bacteroidota; (C) the linear discriminant-analysis effect size (LEfSe) showed that there were 23 differential bacterial taxa between the two groups, and most of them (n=13) belonged to phylum Firmicutes. Con, control; CSDS, chronic social defeat stress; LDA, linear discriminant-analysis.*
## Function predictions of differential bacterial taxa
To find out the potential functions that these differential bacterial taxa were involved in, we used PICRUSt to predict the abundance of functional categories based on the standardized OTU abundance and KEGG database. The results showed that microbial gene functions related to ‘Metabolism’ was the most, accounting for $47.92\%$ of the functional categories in the first level of KEGG pathways (Figure 3A). In the second level of KEGG pathways, we found that amino acid metabolism and lipid metabolism were the top five metabolic pathways among ‘Metabolism’ category. Further analysis found that there were 10 significantly affected metabolic pathways (the third level of KEGG pathways) in CSDS mice (Figure 3B), and tryptophan metabolism was significantly decreased in CSDS mice ($$p \leq 0.0026$$).
**Figure 3:** *Functional predictions of the differential bacterial taxa. (A) the phylogenetic investigation of communities by reconstruction of unobserved states (PICRUSt) analysis showed that the metabolism category ranked the top, with a proportion of 47.92%; (B) in the metabolism category, ten significantly affected metabolism pathways were identified.*
## Correlations between differential bacterial taxa and depressive-like behaviors
In addition, to explore the potential correlations between differential bacterial taxa and depressive-like behaviors, Pearson correlation analysis was used here. The results showed that there were close relationships between depressive-like behaviors and nine differential bacterial taxa ($55.55\%$ belonged to phylum Firmicutes): i) IT was significantly correlated with six differential bacterial taxa; ii) BW was significantly correlated with three differential bacterial taxa under phylum Firmicutes; iii) SPT was significantly correlated with genus Muribaculum under phylum Bacteroidota; and iv) CT was significantly correlated with two differential bacterial taxa (Figure 4).
**Figure 4:** *Correlations between differential bacterial taxa and depressive-like behaviors. SPT, sucrose preference; IT, immobility time; CT, center time (%); BW, body weight.*
## Differential microbial metabolites
The built OPLS model showed that the mice in CSDS group was significantly separated from mice in control group, indicating that there were divergent microbial metabolic phenotypes between CSDS group and control group (Figure 5A). Meanwhile, the results of 399-permutation test showed that the regression line of the Q2-points intersected the vertical axis below zero (Q2=-0.412), demonstrating the valid and not over-fitting of this model (Figure 5B). By analyzing the corresponding loading plots, 120 differential microbial metabolites were identified (VIP>1.0 and p-value<0.05). Among these microbial metabolites, 51 and 69 metabolites were significantly increased and decreased, respectively, in CSDS group than in control group. These differential microbial metabolites mainly belonged to lipid-related metabolism and amino acid-related metabolism. The heat-map built with the relative abundances of differential microbial metabolites showed a consistent clustering pattern within the individual groups (Figure 5C). The detailed information of these differential microbial metabolites was described in Supplementary Table S2. The results of Pearson correlation analysis indicated that depressive-like behaviors were mainly significantly correlated with differential microbial metabolites belonged to lipid-related metabolism and amino acid-related metabolism (Supplementary Figure S1).
**Figure 5:** *Differential microbial metabolites in CSDS mice. (A) the orthogonal partial least squares (OPLS) model showed that the two groups were significantly separated, indicating the divergent microbial metabolic phenotypes in CSDS mice; (B) the results of 399-permutation test suggested that the built OPLS model was valid and not over-fitting; (C) heat-map of the 120 identified differential metabolites.*
## Tryptophan metabolism as bridge between gut and brain
PICRUSt functional prediction of the disordered gut microbiota showed that tryptophan metabolism was significantly decreased in this study. Meanwhile, two metabolites (5-hydroxytryptophan and tryptophan) in tryptophan metabolism were found to be significantly decreased in feces of depressed mice here. The results of correlation analysis indicated that there were strong correlations between tryptophan and differential bacterial taxa under Firmicutes, especially genus Lactobacillus ($r = 0.801$, $$p \leq 0.0002$$) (Figure 6), suggested that the disorder of tryptophan metabolism was closely related to the disturbance of phylum Firmicutes. Significantly correlations also existed between tryptophan and IT (r=-0.540, $$p \leq 0.031$$), 5-hydroxytryptophan (5-HTP) and SPT (r=-0.562, $$p \leq 0.023$$), genus Muribaculum and 5-HTP (r=-0.632, $$p \leq 0.009$$). The detailed information of correlations was described in Supplementary Table S3. In addition, we found that there was significantly decreased level of 5-hydroxytryptamine (5-HT) in hippocampus of depressed mice ($$p \leq 0.015$$). Thus, we suggested that tryptophan metabolism might act as a bridge in the crosstalk of gut and brain (Figure 7).
**Figure 6:** *Correlations between differential bacterial taxa, 5-HTP, Tryptophan and depressive-like behaviors. SPT, sucrose preference; IT, immobility time; 5-HTP, 5-hydroxytryptophan.* **Figure 7:** *Tryptophan metabolism acting as a bridge in the crosstalk of gut and brain. 5-HTP, 5-hydroxytryptophan; 5-HT, 5-hydroxytryptamine.*
## Discussion
As a psychosocial stress model based on social conflict, CSDS has been widely used in depression studies (Lu et al., 2021; Liu et al., 2022). In the present study, we observed that the mice in the CSDS group showed obvious depressive-like behaviors, suggesting that the depression mice model was successfully established. By 16S rRNA gene sequencing analysis, we identified 24 differential bacterial taxa in CSDS mice, and most of them belonged to phylum Firmicutes. By LC-MS, we found 120 differential microbial metabolites, which mainly belonged to lipid-related metabolism and amino acid-related metabolism. Correlation analysis showed that depressive-like behaviors were closely related to differential bacterial taxa under phylum Firmicutes and differential microbial metabolites that mainly belonged to lipid-related metabolism and amino acid-related metabolism. Our results would be helpful for future exploring the role of gut microbiota in the pathogenesis of depression.
The homeostasis of gut microbiota is important for host’s health, and its disturbances were related with many diseases (Li et al., 2022; Li et al., 2022; Zhang et al., 2022). Firmicutes are bacterial phyla that dominate the entire human digestive tract, and its disturbance is usually viewed as one of hallmark in depression. But the specific molecular mechanisms between depression and phylum Firmicutes are still unclear. Lactobacillus is one of beneficial bacteria under phylum Firmicutes. Bravo et al. reported that Lactobacillus supplementation could regulate the emotional behavior of mice via the vagus nerve (Bravo et al., 2011). Dong et al. observed the significantly decreased level of Lactobacillus in depressed mice, and its level was restored after treating with antidepressants (Dong et al., 2022). Schaub et al. found that the increase level of Lactobacillus was associated with decreased depressive symptoms in depression patients receiving probiotics (Schaub et al., 2022). Here, we found the significantly decreased level of Lactobacillus in CSDS group and the strong correlation between Lactobacillus and tryptophan. These results highlighted the role of gut mcirobiota in depression and emphasized the potential of microbiota-related treatment approaches for depression.
In this study, we observed the alterations in lipid and amino acid metabolism in CSDS mice. Meanwhile, the significant lower body weight in CSDS mice was found, although the food intake between the two groups was not statistically different. However, it was uncertain whether the lower body weight in CSDS mice was caused by the disturbance of lipid and amino acid metabolism. Using chronic restraint stress-induced depression model, we found the significantly lower body weight and alterations in lipid metabolism in depressed mice (Tian et al., 2022). But using chronic unpredictable mild stress-induced depression model, we observed the similar body weight between the two groups and alterations in lipid metabolism in depressed mice (Xie et al., 2023). Previous studies reported that gut microbiota was also closely involved in regulating body weight homeostasis (Łoniewski et al., 2022; Van Hul and Cani, 2023). These results indicated that the body weight might be related with many factors, such as gut microbiota and host’s metabolism. Future studies were needed to further explore the potential mechanism under the changes of body weight in depressed mice.
Forced swim test was one of the classic behavioral tests for depression, and immobility time was the evaluation indicator. Here, the significantly higher immobility time was observed in CSDS mice compared to control mice. Anhedonia was one of core symptoms of depression, and sucrose preference test was the most frequently used method for measuring anhedonia. In this study, we identified the significantly lower sucrose preference in CSDS mice compared to control mice. Similar results about immobility time and sucrose preference between depressed mice and control mice were also observed in our previous studies (Tian et al., 2022; Xie et al., 2023). In one study, we found that the significantly decreased level of 5-HT, one of neurotransmitter in tryptophan metabolism, was significantly correlated with immobility time and sucrose preference (Tian et al., 2022). Meanwhile, we also found that two main neurotransmitters in tryptophan metabolism (tryptophan and 5-HT) were significantly decreased in plasma of depression patients (Pan et al., 2018). These results suggested that tryptophan metabolism might have an important role in the onset of depressive-like behaviors.
Tryptophan is the sole precursor of 5-HT, which is an important monoamine neurotransmitter. Many studies have reported that the level of 5-HT was significantly decreased in depression (Cai et al., 2022; Zhu et al., 2022). Our previous studies also found the significantly decreased level of 5-HT in depression patients’ plasma and brain areas of depressed mice (Wu et al., 2020; Tian et al., 2022; Xie et al., 2023). 5-HT is mainly produced in gut, but it cannot cross the blood-brain barrier. However, its direct biosynthetic precursor, 5-HTP can cross blood-brain barrier. In this study, we found the significantly decreased levels of tryptophan and 5-HTP in feces of depressed mice. Considering the decreased level of 5-HT in hippocampus, we deduced that the disordered gut microbiota, especially Lactobacillus, resulted in the significantly decreased tryptophan metabolism, which caused the decreased level of 5-HTP; then the decreased 5-HTP level caused the lower level of 5-HT in hippocampus, and eventually the mice showed depressive-like behaviors. These results demonstrated that tryptophan metabolism might be a bridge between gut microbiota and brain in depressed mice.
## Conclusion
In conclusion, the CSDS successfully induced depressive-like behaviors in mice, and the depressed mice showed significantly different gut microbiota compositions compared to control mice. Most of the differential bacterial taxa belonged to phylum Firmicutes. Meanwhile, there were divergent microbial metabolic phenotypes between control mice and CSDS mice. The identified differential microbial metabolites mainly belonged to lipid-related metabolism and amino acid-related metabolism. Further analysis showed that tryptophan metabolism was significantly decreased, and there was strong correlation between genus Lactobacillus under Firmicutes and tryptophan. Considering the significantly decreased level of 5-HTP in feces and 5-HT in hippocampus, our findings indicated that Firmicutes, especially genus Lactobacillus, might play a key role in the onset of depression via tryptophan metabolism.
## Data availability statement
The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found below: https://www.ncbi.nlm.nih.gov/bioproject/PRJNA909169.
## Ethics statement
This study was performed according to the National Institutes of Health’s Animal Research Guide, and Ethics Committee of Chongqing Medical University reviewed and approved this study.
## Author contributions
YW, JX, W-TW and J-JC performed material preparation, data collection and analysis. JX, W-TW and J-JC wrote the first draft of the manuscript. QZ, DW, LN, SW and YZ performed model built, software and methodology. YW conducted writing-reviewing and editing. All authors contributed to the article and approved the submitted version.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fcimb.2023.1121445/full#supplementary-material
## References
1. Affatato O., Moulin T. C., Pisanu C., Babasieva V. S., Russo M., Aydinlar E. I.. **High efficacy of onabotulinumtoxinA treatment in patients with comorbid migraine and depression: a meta-analysis**. *J. Transl. Med.* (2021) **19** 133. DOI: 10.1186/s12967-021-02801-w
2. Bai S., Xie J., Bai H., Tian T., Zou T., Chen J. J.. **Gut microbiota-derived inflammation-related serum metabolites as potential biomarkers for major depressive disorder**. *J. Inflammation Res.* (2021) **14** 3755-3766. DOI: 10.2147/JIR.S324922
3. Bravo J. A., Forsythe P., Chew M. V., Escaravage E., Savignac H. M., Dinan T. G.. **Ingestion of lactobacillus strain regulates emotional behavior and central GABA receptor expression in a mouse**. *Proc. Natl. Acad. Sci. U S A.* (2011) **108** 16050-16055. DOI: 10.1073/pnas.1102999108
4. Cai W., Wang X. F., Wei X. F., Zhang J. R., Hu C., Ma W.. **Does urinary metabolite signature act as a biomarker of post-stroke depression**. *Front. Psychiatry* (2022) **13**. DOI: 10.3389/fpsyt.2022.928076
5. Cai T., Zheng S. P., Shi X., Yuan L. Z., Hu H., Zhou B.. **Therapeutic effect of fecal microbiota transplantation on chronic unpredictable mild stress-induced depression**. *Front. Cell Infect. Microbiol.* (2022) **12**. DOI: 10.3389/fcimb.2022.900652
6. Carlessi A. S., Borba L. A., Zugno A. I., Quevedo J., Réus G. Z.. **Gut microbiota-brain axis in depression: The role of neuroinflammation**. *Eur. J. Neurosci.* (2021) **53** 222-235. DOI: 10.1111/ejn.14631
7. Chen J. J., He S., Fang L., Wang B., Bai S. J., Xie J.. **Age-specific differential changes on gut microbiota composition in patients with major depressive disorder**. *Aging (Albany NY).* (2020) **12** 2764-2776. DOI: 10.18632/aging.102775
8. Chojnacki C., Konrad P., Błońska A., Medrek-Socha M., Przybylowska-Sygut K., Chojnacki J.. **Altered tryptophan metabolism on the kynurenine pathway in depressive patients with small intestinal bacterial overgrowth**. *Nutrients* (2022) **14** 3217. DOI: 10.3390/nu14153217
9. Dong Z., Xie Q., Xu F., Shen X., Hao Y., Li J.. **Neferine alleviates chronic stress-induced depression by regulating monoamine neurotransmitter secretion and gut microbiota structure**. *Front. Pharmacol.* (2022) **13**. DOI: 10.3389/fphar.2022.974949
10. Gong X., Huang C., Yang X., Chen J., Pu J., He Y.. **Altered fecal metabolites and colonic glycerophospholipids were associated with abnormal composition of gut microbiota in a depression model of mice**. *Front. Neurosci.* (2021) **15**. DOI: 10.3389/fnins.2021.701355
11. Grover S., Adarsh H.. **A comparative study of prevalence of mixed features in patients with unipolar and bipolar depression**. *Asian J. Psychiatr.* (2022) **81** 103439. DOI: 10.1016/j.ajp.2022.103439
12. He L., Zheng Y., Huang L., Ye J., Ye Y., Luo H.. **Nrf2 regulates the arginase 1+ microglia phenotype through the initiation of TREM2 transcription, ameliorating depression-like behavior in mice**. *Transl. Psychiatry* (2022) **12** 459. DOI: 10.1038/s41398-022-02227-y
13. Huang F., Wu X.. **Brain neurotransmitter modulation by gut microbiota in anxiety and depression**. *Front. Cell Dev. Biol.* (2021) **9**. DOI: 10.3389/fcell.2021.649103
14. Jiang W., Gong L., Liu F., Ren Y., Mu J.. **Alteration of gut microbiome and correlated lipid metabolism in post-stroke depression**. *Front. Cell Infect. Microbiol.* (2021) **11**. DOI: 10.3389/fcimb.2021.663967
15. Li Y., Han M., Song J., Liu S., Wang Y., Su X.. **The prebiotic effects of soluble dietary fiber mixture on renal anemia and the gut microbiota in end-stage renal disease patients on maintenance hemodialysis: A prospective, randomized, placebo-controlled study**. *J. Transl. Med.* (2022) **20** 599. DOI: 10.1186/s12967-022-03812-x
16. Li X., Li R., Ji B., Zhao L., Wang J., Yan T.. **Integrative metagenomic and metabolomic analyses reveal the role of gut microbiota in antibody-mediated renal allograft rejection**. *J. Transl. Med.* (2022) **20** 614. DOI: 10.1186/s12967-022-03825-6
17. Liu X., Gu X. H., Zheng L. L., Xu L. J., Yang Y. J., Yang G.. **Autophagy promotes membrane trafficking of NR2B to alleviate depression by inhibiting AQP4 expression in mice**. *Exp. Cell Res.* (2022) **419** 113298. DOI: 10.1016/j.yexcr.2022.113298
18. Liu Y., Wang H., Gui S., Zeng B., Pu J., Zheng P.. **Proteomics analysis of the gut-brain axis in a gut microbiota-dysbiosis model of depression**. *Transl. Psychiatry* (2021) **11** 568. DOI: 10.1038/s41398-021-01689-w
19. Łoniewski I., Szulińska M., Kaczmarczyk M., Podsiadło K., Styburski D., Skonieczna-Żydecka K.. **Analysis of correlations between gut microbiota, stool short chain fatty acids, calprotectin and cardiometabolic risk factors in postmenopausal women with obesity: A cross-sectional study**. *J. Transl. Med.* (2022) **20** 585. DOI: 10.1186/s12967-022-03801-0
20. Lu J., Gong X., Yao X., Guang Y., Yang H., Ji R.. **Prolonged chronic social defeat stress promotes less resilience and higher uniformity in depression-like behaviors in adult male mice**. *Biochem. Biophys. Res. Commun.* (2021) **553** 107-113. DOI: 10.1016/j.bbrc.2021.03.058
21. Lu C., Wei Z., Wang Y., Li S., Tong L., Liu X.. **Soy isoflavones alleviate lipopolysaccharide-induced depressive-like behavior by suppressing neuroinflammation, mediating tryptophan metabolism and promoting synaptic plasticity**. *Food Funct.* (2022) **13** 9513-9522. DOI: 10.1039/D2FO01437H
22. Mojtabavi H., Shaka Z., Momtazmanesh S., Ajdari A., Rezaei N.. **Circulating brain-derived neurotrophic factor as a potential biomarker in stroke: A systematic review and meta-analysis**. *J. Transl. Med.* (2022) **20** 126. DOI: 10.1186/s12967-022-03312-y
23. Ning H., Zhou H., Ren J., Zhou G., Yang N., Wang Z.. **Zishen pingchan granules combined with pramipexole in the improvement of depressive symptoms in parkinson's disease: A prospective, multicenter, randomized, double-blind, controlled clinical study**. *J. Transl. Med.* (2022) **20** 357. DOI: 10.1186/s12967-022-03551-z
24. Pan J. X., Xia J. J., Deng F. L., Liang W. W., Wu J., Yin B. M.. **Diagnosis of major depressive disorder based on changes in multiple plasma neurotransmitters: A targeted metabolomics study**. *Transl. Psychiatry* (2018) **8** 130. DOI: 10.1038/s41398-018-0183-x
25. Pu J., Liu Y., Gui S., Tian L., Yu Y., Wang D.. **Effects of pharmacological treatment on metabolomic alterations in animal models of depression**. *Transl. Psychiatry* (2022) **12** 175. DOI: 10.1038/s41398-022-01947-5
26. Pu Y., Zhang Q., Tang Z., Lu C., Wu L., Zhong Y.. **Fecal microbiota transplantation from patients with rheumatoid arthritis causes depression-like behaviors in mice through abnormal T cells activation**. *Transl. Psychiatry* (2022) **12** 223. DOI: 10.1038/s41398-022-01993-z
27. Schaub A. C., Schneider E., Vazquez-Castellanos J. F., Schweinfurth N., Kettelhack C., Doll J. P. K.. **Clinical, gut microbial and neural effects of a probiotic add-on therapy in depressed patients: A randomized controlled trial**. *Transl. Psychiatry* (2022) **12** 227. DOI: 10.1038/s41398-022-01977-z
28. Song J., Zhou B., Kan J., Liu G., Zhang S., Si L.. **Gut microbiota: Linking nutrition and perinatal depression**. *Front. Cell Infect. Microbiol.* (2022) **12**. DOI: 10.3389/fcimb.2022.932309
29. Tian T., Mao Q., Xie J., Wang Y., Shao W. H., Zhong Q.. **Multi-omics data reveals the disturbance of glycerophospholipid metabolism caused by disordered gut microbiota in depressed mice**. *J. Adv. Res.* (2022) **39** 135-145. DOI: 10.1016/j.jare.2021.10.002
30. Tian T., Qin Y., Wu M., Wang W., Song T., Deng X.. **Differential gut microbiota and microbial metabolites in adolescents with depression**. *Asian J. Psychiatr.* (2023) **83** 103496. DOI: 10.1016/j.ajp.2023.103496
31. Van Hul M., Cani P. D.. **The gut microbiota in obesity and weight management: Microbes as friends or foe**. *Nat. Rev. Endocrinol.* (2023) **17**. DOI: 10.1038/s41574-022-00794-0
32. Wang W., Guo H., Zhang S. X., Li J., Cheng K., Bai S. J.. **Targeted metabolomic pathway analysis and validation revealed glutamatergic disorder in the prefrontal cortex among the chronic social defeat stress mice model of depression**. *J. Proteome Res.* (2016) **15** 3784-3792. DOI: 10.1021/acs.jproteome.6b00577
33. Wu M., Tian T., Mao Q., Zou T., Zhou C. J., Xie J.. **Associations between disordered gut microbiota and changes of neurotransmitters and short-chain fatty acids in depressed mice**. *Transl. Psychiatry* (2020) **10** 350. DOI: 10.1038/s41398-020-01038-3
34. Xiao W., Li J., Gao X., Yang H., Su J., Weng R.. **Involvement of the gut-brain axis in vascular depression**. *Exp. Neurol.* (2022) **358** 114225. DOI: 10.1016/j.expneurol.2022.114225
35. Xie J., Wang Y., Zhong Q., Bai S. J., Zhou C. J., Tian T.. **Associations between disordered microbial metabolites and changes of neurotransmitters in depressed mice**. *Front. Cell Infect. Microbiol.* (2022) **12**. DOI: 10.3389/fcimb.2022.906303
36. Xie J., Zhong Q., Wu W. T., Chen J. J.. **Multi-omics data reveals the important role of glycerophospholipid metabolism in the crosstalk between gut and brain in depression**. *J. Transl. Med.* (2023) **21** 93. DOI: 10.1186/s12967-023-03942-w
37. Zhang P., Tang A., Geng Y., Lai J., Gao X., Pan Y.. **Gut microbial trajectory in patients with bipolar depression: A longitudinal study**. *Asian J. Psychiatr.* (2022) **73** 103098. DOI: 10.1016/j.ajp.2022.103098
38. Zheng P., Zeng B., Zhou C., Liu M., Fang Z., Xu X.. **Gut microbiome remodeling induces depressive-like behaviors through a pathway mediated by the host's metabolism**. *Mol. Psychiatry* (2016) **21** 786-796. DOI: 10.1038/mp.2016.44
39. Zhong Q., Chen J. J., Wang Y., Shao W. H., Zhou C. J., Xie P.. **Differential gut microbiota compositions related with the severity of major depressive disorder**. *Front. Cell Infect. Microbiol.* (2022) **12**. DOI: 10.3389/fcimb.2022.907239
40. Zhu Y., He J., Wu C., Wu J., Cheng Z., Chen Y.. **Transcranial ultrasound stimulation relieves depression in mice with chronic restraint stress**. *J. Neural Eng.* (2022). DOI: 10.1088/1741-2552/ac8bfd
|
---
title: 'Interaction effects of anxiety and outdoor activity spaces on frailty among
nursing home residents in Jinan, China: Is there a gender difference?'
authors:
- Meng Zhao
- Tiange Qu
- Yang Li
- Yaqi Wang
- Ming Li
- Kefang Wang
journal: Frontiers in Public Health
year: 2023
pmcid: PMC9999001
doi: 10.3389/fpubh.2023.1133340
license: CC BY 4.0
---
# Interaction effects of anxiety and outdoor activity spaces on frailty among nursing home residents in Jinan, China: Is there a gender difference?
## Abstract
### Background
Anxiety and the physical environment are critical factors influencing frailty among older adults; however, the interaction effect of anxiety and the physical environment, such as outdoor activity spaces, on frailty has not been examined. This study aimed to investigate the interaction effect of anxiety and outdoor activity spaces on frailty and to identify differences by gender.
### Methods
A total of 353 nursing home residents (197 women; 156 men; age ≥ 60 years) from 27 Chinese nursing homes were included in the analysis. Anxiety and frailty were analyzed using the Generalized Anxiety Disorder Scale and the FRAIL-NH Scale, respectively. Outdoor activity spaces were assessed through on-site observations using self-designed items. Demographic and socioeconomic information and health-related covariates were also collected. Interaction effect analyses were conducted using multilevel mixed-effects linear models.
### Results
Anxiety and outdoor activity spaces had an interaction effect on frailty among nursing home residents (β = −1.32, $95\%$ CI: −2.44, −0.20). However, further analysis demonstrated that this interaction effect was only significant in older women (β = −1.60, $95\%$ CI: −2.93, −0.27) but not in older men (β = −0.23, $95\%$ CI: −2.29, 1.82).
### Conclusions
This study highlighted that gender differences should be considered when preventing frailty in older adults with anxiety. Furthermore, it may be beneficial for nursing homes to provide outdoor activity spaces and create a supportive living environment to help delay or reverse frailty among female nursing home residents.
## 1. Introduction
Frailty is a potentially reversible state characterized by declined physiological reserves across multiple systems and accompanied by increased vulnerability to stressors [1]. Individuals with frailty have an elevated risk of adverse health outcomes, including falls, hospitalization, disability, and mortality [1, 2]. Compared with community-dwelling older adults, individuals residing in nursing homes might be more vulnerable and tend to simultaneously have multiple risk factors for frailty, such as comorbidities and malnutrition. The pooled prevalence of frailty in nursing homes residents is $52.3\%$, which is much higher than the $10.7\%$ reported in community settings [3, 4]. Accordingly, the prevention and management of frailty could be more challenging in nursing homes. Identifying modifiable frailty factors is the first step toward formulating primary prevention and restorative strategies.
Anxiety is strongly associated with accelerated frailty and poorer health status in later life [5, 6]. Anxiety has been described as a “silent geriatric giant,” [7] as it is highly prevalent among older adults. Approximately 6.5–$58.4\%$ of nursing home residents are reported to experience anxiety [8, 9]; however, they rarely seek help for it [10]. Older adults with anxiety are more likely to report and experience health problems such as falls, pain, and chronic illnesses [11, 12], which further contribute to the occurrence of frailty. A prospective cohort study in western China including 4,103 community-dwelling older adults demonstrated that individuals with comorbid depressive and anxiety symptoms were seven times more likely to become frail than those without [5]. Another study in older surgical patients revealed that anxiety was related to frailty [6]. Previous studies have shown that the impact of anxiety on frailty may be reversible and avoidable [13], suggesting that anxiety is potentially a remedial risk factor for frailty. However, few studies have investigated the protective factors that help reduce frailty in older adults with anxiety. As such, it is valuable to consider factors that may aid in buffering the detrimental effect of anxiety on frailty in later life.
In more recent years, contexts have received increasing attention for their role in frailty. An emerging body of research has focused on exploring the association between neighborhood environments and frailty. Specifically, life spaces, walking environments, aesthetic quality, accessible exercise facilities, and basic infrastructure in neighborhoods are important contributors to the level of frailty (14–16). For nursing home residents, the living environment is also recognized as a vital context for residents' health. This is because they typically spend a great deal of time there and rely heavily on social connections and resources to maintain health because of their limited mobility and functioning. Particularly, during the coronavirus disease 2019 pandemic, older adults were restricted from leaving nursing homes, and institutional outdoor spaces were one of the few places for older adults to spend time outside. Hence, high exposure to outdoor environments may potentially affect the health status of nursing home residents. Theories of environmental gerontology state that individuals are influenced by an ongoing interaction between individual, social, and physical environments [17]. As a consistent and proximate aspect of the physical environment, outdoor activity spaces may mitigate the effects of anxiety on frailty.
Furthermore, no studies have assessed whether the interaction effect of anxiety and outdoor activity spaces on frailty may vary by gender. Numerous studies have revealed significant gender differences in anxiety and frailty, with women having higher odds of being anxious and frail than men (18–20). Significant gender differences were also found in the association between anxiety and frailty. Compared with older men without comorbid depression and anxiety, women with anxiety alone had a higher prevalence of frailty [11]. Previous literature has reported gender differences in the effect of the neighborhood environment on health. For instance, Stafford et al. [ 21] showed that physical characteristics of the neighborhood were more strongly associated with women's than men's health. They suggest that the residential environment may be more important for women's health, perhaps because women have greater exposure to their neighborhood environment or are more vulnerable to its effects. Given these findings, we speculate that the interaction effect of outdoor activity spaces and anxiety on frailty is greater for women than men.
Therefore, we hypothesized that outdoor activity spaces play a role in buffering against the adverse effects of anxiety on frailty, which may exert a stronger buffering role in older women than in male residents. This study aimed to explore the interaction effect of outdoor activity spaces and anxiety on frailty among nursing home residents in China and analyse potential gender differences.
## 2.1. Participants
This cross-sectional, descriptive study was conducted among nursing homes in five districts (Lixia, Tianqiao, Huaiyin, Shizhong, and Licheng) in Jinan, Shandong Province, China, from March to June 2018. Twenty-seven nursing homes were sampled for the study from the 69 nursing homes registered at the Civil Affairs Bureau with more than 30 beds and that have operated for longer than a year. Forty-two were excluded for the following reasons: refusing to participate ($$n = 28$$), relocating or renovating ($$n = 6$$), and having missing contact details ($$n = 8$$).
Only residents aged ≥60 years who had been residing in a nursing home for ≥3 months were included in this study. Exclusion criteria were (i) hearing impairments, communication disorders, comatose, or end-stage diseases; (ii) not residing in a nursing home during the study; and (iii) severe cognitive dysfunction as determined by a Mini-Mental State Examination (MMSE) score <10 [22]. In total, 353 eligible residents were invited to participate. Details on the participant enrolment process are shown in Figure 1.
**Figure 1:** *Flowchart depicting participant selection process.*
All participants were fully informed regarding the study and provided written informed consent. The study was approved by the Ethics Committee of the researchers' university (approval number 2017-R-112).
## 2.2. Data collection
The collection was collected anonymously. Prior to the survey, research assistants (well-trained nursing postgraduates and undergraduates) received uniform training on conducting structured face-to-face interviews and physical performance measurements, and they were asked to follow a standardized questioning sequence. After passing a minimum of 6 h of training, the research assistants were allowed to conduct the survey independently.
## 2.3.1. Exposure of interest: Anxiety
Anxiety was measured using the two-item Generalized Anxiety Disorder Scale [23]. Participants were asked how frequently symptoms of anxiety bothered them over the past 2 weeks (0 = “not at all;” 1 = “several days;” 2 = “more than half the days;” 3 = “nearly every day.” A total score of 0 to 2 was defined as “no anxiety,” whereas a score of 3–6 was defined as “anxiety”).
## 2.3.2. Outcome of interest: Frailty
Frailty was defined using the Chinese version of the FRAIL-NH scale [24]. The FRAIL-NH scale, which includes core elements of the frailty phenotype and frailty index, is a specific measurement tool for nursing home residents. It comprises seven components: fatigue, resistance, ambulation, incontinence, weight loss, nutritional approach, and help with dressing. Each component is graded as 0, 1, or 2. The total score ranges from 0 to 14, with higher scores indicating a higher likelihood of frailty.
## 2.3.3. Moderator of interest: Outdoor activity spaces
Outdoor activity spaces were investigated through on-site observations by research assistants. Nursing homes were considered to provide outdoor activity spaces if they contained basic and durable fitness amenities or recreational facilities, usually installed in open spaces, such as outdoor courtyards, including spacewalk machines, leg presses, treadmills, and rotary torso machines.
## 2.3.4. Covariates
A priori, we identified potential covariates for adjustment based on the knowledge of factors that might causally affect the study exposure and study outcome independent of the exposure. The demographic and socioeconomic covariates were age (years), years of education (years), marital status (married vs. single/divorced /widowed), and self-reported economic conditions (good vs. poor).
The health-related covariates included comorbidities, cognitive impairment, loneliness, and nutritional status. Comorbidities were defined as the presence of two or more chronic diseases [25]. Cognitive status was assessed using the MMSE, with scores <24 indicating cognitive impairment [22]. Loneliness was measured using a common five-point Likert scale that asked residents how often they felt lonely [26]. This variable was dichotomised prior to statistical analyses: “sometimes,” “often,” and “always” represented loneliness, whereas “seldom” or “never” represented no loneliness. Nutritional status was determined using the Mini Nutritional Assessment-Short Form [27]. The total scores for this assessment ranged from 0 to 14, with higher scores denoting better nutritional status.
## 2.4. Statistical analyses
Participant characteristics were presented as means (standard deviations) for continuous variables and frequencies (percentages) for categorical variables. Independent sample t-tests for continuous variables and the chi-squared or Fisher's exact tests for categorical variables were used to examine the differences in characteristics between men and women. As residents were clustered within nursing homes, multilevel mixed-effects linear regression models were constructed based on three models to test the proposed hypotheses. Model 1 was unadjusted, Model 2 included sociodemographic covariates, and Model 3 was additionally adjusted for health-related covariates. Furthermore, a margin plot was utilized to illustrate the interaction effect of anxiety and outdoor activity spaces. All analyses were performed using stratified analyses of gender to allow for possible differences in the subgroups.
We performed sensitivity analysis to evaluate the consistency of the results. To reduce the potential for reverse causality between anxiety and frailty, multilevel mixed-effects logistic regression models were re-estimated to identify the potential interaction effect between frailty and outdoor activity spaces on anxiety as well as gender differences.
Analyses were performed using Stata, version 14.1 (Stata Corp, College Station, TX). All statistical tests were two-sided, and $p \leq 0.05$ was considered significant.
## 3. Results
Table 1 shows the descriptive characteristics of the study population. The age range of the participants was 60–103 years, with a mean age of 79.01 years, and $55.81\%$ were women. Fifty-five participants were unable to ambulate. In total, $19.29\%$ of women and $14.10\%$ of men participants reported anxiety. Of the 27 nursing homes, only 15 had outdoor activity spaces. The mean frailty scores were ~2.37 and 2.26 for women and men, respectively. Women were generally more likely to be older and single than men. Moreover, women had lower education levels, better economic conditions, worse cognitive impairment, less loneliness, and worse nutritional status than men.
**Table 1**
| Variables | Overall (n = 353) | Female (n = 197) | Male (n = 156) | p |
| --- | --- | --- | --- | --- |
| Variables | Mean (SD) or n (%) | Mean (SD) or n (%) | Mean (SD) or n (%) | |
| Age (years) | 79.01 (8.80) | 80.93 (7.82) | 76.58 (9.38) | < 0.001 |
| Years of education | 5.26 (4.84) | 4.52 (4.52) | 8.47 (4.33) | <0.001 |
| Marital status | | | | 0.001 |
| Married | 63 (17.8) | 23 (11.68) | 40 (25.64) | |
| Single/divorced/widowed | 290 (82.2) | 174 (88.32) | 116 (74.36) | |
| Economic conditions | | | | 0.049 |
| Good | 131 (37.1) | 82 (41.62) | 49 (31.41) | |
| Poor | 222 (62.9) | 115 (58.38) | 107 (68.59) | |
| Comorbidities | | | | 0.078 |
| Yes | 265 (75.1) | 155 (78.68) | 110 (70.51) | |
| No | 88 (24.9) | 42 (21.32) | 46 (29.49) | |
| Cognitive impairment* | | | | <0.001 |
| Yes | 212 (60.2) | 95 (48.73) | 45 (28.85) | |
| No | 140 (39.8) | 101 (51.27) | 111 (71.15) | |
| Loneliness* | | | | 0.009 |
| Yes | 103 (29.4) | 46 (24.87) | 57 (36.54) | |
| No | 247 (70.6) | 148 (75.13) | 99 (63.46) | |
| Nutritional status | 9.35 (2.19) | 8.99 (2.28) | 9.79 (2.01) | 0.001 |
| Anxiety | | | | 0.254 |
| Yes | 60 (17.0) | 38 (19.29) | 22 (14.10) | |
| No | 293 (83.0) | 159 (80.71) | 134 (85.90) | |
| Outdoor activity spaces | | | | 0.255 |
| Provided | 233 (66.0) | 125 (63.45) | 108 (69.23) | |
| | 120 (34.0) | 72 (36.55) | 48 (30.77) | |
| Frailty* | 2.32 (2.49) | 2.37 (2.54) | 2.26 (2.43) | 0.695 |
The results for the association between anxiety and frailty are presented in Table 2. In the overall sample, individuals with anxiety had an increased risk of frailty compared with participants without anxiety in both unadjusted and adjusted models. Further stratification of the association according to gender revealed that men and women who experienced anxiety were at greater risk of frailty in Models 1 and 2 (women: β = $\frac{1.91}{1.92}$, $95\%$ CI: $\frac{1.07}{1.09}$, $\frac{2.75}{2.75}$; men: β = $\frac{1.65}{1.50}$, $95\%$ CI: $\frac{0.59}{0.44}$, $\frac{2.71}{2.57}$). After adjusting for all identified sociodemographic and health-related covariates, women who reported anxiety had a significantly higher risk of frailty (β = 1.25, $95\%$ CI: 0.59, 1.90), whereas men did not (β = 0.76, $95\%$ CI: −0.12, 1.63).
**Table 2**
| Model | Overall (n = 353) | Female (n = 197) | Male (n = 156) |
| --- | --- | --- | --- |
| | β (95% CI) | β (95% CI) | β (95% CI) |
| Model 1 | Model 1 | Model 1 | Model 1 |
| Anxiety (ref. no) | Anxiety (ref. no) | Anxiety (ref. no) | Anxiety (ref. no) |
| Yes | 1.86 (1.20, 2.52)‡ | 1.91 (1.07, 2.75)‡ | 1.65 (0.59, 2.71)† |
| Model 2 | Model 2 | Model 2 | Model 2 |
| Anxiety (ref. no) | Anxiety (ref. no) | Anxiety (ref. no) | Anxiety (ref. no) |
| Yes | 1.80 (1.14, 2.46)‡ | 1.92 (1.09, 2.75)‡ | 1.50 (0.44, 2.57)† |
| Model 3 | Model 3 | Model 3 | Model 3 |
| Anxiety (ref. no) | Anxiety (ref. no) | Anxiety (ref. no) | Anxiety (ref. no) |
| Yes | 1.09 (0.56, 1.62)‡ | 1.25 (0.59, 1.90)‡ | 0.76 (-0.12, 1.63) |
Table 3 and Figures 2–4 show the interaction effects of anxiety and outdoor activity spaces on frailty as well as gender differences. For the overall sample and for women, the significant interaction term suggested that outdoor activity spaces played a moderating role in anxiety and frailty (β = −1.32/−1.60, $95\%$ CI: −2.44/−2.93, −0.20/−0.27). For instance, if nursing homes provided outdoor activity spaces, women with anxiety were less likely to develop frailty. However, among men, there was no significant interaction effect between anxiety and outdoor activity spaces (β = −0.23, $95\%$ CI: −2.29, 1.82).
The sensitivity analysis revealed no interaction effect between frailty and outdoor activity spaces on anxiety and no gender differences (Supplementary Table 1).
## 4. Discussion
With an aging population in many countries, frailty has increasingly become an emerging public health issue. Environmental and individual factors are fundamental causes of frailty [2, 16]. To the best of our knowledge, only a few studies to date have explored the cross-level interaction effects of individual factors and environmental elements on frailty as well as corresponding gender differences. This study investigated the interaction effect of anxiety and outdoor activity spaces on frailty from the perspective of individual and environment interactions and examined potential gender differences. The results demonstrated that the interaction between anxiety and outdoor activity spaces was a significant predictor of frailty. However, the interaction effect was only observed among older women with anxiety and not older men with anxiety. This study provides practical guidance that may help nursing homes take measures to prevent and mitigate frailty among residents with anxiety. We hope that this study will encourage more researchers to explore frailty from the perspective of the interaction between the individual and the environment.
Consistent with previous findings [11, 12], our results demonstrated that individuals with anxiety were more likely to develop frailty. Although the exact mechanisms remain unclear, there are several possible underlying pathophysiological mechanisms. Growing evidence supports a positive association between anxiety and inflammatory cytokines, such as interleukin-6 and C-reactive protein [28, 29], which are known to be elevated in individuals with frailty [2, 30]. Another explanation is hypothalamic-pituitary-adrenal (HPA) axis dysregulation. A study with older adults in Spain suggested that serum cortisol concentration was related to increasing frailty burden [31]. Another study with residents of long-stay institutions in Brazil reported that salivary cortisol levels were positively associated with frailty [32]. In addition, a population-based study in the Netherlands reported that older adults with anxiety had a lower cortisol awakening response than those without [33]. These findings suggest that HPA axis dysregulation may increase the vulnerability of older adults to anxiety and frailty. However, in the fully adjusted model, which considered gender differences, the association between anxiety and frailty was significant only among women. Similarly, several extant studies indicated that women with mental disorders had higher levels of frailty than men [11, 34]. One possible explanation for gender differences is that women in China have lower incomes and education levels and are more likely to be widowed than men [35, 36]. These factors could contribute to anxiety. Furthermore, psychological distress may increase the likelihood of frailty in women. Another possible explanation primarily linked to gender differences is biological susceptibility. Compared with men, women with anxiety reported higher serum high-sensitivity C-reactive protein [29] and diurnal cortisol levels [37], leading to loss of muscle mass, muscle strength, weight loss, and reduced energy expenditure [30], all of which are key clinical features of frailty.
We observed a significant buffering effect of outdoor activity spaces on the association between anxiety and frailty, which is supported by the aforementioned theories of environmental gerontology [17]. Outdoor activity spaces provide an incentive for being physically active [38, 39], and serve as places for individuals to walk, run, dance, and perform other activities. This, in turn, may contribute to reducing anxiety and frailty risks by promoting a healthy lifestyle. Further, stress reduction theory indicates that exposure to outdoor environments may trigger the parasympathetic nervous system to reduce negative mental health outcomes, such as stress and anxiety [40]. Studies have reported that access to outdoor environments is psychologically restorative and promotes mental health [39]. Moreover, outdoor activity spaces create a platform for older adults to communicate and interact with others. Through the outdoor activity spaces provided by nursing homes, older adults can get out of their small living spaces and engage in social contact, which can reduce the anxiety elicited by the new environment and delay frailty. Thus, outdoor activity spaces could reduce the likelihood of frailty among older adults with anxiety.
Gender-stratified analysis indicated that outdoor activity spaces seemed to only reduce the harmful effects of anxiety on frailty in women. Self-construal theory contains an important factor that may account for this finding [41, 42]. According to this theory [41, 42], men are more likely to develop and maintain an independent self-construal, in which others are represented as separate from the self, whereas women tend to develop and maintain an interdependent self-construal, in which others are represented as part of the self [41, 42]. These gender differences in self-construal could lead to divergent coping behaviors in response to psychological symptoms. Specifically, when men experience anxiety, they are more likely to actively self-regulate and cope independently and assertively. Conversely, women with anxiety are sensitive and tend to seek emotional support and social connections from interaction with others. Strong evidence exists that frailty can be prevented by an increase in social contacts [2, 19], which is more likely to occur with outdoor activity spaces, as they can facilitate social interaction. In addition, although we did not collect any information regarding the usage of outdoor activity spaces, our supplementary analysis revealed that women had a higher level of physical activity and were more likely than men to have spent their leisure time outdoors (data not shown); therefore, women could benefit more from outdoor activity spaces than men. Moreover, a recent study demonstrated that female nursing home residents had significantly higher engagement in physical activity than men [43]. Physical activity is considered a promising method to reverse frailty [2]; hence, outdoor activity spaces are more likely to buffer the risk of frailty in older women with anxiety.
Several limitations of this study should be acknowledged when interpreting the findings. First, the data were collected at a single time point, providing useful information about their associations but precluding any assertions of causality. Further longitudinal studies are required to establish causality by using measures at various time points. Second, participants were selected from one relatively economically developed city in China, which limits the generalisability of the study findings. Future studies are required to replicate these findings with a larger, more diverse sample of older adults, considering the great diversity in the levels of economic development, such as between urban and rural areas. Third, although we collected information on outdoor activity spaces, data on various facilities, hygiene practices, and aesthetic features of outdoor activity spaces were not available for analysis, which may affect the association between anxiety and frailty. These factors merit attention in future studies.
Despite these limitations, this study provides a novel perspective for improving the wellbeing of older adults and promoting healthy aging. Providing outdoor activity spaces could be an effective mechanism to prevent frailty among older women with anxiety. We suggest that policymakers and local governments supervise and guide nursing homes to equip outdoor activity spaces and implement activity plans, as our results found that only 15 of the 27 surveyed nursing homes provided outdoor activity spaces. In addition, some nursing homes had a small area of per capita activity space, and some were reconstructed from old buildings without consideration for outdoor activity spaces, which seriously limited residents' activities. Outdoor activity spaces could also be considered as indicators for nursing home quality evaluation. Moreover, we suggest that health providers and nursing home staff encourage older women with anxiety to visit and relax in outdoor activity spaces, as nursing home residents spend up to $65\%$ of their time alone and are often physically inactive in their rooms [44].
In conclusion, this study represents an important first step in providing generalizable evidence regarding the effect of outdoor activity spaces on the relationship between anxiety and frailty among older men and women. Our findings highlighted the buffering effect of outdoor activity spaces on frailty among older women with anxiety but not among men. Moreover, the study results suggest that gender differences should be considered in the prevention of frailty in older adults with anxiety. Furthermore, it may be beneficial for nursing homes to provide outdoor activity spaces and create a supportive living environment to delay or reverse frailty among female nursing home residents.
## 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 Shandong University (approval number 2017-R-112). The patients/participants provided their written informed consent to participate in this study.
## Author contributions
ML and KW: supervision, writing-reviewing and editing, and critical revision. MZ: funding acquisition, conceptualization, methodology, investigation, data curation, writing-original draft preparation, and writing-reviewing and editing. TQ: investigation and methodology. YL: investigation. YW: conceptualization, investigation, and methodology. All authors contributed to the article and approved the submitted version.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher's note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpubh.2023.1133340/full#supplementary-material
## References
1. Fried LP, Tangen CM, Walston J, Newman AB, Hirsch C, Gottdiener J. **Frailty in older adults: evidence for a phenotype**. *J Gerontol A Biol Sci Med Sci.* (2001) **56** M146-56. DOI: 10.1093/gerona/56.3.M146
2. Dent E, Martin FC, Bergman H, Woo J, Romero-Ortuno R, Walston JD. **Management of frailty: opportunities, challenges, and future directions**. *Lancet.* (2019) **394** 1376-86. DOI: 10.1016/S0140-6736(19)31785-4
3. Collard RM, Boter H, Schoevers RA, Oude Voshaar RC. **Prevalence of frailty in community-dwelling older persons: a systematic review**. *J Am Geriatr Soc.* (2012) **60** 1487-92. DOI: 10.1111/j.1532-5415.2012.04054.x
4. Kojima G. **Prevalence of frailty in nursing homes: A systematic review and meta-analysis**. *J Am Med Dir Assoc.* (2015) **16** 940-5. DOI: 10.1016/j.jamda.2015.06.025
5. Zhao W, Zhang Y, Liu X, Yue J, Hou L, Xia X. **Comorbid depressive and anxiety symptoms and frailty among older adults: findings from the West China health and aging trend study**. *J Affect Disord.* (2020) **277** 970-6. DOI: 10.1016/j.jad.2020.08.070
6. Wang S, Cardieri B, Mo Lin H, Liu X, Sano M, Deiner SG. **Depression and anxiety symptoms are related to pain and frailty but not cognition or delirium in older surgical patients**. *Brain Behav.* (2021) **11** e02164. DOI: 10.1002/brb3.2164
7. Cassidy KA, Rector NA. **The silent geriatric giant: anxiety disorders in late life**. *Geriatr Aging.* (2008) **11** 150-6. DOI: 10.3329/BJPP.V29I1-2.20065
8. Kang H, Smith M, Buckwalter KC, Ellingrod V, Schultz SK. **Anxiety, depression, and cognitive impairment in dementia-specific and traditional assisted living**. *J Gerontol Nurs.* (2010) **36** 18-30. DOI: 10.3928/00989134-20091204-01
9. Sandberg O, Gustafson Y, Brännström B, Bucht G. **Prevalence of dementia, delirium and psychiatric symptoms in various care settings for the elderly**. *Scand J Soc Med.* (1998) **26** 56-62. DOI: 10.1177/14034948980260011201
10. Scott T, Mackenzie CS, Chipperfield JG, Sareen J. **Mental health service use among Canadian older adults with anxiety disorders and clinically significant anxiety symptoms**. *Aging Ment Health.* (2010) **14** 790-800. DOI: 10.1080/13607861003713273
11. Clarke DM, Currie KC. **Depression, anxiety and their relationship with chronic diseases: a review of the epidemiology, risk and treatment evidence**. *Med J Aust.* (2009) **190** S54-60. DOI: 10.5694/j.1326-5377.2009.tb02471.x
12. Yohannes AM, Willgoss TG, Baldwin RC, Connolly MJ. **Depression and anxiety in chronic heart failure and chronic obstructive pulmonary disease: prevalence, relevance, clinical implications and management principles**. *Int J Geriatr Psychiatry.* (2010) **25** 1209-21. DOI: 10.1002/gps.2463
13. Borges MK, Jeuring HW, Marijnissen RM, van Munster BC, Aprahamian I, van den Brink RHS. **Frailty and affective disorders throughout adult life: a 5-year follow-up of the Lifelines Cohort Study. J Am Geriatr Soc**. (2022) **2022** 1-12. DOI: 10.1111/jgs.18021
14. Xue QL, Fried LP, Glass TA, Laffan A, Chaves PH. **Life-space constriction, development of frailty, and the competing risk of mortality: the Women's Health and Aging Study I**. *Am J Epidemiol.* (2008) **167** 240-8. DOI: 10.1093/aje/kwm270
15. Liu H, Chen B, Li Y, Morrow-Howell N. **Neighborhood resources associated with frailty trajectories over time among community-dwelling older adults in China**. *Health Place.* (2022) **74** 102738. DOI: 10.1016/j.healthplace.2021.102738
16. Ye B, Gao J, Fu H. **Associations between lifestyle, physical and social environments and frailty among Chinese older people: a multilevel analysis**. *BMC Geriatr.* (2018) **18** 314. DOI: 10.1186/s12877-018-0982-1
17. Wahl HW, Oswald F, Dannefer D, Phillipson C. **Environmental perspectives on aging**. *The SAGE Handbook of Social Gerontology* (2010) 111-24. DOI: 10.4135/9781446200933.n8
18. Gordon EH, Peel NM, Samanta M, Theou O, Howlett SE, Hubbard RE. **Sex differences in frailty: a systematic review and meta-analysis**. *Exp Gerontol.* (2017) **89** 30-40. DOI: 10.1016/j.exger.2016.12.021
19. Leme DEDC, Alves EVDC, Fattori A. **Relationships between social, physical, and psychological factors in older persons: frailty as an outcome in network analysis**. *J Am Med Dir Assoc.* (2020) **21** 1309-1315. DOI: 10.1016/j.jamda.2020.02.005
20. Lin H, Xiao S, Shi L, Zheng X, Xue Y, Yun Q. **Impact of multimorbidity on symptoms of depression, anxiety, and stress in older adults: is there a sex difference?**. *Front Psychol.* (2021) **12** 762310. DOI: 10.3389/fpsyg.2021.762310
21. Stafford M, Cummins S, Macintyre S, Ellaway A, Marmot M. **Gender differences in the associations between health and neighbourhood environment**. *Soc Sci Med.* (2005) **60** 1681-92. DOI: 10.1016/j.socscimed.2004.08.028
22. Folstein MF, Folstein SE, McHugh PR. **“Mini-mental state”: a practical method for grading the cognitive state of patients for the clinician**. *J Psychiatr Res.* (1975) **12** 189-98. DOI: 10.1016/0022-3956(75)90026-6
23. Kroenke K, Spitzer RL, Williams JB, Löwe B. **The patient health questionnaire somatic, anxiety, and depressive symptom scales: a systematic review**. *Gen Hosp Psychiatry.* (2010) **32** 345-59. DOI: 10.1016/j.genhosppsych.2010.03.006
24. Zhao M, Mou H, Zhu S, Li M, Wang K. **Cross-cultural adaptation and validation of the FRAIL-NH scale for Chinese nursing home residents: a methodological and cross-sectional study**. *Int J Nurs Stud.* (2020) **105** 103556. DOI: 10.1016/j.ijnurstu.2020.103556
25. Fried LP, Ferrucci L, Darer J, Williamson JD, Anderson G. **Untangling the concepts of disability, frailty, and comorbidity: implications for improved targeting and care**. *J Gerontol A Biol Sci Med Sci.* (2004) **59** 255-63. DOI: 10.1093/gerona/59.3.M255
26. Luo Y, Waite LJ. **Loneliness and mortality among older adults in China**. *J Gerontol B Psychol Sci Soc Sci.* (2014) **69** 633-45. DOI: 10.1093/geronb/gbu007
27. Rubenstein LZ, Harker JO, Salvà A, Guigoz Y, Vellas B. **Screening for undernutrition in geriatric practice: developing the short-form mini-nutritional assessment (MNA-SF)**. *J Gerontol A Biol Sci Med Sci.* (2001) **56** M366-72. DOI: 10.1093/gerona/56.6.M366
28. Pitsavos C, Panagiotakos DB, Papageorgiou C, Tsetsekou E, Soldatos C, Stefanadis C. **Anxiety in relation to inflammation and coagulation markers, among healthy adults: the Attica study**. *Atherosclerosis.* (2006) **185** 320-6. DOI: 10.1016/j.atherosclerosis.2005.06.001
29. Yang QQ, Shao D, Li J, Yang CL, Fan MH, Cao FL. **Positive association between serum levels of high-sensitivity C-reactive protein and depression/anxiety in female, but not male, patients with type 2 diabetes mellitus**. *Biol Res Nurs.* (2020) **22** 178-87. DOI: 10.1177/1099800419894641
30. Clegg A, Young J, Iliffe S, Rikkert MO, Rockwood K. **Frailty in elderly people**. *Lancet.* (2013) **381** 752-62. DOI: 10.1016/S0140-6736(12)62167-9
31. Marcos-Pérez D, Sánchez-Flores M, Maseda A, Lorenzo-López L, Millán-Calenti JC, Pásaro E. **Serum cortisol but not oxidative stress biomarkers are related to frailty: results of a cross-sectional study in Spanish older adults**. *J Toxicol Environ Health A.* (2019) **82** 815-25. DOI: 10.1080/15287394.2019.1654639
32. Holanda CMA, Guerra RO, Nóbrega PVN, Costa HF, Piuvezam MR, Maciel ÁCC. **Salivary cortisol and frailty syndrome in elderly residents of long-stay institutions: a cross-sectional study**. *Arch Gerontol Geriatr.* (2012) **54** e146-51. DOI: 10.1016/j.archger.2011.11.006
33. Hek K, Direk N, Newson RS, Hofman A, Hoogendijk WJ, Mulder CL. **Anxiety disorders and salivary cortisol levels in older adults: a population-based study**. *Psychoneuroendocrinology.* (2013) **38** 300-5. DOI: 10.1016/j.psyneuen.2012.06.006
34. Mutz J, Choudhury U, Zhao J, Dregan A. **Frailty in individuals with depression, bipolar disorder and anxiety disorders: longitudinal analyses of all-cause mortality**. *BMC Med.* (2022) **20** 274. DOI: 10.1186/s12916-022-02474-2
35. Girgus JS, Yang K, Ferri CV. **The gender difference in depression: are elderly women at greater risk for depression than elderly men?**. *Geriatrics.* (2017) **2** 35. DOI: 10.3390/geriatrics2040035
36. Mwinyi J, Pisanu C, Castelao E, Stringhini S, Preisig M, Schiöth HB. **Anxiety disorders are associated with low socioeconomic status in women but not in men**. *Womens Health Issues.* (2017) **27** 302-7. DOI: 10.1016/j.whi.2017.01.001
37. Chaudieu I, Beluche I, Norton J, Boulenger JP, Ritchie K, Ancelin ML. **Abnormal reactions to environmental stress in elderly persons with anxiety disorders: evidence from a population study of diurnal cortisol changes**. *J Affect Disord.* (2008) **106** 307-13. DOI: 10.1016/j.jad.2007.07.025
38. Humpel N, Owen N, Leslie E. **Environmental factors associated with adults' participation in physical activity: a review**. *Am J Prev Med.* (2002) **22** 188-99. DOI: 10.1016/S0749-3797(01)00426-3
39. Jimenez MP, Elliott EG, DeVille NV, Laden F, Hart JE, Weuve J. **Residential green space and cognitive function in a large cohort of middle-aged women**. *JAMA Netw Open.* (2022) **5** e229306. DOI: 10.1001/jamanetworkopen.2022.9306
40. Ulrich RS. **View through a window may influence recovery from surgery**. *Science.* (1984) **4647** 420-1. DOI: 10.1126/science.6143402
41. Cross SE, Hardin EE, Gercek-Swing B. **The what, how, why, and where of self–construal**. *Pers Soc Psychol Rev.* (2011) **15** 142-79. DOI: 10.1177/1088868310373752
42. Cross SE, Madson L. **Models of the self: self-construals and gender**. *Psychol Bull.* (1997) **122** 5-37. DOI: 10.1037/0033-2909.122.1.5
43. Neils-Strunjas J, Crandall KJ, Ding X, Gabbard A, Rassi S, Otto S. **Facilitators and barriers to attendance in a nursing home exercise program**. *J Am Med Dir Assoc.* (2021) **22** 803-8. DOI: 10.1016/j.jamda.2020.09.023
44. Lorenz RA, Gooneratne N, Cole CS, Kleban MH, Kalra GK, Richards KC. **Exercise and social activity improve everyday function in long-term care residents**. *Am J Geriatr Psychiatry.* (2012) **20** 468-76. DOI: 10.1097/JGP.0b013e318246b807
|
---
title: 'Clinical symptoms, thyroid dysfunction, and metabolic disturbances in first-episode
drug-naïve major depressive disorder patients with suicide attempts: A network perspective'
authors:
- Pu Peng
- Qianjin Wang
- Xiaoe Lang
- Tieqiao Liu
- Xiang-Yang Zhang
journal: Frontiers in Endocrinology
year: 2023
pmcid: PMC9999007
doi: 10.3389/fendo.2023.1136806
license: CC BY 4.0
---
# Clinical symptoms, thyroid dysfunction, and metabolic disturbances in first-episode drug-naïve major depressive disorder patients with suicide attempts: A network perspective
## Abstract
### Backgrounds
Co-occurrence of thyroid dysfunction, metabolic disturbances, and worsening clinical symptoms in major depressive disorder (MDD) patients with suicidal attempts (SA) are common. However, their relationship in SA patients remains unexplored. We aimed to [1] determine the independent association of thyroid dysfunction, clinical symptoms, and metabolic disturbances with SA; and [2] identify their interactions in SA patients via the network approach.
### Methods
1718 FEDN MDD patients were recruited. Depressive, anxiety, and psychotic symptoms were assessed by the Hamilton Rating Scale for Depression (HAMD), the Hamilton Rating Scale for Anxiety (HAMA), and the Positive and Negative Syndrome Subscale positive subscale, respectively. The serum levels of thyroid hormones and other metabolic parameters were assessed. Logistic regression model was applied to determine the correlates of SA. Network analysis was applied to determine the interaction between thyroid dysfunction, clinical symptoms, and metabolic disturbances.
### Results
SA patients had significant worse metabolic disturbances, thyroid dysfunction, and clinical symptoms than non-SA patients. Thyroid peroxidases antibody, thyroid stimulating hormone (TSH), HAMD scores, HAMA scores, and systolic blood pressure was independently associated with SA. Network analysis suggested that TSH was the hub of the network, exhibiting substantial associations with metabolic disturbances, anxiety, and psychotic symptoms in SA patients.
### Conclusions
Our work highlights the predominant role of serum TSH levels in the pathophysiology of SA. Regular thyroid function tests might help early detect SA. Targeting increased TSH levels may help reduce metabolic disturbances and clinical symptoms in SA patients.
## Introduction
Suicide is the most devastating consequence of patients with major depressive disorder (MDD). The lifetime prevalence of suicidal ideation, suicidal planning, and suicidal attempts (SA) in MDD patients is $37.7\%$, $15.1\%$, and $23.7\%$, respectively [1, 2]. A recent meta-analysis shows that MDD patients are approximately 7 times more likely to have SA in the past year than healthy individuals [3]. The high prevalence of SA highlights the strong need to identify the potential risk factors for SA, which is valuable for SA screening and intervening in MDD patients.
Although numerous studies have identified demographic and clinical risk factors for SA in MDD patients [4, 5], the biological correlates of SA remain largely unexplored [5]. Several studies have suggested that thyroid hormones and metabolic parameters can be potential biomarkers for SA (6–9). However, the reported results are inconsistent. For example, one meta-analysis published in 2020 demonstrated that SA was associated with low serum levels of total cholesterol (TC) and low-density lipoprotein cholesterol (LDL-C) in 7068 patients with MDD [7]. However, this view has been challenged by two recent large-scale studies ($$n = 1279$$ and $$n = 580$$) [10, 11], which found higher concentrations of TC and LDL-C in MDD patients with SA. Similarly, the relationship between thyroid dysfunction and SA in patients with MDD was also in debate. Several studies indicated that elevated TSH increased the risk of suicide [11], whilst some found an inverse association [12] or no association [13]. The inconsistency in the previous studies may be due to the different study samples. Studies have shown that disease duration, comorbidities, and medication may have a substantial impact on thyroid function and metabolism (14–16), which may obscure their association with SA. Therefore, assessing the relationship between thyroid dysfunction, metabolic disturbances, and SA in first-episode drug-naïve (FEND) MDD patients may provide more solid evidence.
Studies have confirmed that the co-occurrence of thyroid dysfunction and metabolic disturbances is very common (17–20), especially in patients with MDD [21]. For example, Kim et al. found that subclinical hypothyroidism increased the risk of metabolic syndrome by 7 times among individuals with depression [21]. Two recent studies also found that MDD patients with SA exhibited more severe clinical symptoms, metabolic disturbances, and thyroid dysfunction than those without [11, 22]. However, no prior study directly evaluated their relationship in patients with SA. Clarifying whether and how thyroid dysfunction, metabolic disorders, and clinical symptoms are interconnected in patients with SA may provide new insights into the pathophysiology of SA.
Network analysis, as an emerging tool, has advantages over traditional methods such as regression models in visualizing and describing independent associations between variables [23, 24]. In a network model, a variable is visualized as a “node”. After sufficient adjustment for other variables within the network, the unique association between two variables is visualized as an “edge”. In addition to identifying correlations between variables, network analysis identifies the most influential variables that are most closely linked to the other variables in the network (i.e., central variables). The central variable is considered to play an important role in triggering and maintaining the network [25]. Hence, the central variable may be a promising target for clinical interventions to reduce thyroid dysfunction, metabolic disturbances, and clinical symptoms in MDD patients with SA.
To date, emerging studies have applied network analysis to assess associations between variables in clinical medicine [26, 27]. For example, Jia et al. have assessed the association of lipid markers with cognition performance and depression through a network approach [27]. A recent study also determined the networks of lipid metabolism, inflammation, and depressive symptoms [26]. However, there are no previous studies evaluating the network of clinical symptoms, thyroid dysfunction, and metabolic disturbance in MDD patients with SA, which gave us the motivation to conduct the present study. We recruited a large sample of FEDN MDD patients and evaluated SA, metabolic parameters, thyroid hormones, and clinical symptoms. We have two main aims [1]: to determine the association of SA with clinical symptoms, metabolic disturbances, and thyroid dysfunction in first-episode drug-naïve patients with MDD; and [2] to determine the inter-relationship between metabolic disturbances, thyroid dysfunction, and clinical symptoms in patients with SA via the network approach.
## Study procedure and participants
Participants were recruited at the psychiatric outpatient department of the First Hospital of Shanxi Medical University from 2015 to 2017. Inclusion criteria were as follows: [1] fulfilling DSM-IV criteria for MDD, diagnosed by two trained psychiatrists using the Structured Clinical Interview for DSM-IV Disorders (SCID); [2] 17-item Hamilton Depression Scale (HAMD) score of more than 23; [3] age 18-60 years old, Han nationality; [4] no prior medication, including antidepressant, antipsychotic drugs, thyroid hormone therapy, hypoglycemic agents, antihypertensive and lipid-lowering drugs; and [5] depression symptoms were first-episode and the disease duration of no more than 24 months. Exclusion criteria included: [1] pregnant or breastfeeding women; [2] concurrent DSM-IV axis I disorder including bipolar disorder, schizophrenia, and schizoaffective or severe medical conditions; [3] substance use disorder except for tobacco; and [4] unwillingness to provide informed consent.
All participants provided written informed consent. This study was approved by the Institutional Review Board (IRB) of the First Hospital of Shanxi Medical University (No. 2016-Y27).
## Interview and clinical assessments
We collected basic information, including age, gender, education, onset, and duration of MDD, and married status through a self-designed questionnaire. All participants were independently interviewed face-to-face by two trained psychiatrists via the SCID. Two psychiatrists independently assessed each participant’s depression, anxiety, and psychotic symptoms by the HAMD, Hamilton Anxiety Scale (HAMA), and the positive subscale of Positive and Negative Syndrome Subscale (PANSS), respectively. HAMD score ranges from 0-52, with a cutoff point of 24 being used to determine severe depression [28]. HAMA consists of 14 items, measuring psychological and somatic anxiety symptoms [29]. It applied the 5-Likert scale, with a total score ranging from 0-56. The PANSS positive subscale assesses seven positive symptoms [30]. The PANSS-positive subscale score ranges from 7-49. Higher scores on the HAMA, HAMD, and PANSS indicate more severe symptoms. These three scales have been validated and widely used in the Chinese population (31–33). According to previous studies [34, 35], HAMA score >20 and PANSS positive subscale score >14 indicate significant anxiety and psychotic symptoms, respectively. The correlation coefficients between the two psychiatrists’ scores on all three scales were higher than 0.8.
We assessed SA through face-face interviews. All participants were asked the question: “In your lifetime, did you ever try to kill yourself?”. This single item has been validated and used widely in previous epidemiological studies for the detection of SA [36, 37]. Those who answered “yes” were considered to have lifetime SA. We further asked them about the timing and frequency of SA. We contacted the family members of the participants for the details of SA when patients were unable to provide definitive information.
## Biochemical indicators
Blood samples were collected in the morning after an overnight fast before participants received any medical treatment. Serum levels of free triiodothyronine (FT3), free thyroxine (FT4), thyroid stimulating hormone (TSH), antithyroglobulin (TgAb), thyroid peroxidase antibody (TPOAb), TC, TG, high-density lipoprotein (HDL-C), low-density lipoprotein (LDL-C), and glucose were assessed. Lipid markers (TC, TG, HDL-C, LDL-C) and glucose were measured on a Cobas E610 (Roche, Basel, Switzerland). Thyroid hormones were assayed on a Roche C6000 Electrochemiluminescence Immunoassay Analyzer (Roche Diagnostics, Indianapolis, IN, USA). Measurements were conducted in the laboratory of the First Hospital, Shanxi Medical University. The nurses measured the patients’ weight, height, and blood pressure. We calculated body mass index (BMI) according to the following formula: BMI = Weight (kg)/Height (m) 2.
According to previous studies in the Chinese population [38, 39], metabolic disturbances and thyroid dysfunction were defined as follows: [1] overweight or obesity: BMI≥24; [2] hyperglycemia: glucose≥6.1mmol/L; [3] hypertension: SBP≥140 mmHg and/or DBP≥90mmHg; [4] hypertriglyceridemia: TG≥2.3 mmol/L; [5] low HDL: HDL-C ≤ 1.0 mmol/L; [6] hypercholesterolemia: TC≥6.2 mmol/L or LDL-C≥4.1 mmol/L; [7]abnormal TgAb: TgAb≥115 IU/L; [8] abnormal TPOAb: TPOAb ≥34 IU/L; [9] subclinical hypothyroidism (SCH): TSH >4.2 mIU/L with normal fT4 concentration (10–23 pmol/L); [10] hyperthyroidism: TSH<0.27 mIU/L and FT4 >23 pmol/L, and [11] hypothyroidism: TSH >4.2 mIU/L with low FT4 concentration (<10 pmol/L).
## Data processing
According to the Shapiro-Wilk test, the continuous data in our study were not normally distributed. Therefore, we expressed the continuous data as the median and interquartile range (IRQ; 25-$75\%$) and the categorical data as frequencies and percentages. All statistical analyses were conducted on R (ver. 4.20). We adopt two-tailed tests with $p \leq 0.05$ indicating statistical significance.
## Univariate and multiple analyses
We assessed differences in metabolic disturbances, clinical symptoms, and thyroid dysfunction between MDD patients with and without SA by chi-square test, Fisher’s exact test, and Whitney U test, as appropriate. Bonferroni correction was employed for multiple testing (p’=$\frac{0.05}{40}$ = 0.00125). A multiple logistic regression model was conducted to identify independent correlates of SA. Variables with $P \leq 0.05$ in univariate tests were included in the multiple logistic regression analysis using the stepwise method.
## Network analysis
Clinical symptoms (HAMA, HAMD, and PANSS scores), metabolic parameters (TC, TG, LDL-C, HDL-C, SBP, DBP, and BMI), and thyroid hormones (TSH, TPOAb, TgAb, FT3, and FT4) were included in the network. Following a previous study [40], we performed nonparanormal transformations using Rpackage “huge” because the data were not normally distributed. We estimated and visualized the network using Rpackage “qgraph” and “bootnet” [41]. We estimated the network using the default of the EBICglasso model, which was widely used in psychological network models [42].γ was set to 0.5, which made the network more sparse and strikes a balance between sensitivity and specificity in preserving true edges. The network consisted of “nodes” (i.e., metabolic parameters, thyroid function, and clinical symptoms) and “edges” (i.e., pairwise correlations between two nodes after controlling for other variables within the network). Thicker edges implied a greater association [43]. Red edges indicated negative associations, while blue edges indicated positive associations. We calculated the centrality index “strength” to quantify the importance of the nodes. Nodes with higher strength were considered to exhibit strong associations and impacts on other nodes within the network. We also calculated the predictability of the nodes by Rpackage “MGM” [44]. Similar to the R2 in the regression model, predictability referred to the extent to which the variance of a node can be explained by other nodes in the network [45].
Finally, we evaluated the stability and accuracy of our network by Rpackage “bootnet”. Bootstrap procedures were performed with 1000 bootstrap samples to determine the accuracy of the estimated edges. We conducted a case-dropping procedure to evaluate the stability of the network. The correlation stability coefficient (CS-C) was calculated, and a CS-C above 0.5 implied reasonable stability.
## Sample characteristics
We recruited 1718 FEDN MDD patients (Table 1). The majority of the participants were female (1130, $66\%$), married (1216, $71\%$), and had a degree below college (1173, $68\%$). One-fifth of the participants (346, $20\%$) had lifetime SA. 235 ($14\%$) had SA in the past two weeks.
**Table 1**
| Variable | Overall, N = 1,7181 | Without SA, N = 1,3721 | With SA, N = 3461 | p-value2 |
| --- | --- | --- | --- | --- |
| Age, year | 34 (23, 45) | 33 (23, 45) | 35 (25, 47) | 0.023 |
| Duration, month | 5 (3, 8) | 5 (3, 8) | 6 (3, 9) | <0.001 |
| Onset, year | 34 (23, 45) | 33 (23, 45) | 34 (25, 47) | 0.026 |
| Gender | | | | 0.4 |
| Male | 588 (34%) | 476 (35%) | 112 (32%) | |
| Female | 1,130 (66%) | 896 (65%) | 234 (68%) | |
| Education | | | | 0.5 |
| Below college | 1,173 (68%) | 932 (68%) | 241 (70%) | |
| College or above | 545 (32%) | 440 (32%) | 105 (30%) | |
| Married | 1,216 (71%) | 965 (70%) | 251 (73%) | 0.4 |
| PANSS | 7 (7, 7.8) | 7 (7, 7) | 8 (7, 17.8) | <0.001 |
| Psychotic symptom | 171 (10.0%) | 83 (6.0%) | 88 (25%) | <0.001 |
| HAMD | 30 (28, 32) | 30 (28, 32) | 32 (30, 34) | <0.001 |
| HAMA | 21.0 (18.0, 23.0) | 20.0 (18.0, 22.0) | 23.0 (21.0, 26.0) | <0.001 |
| Anxiety | 894 (52%) | 610 (44%) | 284 (82%) | <0.001 |
| TSH, uIU/L | 4.91 (3.11, 6.66) | 4.63 (2.89, 6.14) | 6.76 (4.54, 8.89) | <0.001 |
| TgAb, IU/L | 21 (14, 44) | 20 (14, 32) | 28 (18, 144) | <0.001 |
| TPOAb, IU/L | 17 (12, 35) | 16 (12, 29) | 29 (14, 171) | <0.001 |
| FT3, pmol/L | 4.92 (4.38, 5.41) | 4.91 (4.39, 5.40) | 4.92 (4.34, 5.44) | >0.9 |
| FT4, pmol/L | 16.5 (14.4, 18.7) | 16.5 (14.4, 18.8) | 16.5 (14.4, 18.6) | 0.9 |
| Glucose, mmol/L | 5.34 (4.94, 5.80) | 5.28 (4.92, 5.71) | 5.56 (5.05, 6.10) | <0.001 |
| TC, mmol/L | 5.22 (4.46, 6.00) | 5.11 (4.36, 5.81) | 5.72 (4.95, 6.59) | <0.001 |
| HDLC, mmol/L | 1.23 (1.01, 1.42) | 1.25 (1.05, 1.44) | 1.13 (0.89, 1.30) | <0.001 |
| TG, mmol/L | 1.97 (1.40, 2.77) | 1.94 (1.37, 2.74) | 2.16 (1.46, 2.93) | 0.004 |
| LDLC, mmol/L | 2.96 (2.38, 3.52) | 2.90 (2.30, 3.42) | 3.21 (2.60, 3.74) | <0.001 |
| BMI, kg/m2 | 24.23 (23.22, 25.60) | 24.23 (23.23, 25.60) | 24.27 (23.18, 25.99) | 0.8 |
| SBP, mmHg | 120 (112, 127) | 120 (111, 126) | 125 (116, 134) | <0.001 |
| DBP, mmHg | 76 (70, 80) | 75 (70, 80) | 78 (74, 84) | <0.001 |
| Abnormal TgAb | 297 (17%) | 191 (14%) | 106 (31%) | <0.001 |
| Abnormal TPOAb | 438 (25%) | 282 (21%) | 156 (45%) | <0.001 |
| SCH | 1,041 (61%) | 778 (57%) | 263 (76%) | <0.001 |
| Hyperthyroidism | 5 (0.3%) | 5 (0.4%) | 0 (0%) | 0.6 |
| Hypothyroidism | 3 (0.2%) | 2 (0.1%) | 1 (0.3%) | 0.5 |
| Hyperglycemia | 241 (14%) | 153 (11%) | 88 (25%) | <0.001 |
| Low HDL | 429 (25%) | 306 (22%) | 123 (36%) | <0.001 |
| Overweight or obesity | 1,026 (60%) | 825 (60%) | 201 (58%) | 0.5 |
| High SBP | 53 (3.1%) | 16 (1.2%) | 37 (11%) | <0.001 |
| High DBP | 74 (4.3%) | 38 (2.8%) | 36 (10%) | <0.001 |
| Hypertriglyceridemia | 668 (39%) | 512 (37%) | 156 (45%) | 0.008 |
| Abnormal TC | 357 (21%) | 225 (16%) | 132 (38%) | <0.001 |
| Abnormal LDL-C | 185 (11%) | 125 (9.1%) | 60 (17%) | <0.001 |
| Hypertension | 92 (5.4%) | 42 (3.1%) | 50 (14%) | <0.001 |
| Hypercholesterolemia | 421 (25%) | 277 (20%) | 144 (42%) | <0.001 |
## The difference in metabolic disturbances, thyroid function, and clinical symptoms in FEDN MDD patients with and without SA
SA patients tended to be older, had a longer duration of disease, and had a later onset (Table 1). Compared with non-SA patients, SA patients had significantly more severe metabolic disturbances, thyroid dysfunction, and psychological distress than non-SA patients, showing higher scores on HAMD, HAMA, and PANSS positive subscale. The prevalence rates of SCH, abnormal TgAb, abnormal TPOAb, hyperglycemia, abnormal TC, abnormal LDL-C, low HDL, hypertension, and hypercholesterolemia were significantly higher in SA patients than in non-SA patients. Their associations remained significant after the Bonferroni correction. In addition, SA patients were also more likely to have hypertriglyceridemia. However, the association between hypertriglyceridemia and SA was no longer significant after multiple testing.
## Independent correlates of SA in FEDN MDD patients
We conducted a multiple logistic regression model in variables showing $p \leq 0.05$ in univariate analysis (i.e., age, duration, and the onset of MDD, HAMD, HAMA, PANSS, TSH, TPOAb, TgAb, TC, TG, HDL-C, LDL-C, glucose, SBP, and DBP). Table 2 summarizes the results of the logistic regression model. HAMD (Odds ratio, OR, 1.081, $95\%$ confidence intervals, $95\%$ CI, 1.016-1.151, $$p \leq 0.014$$), HAMA (OR, 1.251, $95\%$CI, 1.189-1.316, $p \leq 0.001$), TSH (OR, 1.115, $95\%$CI, 1.047-1.187, $$p \leq 0.001$$), TPOAb (OR, 1.002, $95\%$CI, 1.001-1.003, $p \leq 0.001$), and SBP (OR, 1.023, $95\%$CI, 1.008-1.038, $$p \leq 0.002$$) were independently associated with SA in FEDN MDD patients.
**Table 2**
| Characteristic | OR1 | 95% CI1 | p-value |
| --- | --- | --- | --- |
| HAMD | 1.081 | 1.016, 1.151 | 0.014 |
| HAMA | 1.251 | 1.189, 1.316 | <0.001 |
| TSH | 1.115 | 1.047, 1.187 | 0.001 |
| TPOAb | 1.002 | 1.001, 1.003 | <0.001 |
| SBP | 1.023 | 1.008, 1.038 | 0.002 |
## Network of thyroid dysfunction, metabolic disturbances, and clinical symptoms in MDD patients with SA
Figure 1 illustrates the network of thyroid dysfunction, metabolic disturbances, and clinical profiles in MDD patients with SA. The network was composed of 16 nodes and 32 edges. Visually, TSH was in the center of the network. It exhibited a strong positive association with metabolic parameters including SBP, TC, and glucose. TSH was also positively correlated with PANSS and HAMA. In contrast, BMI, FT3, and FT4 were at the margin of the network, exhibiting a very weak association with clinical symptoms. We also observed a strong association between PANSS, HAMA, and HAMD. The correlation matrix between the nodes is presented in Table S1.
**Figure 1:** *The network of thyroid-dysfunction, metabolic disturbances, and clinical symptoms in FEDN MDD patients with suicidal attempts. Blue, orange, and green nodes represented thyroid hormones, clinical symptoms, and metabolic parameters, respectively. Blue and red edges indicated positive and negative associations, respectively. Thicker edges suggested stronger associations. HAMD = Hamilton Depression Rating Scale, HAMA = Hamilton Anxiety Rating Scale, PANSS = the Positive and Negative Syndrome Scale, TSH = thyroid-stimulating hormone, FT3 = free triiodothyronine, FT4 = free thyroxine, TgAb = antithyroglobulin, TPOAb = thyroid peroxidases antibody, TC = total cholesterol, HDL-C = high-density lipoprotein, LDL-C = low-density lipoprotein, TG = total triglycerides, BMI = body mass index.*
The centrality plot (Figure 2) confirmed that TSH was the central node of the network, followed by TC and PANSS scores. Table S2 displays the predictability of the nodes in the network. The predictability of TSH was the highest (0.57). The predictability of clinical symptoms was 0.53 for PANSS, 0.50 for HAMA, and 0.45 for HAMD. These results indicated that half of the variance of clinical symptoms could be explained by the nodes in the network. The lowest predictability was found for FT3, FT4, and BMI.
**Figure 2:** *The centrality plot of the network. The X-rays represented the strength of each node. Nodes with higher strength have stronger impact in other nodes within the network. HAMD = Hamilton Depression Rating Scale, HAMA = Hamilton Anxiety Rating Scale, PANSS = the Positive and Negative Syndrome Scale, TSH = thyroid-stimulating hormone, FT3 = free triiodothyronine, FT4 = free thyroxine, TgAb = antithyroglobulin, TPOAb = thyroid peroxidases antibody, TC = total cholesterol, HDL-C = high-density lipoprotein, LDL-C = low-density lipoprotein, TG = total triglycerides, BMI = body mass index.*
The network had reasonable stability with a value of 0.671 for CS-C (Figure S1), indicating that after omitting $67\%$ of the raw data, the network remained highly correlated with the original network ($r = 0.7$). The bootstrap procedure also demonstrated high accuracy of the estimated edges within the network (Figure S2).
## Discussion
To our knowledge, this is the first study to explore the relationship between thyroid dysfunction, metabolic disturbances, and clinical symptoms in SA patients through a network approach. Our main findings included [1]: SA MDD patients exhibited more severe metabolic disturbances, thyroid dysfunction, and clinical symptoms compared to non-SA MDD patients; [2] the severity of anxiety and depression symptoms, SBP, TSH, and TPOAb were independently associated with SA in FEDN MDD patients; and [3] TSH played an important role in the network of thyroid dysfunction, metabolic disturbances, and clinical symptoms in SA patients. Taken together, our work highlights the predominant role of serum TSH levels in the pathophysiology of SA. In addition to being a potential biomarker for SA in MDD patients, the serum TSH level is closely associated with SA-related metabolic disturbances and clinical symptoms. Hence, regular thyroid function tests might help early detect SA. Targeting increased TSH levels may help to reduce metabolic disturbances and clinical symptoms in MDD patients with SA.
Consistent with previous studies (46–48), our study demonstrated a very high metabolic burden and thyroid dysfunction in patients with SA, which called for regular metabolic and thyroid function tests in this population. There are a few explanations for the biological changes in SA patients. First, SA patients have more severe depressive symptoms, which may lead to an unhealthy lifestyle, such as irregular sleep and diet, resulting in metabolic disturbances and thyroid dysfunction [49]. Second, inflammation may act as a bridge between SA and metabolic disorders. Emerging studies have found that inflammation plays an important role in MDD and its associated SA (50–54). Metabolic disorders were found to be associated with a chronic inflammatory state [55] and therefore may contribute to SA. Third, thyroid dysfunction was tightly associated with abnormal neurotransmitters (e.g., 5-hydroxytryptamine and norepinephrine), which played an important role in SA [56]. Fourth, the high level of TPOAb might indicate the autoimmune status of MDD patients with SA. The disturbances in the kynurenine pathway and hypothalamic-pituitary-adrenal axis in autoimmune status might contribute to the SA [57].
Network analysis suggested that thyroid dysfunction, metabolic disturbance, and clinical symptoms were highly correlated among SA patients. High TSH levels were found to be the central variables within the network, which were tightly associated with both metabolic disturbances (impaired glucose metabolism, lipid metabolism, and hypertension) and clinical symptoms (psychotic and anxiety symptoms) in SA patients. The strong association of TSH with metabolic disturbances replicates findings in the general population [17], which can be explained by the following points. First, serum TSH levels can regulate lipid metabolism in various ways [58]. High TSH levels can regulate cholesterol metabolism by binding to TSH receptors on the surface of hepatocytes [59]. It can accelerate cholesterol synthesis and reduce cholesterol clearance [58], which can lead to dyslipidemia and obesity. Second, TSH levels may also play a role in insulin resistance and glucose tolerance [60]. Studies have shown that high TSH levels are associated with the impairment of glucose transport [61].
Emerging studies suggested that thyroid dysfunction could predict several negative consequences in patients with MDD, including long-term readmission, conversion to bipolar disorder, and anxiety (62–64). However, most of these association was observed in the context of overt hypothyroidism. The relationship between SCH and clinical symptoms in patients with MDD remained controversial. Meta-analysis suggested that SCH exhibited a rather weak association with depressive symptoms [65]. One population-based study demonstrated a negative association of TSH levels with anxiety [66]. Interestingly, Liu et al. reported the same results as ours [11], finding higher serum TSH levels were associated with anxiety and psychosis among 1279 patients with MDD. The different results may be due to differences in sample characteristics (MDD patients versus community samples). Unfortunately, the relationship between SCH and clinical symptoms in patients with MDD was mostly studied in the cross-sectional study. The biological mechanism remained largely unexplored. Further studies are needed to validate our findings and to assess the possible mechanisms.
Our study has several important clinical implications. First, our study showed a high prevalence of metabolic disorders, thyroid dysfunction, anxiety, and psychotic symptoms in MDD patients with SA. Therefore, screening for these problems is crucial in this particular population. Second, our study suggested the severity of anxiety and depression, TSH level, TPOAb level, and SBP were independently associated with SA. Regular monitoring of these clinical variables might help early detect and prevent SA. Third, our study highlighted the predominant role of TSH in the pathophysiology of SA. Targeting TSH may be valuable in reducing metabolic disorders, clinical symptoms, and thyroid dysfunction associated with SA. To date, a few studies have shown that thyroid hormone therapy is effective in improving lipid metabolism in patients with SCH [67]. Some studies have also documented its effectiveness in the treatment of MDD and bipolar depression [68, 69], but the results are inconsistent [70, 71]. Therefore, more studies are in need to test our hypothesis.
Our study has several limitations. First, we used a cross-sectional study design, which prevented us from drawing causal relationships. Second, this study is monocentric and includes only the Han Chinese population. It remains unknown whether our findings can be generalized to other populations. Third, we did not collect several important sociocultural risk factors for SA, such as stressful life events and economic hardship [5]. In addition, we did not collect lifestyle factors, such as smoking and exercise, as well as diet, which are strongly associated with metabolic disturbances and thyroid dysfunction, and this should be remedied in future studies. Fourth, our study is mainly descriptive and the underlying biological mechanisms are unknown. Fifth, we assessed suicide attempts by a single item only. Application of a specific suicide rating scale may better assess various aspects of suicidality (suicidal ideation, suicide planning, and SA) and their relationship with clinical symptoms, metabolic disturbances, and thyroid dysfunction. Further longitudinal studies with a more comprehensive assessment of confounding factors and suicidality are needed to validate our findings.
In conclusion, our study demonstrates that MDD patients with SA have severe thyroid dysfunction, metabolic disturbances, and clinical symptoms. Anxiety, depression, TSH, TPOAb, and SBP were independently associated with SA in FEDN MDD patients. Targeting increased TSH in MDD patients with SA may help reduce metabolic disturbances, clinical symptoms, and thyroid dysfunction in SA patients.
## Data availability statement
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.
## Ethics statement
All participants provided written informed consent. This study was approved by the Institutional Review Board (IRB) of the First Hospital of Shanxi Medical University (No. 2016-Y27). The patients/participants provided their written informed consent to participate in this study.
## Author contributions
PP, formal analysis, writing - original draft. QW and XL, writing – review and editing. TL and X-YZ, conceptualization, writing–review and editing. All authors contributed to the article and approved the submitted version.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fendo.2023.1136806/full#supplementary-material
## References
1. Cai H, Jin Y, Liu S, Zhang Q, Zhang L, Cheung T. **Prevalence of suicidal ideation and planning in patients with major depressive disorder: A meta-analysis of observation studies**. *J Affect Disord* (2021) **293**. DOI: 10.1016/j.jad.2021.05.115
2. Dong M, Wang S-B, Li Y, Xu D-D, Ungvari GS, Ng CH. **Prevalence of suicidal behaviors in patients with major depressive disorder in China: A comprehensive meta-analysis**. *J Affect Disord* (2018) **225**. DOI: 10.1016/j.jad.2017.07.043
3. Cai H, Xie X-M, Zhang Q, Cui X, Lin J-X, Sim K. **Prevalence of suicidality in major depressive disorder: A systematic review and meta-analysis of comparative studies**. *Front Psychiatry* (2021) **12**. DOI: 10.3389/fpsyt.2021.690130
4. Lin J-Y, Huang Y, Su Y-A, Yu X, Lyu X-Z, Liu Q. **Association between perceived stressfulness of stressful life events and the suicidal risk in Chinese patients with major depressive disorder**. *Chin Med J (Engl)* (2018) **131**. DOI: 10.4103/0366-6999.229898
5. Li X, Mu F, Liu D, Zhu J, Yue S, Liu M. **Predictors of suicidal ideation, suicide attempt and suicide death among people with major depressive disorder: A systematic review and meta-analysis of cohort studies**. *J Affect Disord* (2022) **302**. DOI: 10.1016/j.jad.2022.01.103
6. Fu X-L, Li X, Ji J-M, Wu H, Chen H-L. **Blood hormones and suicidal behaviour: A systematic review and meta-analysis**. *Neurosci Biobehav Rev* (2022) **139**. DOI: 10.1016/j.neubiorev.2022.104725
7. Li H, Zhang X, Sun Q, Zou R, Li Z, Liu S. **Association between serum lipid concentrations and attempted suicide in patients with major depressive disorder: A meta-analysis**. *PloS One* (2020) **15**. DOI: 10.1371/journal.pone.0243847
8. Zhang Q, Zhao S, Liu Z, Luo B, Yang Y, Shi Y. **Association of thyroid-stimulating hormone and lipid levels with suicide attempts among adolescents with major depressive disorder in China**. *Front Psychiatry* (2022) **13**. DOI: 10.3389/fpsyt.2022.1031945
9. Wu S, Ding Y, Wu F, Xie G, Hou J, Mao P. **Serum lipid levels and suicidality: A meta-analysis of 65 epidemiological studies**. *J Psychiatry Neurosci* (2016) **41** 56-69. DOI: 10.1503/jpn.150079
10. Zhou S, Zhao K, Shi X, Sun H, Du S, Miao X. **Serum lipid levels and suicide attempts within 2 weeks in patients with major depressive disorder: Is there a relationship**. *Front Psychiatry* (2021) **12**. DOI: 10.3389/fpsyt.2021.676040
11. Liu W, Wu Z, Sun M, Zhang S, Yuan J, Zhu D. **Association between fasting blood glucose and thyroid stimulating hormones and suicidal tendency and disease severity in patients with major depressive disorder**. *Bosn J Basic Med Sci* (2022) **22**. DOI: 10.17305/bjbms.2021.6754
12. Peng R, Dai W, Li Y. **Low serum free thyroxine level is correlated with lipid profile in depressive patients with suicide attempt**. *Psychiatry Res* (2018) **266**. DOI: 10.1016/j.psychres.2018.05.059
13. Ma Y-J, Wang D-F, Yuan M, Zhang X-J, Long J, Chen S-B. **The prevalence, metabolic disturbances and clinical correlates of recent suicide attempts in Chinese inpatients with major depressive disorder**. *BMC Psychiatry* (2019) **19** 144. DOI: 10.1186/s12888-019-2131-6
14. Bou Khalil R, Richa S. **Thyroid adverse effects of psychotropic drugs: a review**. *Clin Neuropharmacol* (2011) **34**. DOI: 10.1097/WNF.0b013e31823429a7
15. Gehlawat P, Gupta R, Rajput R, Gahlan D, Gehlawat VK. **Diabetes with comorbid depression: role of SSRI in better glycemic control**. *Asian J Psychiatr* (2013) **6**. DOI: 10.1016/j.ajp.2013.03.007
16. Köhler-Forsberg O, Otte C, Gold SM, Østergaard SD. **Statins in the treatment of depression: Hype or hope**. *Pharmacol Ther* (2020) **215**. DOI: 10.1016/j.pharmthera.2020.107625
17. Liu F-H, Hwang J-S, Kuo C-F, Ko Y-S, Chen S-T, Lin J-D. **Subclinical hypothyroidism and metabolic risk factors association: A health examination-based study in northern Taiwan**. *BioMed J* (2018) **41**. DOI: 10.1016/j.bj.2018.02.002
18. Li R-C, Zhang L, Luo H, Lei Y, Zeng L, Zhu J. **Subclinical hypothyroidism and anxiety may contribute to metabolic syndrome in sichuan of China: A hospital-based population study**. *Sci Rep* (2020) **10** 2261. DOI: 10.1038/s41598-020-58973-w
19. Ding X, Zhao Y, Zhu C-Y, Wu L-P, Wang Y, Peng Z-Y. **The association between subclinical hypothyroidism and metabolic syndrome: an update meta-analysis of observational studies**. *Endocr J* (2021) **68**. DOI: 10.1507/endocrj.EJ20-0796
20. Nakajima Y, Yamada M, Akuzawa M, Ishii S, Masamura Y, Satoh T. **Subclinical hypothyroidism and indices for metabolic syndrome in Japanese women: one-year follow-up study**. *J Clin Endocrinol Metab* (2013) **98**. DOI: 10.1210/jc.2013-1353
21. Kim M-D, Yang H-J, Kang NR, Park JH, Jung Y-E. **Association between subclinical hypothyroidism and metabolic syndrome among individuals with depression**. *J Affect Disord* (2020) **264**. DOI: 10.1016/j.jad.2019.11.080
22. Liu F, Huang L, Zhu R, Peng C, Deng X. **The thyroid dysfunction of suicide attempts in major depression**. *Annales Médico-psychologiques Rev psychiatrique* (2021) **179**. DOI: 10.1016/j.amp.2020.12.022
23. McInerney AM, Lindekilde N, Nouwen A, Schmitz N, Deschênes SS. **Diabetes distress, depressive symptoms, and anxiety symptoms in people with type 2 diabetes: A network analysis approach to understanding comorbidity**. *Diabetes Care* (2022) **45** 1715-23. DOI: 10.2337/dc21-2297
24. Borsboom D, Deserno MK, Rhemtulla M, Epskamp S, Fried EI, McNally RJ. **Network analysis of multivariate data in psychological science**. *Nat Rev Methods Primers* (2021) **1** 1-18. DOI: 10.1038/s43586-021-00055-w
25. McNally RJ. **Can network analysis transform psychopathology**. *Behav Res Ther* (2016) **86** 95-104. DOI: 10.1016/j.brat.2016.06.006
26. Zainal NH, Newman MG. **Prospective network analysis of proinflammatory proteins, lipid markers, and depression components in midlife community women**. *Psychol Med* (2022) 1-12. DOI: 10.1017/S003329172200232X
27. Jia Q-F, Yang H-X, Zhuang N-N, Yin X-Y, Zhu Z-H, Yuan Y. **The role of lipoprotein profile in depression and cognitive performance: a network analysis**. *Sci Rep* (2020) **10** 20704. DOI: 10.1038/s41598-020-77782-9
28. Zimmerman M, Martinez JH, Young D, Chelminski I, Dalrymple K. **Severity classification on the Hamilton depression rating scale**. *J Affect Disord* (2013) **150**. DOI: 10.1016/j.jad.2013.04.028
29. Thompson E. **Hamilton Rating scale for anxiety (HAM-a)**. *Occup Med (Lond)* (2015) **65** 601. DOI: 10.1093/occmed/kqv054
30. Kay SR, Fiszbein A, Opler LA. **The positive and negative syndrome scale (PANSS) for schizophrenia**. *Schizophr Bull* (1987) **13**. DOI: 10.1093/schbul/13.2.261
31. Duan Y, Wei J, Geng W, Jiang J, Zhao X, Li T. **Research on cognitive function in anxious depression patients in China**. *J Affect Disord* (2021) **280**. DOI: 10.1016/j.jad.2020.11.078
32. Li X, Liu H, Hou R, Baldwin DS, Li R, Cui K. **Prevalence, clinical correlates and IQ of suicidal ideation in drug naïve Chinese han patients with major depressive disorder**. *J Affect Disord* (2019) **248** 59-64. DOI: 10.1016/j.jad.2018.12.017
33. Li Z, Xue M, Zhao L, Zhou Y, Wu X, Xie X. **Comorbid major depression in first-episode drug-naïve patients with schizophrenia: Analysis of the depression in schizophrenia in China (DISC) study**. *J Affect Disord* (2021) **294**. DOI: 10.1016/j.jad.2021.06.075
34. Peng P, Wang Q, Ren H, Zhou Y, Hao Y, Chen S. **Association between thyroid hormones and comorbid psychotic symptoms in patients with first-episode and drug-naïve major depressive disorder**. *Psychiatry Res* (2023) **320** 115052. DOI: 10.1016/j.psychres.2023.115052
35. Zhou Y, Wang Q, Ren H, Yang WFZ, Ma Y, Wu Q. **Prevalence and related factors of anxiety in first episode and drug naïve Chinese han outpatients with psychotic major depression**. *J Affect Disord* (2022) **316**. DOI: 10.1016/j.jad.2022.08.018
36. Shen Y, Chan BSM, Liu J, Meng F, Yang T, He Y. **Estimated prevalence and associated risk factors of attention deficit hyperactivity disorder (ADHD) among medical college students in a Chinese population**. *J Affect Disord* (2018) **241**. DOI: 10.1016/j.jad.2018.08.038
37. Li Y-C, Bai W, Cai H, Wu Y, Zhang L, Ding Y-H. **Suicidality in clinically stable bipolar disorder and schizophrenia patients during the COVID-19 pandemic**. *Transl Psychiatry* (2022) **12** 303. DOI: 10.1038/s41398-022-02045-2
38. He J, Lai Y, Yang J, Yao Y, Li Y, Teng W. **The relationship between thyroid function and metabolic syndrome and its components: A cross-sectional study in a Chinese population**. *Front Endocrinol (Lausanne)* (2021) **12**. DOI: 10.3389/fendo.2021.661160
39. Liu Y, Sheng C, Feng W, Sun F, Zhang J, Chen Y. **A multi-center study on glucometabolic response to bariatric surgery for different subtypes of obesity**. *Front Endocrinol (Lausanne)* (2022) **13**. DOI: 10.3389/fendo.2022.989202
40. Skjerdingstad N, Johnson MS, Johnson SU, Hoffart A, Ebrahimi OV. **Feelings of worthlessness links depressive symptoms and parental stress: A network analysis during the COVID-19 pandemic**. *Eur Psychiatry* (2021) **64** e50. DOI: 10.1192/j.eurpsy.2021.2223
41. Epskamp S, Borsboom D, Fried EI. **Estimating psychological networks and their accuracy: A tutorial paper**. *Behav Res* (2018) **50** 195-212. DOI: 10.3758/s13428-017-0862-1
42. Peng P, Liang M, Wang Q, Lu L, Wu Q, Chen Q. **Night shifts, insomnia, anxiety, and depression among Chinese nurses during the COVID-19 pandemic remission period: A network approach**. *Front Public Health* (2022) **10**. DOI: 10.3389/fpubh.2022.1040298
43. Burger J, Isvoranu A-M, Lunansky G, Haslbeck JMB, Epskamp S, Hoekstra RHA. **Reporting standards for psychological network analyses in cross-sectional data**. *psychol Methods* (2022). DOI: 10.1037/met0000471
44. Bai W, Xi H-T, Zhu Q, Ji M, Zhang H, Yang B-X. **Network analysis of anxiety and depressive symptoms among nursing students during the COVID-19 pandemic**. *J Affect Disord* (2021) **294**. DOI: 10.1016/j.jad.2021.07.072
45. Ren L, Wang Y, Wu L, Wei Z, Cui L-B, Wei X. **Network structure of depression and anxiety symptoms in Chinese female nursing students**. *BMC Psychiatry* (2021) **21** 279. DOI: 10.1186/s12888-021-03276-1
46. Zhao K, Zhou S, Shi X, Chen J, Zhang Y, Fan K. **Potential metabolic monitoring indicators of suicide attempts in first episode and drug naive young patients with major depressive disorder: a cross-sectional study**. *BMC Psychiatry* (2020) **20** 387. DOI: 10.1186/s12888-020-02791-x
47. Chang J-C, Yen AM-F, Lee C-S, Chen SL-S, Chiu SY-H, Fann JC-Y. **Metabolic syndrome and the risk of suicide: a community-based integrated screening samples cohort study**. *Psychosom Med* (2013) **75**. DOI: 10.1097/PSY.0000000000000014
48. Ko J-K, Han K-M, Shin C, Lee S-H, Han C, Kim Y-K. **Association of metabolic syndrome and its components with suicidal ideation and depression in adults: A nationally representative sample of the Korean population**. *J Affect Disord* (2019) **249**. DOI: 10.1016/j.jad.2019.02.049
49. Marazziti D, Rutigliano G, Baroni S, Landi P, Dell’Osso L. **Metabolic syndrome and major depression**. *CNS Spectr* (2014) **19** 293-304. DOI: 10.1017/S1092852913000667
50. Beurel E, Toups M, Nemeroff CB. **The bidirectional relationship of depression and inflammation: Double trouble**. *Neuron* (2020) **107**. DOI: 10.1016/j.neuron.2020.06.002
51. Dolsen MR, Prather AA, Lamers F, Penninx BWJH. **Suicidal ideation and suicide attempts: associations with sleep duration, insomnia, and inflammation**. *Psychol Med* (2021) **51** 2094-2103. DOI: 10.1017/S0033291720000860
52. Fernández-Sevillano J, González-Ortega I, MacDowell K, Zorrilla I, López MP, Courtet P. **Inflammation biomarkers in suicide attempts and their relation to abuse, global functioning and cognition**. *World J Biol Psychiatry* (2022) **23**. DOI: 10.1080/15622975.2021.1988703
53. Cáceda R, Griffin WST, Delgado PL. **A probe in the connection between inflammation, cognition and suicide**. *J Psychopharmacol* (2018) **32**. DOI: 10.1177/0269881118764022
54. Kappelmann N, Arloth J, Georgakis MK, Czamara D, Rost N, Ligthart S. **Dissecting the association between inflammation, metabolic dysregulation, and specific depressive symptoms: A genetic correlation and 2-sample mendelian randomization study**. *JAMA Psychiatry* (2021) **78**. DOI: 10.1001/jamapsychiatry.2020.3436
55. Chan KL, Cathomas F, Russo SJ. **Central and peripheral inflammation link metabolic syndrome and major depressive disorder**. *Physiol (Bethesda)* (2019) **34**. DOI: 10.1152/physiol.00047.2018
56. Toloza FJK, Mao Y, Menon L, George G, Borikar M, Thumma S. **Association of thyroid function with suicidal behavior: A systematic review and meta-analysis**. *Medicina (Kaunas)* (2021) **57**. DOI: 10.3390/medicina57070714
57. Brundin L, Bryleva EY, Thirtamara Rajamani K. **Role of inflammation in suicide: From mechanisms to treatment**. *Neuropsychopharmacology* (2017) **42**. DOI: 10.1038/npp.2016.116
58. Liu H, Peng D. **Update on dyslipidemia in hypothyroidism: The mechanism of dyslipidemia in hypothyroidism**. *Endocr Connect* (2022) **11**. DOI: 10.1530/EC-21-0002
59. Zhang W, Tian L, Han Y, Ma H, Wang L, Guo J. **Presence of thyrotropin receptor in hepatocytes: not a case of illegitimate transcription**. *J Cell Mol Med* (2009) **13** 4636. DOI: 10.1111/j.1582-4934.2008.00670.x
60. Chaker L, Ligthart S, Korevaar TIM, Hofman A, Franco OH, Peeters RP. **Thyroid function and risk of type 2 diabetes: A population-based prospective cohort study**. *BMC Med* (2016) **14** 150. DOI: 10.1186/s12916-016-0693-4
61. Maratou E, Hadjidakis DJ, Kollias A, Tsegka K, Peppa M, Alevizaki M. **Studies of insulin resistance in patients with clinical and subclinical hypothyroidism**. *Eur J Endocrinol* (2009) **160**. DOI: 10.1530/EJE-08-0797
62. Yang L, Yang X, Yang T, Wu X, Sun P, Zhu Y. **The effect of thyroid function on the risk of psychiatric readmission after hospitalization for major depressive disorder**. *Psychiatry Res* (2021) **305**. DOI: 10.1016/j.psychres.2021.114205
63. Niu Z, Wu X, Zhu Y, Yang L, Shi Y, Wang Y. **Early diagnosis of bipolar disorder coming soon: Application of an oxidative stress injury biomarker (BIOS) model**. *Neurosci Bull* (2022) **38**. DOI: 10.1007/s12264-022-00871-4
64. Zhu Y, Wu X, Zhou R, Sie O, Niu Z, Wang F. **Hypothalamic-Pituitary-End-Organ axes: Hormone function in female patients with major depressive disorder**. *Neurosci Bull* (2021) **37**. DOI: 10.1007/s12264-021-00689-6
65. Bode H, Ivens B, Bschor T, Schwarzer G, Henssler J, Baethge C. **Association of hypothyroidism and clinical depression: A systematic review and meta-analysis**. *JAMA Psychiatry* (2021) **78**. DOI: 10.1001/jamapsychiatry.2021.2506
66. Panicker V, Evans J, Bjøro T, Åsvold BO, Dayan CM, Bjerkeset O. **A paradoxical difference in relationship between anxiety, depression and thyroid function in subjects on and not on T4: findings from the HUNT study**. *Clin Endocrinol* (2009) **71**. DOI: 10.1111/j.1365-2265.2008.03521.x
67. Zhao C, Wang Y, Xiao L, Li L. **Effect of levothyroxine on older patients with subclinical hypothyroidism: A systematic review and meta-analysis**. *Front Endocrinol (Lausanne)* (2022) **13**. DOI: 10.3389/fendo.2022.913749
68. George S, Maiti R, Mishra A, Ranjan Mishra B, Jena M. **Efficacy and safety of supraphysiologic doses of levothyroxine for patients with bipolar depression in adults: A systematic review**. *J Psychopharmacol* (2022) **36**. DOI: 10.1177/02698811221078758
69. Bauer M, Whybrow PC. **Role of thyroid hormone therapy in depressive disorders**. *J Endocrinol Invest* (2021) **44**. DOI: 10.1007/s40618-021-01600-w
70. Zhao T, Chen BM, Zhao XM, Shan ZY. **Subclinical hypothyroidism and depression: A meta-analysis**. *Transl Psychiatry* (2018) **8** 239. DOI: 10.1038/s41398-018-0283-7
71. Wildisen L, Feller M, Del Giovane C, Moutzouri E, Du Puy RS, Mooijaart SP. **Effect of levothyroxine therapy on the development of depressive symptoms in older adults with subclinical hypothyroidism: An ancillary study of a randomized clinical trial**. *JAMA Netw Open* (2021) **4**. DOI: 10.1001/jamanetworkopen.2020.36645
|
---
title: Seasonal differences in intestinal flora are related to rats’ intestinal water
metabolism
authors:
- Jing Li
- Yike Sun
- Ruochong Wang
- Shuran Ma
- Lei Shi
- Kai Wang
- Hairong Zhang
- Tong Wang
- Leilei Liu
journal: Frontiers in Microbiology
year: 2023
pmcid: PMC9999011
doi: 10.3389/fmicb.2023.1109696
license: CC BY 4.0
---
# Seasonal differences in intestinal flora are related to rats’ intestinal water metabolism
## Abstract
Many studies have reported obvious seasonal differences in the intestinal flora of rats, and this stable distribution of the seasonal flora helps in maintaining the normal physiological function of the host. However, the mechanism underlying these seasonal differences in intestinal flora remains unclear. To explore the correlation among seasonal factors and intestinal water metabolism and intestinal flora, 20 Sprague Dawley (SD) rats were divided into spring, summer, autumn, and winter groups. The environment for the four seasons was simulated using the Balanced Temperature and Humidity Control system. The intestinal water metabolism was evaluated by determining the intestinal transmission function, fecal water content, water content of colonic tissue, and the colonic expression levels of AQP3, AQP4, and AQP8. The composition and relative abundance of intestinal microflora in rats in each season were assessed through 16S rDNA amplifier sequencing, and the relationship between the dominant flora and intestinal water metabolism in each season was analyzed using Spearman correlation analysis. The high temperature and humidity season could lead to an increase in intestinal water metabolism and intestinal water content in rats, whereas the low temperature and humidity season could lead to a decrease, which was closely related to the change in microflora. To explore the molecular mechanism of seasonal changes in intestinal water metabolism, the concentration of colonic 5-HT, VIP, cAMP, and PKA associated with intestinal water metabolism in rats were also examined. Seasonal changes could affect the concentration of colonic 5-HT and VIP in rats, and then regulate AQPs through cAMP/PKA pathway to affect the intestinal water metabolism. These results suggest that seasonal factors affect the level of intestinal water metabolism in rats and result in seasonal differences in intestinal flora.
## Introduction
The gut has hundreds of millions of flora, and they are closely linked to the host’s metabolism. Research on the relationship between intestinal flora and body health and disease has recently received considerable attention by the scientific community, and the composition of intestinal flora is closely related to the host environment has been confirmed (Ver Heul et al., 2019; Rutsch et al., 2020). Most current studies have focused on the effects of the gut flora on the host’s physiological function and the role they play in disease formation and the development. For example, physiologically, intestinal flora metabolites can act on the body’s emotional center through the intestine–cerebral axis, thereby affecting the host’s emotional activity (Jenkins et al., 2016); the intestinal flora can also affect the function of intestinal macrophages through n-butyrate secretion and participate in intestinal immunity (Chang et al., 2014). Pathologically, intestinal flora disorder is closely related to the formation and development of various intestinal diseases, such as irritable bowel syndrome (IBS; Wang et al., 2020), inflammatory bowel disease (Du et al., 2022), and colorectal cancer (Zou et al., 2018). Abnormal intestinal flora can also cause many parenteral diseases, such as asthma (Salameh et al., 2020), diabetes (Du et al., 2022), and nonalcoholic fatty liver (Di Ciaula et al., 2022). Therefore, researchers are keenly interested in understanding the reasons for the change in the intestinal flora composition.
The alimentary canal is directly connected to the external world, and external environmental factors, such as circadian rhythm (Liang et al., 2015), seasonal rhythm (Liu et al., 2018), air pollution (Ran et al., 2021), and diet (De Angelis et al., 2019), can easily affect the intestinal flora, thereby leading to changes in the host’s physiological functions. Of these factors, the effect of seasonal changes on intestinal flora is among the current research hotspots (Segawa et al., 2005; Zhu et al., 2020), through changes in composition and diversity, microbiota can adapt to seasonal changes in the environment and maintain host health (Davenport et al., 2014; You et al., 2022). However, how seasonal factors act on the flora and lead to changes in its composition remains unclear. After literature review, we found that the aqueous environment, as among the necessary conditions for the survival of flora, is closely related to the changes in the composition of flora (Or et al., 2007; Dechesne et al., 2008; Young et al., 2008; Dechesne et al., 2010; Vos et al., 2013). On the other hand, both the aqueous environment and the flora composition vary seasonally (Lv et al., 2021). Therefore, the present study mainly attempts to understand the relationship between host’s intestinal water metabolism and seasonal differences in intestinal flora.
Temperature and humidity are common external environmental factors affecting body’s water metabolism. For example, high temperatures (30°C ± 1°C) increase moisture loss caused by respiration of insects and epidermal transpiration (Kleynhans et al., 2014), and high humidity ($90\%$ ± $2\%$) can increase blood urea nitrogen and antidiuretic hormone secretion, leading to water metabolism disorders (Yin et al., 2022). Clinical studies have also reported that diarrheal diseases are frequently due to relatively high temperatures and humidity in summer and autumn (Phung et al., 2015; Anwar et al., 2019), and these diseases are often accompanied by intestinal flora disorders (Li et al., 2021). Thus, environmental factors, especially temperature and humidity, are reasonably speculated to affect intestinal water metabolism, thereby causing changes in the intestinal flora composition. Based on the environmental characteristics of the four seasons in Beijing, this study strictly controlled the environmental variables and mainly explored the intrinsic connection between seasonal changes in rats’ intestinal flora and intestinal water metabolism ability from the perspective of temperature and humidity.
## Animal and experimental grouping
Twenty male specific-pathogen free (SPF)-grade Sprague–Dawley (SD) rats (age: 6 weeks; weight: 200 ± 20 g) were purchased from SPF (Beijing) Biotechnology Co., Ltd. Number of animal license: SYXK (Beijing) 2019–0010. This study was approved by the Ethics Committee of Beijing University of Chinese Medicine (approval number: BUCM-4-2,021,032,603-1,059).
All rats were adaptability housed for 7 days and randomly divided into four groups ($$n = 5$$ rats each) according to the method of the random number table. All rats were fed sterilized feed and deionized water in an artificial climate simulator (NHRHG6, Chongqing Hongrui Experimental Instrument Co., Ltd).
The artificial climate simulator uses a Balanced Temperature and Humidity Control (BTHC) system to control temperature (T) and relative humidity (RH). One can also design the corresponding parameters according to the actual environmental and climatic characteristics of the area and simulate the local climate. By strictly controlling environmental variable factors, the interference of other factors in experimental results can be reduced. Referring to the data1 displayed by the Spanish National Meteorological Institute (Agencia Estatal de Meteorología, AEMET), select the China-Beijing area, we established the environment temperature for the four seasons based on the information about the average monthly maximum temperature for 2020. The temperatures of the four seasons were 14.8°C in spring, 26.2°C in summer, 12.9°C in autumn, and − 2.3°C in winter. Then, the relative humidity was established with reference to the average humidity of the natural season in 2020. Accordingly, the humidity of the four seasons was $40.3\%$ in spring, $64.3\%$ in summer, $52.2\%$ in autumn, and $41.2\%$ in winter.
## 16S rDNA amplicon sequencing of intestinal flora
The experiment ended after 4 groups of rats were fed in the artificial climate simulator for 1 month. After the experiment ended, 3 rats were randomly selected from each group and their perianal skin was disinfected with $75\%$ alcohol. The feces of each rat was collected using sterilized forceps and placed into a sterile 2-mL Eppendorf tube. The collected feces were immediately stored in an ultra-low temperature refrigerator at −80°C.
The microbial community genomic total DNA was extracted by E.Z.N.A. ® Stool DNA Kit (Omega Bio-Tek, United States). PCR amplification was performed using 16S V4 region primers (515F and 806R) according to the requirement of Phusion® High-Fidelity PCR Master Mix amplification kit (New England Biolabs). The purity of the PCR amplified sample was tested by agarose gel electrophoresis. After passing the test, the instructions of the TruSeq® DNA PCR-Free Sample Preparation Kit (New England Biolabs) were followed to build a sequencing library. After building and testing the library, the NovaSeq6000 sequencing platform was used by Beijing Novogene Technology Co., Ltd., to perform the sequencing of the flora. After sorting out the original data, effective tags were clustered, and the sequences were clustered into operational taxonomic units (OTUs) with $97\%$ consistency. Then, using the representative sequences of OTUs, species annotation (Venn Graph, species relative abundance column accumulation map, species abundance cluster heat map), alpha diversity analysis (Shannon index, Chao1 index), beta diversity analysis (principal co-ordinate analysis, PCoA), and between-group difference species analysis (LDA effect size analysis) were performed.
## Fecal water content test
After the experiment, each rat was placed in a metabolic cage, and their feces were collected for 24 h. The feces morphology was assessed according to the Bristol Stool Form Scale (BSFS; Table 1).The feces weight was recorded as the fecal wet weight. Then, all the fecal samples were dried with a filter paper, placed in an electric blast drying oven (GZX-9023MBE, Shanghai Boxun Industrial Co., Ltd), and dried at 60°C. Then, the fecal dry weight was recorded. The fecal water content was calculated as follows: the fecal water content = (fecal wet weight – fecal dry weight)/fecal wet weight × $100\%$.
**Table 1**
| Type | Form |
| --- | --- |
| I | Separate hard lumps, like nuts |
| II | Sausage-shaped but lumpy |
| III | Like a sausage or snake but with cracks on its surface |
| IV | Like a sausage or snake, smooth and soft |
| V | Soft blobs with clear-cut edges |
| VI | Fluffy pieces with ragged edges, a mushy stool |
| VII | Watery, no solid pieces |
## Detection of intestinal transmission function
Distilled water (800 ml) was added to 100 g gum arabic (A108975-500G, Beijing Meikang Instrument Equipment Co., Ltd). Boiled the solution until transparent. Then, 50 g of activated carbon (C139601, Beijing Meikang Instrument Equipment Co., Ltd) was added to the solution and boiled 3 times. After the solution was cooled, distilled water was subsequently added to make the final volume of the solution to 1,000 ml. After each rat was placed in a metabolic cage, they received a gavage of 2 ml of the 100 g/l activated carbon suspension. The time from the completion of the activated carbon gavage to the excretion of the rat’s first black stool was recorded as the fecal excretion time.
## Water content of colonic tissue test
All the rats were sacrificed through cervical dislocation. Then, 2 cm of the colonic tissue above the anus of each rat was cut off with sterilized surgical scissors and weighed after being dried with a filter paper. This weight was recorded as the wet weight of colonic tissue. The collected tissue was then dried in the electric blast drying oven at 60°C and weighed. This weight was recorded as the dry weight, and the water content of the colonic tissue was calculated as follows: water content of colonic tissue = (wet weight of colonic tissue − dry weight of colonic tissue)/wet weight of colonic tissue × $100\%$.
## Colonic AQP, 5-HT, VIP, cAMP, and PKA expression level detection
The colonic tissue (2 cm) of the rats was frozen in liquid nitrogen and stored at −80°C for later use. The tissue was then made into homogenates, and the supernatant was used in accordance with the instructions of Aquaporin 3 (MB-2017A, Jiangsu Enzyme Biotechnology Co., Ltd), Aquaporin 4 (MB-2016A, Jiangsu Enzyme Biotechnology Co., Ltd), Aquaporin 8 (MB-7036A, Jiangsu Enzyme Biotechnology Co., Ltd), 5-Hydroxytryptamine (UK Abcam), Vasoactive intestinal polypeptide (American Raybio), Cathelicidin Antimicrobial Peptide (Shanghai Bluegene Biotechnology Co., Ltd) and Protein kinase A (Shanghai Bluegene Biotechnology Co., Ltd) ELISA kits. Later, the absorbance (optical density (OD) value) of each well was measured sequentially at 450 nm. The standard curves were drawn according to the standard concentration and OD value given in the instructions, the formula was established, and the concentration of each sample was calculated using the formula.
## Correlation analysis
The correlation between the relevant indicators of intestinal water metabolism (fecal excretion time, dry and wet weights of colonic tissue, water content of colonic tissue, fecal dry and wet weights, fecal water content, AQP3, AQP4, and AQP8) and the abundance of the dominant intestinal flora in each season was studied through Spearman correlation analysis, and the relevance heatmap was drawn according to the obtained results.
## Data analysis
SPSS20.0 software (IBM, United States) was used for statistically analyzing the data. All the data were expressed as mean ± SD. The Shapiro–Wilk test was first used for determining normality. When the data of each group followed a normal distribution and the variances were equal, one-way ANOVA was performed to compare multiple groups, and the two–two comparisons among the groups were performed using the LSD test at the same time. However, when the data of each group followed a normal distribution but the variance was unequal, the *Welch analysis* was performed to compare multiple groups, and the two–two comparisons among the groups were performed using the Dunnett T3 test at the same time. When the data of the groups did not follow a normal distribution, the Kruskal–Wallis test was performed to compare multiple groups, and the two–two comparisons among the groups were performed using the Mann–Whitney test at the same time. α = 0.05 was considered the test level, and $p \leq 0.05$ indicated that the difference was statistically significant, and $p \leq 0.01$ indicated that the difference was extremely significant.
## Effects of different seasons on intestinal water metabolism in rats
The feces of all groups were type IV in BSFS, the length of the feces was uniform and the color was brownish-black, and no mushy or loose stools.
## Effects of different seasons on fecal water content and water content of colonic tissue in rats
Figure 1 presents the fecal excretion time. The fecal excretion time of the winter group increased significantly compared with the summer group ($p \leq 0.05$). Although the fecal excretion time in the summer group was not statistically different from that in the spring and autumn groups, an obvious decrease trend was observed in the summer group.
**Figure 1:** *Fecal excretion time. #P < 0.05 compared with the summer group.*
Figure 2 presents the dry and wet weights and water content of colonic tissue in different seasons. Although no statistical difference was observed in the dry weight of colonic tissue, an increase trend was observed in the spring group compared with the other groups, and a decrease trend was observed in the summer group compared with other groups. The wet weight of colonic tissue in the summer group was the highest in four seasons, and was significantly different from that in the spring group ($p \leq 0.05$). Compared with the autumn group and the winter group, the wet weight of colonic tissue in the spring group showed a decrease trend. The water content of the colonic tissue in the summer group was the highest in four seasons, and was significantly different from that in the spring group ($p \leq 0.05$).
**Figure 2:** *(A) Dry weight of colonic tissue. (B) Wet weight of colonic tissue. (C) Water content of colonic tissue. *p < 0.05 compared with the spring group.*
Figure 3 illustrates the dry and wet weights and water content of feces in different seasons. The fecal dry weight was the lowest in the summer group and was significantly different from that in the spring group ($p \leq 0.05$). The fecal wet weight was the highest in the summer group and the lowest in the winter group. The fecal wet weight in the winter group was significantly different from that in the summer and autumn groups ($p \leq 0.05$, $p \leq 0.05$, respectively). The fecal wet weight in the autumn group increased and was significantly different from that in the spring group ($p \leq 0.05$). The fecal water content in the summer group was the highest and was significantly different from that in the spring, autumn, and winter groups ($p \leq 0.01$, $p \leq 0.05$, $p \leq 0.01$, respectively). The fecal water content in the autumn group was the second highest and was statistically different from that in the spring and winter groups ($p \leq 0.05$, $p \leq 0.05$, respectively).
**Figure 3:** *(A) Fecal dry weight. (B) Fecal wet weight. (C) Fecal water content. *p < 0.05 compared with the spring group; #p < 0.05 compared with the summer group; ##p < 0.01 compared with the summer group; ∧p < 0.05 compared with the autumn group.*
## Effects of different seasons on AQP3, AQP4, and AQP8 in rat colons
The contents of AQP3, AQP4, and AQP8 in rat colons in different seasons are presented shown in Figure 4. The content of colonic AQP3 was the highest in the summer group and the lowest in the spring group, and the difference between these two groups was statistically significant ($p \leq 0.05$). The content of colonic AQP4 was the lowest in the spring group and the highest in the winter group. The spring group had lower AQP4 content than the autumn and winter groups ($p \leq 0.05$, $p \leq 0.05$, respectively). The spring group had the lowest AQP8 content. The AQP8 content in the spring group was significantly different from that in the summer, autumn, and winter groups ($p \leq 0.01$, $p \leq 0.01$, $p \leq 0.05$, respectively). The autumn group had the highest AQP8 content, and the AQP8 content in the autumn group was higher than that in the summer and winter groups ($p \leq 0.05$, $p \leq 0.05$, respectively).
**Figure 4:** *(A) AQP3 content of colon. (B) AQP4 content of colon. (C) AQP8 content of colon. *p < 0.05 compared with the spring group; #p < 0.05 compared with the summer group; **p < 0.01 compared with the summer group; ∧p < 0.05 compared with the autumn group.*
## Effects of different seasons on 5-HT, VIP, cAMP and PKA in rat colons
The contents of 5-HT, VIP, cAMP and PKA in rat colons in different seasons are presented shown in Figure 5. The content of colonic 5-HT was the highest in the summer group. The 5-HT content in the summer group was significantly different from that in the spring, autumn, and winter groups ($p \leq 0.01$, $p \leq 0.05$, $p \leq 0.05$, respectively). A decrease trend was observed in the 5-HT content in the spring group compared with the autumn, and winter groups. The content of colonic VIP was the highest in the summer group and the lowest in the spring group, and the difference between these two groups was statistically significant ($p \leq 0.01$). The summer group had the highest cAMP content. The cAMP content in the summer group was significantly different from that in the spring, autumn, and winter groups ($p \leq 0.01$, $p \leq 0.01$, $p \leq 0.01$, respectively). A decrease trend was observed in the cAMP content in the spring group compared with the autumn, and winter groups. The content of colonic PKA was the highest in the summer group and the lowest in the spring group. The PKA content in the spring group was significantly different from that in the summer, autumn, and winter groups ($P \leq 0.05$, $P \leq 0.05$, $P \leq 0.01$, respectively).
**Figure 5:** *(A) 5-HT content of colon. (B) VIP content of colon. (C) cAMP content of colon. (D) PKA content of colon. *p < 0.05 compared with the spring group; **P<0.01 compared with the spring group; #P<0.05 compared with the summer group; ##P<0.01 compared with the summer group.*
## Effect of seasonal humidity and temperature difference on intestinal flora in rats
We obtained 665, 710, 613, and 735 OTUs in the spring, summer, autumn, and winter groups, respectively. The Venn diagram showed that all four groups shared 578 OTUs. Among them, the summer group and the winter group shared 774 OTUs, the spring group and the winter group shared 760 OTUs, the spring group and the autumn group shared 720 OTUs, the autumn group and the winter group shared 714 OTUs, the spring group and the summer group shared 713 OTUs, the summer group and the autumn group shared 700 OTUs. It was shown that in terms of intestinal flora structure, the summer group and winter group were closer (Figure 6A). The Chao1 and Shannon indices were used to calculate community richness and community diversity of the intestinal flora, respectively. The alpha diversity analysis revealed that the winter group had the highest Chao1 index, followed by the summer and spring groups, whereas the autumn group had the lowest Chao1 index (Figure 6C). The analysis also revealed that the summer group had the highest Shannon index, followed by the spring, winter, and autumn groups (Figure 6D). However, no statistically significant difference was observed in these changes. The PCoA based on OTU abundance revealed that the sample distance in each group was relatively close, which indicated that the species composition structure of each sample in each group was similar. In terms of differences between groups, the distance between the spring and autumn groups was lower than that between the other groups, indicating that the composition of intestinal flora in the spring and autumn was more similar. By contrast, the distance between the summer and the other groups was relatively greater, suggesting that the rat’s intestinal flora in summer was more distinct (Figure 6B).
**Figure 6:** *Different seasons affect the structure and composition of the intestinal flora. (A) The Venn diagram of spring, summer, autumn, and winter groups. (B) PCoA score. (C) Chao index. (D) Shannon index. (E) Columnar accumulation plot of relative species abundance at the phylum level: top 6 species according to the abundance were selected for each group. (F) Heatmap of gut microbiota at the genus level: top 25 genera according to the abundance were selected for each group (the abundance of each group was the average of all samples in the group). (G) The LEfSe analysis.*
Regarding the phyla, Firmicutes and Bacteroidetes were the two most dominant phyla among all samples. Firmicutes accounted for more than $75\%$ of the flora and was the main flora. Compared with the other three seasons, the proportion of Firmicutes was relatively low in summer ($77.25\%$; Figure 6E). Table 2 presents the detailed data of the abundance of the top 6 species in each season at the phylum level. The species abundance cluster heatmap (Figure 6F) and Table 3 show the microbes with a higher relative abundance. The microbes with a relatively high abundance in spring were Lactobacillus and Blautia. The microbes with a relatively high abundance in autumn were Lactobacillus and Faecalibaculum. The microbes with a relatively high abundance in winter and summer were Lactobacillus and Faecalibaculum. The subsequent LEfSe analysis revealed that the biomarkers in the spring group were Lactobacillus, Blautia, etc. The biomarkers in the summer group were Rothia, Helicobacter, etc. The biomarkers in the autumn group were Faecalibaculum, Methanobrevibacter, etc. The biomarkers in the winter group were Bacteria, Romboutsia, Turicibacter, and Alistipes (Figure 6G).
## Correlation analysis of intestinal water metabolism and composition of intestinal flora
To understand whether seasonal differences in the intestinal flora are related to intestinal water metabolism, we investigated the correlation between fecal excretion time, fecal dry and wet weights, fecal water content, dry and wet weights of colonic tissue, water content of colonic tissue, content of AQP3, AQP4, and AQP8, and dominant intestinal bacteria in the intestinal environment during the four seasons. This correlation was determined through Spearman correlation analysis. According to the results, the abundance of the top 25 flora in rat feces was correlated with changes in intestinal water metabolism (Figure 7A).
**Figure 7:** *(A) Correlation analysis between the abundance of the top 25 microbes at the genus level and the related indexes of intestinal water metabolism. (B) Correlation analysis between the dominant microbes in different seasons and the related indexes of intestinal water metabolism. The ordinate is intestinal water metabolism and information on AQPs (CD, dry weight of colonic tissue; CW, wet weight of colonic tissue; CM, water content of colonic tissue; FD, fecal dry weight; FW, fecal wet weight; FC, fecal water content; FT, fecal excretion time). The abscissa is the genus level-related species information obtained through clustering. The Spearman correlation coefficient r(−1 ≤ r ≤ 1) is expressed in a color shade. The darker red heatmap color indicates that r is closer to 1. When r < 0, the environmental factor is considered negatively correlated with the flora, and when r > 0, a positive correlation is considered, *p < 0.05, **p < 0.01, ***p < 0.001.*
As shown in Figure 7, the dominant microbes Lactobacillus and Blautia in the spring group exhibited a certain correlation with intestinal water metabolism. The abundance of Lactobacillus was negatively correlated with the AQP3 content (r < 0, $p \leq 0.05$) and the AQP4 content (r < 0, $p \leq 0.05$); it was positively correlated with fecal dry weight (r > 0, $p \leq 0.05$). The abundance of Lactobacillus had a negative correlation trend with both fecal water content and water content of colonic tissue. The abundance of Blautia was significantly and negatively correlated with the AQP4 content (r < 0, $p \leq 0.01$), and the AQP8 content (r < 0, $p \leq 0.05$). The abundance of Blautia was positively correlated with fecal dry weight (r > 0, $p \leq 0.01$) and the water content of colonic tissue (r > 0, $p \leq 0.05$). The abundance of Blautia exhibited a significant negative correlation with fecal water content (r < 0, $p \leq 0.01$) and a significant positive correlation with fecal excretion time (r > 0, $p \leq 0.01$).
The dominant microbes Romboutsia and *Helicobacter in* the summer group were related to intestinal water metabolism. The abundance of Romboutsia was positively correlated with the AQP3 content (r > 0, $p \leq 0.05$) and negatively correlated with fecal dry weight (r < 0, $p \leq 0.05$). The abundance of Romboutsia also had a positive correlation trend with both fecal water content and water content of colonic tissue. The abundance of *Helicobacter was* significantly positively correlated with the AQP3 content (r > 0, $p \leq 0.01$), fecal wet weight, wet weight of colonic tissue (r > 0, $p \leq 0.05$), and fecal water content (r > 0, $p \leq 0.01$). The abundance of *Helicobacter was* negatively correlated with fecal excretion time (r < 0, $p \leq 0.05$).
The dominant microbes Faecalibaculum and Methanobrevibacter in the autumn group were related to intestinal water metabolism. The abundance of Faecalibaculum was positively correlated with the AQP3 (r < 0, $p \leq 0.05$) and AQP8 contents (r > 0, $p \leq 0.01$). The abundance of Faecalibaculum had a negative correlation trend with the water content of colonic tissue. The abundance of Methanobrevibacter was positively correlated with the AQP8 content (r > 0, $p \leq 0.05$), and it had a negative correlation trend with the water content of colonic tissue.
The dominant microbes Alistipes and Turicibacter in the winter group were associated with intestinal water metabolism. The abundance of Alistipes was negatively correlated with fecal wet weight (r < 0, $p \leq 0.05$) and fecal water content. The AQP4 content exhibited a positive correlation trend with the abundance of Alistipes. The abundance of Turicibacter was negatively correlated with fecal wet weight (r < 0, $p \leq 0.05$), and it had a positive correlation trend with AQP4 content and had a negative correlation trend with fecal water content (Figure 7B).
## Discussion
The intestinal barrier is composed of four barriers, namely the mechanical, chemical, immune, and microbial barriers, that are closely related and influence each other. The intestinal flora has a crucial role as the microbial barrier, while intestinal water metabolism is related to epithelial cells in the intestinal mechanical barrier. Colonic epithelial cells participate in water and fluid metabolism by regulating luminal secretion and fluid and ion absorption (Negussie et al., 2022). Therefore, changes in the water content of colonic tissue and fecal water content directly reflect changes in intestinal water metabolism. A study on the water balance within the Greek population revealed significant differences in water loss among different seasons under the same water balance distribution (Malisova et al., 2013). Our study showed that compared with other seasons, the water content of colonic tissue and fecal water content of rats in the summer group were higher, suggesting that intestinal water metabolism was enhanced. By contrast, the water content of colonic tissue and fecal water content of rats were lower in the spring group, which suggested that the weakening of intestinal water metabolism. This confirms that a seasonal difference exists in the level of intestinal water metabolism in rats, which may be related to the seasonal change in the intestinal water reabsorption capacity.
According to a report, intestinal water transport is related to aquaporin expression on the surface of epithelial cells (Zhang et al., 2019). AQPs is a series of highly selective transmembrane channels, in which AQP3, AQP4, and AQP8 are mainly expressed in the colonic epithelium (Mobasheri et al., 2005; He and Yang, 2019). They can regulate colonic water transport and directly affect the change in intestinal water content and the level of intestinal water metabolism. Therefore, AQP3、AQP4 and AQP8 have always been the core index to investigate the mechanism of intestinal water metabolism. The increased AQP3 expression level allows the transport of a large amount of water to the lumen side (Ikarashi et al., 2016) and increases the water content in the intestinal cavity. The increased AQP4 and AQP8 expression reduces the intestinal fluid content by reducing the fecal water content, mucus secretion, and intestinal peristalsis (Hu et al., 2019). For example, AQP3 protein expression in the colon of rotavirus-infected diarrhea mice is significantly increased, whereas AQP4 and AQP8 protein expression is downregulated (Cao et al., 2014). The decreased AQP8 expression level in the mouse colon can alleviate the symptoms of constipation (Wang L. et al., 2022). Our results also confirmed that the differential expression of AQP3, AQP4, and AQP8 in the colon is often consistent with the change in the colonic water content or fecal water content. For example, AQP3 expression in the colon of summer group rats increased, whereas AQP4 and AQP8 expression decreased, which was consistent with the increased the fecal water content and intestinal water content. However, in the spring group, the expression of AQP3, AQP4, and AQP8 decreased and the level of intestinal water metabolism decreased, which was also consistent with the decreased colonic water content in rats. Thus, the seasonal difference in water metabolism in the intestinal tract is related to the seasonal change in the AQP expression level.
Cerebroenteric peptide is an important neurotransmitter or peptide hormone transmitted between the central nervous system and the enteric nervous system. As an important factor in regulating gastrointestinal fluid balance, its role in the development of intestinal diseases has been increasingly valued. Among them, vasoactive intestinal peptide (VIP) is an effective stimulant that secretes water and electrolytes through the intestinal mucosa. Increased serum VIP can cause diarrhea (Xiong et al., 2018). 5-hydroxytryptamine (5-HT) plays an important role in regulating gastrointestinal motility and secretion. Increased secretion of 5-HT is strongly associated with the development of IBS-D (Gao et al., 2022). Studies have shown that 5-HT and VIP have a regulatory effect on AQPs, which may be achieved by cAMP/PKA signaling mechanisms (Kon et al., 2015; Tan et al., 2021). 5-HT activates adenylate cyclase (AC) by stimulating the protein Gs, which can increase cyclic adenosine monophosphate (cAMP), and then activate protein kinase A (PKA)(Galligan, 2021). The upregulated cAMP / PKA signaling contributes to phosphorylation and expression of AQP3, AQP4, AQP8 (Soria et al., 2009; Chen et al., 2014; Zhu, 2015). VIP can regulate the expression of AQP3, AQP4, AQP8 through the cAMP-PKA signaling pathway (Soria et al., 2009; Chen et al., 2014; Tan et al., 2021), thereby regulating the permeability of cell membranes to water. Our results also confirmed this view by showing the decreased VIP, 5-HT, cAMP, PKA, AQP3, AQP4 and AQP8 in the spring group.
In addition, 5-HT, VIP, cAMP, PKA have seasonal variation (Pfister and Storey, 2006; Mishra et al., 2018; Ciani et al., 2019; Li et al., 2020). For example, the concentration of 5-HT in the hippocampus of rats increases in summer and decreases in winter (Li et al., 2020). Therefore, the seasonal difference in intestinal water metabolism may be caused by seasonal changes of 5-HT and VIP, which regulate cAMP/PKA and affect the expression of AQPs. Our results also confirmed that the increased intestinal water volume in rats in summer may be caused by the increased 5-HT and VIP, and the upregulation of AQP3, AQP4 and AQP8 after cAMP/PKA activation. The decreased intestinal water volume in rats in spring may be caused by the decreased 5-HT and VIP, which inhibited cAMP/PKA, and then downregulated the expression of AQP3, AQP4 and AQP8. Combined with the results of this experiment, we believe that seasonal changes in intestinal water metabolism may be closely related to the changes in the regulation of AQPs by 5-HT and VIP through the cAMP-PKA signaling pathway.
On the other hand, the change in the water–liquid environment can affect microflora composition and abundance. Related studies have reported that the content and availability of water in the environment not only directly determines the lifespan of microorganisms (Or et al., 2007; Young et al., 2008) but also affects the nutrient diffusion and distribution range of microorganisms (Dechesne et al., 2008, 2010; Vos et al., 2013). For example, compared with the wet soil, the microbial diversity and abundance of Enterobacterales, Clostridiales, Lactobacillales, and Bacteroidales decreased significantly when the soil was dry, suggesting that the change in the water–liquid environment affects microflora composition and abundance (Šťovíček et al., 2017). Similarly, other studies have confirmed that the changes in intestinal water metabolism are consistent with the changes in microflora. For example, in the rat model of diarrhea-predominant IBS, AQP8 expression decreased, whereas intestinal flora imbalance appeared; in this case, the abundance of Lactobacillus significantly decreased, whereas that of Clostridiales_bacterium increased (Zhao et al., 2022).
Through the detection of rat fecal microbiota, differences in the abundance and composition of rat intestinal microflora were observed in different seasonal environments. The change in the season is mainly reflected through changes in temperature and humidity. Temperature and humidity have a close impact on intestinal microbiota (Sun et al., 2017; Deng et al., 2020). Therefore, we here mainly controlled temperature and humidity to simulate the environment of the four seasons. At the phylum level, the intestinal microflora of rats is similar to that of other mammals (including humans), including Firmicutes and Bacteroidetes (Ley et al., 2008). Our study revealed that no significant difference was observed in the flora structure of rats in different seasons at the gate level, and the two dominant bacteria were Firmicutes and Bacteroides, which may be due to the normal physiological state of rats. Under physiological conditions, some seasonal differences were observed in α and β diversities of the rat intestinal flora, although no statistically significant difference was noted. However, according to the PCoA results, the flora structure of the spring and autumn groups was similar. The proportion of Firmicutes and Bacteroidetes in the spring and autumn groups was similar. This may be due to similar temperature and humidity in spring and autumn. However, at the genus level, the dominant flora in different seasons exhibited significant differences, and numerous studies have confirmed this finding. For example, in a 2-year study of intestinal microflora of wild mice, a strong seasonal change was observed in the structure of mouse intestinal microflora. The abundance of Lactobacillus increased significantly in late spring/early summer and decreased significantly in late summer/early autumn. By contrast, the abundance of *Helicobacter and* Alistipes showed seasonal characteristics opposite to those of Lactobacillus, which increased significantly in late summer/early autumn and decreased significantly in late spring/early summer. However, the study attributed seasonal differences in bacteria to seasonal differences in the diet type (Maurice et al., 2015). In this study, we used laboratory rats as research objects, and diet types in different seasons were the same. However, our analysis of the genera level of intestinal flora in each group revealed that significant seasonal differences were still observed. For example, the abundance of Lactobacillus increased significantly in spring, while that of Romboutsia increased significantly in winter, indicating that the effect of season on intestinal flora is related to not only diet but also other factors. For example, some studies have shown that the amount of water and food intake of adult African giant rats is significantly higher in cold and dry seasons than in hot seasons (Dzenda et al., 2013), suggesting that different seasonal conditions affect water and food intake in animals (Ebling and Barrett, 2008; Ebling, 2015; Ahlberg et al., 2018). The difference in drinking water or food intake directly affects the digestion and metabolism of the host, including water and fluid metabolism in the intestinal tract. Therefore, this study directly focused on the difference in intestinal water metabolism with change in seasons, and analyzed the relationship between intestinal water metabolism and intestinal flora.
To better study the relationship between intestinal water metabolism and intestinal flora, on the basis of some related indicators of intestinal water metabolism, we analyzed the correlation between their content and the abundance of the top 25 bacteria at the genus level. To further determine the relationship between the seasonal difference in intestinal water metabolism and that in intestinal flora, we analyzed the correlation between the content of relevant indices of intestinal water metabolism and the dominant flora in different seasons. According to the results, the abundance of Lactobacillus, Blautia, Faecalibaculum, Methanobrevibacter, Turicibacter, Alistipes, and other bacteria in spring, autumn, and winter was negatively correlated with intestinal water content. The abundance of Romboutsia, Helicobacter, and other bacteria in summer was positively correlated with intestinal water content.
The therapeutic effect of Lactobacillus, a probiotic, in improving diarrhea has been widely reported (Li et al., 2021). The intestinal water content is higher and the abundance of *Lactobacillus is* significantly lower in IBS patients with diarrhea than in the healthy control group (Wang et al., 2020). Blautia, an anaerobic bacterium, widely exists in the feces and intestines of mammals, and 37°C is most suitable for its survival (Liu et al., 2021). The relative abundance of Blautia significantly increased with a decrease in intestinal moisture in patients with constipation (Botelho et al., 2020). By contrast, in patients with irritable bowel disease, the abundance of Blautia significantly decreased with the occurrence of mucous pus and bloody stool (Liu et al., 2021). Being a common bacterium in the gastrointestinal tract, *Faecalibaculum is* closely related to the occurrence of intestinal metabolic diseases. The relative abundance of Faecalibaculum was significantly decreased in diarrhea-predominant IBS patients, but significantly increased in mice with constipation (Liu et al., 2020; Chai et al., 2021). The abundance of Methanobrevibacter, a beneficial bacterium, was decreased in diarrhea pigs. After capsulized fecal microbiota transplantation, the diarrhea symptoms of piglets improved, and the abundance of Methanobrevibacter increased with a decrease in intestinal water content (Tang et al., 2020). Turicibacter is a gram-positive, anaerobic, non-spore-forming bacterium (Wang X. et al., 2021). After weaning calves recovered from bovine coronavirus-mediated diarrhea, the abundance of Turicibacter increased significantly (Kwon et al., 2021). Alistipes is a relatively new bacterial genus, typically isolated from the human gut microbiome (Parker et al., 2020). It requires high survival conditions when cultured in vitro, and therefore, its survival in liquid medium is difficult. Some studies have shown that the abundance of intestinal Alistipes in patients with Sjogren’s syndrome was significantly increased compared with that in patients with xerophthalmia (Moon et al., 2020). However, Alistipes was not observed in the intestine of diarrhea piglets (Huang et al., 2019), which may be because the dry environment of the intestine is more suitable for Alistipes survival. This, from the pathological perspecpective, confirms our results that the abundance of these bacteria is negatively correlated with intestinal water content. Romboutsia is a type of idiopathic anaerobe that is usually detected in the gastrointestinal tract of many vertebrates, including poultry (Maki et al., 2020). Studies have shown that after functional constipation is relieved with chrysanthemum polysaccharide, the relative abundance of Romboutsia increases with the recovery of the intestinal fluid (Wang J. et al., 2021). Similarly, *Helicobacter is* considered a pathogen and is associated with chronic diarrhea (Queiroz et al., 2013). This also confirms our findings from a pathological point of view that the abundance of these bacteria is positively correlated with the water content in the intestinal tract.
Therefore, we speculate that seasonal factors may cause seasonal changes in intestinal flora by affecting the intestinal water metabolism level in rats. Romboutsia and *Helicobacter were* positively correlated with intestinal water content. High temperature and humidity in summer enhanced the intestinal water metabolism level and intestinal water volume in rats, which may have led to an increase in the abundance of Romboutsia and *Helicobacter as* intestinal dominant flora in summer. However, Lactobacillus, Faecalibaculum, Turicibacter, and other bacteria were negatively correlated with intestinal water content. The temperature and humidity were relatively reduced in spring, autumn and winter, the intestinal water metabolism level and the intestinal water volume in rats decreased. At the same time, the abundance of Lactobacillus, Faecalibaculum, Turicibacter, and other bacteria increased, which also confirmed our speculation.
Notably, our present study only explores the difference in intestinal water metabolism caused by changes in environmental factors, such as temperature and humidity, with changes in in seasons. A certain correlation exists between this difference and the seasonal difference in intestinal flora. However, this correlation is not a simple linear relationship. For example, the abundance of Romboutsia exhibited a positive correlation with the AQP3 content, and had a positive correlation trend with fecal water content, but the abundance of Romboutsia increased in the winter environment relative to other seasons. In fact, the intestinal water metabolism level was weaker in winter than in summer. The factors and specific mechanisms affecting colonic water metabolism need to be further explored. At present, combined with studies in the same field, it has been found that mitophagy and oxidative stress participate in the seasonal physiological regulation of the body (Wang Y. et al., 2021). Stress or environmental changes can affect mitochondrial energy metabolism (Chang et al., 2023), and microbiota metabolites can also directly affect mitochondrial oxidative stress and the formation of mitochondrial autophagic lysosomes (Wang J. et al., 2022). Therefore, in the future, we will further explore the relationship between mitochondrial energy metabolism of intestinal cells and intestinal water metabolism, and explore the deep mechanism of adaptive changes in intestinal flora.
## Data availability statement
The data presented in the study are deposited in the NCBI-SRA repository (https://www.ncbi.nlm.nih.gov/bioproject/PRJNA908739), accession number PRJNA908739.
## Ethics statement
The animal study was reviewed and approved by Beijing University of Chinese Medicine Animal Ethics Committee (Approval number: BUCM-4-2,021,032,603-1,059).
## Author contributions
LL, TW, and SM performed the experiments. LL, JL, YS, and RW wrote the manuscript. TW, LS, KW, and HZ revised the manuscript. LL designed and supervised the study. All authors contributed to the article and approved the submitted version.
## Funding
This work was supported a grant provided by National Natural Science Foundation of China (Grant number: 8210152377).
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb.2023.1109696/full#supplementary-material
## References
1. Ahlberg C. M., Allwardt K., Broocks A., Bruno K., McPhillips L., Taylor A.. **Environmental effects on water intake and water intake prediction in growing beef cattle**. *J. Anim. Sci.* (2018) **96** 4368-4384. DOI: 10.1093/jas/sky267
2. Anwar M. Y., Warren J. L., Pitzer V. E.. **Diarrhea patterns and climate: a spatiotemporal Bayesian hierarchical analysis of diarrheal disease in Afghanistan**. *Am. J. Trop. Med. Hyg.* (2019) **101** 525-533. DOI: 10.4269/ajtmh.18-0735
3. Botelho P. B., Ferreira M. V. R., Araújo A. M., Mendes M. M., Nakano E. Y.. **Effect of multispecies probiotic on gut microbiota composition in individuals with intestinal constipation: a double-blind, placebo-controlled randomized trial**. *Nutrition* (2020) **78** 110890. DOI: 10.1016/j.nut.2020.110890
4. Cao M., Yang M., Ou Z., Li D., Geng L., Chen P.. **Involvement of aquaporins in a mouse model of rotavirus diarrhea**. *Virol. Sin.* (2014) **29** 211-217. DOI: 10.1007/s12250-014-3469-z
5. Chai M., Wang L., Li X., Zhao J., Zhang H., Wang G.. **Different B**. *Food Funct.* (2021) **12** 6058-6069. DOI: 10.1039/d1fo00559f
6. Chang P. V., Hao L., Offermanns S., Medzhitov R.. **The microbial metabolite butyrate regulates intestinal macrophage function via histone deacetylase inhibition**. *Proc. Natl. Acad. Sci. U. S. A.* (2014) **111** 2247-2252. DOI: 10.1073/pnas.1322269111
7. Chang X., Li Y., Liu J., Wang Y., Guan X., Wu Q.. **ß-tubulin contributes to Tongyang Huoxue decoction-induced protection against hypoxia/reoxygenation-induced injury of sinoatrial node cells through SIRT1-mediated regulation of mitochondrial quality surveillance**. *Phytomedicine* (2023) **108** 154502. DOI: 10.1016/j.phymed.2022.154502
8. Chen S.-J., Yang J.-F., Kong F.-P., Ren J.-L., Hao K., Li M.. **Overactivation of corticotropin-releasing factor receptor type 1 and aquaporin-4 by hypoxia induces cerebral edema**. *Proc. Natl. Acad. Sci.* (2014) **111** 13199-13204. PMID: 25146699
9. Ciani E., Fontaine R., Maugars G., Mizrahi N., Mayer I., Levavi-Sivan B.. **Melatonin receptors in Atlantic salmon stimulate cAMP levels in heterologous cell lines and show season-dependent daily variations in pituitary expression levels**. *J. Pineal Res.* (2019) **67** e12590. DOI: 10.1111/jpi.12590
10. Davenport E. R., Mizrahi-Man O., Michelini K., Barreiro L. B., Ober C., Gilad Y.. **Seasonal variation in human gut microbiome composition**. *PLoS One* (2014) **9** e90731. DOI: 10.1371/journal.pone.0090731
11. De Angelis M., Garruti G., Minervini F., Bonfrate L., Portincasa P., Gobbetti M.. **The food-gut human Axis: the effects of diet on gut microbiota and metabolome**. *Curr. Med. Chem.* (2019) **26** 3567-3583. DOI: 10.2174/0929867324666170428103848
12. Dechesne A., Or D., Smets B. F.. **Limited diffusive fluxes of substrate facilitate coexistence of two competing bacterial strains**. *FEMS Microbiol. Ecol.* (2008) **64** 1-8. DOI: 10.1111/j.1574-6941.2008.00446.x
13. Dechesne A., Wang G., Gülez G., Or D., Smets B. F.. **Hydration-controlled bacterial motility and dispersal on surfaces**. *Proc. Natl. Acad. Sci. U. S. A.* (2010) **107** 14369-14372. DOI: 10.1073/pnas.1008392107
14. Deng L., Xu H., Liu P., Wu S., Shi Y., Lv Y.. **Prolonged exposure to high humidity and high temperature environment can aggravate influenza virus infection through intestinal flora and nod/RIP2/NF-κB signaling pathway**. *Vet. Microbiol.* (2020) **251** 108896. DOI: 10.1016/j.vetmic.2020.108896
15. Di Ciaula A., Bonfrate L., Portincasa P.. **The role of microbiota in nonalcoholic fatty liver disease**. *Eur. J. Clin. Investig.* (2022) **52** e13768. DOI: 10.1111/eci.13768
16. Du L., Li Q., Yi H., Kuang T., Tang Y., Fan G.. **Gut microbiota-derived metabolites as key actors in type 2 diabetes mellitus**. *Biomed. Pharmacother.* (2022) **149** 112839. DOI: 10.1016/j.biopha.2022.112839
17. Dzenda T., Ayo J. O., Lakpini C. A., Adelaiye A. B.. **Seasonal, sex and live weight variations in feed and water consumptions of adult captive African Giant rats (**. *J. Anim. Physiol. Anim. Nutr.* (2013) **97** 465-474. DOI: 10.1111/j.1439-0396.2012.01287.x
18. Ebling F. J. P.. **Hypothalamic control of seasonal changes in food intake and body weight**. *Front. Neuroendocrinol.* (2015) **37** 97-107. DOI: 10.1016/j.yfrne.2014.10.003
19. Ebling F. J. P., Barrett P.. **The regulation of seasonal changes in food intake and body weight**. *J. Neuroendocrinol.* (2008) **20** 827-833. DOI: 10.1111/j.1365-2826.2008.01721.x
20. Galligan J. J.. **Colonic 5-HT 4 receptors are targets for novel prokinetic drugs**. *Neurogastroenterol. Mot.* (2021) **33** e14125. DOI: 10.1111/nmo.14125
21. Gao J., Xiong T., Grabauskas G., Owyang C.. **Mucosal serotonin reuptake transporter expression in irritable bowel syndrome is modulated by gut microbiota via mast cell–prostaglandin E2**. *Gastroenterology* (2022) **162** 1962-1974.e6. DOI: 10.1053/j.gastro.2022.02.016
22. He J., Yang B.. **Aquaporins in renal diseases**. *Int. J. Mol. Sci.* (2019) **20** 366. DOI: 10.3390/ijms20020366
23. Hu T. G., Wen P., Fu H. Z., Lin G. Y., Liao S. T., Zou Y. X.. **Protective effect of mulberry (**. *Food Funct.* (2019) **10** 1513-1528. DOI: 10.1039/c9fo00132h
24. Huang A., Cai R., Wang Q., Shi L., Li C., Yan H.. **Dynamic change of gut microbiota during porcine epidemic diarrhea virus infection in suckling piglets**. *Front. Microbiol.* (2019) **10** 322. DOI: 10.3389/fmicb.2019.00322
25. Ikarashi N., Kon R., Sugiyama K.. **Aquaporins in the colon as a new therapeutic target in diarrhea and constipation**. *Int. J. Mol. Sci.* (2016) **17** 1172. DOI: 10.3390/ijms17071172
26. Jenkins T. A., Nguyen J. C., Polglaze K. E., Bertrand P. P.. **Influence of tryptophan and serotonin on mood and cognition with a possible role of the gut-brain Axis**. *Nutrients* (2016) **8** E56. DOI: 10.3390/nu8010056
27. Kleynhans E., Conlong D. E., Terblanche J. S.. **Direct and indirect effects of development temperature on adult water balance traits of**. *J. Insect Physiol.* (2014) **68** 69-75. DOI: 10.1016/j.jinsphys.2014.06.018
28. Kon R., Ikarashi N., Hayakawa A., Haga Y., Fueki A., Kusunoki Y.. **Morphine-induced constipation develops with increased Aquaporin-3 expression in the colon via increased serotonin secretion**. *Toxicol. Sci.* (2015) **145** 337-347. DOI: 10.1093/toxsci/kfv055
29. Kwon M. S., Jo H. E., Lee J., Choi K. S., Yu D., Oh Y. S.. **Alteration of the gut microbiota in post-weaned calves following recovery from bovine coronavirus-mediated diarrhea**. *J. Anim. Sci. Technol.* (2021) **63** 125-136. DOI: 10.5187/jast.2021.e20
30. Ley R. E., Hamady M., Lozupone C., Turnbaugh P. J., Ramey R. R., Bircher J. S.. **Evolution of mammals and their gut microbes**. *Science* (2008) **320** 1647-1651. DOI: 10.1126/science.1155725
31. Li T., Wang H., Zhang H., Liu L., Li P., Ma S.. **Effect of the pineal gland on 5-hydroxytryptamine and γ-aminobutyric acid secretion in the hippocampus of male rats during the summer and winter**. *J. Trad. Chin. Medi. Sci.* (2020) **7** 283-290. DOI: 10.1016/j.jtcms.2020.07.004
32. Li Y., Xia S., Jiang X., Feng C., Gong S., Ma J.. **Gut microbiota and diarrhea: An updated review**. *Front. Cell. Infect. Microbiol.* (2021) **11** 625210. DOI: 10.3389/fcimb.2021.625210
33. Liang X., Bushman F. D., FitzGerald G. A.. **Rhythmicity of the intestinal microbiota is regulated by gender and the host circadian clock**. *Proc. Natl. Acad. Sci. U. S. A.* (2015) **112** 10479-10484. DOI: 10.1073/pnas.1501305112
34. Liu T., Gu X., Li L. X., Li M., Li B., Cui X.. **Microbial and metabolomic profiles in correlation with depression and anxiety co-morbidities in diarrhoea-predominant IBS patients**. *BMC Microbiol.* (2020) **20** 168. DOI: 10.1186/s12866-020-01841-4
35. Liu X., Mao B., Gu J., Wu J., Cui S., Wang G.. *Gut Microbes* (2021) **13** 1-21. DOI: 10.1080/19490976.2021.1875796
36. Liu J., Song Y., Lu X., Chen T., Guo W., Fan Z.. **Seasonal variation influences on intestinal microbiota in rats**. *Curr. Microbiol.* (2018) **75** 1006-1010. DOI: 10.1007/s00284-018-1480-6
37. Lv J., Niu Y., Yuan R., Wang S.. **Different responses of bacterial and archaeal communities in river sediments to water diversion and seasonal changes**. *Microorganisms.* (2021) **9** 782. DOI: 10.3390/microorganisms9040782
38. Maki J. J., Nielsen D. W., Looft T.. **Complete genome sequence and annotation for**. *Microbiol. Resour. Announc.* (2020) **9** e00382-e00320. DOI: 10.1128/MRA.00382-20
39. Malisova O., Bountziouka V., Panagiotakos D. Β., Zampelas A., Kapsokefalou M.. **Evaluation of seasonality on total water intake, water loss and water balance in the general population in Greece**. *J. Hum. Nutr. Diet.* (2013) **26** 90-96. DOI: 10.1111/jhn.12077
40. Maurice C. F., Knowles S. C., Ladau J., Pollard K. S., Fenton A., Pedersen A. B.. **Marked seasonal variation in the wild mouse gut microbiota**. *ISME J.* (2015) **9** 2423-2434. DOI: 10.1038/ismej.2015.53
41. Mishra I., Singh D., Kumar V.. **Temporal expression of c-fos and genes coding for neuropeptides and enzymes of amino acid and amine neurotransmitter biosynthesis in retina, pineal and hypothalamus of a migratory songbird: evidence for circadian rhythm-dependent seasonal responses**. *Neuroscience* (2018) **371** 309-324. DOI: 10.1016/j.neuroscience.2017.12.016
42. Mobasheri A., Wray S., Marples D.. **Distribution of AQP2 and AQP3 water channels in human tissue microarrays**. *J. Mol. Histol.* (2005) **36** 1-14. DOI: 10.1007/s10735-004-2633-4
43. Moon J., Choi S. H., Yoon C. H., Kim M. K.. **Gut dysbiosis is prevailing in Sjögren’s syndrome and is related to dry eye severity**. *PLoS One* (2020) **15** e0229029. DOI: 10.1371/journal.pone.0229029
44. Negussie A. B., Dell A. C., Davis B. A., Geibel J. P.. **Colonic fluid and electrolyte transport 2022: An update**. *Cells* (2022) **11** 1712. DOI: 10.3390/cells11101712
45. Or D., Smets B. F., Wraith J. M., Dechesne A., Friedman S. P.. **Physical constraints affecting bacterial habitats and activity in unsaturated porous media–a review**. *Adv. Water Resour.* (2007) **30** 1505-1527. DOI: 10.1016/j.advwatres.2006.05.025
46. Parker B. J., Wearsch P. A., Veloo A. C. M., Rodriguez-Palacios A.. **The genus**. *Front. Immunol.* (2020) **11** 906. DOI: 10.3389/fimmu.2020.00906
47. Pfister T. D., Storey K. B.. **Insect freeze tolerance: roles of protein phosphatases and protein kinase a**. *Insect Biochem. Mol. Biol.* (2006) **36** 18-24. DOI: 10.1016/j.ibmb.2005.10.002
48. Phung D., Huang C., Rutherford S., Chu C., Wang X., Nguyen M.. **Association between climate factors and diarrhoea in a Mekong Delta area**. *Int. J. Biometeorol.* (2015) **59** 1321-1331. DOI: 10.1007/s00484-014-0942-1
49. Queiroz D. M., Rocha A. M., Crabtree J. E.. **Unintended consequences of**. *Gut Microbes* (2013) **4** 494-504. DOI: 10.4161/gmic.26277
50. Ran Z., An Y., Zhou J., Yang J., Zhang Y., Yang J.. **Subchronic exposure to concentrated ambient PM2.5 perturbs gut and lung microbiota as well as metabolic profiles in mice**. *Environ. Pollut.* (2021) **272** 115987. DOI: 10.1016/j.envpol.2020.115987
51. Rutsch A., Kantsjö J. B., Ronchi F.. **The gut-brain Axis: how microbiota and host Inflammasome influence brain physiology and pathology**. *Front. Immunol.* (2020) **11** 604179. DOI: 10.3389/fimmu.2020.604179
52. Salameh M., Burney Z., Mhaimeed N., Laswi I., Yousri N. A., Bendriss G.. **The role of gut microbiota in atopic asthma and allergy, implications in the understanding of disease pathogenesis**. *Scand. J. Immunol.* (2020) **91** e12855. DOI: 10.1111/sji.12855
53. Segawa T., Miyamoto K., Ushida K., Agata K., Okada N., Kohshima S.. **Seasonal change in bacterial Flora and Biomass in mountain snow from the Tateyama Mountains, Japan, analyzed by 16S rRNA gene sequencing and real-time PCR**. *Appl. Environ. Microbiol.* (2005) **71** 123-130. DOI: 10.1128/AEM.71.1.123-130.2005
54. Soria L. R., Gradilone S. A., Larocca M. C., Marinelli R. A.. **Glucagon induces the gene expression of aquaporin-8 but not that of aquaporin-9 water channels in the rat hepatocyte**. *Am. J. Phys. Regul. Integr. Comp. Phys.* (2009) **296** R1274-R1281. DOI: 10.1152/ajpregu.90783.2008
55. Šťovíček A., Kim M., Or D., Gillor O.. **Microbial community response to hydration-desiccation cycles in desert soil**. *Sci. Rep.* (2017) **7** 45735. DOI: 10.1038/srep45735
56. Sun Z., Kumar D., Cao G., Zhu L., Liu B., Zhu M.. **Effects of transient high temperature treatment on the intestinal flora of the silkworm**. *Sci. Rep.* (2017) **7** 3349. DOI: 10.1038/s41598-017-03565-4
57. Tan Q., Hu J., Zhou Y., Wan Y., Zhang C., Liu X.. **Inhibitory effect of**. *Drug Des. Devel. Ther.* (2021) **15** 1971-1980. DOI: 10.2147/DDDT.S309675
58. Tang W., Chen D., Yu B., He J., Huang Z., Zheng P.. **Capsulized faecal microbiota transplantation ameliorates post-weaning diarrhoea by modulating the gut microbiota in piglets**. *Vet. Res.* (2020) **51** 55. DOI: 10.1186/s13567-020-00779-9
59. Ver Heul A., Planer J., Kau A. L.. **The human microbiota and asthma**. *Clin. Rev. Allergy Immunol.* (2019) **57** 350-363. DOI: 10.1007/s12016-018-8719-7
60. Vos M., Wolf A. B., Jennings S. J., Kowalchuk G. A.. **Micro-scale determinants of bacterial diversity in soil**. *FEMS Microbiol. Rev.* (2013) **37** 936-954. DOI: 10.1111/1574-6976.12023
61. Wang L., Alammar N., Singh R., Nanavati J., Song Y., Chaudhary R.. **Gut microbial dysbiosis in the irritable bowel syndrome: a systematic review and meta-analysis of case-control studies**. *J. Acad. Nutr. Diet.* (2020) **120** 565-586. DOI: 10.1016/j.jand.2019.05.015
62. Wang L., Chai M., Wang J., Yu Q., Wang G., Zhang H.. *Food Funct.* (2022) **13** 5037-5049. DOI: 10.1039/D1FO04151G
63. Wang J., Chen P., Cao Q., Wang W., Chang X.. **Traditional Chinese medicine ginseng Dingzhi decoction ameliorates myocardial fibrosis and high glucose-induced Cardiomyocyte injury by regulating intestinal Flora and Mitochondrial dysfunction**. *Oxidative Med. Cell. Longev.* (2022) **2022** 1-33. DOI: 10.1155/2022/9205908
64. Wang J., Liang Q., Zhao Q., Tang Q., Ahmed A. F., Zhang Y.. **The effect of microbial composition and proteomic on improvement of functional constipation by**. *Food Chem. Toxicol.* (2021) **153** 112305. DOI: 10.1016/j.fct.2021.112305
65. Wang Y., Su R., Liu P., Yuan Z., Han Y., Zhang H.. **Seasonal changes of mitochondrial autophagy and oxidative response in the testis of the wild ground squirrels (Spermophilus dauricus)**. *Am. J. Phys. Regul. Integr. Comp. Phys.* (2021) **321** R625-R633. DOI: 10.1152/ajpregu.00105.2021
66. Wang X., Yi W., He L., Luo S., Wang J., Jiang L.. **Abnormalities in gut microbiota and metabolism in patients with chronic spontaneous Urticaria**. *Front. Immunol.* (2021) **12** 691304. DOI: 10.3389/fimmu.2021.691304
67. Xiong R., Li W., Li Y., Zheng K., Zhang T., Gao M.. **Er Shen Wan extract reduces diarrhea and regulates AQP 4 and NHE 3 in a rat model of spleen-kidney Yang deficiency–induced diarrhea**. *Biomed. Pharmacother.* (2018) **98** 834-846. DOI: 10.1016/j.biopha.2018.01.023
68. Yin H., Zhong Y., Wang H., Hu J., Xia S., Xiao Y.. **Short-term exposure to high relative humidity increases blood urea and influences colonic urea-nitrogen metabolism by altering the gut microbiota**. *J. Adv. Res.* (2022) **35** 153-168. DOI: 10.1016/j.jare.2021.03.004
69. You Z., Deng J., Liu J., Fu J., Xiong H., Luo W.. **Seasonal variations in the composition and diversity of gut microbiota in white-lipped deer (**. *PeerJ.* (2022) **10** e13753. DOI: 10.7717/peerj.13753
70. Young I. M., Crawford J. W., Nunan N., Otten W., Spiers A., Sparks D. L.. **Microbial distribution in soils: physics and scaling**. *Advances in agronomy* (2008) 81-121
71. Zhang J., Li S., Deng F., Baikeli B., Yu W., Liu G.. **Distribution of aquaporins and sodium transporters in the gastrointestinal tract of a desert hare,**. *Sci. Rep.* (2019) **9** 16639-16714. DOI: 10.1038/s41598-019-53291-2
72. Zhao L., Ren P., Wang M., Wang J., He X., Gu J.. **Changes in intestinal barrier protein expression and intestinal flora in a rat model of visceral hypersensitivity**. *Neurogastroenterol. Motil.* (2022) **34** e14299. DOI: 10.1111/nmo.14299
73. Zhu X.. **Expression of AQP3 protein in hAECs is regulated by camp-PKA-CREB signaling pathway**. *Front. Biosci.* (2015) **20** 1047-1055. DOI: 10.2741/4357
74. Zhu Z., Sun Y., Zhu F., Liu Z., Pan R., Teng L.. **Seasonal variation and sexual dimorphism of the microbiota in wild blue sheep (**. *Front. Microbiol.* (2020) **11** 1260. DOI: 10.3389/fmicb.2020.01260
75. Zou S., Fang L., Lee M.-H.. **Dysbiosis of gut microbiota in promoting the development of colorectal cancer**. *Gastroenterol. Rep.* (2018) **6** 1-12. DOI: 10.1093/gastro/gox031
|
---
title: Visceral fat correlates with insulin secretion and sensitivity independent
of BMI and subcutaneous fat in Chinese with type 2 diabetes
authors:
- Haishan Huang
- Xiaobin Zheng
- Xiaoming Wen
- Jingyi Zhong
- Yanting Zhou
- Lingling Xu
journal: Frontiers in Endocrinology
year: 2023
pmcid: PMC9999013
doi: 10.3389/fendo.2023.1144834
license: CC BY 4.0
---
# Visceral fat correlates with insulin secretion and sensitivity independent of BMI and subcutaneous fat in Chinese with type 2 diabetes
## Abstract
### Aim
Clinical heterogeneity exists in overall obesity and abdominal obesity in terms of insulin secretion and sensitivity. Further, the impact of visceral fat (VF) on the first- and second-phase insulin secretion (FPIS and SPIS) is controversial. We aim to investigate insulin secretion and sensitivity in Chinese patients with T2DM according to different BMI and VF levels.
### Methods
This study enrolled 300 participants. A dual bioelectrical impedance analyzer was used to assess the visceral and subcutaneous fat area (VFA and SFA). VF levels were categorized as normal or high, with the cutoff value of 100 cm2. FPIS and SPIS were evaluated by arginine stimulation test and standardized steamed bread meal tolerance test, respectively. β-cell function (HOMA2-β), insulin resistance (HOMA2-IR), and Gutt’s insulin sensitivity index (Gutt-ISI) were also calculated. Spearman’s correlation analysis and multivariate linear regression analysis were adopted for statistical analysis.
### Results
Participants were categorized into four groups: normal weight-normal VF, normal weight-high VF, overweight/obese-normal VF and overweight/obese-high VF. Multivariate linear regression showed that both VFA and SFA were correlated with FPIS, HOMA2-IR and Gutt-ISI after controlling for gender and diabetes duration. After further adjustment for BMI and VFA, some associations of SFA with insulin secretion and sensitivity disappeared. After adjustment for gender, diabetes duration, BMI and SFA, VFA was positively correlated with FPIS, SPIS and HOMA2-IR. Subjects with overweight/obese-high VF were more likely to have higher FPIS, HOMA2-IR and lower Gutt-ISI (all $p \leq 0.05$).
### Conclusion
VF affects both FPIS and SPIS, and worsens insulin sensitivity independent of BMI and subcutaneous fat in Chinese patients with T2DM.
### Clinical trial registration
http://www.chictr.org.cn, identifier ChiCTR2200062884.
## Introduction
Type 2 diabetes mellitus (T2DM) has become a major public health problem worldwide. The pathophysiology of T2DM is characterized by insulin resistance (IR) and β-cell dysfunction [1, 2]. After disease onset, β-cell function progressively declines over time. Thus, exploring potential risk factors of β-cell dysfunction is important for preventing or delaying the development of diabetes [3, 4].
Obesity is a major risk factor for IR and T2DM [5]. Although body mass index (BMI) is an internationally recognized index for diagnosing obesity, some studies have shown that obesity defined by BMI is remarkably heterogenous, and people with similar BMI do not have the same level of T2DM risk [6, 7]. Abdominal obesity, specifically visceral adipose tissue (VAT), is associated with a greater risk of developing T2DM than peripheral obesity because expanded visceral fat stores affect insulin metabolism by releasing free fatty acids into the portal circulation, which may reduce the hepatic clearance of insulin, thus leading to IR and hyperinsulinemia [8]. Further, Chinese have more visceral fat (VF) than Caucasians with the same BMI [9, 10]. Therefore, it is significant to discover the differences of insulin secretion and sensitivity in Chinese with different types of obesity.
Measurement of VF accumulation is essential for the diagnosis of obesity. The visceral fat area (VFA) measured by dual bioelectrical impedance analysis (dual-BIA) is a simple and reliable method to estimate VF accumulation. Dual-BIA measures the bioelectrical impedance of the entire abdomen and its surface with a dual current path, which is considered better than the conventional BIA using only one current path and has high correlation with computed tomography (CT), a gold standard for VF accumulation [11, 12]. To our knowledge, no population-based studies have examined the associations of β-cell function with VAT evaluated by dual-BIA.
The aim of this study was to examine the association of abdominal obesity assessed by dual-BIA with basal and post-load β-cell function, and clarify whether VAT and subcutaneous adipose tissue (SAT) have the same predictive effect on insulin secretion and sensitivity in Chinese patients with T2DM.
## Study population
This cross-sectional study recruited individuals hospitalized at the Department of Endocrinology of Shenzhen hospital, Southern Medical University, between August 2022 and November 2022. Inclusion criteria were Chinese participants who met the criteria of T2DM diagnosis based on the WHO consulting group [13]; aged ≥ 18 years; BMI ≥ 18.5 kg/m2. Those with infectious diseases, cancer, or recent acute diabetic complications were excluded.
The study was approved by the Medical Ethics Committee, Shenzhen Hospital, Southern Medical University (NYSZYYEC202200017), and was registered at the Chinese Clinical Trials Registry (ChiCTR2200062884). All subjects signed informed consent before the investigation.
## Clinical measurements
Basic information, including age, gender, history of diabetes, history of diabetic complications and co-morbidities, drug use history, and other events in the exclusion criteria were collected. All participants underwent physical examination, which included measuring systolic and diastolic blood pressure (SBP and DBP), height (m), and weight (kg). BMI was calculated as weight (kg) divided by height (m) squared. Normal weight was defined as 18.5 ≤ BMI < 24 kg/m2, and overweight/obese was BMI ≥ 24 kg/m2 according to the Working Group on Obesity in China (WGOC) [2002] [14].
## Visceral fat area measurement by dual-BIA
VFA, along with subcutaneous fat area (SFA), was measured at the umbilical level by a dual bioelectrical impedance analyzer (Omron HDS-2000 DUALSCAN, Omron Healthcare Co, Kyoto, Japan), an equipment mainly designed to assess the abdominal fat area, as previously described [11, 15]. Briefly, eight-point tactile electrode method was utilized according to the protocol. Resistance at five specific frequencies (1, 50, 250, 500 kHz, and 1 MHz) and reactance at three specific frequencies (5, 50 and 250 kHz) were measured to obtain the reading of VFA (cm2) and SFA (cm2) on the screen. The ratio of VFA and SFA (V/S ratio) was evaluated. All measurements were performed by the same experienced technician.
We used cutoff value of 100 cm2 in VFA to define visceral adiposity for both men and women [16]. Thereafter, participants were categorized into four groups based on combinations of BMI and VF categories as follows: [1] normal weight-normal VF (18.5 kg/m2 BMI < 24 kg/m2 and VFA < 100 cm2), [2] normal weight-high VF (18.5 kg/m2 BMI < 24 kg/m2 and VFA ≥ 100 cm2), [3] overweight/obese-normal VF (BMI ≥ 24 kg/m2 and VFA <100 cm2), [4] overweight/obese-high VF (BMI ≥ 24 kg/m2 and VFA ≥ 100 cm2).
## Arginine stimulation test
AST was used to assess first-phase insulin secretion (FPIS) after overnight fasting for at least eight hours. After a baseline blood sample was collected, a $10\%$ (wt/vol.) solution of arginine hydrochloride (5 g) (Shanghai Xinyi Jinzhu Pharmaceutical Co., Ltd., Shanghai, China) was injected intravenously within 30-45 s. Blood samples were obtained at 2, 4, and 6 min after injection [17]. All anti-diabetic therapy was paused during the test.
## Standardized steamed bread meal tolerance test and biochemical measurements
The standardized steamed bread meal was made of 100 g flour, which contained carbohydrates approximately equivalent to 75 g glucose. The Chinese Islet Beta-Cell Function Collaborative Research Group showed that standardized steamed bread meal tolerance test (BMTT) was reproducible and was better tolerated when compared to oral glucose tolerance test (OGTT) to assess β-cell function in healthy subjects [18]. Therefore, in China, BMTT is often used in clinical practice instead of OGTT to evaluate β-cell function in patients previously diagnosed with diabetes [19]. Thus, we used BMTT to assess the second phase insulin secretion (SPIS).
Blood samples were collected in the morning under fasting conditions. Glycated hemoglobin (HbA1c), fasting plasma glucose (FPG), fasting insulin (FINS), fasting C-peptide (FCP), total cholesterol (TC), triglycerides (TG), high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C) and high-sensitivity C-reactive protein (Hs-CRP) were measured. Post-load blood samples were collected to assess 2 h plasma glucose (PG2h), 2 h insulin (INS2h) and 2 h C-peptide (CP2h) after the patients ate a 100 g steamed bread.
## Evaluation of insulin secretion and sensitivity
Since C-peptide response was equal to insulin response [20], the first and second-phase insulin release were separately calculated using the following formulate: FPIS = [(CP2min + CP4min + CP6min)/3 - CP0min]/[(PG2min + PG4min + PG6min)/3 - PG0min] and SPIS = (CP2h – FCP)/(PG2h – FPG) [21, 22]. Basal β-cell function and IR were determined by employing updated Homeostasis Model Assessment (HOMA2) model of HOMA2-β and HOMA2-IR, which could be calculated by entering FPG and FCP into the HOMA Calculator software v2.2.3 [23]. The postprandial insulin sensitivity index (ISI) was estimated according to the computation proposed by Gutt’s et al. [ 24]. Therefore, we generated the following five indices: FPIS, SPIS, HOMA2-β, HOMA2-IR and Gutt-ISI.
## Statistical analyses
Data were analyzed using the SPSS software package (version 24.0; SPSS Inc, Chicago, IL, USA). Continuous variables were presented as means ± standard deviation (SD) for normal distribution or median with interquartile ranges for non-normal distribution. Categorical variables were presented as frequency (percentages). The Kolmogorov-Smirnov test was used to verify the normal distribution of continuous variables. The χ² test, one-way ANOVA or Kruskal-Wallis rank sum test were used to compare differences in categorical or continuous variables across the four groups, as appropriate. Relationships between abdominal fat distribution and β-cell function were analyzed using Spearman’s correlation analysis. All the covariates were tested for collinearity; the tolerance was > 0.1, and variance inflation factor did not > 5.0. Multivariate linear regression was used to assess the association of abdominal fat distribution with insulin secretion and sensitivity. A p value < 0.05 (two-sided) was considered as statistically significant.
## Results
The basic clinical characteristics of participants are summarized in Table 1. A total of 300 patients, 221 ($73.67\%$) men and 79 ($26.33\%$) women, with a mean age of 51.25 ± 11.96 years were included for data analysis. The median (25th, 75th percentile) VFA and SFA of the subjects were 100.00 (77.00, 127.00) cm2 and 171.00 (138.25, 211.00) cm2, respectively.
**Table 1**
| Unnamed: 0 | Total (n=300) | Normal weight-normal VF (n = 99) | Normal weight-high VF (n = 31) | Overweight/obese-normal VF (n = 49) | Overweight/obese-high VF (n = 121) | p value |
| --- | --- | --- | --- | --- | --- | --- |
| Age (years) | 51.25 ± 11.96 | 53.04 ± 11.59 | 57.32 ± 10.30 | 51.86 ± 10.84 | 47.99 ± 12.26 **†† | < 0.001 |
| Male/female (%) | 221/79 (73.67/26.33) | 61/38 (61.62/38.38) | 22/9 (70.97/29.03) | 35/14 (71.43/28.57) | 103/18 (85.12/14.88)* | 0.001 |
| BMI (kg/m2) | 24.71 (22.68, 26.68) | 22.27 (21.11, 23.23) | 22.84 (21.30, 23.77) | 25.15 (24.70, 26.17) ***††† | 27.30 (25.62, 29.10) ***†††‡‡ | < 0.001 |
| Diabetes duration (years) | 6.58 (1.00, 12.00) | 8.00 (2.00, 14.00) | 8.00 (2.00, 17.00) | 8.00 (1.50, 14.50) | 3.00 (0.67, 10.00) ** | 0.003 |
| SBP (mmHg) | 126.07 ± 17.20 | 120.81 ± 17.00 | 126.55 ± 19.54 | 124.59 ± 16.26 | 130.84 ± 15.93*** | < 0.001 |
| DBP (mmHg) | 79.00 (72.25, 85.00) | 75.00 (69.00, 80.00) | 78.00 (71.00, 85.00) | 77.00 (71.50, 84.00) | 83.00 (76.00, 89.50) ***‡ | < 0.001 |
| Hs-CRP (mg/L) | 1.31 (0.60, 3.13) | 1.06 (0.40, 2.94) | 1.30 (0.68, 2.76) | 0.86 (0.29, 2.24) | 1.91 (0.92, 3.73) **††‡‡‡ | < 0.001 |
| TG (mmol/L) | 1.52 (1.04, 2.33) | 1.23 (0.92, 2.00) | 1.48 (0.92, 2.38) | 1.51 (0.97, 2.17) | 1.85 (1.27, 2.68) *** | < 0.001 |
| TC (mmol/L) | 4.41 (3.73, 5.09) | 4.46 (3.52, 5.05) | 4.26 (3.27, 4.90) | 4.34 (3.77, 5.15) | 4.42 (3.87, 5.11) | NS |
| LDL (mmol/L) | 2.80 ± 0.98 | 2.86 ± 1.02 | 2.55 ± 0.88 | 2.81 ± 1.02 | 2.80 ± 0.94 | NS |
| HDL (mmol/L) | 1.06 (0.90, 1.26) | 1.09 (0.95, 1.36) | 1.12 (0.91, 1.37) | 1.04 (0.86, 1.30) | 0.99 (0.87, 1.17) ** | 0.006 |
| HbA1c (%) | 9.00 (7.30, 11.08) | 9.50 (7.50, 11.20) | 8.60 (7.20, 10.80) | 8.70 (6.60, 11.25) | 9.00 (7.70, 10.85) | NS |
| FPG (mmol/L) | 7.31 (5.91, 9.38) | 7.09 (5.87, 9.27) | 7.27 (5.43, 8.30) | 6.59 (5.24, 8.06) | 8.12 (6.36, 9.97) ‡ | 0.010 |
| PG2h (mmol/L) | 15.01 ± 4.73 | 15.24 ± 5.03 | 14.60 ± 4.62 | 14.40 ± 4.74 | 15.18 ± 4.52 | NS |
| FCP (ng/mL) | 2.01 (1.26, 2.78) | 1.47 (1.02, 2.17) | 2.08 (1.59, 2.52) | 1.80 (1.10, 2.59) | 2.54 (1.60, 3.31) ***‡‡ | < 0.001 |
| CP2h (ng/mL) | 4.87 (3.07, 7.21) | 3.83 (2.46, 5.69) | 4.58 (3.33, 7.92) | 4.57 (3.33, 7.57) | 6.15 (4.04, 8.27) *** | < 0.001 |
| FINS (μU/mL) | 6.88 (4.05, 11.08) | 5.09 (3.22, 6.79) | 6.85 (4.30, 9.81) | 6.99 (3.62, 11.09) | 10.21 (6.19, 14.89) ***†‡‡ | < 0.001 |
| INS2h (μU/mL) | 27.98 (14.83, 48.59) | 21.81 (11.10, 34.44) | 20.65 (14.02, 46.26) | 28.55 (16.14, 51.11) | 35.67 (21.78, 62.48) *** | < 0.001 |
| VFA (cm/2) | 100.00 (77.00, 127.00) | 73.00 (52.00, 82.00) | 119.00 (110.00, 130.00) *** | 87.00 (71.50, 94.00) ††† | 129.00 (114.50, 153.50) ***†††‡‡‡ | < 0.001 |
| SFA (cm/2) | 171.00 (138.25, 211.00) | 133.00 (110.00, 153.00) | 151.00 (130.00, 170.00) | 174.00 (156.00, 193.50) *** | 219.00 (186.50, 267.00) ***†††‡‡‡ | < 0.001 |
| V/S ratio | 0.58 ± 0.18 | 0.50 ± 0.15 | 0.82 ± 0.16*** | 0.48 ± 0.10††† | 0.61 ± 0.15***†††‡‡‡ | < 0.001 |
| FPIS | 4.44 (2.22, 8.64) | 3.22 (1.24, 5.23) | 4.29 (2.52, 9.45) | 4.05 (2.42, 6.68) | 5.48 (3.05, 14.27) *** | < 0.001 |
| SPIS | 0.40 (0.19, 0.77) | 0.32 (0.16, 0.63) | 0.43 (0.19, 0.90) | 0.42 (0.22, 1.00) | 0.47 (0.24, 0.79) * | 0.023 |
| HOMA2-β (%) | 57.05 (32.95, 97.35) | 54.10 (29.30, 81.90) | 57.60 (38.90, 129.90) | 71.50 (33.15, 107.45) | 58.90 (37.95, 98.05) | NS |
| HOMA2-IR (%) | 1.67 (1.10, 2.30) | 1.27 (0.88, 1.75) | 1.78 (1.37, 2.10) | 1.55 (0.90, 2.08) | 2.13 (1.45, 2.83) ***‡‡ | < 0.001 |
| Gutt-ISI | 45.78 (36.28, 59.52) | 51.92 (41.37, 65.29) | 48.43 (40.15, 65.50) | 47.93 (35.08, 71.04) | 39.77 (33.33, 51.59) ***†‡ | < 0.001 |
Table 1 also shows the characteristics of subjects according to BMI and VFA levels. Among the 130 participants with normal weight, 99 had normal VF and 31 had high VF. Among the 170 participants who were overweight/obese, 49 had normal VF and 121 had high VF. The participants in the overweight/obese-high VF group were younger, had higher SBP, DBP, Hs-CRP, TG, FINS, INS2h, FCP, CP2h, FPIS, SPIS, HOMA2-IR, shorter diabetes duration, lower HDL and Gutt-ISI compared to the normal weight-normal VF group (all $p \leq 0.05$). In addition, the participants in the overweight/obese-high VF group had higher DBP, Hs-CRP, FPG, FCP, HOMA2-IR and lower Gutt-ISI than those with normal VF (all $p \leq 0.05$).
Spearman’s correlation analysis (Table 2) showed that BMI, VFA and SFA were positively correlated with FPIS ($r = 0.302$, 0.318, 0.294, all $p \leq 0.001$), SPIS ($r = 0.152$, 0.170 and 0.144, $$p \leq 0.008$$, 0.003 and 0.013), HOMA2-β ($r = 0.119$, 0.135 and 0.121, $$p \leq 0.039$$, 0.019 and 0.037), HOMA2-IR ($r = 0.388$, 0.418 and 0.432, all $p \leq 0.001$), and negatively correlated with Gutt-ISI (r = -0.271, -0.292 and -0.308, all $p \leq 0.001$), while there was no significant correlation between V/S ratio and the indexes of insulin secretion and sensitivity.
**Table 2**
| Unnamed: 0 | BMI | Unnamed: 2 | VFA | Unnamed: 4 | SFA | Unnamed: 6 | V/S ratio | V/S ratio.1 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| | r | p | r | p | r | p | r | p |
| FPIS | 0.302 | < 0.001 | 0.318 | < 0.001 | 0.294 | < 0.001 | 0.107 | NS |
| SPIS | 0.152 | 0.008 | 0.170 | 0.003 | 0.144 | 0.013 | 0.095 | NS |
| HOMA2-β | 0.119 | 0.039 | 0.135 | 0.019 | 0.121 | 0.037 | 0.087 | NS |
| HOMA2-IR | 0.388 | <0.001 | 0.418 | <0.001 | 0.432 | <0.001 | 0.084 | NS |
| Gutt-ISI | -0.271 | <0.001 | -0.292 | <0.001 | -0.308 | <0.001 | -0.051 | NS |
Multivariate linear regression was used to analyze the relationship between dependent (FPIS, SPIS, HOMA2-β, HOMA2-IR and Gutt-ISI) and predictor variables (VFA and SFA) (Table 3). After adjusted for gender and diabetes duration, either VFA or SFA was significantly and positively associated with FPIS (standard β = 0.229 and 0.149, $p \leq 0.001$ and $$p \leq 0.012$$), HOMA2-IR (standard β = 0.400 and 0.393, both $p \leq 0.001$) but negatively associated with Gutt-ISI (standard β = -0.199 and -0.158, $$p \leq 0.001$$ and $$p \leq 0.007$$). After further adjustment for BMI and VFA, the positive association of SFA with FPIS, HOMA2-IR and the inverse association of SFA with Gutt-ISI disappeared. After further adjustment of BMI and SFA, the positive association of VFA with FPIS (standard β = 0.225, $$p \leq 0.010$$), SPIS (standard β = 0.198, $$p \leq 0.024$$), HOMA2-IR (standard β = 0.211, $$p \leq 0.008$$) still remained significant but the negative association of VFA with Gutt-ISI disappeared.
**Table 3**
| Dependent variable | Predictor variables | Standard β (CI) | p |
| --- | --- | --- | --- |
| FPIS | VFA, adjusted for gender, diabetes duration | 0.229 (0.091, 0.282) | < 0.001 |
| | SFA, adjusted for gender, diabetes duration | 0.149 (0.018, 0.139) | 0.012 |
| | VFA, adjusted for gender, diabetes duration, BMI and SFA | 0.225 (0.045, 0.321) | 0.010 |
| | SFA, adjusted for gender, diabetes duration, BMI and VFA | -0.047 (-0.136, 0.087) | NS |
| SPIS | VFA, adjusted for gender, diabetes duration | 0.066 (-0.005, 0.019) | NS |
| | SFA, adjusted for gender, diabetes duration | -0.020 (-0.009, 0.006) | NS |
| | VFA, adjusted for gender, diabetes duration, BMI and SFA | 0.198 (0.003, 0.038) | 0.024 |
| | SFA, adjusted for gender, diabetes duration, BMI and VFA | 0.004 (-0.014, 0.014) | NS |
| HOMA2-β | VFA, adjusted for gender, diabetes duration | 0.122 (0.003, 0.298) | 0.046 |
| | SFA, adjusted for gender, diabetes duration | 0.079 (-0.030, 0.156) | NS |
| | VFA, adjusted for gender, diabetes duration, BMI and SFA | 0.085 (-0.107, 0.317) | NS |
| | SFA, adjusted for gender, diabetes duration, BMI and VFA | -0.129 (-0.274, 0.068) | NS |
| HOMA2-IR | VFA, adjusted for gender, diabetes duration | 0.400 (0.008, 0.014) | < 0.001 |
| | SFA, adjusted for gender, diabetes duration | 0.393 (0.005, 0.009) | < 0.001 |
| | VFA, adjusted for gender, diabetes duration, BMI and SFA | 0.211 (0.001, 0.010) | 0.008 |
| | SFA, adjusted for gender, diabetes duration, BMI and VFA | 0.178 (0.000, 0.006) | NS |
| Gutt-ISI | VFA, adjusted for gender, diabetes duration | -0.199 (-0.316, -0.081) | 0.001 |
| | SFA, adjusted for gender, diabetes duration | -0.158 (-0.176, -0.028) | 0.007 |
| | VFA, adjusted for gender, diabetes duration, BMI and SFA | -0.154 (-0.324, 0.017) | NS |
| | SFA, adjusted for gender, diabetes duration, BMI and VFA | -0.011 (-0.144, 0.130) | NS |
To understand the indexes of insulin secretion and sensitivity associated with different obesity patterns, we developed multiple linear regression models (Table 4). Independent variables were normal weight-normal VF, normal weight-high VF, overweight/obese-normal VF, and overweight/obese-high VF. Dependent variables were FPIS, SPIS, HOMA2-β, HOMA2-IR and Gutt-ISI. After adjustment of gender and diabetes duration, FPIS, HOMA2-IR were higher and Gutt-ISI was lower in the overweight/obese-high VF group when compared to the normal weight-normal VF group and overweight/obese-normal VF group, respectively.
**Table 4**
| Unnamed: 0 | FPIS | Unnamed: 2 | SPIS | Unnamed: 4 | HOMA2-β | Unnamed: 6 | HOMA2-IR | Unnamed: 8 | Gutt-ISI | Unnamed: 10 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| | Standard β (CI) | p | Standard β (CI) | p | Standard β (CI) | p | Standard β (CI) | p | Standard β (CI) | p |
| Model 1 | Model 1 | Model 1 | Model 1 | Model 1 | Model 1 | Model 1 | Model 1 | Model 1 | Model 1 | Model 1 |
| Normal weight-normal VF (Reference) | 0 | | 0 | | 0 | | 0 | | 0 | |
| Normal weight-high VF | 0.085(-3.989, 22.384) | NS | 0.023(-1.368, 1.996) | NS | 0.079(-7.223, 33.207) | NS | 0.075(-0.148, 0.684) | NS | -0.033(-20.420, 11.827) | NS |
| Overweight/obese-normal VF | 0.029(-8.604, 13.797) | NS | 0.013(-1.282, 1.575) | NS | 0.121(-.875, 33.468) | NS | 0.059(-0.180, 0.527) | NS | -0.022(-16.124, 11.267) | NS |
| Overweight/obese-high VF | 0.215(5.369, 23.393) | 0.002 | 0.027(-0.921, 1.377) | NS | 0.125(-1.146, 26.487) | NS | 0.369(0.529, 1.097) | < 0.001 | -0.227(-29.616, -7.577) | 0.001 |
| Model 2 | Model 2 | Model 2 | Model 2 | Model 2 | Model 2 | Model 2 | Model 2 | Model 2 | Model 2 | Model 2 |
| Normal weight-normal VF | -0.037(-13.797, 8.604) | NS | -0.017(-1.575, 1.282) | NS | -0.154(-33.468, 0.875) | NS | -0.075(-0.527, 0.180) | NS | 0.028(-11.267, 16.124) | NS |
| Normal weight-high VF | 0.061(-8.072, 21.274) | NS | 0.012(-1.704, 2.039) | NS | -0.020(-25.800, 19.191) | NS | 0.027(-0.368, 0.557) | NS | -0.014(-19.809, 16.074) | NS |
| Overweight/obese-normal VF(Reference) | 0 | | 0 | | 0 | | 0 | | 0 | |
| Overweight/obese-high VF | 0.176(0.779, 22.790) | 0.036 | 0.010(-1.322, 1.485) | NS | -0.036(-20.498, 13.246) | NS | 0.290(0.293, 0.987) | < 0.001 | -0.197(-29.625, -2.711) | 0.019 |
## Discussion
This study examined the cross-sectional associations of abdominal fat distribution with basal and post-load insulin secretion and sensitivity in Chinese patients with T2DM based on different BMI and VF levels. The results demonstrated that [1] after adjustment for gender, diabetes duration, BMI and VFA, some associations of SFA with insulin secretion and sensitivity indices disappeared. Of note, after adjustment of gender, diabetes duration, BMI and SFA, the association of VFA with FPIS, SPIS and HOMA2-IR still remained significant; [2] overweight/obese-high VF patients were more likely to have higher FPIS, HOMA2-IR and lower Gutt-ISI.
Our study found that VFA, rather than SFA, was associated with HOMA2-β, an index reflected basal insulin secretion; VFA was also correlated with post-load insulin secretion, including FPIS and SPIS, independent of SFA and BMI. The influence of obesity on insulin secretion is controversial. Kautzky-Willer et al. [ 25] found that there was no difference in the dynamic sensitivities to glucose of FPIS and SPIS as studied by the C-peptide minimal model. Bonadonna et al. [ 26] found an increase in both FPIS and SPIS as assessed by hyperglycemic clamp. Walton [27] et al. and Macor et al. [ 28] reported that an increased centrality of fat distribution was associated with an elevated SPIS rather than FPIS. Walton [27] et al. and Macor et al. [ 28] assessed FPIS by intravenous glucose tolerance test (IVGTT) while we assessed FPIS by AST in this study. Progressive impairment in the FPIS to glucose was evident with increasing severity of glucose intolerance; however, patients with T2DM may still have residual β-cell function in response to non-glucose stimulation [29], which may partly explain the association of VFA with FPIS in our study.
In addition to insulin secretion assessed by AST and BMTT, we also used HOMA2-IR and Gutt-ISI to assess insulin sensitivity. *In* general, HOMA2-IR are derived in the basal state and can therefore be considered to reflects basal or hepatic insulin sensitivity [30], whereas Gutt-ISI is a measure of post–glucose loading insulin resistance and represents both peripheral and hepatic insulin sensitivity, which have a higher correlation with the gold standard method for measuring insulin sensitivity: the euglycemic hyperinsulinemic clamp [31]. The results that VFA correlated with HOMA2-IR independent to SFA and BMI suggested that VFA plays important roles in hepatic insulin sensitivity. The association of Gutt-ISI with VFA was disappeared when the confounders were added to SFA and BMI, which indicated that peripheral insulin sensitivity may also affected by SAT. Once SAT reaches its maximal expanding capacity, fatty acids redistribute ectopically in VAT and non-adipose tissues (32–34). Increased VAT leads to an increase in systemic release in resistin and possibly interleukins, and elevated circulating cytokines may play a role in the impairment of muscle insulin response [35]. HOMA2-IR was higher and Gutt-ISI was lower in subjects of the overweight/obese-high VF group which confirmed that in subjects with higher VF, even though insulin secretion is higher, their β-cells could not compensate fully for decreased insulin sensitivity, thus leading to diabetes.
This study had several strengths. First, we assessed FPIS and SPIS by AST and BMTT, both of which generate a supraphysiologic insulin secretory response and are less technically demanding than methodologies such as hyperglycemic clamp. AST provides a measure of near-maximal insulin secretion (insulin secretory reserve) [36], while BMTT is easy to administer and is more suitable in β-cell function assessment than OGTT for subjects who have confirmed diabetes. Second, we assessed basal and postprandial insulin sensitivity by employing HOMA2 model and Gutt’s equation, respectively. Although hyperglycemic clamp is the gold standard to assess insulin sensitivity, it is technically challenging, while the indexes in our study are easy to calculate and are suitable for large sample size studies. Third, we divided participants into four groups based on BMI and VF levels, and found that even if subjects had same BMI levels, only those with high VF were associated with the indexes of insulin secretion and sensitivity.
This study had some limitations. First, the number of normal weight-high VF and overweight/obese-normal VF participants in this study was relatively limited; further, T2DM patients with the same BMI may have gender differences in the association of abdominal obesity with β-cell function though we did not find in the present study, a larger population study subgrouped by male and female is needed to confirm our results. Second, the levels of VAT and SAT were measured by dual-BIA rather than CT, which is a gold standard [37, 38]. However, CT has problems of complexity, cost and X-ray exposure while dual-BIA, mainly designed to assess VFA and SFA, is simple and may have comparable effectiveness as CT [11, 12]. Third, this was a single-center study, and the results might be applicable only to adults with T2DM in southern China.
In summary, the current study suggested that VAT affected basal and post-load insulin secretion and sensitivity. Hence, practitioners should not undermine the risk of IR and β-cell dysfunction in their patients entirely based on BMI, but consider fat distribution as well. For overweight/obese T2DM patients, especially those with VF accumulation, early intervention is needed to reduce VF, in order to delay β-cell dysfunction.
## Data availability statement
The original contributions presented in the study are included in the article/supplementary material. Further inquiries can be directed to the corresponding author.
## Ethics statement
The studies involving human participants were reviewed and approved by Medical Ethics Committee, Shenzhen Hospital, Southern Medical University (NYSZYYEC202200017). The patients/participants provided their written informed consent to participate in this study.
## Author contributions
HH conducted the statistical analyses and wrote the first draft of the manuscript. XZ and XW were involved in the interpretation of data. JZ and YZ contributed to the acquisition of data. LX designed the study and is the guarantor of this work. All authors contributed to the article and approved the submitted version.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## References
1. Porte D. **Banting lecture 1990**. *Beta-cells Type II Diabetes Mellitus Diabetes* (1991) **40**. DOI: 10.2337/diab.40.2.166
2. Kahn SE. **The relative contributions of insulin resistance and beta-cell dysfunction to the pathophysiology of type 2 diabetes**. *Diabetologia* (2003) **46** 3-19. DOI: 10.1007/s00125-002-1009-0
3. **U.K. prospective diabetes study 16. overview of 6 years’ therapy of type II diabetes: a progressive disease. U.K. prospective diabetes study group**. *Diabetes* (1995) **44**. DOI: 10.2337/diab.44.11.1249
4. Knowler WC, Barrett-Connor E, Fowler SE, Hamman RF, Lachin JM, Walker EA. **Reduction in the incidence of type 2 diabetes with lifestyle intervention or metformin**. *New Engl J Med* (2002) **346** 393-403. DOI: 10.1056/NEJMoa012512
5. Kahn SE, Hull RL, Utzschneider KM. **Mechanisms linking obesity to insulin resistance and type 2 diabetes**. *Nature* (2006) **444**. DOI: 10.1038/nature05482
6. Tchernof A, Després JP. **Pathophysiology of human visceral obesity: An update**. *Physiol Rev* (2013) **93** 359-404. DOI: 10.1152/physrev.00033.2011
7. González-Muniesa P, Mártinez-González MA, Hu FB, Després JP, Matsuzawa Y, Loos RJF. **Obesity**. *Nat Rev Dis Primers* (2017) **3** 17034. DOI: 10.1038/nrdp.2017.34
8. Björntorp P. **Metabolic implications of body fat distribution**. *Diabetes Care* (1991) **14**. DOI: 10.2337/diacare.14.12.1132
9. Deurenberg P, Yap M, van Staveren WA. **Body mass index and percent body fat: a meta analysis among different ethnic groups**. *Int J Obes Related Metab Disorders: J Int Assoc Stud Obes* (1998) **22**. DOI: 10.1038/sj.ijo.0800741
10. Deurenberg P, Deurenberg-Yap M, Guricci S. **Asians are different from caucasians and from each other in their body mass index/body fat per cent relationship**. *Obes Rev: An Off J Int Assoc Stud Obes* (2002) **3**. DOI: 10.1046/j.1467-789X.2002.00065.x
11. Yamakage H, Ito R, Tochiya M, Muranaka K, Tanaka M, Matsuo Y. **The utility of dual bioelectrical impedance analysis in detecting intra-abdominal fat area in obese patients during weight reduction therapy in comparison with waist circumference and abdominal CT**. *Endocrine J* (2014) **61**. DOI: 10.1507/endocrj.EJ14-0092
12. Shiga T, Hamaguchi T, Oshima Y, Kanai H, Nakao K. **A new simple measurement system of visceral fat accumulation by bioelectrical impedance analysis**. *Springer Berlin Heidelberg.* (2009) **25**. DOI: 10.1007/978-3-642-03885-3_94
13. Alberti KG, Zimmet PZ. **Definition, diagnosis and classification of diabetes mellitus and its complications. part 1: Diagnosis and classification of diabetes mellitus provisional report of a WHO consultation**. *Diabetic medicine: J Br Diabetic Assoc* (1998) **15**. DOI: 10.1002/(SICI)1096-9136(199807)15:7<539::AID-DIA668>3.0.CO;2-S
14. Zhou BF. **Predictive values of body mass index and waist circumference for risk factors of certain related diseases in Chinese adults–study on optimal cut-off points of body mass index and waist circumference in Chinese adults**. *Biomed Environ sciences: BES.* (2002) **15** 83-96. DOI: 10.1046/j.1440-6047.11.s8.9.x
15. Lin X, Chen Z, Huang H, Zhong J, Xu L. **Diabetic kidney disease progression is associated with decreased lower-limb muscle mass and increased visceral fat area in T2DM patients**. *Front Endocrinol* (2022) **13**. DOI: 10.3389/fendo.2022.1002118
16. Nagai M, Komiya H, Mori Y, Ohta T. **Developments in estimating visceral fat area from medical examination data**. *J Atheroscl Thrombosis* (2008) **15**. DOI: 10.5551/jat.E566
17. Hu C, Wang C, Zhang R, Ma X, Wang J, Lu J. **Variations in KCNQ1 are associated with type 2 diabetes and beta cell function in a Chinese population**. *Diabetologia* (2009) **52**. DOI: 10.1007/s00125-009-1335-6
18. Chen J, Zhang D, Xu M, Luo B, Cai Y, Chen S. **. Plasma glucose, insulin and c-peptide changes in normal subjects during steamed-bread meal test**. *Zhonghua yi xue za zhi* (1982) **62**
19. Yang W, Lu J, Weng J, Jia W, Ji L, Xiao J. **Prevalence of diabetes among men and women in China**. *New Engl J Med* (2010) **362**. DOI: 10.1056/NEJMoa0908292
20. Jimenez-Feltstrom J, Lundquist I, Obermuller S, Salehi A. **Insulin feedback actions: Complex effects involving isoforms of islet nitric oxide synthase**. *Regul Peptides* (2004) **122**. DOI: 10.1016/j.regpep.2004.06.004
21. Shen Y, Si Y, Lu J, Ma X, Zhang L, Mo Y. **Association between 1,5-anhydroglucitol and acute c peptide response to arginine among patients with type 2 diabetes**. *J Diabetes Res* (2020) **2020** 4243053. DOI: 10.1155/2020/4243053
22. Utzschneider KM, Prigeon RL, Tong J, Gerchman F, Carr DB, Zraika S. **Within-subject variability of measures of beta cell function derived from a 2 h OGTT: implications for research studies**. *Diabetologia* (2007) **50**. DOI: 10.1007/s00125-007-0819-5
23. Levy JC, Matthews DR, Hermans MP. **Correct homeostasis model assessment (HOMA) evaluation uses the computer program**. *Diabetes Care* (1998) **21**. DOI: 10.2337/diacare.21.12.2191
24. Gutt M, Davis CL, Spitzer SB, Llabre MM, Kumar M, Czarnecki EM. **Validation of the insulin sensitivity index (ISI(0,120)): Comparison with other measures**. *Diabetes Res Clin Practice* (2000) **47**. DOI: 10.1016/S0168-8227(99)00116-3
25. Kautzky-Willer A, Pacini G, Ludvik B, Schernthaner G, Prager R. **Beta-cell hypersecretion and not reduced hepatic insulin extraction is the main cause of hyperinsulinemia in obese nondiabetic subjects**. *Metabolism: Clin Experimental* (1992) **41**. DOI: 10.1016/0026-0495(92)90100-O
26. Bonadonna RC, Groop L, Kraemer N, Ferrannini E, Del Prato S, DeFronzo RA. **Obesity and insulin resistance in humans: A dose-response study**. *Metabolism: Clin Experimental* (1990) **39**. DOI: 10.1016/0026-0495(90)90002-T
27. Walton C, Godsland IF, Proudler AJ, Felton CV, Wynn V. **Effect of body mass index and fat distribution on insulin sensitivity, secretion, and clearance in nonobese healthy men**. *J Clin Endocrinol Metab* (1992) **75**. DOI: 10.1210/jcem.75.1.1619007
28. Macor C, Ruggeri A, Mazzonetto P, Federspil G, Cobelli C, Vettor R. **Visceral adipose tissue impairs insulin secretion and insulin sensitivity but not energy expenditure in obesity**. *Metabolism: Clin Experimental* (1997) **46**. DOI: 10.1016/S0026-0495(97)90288-2
29. Larsson H, Ahrén B. **Glucose-dependent arginine stimulation test for characterization of islet function: Studies on reproducibility and priming effect of arginine**. *Diabetologia* (1998) **41**. DOI: 10.1007/s001250050986
30. Meyer C, Dostou J, Nadkarni V, Gerich J. **Effects of physiological hyperinsulinemia on systemic, renal, and hepatic substrate metabolism**. *Am J Physiol* (1998) **275**. DOI: 10.1152/ajprenal.1998.275.6.F915
31. Soonthornpun S, Setasuban W, Thamprasit A, Chayanunnukul W, Rattarasarn C, Geater A. **Novel insulin sensitivity index derived from oral glucose tolerance test**. *J Clin Endocrinol Metab* (2003) **88**. DOI: 10.1210/jc.2002-021127
32. Després JP, Lemieux I. **Abdominal obesity and metabolic syndrome**. *Nature* (2006) **444**. DOI: 10.1038/nature05488
33. Frayn KN. **Adipose tissue as a buffer for daily lipid flux**. *Diabetologia* (2002) **45**. DOI: 10.1007/s00125-002-0873-y
34. Unger RH. **Lipid overload and overflow: Metabolic trauma and the metabolic syndrome**. *Trends Endocrinol Metabolism: TEM* (2003) **14** 398-403. DOI: 10.1016/j.tem.2003.09.008
35. Borst SE, Conover CF, Bagby GJ. **Association of resistin with visceral fat and muscle insulin resistance**. *Cytokine* (2005) **32** 39-44. DOI: 10.1016/j.cyto.2005.07.008
36. Shankar SS, Vella A, Raymond RH, Staten MA, Calle RA, Bergman RN. **Standardized mixed-meal tolerance and arginine stimulation tests provide reproducible and complementary measures of β-cell function: Results from the foundation for the national institutes of health biomarkers consortium investigative series**. *Diabetes Care* (2016) **39**. DOI: 10.2337/dc15-0931
37. Miyawaki T, Hirata M, Moriyama K, Sasaki Y, Aono H, Saito N. **Metabolic syndrome in Japanese diagnosed with visceral fat measurement by computed tomography**. *Proc Japan Academy Ser B* (2005) **81**. DOI: 10.2183/pjab.81.471
38. Hayashi T, Boyko EJ, McNeely MJ, Leonetti DL, Kahn SE, Fujimoto WY. **Minimum waist and visceral fat values for identifying Japanese americans at risk for the metabolic syndrome**. *Diabetes Care* (2007) **30**. DOI: 10.2337/dc06-0739
|
---
title: Antagonism between Prdm16 and Smad4 specifies the trajectory and progression
of pancreatic cancer
authors:
- Eric Hurwitz
- Parash Parajuli
- Seval Ozkan
- Celine Prunier
- Thien Ly Nguyen
- Deanna Campbell
- Creighton Friend
- Allyn Austin Bryan
- Ting-Xuan Lu
- Steven Christopher Smith
- Mohammed Shawkat Razzaque
- Keli Xu
- Azeddine Atfi
journal: The Journal of Cell Biology
year: 2023
pmcid: PMC9999015
doi: 10.1083/jcb.202203036
license: CC BY 4.0
---
# Antagonism between Prdm16 and Smad4 specifies the trajectory and progression of pancreatic cancer
## Abstract
Pancreatic ductal adenocarcinoma (PDAC) is a deadly disease whose molecular etiology remains mostly enigmatic. By discovering an antagonistic relationship between the tumor suppressors Prdm16 and Smad4 in PDAC, this study paves the way for innovative frameworks with potential therapeutic implications.
The transcription factor Prdm16 functions as a potent suppressor of transforming growth factor-beta (TGF-β) signaling, whose inactivation is deemed essential to the progression of pancreatic ductal adenocarcinoma (PDAC). Using the KrasG12D-based mouse model of human PDAC, we surprisingly found that ablating Prdm16 did not block but instead accelerated PDAC formation and progression, suggesting that Prdm16 might function as a tumor suppressor in this malignancy. *Subsequent* genetic experiments showed that ablating Prdm16 along with Smad4 resulted in a shift from a well-differentiated and confined neoplasm to a highly aggressive and metastatic disease, which was associated with a striking deviation in the trajectory of the premalignant lesions. Mechanistically, we found that Smad4 interacted with and recruited Prdm16 to repress its own expression, therefore pinpointing a model in which Prdm16 functions downstream of Smad4 to constrain the PDAC malignant phenotype. Collectively, these findings unveil an unprecedented antagonistic interaction between the tumor suppressors Smad4 and Prdm16 that functions to restrict PDAC progression and metastasis.
## Introduction
Pancreatic ductal adenocarcinoma (PDAC) is the most aggressive type of pancreatic cancer, currently ranked as the fourth leading cause of cancer-related deaths in the United States (Connor and Gallinger, 2021; Hidalgo, 2010). Most of PDAC patients present with both locally invasive tumors and widespread metastasis, thus rendering ineffective the resection of the primary tumor as well as the applicability of the dismal therapeutic options available (Hidalgo, 2010; Stathis and Moore, 2010). Consequently, the outcome of PDAC patients remains extremely poor, with an overall 5-yr survival rate of less than $11\%$.
PDAC tumors emerge through three types of distinct precursor lesions called pancreatic intraepithelial neoplasia (PanIN), intraductal papillary mucinous neoplasia (IPMN), and mucinous cystic neoplasia (MCN), respectively (Connor and Gallinger, 2021; Yonezawa et al., 2008). These early-stage lesions harbor various genetic alterations, the earliest and most pervasive of which are activating mutations in KRAS, occurring in ∼$90\%$ of PDAC tumors (Hayashi et al., 2021). The current model posits that mutational activation of KRAS represents an essential initiating event, whereas subsequent accumulation of inactivating mutations in the tumor suppressor genes p16INK4a, SMAD4, and TP53 is necessary for PDAC to progress and metastasize (Hayashi et al., 2021; Iacobuzio-Donahue, 2012). Significant efforts have been made over the past two decades to create genetically engineered mouse models (GEMMs) that faithfully recapitulate the prominent features of human PDAC. For instance, pancreas-specific expression of KrasG12D in mice is sufficient to initiate PanINs, which occasionally progress into invasive PDAC following a long latency period, supporting the general notion that oncogenic activation of KRAS represents the main initiating genetic event in PDAC (Buscail et al., 2020; Hingorani et al., 2005; Tuveson et al., 2004; Westphalen and Olive, 2012). Concomitant expression of KrasG12D and deletion of any of the three cardinal tumor suppressors, e.g., Trp53, p16Ink4a, Smad4, accelerate PDAC progression, though the nature and final outcome of the tumors might differ. Indeed, while mice with the combined expression of KrasG12D and deletion of Trp53 (KPC) or p16Ink4a (KIC) develop PanINs that progress very rapidly to highly aggressive and metastatic PDAC, mice with the combined expression of KrasG12D and deletion of Smad4 (KSC) develop mostly IPMNs, which also progress to invasive PDAC, but the terminal disease develops with a much slower onset and manifests an attenuated metastatic phenotype (Bardeesy et al., 2006a; Bardeesy et al., 2006b; Hingorani et al., 2005; Izeradjene et al., 2007). Other examples of PDAC GEMMs include KTβC mice, which harbor KrasG12D and deletion of the transforming growth factor-beta (TGF-β) type II receptor (TβRII) gene, the latter being inactivated by mutations or deletions in $4\%$ of PDAC (Iacobuzio-Donahue, 2012; Ijichi et al., 2006).
TGF-β signaling regulates a wide array of biological processes vital for normal cell growth, function, and homeostasis (David and Massagué, 2018; Massagué, 2008). TGF-β initiates signaling by inducing the assembly of a receptor complex composed of two types of transmembrane serine/threonine kinases called TβRI and TβRII. In that complex, the constitutive kinase of TβRII phosphorylates and activates the kinase activity of TβRI, which then propagates the signal to the nucleus through phosphorylation of Smad2 and Smad3 (David and Massagué, 2018; Feng and Derynck, 2005; Massagué et al., 2005). Once phosphorylated, Smad2 or Smad3 associates with Smad4, and the two complexes accumulate in the nucleus to regulate the expression of TGF-β target genes through cooperative interactions with transcriptional cofactors or corepressors (David and Massagué, 2018; Feng and Derynck, 2005; Massagué, 2008; Massagué et al., 2005).
Because of the widespread roles of TGF-β signaling in cellular functions, there must be multiple levels of positive and negative regulations to fine-tune initiation, magnitude, or termination of the response depending on the cell type or physiological context. One example of the mechanisms that limit TGF-β signaling involves the transcription factor PR domain containing 16 (Prdm16). Upon accumulation in the nucleus, the Smad3/Smad4 complex associates with the general transcriptional co-activators CBP and p300 to activate transcription of TGF-β target genes (David and Massagué, 2018; Feng and Derynck, 2005; Massagué, 2008; Massagué et al., 2005). Conversely, the Smad complex can also associate with Prdm16 and its partner c-Ski, which leads to the recruitment of general transcriptional corepressor complexes containing histone deacetylases and concomitant displacement of CBP and p300, thereby resulting in transcriptional repression (Takahata et al., 2009).
In addition to its function as a suppressor of TGF-β signaling, Prdm16 has been shown to play key roles in a number of biological processes, including differentiation of brown fat and specification of hematopoietic and neuronal stem cell fate (Chi and Cohen, 2016; Seale et al., 2007; Shimada et al., 2017). Prdm16 possesses a methyltransferase activity that catalyzes the methylation of Lysine-9 on histone-3 (H3K9), a mark associated with heterochromatin formation and gene expression (Jambhekar et al., 2019; Pinheiro et al., 2012). Recently, Prdm16 loss-of-function has been shown to play an instrumental role in leukemia driven by the MLL fusion oncoprotein (Zhou et al., 2016). Because the MLL gene encodes a histone-3 Lysine-4 (H3K4) methyltransferase that is critical in promoting gene expression during hematopoiesis (Xue et al., 2019), it has been postulated that Prdm16 might suppress leukemia pathogenesis owing to its ability to drive heterochromatin formation (Pinheiro et al., 2012; Zhou et al., 2016). At present, whether Prdm16 has any role in cancer pathogenesis and progression that is linked to its function in TGF-β signaling is still unknown. Here, we combined several orthogonal approaches and GEMMs to demonstrate that Prdm16 functions downstream of Smad4 to suppress PDAC progression and metastasis. As such, our findings unveil a previously uncharacterized mechanism that orchestrates Prdm16 tumor-suppressive function, and further shed new insights into the molecular etiology of PDAC, a fatal disease for which no effective therapeutics are currently available.
## Transient expression of Prdm16 during PDAC progression
To explore the possible involvement of Prdm16 in PDAC, we conducted Kaplan-*Meier analysis* using The Cancer Genome Atlas (TCGA) dataset. As shown in Fig. 1 A, low PRDM16 expression is associated with poor survival, providing an initial hint that Prdm16 might function as a tumor suppressor in PDAC. To substantiate this finding, we analyzed Prdm16 expression by immunohistochemistry (IHC) using large human tissue microarrays (TMAs) comprising samples with tumor lesions at various stages (e.g., PanIN1, PanIN2, PanIN3, PDAC) and normal tissues. Using a highly specific antibody to Prdm16 (see Fig. S2 A), we detected Prdm16 expression in both cancerous lesions and stromal areas (Fig. 1 B). Interestingly, Prdm16 expression appeared to fluctuate significantly during PDAC progression, commencing with a relatively low level in normal tissue, then rising in early PanINs, and finally declining to the background level in invasive PDAC (Fig. 1 B). Although this finding fits well with the notion that Prdm16 expression might be downregulated because of the accumulation of late genetic or epigenetic alterations, it did not shed light on the mechanisms leading to its transient expression during PDAC progression. To address this issue rigorously, we sought to utilize GEMMs that faithfully recapitulate the human PDAC in a uniform genetic background (Bardeesy et al., 2006a; Bardeesy et al., 2006b; Hingorani et al., 2005; Izeradjene et al., 2007; Tuveson et al., 2004). We initially utilized mice with pancreas-specific expression of KrasG12D alone (KC) and detected transient expression of the Prdm16 protein during PDAC progression, similar to what was observed in human PDAC, being relatively high in PanINs and very modest to low in normal tissue and invasive PDAC (Fig. 1 C). Confirmation of this result was obtained by comparative qRT-PCR experiments using cohorts of KC mice at the age of 3 mo when they experience mostly PanINs and 10 mo when they display visible signs of terminal PDAC (Fig. 1 D; Parajuli et al., 2020; Parajuli et al., 2019). To understand this phenomenon more deeply, we generated mice with KrasG12D together with deletion of Trp53 (KPC), p16Ink4a (KIC), or Smad4 (KSC). Noteworthy, we found that KSC mice had a longer survival rate than KIC and KPC mice, while the two latter had almost similar survival (Fig. S1 A). With regard to Prdm16 expression, we found that KIC and KPC mice behaved similarly to KC mice (Fig. S1, B and C), suggesting that transient expression of Prdm16 might take place even under the presence of the most common and aggressive genetic alterations that facilitate PDAC progression (Hayashi et al., 2021; Iacobuzio-Donahue, 2012). But most appealing was the fact that Prdm16 expression in KSC mice did not follow this transient pattern of Prdm16 expression, increasing markedly in IPMN lesions but thereafter remaining constant in PDAC lesions (Fig. 1 E), suggesting that Smad4 might influence Prdm16 expression during the progression from IPMN to full-blown PDAC. Co-immunofluorescence assays using anti-Prdm16 antibody together with antibodies to E-cadherin (epithelial marker) or vimentin (mesenchymal marker) showed that Prdm16 expression remained very high in E-cadherin + cells as compared to vimentin + cells (Fig. S1 D). Consistent with these findings, we found that patients with low expression of Prdm16 had the worst survival if they carry SMAD4 mutations (Fig. S1 E). Moreover, interrogating the TCGA dataset revealed that samples with deleterious genetic alterations in SMAD4 display higher expression of PRDM16 as compared to samples with wild-type SMAD4 (Fig. 1 F). As such, these data hint at the existence of an antagonistic association between Smad4 and Prdm16 during PDAC progression; we will return to this notion later.
**Figure 1.:** *Transient expression of Prdm16 during PDAC progression. (A) Kaplan-Meier survival of PDAC patients based on high versus low PRDM16 expression was conducted using the TCGA dataset. Statistical power was assessed by log-rank test for significance. (B) Prdm16 protein expression was analyzed by IHC using human PDAC TMAs containing both normal tissues and PanIN/PDAC lesions (n = 152). Representative pictures of normal, PanIN, and PDAC areas are shown. Scale bars: 50 μm (left). Relative Prdm16 expression in normal, PanIN, and PDAC areas (right). Data are expressed as mean ± SEM. (C) FFPE pancreatic sections from 4-mo-old control and KC mice (n = 31 to 33) were stained with H&E or immunostained with anti-Prdm16 antibody and subjected to IHC. Representative pictures of normal, PanIN, and PDAC areas are shown. Scale bars: 50 μm (left). Relative Prdm16 expression in normal, PanIN, and PDAC areas are shown (right). Data are expressed as mean ± SEM. (D) Expression of Prdm16 mRNA in pancreas from 3-mo-old control or KC mice (n = 6) with PanIN or 10-mo-old KC mice with terminal PDAC was analyzed by qRT-PCR. Data are expressed as mean ± SEM. (E) FFPE pancreatic sections from 4-mo-old control and KSC mice (n = 31–45) were stained with H&E or immunostained with anti-Prdm16 antibody and subjected to IHC. Representative pictures of normal, IPMN and PDAC areas are shown. Scale bars: 50 μm (left). Relative Prdm16 expression in normal, IPMN and PDAC areas are shown (right). Data are expressed as mean ± SEM. (F) Relative expression of PRDM16 in human samples with wild-type or mutated SMAD4 was conducted using the TCGA dataset. Data are presented as a violin plot. Statistical power in B–F was assessed by a two-tailed, unpaired Mann–Whitney test.* **Figure S1.:** *Transient expression of Prdm16 during PDAC progression. (A) Kaplan-Meier survival analysis of control, KSC, KIC, and KPC mice (n = 31–39). Statistical power was assessed by a log-rank test for significance. (B) FFPE pancreatic sections from 4-mo-old control and KIC mice (n = 38–39) were stained with H&E or immunostained with anti-Prdm16 antibody and subjected to IHC. Representative pictures of normal, PanIN and PDAC areas are shown. Scale bars: 50 μm (left). Relative Prdm16 expression in normal tissue and PanIN or PDAC lesions are shown (right). Data are expressed as mean ± SEM. (C) FFPE pancreatic sections from 3-mo-old control and KPC mice (n = 31–39) were stained with H&E or immunostained with anti-Prdm16 antibody and subjected to IHC. Representative pictures of normal, PanIN, and PDAC areas are shown. Scale bars: 50 μm (left). Relative Prdm16 expression in normal tissue and PanIN or PDAC lesions are shown (right). Data are expressed as mean ± SEM. (D) FFPE pancreatic sections from control and KSC mice (n = 31–45) were subjected to co-IF using antibodies to Prdm16 and E-cadherin or vimentin. Representative pictures of normal tissue and IPMN or PDAC areas are shown. Scale bars: 50 μm. (E) Kaplan-Meier survival analysis of patients with wild-type or mutated SMAD4 based on high versus low PRDM16 expression was conducted using the TCGA dataset. Statistical power was assessed by log-rank test for significance. Statistical power in B and C were assessed by a two-tailed, unpaired Mann–Whitney test.*
## Prdm16 accelerates KrasG12D-driven PDAC
The aforementioned data prompted us to investigate whether Prdm16 could contribute to PDAC initiation, progression, or both. To do so, we generated mice with pancreas-specific deletion of Prdm16 (Prdm16KO) by crossing mice bearing a floxed allele of Prdm16 with Pdx1-Cre mice, which express Cre recombinase in all pancreatic progenitor cells that give rise to ductal, acinar, and islets compartments very early (E8.5) during development (Gu et al., 2003). Prdm16KO mice were born with the normal Mendelian frequency, develop normally without any signs of anatomic abnormalities, and were fertile. Effective deletion of Prdm16 in the pancreatic epithelium was confirmed by RT-PCR and IHC (Fig. 2 A and Fig. S2 A). To investigate whether Prdm16 deficiency could affect pancreas histology or function, we conducted a comprehensive analysis of pancreatic sections either by hematoxylin and eosin (H&E) staining, IHC or immunofluorescence (IF) encompassing all major tissue compartments, including duct (cytokeratin 19, CK19), acini (amylase), stroma (α-SMA), and islet (insulin, glucagon, chromogranin-A). We were not able to detect any noticeable changes in all three compartments irrespective of the age of mice analyzed (Fig. 2, B–D; and Fig. S2, B and C). Congruently, there was also no difference in fasting blood glucose between wild-type and Prdm16KO mice (Fig. S2 D). Thus, inactivation of Prdm16 throughout embryonic development and postnatal life was insufficient to perturb pancreas homeostasis or drive sporadic pancreatic cancers.
**Figure 2.:** *Prdm16 inactivation did not affect pancreas histology or function. (A) Prdm16 mRNA expression in 3-mo-old control and Prdm16KO mice was measured by qRT-PCR (n = 6). Data are expressed as mean ± SEM, and statistical power was assessed by a two-tailed, unpaired t test. (B) FFPE pancreatic sections from control or Prdm16KO mice (n = 8) at 6 or 20 wk-old were stained with H&E. Scale bars: 200 μm for “whole” pictures and 50 μm for all other pictures. (C) FFPE pancreatic sections from 15-wk-old control and Prdm16KO mice (n = 8 to 31) were immunostained with anti-CK19 or anti-Chromogranin A antibody and subjected to IHC. Scale bars: 50 μm. (D) FFPE pancreatic sections from 15-wk-old control and Prdm16KO mice (n = 8) were immunoreacted with antibodies to amylase or α-SMA before being subjected to immunofluorescence. Scale bars: 50 μm.* **Figure S2.:** *Prdm16KO mice display normal pancreatic endocrine function. (A) FFPE pancreatic sections from 15-wk-old control and Prdm16KO mice (n = 8–31) were immunostained with anti-Prdm16 antibody and subjected to IHC. Scale bars: 50 μm. (B) FFPE pancreatic sections from 15-wk-old control and Prdm16KO mice (n = 8–31) were immunoreacted with antibodies to insulin or glucagon and subjected to IHC. Scale bars: 50 μm. (C) FFPE pancreatic sections from 15-wk-old control and Prdm16KO mice were subjected to IHC using antibodies to amylase or α-SMA. Scale bars: 50 μm. (D) Blood glucose of 15-wk-old control or Prdm16KO mice (n = 8). Statistical power was assessed by a two-tailed, unpaired Mann–Whitney test.*
Next, we sought to investigate whether Prdm16 could influence PDAC progression initiated through activation of Kras signaling. The salient genetic features of PDAC originate with the near-ubiquitous gain of function mutations in KRAS in their incipient stage. However, progression to invasive PDAC in KrasG12D-bearing mice has proved to be either a protracted or unachieved process, as a small fraction of mice succumb directly to PDAC following a very long latency period (Hingorani et al., 2005; Parajuli et al., 2020; Parajuli et al., 2019; Tuveson et al., 2004). It is widely believed that the acquisition of secondary mutations in certain tumor suppressors can endow transformed cells with the growth advantage needed for disease progression. For instance, combining KrasG12D with deletion of Smad4 or TβRII has been shown to accelerate the progression of PDAC, which was thought to be conferred through disruption of TGF-β cytostatic signaling (Bardeesy et al., 2006b; Ijichi et al., 2006; Izeradjene et al., 2007). Given its role as an inhibitor of Smad signaling, we surmised that Prdm16 inactivation might suppress PDAC development and/or progression owing to the de-repression of TGF-β/Smad signaling. To probe this possibility, we generated mice harboring KrasG12D alone (KC) or in combination with conditional deletion of both alleles of Prdm16 (KPrC) and conducted comparative studies to analyze their PDAC phenotypes. Consistent with previous studies (Parajuli et al., 2020; Parajuli et al., 2019; Tuveson et al., 2004), KC mice maintained uniformly good health until around the age of 20 wk, and thereafter a fraction of mice became suddenly morbid and succumbed within days to an aggressive PDAC. Contrary to our prediction, combining Prdm16 deletion with KrasG12D instead resulted in a marked acceleration of PDAC. Kaplan-*Meyer analysis* showed a significant decrease in the median survival of KPrC mice as compared to KC mice (Fig. 3 A). During an observation period of 6 mo, $70\%$ of KPrC mice succumbed to PDAC, whereas more than $76\%$ of KC mice survived and remained free of invasive PDAC (Fig. 3 A). To confirm this finding, we conducted histopathological analysis with pancreatic sections from KPrC and KC mice of the same age that showed either relatively healthy appearance or signs of morbidity characteristic of invasive PDAC at the time of necropsy. At early stages of tumorigenesis, KPrC mice displayed a significant increase in PanIN lesions compared to KC mice, as assessed by H&E and IHC using anti-CK19 antibody (Fig. 3 B). A similar conclusion could be drawn while analyzing another ductal marker, MUC5AC, either by IHC or Alcian blue staining (Fig. S3). KPrC and KC mice with invasive PDAC also showed clear difference in both tumor architecture and reactivity to the anti-CK19 and anti-Mu5AC antibodies as well as to Alcian blue (Fig. 3 B and Fig. S3). Moreover, IHC analysis using anti-α-SMA antibody showed more extensive stroma both within and outside PDAC lesions in KPrC mice relative to KC mice (Fig. S3). An automatic-guided quantification confirmed the increase in the surface areas of PanIN and PDAC lesions in KPrC mice as compared to KC mice (Fig. 3 B). Thus, Prdm16 inactivation appeared to accelerate PDAC once it has been initiated through activation of oncogenic KrasG12D signaling.
**Figure 3.:** *Prdm16 inactivation accelerates KrasG12D-driven PDAC. (A) Kaplan-Meier survival of control, Prdm16KO, KC and KPrC mice. A two-color line (black and blue bold) was used to differentiate between control and Prdm16KO mice. Statistical power was assessed by log-rank test for significance (left). The percentage of survival at the end of the observation period (right). (B) FFPE pancreatic sections from 4-m-old control, Prdm16KO, KC, and KPrC mice (n = 8–33) were stained with H&E or immunostained with antibodies to CK19 and subjected to IHC. Representative pictures are shown (top). Scale bars: 50 μm. Relative PanIN and PDAC surface areas, number of PanIN and PDAC lesions and CK19 intensity (bottom) are shown. Data are expressed as mean ± SEM, and statistical power was assessed by a two-tailed, Mann–Whitney test.* **Figure S3.:** *Prdm16 ablation accelerates PDAC driven by KrasG12D. FFPE pancreatic sections from 4-mo-old control, Prdm16KO, KC, and KPrC mice (n = 8–33) were stained with H&E or Alcian Blue or immunostained with antibodies to Mu5AC or α-SMA and subjected to IHC. Representative pictures are shown. Scale bars: 50 μm (top). Relative intensity of Muc5AC, α-SMA, and Alcian blue staining in areas of PanIN and PDAC lesions are shown (n = 13–33). Data are expressed as mean ± SEM (bottom), and statistical power was assessed by a two-tailed, unpaired Mann–Whitney test.*
## Requirement of Prdm16 for IPMN-to-PDAC progression
Given the inverse association between Smad4 and Prdm16 that we noticed earlier during PDAC progression (Fig. 1, E and F), we sought to extend our genetic approaches to explore whether Prdm16 could play a role, if any, in PDAC that depends on its function in TGF-β/Smad signaling. Accordingly, we generated mice with the combined deletion of Prdm16 and Smad4 in a KrasG12D background (KSPrC). KPrC, KSC, KC, and wild-type mice were used as controls. KSPrC mice were born with Mendelian frequencies, and no phenotypic differences between KSPrC and KSC mice were observed. Strikingly, however, the vast majority of KSPrC mice became stunted and morbid in appearance within 2 to 3 wk of weaning, and only $25\%$ of them survived beyond 3 mo (Fig. 4 A). During this period, most of KPrC and KSC mice (84 and $90\%$, respectively) did not develop or succumb to PDAC. To elucidate the mechanism causing the acceleration of PDAC in KSPrC mice, we conducted histopathological analyses to study different stages of PDAC from the premalignant lesions to invasive adenocarcinomas. We found that KSC pancreas displayed predominantly macroscopic cystic lesions reminiscent of IPMN, as evidenced by the overall architecture as well as the high reactivity to the anti-Muc5AC and anti-CK19 antibodies as well as Alcian blue (Fig. 4 B and Fig. S4 A). In contrast, KSPrC pancreas displayed none to very few IPMN lesions (Fig. 4 B and Fig. S4 A). At the stage of full PDAC, KSPrC tumors were poorly differentiated adenocarcinomas, characterized by loss of the epithelial marker E-cadherin and acquisition of the mesenchymal marker vimentin (Fig. S3 B), which could be due either to increased accumulation of cancer associated fibroblasts or epithelial to mesenchymal transition (EMT), the latter being a general hallmark of metastasis (Pei et al., 2019). In marked contrast, KSC tumors were well differentiated with little or no change in E-cadherin or vimentin expression (Fig. S4 B), in line with previous studies that KSC mice are resistant to metastasis (Bardeesy et al., 2006b; Izeradjene et al., 2007; Whittle et al., 2015). As concomitant inactivation of Prdm16 appeared to shift the evolution of the IPMN-to-PDAC progression sequence toward the PanIN-to-PDAC progression sequence, it is tempting to speculate that Prdm16 might function at the stage of early preneoplastic lesions to influence PDAC development and progression. In support of this notion, deleting PRDM16 in two fully-transformed human PDAC cell lines (i.e., PANC-1, sufficient for SMAD4 and BxPC-3, deficient for SMAD4) did not affect their proliferative or invasive behaviors, as gauged by a combination of in vivo and in vitro assays (Fig. S4, C–F).
**Figure 4.:** *Concomitant inactivation of Prdm16 and Smad4 shifts the progression trajectory of PDAC. (A) Kaplan-Meier survival analysis of control, KC, KPrC, KSC, and KSPrC mice (n = 13–45). Statistical power was assessed by a log-rank test for significance (left). The percentage of survival at the end of the observation period (right). (B) FFPE pancreatic sections from 4-m-old control, KC, KPrC, KSC, and KSPrC mice (n = 13–45) were stained with H&E or Alcian blue or immunostained with antibodies to CK19 or Muc5AC and subjected to IHC. Representative pictures are shown. Scale bars: 50 μm. (C) FFPE lung sections from 4-mo-old control, KPrC, KSC and KSPrC mice (n = 13–45) were stained with H&E or immunostained with anti-CK19 antibody. Metastatic lesions are highlighted by blue dot-circles. Representative pictures are shown (left). Scale bars: 50 μm. Relative CK19 intensity from lung sections are shown (right). Data are expressed as mean ± SEM, and statistical power was assessed by a two-tailed Mann–Whitney test.* **Figure S4.:** *Prdm16 inactivation in KSC mice resulted in acceleration of PDAC. (A) FFPE pancreatic sections from 4-mo-old control, KC, KPrC, KSC, and KSPrC mice (n = 13–45) were stained with H&E or Alcian blue or immunostained with antibodies to CK19 or Muc5AC and subjected to IHC. Relative abundance of IPMN lesions (top left) or intensity of CK19 (top right), Muc5AC (bottom left), or Alcian blue (bottom right) are shown. Data are expressed as mean ± SEM, and statistical power was assessed by a two-tailed, unpaired Mann–Whitney test. (B) FFPE pancreatic sections from control, KPrC, KSC, and KSPrC mice (n = 13–45) were subjected to IF using antibodies to E-cadherin or vimentin. Representative pictures are shown. Scale bars: 50 μm. (C) Pictures of tumors harvested from NSG mice injected with isogenic PANC-1 and BxPC-3 cell lines stably expressing control or PRDM16 gRNA (n = 3). (D) FFPE liver and lung sections from NSG mice injected with isogenic PANC-1 and BxPC-3 cell lines stably expressing control or PRDM16 gRNA were stained with H&E (n = 3). Representative pictures are shown. Scale bars: 50 μm. (E) Representative pictures of soft-agar colonies formed by isogenic PANC-1 cell lines stably expressing control or PRDM16 gRNA. BxPC-3 stably expressing control or PRDM16 gRNA did not form colonies. (F) Cell proliferation assay of isogenic PANC-1 (day 3) and BxPC-3 (day 6) cell lines stably expressing control or PRDM16 gRNA. The fold increase in cell number at the end of the experiment relative to the seeding density is shown (n = 3). Data are expressed as mean ± SEM and statistical power was assessed by a two-tailed, unpaired Mann–Whitney test.*
The poor prognosis for human PDAC is mainly due to quasi-inevitable metastasis affecting the liver and lung at the time of diagnosis (Connor and Gallinger, 2021; Hidalgo, 2010). Due to the severity of PDAC in KSPrC mice, we wondered whether concomitant deletion of Prdm16 could confer metastatic ability to the otherwise non-metastasizing PDAC tumors that typically develop in KSC mice (Bardeesy et al., 2006b; Izeradjene et al., 2007; Whittle et al., 2015). Indeed, we consistently observed the presence of metastatic lesions in the lung in all KSPrC mice that developed invasive tumors but survived until necropsy (Fig. 4 C). In contrast, no metastatic lesions were detected in KSC mice even with terminal PDAC (Fig. 4 C), as previously described (Bardeesy et al., 2006b; Izeradjene et al., 2007; Whittle et al., 2015). Confirmation of these results was obtained by IHC using an antibody to the PDAC marker CK19 (Fig. 4 C). Collectively, these data demonstrate that concomitant inactivation of Prdm16 was sufficient to confer metastatic properties on non-metastatic KSC tumors, a phenomenon that is associated with a shift from the IPMN-to-PDAC phenotype to the PanIN-to-PDAC phenotype.
## Repression of Prdm16 expression by Smad4
To investigate the molecular mechanisms by which Prdm16 controls PDAC progression and metastasis in the context of a Smad4 null background, we took advantage of our earlier IHC analysis showing that Smad4 deficiency in KSC mice was associated with a persistent de-repression of Prdm16 during the progression from IPMN to PDAC (Fig. 1 E). We surmised that Smad4 might function either directly or indirectly to repress Prdm16 expression, which in turn impacts the progression trajectory of PDAC. We initially conducted qRT-PCR experiments using KSC mice, and found that the increase in Prdm16 expression was mediated at least via gene expression (Fig. 5 A). Because Smad4 functions as an essential component of TGF-β signaling (David and Massagué, 2018; Feng and Derynck, 2005; Massagué, 2008), we next wondered whether activation of TGF-β signaling could repress Prdm16 expression, as does Smad4. To our surprise, treating mouse PDAC cells KPC1 or human PDAC cells PANC-1 with TGF-β1 instead elicited a marked increase in Prdm16 expression (Fig. 5, B–D). As a specificity control, TGF-β1 treatment failed to induce Prdm16 expression in the human PDAC cell line MIA-PaCa-2 (Fig. 5 E), which lacks a functional TGF-β receptor (Freeman et al., 1995). To determine whether the effect of TGF-β1 is mediated via Smad4, we conducted comparative experiments using PANC-1 cells deleted of SMAD4 by CRISPR/CAS9. We found that ablating SMAD4 resulted in a marked increase in the steady-state expression of Prdm16 mRNA and protein (Fig. 5 F, see also Fig. 6 F), confirming the ability of endogenous Smad4 to repress PRDM16 expression in human cells. Intriguingly, challenging cells with TGF-β1 did not further increase Prdm16 expression in cells deleted of SMAD4 as compared to cells expressing the control gRNA (Fig. 5 F; see also Fig. 6 F), implying that SMAD4 inactivation is sufficient to mimic the effects of TGF-β1 stimulation.
**Figure 5.:** *Smad4 represses Prdm16 expression. (A) Expression of Prdm16 (left) or Smad4 (right) in pancreas from 4-mo-old control and KSC mice (n = 6) was analyzed by qRT-PCR. (B) Expression of Prdm16 mRNA in KPC1 cells cultured in the presence or absence of TGF-β1 for various times was analyzed by qRT-PCR (n = 6). (C) Expression of PRDM16 mRNA in PANC-1 cells treated with TGF-β1 for various times was analyzed by qRT-PCR (n = 6). (D) Expression of Prdm16 protein in PANC-1 cells treated with TGF-β1 for various times was analyzed by immunoblotting. (E) Expression of PRDM16 mRNA in MIA-PaCa-2 cells cultured in the presence or absence of TGF-β1 for various times was analyzed by qRT-PCR (n = 6). (F) Expression of PRDM16 mRNA in isogenic PANC-1 cell lines stably transduced with control or SMAD4 gRNA lentiviruses and cultured in the presence or absence of TGF-β1 was analyzed by qRT-PCR (n = 6). Data in A–C, E, and F are expressed as mean ± SEM, and statistical power was assessed by a two-tailed, unpaired t test. Source data are available for this figure: SourceData F5.* **Figure 6.:** *Smad4 interacts with Prdm16 on the PRDM16 promoter to repress its own expression. (A) Pancreatic chromatin from PANC-1 or BxPC-3 cells (n = 6) cultured in the presence or absence of TGF-β1 was analyzed for the binding of Smad4 to the PRDM16 or JUNB promoter by ChIP and agarose gel (left) and qPCR (right). (B) Pancreatic chromatin from PANC-1 cells (n = 6) cultured in the presence or absence of TGF-β1 was analyzed for the binding of Smad2 and Smad3 to the PRDM16 promoter by ChIP and agarose gel (left) and qPCR (right). (C) PANC-1 expressing control or SMAD2/3 gRNAs were cultured in the presence or absence of TGF-β1 for 48 h and analyzed for the expression of Prdm16 and Smad2/Smad3 by direct immunoblotting. (D) PANC-1 cells were transfected with the CAGA9-Lux gene reporter and increasing amounts of pcDNA3.1-Prdm16. 24 h after transfection, cells were treated with TGF-β1 for 16 h and then assessed for luciferase activity and normalized. (E) Pancreatic chromatin from PANC-1 or BxPC-3 cells (n = 6) cultured in the presence or absence of TGF-β1 was analyzed for the binding of Prdm16 to the PRDM16 promoter by ChIP and agarose gel (left) and qPCR (right). (F) PANC-1 expressing control or SMAD4 gRNA were treated with TGF-β1 for 48 h and then analyzed for the interaction of Prdm16 with Smad4 by co-immunoprecipitation (IP) followed by immunoblotting (WB). Expression of Prdm16 was also analyzed by direct immunoblotting. (G) PANC-1 cells were transfected with the wild-type (left) or mutated Prdm16-Lux (right) reporter together with empty vector, pcDNA3.1-Prdm16 or pCMV5-HA-Smad4. 48 h after transfection, cells were assessed for luciferase activity and normalized (n = 6). (H) BxPC-3 cells were transfected with the wild-type Prdm16-Lux reporter together with the indicated combinations of empty vector (EV), pcDNA3.1-Prdm16, and pCMV5-HA-Smad4. 48 h after transfection, cells were assessed for luciferase activity and normalized (n = 6). Data in A, B, D, E, G, and H are expressed as mean ± SEM, and statistical power was assessed by a two-tailed, unpaired t test. Source data are available for this figure: SourceData F6.*
Previous studies have shown that Smad proteins can stimulate or repress expression of TGF-β responsive genes through direct binding to their promoter (David and Massagué, 2018; Feng and Derynck, 2005; Massagué, 2008). In addition, a substantial fraction of Smad4 has been shown to localize in the nucleus in the absence TGF-β stimulation, but the physiopathological significance of this phenomenon remains unknown (Pierreux et al., 2000). Because SMAD4 ablation in PANC-1 cells was sufficient to recapitulate the stimulatory effects of TGF-β signaling on PRDM16 expression, we initially reasoned that Smad4 might bind to and repress the PRDM16 promoter at steady state, and that TGF-β signaling activation might displace Smad4 from the PRDM16 promoter. Accordingly, we conducted ChIP experiments, focusing on Smad conserved binding elements (SBE) within the PRDM16 promoter that we identified through an in-silico analysis. Using chromatin from PANC-1 cells, we detected a strong binding of Smad4 to the PRDM16 promoter at steady state (Fig. 6 A). This binding is specific, as there was no signal in the human PDAC cell line BxPC-3 (Fig. 6 A), which bears natural homozygous deletion of SMAD4 (Duda et al., 2003). Interestingly, treating PANC-1 cells with TGF-β1 had little or no effect on the binding of Smad4 to the PRDM16 promoter despite eliciting a strong activation of this pathway, as assessed by the increased binding of Smad4 to the promoter of JUNB (Fig. 6 A), a well-characterized TGF-β target gene (Sundqvist et al., 2018). This observation, together with our gene expression experiments, strongly suggests that TGF-β signaling might involve other players that act in partnership with Smad4 to repress Prdm16 expression. To explore this possibility, we conducted ChIP experiments using antibodies to Smad2 and Smad3, as both transcription factors are known to interact with Smad4 in response to TGF-β signaling (David and Massagué, 2018; Feng and Derynck, 2005; Massagué, 2008; Massagué et al., 2005). We detected a slight but significant increase in the binding to Smad2 to the PRDM16 promoter in PANC-1 cells upon stimulation with TGF-β1 (Fig. 7 B). In contrast, TGF-β1 stimulation induced a massive increase in the binding of Smad3 to the PRDM16 promoter (Fig. 6 B). Concomitant deletion of SMAD2 and SMAD3 in PANC-1 cells resulted in almost complete blockade in TGF-β-induced Prdm16 expression (Fig. 6 C). Although this finding provides a potential mechanism by which TGF-β signaling could induce Prdm16 expression, it failed to explain why deletion of Smad4 in KSC mice leads to the derepression of Prdm16. Based on the literature (Chuikov et al., 2010; Stine et al., 2019; Takahata et al., 2009) and our data that Prdm16 can repress Smad transcriptional activity in human PANC-1 cells (Fig. 6 D), we considered the possibility that Smad4 might recruit Prdm16 to its own promoter, thereby leading to Prdm16 repression. Indeed, we detected a strong binding of Prdm16 to its promoter in PANC-1 cells at steady state, and this was almost completely suppressed upon TGF-β1 stimulation (Fig. 6 E), strongly suggesting that TGF-β signaling activation might dislodge Prdm16 from its promoter. In comparison, we were not able to detect any binding of Prdm16 to its promoter in BxPC-3 cells (Fig. 6 E), attesting to the specificity of our experiments, and further providing strong evidence supporting the notion that Smad4 functions to recruit Prdm16 to its promoter to repress its expression. To corroborate these findings, we conducted co-immunoprecipitation assays using PANC-1 cells, and detected a strong interaction between Prdm16 and Smad4, which was inhibited upon treatment of cells with TGF-β1 (Fig. 6 F). Such interaction was not detected in PANC-1 cells deleted of SMAD4 (Fig. 6 F), attesting to the specificity of the approach. Finally, we generated a reporter construct in which luciferase expression is under the control of either wild-type or mutated (SBE) PRDM16 promoter (Prdm16-Lux). We found that Smad4 was able to repress expression from the wild-type PRDM16 promoter in PANC-1 cells (Fig. 6 G). More importantly, expression of Prdm16 was also able to suppress luciferase expression from the wild-type PRDM16 promoter, and this effect was completely lost when the SBE mutated promoter was used in the assay (Fig. 6 G). Finally, expression of Prdm16 was able to repress the wild-type PRDM16 promoter in BxPC-3 cells only when Smad4 was co-expressed (Fig. 6 H). Overall, these findings revealed that Smad4 functions as a potent repressor of Prdm16, therefore providing a mechanistic explanation as to why KSC mice display high expression of Prdm16.
## Concomitant inactivation of Prdm16 and Smad4 recapitulates the global inactivation of TGF-β signaling
Both SMAD4 and TβRII are frequently inactivated in human PDAC, and landmark genetic experiments have shown that inactivation of either Smad4 or TβRII accelerates KrasG12D-driven PDAC (Bardeesy et al., 2006b; Ijichi et al., 2006; Izeradjene et al., 2007). To date, it remains largely unknown whether inactivation of Smad4 or TβRII could differentially impact the dynamics or trajectory of PDAC progression. Our mechanistic data that inactivation of Smad4 recapitulates the effects of TGF-β signaling on Prdm16 expression provided us with a unique platform to address this issue. To do so, we conducted an in-depth comparative analysis of the PDAC phenotypes in mice with homozygous deletion of TβRII (KTβC), KSC, and KSPrC mice side-by-side. KC and wild-type mice were used as controls. Kaplan-*Meier analysis* showed that KTβC mice developed lethal PDAC much more earlier than KSC mice, often succumbing to the disease within 4 wk of age and none survived beyond 17 wk, whereas $60\%$ of KSC mice survived within this observation period (Fig. 7 A). This observation indicates that global inactivation of TGF-β signaling through TβRII ablation is more efficient at deepening PDAC progression than inactivation of canonical TGF-β/Smad signaling through Smad4 ablation. More importantly, we found that KSPrC mice succumbed to lethal PDAC with kinetics approaching that of KTβC mice (Fig. 7 A), suggesting that simultaneous inactivation of Smad4 and Prdm16 might be sufficient to recapitulate the global inactivation of TGF-β signaling.
**Figure 7.:** *Concomitant ablation of Prdm16 and Smad4 recapitulates the global inactivation of TGF-β signaling through ablation of TβRII. (A) Kaplan-Meier survival analysis of control, KC, KSC, KSPrC, and KTβC mice (n = 19–45). Statistical power was assessed by a log-rank test for significance (left). The percentage of survival at the end of the observation period (right). (B) FFPE pancreatic sections from 1- to 4-mo-old control, KC, KSC, KSPrC, and KTβC mice (n = 19–45) were stained with H&E or Alcian blue, or immunostained with antibodies to CK19 or Muc5AC and subjected to IHC. Scale bars: 50 μm (top). Relative intensity of CK19, Muc5AC, and Alcian Blue or relative IPMN abundance are shown (bottom). Data are expressed as mean ± SEM, and statistical power was assessed by a two-tailed, unpaired Mann–Whitney test.*
Next, in light of our earlier findings that concomitant inactivation of Prdm16 was able to shift the evolution of the IPMN-to-PDAC phenotype in KSC mice toward the PanIN-to-PDAC phenotype, we wondered whether ablation of TβRII or Smad4 could differentially affect the nature of the premalignant lesions leading to PDAC, and if so, whether this event depends on Prdm16. Thus, we conducted histopathological analyses to compare the PDAC phenotypes in KSPrC, KTβC and KSC mice both at the levels of pre-malignant and full-blown PDAC lesions. H&E staining showed that KTβC tumors displayed uniformly poorly differentiated architecture, which is consistent with the rapid development of invasive PDAC in these mice (Fig. S5). Nevertheless, using KTβC mice before displaying signs of invasive PDAC, we consistently noticed the presence of premalignant lesions that display the classical features of PanINs, as gauged by IHC using antibodies to CK19 and Muc5AC (Fig. 7 B). Interestingly, KSPrC mice displayed similar cancerous phenotype as KTβC mice, both in terms of PanIN and PDAC lesions (Fig. 7 B and Fig. S5). In contrast, KSC mice consistently showed abundant and large IPMN lesions that exhibit high reactivity to the anti-Muc5AC antibody and Alcian blue (Fig. 7 B), which is in agreement with previous studies that KSC mice develop IPMN premalignant lesions rather than PanIN lesions (Bardeesy et al., 2006b; Izeradjene et al., 2007; Whittle et al., 2015). Taken together, these findings strongly suggest that inactivation of the entire TGF-β/Smad pathway promotes PanIN-to-PDAC progression, whereas inactivation of Smad4 promotes IPMN-to-PDAC progression. In addition, since concomitant ablation of Prdm16 and Smad4 resulted in highly aggressive PDAC similar to what was observed in KTβC mice, we suggested that global inactivation of TGF-β signaling might simultaneously inactivate both Smad4 and Prdm16.
**Figure S5.:** *Alterations in TGF-β signaling result in different PDAC trajectories. FFPE pancreatic sections from 1–4-mo-old KC, KSC, KSPrC, and KTβC mice (n = 19–45) with full-blown PDAC were stained with H&E or Alcian blue, or immunostained with antibodies to CK19 or Muc5AC and subjected to IHC. Representative pictures are shown. Scale bars: 50 μm (top left). Relative number of PDAC lesions (top right). Relative intensity of CK19, Muc5AC, and Alcian blue staining in PDAC lesions (bottom). Data are expressed as mean ± SEM, and statistical power was assessed by a two-tailed, unpaired Mann–Whitney test.*
## Discussion
Prdm16 belongs to the PR domain-containing protein family of transcription factors, which control a plethora of essential cellular processes, including specification of cell lineage during development (Chi and Cohen, 2016). Prdm16 was first identified in leukemia, where truncation mutants lacking functional domains behaved as oncogenic (Zhou et al., 2016), providing the first indication that Prdm16 might function as a tumor suppressor. In addition its involvement in leukemia, several studies have subsequently shown that Prdm16 controls brown fat cell differentiation as well as dedifferentiation of white fat to beige fat (Harms et al., 2015; Hiraike et al., 2017; Seale et al., 2008; Seale et al., 2007). Moreover, Prdm16 is required for stemness in multiple tissues, including hematopoietic and nervous systems (Chi and Cohen, 2016). Germline deletion of Prdm16 in mice impairs the maintenance of neural and hematopoietic stem cells during fetal development, resulting in neonatal death (Shimada et al., 2017). As such, this lethal phenotype hampered any further investigation to delineate a possible role of Prdm16 in cell fate determination in other organ systems, such as pancreas, where the same progenitor cells give raise to all pancreas lineages, e.g., ductal, acinar, and islet (Gu et al., 2003). In this study, we found that conditional deletion of Prdm16 in early pancreatic progenitor cells had no discernible impact on animal health or pancreas physiology, indicating that Prdm16 is dispensable for pancreas development and function. Because mutational inactivation of PRDM16 has been shown to be associated with leukemia (Zhou et al., 2016), we went on to explore whether Prdm16 could contribute to the pathogenesis and/or progression of PDAC, in which acquisition of oncogenic KRAS endows acinar cells with stemness traits that facilitate their differentiation toward a ductal-like lineage, thereby culminating in acinar-to-ductal metaplasia and attendant emergence of premalignant lesions (Bardeesy et al., 2006a; Gu et al., 2003; Park et al., 2008; Tuveson et al., 2004). Progression of premalignant lesions either follows the PanIN-to-PDAC sequence, MCN-to-PDAC, or IPMN-to-PDAC sequence, depending on the nature of the secondary genetic events (Bardeesy et al., 2006a; Bardeesy et al., 2006b; Gu et al., 2003; Tuveson et al., 2004). Yet, among the most studied secondary genetic alterations in PDAC, only Smad4 inactivation stood out as the main mechanism that enables progression through the IPMN-to-PDAC sequence (Bardeesy et al., 2006b; Whittle et al., 2015). To the best of our knowledge, how Smad4 inactivation facilitates this IPMN-to-PDAC transition phenotype has never been addressed experimentally. Using the KrasG12D-based mouse model of PDAC, we confirmed that KSC mice develop mostly IPMN lesions as described initially (Bardeesy et al., 2006b) rather than MCN lesions described in a subsequent study (Izeradjene et al., 2007). Most importantly, we found that concomitant ablation of Prdm16 and Smad4 (KSPrC) resulted in highly aggressive tumors, which develop with very short latencies to the full-blown PDAC and frequently metastasize to the lung, a site associated with the human disease (Connor and Gallinger, 2021; Hidalgo, 2010). Comprehensive histopathological analyses revealed that these tumors follow the PanIN-to-PDAC progression route rather than the IPMN-to-PDAC progression route that proceeds with ablation of Smad4 alone. Because inactivating Smad4 led to the increased expression of Prdm16, we proposed a model in which Prdm16 functions as a molecular switch to dictate whether the malignant transformation process follows the IPMN-to-PDAC route or the PanIN-to-PDAC route (Fig. 8). This model also posits that Prdm16 might function to suppress PDAC pathogenesis at very early stages of the malignancy. In further support of this notion, we found that ablating PRDM16 in the human PDAC cancer cell lines BxPC-3 and PANC-1 did not influence their proliferative or metastatic behaviors, as evidenced using a variety of in vivo and in vitro tumor growth and invasion assays. In light of these findings, a more comprehensive investigation using genetic and histological approaches are needed to firmly establish whether Prdm16 indeed elicits its tumor suppressor activity at early stages, and if so, whether this occurs through direct effects on cancer cell growth or tumor microenvironment reprogramming. As such, our findings open up unique frameworks that would ultimately leverage general efforts to unravel mechanistic paradigms of PDAC, for which very limited therapeutic interventions are currently available.
**Figure 8.:** *Model for the functional interaction between Smad4 and Prdm16 during PDAC formation and progression.*
Accumulating evidence suggests that Prdm16 functions as a potent inhibitor of TGF-β/Smad signaling under various physiological contexts (Chuikov et al., 2010; Stine et al., 2019; Takahata et al., 2009). TGF-β/Smad signaling is well known to play a dual role during cancer progression, functioning at early stages as a tumor suppressor to restrict the malignant transformation, and at late stages as a tumor promoter to facilitate cell invasion and metastasis (Feng and Derynck, 2005). To date, the most appealing speculations as to TGF-β dual function during PDAC progression have been that loss of the TGF-β cytostatic function enables cells to escape growth-inhibitory regulation, which would ultimately culminate in malignant transformation (David et al., 2016; Feng and Derynck, 2005; Massagué, 2008). Once the tumor has developed, other TGF-β responses unrelated to its cytostatic function then supposedly prevail presumably in a manner that facilitates PDAC invasion and metastasis (Bardeesy et al., 2006b; Feng and Derynck, 2005; Ijichi et al., 2006; Massagué, 2008). Interestingly, high levels of TGF-β expression in human PDAC strongly correlates with poor prognosis (Friess et al., 1993; Parajuli et al., 2019), which raises a conundrum as to whether activation of TGF-β signaling could contribute directly to malignant transformation in addition to driving cell invasion and metastasis. However, subsequent studies have shown that Smad4 inactivation in the context of KrasG12D (KSC) led to the acceleration of PDAC (Bardeesy et al., 2006b; Izeradjene et al., 2007), unequivocally confirming the tumor suppressor role of TGF-β signaling in PDAC. Nevertheless, the tumors deficient for Smad4 retained epithelial differentiation and manifested an attenuated metastatic potential (Bardeesy et al., 2006b; Whittle et al., 2015), which is also in favor of a tumor promoter role of TGF-β signaling. So far, definitive experimental evidence on whether inactivation of canonical TGF-β/Smad signaling per se is sufficient to suppress PDAC invasion and metastasis in an irreversible manner is still lacking. Here, we found that ablating Prdm16 in a Smad4 null-background was sufficient to render the PDAC tumors again highly invasive and metastatic. Intriguingly, concomitant ablation of Prdm16 in KSC mice also resulted in a shift from IPMN to PanIN, which could conceivably contribute to metastasis in KSPrC mice, as the vast majority of PDAC GEMMs that develop PanINs also develop highly metastatic PDAC, including KSPC mice (Smad4 deletion and p53.R172H expression), which behave similarly to our KSPrC mice (Bardeesy et al., 2006a; Bardeesy et al., 2006b; Tuveson et al., 2004; Whittle et al., 2015). These findings, together with the observation that Prdm16 expression is lost during late stages of PDAC, highlight Prdm16 as a key player in PDAC progression and metastasis when Smad4 is inactivated. Because TGF-β signaling activation leads to the accumulation of Prdm16 through the suppression of Smad4 inhibitory effects, one would speculate that Smad4 and Prdm16 might function in the same signaling network that integrates the TGF-β tumor promoter effects during PDAC progression. However, it is also conceivable that Prdm16 might function to suppress metastasis induced by other TGF-β superfamily members, such as Activins and BMPs, which are known to signal through Smad4, and can enhance malignancy and promote cancer metastasis in a variety of human malignancies (Attisano and Wrana, 2000; Feng and Derynck, 2005; Pickup et al., 2017). As such, a comprehensive investigation of the mechanisms by which Smad4 and Prdm16 interact to influence PDAC progression may uncover the existence of additional key players and/or pathways that are amenable to therapeutic interventions.
Perhaps the most intriguing finding in this study was the persistent increase in Prdm16 expression during the progression from IPMN to PDAC in KSC mice, which at first glance seems to support a hypothesis in which Smad4 might function as a repressor of Prdm16 during PDAC progression, and hence conceivably that canonical TGF-β/Smad signaling might also repress Prdm16 expression. Quite unexpectedly, we found that activation of TGF-β signaling did not repress Prdm16 expression, but rather resulted in a strong accumulation of both Prdm16 mRNA and protein both in KSC mice and human PANC-1 cells. Noteworthy, we also detected relatively high expression in the stromal compartment, which likely occurs because of the increased TGF-β signaling, which is known to take place during PDAC progression and contribute to the desmoplastic stroma of this malignancy (Friess et al., 1993). In efforts to probe the underlying mechanisms, we found that inactivating Smad4 was sufficient to recapitulate the effects of TGF-β signaling, inducing Prdm16 expression to an extent similar to that elicited by TGF-β1. Based on these observations, we reasoned that activation of TGF-β signaling might relieve the transcriptional repression imposed by Smad4 on the Prdm16 promoter. However, although we found that Smad4 associated strongly with the PRDM16 promoter at steady state, this binding was not affected by the activation of TGF-β1 signaling, indicating that other factors are involved in TGF-β-mediated Prdm16 expression. Probing this possibility, we detected a strong binding of Prdm16 to its own promoter at steady state, which was almost completely abolished by TGF-β stimulation, suggesting that activation of TGF-β signaling might dislodge Prdm16 from its own promoter. Of note, Prdm16 failed to bind to its promoter in cells deficient for SMAD4, suggesting that Smad4 might associate with and recruit Prdm16 to the PRDM16 promoter. Because Prdm16 has been shown to function as a potent transcriptional repressor in various contexts (Pinheiro et al., 2012; Seale et al., 2008; Seale et al., 2007; Stine et al., 2019; Takahata et al., 2009), we proposed a model in which Prdm16 mediates its own repression once it has been recruited to its promoter by Smad4. While these data demonstrate for the first time that Prdm16 can repress its own expression, we cannot exclude the possibility that other mechanisms might also contribute to this phenomenon. Despite this limitation, our study sheds light on a previously uncharacterized interplay between Smad4 and Prdm16, which appears to dictate the progression trajectory of PDAC. Going forward, we anticipated that our discovery will guide forthcoming studies seeking to understand mechanistic paradigms of PDAC, which could ultimately pave the way for innovative therapeutic breakthroughs to curb this deadly disease.
## Plasmids
The CAGA9-*Lux* gene reporter construct was previously described (Seo et al., 2006). The expression vector pcDNA3.1-Prdm16 was a gift from Dr. Bruce Spiegelman (#15503; Addgene). The expression vector pCMV5-HA-Smad4 was a gift from Dr. Joan Massague. *To* generate the Prdm16-Lux reporter, genomic fragments (1,391 bp) upstream of the transcription initiation site (SST) of the PRDM16 gene (based on gene association NM_022114 and Eukaryotic Promoter Database, epd.epfl.ch) was amplified by the Genomic-GC PCR amplification kit (BD Biosciences) using human genomic DNA obtained from PANC-1 cells. Unique KpnI and XhoI sites were incorporated at the 5′ and 3′ ends of the sequence, respectively, to simplify directional cloning into KpnI and XhoI sites in the reporter plasmid, pGL3-basic (Promega). Introduction of inactivating mutation into the SBE sequence (−41 bp from SST) was generated by PCR using the QuickChange Site-Directed Mutagenesis kit according to the manufacturer’s instructions (Stratagene). The lentiCRISPRV2 expression vectors encoding SMAD4 and PRDM16 gRNAs were purchased from GenScript. The lentiCRISPRV2 expression vectors encoding SMAD2 and SMAD3 gRNAs were generated using lentiCRISPRV2 hygro (#98291; Addgene) and primers with sequences generated using the Synthego Design tool. All cloned cDNAs and their corresponding mutants were checked by sequencing.
## SMAD4
5′-TTCTTCCTAAGGTTGCACAT-3′; 5′-AATACACTTACCAGGATGAT-3′.
## PRDM16
5′-CTCGTACGGCGAGCCCTCCT-3′; 5′-AGGGGTCTTACCGTCCAGGC-3′.
## SMAD2
5′-TGGCGGCGTGAATGGCAAGA-3′; 5′-TTCACAACTGGCGGCGTGAA-3′.
## SMAD3
5′-CACCTGCAACCGGCCATCCA-3′; 5′-ACACCTGCAACCGGCCATCC-3′.
## Antibodies
Chromatin immunoprecipitation (ChIP), immunoblotting, immunofluorescence, or immunohistochemistry were performed using the following antibodies: anti-α-SMA (#19245T; Cell Signaling); anti-β-Actin (#64225332; Bio-Rad), anti-amylase (#ab21156; Abcam), anti-chromogranin-A (#ab45179; Abcam), anti-cytokeratin 19 (#ab52625; Abcam), anti-E-cadherin (#3195S; Cell Signaling), anti-glucagon (#2760; Cell Signaling), anti-insulin (#4590; Cell Signaling), anti-JunB (#3753; Cell Signaling), anti-Muc5AC (#ab3649; Abcam), anti-Prdm16 (#ab202344 and #ab106410; Abcam), anti-Smad2 (#5339; Cell Signaling), anti-Smad3, (#9523; Cell Signaling), anti-Smad4, (#46535; Cell Signaling), anti-Smad4 (#sc-7966; Santa Cruz), anti-Smad$\frac{2}{3}$ (#8685; Cell Signaling), and anti-vimentin (#5741S; Cell Signaling).
## Cell lines and culture
HEK293T, MIA-PaCa-2, BxPC-3, and PANC-1 cell lines were obtained from the American Type Culture Collection (ATCC). They were cultured in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with $10\%$ fetal bovine serum (#S11150; FBS, Atlanta Biologicals), antibiotics (#P4458; Gibco) and L-glutamine (#17921004; Corning). The murine pancreatic cancer cell line KPC1 was originally described in our recent publication (Parajuli et al., 2018). The cell line was established from a KP53 mouse, which harbored KrasG12D and one conditional allele of Trp53 (LSL-KrasG12D;LSL-Trp53fl/+;Pdx1-Cre). Freshly isolated specimen from the KP53 mouse with terminal PDAC was gently dissected, minced with scissors, and digested with Dispase II at 2.4 U/ml (#4942078001; Sigma-Aldrich) and Collagenase D at 0.5 mg/ml (#11088858001; Sigma-Aldrich) for 1 h at 37°C in an atmosphere of $5\%$ CO2. Then, cells were washed three times with PBS, suspended in RPMI 1540 containing $20\%$ FCS, and seeded on fibronectin-coated plates. Cell colonies were subsequently passaged by trypsinization, pooled, and propagated in DMEM supplemented with $10\%$ FBS, antibiotics, and L-glutamine. *To* generate the PANC-1-SMAD4KO and PANC-1-SMAD$\frac{2}{3}$KO cell lines, cells were transduced with the corresponding lentiCRISPRV2-gRNA lentiviruses, selected with puromycin (for SMAD4) or hygromycin (for SMAD$\frac{2}{3}$), and all resistant clones were pooled and expanded as a single population. Lentiviruses were produced by transfecting HEK293T cells with lentiviral constructs and the One-Step Lentivirus Packaging System as described by the manufacturer (#631275; Takara). Lentiviral particles in the conditioned media were harvested after a period of 48–72 h. The conditioned media were then cleaned of cell debris by centrifugation at 5,000×g for 15 min, filtered through a 0.45-μm filter, and used immediately for cell transduction.
## In vitro and in vivo cell proliferation assays
For the soft agar assay, cell culture dishes (p60) were first prepared using complete DMEM media containing $0.6\%$ agarose (#16500500; Thermo Fisher Scientific) and allowed to solidify at room temperature for 2 h. Then, cells suspended in complete DMEM media containing $0.3\%$ agarose were added to the dishes preloaded with the $0.6\%$ agarose layer. PANC-1 and BxPC-3 stably expressing control or PRDM16 gRNA were plated at a density of 1,000 cells per dish and grown for ∼2 mo. Within this time frame, PANC-1 isogenic cell lines developed small but similar colonies in size, whereas neither of the BXPC-3 isogenic cell lines developed colonies. Colonies were visualized and counted using an Olympus CKX53 microscope with the UPlanFL N 4×/0.13 iPC objective.
For the cell proliferation assay, isogenic PANC-1 (50,000 cells/well) and BxPC-3 (100,000 cells/well) cell lines stably expressing control or PRDM16 gRNA were inoculated into 6-well plates. Three (for PANC-1) and six (for BxPC-3) days after inoculation, cells were trypsinized and mixed with equal volumes of trypan blue (#T10282; Invitrogen) before being counted using an automatic cell counter (#AMQAF2000; Invitrogen Countess 3 FL). Each well was counted twice and averaged to ensure accurate cell counts were obtained.
For the in vivo growth assay, NOD scid gamma (NSG) mice were injected subcutaneously with isogenic PANC-1 and BxPC-3 cell lines stably expressing control or PRDM16 gRNA (106 cells) under septic conditions. During the observation period of ∼2 mo, mice were maintained in sterile conditions and sacrificed if they displayed any symptoms of illness. At the end of the observation period, tumors were dissected, weighted, and imaged using a 12-megapixel f/1.8 aperture camera.
## Mice
NOD scid gamma (NSG), Prdm16fl/fl, Smad4fl/fl, TβR2fl/fl and Trp53fl/fl mice were obtained from Jackson Laboratories. Loxp-Stop-Loxp-KrasG12D (LSL-KrasG12D) and Pdx1-Cre mice were obtained from the NCI Mouse Repository. p16Ink4A-Luciferase (p16Luc) was kindly provided by Dr. Sharpless (Burd et al., 2013). All PDAC mouse models were generated through successive crossbreeding of Prdm16fl/fl, Smad4fl/fl, TβR2fl/fl, p16Luc, Trp53fl/fl, LSL-KrasG12D and Pdx1-Cre mice as appropriate. Full descriptions of the genotypes of mice used throughout the study are: KC: LSL-KrasG12D;Pdx1-Cre; Prdm16KO: Prdm16fl/fl;Pdx1-Cre; KPrC: LSL-KrasG12D;Prdm16fl/fl;Pdx1-Cre; KSC: LSL-KrasG12D;Smad4fl/fl;Pdx1-Cre; KSPrC: LSL-KrasG12D;Smad4fl/fl;Prdm16fl/fl;Pdx1-Cre; KPC: LSL-KrasG12D;LSL-Trp53fl/fl;Pdx1-Cre; KIC: LSL-KrasG12D;p16Luc+/+;Pdx1-Cre; KTβC: LSL-KrasG12D;TβR2fl/fl;Pdx1-Cre.
The Institutional Animal Care and Use Committee (IACUC) of the University of Mississippi Medical Center (UMMC) or Virginia Commonwealth University (VCU) approved all animal experiments. All experiments with transgenic mouse models (including KSPrC mice) were initiated at UMMC and continued at VCU. We did not see any significant difference in the onset of tumor formation or survival in mice generated or maintained in both sites.
All mice were maintained on a mixed C57BL/6 and FVB/N genetic background. Mice were maintained in twelve-hour light/dark cycles (6:00 AM–6:00 PM) at 22°C and fed a standard rodent chow diet. Formation of PDAC in all mice enrolled in the study was confirmed using pancreatic tissue sections stained with hematoxylin and eosin (H&E) or immunostained with an anti-cytokeratin 19 antibody. Blood glucose levels were measured with blood collected from the tail vein using the ReliON Prime blood glucose strips. The average of one measurement from 2 to 3 different blood ReliON meters was used for each mouse.
## Clinical samples
Human tissue micro arrays for pancreatic tissues (#PA242b, $$n = 24$$; #PA483c, $$n = 48$$; #PA805c, $$n = 80$$) were purchased from US Biomax, Inc.
## Kaplan-Meier survival analysis in patients with wild-type or mutant SMAD4
In order to compare the survival of patients with high versus low expression of PRDM16 in the context of wild-type or mutated SMAD4, PRDM16 expression data (mRNA expression z-scores relative to all samples (log RNA Seq V2 RSEM) were first downloaded from the TCGA PanCancer Atlas in cBioPortal. Then, patients of the TCGA-PAAD cohort with wild-type ($$n = 140$$) or mutated ($$n = 26$$) SMAD4 were identified using the COSMIC database. Next, patients were classified as having high or low PRDM16 expression based on whether they were above or below the top and bottom quartile of PRDM16 expression of the TCGA-PAAD cohort, respectively. Lastly, each patient was matched with the corresponding expression of PRDM16 and SMAD4 mutational status as well as the time to death or to last follow up (depending on their vital status) to create a Kaplan-Meier survival curve.
## PRDM16 expression in patients with or without SMAD4 mutations
To assess the expression of PRDM16 in patients in the context of wild-type or mutated SMAD4, PRDM16 expression data (mRNA expression z-scores relative to all samples, log RNA Seq V2 RSEM) were first downloaded from the TCGA PanCancer Atlas in cBioPortal. Then, patients of the TCGA-PAAD cohort with different types of SMAD4 mutations were identified using the cBioPortal interface. Patients were then filtered based on those with no alteration or with truncating mutations in SMAD4 in order to create a violin plot comparing the normalized PRDM16 expression between these two groups.
## qRT-PCR
Total RNA was extracted from frozen mouse tissue samples using TRIzol (#15596018; Ambion) and purified with chloroform (#066903; Thermo Fisher Scientific) and ethanol (#BP2818; Thermo Fisher Scientific). The RNA was then reverse-transcribed using a High-Capacity cDNA Reverse Transcription kit (#4368814; Applied Biosystems). The cDNA product was analyzed by qRT-PCR. Briefly, 25 ng cDNA and 150 nmol of each primer were mixed together with the SsoFast EvaGreen Supermix (#1725200; BioRad). PCR reactions were conducted using a CFX96 Real-Time System (BioRad) in a 96-well plate. The relative mRNA levels were calculated with the comparative CT method and normalized to GAPDH mRNA.
## Primers used for human samples
PRDM16-For 5′-CTTTGACCACACCCGAAGGT-3′; PRDM16-Rev 5′-TGTGGAGAGGAGTGTCTTCG-3′; JUNB-For 5′-CCTGGACGATCTGCACAAGA-3′; JUNB-Rev 5′-GGTTGGTGTAAACGGGAGGT-3′; GAPDH-For 5′-CCATGGGGAAGGTGAAGGTC-3′; GAPDH-Rev 5′-AGTGATGGCATGGACTGTGG-3′.
## Primers used for mouse samples
Prdm16-For 5′-TCCCACCAGACTTCGAGCTA-3′; Prdm16-Rev 5′-AAAGTCGGCCTCCTTCAGTG-3′; Gapdh-For 5′-CACCATCTTCCAGGAGCGAG-3′; Gapdh-Rev 5′-CACCATCTTCCAGGAGCGAG-3′.
## Chromatin immunoprecipitation assay (ChIP)
ChIP assays were performed using a kit following the manufacturer’s instructions (#17-295; Millipore). Accordingly, cells were first treated with $1\%$ formaldehyde and incubated at 37°C for 10 min. Next, cells were washed twice with ice-cold PBS containing protease inhibitors. Cells were then scraped and pelleted by centrifugation at 2,000 RPM for 4 min at 4°C. Then, cells were resuspended in SDS Lysis Buffer (Millipore, #20-163) and incubated for 10 min on ice. After samples were centrifuged for 10 min at 13,000 RPM at 4°C, the supernatants were diluted 10 times by adding ChIP Dilution Buffer (#20-153; Millipore) containing protease inhibitors. The diluted supernatants were then treated with 75 μl of a $50\%$ slurry of Protein-A Agarose/Salmon Sperm DNA (#16-157C; Millipore) at 4°C for 30 min with agitation. After centrifugation, supernatants were immunoprecipitated with antibodies against Smad4, Smad2, Smad3, Prdm16, GAPDH or isotype-matched control IgG and 60 μl of a $50\%$ slurry of Protein-A Agarose/Salmon Sperm DNA at 4°C for 1 h with rotation. Agarose was pelleted using centrifugation at 1,000 RPM for 1 min at 4°C. The pellets were washed for 5 min in Low Salt Immune Complex Wash Buffer (#20-154; Millipore) once, High Salt Immune Complex Wash Buffer (#20-155; Millipore) once, LiCl Immune Complex Wash Buffer (#20-156; Millipore) once, and TE Buffer (#20-157; Millipore) twice. To amplify DNA bound to the immunoprecipitates, elution buffer ($1\%$ SDS, 0.1 M NaHCO3) was added to each sample followed by agitation and incubation for 15 min with rotation at room temperature. Eluates were then mixed with NaCl (final concentration of 0.2 M) and incubated for 4 h at 65°C followed by adding EDTA (0.01 M), Tris-HCl, pH 6.5 (0.04 M), and Proteinase K (0.04 mg/ml). Samples were then incubated for 1 h at 45°C, and DNA was recovered using phenol/chloroform extraction coupled with ethanol precipitation. Pellets were washed with $70\%$ ethanol and air-dried. Lastly, pellets were resuspended in an appropriate buffer for PCR, and PCR products were analyzed on a $2\%$ agarose gel. The immunoprecipitated DNA was also analyzed by qPCR using locus specific primers and normalized to input DNA. Relative fold enrichment in each locus was quantified relative to the control as described above (qRT-PCR) as well as in our published studies (Parajuli et al., 2018; Zhang et al., 2015). The following primers were used: PRDM16-For 5′-CATCTCCCCAGCATTGTCAGT-3′; PRDM16-Rev 5′-GGAGCGCCGAACACGGAATG-3′; JUNB-For 5′-GGCAAAGCCCAGGGTCAATA-3′; JUNB-Rev 5′-AAAGCTAGTAAGCGGCCTGG-3′; GAPDH-For 5′-CGGGATTGTCTGCCCTAATTAT-3′; GAPDH-Rev 5′-GCACGGAAGGTCACGATGT-3′.
## Luciferase reporter assay
PANC-1 cells were plated in 6-well plates and transfected with the CAGA9-Lux or Prdm16-Lux reporter in the presence of pcDNA3.1-Prdm16, pCMV5-HA-Smad4, or empty vector (pcDNA3.1 or pCMV5-HA as appropriate) using X-tremeGENE9 (#0635779001; Sigma-Aldrich). The pRL-SV40 plasmid (#AF025845; Promega) was cotransfected to normalize for transfection efficiency. For CAGA9-Lux assays, cells were incubated for 24 h with the transfection mixtures and then treated with 5 ng/ml TGF-β1 (#7754-BH; R&D Systems) for 24 h before measuring luciferase activity using the Dual-Luciferase Reporter Assay System (#E1910; Promega). For Prdm16-Lux assays, cells were incubated for 48 h with the transfection mixtures and then processed for luciferase activity as described for CAGA9-Lux. Firefly Luciferase activity was normalized based on Renilla luciferase expressed from pRL-SV40 plasmid.
## Co-immunoprecipitation
Cell lysates were prepared in lysis buffer (25 mM Tris-HCl, pH 7.2, 150 mM NaCl, 5 mM MgCl2, $5\%$ glycerol and $1\%$ NP40) supplemented with phosphatase inhibitors (#P5726; Sigma-Aldrich) and EDTA-free protease inhibitors (#P8340; Sigma-Aldrich). Cells were lysed with 1 ml of lysis buffer for 10 min on ice and protein concentrations were determined using the BCA reagent (#23227; Thermo Fisher Scientific). Then, $90\%$ of the pre-cleared lysates were added to anti-Smad4 antibody for 1 h at 4°C under constant rocking, and then protein A magnetic beads (#G8781; Promega) were added for an additional 1 h at 4°C. The beads were subsequently pelleted and washed five times with lysis buffer and eluted for immunoblotting using 1X SDS-PAGE sample buffer (#NP0007; Thermo Fisher Scientific). The other remaining $10\%$ of lysate was used to determine total protein levels by direct immunoblotting.
## Immunoblotting
Cell extracts were prepared in lysis buffer containing 20 mM Tris HCl (pH 7.5), 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, $1\%$ Triton, $2.5\%$ sodium pyrophosphate, 1 mM β-glycerophosphate, 1 mM Na3VO4, 1 µg/µL leupeptin, protease inhibitors (#P8340; Sigma-Aldrich) and phosphatase inhibitors (#P5726; Sigma-Aldrich). Protein concentrations were determined using the BCA reagent as described earlier, and samples were denatured using SDS sample buffer (#1610747; BioRad). Samples were loaded into a Criterion Tris-Glycine Extended Gel (#5671124; BioRad) and separated by electrophoreses at 60 mA. The gels were then transferred onto a nitrocellulose membrane (#1620115; BioRad) by a wet transfer system (BioRad) at 100V for 1 h at room temperature. All membranes were then blocked by incubation with $5\%$ dry milk in TBST (TBS with $0.1\%$ Tween20) for 1 h at room temperature. Membranes were probed with the primary antibody overnight at 4°C in the blocking buffer, washed with TBST, and incubated with the peroxidase-conjugated secondary antibody. Enhanced chemiluminescence (ECL) Western blotting substrates (#170-5061; BioRad) were used for the visualization of the results. The acquisition of images was performed using the ChemiDoc MP Imaging System (BioRad).
## Histology, immunohistochemistry, and immunofluorescence
Tissue samples were fixed in $10\%$ formalin and embedded in paraffin. For pancreatic tissue histology, paraffin sections were stained with hematoxylin and eosin (H&E) using standard techniques. Briefly, sections were deparaffinized with xylene and rehydrated in a graded series of ethanol. They were then successively immersed in a hematoxylin solution (HHS128-4L; Sigma-Aldrich) for 2 min, a clarifier solution (7402L; Epredia) for 15 s, and blueing reagent solution (7301L; Epredia) for 1 min. Between each of the three steps, sections were immersed in water for 1 min. Next, slides were immersed in an eosin solution (HT110280-2.5L; Sigma-Aldrich) for 3 min before being dehydrated 3 times for 3 min in $100\%$ ethanol (89370-088; VWR) followed by xylene (V1001; Koptec). For immunofluorescence and immunohistochemistry, tissue sections were deparaffinized with xylene and rehydrated in a graded series of ethanol. Antigen retrieval was performed for 30 min at high temperature in citrate buffer. Then, slides were blocked and incubated overnight with anti-insulin, anti-glucagon, anti-Prdm16, anti-Muc5AC, anti-chromogranin-A, anti-αSMA, anti-E-cadherin, anti-vimentin or IgG-matched isotype control antibody (negative control) at 4°C. For immunofluorescence, slides were incubated with the secondary antibodies conjugated to Alexa-Fluor568 (#A11011; Invitrogen) or Alex-Fluor488 (#A11088; Invitrogen), co-stained with DAPI (#H1800; Vector Laboratories), and viewed on a Nikon Ti-E fluorescence microscope. Immunohistochemistry was done with the VECTASTAIN Elite ABC HRP kit (rabbit, #PK6101 or mouse, #PK-6102; Vector Laboratories) as per manufacturer’s instructions. Tissue sections were incubated for 30 min in the secondary antibody followed by the VECTASTAIN ABC reagent. Color development was done with the DAB Peroxidase Substrate kit (#SK-4100; Vector Laboratories) with or without Nickel added enhancement as appropriate.
To quantify Prdm16 expression in human samples, the TMA was scanned using the PlanApo 40 × $\frac{0.95}{0.25}$–0.17 mm objective on the Keyence BZ-X810 automated microscope and characterized using the BZ-X800 Analyzer software from Keyence. The expression intensity of six images of normal areas and PanIN stages 1, 2, 3 and PDAC lesions were chosen in a random manner. The intensity of Prdm16 expression was obtained automatically using the BZ-X800 Analyzer software from Keyence and the means of each stage (normal, PanIN stage 1, 2, 3, PDAC) were calculated. Each area/lesion was individually quantified using the area directly around the lesion.
To quantify Prdm16 expression in mouse tissues, slides chosen in a random manner from all mice under study were scanned using the PlanApo 10 × $\frac{0.45}{4.00}$ mm objective on the Keyence BZ-X810 automated microscope and characterized using the BZ-X800 Analyzer software from Keyence. Random images of PanIN and PDAC lesions were taken and quantified only using the area directly around the lesions. Each lesion was individually quantified and the mean ± SEM of six independent lesions was presented in figures.
The quantifications of Alcian blue staining or CK19, Muc5AC, and α-SMA immunostaining were conducted by first taking images of six normal areas or PanIN/PDAC lesions from all mice under study in a random manner using the PlanApo 40 × $\frac{0.95}{0.25}$–0.17 mm objective on the Keyence BZ-X800 microscope. We then individually quantified each image using the Keyence BZ-X800 analyzer software from Keyence. Lastly, the mean ± SEM was calculated for each genotype.
To quantify the distribution of PDAC lesions, mouse tissue slides were scanned using the PlanApo 10 × $\frac{0.45}{4.00}$ mm objective on the Keyence BZ-X810 automated microscope and characterized using the BZ-X800 Analyzer software from Keyence. PanIN and IPMN lesions were counted and characterized as either PanIN or IPMN. The surface area for each lesion was obtained from the Keyence software and the mean sum of the surface area for all PanIN or IPMN lesions were calculated for each genotype, including all mice recruited. The percentage of stroma was identified using the BZ-X800 Analyzer software from Keyence. The distribution of PDAC lesions was then calculated by multiplying the percentage of PanIN surface area divided by the total non-PDAC surface area of the tissue. The same process was repeated for IPMN lesions.
All images were taken using the Zeiss Axio Lab. A1 upright (Zeiss EC Plan-NEOFLUAR 40×/0.9 Pol and Zeiss A-Plan 10×/0.25 objectives), Zeiss Observer. A1 inverted (Zeiss LDA-Plan 40×/0.55 Ph1 objective), or Leica DM1000LED upright microscopes (Leica HI PLAN 40×/0.65 objective). The numerical aperture of the objective lenses are 0.9 and 0.25, 0.55, and 0.65, respectively, with a temperature of 1 (10 Kelvin) with an imaging medium of air. The fluorochromes analyzed in immunofluorescence experiments were were Alexa-Fluor568 (red), Alexa-Fluor488 (green) and DAPI (blue). The cameras used were the Axiocam ICc5, Axiocam503mono, and LeicaDM2900 with the acquisition software was ZEN 2 lite for both Zeiss microscopes and LAS X for the Leica microscope. Subsequent software used for incorporating images into figures was Adobe Photoshop followed Microsoft PowerPoint.
## Statistical analysis
The values are expressed as mean ± SEM. The error bars (SEM) shown for all results were derived from biological replicates, not technical replicates. Significant differences between two groups were evaluated using either a two-tailed, unpaired Mann–Whitney test or two-tailed, unpaired t test, which was found to be appropriate for the statistics, as the sample groups displayed a normal distribution and comparable variance. Statistical significance of survival differences was determined by log-rank test.
## Online supplemental material
Fig. S1 shows that Prdm16 is transiently expressed in the premalignant lesions. Fig. S2 shows that Prdm16KO mice display normal insulin and glucagon expression and distribution as well as normal blood glucose levels. Fig. S3 provides additional data demonstrating that inactivation of Prdm16 accelerates KrasG12D-driven PDAC. Fig. S4 displays that Prdm16 is required for IPMN-to-PDAC progression and that Prdm16 deletion led to the accumulation of cells with high vimentin expression. In addition, Fig. S4 shows the effects of deleting PRDM16 on the proliferation of PANC-1 and BxPC-3 cell lines. Fig. S5 further expands upon the notion that concomitant inactivation of Prdm16 and Smad4 mimics the phenotype of complete TGF-β signaling inactivation through TβRII ablation.
## References
1. Attisano L., Wrana J.L.. **Smads as transcriptional co-modulators**. *Curr. Opin. Cell Biol.* (2000) **12** 235-243. DOI: 10.1016/S0955-0674(99)00081-2
2. Bardeesy N., Aguirre A.J., Chu G.C., Cheng K.H., Lopez L.V., Hezel A.F., Feng B., Brennan C., Weissleder R., Mahmood U.. **Both p16(Ink4a) and the p19(Arf)-p53 pathway constrain progression of pancreatic adenocarcinoma in the mouse**. *Proc. Natl. Acad. Sci. USA.* (2006a) **103** 5947-5952. DOI: 10.1073/pnas.0601273103
3. Bardeesy N., Cheng K.H., Berger J.H., Chu G.C., Pahler J., Olson P., Hezel A.F., Horner J., Lauwers G.Y., Hanahan D., DePinho R.A.. **Smad4 is dispensable for normal pancreas development yet critical in progression and tumor biology of pancreas cancer**. *Genes Dev.* (2006b) **20** 3130-3146. DOI: 10.1101/gad.1478706
4. Burd C.E., Sorrentino J.A., Clark K.S., Darr D.B., Krishnamurthy J., Deal A.M., Bardeesy N., Castrillon D.H., Beach D.H., Sharpless N.E.. **Monitoring tumorigenesis and senescence in vivo with a p16(INK4a)-luciferase model**. *Cell* (2013) **152** 340-351. DOI: 10.1016/j.cell.2012.12.010
5. Buscail L., Bournet B., Cordelier P.. **Role of oncogenic KRAS in the diagnosis, prognosis and treatment of pancreatic cancer**. *Nat. Rev. Gastroenterol. Hepatol.* (2020) **17** 153-168. DOI: 10.1038/s41575-019-0245-4
6. Chi J., Cohen P.. **The multifaceted roles of PRDM16: Adipose biology and beyond**. *Trends Endocrinol. Metab.* (2016) **27** 11-23. DOI: 10.1016/j.tem.2015.11.005
7. Chuikov S., Levi B.P., Smith M.L., Morrison S.J.. **Prdm16 promotes stem cell maintenance in multiple tissues, partly by regulating oxidative stress**. *Nat. Cell Biol.* (2010) **12** 999-1006. DOI: 10.1038/ncb2101
8. Connor A.A., Gallinger S.. **Pancreatic cancer evolution and heterogeneity: Integrating omics and clinical data**. *Nat. Rev. Cancer* (2021) **22** 131-142. DOI: 10.1038/s41568-021-00418-1
9. David C.J., Huang Y.H., Chen M., Su J., Zou Y., Bardeesy N., Iacobuzio-Donahue C.A., Massagué J.. **TGF-Β tumor suppression through a lethal EMT**. *Cell* (2016) **164** 1015-1030. DOI: 10.1016/j.cell.2016.01.009
10. David C.J., Massagué J.. **Contextual determinants of TGFβ action in development, immunity and cancer**. *Nat. Rev. Mol. Cell Biol.* (2018) **19** 419-435. DOI: 10.1038/s41580-018-0007-0
11. Duda D.G., Sunamura M., Lefter L.P., Furukawa T., Yokoyama T., Yatsuoka T., Abe T., Inoue H., Motoi F., Egawa S.. **Restoration of SMAD4 by gene therapy reverses the invasive phenotype in pancreatic adenocarcinoma cells**. *Oncogene* (2003) **22** 6857-6864. DOI: 10.1038/sj.onc.1206751
12. Feng X.H., Derynck R.. **Specificity and versatility in tgf-beta signaling through Smads**. *Annu. Rev. Cell Dev. Biol.* (2005) **21** 659-693. DOI: 10.1146/annurev.cellbio.21.022404.142018
13. Freeman J.W., Mattingly C.A., Strodel W.E.. **Increased tumorigenicity in the human pancreatic cell line MIA PaCa-2 is associated with an aberrant regulation of an IGF-1 autocrine loop and lack of expression of the TGF-beta type RII receptor**. *J. Cell. Physiol.* (1995) **165** 155-163. DOI: 10.1002/jcp.1041650118
14. Friess H., Yamanaka Y., Büchler M., Ebert M., Beger H.G., Gold L.I., Korc M.. **Enhanced expression of transforming growth factor beta isoforms in pancreatic cancer correlates with decreased survival**. *Gastroenterology* (1993) **105** 1846-1856. DOI: 10.1016/0016-5085(93)91084-U
15. Gu G., Brown J.R., Melton D.A.. **Direct lineage tracing reveals the ontogeny of pancreatic cell fates during mouse embryogenesis**. *Mech. Dev.* (2003) **120** 35-43. DOI: 10.1016/S0925-4773(02)00330-1
16. Harms M.J., Lim H.W., Ho Y., Shapira S.N., Ishibashi J., Rajakumari S., Steger D.J., Lazar M.A., Won K.J., Seale P.. **PRDM16 binds MED1 and controls chromatin architecture to determine a brown fat transcriptional program**. *Genes Dev.* (2015) **29** 298-307. DOI: 10.1101/gad.252734.114
17. Hayashi A., Hong J., Iacobuzio-Donahue C.A.. **The pancreatic cancer genome revisited**. *Nat. Rev. Gastroenterol. Hepatol.* (2021) **18** 469-481. DOI: 10.1038/s41575-021-00463-z
18. Hidalgo M.. **Pancreatic cancer**. *N. Engl. J. Med.* (2010) **362** 1605-1617. DOI: 10.1056/NEJMra0901557
19. Hingorani S.R., Wang L., Multani A.S., Combs C., Deramaudt T.B., Hruban R.H., Rustgi A.K., Chang S., Tuveson D.A.. **Trp53R172H and KrasG12D cooperate to promote chromosomal instability and widely metastatic pancreatic ductal adenocarcinoma in mice**. *Cancer Cell* (2005) **7** 469-483. DOI: 10.1016/j.ccr.2005.04.023
20. Hiraike Y., Waki H., Yu J., Nakamura M., Miyake K., Nagano G., Nakaki R., Suzuki K., Kobayashi H., Yamamoto S.. **NFIA co-localizes with PPARγ and transcriptionally controls the brown fat gene program**. *Nat. Cell Biol.* (2017) **19** 1081-1092. DOI: 10.1038/ncb3590
21. Iacobuzio-Donahue C.A.. **Genetic evolution of pancreatic cancer: Lessons learnt from the pancreatic cancer genome sequencing project**. *Gut* (2012) **61** 1085-1094. DOI: 10.1136/gut.2010.236026
22. Ijichi H., Chytil A., Gorska A.E., Aakre M.E., Fujitani Y., Fujitani S., Wright C.V., Moses H.L.. **Aggressive pancreatic ductal adenocarcinoma in mice caused by pancreas-specific blockade of transforming growth factor-beta signaling in cooperation with active Kras expression**. *Genes Dev.* (2006) **20** 3147-3160. DOI: 10.1101/gad.1475506
23. Izeradjene K., Combs C., Best M., Gopinathan A., Wagner A., Grady W.M., Deng C.X., Hruban R.H., Adsay N.V., Tuveson D.A., Hingorani S.R.. **Kras(G12D) and Smad4/Dpc4 haploinsufficiency cooperate to induce mucinous cystic neoplasms and invasive adenocarcinoma of the pancreas**. *Cancer Cell* (2007) **11** 229-243. DOI: 10.1016/j.ccr.2007.01.017
24. Jambhekar A., Dhall A., Shi Y.. **Roles and regulation of histone methylation in animal development**. *Nat. Rev. Mol. Cell Biol.* (2019) **20** 625-641. DOI: 10.1038/s41580-019-0151-1
25. Massagué J.. **TGFbeta in cancer**. *Cell* (2008) **134** 215-230. DOI: 10.1016/j.cell.2008.07.001
26. Massagué J., Seoane J., Wotton D.. **Smad transcription factors**. *Genes Dev.* (2005) **19** 2783-2810. DOI: 10.1101/gad.1350705
27. Parajuli P., Kumar S., Loumaye A., Singh P., Eragamreddy S., Nguyen T.L., Ozkan S., Razzaque M.S., Prunier C., Thissen J.P., Atfi A.. **Twist1 activation in muscle progenitor cells causes muscle loss akin to cancer cachexia**. *Dev. Cell* (2018) **45** 712-725.e6. DOI: 10.1016/j.devcel.2018.05.026
28. Parajuli P., Nguyen T.L., Prunier C., Razzaque M.S., Xu K., Atfi A.. **Pancreatic cancer triggers diabetes through TGF-β-mediated selective depletion of islet β-cells**. *Life Sci. Alliance* (2020) **3** 3. DOI: 10.26508/lsa.201900573
29. Parajuli P., Singh P., Wang Z., Li L., Eragamreddi S., Ozkan S., Ferrigno O., Prunier C., Razzaque M.S., Xu K., Atfi A.. **TGIF1 functions as a tumor suppressor in pancreatic ductal adenocarcinoma**. *EMBO J.* (2019) **38**. DOI: 10.15252/embj.2018101067
30. Park S.W., Davison J.M., Rhee J., Hruban R.H., Maitra A., Leach S.D.. **Oncogenic KRAS induces progenitor cell expansion and malignant transformation in zebrafish exocrine pancreas**. *Gastroenterology* (2008) **134** 2080-2090. DOI: 10.1053/j.gastro.2008.02.084
31. Pei D., Shu X., Gassama-Diagne A., Thiery J.P.. **Mesenchymal-epithelial transition in development and reprogramming**. *Nat. Cell Biol.* (2019) **21** 44-53. DOI: 10.1038/s41556-018-0195-z
32. Pickup M.W., Owens P., Moses H.L.. **TGF-Β, bone morphogenetic protein, and activin signaling and the tumor microenvironment**. *Cold Spring Harb. Perspect. Biol.* (2017) **9** 9. DOI: 10.1101/cshperspect.a022285
33. Pierreux C.E., Nicolás F.J., Hill C.S.. **Transforming growth factor beta-independent shuttling of Smad4 between the cytoplasm and nucleus**. *Mol. Cell. Biol.* (2000) **20** 9041-9054. DOI: 10.1128/MCB.20.23.9041-9054.2000
34. Pinheiro I., Margueron R., Shukeir N., Eisold M., Fritzsch C., Richter F.M., Mittler G., Genoud C., Goyama S., Kurokawa M.. **Prdm3 and Prdm16 are H3K9me1 methyltransferases required for mammalian heterochromatin integrity**. *Cell* (2012) **150** 948-960. DOI: 10.1016/j.cell.2012.06.048
35. Seale P., Bjork B., Yang W., Kajimura S., Chin S., Kuang S., Scimè A., Devarakonda S., Conroe H.M., Erdjument-Bromage H.. **PRDM16 controls a brown fat/skeletal muscle switch**. *Nature* (2008) **454** 961-967. DOI: 10.1038/nature07182
36. Seale P., Kajimura S., Yang W., Chin S., Rohas L.M., Uldry M., Tavernier G., Langin D., Spiegelman B.M.. **Transcriptional control of brown fat determination by PRDM16**. *Cell Metab.* (2007) **6** 38-54. DOI: 10.1016/j.cmet.2007.06.001
37. Seo S.R., Ferrand N., Faresse N., Prunier C., Abécassis L., Pessah M., Bourgeade M.F., Atfi A.. **Nuclear retention of the tumor suppressor cPML by the homeodomain protein TGIF restricts TGF-beta signaling**. *Mol. Cell* (2006) **23** 547-559. DOI: 10.1016/j.molcel.2006.06.018
38. Shimada I.S., Acar M., Burgess R.J., Zhao Z., Morrison S.J.. **Prdm16 is required for the maintenance of neural stem cells in the postnatal forebrain and their differentiation into ependymal cells**. *Genes Dev.* (2017) **31** 1134-1146. DOI: 10.1101/gad.291773.116
39. Stathis A., Moore M.J.. **Advanced pancreatic carcinoma: Current treatment and future challenges**. *Nat. Rev. Clin. Oncol.* (2010) **7** 163-172. DOI: 10.1038/nrclinonc.2009.236
40. Stine R.R., Sakers A.P., TeSlaa T., Kissig M., Stine Z.E., Kwon C.W., Cheng L., Lim H.W., Kaestner K.H., Rabinowitz J.D., Seale P.. **PRDM16 maintains homeostasis of the intestinal epithelium by controlling region-specific metabolism**. *Cell Stem Cell* (2019) **25** 830-845.e8. DOI: 10.1016/j.stem.2019.08.017
41. Sundqvist A., Morikawa M., Ren J., Vasilaki E., Kawasaki N., Kobayashi M., Koinuma D., Aburatani H., Miyazono K., Heldin C.H.. **JUNB governs a feed-forward network of TGFβ signaling that aggravates breast cancer invasion**. *Nucleic Acids Res.* (2018) **46** 1180-1195. DOI: 10.1093/nar/gkx1190
42. Takahata M., Inoue Y., Tsuda H., Imoto I., Koinuma D., Hayashi M., Ichikura T., Yamori T., Nagasaki K., Yoshida M.. **SKI and MEL1 cooperate to inhibit transforming growth factor-beta signal in gastric cancer cells**. *J. Biol. Chem.* (2009) **284** 3334-3344. DOI: 10.1074/jbc.M808989200
43. Tuveson D.A., Shaw A.T., Willis N.A., Silver D.P., Jackson E.L., Chang S., Mercer K.L., Grochow R., Hock H., Crowley D.. **Endogenous oncogenic K-ras(G12D) stimulates proliferation and widespread neoplastic and developmental defects**. *Cancer Cell* (2004) **5** 375-387. DOI: 10.1016/S1535-6108(04)00085-6
44. Westphalen C.B., Olive K.P.. **Genetically engineered mouse models of pancreatic cancer**. *Cancer J.* (2012) **18** 502-510. DOI: 10.1097/PPO.0b013e31827ab4c4
45. Whittle M.C., Izeradjene K., Rani P.G., Feng L., Carlson M.A., DelGiorno K.E., Wood L.D., Goggins M., Hruban R.H., Chang A.E.. **RUNX3 controls a metastatic switch in pancreatic ductal adenocarcinoma**. *Cell* (2015) **161** 1345-1360. DOI: 10.1016/j.cell.2015.04.048
46. Xue H., Yao T., Cao M., Zhu G., Li Y., Yuan G., Chen Y., Lei M., Huang J.. **Structural basis of nucleosome recognition and modification by MLL methyltransferases**. *Nature* (2019) **573** 445-449. DOI: 10.1038/s41586-019-1528-1
47. Yonezawa S., Higashi M., Yamada N., Goto M.. **Precursor lesions of pancreatic cancer**. *Gut Liver* (2008) **2** 137-154. DOI: 10.5009/gnl.2008.2.3.137
48. Zhang M.Z., Ferrigno O., Wang Z., Ohnishi M., Prunier C., Levy L., Razzaque M., Horne W.C., Romero D., Tzivion G.. **TGIF governs a feed-forward network that empowers Wnt signaling to drive mammary tumorigenesis**. *Cancer Cell* (2015) **27** 547-560. DOI: 10.1016/j.ccell.2015.03.002
49. Zhou B., Wang J., Lee S.Y., Xiong J., Bhanu N., Guo Q., Ma P., Sun Y., Rao R.C., Garcia B.A.. **PRDM16 suppresses MLL1r leukemia via intrinsic histone methyltransferase activity**. *Mol. Cell* (2016) **62** 222-236. DOI: 10.1016/j.molcel.2016.03.010
|
---
title: 'High hemoglobin glycation index is associated with increased risk of diabetes:
A population-based cohort study in China'
authors:
- Lu Lin
- Anping Wang
- Xiaomeng Jia
- Haibin Wang
- Yan He
- Yiming Mu
- Jingtao Dou
journal: Frontiers in Endocrinology
year: 2023
pmcid: PMC9999023
doi: 10.3389/fendo.2023.1081520
license: CC BY 4.0
---
# High hemoglobin glycation index is associated with increased risk of diabetes: A population-based cohort study in China
## Abstract
### Purpose
The hemoglobin glycation index (HGI) quantifies the mismatch between glycated hemoglobin A1c and average glycemia among individuals. Currently, it is unknown the potential role of HGI in exhaustively evaluating the progression of glucose metabolism/the risk of developing diabetes mellitus. Therefore, this study aimed to investigate the association between HGI and the risk of incident diabetes.
### Methods
A total of 7,345 participants aged at least 40 years and without diabetes were divided into three groups according to the tertile of their baseline HGI level and followed for a median of 3.24 years to track new-onset diabetes. Using multivariate Cox regression analyses, we explored the association between the HGI, both categorized and continuous, and incident diabetes.
### Results
During follow-up, 742 subjects (263 males and 479 females) developed diabetes mellitus. Higher HGI was associated with an increased risk of diabetes, even when adjusted for confounding factors, and every standard deviation increase in HGI was associated with a significant risk increase of $30.6\%$ for diabetes (hazard ratio 1.306, $95\%$ confidence interval 1.232–1.384).
### Conclusions
Participants with a higher HGI were at a higher risk of future diabetes, irrespective of their glycemic conditions. Consequently, HGI may be employed to identify individuals at high risk for diabetes.
## Introduction
During the past two decades in clinical practice, glycated hemoglobin (HbA1c) has been universally applied for screening, diagnosis, and glycemic control monitoring in patients with diabetes [1, 2]. HbA1c is considered an important biomarker indicating average blood glucose levels over a period of 8–12 weeks [3]. However, HbA1c is not a one-size-fits-all indicator for assessing chronic glycemia, which ignores inter-individual variations in the relationship between HbA1c and average glucose [4, 5]. In 2002, Hempe et al. developed and validated the hemoglobin glycation index (HGI) to quantify the inter-individual consistent disparity between HbA1c and the mean blood glucose (MBG) level [6].
The HGI is a metric describing inter-individual biological variation of HbA1c or individual propensity for glycation of hemoglobin, which is another major factor affecting HbA1c results besides blood glucose concentration [3, 7]. The HGI is computed as measured HbA1c minus predicted HbA1c. The predicted HbA1c was initially calculated by inserting the date-matched MBG into a linear regression equation derived from the measured HbA1c and MBG. Some studies have confirmed that using fasting plasma glucose (FPG) to assess predicted HbA1c and calculate HGI is feasible [8, 9]. Most prior studies have primarily reported the association between HGI and diabetes complications. An elevated HGI might promote diabetes complications through inflammation and the formation of advanced glycation end products (AGEs) [10, 11]. Existing literature suggests that the HGI is a strong predictor of the risk of diabetes complications among patients with either type 1 or type 2 diabetes [12, 13].
However, only a few studies have investigated the impact of HGI on glucose metabolism. A recent study in Italy demonstrated that patients without diabetes with a high HGI had higher fasting insulin levels and more severe insulin resistance than those with a low HGI [14]. Moreover, a higher HGI appears to be related to older age, obesity, and dyslipidemia, which are risk factors for diabetes [10, 15, 16]. Based on previous studies (10, 14–16), we hypothesized that individuals with a high HGI might have an increased risk of developing diabetes. Therefore, the current study aimed to investigate the association between the HGI and the incidence of diabetes among a population in China, using a prospective cohort study design.
## Study design and participants
The Risk Evaluation of Cancers in Chinese Diabetic Individuals: A longitudinal (REACTION) study was a multicenter population-based prospective cohort study investigating the association between abnormal glucose metabolism and increased risk of cancer among the population of China. The experimental design of the REACTION study has been published elsewhere [17, 18].
This present study was retrospective, and analysis was performed based on the data from a subcenter (the Pingguoyuan community of Beijing) of the REACTION study, which was conducted by the Department of Endocrinology of First Medical Center of Chinese PLA General Hospital. From January to August 2012, 10,126 individuals aged ≥40 years were recruited from the study location, of whom 2,749 participants did not meet the inclusion criteria and were therefore excluded at the baseline visit. In total, 7,467 eligible participants without diabetes were consecutively enrolled and followed up from April to October 2015. Of the 7,467 participants, 82 ($1.1\%$) were lost to follow-up; thus, the study achieved a response rate of $98.6\%$. In addition, 40 patients with missing data on FPG, 2-h plasma glucose (2hPG), or HbA1c at follow-up were excluded. Consequently, 7,345 participants were included in the final evaluation (Figure 1). This REACTION study program was approved by the Medical Ethics Committee of Shanghai Jiaotong University (No. 2011-14). All procedures were performed according to the tenets of the revised [1983] Declaration of Helsinki. All participants provided written informed consent at each study visit.
**Figure 1:** *Study flowchart of participants. FPG, fasting plasma glucose; 2hPG, 2-h plasma glucose; HbA1c, hemoglobin A1c; HGI, hemoglobin glycation index; OGTT, oral glucose tolerance test.*
## Data collection and laboratory measurements
Following the standardized study protocol, data were collected via detailed questionnaires, physical examination, and blood samples during baseline and follow-up visits at a local clinic. Trained clinical staff administered standardized questionnaires and conducted in-person interviews to collect information on demographic data and medical, family, and medication history. Blood pressure, height, body weight, and waist circumference were measured according to standard procedures, and body mass index (BMI) was calculated as weight (kg)/height squared (m2).
At each clinic visit, a 75-g oral glucose tolerance test (OGTT) was performed after collecting a fasting blood sample. FPG, HbA1c, fasting lipids, and 2hPG levels were also measured. After collecting venous blood, all samples were immediately placed on ice to maintain stability. Thereafter, the samples were instantly transported to the laboratory at the First Medical Centre of Chinese PLA General Hospital and processed within 2 h of blood collection. For plasma glucose (including FPG and 2hPG), blood samples were collected in tubes containing sodium fluoride and measured using the hexokinase method. HbA1c was measured using high-performance liquid chromatography (VARIANT II system, Bio-Rad, Hercules, CA). Total cholesterol (TC), triglycerides (TG), high-density lipoprotein cholesterol (HDL-C), and low-density lipoprotein cholesterol (LDL-C) levels were determined using an auto-analyzer (ARCHITECT c16000 System; Abbott Laboratories, Chicago, IL). The quality control protocol for laboratory assays has been published in detail elsewhere [17].
The diagnosis of dysglycemia (including pre-diabetes or diabetes) was based on OGTT, conforming to the American Diabetes *Association criteria* [1]. Pre-diabetes was defined as follows: FPG: 100–125 mg/dL (5.6–6.9 mmol/L); or 2hPG during 75 g OGTT: 140–199 mg/dL (7.8–11.0 mmol/L). Diabetes was defined as: documented diagnosis of diabetes in medical records or taking glucose-lowering medications; FPG ≥126 mg/dL (7.0 mmol/L); or 2hPG ≥200 mg/dL (11.1 mmol/L) during 75g OGTT. Normoglycemia was described as FPG <100 mg/dl (5.6 mmol/L) with 2hPG <140 mg/dl (7.8 mmol/L) during 75g OGTT. The primary study outcome was the occurrence of diabetes, defined as diagnosed (i.e., physician-diagnosed diabetes or use of antidiabetic medication during follow-up) or undiagnosed (based on the above diabetes criteria).
## Calculation of the HGI
A linear regression equation between FPG and HbA1c was established using data from the baseline population of 8,475 participants who met the inclusion criteria (Figure 2). The predicted HbA1c level was calculated by imputing FPG into the following equation: HbA1c (%) = 3.335 + 0.025 FPG (mg/dL). Then, the baseline HGI was calculated by subtracting the measured HbA1c from the predicted HbA1c level [8, 9]. Tertile cut-off points were determined from the baseline HGI and applied to the study subset of participants without diabetes ($$n = 7$$,345). Subsequently, the final study cohort was classified into low (≤ -0.1480), moderate (-0.1480 to 0.1675), and high (>0.1675) HGI groups.
**Figure 2:** *Scattergram of HbA1c versus FPG. There is a clear linear relation between HbA1c and FPG [HbA1c (%) = 3.335 + 0.025 FPG (mg/dL), P <0.001, R2 = 0.528]. HbA1c: hemoglobin A1c; FPG: fasting plasma glucose.*
## Statistical analysis
Continuous variables were expressed as mean ± standard deviation (SD) for normally distributed variables or as medians (interquartile range) for non-normally distributed variables. Group comparisons were analyzed using a one-way analysis of variance or the Kruskal–Wallis rank sum test for normally or non-normally distributed data, respectively. Proportions for categorical variables were compared using the chi-squared test or Fisher’s exact test. Subgroup analyses were also performed by baseline glucometabolic status (normoglycemia and pre-diabetes based on OGTT results).
The time to diabetes event was defined as the time from baseline to the reported date of diagnosis obtained at the follow-up visit. For incident diabetes cases diagnosed at the follow-up visit, we inserted the estimated time to diabetes event (using baseline and follow-up values) as the first date at which one of the values met the diagnostic criteria. For instance, if a participant was diagnosed with diabetes using an FPG of 9 mmol/L at a follow-up visit and their FPG at baseline was 5 mmol/L, the incident date was set as half the time of follow-up ([7 – 5]/[9 – 5] = $\frac{2}{4}$ = 0.5). The relationship between the incidence of diabetes and the HGI was examined using multivariate Cox proportional hazards regression analyses. The test for trend was performed to ascertain the statistical significance of the trends observed. In all models, covariates were chosen for both their clinical relevance and univariate correlation with the outcome of interest. Owing to the collinearity between the HGI and HbA1c ($r = 0.694$), none of the models included both the HGI and HbA1c. In addition, the HGI and HbA1c were evaluated per SD change to facilitate a comparison of their strengths in association with the incidence of diabetes. Hazard ratios (HRs) were reported with corresponding $95\%$ confidence intervals (CIs), and statistical significance was set at $p \leq 0.05$ (two-sided). The Statistical analyses were performed using the Statistical Package for the Social Sciences (version 26.0; IBM, Armonk, NY).
## Baseline general characteristics of patients in the three HGI groups
The baseline characteristics of the study population according to the HGI category are shown in Table 1. Among the 7,345 participants, the mean (SD) baseline age was 56.5 (7.6) years, and 4,885 ($66.5\%$) were female. Participants with high HGI were more likely to be older, have a higher BMI, and have an adverse lipid profile (all P for trend <0.001). The same trends were also observed among the normal glycemia and pre-diabetes subgroups (Tables S1, S2).Even though more participants in the high HGI group had hypertension, diastolic blood pressure (DBP) decreased with HGI tertile (P for trend <0.001). There was a stepwise increase in the HbA1c levels in the order of increasing HGI levels owing to the method used to calculate the HGI; however, FPG levels decreased. Compared with low and moderate HGI groups, participants in the high HGI group had significantly higher 2hPG levels in the total study population and pre-diabetes subgroup (Tables 1, S2).
**Table 1**
| Variables | Total | Low HGI | Moderate HGI | High HGI | Statistical value | P value |
| --- | --- | --- | --- | --- | --- | --- |
| Participants, n | 7, 345 | 2, 457 | 2, 544 | 2, 344 | | |
| Mean HGI | 0.0010 (-0.2370, 0.2402) | -0.3370 (-0.4920, -0.2357) | 0.0075 (-0.0680, 0.0839) | 0.3597(0.2540, 0.5164) | | |
| Age (years) | 56.5 ± 7.6 | 55.7 ± 7.4 | 56.6 ± 7.7 | 57.4 ± 7.6 | 28.553 | <0.001 |
| Female, n (%) | 4885(66.5) | 1491 (60.7) | 1773 (69.7) | 1621 (69.2) | 56.38 | <0.001 |
| BMI (kg/m2) | 25.5 ± 3.4 | 25.2 ± 3.3 | 25.5 ± 3.3 | 25.8 ± 3.5 | 19.858 | <0.001 |
| Waist circumference (cm) | 83.0 ± 8.8 | 83.1 ± 8.7 | 82.8 ± 8.7 | 83.2 ± 9.0 | 1.12 | 0.298 |
| SBP (mmHg) | 130.3 ± 16.1 | 130.5 ± 16.2 | 129.7 ± 16.0 | 130.8 ± 16.1 | 3.265 | 0.038 |
| DBP (mmHg) | 75.5 ± 9.5 | 76.7 ± 9.6 | 75.2 ± 9.4 | 74.7 ± 9.5 | 31.933 | <0.001 |
| Family history of diabetes, n (%) | 1772 (24.1) | 595(24.2) | 625 (24.6) | 552(23.5) | 1.049 | 0.902 |
| History of hypertension, n (%) | 1971 (26.8) | 653(26.6) | 647(25.4) | 671 (28.6) | 6.463 | 0.039 |
| Antihypertensive medication, n (%) | 1626 (22.1) | 524 (21.3) | 541 (21.3) | 561 (23.9) | 6.445 | 0.040 |
| Lipid-lowering medication, n (%) | 195 (2.7) | 36 (1.5) | 83 (3.3) | 76 (3.2) | 20.221 | <0.001 |
| TC (mmol/L) | 5.23 ± 0.98 | 5.07 ± 0.93 | 5.27 ± 1.00 | 5.35 ± 0.98 | 51.129 | <0.001 |
| TG (mmol/L) | 1.28 (0.91, 1.80) | 1.22 (0.88, 1.71) | 1.27 (0.90, 1.82) | 1.33 (0.95, 1.87) | 27.497 | <0.001 |
| HDL-C (mmol/L) | 1.40 (1.18, 1.66) | 1.40 (1.18, 1.66) | 1.42 (1.20, 1.66) | 1.40 (1.18, 1.68) | 2.05 | 0.359 |
| LDL-C (mmol/L) | 3.20 ± 0.80 | 3.08 ± 0.78 | 3.22 ± 0.79 | 3.29 ± 0.81 | 44.794 | <0.001 |
| HbA1c (%) | 5.8 ± 0.5 | 5.4 ± 0.3 | 5.8 ± 0.2 | 6.2 ± 0.4 | 2537.421 | <0.001 |
| FPG (mg/dL) | 97 ± 9 | 99 ± 10 | 96 ± 9 | 95 ± 9 | 87.036 | <0.001 |
| 2hPG (mg/dL) | 123 ± 29 | 122 ± 29 | 122 ± 29 | 125 ± 31 | 8.899 | <0.001 |
| Pre-diabetes based on OGTT, n (%) | 3239 (44.1) | 1221 (49.7) | 1035 (40.7) | 983 (41.9) | 46.689 | <0.001 |
## Association between baseline HGI and incident diabetes
During the follow-up period (median 3.24 years), 742 participants ($10.1\%$) developed diabetes, with the highest incidence in the high HGI group ($13.1\%$), followed by the moderate ($9.1\%$) and low groups ($8.3\%$). Similarly, the same results were observed in the subgroup analyses by baseline glucometabolic status (Figure 3)
**Figure 3:** *Incidence of diabetes for HGI categories by baseline glucometabolic status # Glucometabolic status is assessed using oral glucose tolerance test results (including fasting plasma glucose and 2-h plasma glucose). *compared with low HGI group, P <0.05 HGI: hemoglobin glycation index.*
To assess for relationships between variables and diabetes incidence, HGI was analyzed both as a categorical and a continuous variable. Table 2 shows the adjusted HRs for incident diabetes by the HGI group, with the low HGI group as the reference. Multivariate analysis revealed that high HGI was associated with an increased risk of diabetes (models 1 and 2) after adjusting for age, sex, BMI, systolic blood pressure (SBP), DBP, history of hypertension, antihypertensive medication use, TC, TG, and LDL-C. Further adjustment for FPG and 2hPG (model 3) strengthened this association, with adjusted HRs for diabetes reaching 1.366 ($95\%$ CI, 1.129-1.653) and 1.903 ($95\%$ CI, 1.584–2.288) for the moderate and top tertiles of HGI, respectively, when compared with the bottom tertile (P for trend <0.001).
**Table 2**
| Group | Case/overall | Crude model | Model 1(HR 95% CI) | Model 2(HR 95% CI) | Model 3(HR 95% CI) |
| --- | --- | --- | --- | --- | --- |
| Low HGI | 203/2457 | Reference | Reference | Reference | Reference |
| Moderate HGI | 232/2544 | 1.105(0.915-1.334) | 1.069(0.885-1.292) | 1.048(0.867-1.267) | 1.366(1.129-1.653) |
| High HGI | 307/2344 | 1.624(1.360-1.939) | 1.539(1.286-1.840) | 1.468(1.226-1.757) | 1.903(1.584-2.288) |
| P for trend | | <0.001 | <0.001 | <0.001 | <0.001 |
Moreover, when considering the HGI as a continuous variable, every 1-SD increase resulted in a $30.6\%$ increase in the risk of diabetes. In contrast, when the HbA1c level was used, every 1-SD increase resulted in a $32.8\%$ increase in the risk of diabetes (Figure 4), and there were no differences between HGI and HbA1c with an overlap in the $95\%$ CIs in the adjusted HRs.
**Figure 4:** *Forest plot of HRs for incident diabetes per SD difference in HGI and HbA1c. Circles represent HGI, and squares represent HbA1c in different Cox regression models. Model 1 was adjusted for age and sex. Model 2 was adjusted for the variables in model 1 plus body mass index, systolic blood pressure, diastolic blood pressure, hypertension, antihypertensive medication, total cholesterol, triglycerides, and low-density lipoprotein cholesterol. Model 3 was adjusted for all variables in model 2 plus the baseline fasting plasma glucose and 2-h plasma glucose at the OGTT. CI, confidence interval; HbA1c, hemoglobin A1c; HGI, hemoglobin glycation index; HR, hazard ratios; OGTT, oral glucose tolerance test.*
## Discussion
In this prospective population-based cohort study, we found that regardless of baseline glycemic status, participants with higher HGI values were at a greater risk of incident diabetes during the 3-year follow-up period. This association remained significant even after adjustment for potential confounders of diabetes. Our study contributes new knowledge about the relative importance of inter-individual biological variation of HbA1c on predicting the risk of incident diabetes.
Pervious investigators have reported that two major factors determined HbA1c levels: the mean blood glucose level over an 8- to 12-week period and biological variation in HbA1c, as assessed by the HGI [3, 6, 7]. The HGI quantifies the persistent individual discrepancy between HbA1c values and similar blood glucose levels, which is a good indicator to investigate the clinical implications of biological variation in HbA1c. To date, several reports have shown that a high HGI is significantly associated with an increased risk of developing both diabetic microvascular and macrovascular complications [12, 13, 19, 20].
However, few studies have focused on the implications of the HGI for the predictive value of incident diabetes. In this study, we used baseline HbA1c and FPG levels to show that a high HGI at baseline was related to a high risk of incident diabetes. At the 3-year follow-up, participants in the high and moderate HGI groups had a 1.903- and 1.366-fold greater risk of diabetes, respectively, than those in the low HGI group. Thus, individuals with a high HGI (i.e., high propensity for glycation) have a greater susceptibility to future diabetes at similar baseline glucose levels than those with a low HGI. Notably, these findings have ruled out the confounding influence of glycemia, which could underestimate or mask the true effect of the HGI on diabetes risk. To the best of our knowledge, this is the first study to explore the impact of high HGI on the progression of hyperglycemia.
When evaluating HGI as a continuous variable, every 1-SD increase was associated with a $30.6\%$ increase in the risk of incident diabetes; using HbA1c instead of the HGI as a predictive variable to compare the effect on diabetes occurrence corroborated this result. After adjustment for identical potential confounders (especially blood glucose levels), every 1-SD increase in HbA1c resulted in a significant $32.8\%$ increase in the risk of future diabetes, which was consistent with the result of the HGI assessment. The adjusted HRs of the HGI and HbA1c values were approximately the same, which supported the concept that the predictive value of the HGI reflected the non-glycemic predictive value of HbA1c. That is, except for the predictive value of glycemic measures, an individual’s propensity for glycation also has distinct effects on the development and progression of diabetes. The underlying mechanism linking a high HGI and glycemic disorders has not yet been fully elucidated. A high HGI, representing an increased degree of intracellular non-enzymatic glycosylation, is correlated with inflammation and the generation of AGEs [10, 11], both of which contribute to insulin resistance and impaired beta-cell function, thus causing elevated blood glucose levels (21–23). However, hyperglycemia could also cause chronic inflammation and increased generation of AGEs (24–26), thereby establishing a vicious cycle. In brief, a high HGI identified individuals with increased susceptibility to diabetes and associated complications at similar blood glucose concentrations.
Previous studies have reported that individuals with a high HGI exhibit similar clinical traits, including older age, a higher BMI, dyslipidemia, and postprandial glycemic excursion (9, 15, 27–29). However, the observations from different studies are not completely concordant. These discrepancies might be attributed to the different geographical populations with different study inclusion criteria and different methods of calculating HGI. Accordingly, further studies are needed to unify and standardize the calculating method of HGI and confirm which clinical features are most highly associated with higher HGI.
HGI is a marker with biological variation in HbA1c or an individual’s propensity for glycation, combined with HbA1c, which might provide a better assessment of a patient’s risk for glycemic progression. This prospective cohort study showed that high HGI was associated with an increased risk for diabetes. Therefore, our findings will facilitate work to understand the linkage between an individual’s high propensity for glycation and predisposition to hyperglycemia.
The potential limitations of our study are as follows: First, the diagnosis of diabetes was based on a single measurement of glucose (FPG or 2hPG during OGTT) or HbA1c, which may lead to misdiagnosis or misclassification for some patients without classic symptoms of hyperglycemia. Second, the study participants originated from a single center in the REACTION study; thus, the results may not be directly generalizable, and further studies are required to confirm our findings.
In summary, our study findings suggest that inter-individual differences in the propensity for glycation assessed by the HGI are associated with the risk of incident diabetes. Hence, more attention should be given to the development and progression of glycemic abnormalities in participants with a high HGI. In addition, the application of HbA1c to assess average glycemia in considering individual HGI levels may help to individualize preventive and therapeutic decision-making. Also, in constructing a population-based prediction model, the HGI measurement may be considered an additional parameter to improve the risk stratification of patients with diabetes. Consequently, future work should focus on investigating the underlying mechanism of how the HGI involves the pathogenesis of diabetes and how to reduce the risk of developing diabetes in a population with high HGIs.
## Data availability statement
The datasets presented in this article are not readily available because of individual privacy reasons, but are available after anonymization from the corresponding author on reasonable request. Requests to access the datasets should be directed to [email protected].
## Ethics statement
The studies involving human participants were reviewed and approved by the Medical Ethics Committee of Shanghai Jiaotong University (No. 2011-14). The patients/participants provided their written informed consent to participate in this study.
## Author contributions
LL and JD contributed to this study’s concept, design, implementation, and rationale. LL performed the statistical analysis with advice from YH. LL drafted and revised the manuscript. JD supervised the study, and revised the paper. All authors contributed to the acquisition, analysis, and interpretation of the data and the review and edits of the drafts. JD is responsible for the integrity of the work as a whole. All authors contributed to the article and approved the submitted version.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fendo.2023.1081520/full#supplementary-material
## References
1. **2. classification and diagnosis of diabetes: Standards of medical care in diabetes-2021**. *Diabetes Care* (2021) **44** S15-s33. DOI: 10.2337/dc21-S002
2. **Guideline for the prevention and treatment of type 2 diabetes mellitus in China (2020 edition)**. *Chin J Diabetes Mellitus* (2021) **13** 315-409. DOI: 10.3760/cma.j.cn115791-20210221-00095
3. Sacks DB. **A1c versus glucose testing: A comparison**. *Diabetes Care* (2011) **34**. DOI: 10.2337/dc10-1546
4. Yudkin JS, Forrest RD, Jackson CA, Ryle AJ, Davie S, Gould BJ. **Unexplained variability of glycated haemoglobin in non-diabetic subjects not related to glycaemia**. *Diabetologia* (1990) **33**. DOI: 10.1007/bf00404798
5. Nayak AU, Holland MR, Macdonald DR, Nevill A, Singh BM. **Evidence for consistency of the glycation gap in diabetes**. *Diabetes Care* (2011) **34**. DOI: 10.2337/dc10-1767
6. Hempe JM, Gomez R, McCarter RJ, Chalew SA. **High and low hemoglobin glycation phenotypes in type 1 diabetes: A challenge for interpretation of glycemic control**. *J Diabetes Complications* (2002) **16**. DOI: 10.1016/s1056-8727(01)00227-6
7. McCarter RJ, Hempe JM, Chalew SA. **Mean blood glucose and biological variation have greater influence on Hba1c levels than glucose instability: An analysis of data from the diabetes control and complications trial**. *Diabetes Care* (2006) **29**. DOI: 10.2337/diacare.29.02.06.dc05-1594
8. Hempe JM, Liu S, Myers L, McCarter RJ, Buse JB, Fonseca V. **The hemoglobin glycation index identifies subpopulations with harms or benefits from intensive treatment in the accord trial**. *Diabetes Care* (2015) **38**. DOI: 10.2337/dc14-1844
9. Hsia DS, Rasouli N, Pittas AG, Lary CW, Peters A, Lewis MR. **Implications of the hemoglobin glycation index on the diagnosis of prediabetes and diabetes**. *J Clin Endocrinol Metab* (2020) **105**. DOI: 10.1210/clinem/dgaa029
10. Liu S, Hempe JM, McCarter RJ, Li S, Fonseca VA. **Association between inflammation and biological variation in hemoglobin A1c in U.S. nondiabetic adults**. *J Clin Endocrinol Metab* (2015) **100**. DOI: 10.1210/jc.2014-4454
11. Felipe DL, Hempe JM, Liu S, Matter N, Maynard J, Linares C. **Skin intrinsic fluorescence is associated with hemoglobin a(1c)and hemoglobin glycation index but not mean blood glucose in children with type 1 diabetes**. *Diabetes Care* (2011) **34**. DOI: 10.2337/dc11-0049
12. McCarter RJ, Hempe JM, Gomez R, Chalew SA. **Biological variation in Hba1c predicts risk of retinopathy and nephropathy in type 1 diabetes**. *Diabetes Care* (2004) **27**. DOI: 10.2337/diacare.27.6.1259
13. van Steen SC, Woodward M, Chalmers J, Li Q, Marre M, Cooper ME. **Haemoglobin glycation index and risk for diabetes-related complications in the action in diabetes and vascular disease: Preterax and diamicron modified release controlled evaluation (Advance) trial**. *Diabetologia* (2018) **61**. DOI: 10.1007/s00125-017-4539-1
14. Marini MA, Fiorentino TV, Succurro E, Pedace E, Andreozzi F, Sciacqua A. **Association between hemoglobin glycation index with insulin resistance and carotid atherosclerosis in non-diabetic individuals**. *PloS One* (2017) **12**. DOI: 10.1371/journal.pone.0175547
15. Lin L, Wang A, He Y, Wang W, Gao Z, Tang X. **Effects of the hemoglobin glycation index on hyperglycemia diagnosis: Results from the reaction study**. *Diabetes Res Clin Pract* (2021) 109039. DOI: 10.1016/j.diabres.2021.109039
16. Fiorentino TV, Marini MA, Succurro E, Andreozzi F, Sciacqua A, Hribal ML. **Association between hemoglobin glycation index and hepatic steatosis in non-diabetic individuals**. *Diabetes Res Clin Pract* (2017) **134** 53-61. DOI: 10.1016/j.diabres.2017.09.017
17. Bi Y, Lu J, Wang W, Mu Y, Zhao J, Liu C. **Cohort profile: Risk evaluation of cancers in Chinese diabetic individuals: A longitudinal (Reaction) study**. *J Diabetes* (2014) **6**. DOI: 10.1111/1753-0407.12108
18. Ning G, Bloomgarden Z. **Diabetes and cancer: Findings from the reaction Study:Reaction**. *J Diabetes* (2015) **7**. DOI: 10.1111/1753-0407.12240
19. Kim MK, Jeong JS, Yun JS, Kwon HS, Baek KH, Song KH. **Hemoglobin glycation index predicts cardiovascular disease in people with type 2 diabetes mellitus: A 10-year longitudinal cohort study**. *J Diabetes its Complications* (2018) **32**. DOI: 10.1016/j.jdiacomp.2018.08.007
20. Wu JD, Liang DL, Xie Y, Chen MY, Chen HH, Sun D. **Association between hemoglobin glycation index and risk of cardiovascular disease and all cause mortality in type 2 diabetic patients: A meta-analysis**. *Front Cardiovasc Med* (2021) **8**. DOI: 10.3389/fcvm.2021.690689
21. Ighodaro OM. **Molecular pathways associated with oxidative stress in diabetes mellitus**. *BioMed Pharmacother* (2018) **108**. DOI: 10.1016/j.biopha.2018.09.058
22. Donath MY. **Targeting inflammation in the treatment of type 2 diabetes: Time to start**. *Nat Rev Drug Discovery* (2014) **13**. DOI: 10.1038/nrd4275
23. Song F, Schmidt AM. **Glycation and insulin resistance: Novel mechanisms and unique targets**. *Arterioscler Thromb Vasc Biol* (2012) **32**. DOI: 10.1161/atvbaha.111.241877
24. Pezhman L, Tahrani A, Chimen M. **Dysregulation of leukocyte trafficking in type 2 diabetes: Mechanisms and potential therapeutic avenues**. *Front Cell Dev Biol* (2021) **9**. DOI: 10.3389/fcell.2021.624184
25. Makita Z, Vlassara H, Rayfield E, Cartwright K, Friedman E, Rodby R. **Hemoglobin-age: A circulating marker of advanced glycosylation**. *Science* (1992) **258**. DOI: 10.1126/science.1411574
26. Friedman EA. **Advanced glycosylated end products and hyperglycemia in the pathogenesis of diabetic complications**. *Diabetes Care* (1999) **22 Suppl 2**
27. Hempe JM, Yang S, Liu S, Hsia DS. **Standardizing the haemoglobin glycation index**. *Endocrinol Diabetes Metab* (2021) **4**. DOI: 10.1002/edm2.299
28. Nagayama D, Watanabe Y, Yamaguchi T, Saiki A, Shirai K, Tatsuno I. **High hemoglobin glycation index is associated with increased systemic arterial stiffness independent of hyperglycemia in real-world Japanese population: A cross-sectional study**. *Diabetes Vasc Dis Res* (2020) **17**. DOI: 10.1177/1479164120958625
29. Yoo JH, Kang YM, Cho YK, Lee J, Jung CH, Park JY. **The haemoglobin glycation index is associated with nonalcoholic fatty liver disease in healthy subjects**. *Clin Endocrinol* (2019) **91**. DOI: 10.1111/cen.14001
|
---
title: The Effect of Intraabdominal Visceral and Subcutaneous Adipose Volume and Muscle
Volume on Lumbar Vertebrae Degeneration
journal: Cureus
year: 2023
pmcid: PMC9999032
doi: 10.7759/cureus.35940
license: CC BY 3.0
---
# The Effect of Intraabdominal Visceral and Subcutaneous Adipose Volume and Muscle Volume on Lumbar Vertebrae Degeneration
## Abstract
Objectives: This study aimed to investigate the effect of the volume of subcutaneous, visceral, and total adipose tissue, and paravertebral muscles in patients with lumbar vertebrae degeneration (LVD) through computerized tomography (CT) images.
Materials and methods: One forty-six patients with a complaint of lower back pain (LBP) between January 2019 and December 2021 were included in the study. CT scans of all patients were analyzed retrospectively for abdominal visceral, subcutaneous, and total fat volume, and also paraspinal muscle volume measurements and analysis of lumbar vertebrae degeneration (LVD) using designated software. In CT images, each intervertebral disc space was evaluated in terms of the presence of osteophytes, loss of disc height, sclerosis in the end plates, and spinal stenosis to investigate the presence of degeneration. Each level was scored according to the presence of findings, with 1 point for each finding. The total score at all levels (L1-S1) was calculated for each patient.
Results: An association was observed between the loss of intervertebral disc height and the amount of visceral, subcutaneous, and total fat volume at all lumbar levels (p˂0.05). The amount of all fat volume measurements also showed association with osteophyte formation (p˂0.05). An association was found between sclerosis and the amount of all fat volume at all lumbar levels (p˂0.05). It was observed that spinal stenosis at the lumbar levels was not associated with the amount of fat (total, visceral, subcutaneous) at any level (p˃0.05). No association was found between the amount of adipose and muscle volumes and vertebral pathologies at any level (p˃0.05).
Conclusion: The abdominal visceral, subcutaneous, and total fat volumes are associated with lumbar vertebral degeneration and loss of disc height. Paraspinal muscle volume does not show an association with vertebral degenerative pathologies.
## Introduction
Obesity is a worldwide problem regarding its increased risk for cardiovascular diseases, stroke, diabetes, cancer, asthma, and metabolic syndrome. It also causes psychosocial disorders, decreased productivity, and economic healthcare burden [1].
Obesity has been recently admitted as a risk factor for lower back pain (LBP) which decreases physical functions, compromises the quality of life, and causes psychological distress [2]. Therefore, the etiology of vertebral disc degeneration is clinically significant. Body mass index (BMI) has been blamed for vertebral disc degeneration among both adolescents and adults [3].
Lumbar vertebrae degeneration (LVD) is a prolonged process of deterioration involving genetically determined and mechanically triggered biological factors [4]. The proceeding phase of the degenerative process is segmental dysfunction and primarily shows impairment in facet joint functions. Although aging is considered to be the only significant contributor to the process, some factors such as inflammation may have a predisposing effect on LVD [5]. As a result of the degeneration, pain, inflammation, and hypomobility originating from the facet joints begin, and the movement segment is restricted [6]. Inflammation may not only emerge as a restriction but also both pain and hypomobility altogether. Hence, cells or tissues with increasing or emerging inflammation have been investigated as a potential risk factor for LVD [7]. In some community-based general studies, higher rates of back pain and disability were detected in individuals with more fat mass, whereas those with higher lean tissue volume had no association with back pain intensity [8]. Moreover, the increased adipose volume has been shown to be associated with the risk of type 2 Modic changes in the spine resulting in back pain, which tends to have a lean mass-protective effect [9].
The relationship between fat mass in the lumbar region and intervertebral disc diseases has been reported in the literature before, but the mechanism remains unclear [10]. Not only BMI but also excessive abdominal fat mass has been associated with lumbar pathologies. There is limited information in the literature about the relationship between subcutaneous and visceral abdominal fat distribution and lumbar vertebrae pathologies [10].
Detection of adipose tissue volume and adiposity varies according to the possibilities of the researchers and the conditions provided. Methods such as densitometry, MRI, and CT are costly although they can present clear results about body fat [11]. In a variety of studies, all these measurement methods were used for the analysis of adipose tissue [2,12,13]. CT shows this complex region's bone anatomy very well and is accepted as one of the best radiological techniques for adipose tissue volume calculations [14]. Since the muscle mass is highest at L3 and L4 levels, the region that is frequently preferred in these measurements is the L3-L4 region [15]. It has also been shown in the literature that visceral fat tissue measured on a single-slice CT scan at the L4 level is significantly associated with total abdominal visceral fat volume [16].
The effects of abdominal fat tissue volume on the spinal canal and vertebrae are still unknown and a comprehensive study on this subject has not been observed in the literature. In the current study, we investigated the effect of subcutaneous adipose tissue volume, visceral adipose tissue volume, and paravertebral muscle mass on LVD through CT images of the L1-S1 vertebral levels.
## Materials and methods
Following the institutional review board approval for the study (number: $\frac{119}{2019}$; Muğla Sıtkı Koçman University Ethical Committee), a retrospective cohort analysis was performed using the medical records of patients. For the current study, patient consent is not required. All procedures executed involving human participants were in accordance with the ethical standards of the institutional ethical committee and with the 1964 Helsinki declaration.
A total of 146 patients who applied to the neurosurgery outpatient clinic with a recent abdominal CT (max three months) because of a lower back pain complaint were included in the study. Patients with a previous history of surgery or a vertebral fracture were excluded. After excluded patients, a total of 146 patients were included in the study, of whom 90 were female ($61.6\%$) and 56 were male ($38.4\%$). The mean age of the patients was 51.42±13.91 [20-82] years.
Lumbar vertebra CT scans of all patients were reviewed retrospectively. CT images at the level from L3-L4 intervertebral disc were analyzed for body composition of fat tissue and muscle mass volume through the dedicated CT software (Syngo.via, SOMATOM Definition Flash: Siemens Healthcare, Forchheim, Germany). The L3-L4 level was selected in sagittal reformat CT images with the software (Figure 1).
**Figure 1:** *Determination of axial image at L3-L4 level and detection of region growing area on sagittal reformat images.*
The density range of -200, -40 HU was selected for the fat density measurement in the cross-section with the "region grooving" application in the angled axial images obtained parallel to the disc plane at this level. First, the fat volume in the whole section was measured (visceral and subcutaneous). Then, only the visceral adipose tissue volume was calculated by drawing borders to exclude subcutaneous adipose tissue (Figure 2). The subcutaneous fat tissue volume was obtained by subtracting the visceral fat tissue volume from the total fat volume (Figure 3).
**Figure 2:** *Drawing the intraabdominal area and calculating the fat volume along the inner surface of the abdominal wall, excluding the muscle planes.* **Figure 3:** *Measurement of total fat volume over fat density by taking the skin line as the border. Calculation of subcutaneous fat tissue volume by subtracting visceral adipose tissue volume from total fat volume.*
With the same application, muscle density was selected and paravertebral muscle tissue volume was calculated (bilateral musculus psoas major, musculus quadratus lumborum, musculus iliocostalis, musculus longissimus, musculus multifidus volumes). A Spearman correlation model was used to analyze visceral adiposity, subcutaneous fat, and muscle mass.
In CT images, each intervertebral disc space was evaluated in terms of the presence of osteophytes, loss of disc height, sclerosis in the end plates, and spinal stenosis (spinal canal narrowing under 15 mm AP diameter) to investigate the presence of degeneration. Each level was scored according to the presence of findings, with 1 point for the presence of osteophytes, loss of disc height, sclerosis in the end plates, and spinal stenosis. The total score at all levels (L1-S1) was calculated for each patient.
Statistical analyses were performed using IBM SPSS version 20.0 software (IBM Corp., Armonk, NY). The conformity of the data to normal distribution was assessed using the Shapiro-Wilk test. Normally distributed variables were presented as mean±standard deviation and those not showing normal distribution as median (minimum-maximum) values. Categorical variables were presented as numbers (n) and percentages (%). The Spearman's rank correlation coefficient test was used to determine the correlation between the measured parameters in various vertebral pathologies. Continuous variables were compared using the Mann-Whitney U test. The receiver operating characteristic (ROC) analysis was used to detect the area under the curve (AUC) and define the cutoff values with their sensitivities and specificities of the measurements. An alpha value of $p \leq 0.05$ was accepted as statistically significant.
## Results
A positive correlation between visceral fat, subcutaneous fat, and total fat was observed. No correlation was detected between any fat volume and total muscle volume ($r = 0.450$-0.867) (Table 1).
**Table 1**
| Variables | n | Mean (cm³) | SD | 1 | 2 | 3 | 4 |
| --- | --- | --- | --- | --- | --- | --- | --- |
| 1. Total fat | 146 | 45.06 | 1.7 | - | 0.789* | 0.867* | 0.115 |
| 2. Visceral fat | 146 | 17.91 | 0.86 | 0.789* | - | 0.450* | 0.158 |
| 3. Subcutaneous fat | 146 | 26.89 | 1.2 | 0.867* | 0.450* | - | 0.027 |
| 4. Total muscle | 146 | 10.13 | 0.27 | 0.115 | 0.158 | 0.027 | - |
The associations between the measured parameters (total, visceral, and subcutaneous fat mass, muscle mass) and the presence of loss of height in the vertebral disc, sclerosis, osteophytes, and spinal stenosis are assessed and presented in Tables 2-5. A positive correlation was found between the loss of intervertebral disc height at all lumbar levels and the amount of fat volumes (total, visceral, subcutaneous) ($$p \leq 0.001$$-0.029). Although the amount of muscle volume was not associated with the loss of intervertebral disc height ($$p \leq 0.057$$-0.417).
As the area under the curve (AUC) values were presented, it was observed that the association between the visceral and total fat volume and degeneration (loss of disc height) scores at all levels was higher than that of subcutaneous fat volume (Table 2). An association was observed between osteophytes at lumbar levels and all the fat volumes (total, visceral, and subcutaneous) ($$p \leq 0.002$$-0.044). Although, the amount of muscle volume was not associated with lumbar osteophytes ($$p \leq 0.263$$-0.995). When the AUC was examined, it was determined that the parameters most associated with the loss of height in the vertebral disc were visceral and total fat masses (Table 3).
An association was found between sclerosis and all the fat volumes at all lumbar levels ($$p \leq 0.001$$-0.021) but again no correlation was observed between the amount of muscle mass and sclerosis presence ($$p \leq 0.081$$-0.582) (Table 4). It was observed that spinal stenosis at the lumbar levels was not associated with the amount of fat (total, visceral, subcutaneous) at any level ($$p \leq 0.055$$-0.990) and also was not associated with the amount of muscle mass ($$p \leq 0.130$$-0.979) (Table 5). In addition, there was a significant difference between the assessed vertebral disorders in terms of fat tissue, but no difference was observed regarding the amount of muscle mass.
## Discussion
The results determined in our study present the association between visceral, subcutaneous, and total fat masses and muscle mass with LVD. Most individuals in the study were overweight and obese ($79\%$), implying a possible change in lumbar disc characteristics due to increased mechanical load, as noted by Iatridis et al. [ 17]. According to the general opinion, excess weight causes degeneration in the intervertebral disc structure at histological and macroscopic levels, leading to an acceleration of the lumbar degenerative process [3]. In an MRI study conducted by Takatalo et al., it was revealed that there is a relationship between degenerated discs and abdominal obesity [6]. The causal relationship between the height of visceral, subcutaneous, and total fat masses detected in our study and disc degeneration is consistent with the results of the study of Takatalo et al. [ 6]. Again, Hershkovich et al. reported a relationship between obesity and disc degeneration in terms of low back pain [18]. In addition, vertebral osteophytes and sclerosis were also examined in our study, and the relationship between vertebral bone degeneration and abdominal fat volumes was also revealed.
Although obesity has been shown to be associated with many endocrine and cardiovascular diseases, its relationship with LVD remains unclear in the current literature. The reason is largely associated with the lack of large epidemiological studies with assumptions resulting from an appropriate study design, inadequate statistical analysis, and limited radiographic interpretation of additional spinal findings that may advance to the degenerative process. In a study conducted in the Netherlands in which direct roentgenograms of 2819 individuals were examined, no correlation was found between increased body mass index and decreased intervertebral disc height [19]. Similarly, in a study conducted in England, it was stated that this relationship was weak [20]. Again, in a study conducted in the USA, this relationship was examined in 187 individuals, facet joint degeneration was more common in individuals with increased body mass index, but no relationship was found with the narrowing of the disc space [21]. As can be seen, when investigating the relationship between disc degeneration and vertebral pathologies, the presence of obesity alone seems insufficient which directed us to investigate more related parameters like visceral, subcutaneous, and total abdominal fat volumes.
Previous studies have reported that high BMI is a risk factor for lower back pain. Excessive adipose tissue has been highly blamed for damage to the spinal structures [22]. Structural damage and pathological changes in the vertebral body are the most prominent changes [23]. However, as stated in previous studies, we made these measurements with the thought that the relationship between vertebral bone degeneration and adipose tissue might be more illuminating since BMI has a weak relationship with degeneration. Furthermore, the distribution of the body adiposity may play a more important role in lumbar disc herniation. It has been found that obesity leads to an increase in the synthesis of proinflammatory cytokines produced from adipose tissue. These adipose cytokines also increase c-reactive protein synthesis from hepatocytes in obese individuals [24]. These reasons play a role in the association of obesity with disc degeneration. In our study, lumbar vertebrae degeneration was significantly associated with adipose mass parameters, while none of the muscle mass measurements were related to disc degeneration. Failure to find a relationship between paravertebral muscle volume and vertebral degeneration may mean that the amount of abdominal fat volume may be more effective on vertebral degeneration than the amount of muscle volume, but broader sample size studies are required to advocate this theory.
The fact that vertebral degeneration can also be seen in asymptomatic individuals has led to further investigation of the relationship between disc degeneration and vertebral anatomical differences. Boden et al. performed MRI examination in 67 patients who never had low back pain (LBP), neurogenic claudication, or sciatica, and found that approximately one-third of these patients had significant vertebral pathologies, such as herniated nucleus pulposus, stenosis, degeneration, and bulging [25]. Although this degeneration is observed in asymptomatic individuals, according to Samartzis et al., existing disc degeneration is guiding and predictive for future LBP [26].
Samartizis et al. stated that obesity is a risk factor for the presence, prevalence, and severity of disc degeneration [27]. Takatalo et al. measured body fat using MRI and found similar measurements of abdominal circumference, therefore they suggested clinical use of this measurement to assess abdominal adiposity as a risk factor for disc degeneration [6]. Other studies such as Han et al. reported that an increase in the amount of fat around the abdomen and high BMI values in patients were associated with chronic low back pain and lumbar disc herniation [28]. In another study in the literature indicating the relationship between lumbar fat mass and lumbar intervertebral pathologies, it was reported that subcutaneous fat mass reliably differentiated patients with chronic low back pain and severe Modic changes at the lumbar level from asymptomatic subjects [9]. Baek et al. argued that decreased muscle mass and increased fat mass are associated with the loss of disc height and spondylolisthesis of consecutive vertebrae in the lumbar region [29]. Considering the results of our study, while the relationship between intervertebral disc degeneration and loss of disc height and increased fat mass was consistent with the study of Baek et al., we could not find a relationship between muscle mass and disc pathologies [29].
Our study seems to be quite compatible with the inferences mentioned in these studies, which stated that vertebral disc degeneration is associated with high abdominal adiposity. In our study, we found that the visceral, subcutaneous, and total adipose tissue volumes that we measured in patients who underwent abdominal CT imaging were correlated with each other, but we did not detect any relationship between these adipose tissue measurements and muscle mass. We found that there was a correlation between the loss of vertebral disc space and adipose tissue volumes at all levels. When examined separately, it was observed that the amount of visceral and total adipose tissue was more associated with degeneration than the amount of subcutaneous adipose tissue. Likewise, it is observed that the amount of adipose tissue is associated with vertebral osteophyte and sclerosis formation, and spinal stenosis. We observed that muscle mass was not associated with any of these pathologies. In previous studies, the weak correlation of body mass index alone with vertebral pathologies led us to examine the effect of abdominal fat and muscle amounts. As a result, it was determined in our study that abdominal visceral, subcutaneous, and total fat volumes are associated with pathologies, such as vertebral degeneration, loss of disc height, sclerosis, and osteophyte formation. However, to prove that this relationship is stronger than BMI, it is necessary to examine the correlation between adipose tissue measurements and BMI, which is among the main targets in our future planned studies.
This study aimed to identify the amount of the body composition components like visceral, subcutaneous, and total fat as well as muscle mass as risk factors for loss of disc height (LDH). We believe that our study will make a substantial contribution to the current literature as one of the studies investigating the etiology of vertebral discopathies.
The first and probably the most important limitation of this study is its cross-sectional design. The measurements of this study require further analysis and verification to visualize whether the patients’ fat and muscle composition can predict future lumbar disc pathologies.
## Conclusions
The amount of visceral, subcutaneous, and total adipose tissue in the abdominal region are components associated with vertebral disc degeneration, sclerosis, and osteophyte formation. Abdominal fat mass can be used in clinical decisions as a risk factor for LVD. These factors should be taken into account when assessing the patient's likelihood of developing vertebral disorders.
## References
1. Flegal KM, Graubard BI, Williamson DF, Gail MH. **Cause-specific excess deaths associated with underweight, overweight, and obesity**. *JAMA* (2007) **298** 2028-2037. PMID: 17986696
2. Shiri R, Karppinen J, Leino-Arjas P, Solovieva S, Viikari-Juntura E. **The association between obesity and low back pain: a meta-analysis**. *Am J Epidemiol* (2010) **171** 135-154. PMID: 20007994
3. Liuke M, Solovieva S, Lamminen A, Luoma K, Leino-Arjas P, Luukkonen R, Riihimäki H. **Disc degeneration of the lumbar spine in relation to overweight**. *Int J Obes (Lond)* (2005) **29** 903-908. PMID: 15917859
4. Taher F, Essig D, Lebl DR, Hughes AP, Sama AA, Cammisa FP, Girardi FP. **Lumbar degenerative disc disease: current and future concepts of diagnosis and management**. *Adv Orthop* (2012) **2012** 970752. PMID: 22567411
5. Lee SH, Park SW, Kim YB, Nam TK, Lee YS. **The fatty degeneration of lumbar paraspinal muscles on computed tomography scan according to age and disc level**. *Spine J* (2017) **17** 81-87. PMID: 27497888
6. Takatalo J, Karppinen J, Taimela S. **Association of abdominal obesity with lumbar disc degeneration--a magnetic resonance imaging study**. *PLoS One* (2013) **8** 0
7. Yang L, Mu L, Huang K. **Abdominal adipose tissue thickness measured using magnetic resonance imaging is associated with lumbar disc degeneration in a Chinese patient population**. *Oncotarget* (2016) **7** 82055-82062. PMID: 27833090
8. Riesenburger R. **Increased Fat Mass Is Associated With High Levels of Low Back Pain Intensity and Disability**. *Yearbook of Neurology and Neurosurgery* (2012) **2012** 271-272
9. Teichtahl AJ, Urquhart DM, Wang Y, Wluka AE, O'Sullivan R, Jones G, Cicuttini FM. **Modic changes in the lumbar spine and their association with body composition, fat distribution and intervertebral disc height - a 3.0 T-MRI study**. *BMC Musculoskelet Disord* (2016) **17**
10. Brady SR, Urquhart DM, Hussain SM, Teichtahl A, Wang Y, Wluka AE, Cicuttini F. **High baseline fat mass, but not lean tissue mass, is associated with high intensity low back pain and disability in community-based adults**. *Arthritis Res Ther* (2019) **21**
11. Eloi JC, Epifanio M, de Gonçalves MM. **Quantification of abdominal fat in obese and healthy adolescents using 3 Tesla magnetic resonance imaging and free software for image analysis**. *PLoS One* (2017) **12**
12. Maillard F, Pereira B, Boisseau N. **Effect of high-intensity interval training on total, abdominal and visceral fat mass: a meta-analysis**. *Sports Med* (2018) **48** 269-288. PMID: 29127602
13. Wewege M, van den Berg R, Ward RE, Keech A. **The effects of high-intensity interval training vs. moderate-intensity continuous training on body composition in overweight and obese adults: a systematic review and meta-analysis**. *Obes Rev* (2017) **18** 635-646. PMID: 28401638
14. Kvist H, Chowdhury B, Grangård U, Tylén U, Sjöström L. **Total and visceral adipose-tissue volumes derived from measurements with computed tomography in adult men and women: predictive equations**. *Am J Clin Nutr* (1988) **48** 1351-1361. PMID: 3202084
15. Danneels LA, Vanderstraeten GG, Cambier DC, Witvrouw EE, De Cuyper HJ. **CT imaging of trunk muscles in chronic low back pain patients and healthy control subjects**. *Eur Spine J* (2000) **9** 266-272. PMID: 11261613
16. Ryo M, Kishida K, Nakamura T, Yoshizumi T, Funahashi T, Shimomura I. **Clinical significance of visceral adiposity assessed by computed tomography: a Japanese perspective**. *World J Radiol* (2014) **6** 409-416. PMID: 25071881
17. Iatridis JC, MacLean JJ, Roughley PJ, Alini M. **Effects of mechanical loading on intervertebral disc metabolism in vivo**. *J Bone Joint Surg Am* (2006) **88 Suppl 2** 41-46
18. Hershkovich O, Friedlander A, Gordon B. **Associations of body mass index and body height with low back pain in 829,791 adolescents**. *Am J Epidemiol* (2013) **178** 603-609. PMID: 23690249
19. de Schepper EI, Damen J, van Meurs JB. **The association between lumbar disc degeneration and low back pain: the influence of age, gender, and individual radiographic features**. *Spine (Phila Pa 1976)* (2010) **35** 531-536. PMID: 20147869
20. Hassett G, Hart DJ, Manek NJ, Doyle DV, Spector TD. **Risk factors for progression of lumbar spine disc degeneration: the Chingford Study**. *Arthritis Rheum* (2003) **48** 3112-3117. PMID: 14613273
21. Kalichman L, Guermazi A, Li L, Hunter DJ. **Association between age, sex, BMI and CT-evaluated spinal degeneration features**. *J Back Musculoskelet Rehabil* (2009) **22** 189-195. PMID: 20023349
22. Sheng B, Feng C, Zhang D, Spitler H, Shi L. **Associations between obesity and spinal diseases: a medical expenditure panel study analysis**. *Int J Environ Res Public Health* (2017) **14**
23. Amin RM, Andrade NS, Neuman BJ. **Lumbar disc herniation**. *Curr Rev Musculoskelet Med* (2017) **10** 507-516. PMID: 28980275
24. Ford ES, Galuska DA, Gillespie C, Will JC, Giles WH, Dietz WH. **C-reactive protein and body mass index in children: findings from the Third National Health and Nutrition Examination Survey, 1988-1994**. *J Pediatr* (2001) **138** 486-492. PMID: 11295710
25. Boden SD, Davis DO, Dina TS, Patronas NJ, Wiesel SW. **Abnormal magnetic-resonance scans of the lumbar spine in asymptomatic subjects. A prospective investigation**. *J Bone Joint Surg Am* (1990) **72** 403-408. PMID: 2312537
26. Samartzis D, Karppinen J, Luk K, Cheung DK. **Baseline MRI characteristics in asymptomatic subjects as predictors for future first-time LBP episode Spineweek**. *Glob Spine J* (2012) **1**
27. Samartzis D, Karppinen J, Cheung JP, Lotz J. **Disk degeneration and low back pain: are they fat-related conditions?**. *Global Spine J* (2013) **3** 133-144. PMID: 24436864
28. Han TS, Schouten JS, Lean ME, Seidell JC. **The prevalence of low back pain and associations with body fatness, fat distribution and height**. *Int J Obes Relat Metab Disord* (1997) **21** 600-607. PMID: 9226492
29. Baek S, Park HW, Kim G. **Associations between trunk muscle/fat composition, narrowing lumbar disc space, and low back pain in middle-aged farmers: a cross-sectional study**. *Ann Rehabil Med* (2022) **46** 122-132. PMID: 35793901
|
---
title: ABC-GOALScl score predicts admission to the intensive care unit and mortality
of COVID-19 patients over 60 years of age
authors:
- María Elena Camacho-Moll
- Zayra Ramírez-Daher
- Brenda Leticia Escobedo-Guajardo
- Julio César Dávila-Valero
- Brenda Ludmila Rodríguez-de la Garza
- Mario Bermúdez de León
journal: BMC Geriatrics
year: 2023
pmcid: PMC9999052
doi: 10.1186/s12877-023-03864-8
license: CC BY 4.0
---
# ABC-GOALScl score predicts admission to the intensive care unit and mortality of COVID-19 patients over 60 years of age
## Abstract
### Background
One of the risk factors for getting seriously ill from COVID-19 and reaching high mortality rates is older age. Older age is also associated with comorbidities, which are risk factors for severe COVID-19 infection. Among the tools that have been evaluated to predict intensive care unit (ICU) admission and mortality is ABC-GOALScl.
### Aim
In the present study we validated the utility of ABC-GOALScl to predict in-hospital mortality in subjects over 60 years of age who were positive for SARS-CoV-2 virus at the moment of admission with the purpose of optimizing sanitary resources and offering personalized treatment for these patients.
### Methods
This was an observational, descriptive, transversal, non-interventional and retrospective study of subjects (≥ 60 years of age), hospitalized due to COVID-19 infection at a general hospital in northeastern Mexico. A logistical regression model was used for data analysis.
### Results
Two hundred forty-three subjects were included in the study, whom 145 ($59.7\%$) passed away, while 98 ($40.3\%$) were discharged. Average age was 71, and $57.6\%$ were male. The prediction model ABC-GOALScl included sex, body mass index, Charlson comorbidity index, dyspnea, arterial pressure, respiratory frequency, SpFi coefficient (Saturation of oxygen/Fraction of inspired oxygen ratio), serum levels of glucose, albumin, and lactate dehydrogenase; all were measured at the moment of admission. The area under the curve for the scale with respect to the variable of discharge due to death was 0.73 (IC $95\%$ = 0.662—0.792).
### Conclusion
The ABC-GOALScl scale to predict ICU admission in COVID-19 patients is also useful to predict in-hospital death in COVID-19 patients ≥ 60 years old.
## Background
In December 2019, the people’s Republic of China reported rare cases of pneumonia of unknown origin. The first cases were traced back to a seafood market called “Huanan” in Wuhan, Hubei province. Clinical symptoms included fever, dyspnea, and bilateral lung infiltrates [1]. In January 2020, the identification of the coronavirus responsible for this disease was described. This was a Betacoronavirus RNA, which was named Severe Acute Respiratory Syndrome Coronavirus-2 (SARS-CoV-2) [2]. By March 11, 2020, this disease was declared a pandemic by the World Health Organization, and to date, there have been more than 5 million cumulative deaths worldwide, with a global lethality of $1.9\%$ [3]. Among the risk factors for infection and developing severe COVID-19, age, sex, comorbidities, obesity, and tobacco smoking have been included [4–8]. In Mexico, to date there are more than 300,000 cumulative deaths, and the prevalence of COVID-19 in older patients represents $19\%$ [3]. According to global reports, most infected people’s age is around 50 years; however, there is increased mortality in people over 60 years old [9]. The average age of fatal cases is 80 years based on Italian reports, where only $1.1\%$ of deaths occurred in people younger than 50 years old [10]. The association of age as a risk factor has also been published elsewhere [11].
Scales to predict possible outcomes are suggested tools for treatment decisions; for instance, the National Institute for Health and Care Excellence guidelines are evidence-based recommendations for health and care in England that assess frailty through a clinical frailty score. This score takes into consideration comorbidity, function, and cognition and classifies patients from 1 (very healthy) to 9 (terminally ill). Those whose score is less than 5 are eligible for complete and invasive support in the intensive care unit (ICU) [12]. ABC-GOALSc is another promising tool that has been demonstrated to be useful to predict ICU admission in COVID-19 patients, with an area under the curve of 0.79 and 0.77 in the development and validation cohorts, respectively. This result has been improved by adding other factors to the model, where ABC-GOALScl had an area under the curve of 0.86 and 0.87 in the development and validation cohorts, respectively, and ABC-GOALSclx had an area under the curve of 0.88 and 0.86 in the development and validation cohorts, respectively [13]. Differences between ABC-GOALScl and ABC-GOALSclx scores are that the latter includes a tomographic image analysis of thorax through the CO-RADS categories. The present paper aims to establish whether ABC-GOALScl is also useful to predict in-hospital mortality in COVID-19 patients over 60 years of age.
## Methods
The study was conducted according to the guidelines of the Declaration of Helsinki and approved by the Institutional Review Board of the Mexican Social Security Institute (protocol 2021–1909-106, August 9, 2021). This is a retrospective and non-interventional study, where medical records were consulted and data in the public repository are de-identified.
This study was carried out with records from patients hospitalized in the Zone General Hospital No. 4 “Villa Guadalupe” located in Guadalupe, Nuevo Leon, Mexico. Patients 60 of years old and older with confirmed COVID-19 diagnosis by RT-PCR test and hospitalized between December 1, 2020 and January 5, 2021 were included. A database was collected that included social security number, age, comorbidities, days of hospitalization, outcome, and date of discharge or death. Database, raw and processed are available at Mendeley Data, V1, https://doi.org/10.17632/z4z22nbmmz.1. ABC-GOALScl, which incorporates clinical and laboratory results, was used and scored as previously described [13]. This model includes sex, systolic arterial pressure (SAP), presence or absence of dyspnea by respiratory frequency (RF), Charlson comorbidity index, glucose serum levels, obesity, albumin serum levels, lactate dehydrogenase (LDH) serum levels, and SpFi coefficient (Saturation of oxygen/fraction of inspired oxygen, SO2/FiO2, ratio). Subjects who had incomplete clinical records, were diagnosed with Acinetobacter spp. infection or Clostridium difficile, had records that came from another unit, or were directly admitted to the ICU were excluded. Files from subjects who voluntarily requested to leave the study were deleted.
The distribution of continuous variables was evaluated with Kolmogorov–Smirnov. Descriptive statistics were used to analyze the data; qualitative variables are described in frequencies and percentages. For comparison of qualitative variables, chi-squares and stepwise multivariate logistic ordinal regression models were run to calculate adjusted odds ratio (OR) and $95\%$ Confidence Interval (CI) for each component of the ABC-GOALScl score. For quantitative data, a t–test and a Mann–Whitney U test were performed. Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and the area under the curve (AUC) were calculated, and a value of $p \leq 0.05$ was considered significant.
## Results
A total of 243 subjects over 60 years of age diagnosed with COVID-19 were included in this study. The average age was 71.7, and males represented $57.6\%$. Average days of hospitalization were 11.53. At least one comorbidity was reported in 206 patients ($84.8\%$). The Charlson comorbidity index had an average score of 4, which would fall into the moderate category. Average Body Mass Index (BMI) was 28.7. Average systolic arterial pressure was 127 mm Hg, average O2 saturation was $80\%$, average FiO2 was $40\%$, and the average SpFi coefficient (SO2/FiO2 ratio) was 294.8. Measured laboratory variables were glucose, albumin, and LDH. The average glucose value was 190.8 mg/dL, 3.2 g/dL for albumin, and 411 U/L for LDH (Table 1). The outcome of 145 subjects ($59.7\%$) was death, whereas 98 recovered ($40.3\%$). Finally, the average ABC-GOALScl result was 8.2 (Table 1). Table 2 summarizes the contributions of factors in the final ABC-GOALS score. Mainly, age, total comorbidities, BMI, respiratory frequency, and SO2/FiO2 ratio were the most determinant factors (Table 2).Table 1Characteristics of subjects ($$n = 243$$) included in the studyN (%)Sex (N) Male140 (57.6) Female103 (42.4)Comorbidities (N) Any206 (84.8) Hypertension181 (74.5) Diabetes mellitus129 (53.1) Chronic kidney disease73 (15.2) COPD15 (6.2) Oncological disease10 (4.1) Dementia6 (2.5) Autoimmune disease3 (1.2) Liver disease2 (0.8)Age (years) 60–6559 (24.3) 66–7059 (24.3) 71–7555 (22.6) ≥ 7670 (28.8)Charlson Comorbidity Index (N) Mild 1–250 (20.6) Moderate 3–4113 (46.5) Severe ≥ 580 (32.9)Outcome (N) Death145 (59.7) Recovery98 (40.3)ABC-GOALScl (N) 0–3 Low risk12 (4.9) 5–9 Moderate risk148 (60.9) ≥ 10 High risk83 (34.2)Mean (Min, Max) Days of hospitalization11.53 [1, 110] Total comorbidities (N)1.6 [0, 4] BMI28.7 (16.7, 59.03) Systolic arterial pressure (mmHg)130.7 [52, 220] Respiratory frequency (breaths/min)25.5 [12, 48] O2 saturation (sO2)0.8 (0.15, 0.99) fraction of inspired oxygen (FiO2)0.4 (0.21, 1.00) sO2/FiO2 ratio294.8 (18.75, 466.67)Laboratory results Glucose (mg/dL)190.8 [32, 1097] Albumin (g/dL)3.2 (1.38, 4.34) LDH (U/L)411 [110, 2950]COPD chronic obstructive pulmonary disease, BMI body mass index, sO2 saturation of oxygen, FiO2 fraction of inspired oxygen, LDH lactate dehydrogenaseTable 2ABC-GOALScl componentsDeathN (%)RecoveryN (%)Chi- squarep-valueOR (Min, Max)p-valueSex (N) Men86 (59.3)54 (55.1)0.521 Women59 (40.7)44 (44.9)0.94 (0.50, 1.78)0.852Age (years) 60–6551 (35.2)19 (19.4)0.0521 66–7032 (22.1)23 (23.4)0.92 (0.40, 2.10)0.843 71–7630 (20.7)29 (29.6)1.15 (0.63, 3.57)0.362 > 7632 (54.2)27 (27.6)3.19 (1.32, 7.68)0.01Total comorbidities (N) 013 (9.0)24 (24.5)0.0211 147 (32.4)26 (26.5)3.35 (1.31, 8.53)0.011 262 (42.8)33 (33.7)3.37 (1.32, 8.65)0.011 320 (13.8)14 (14.3)2.19 (0.72, 6.68)0.169 43 (2.1)1 (1.0)2.02 (0.14, 29.20)0.605BMI < 29.99101 (69.7)79 (80.6)0.0561 > 3044 (30.3)19 (19.4)2.39 (1.13, 5.05)0.023Dyspnoea (N) No7 (4.8)4 (4.1)0.781 Yes138 (95.2)94 (95.9)0.89 (0.21, 3.87)0.876SAP (mm Hg) > 101128 (88.3)88 (89.8)0.711 < 100.917 (10.2)10 (10.2)1.33 (0.41, 4.37)0.635RF (breaths/min) < 2355 (37.9)45 (45.9)0.0011 24–2854 (37.2)47 [48]0.79 (0.42, 1.47)0.456 > 2936 (24.8)6 (6.1)4.30 (1.48, 12.46)0.007Glucose (mg/dL) < 19992 (63.9)78 (79.6)0.0091 > 200.953 (36.6)20 (20.4)1.68 (0.83, 3.37)0.148Albumin (g/dL) > 3.542 (29.0)40 (40.8)0.0551 < 3.49103 (71.0)58 (59.2)1.40 (0.74, 2.67)0.306LDH (U/L) < 199.914 (9.7)16 (16.3)0.121 > 200131 (90.3)82 (83.7)1.88 (0.79, 4.44)0.152sO2/FiO2 ratio > 30065 (44.8)69 (70.4)0.0001 < 299.980 (55.2)29 (29.6)2.32 (1.25, 4.29)0.008BMI Body mass index, OR Odds ratio, SAP Systolic Arterial Pressure, RF Respiratory Frequency, LDH Lactate Dehydrogenase, sO2/FiO2 Saturation of Oxygen/Fraction of Inspired Oxygen A significant relationship between subject outcome and ABC-GOALScl score was observed by the chi-square test. Logistic regression analysis also demonstrated that the ABC-GOALScl score is a useful tool to predict subject outcomes, where subjects classified with moderate risk in the ABC-GOALScl classification had 5 times the probability of death, and subjects classified with high risk had 24.6 times higher probability (Table 3).Table 3ABC-GOALScl chi-square and multivariate linear regression against outcomeABC-GOALSclDeathN (%)RecoveryN (%)Chi-squarep-valueOR (Min, Max)Low risk2 (1.4)10 (10.2)0.0001Moderate risk74 (51.0)74 (75.5)5.00 (1.06, 23.60)*High risk69 (47.6)14 (14.3)24.64 (4.86, 124.93)****, $p \leq 0.05$; ***, $p \leq 0.001.$ OR Odds ratio The ABC-GOALScl model demonstrated good accuracy in estimating the risk of death, with an area under the curve of 0.7, with 0.96 sensitivity and 0.79 for 1–specificity values at 4.5 as the best cutoff point (Fig. 1). PPV and NPV were also calculated, resulting in $83.1\%$ and $52.5\%$, respectively, when grouping low– with moderate–risk compared to high-risk patients based on their ABC-GOALScl score. Fig. 1Estimating the risk of death using ABC-GOALScl. A Receiver operating characteristic (ROC) curve; B curve coordinates; C Area under the curve results for ABC-GOALScl. ***, $p \leq 0.001.$ SD, standard deviation
## Discussion
The present study aimed to establish whether the ABC-GOALScl score is useful to predict in-hospital death in a subject cohort over 60 years of age infected with COVID-19. It is well known that advanced age is a risk factor for COVID-19 infection and mortality. Aging is also a trigger for comorbidity development, which at the same time reduces the probability of a good outcome after COVID-19 infection. Multiple tools such as the ABC-GOALScl score have been validated to predict ICU admission and mortality due to COVID-19 in the general population. We have found a $73\%$ probability that ABC-GOALScl will predict subject outcome.
There are other scores specifically developed for in-hospital mortality. They have been developed by using parameters to evaluate respiratory function, oxygen saturation, and some markers of inflammatory processes, all common events in pulmonary diseases, among other variables that support the original intended goal of the score. Examples include the Clinical Characterization Consortium (ISARIC‐4C) score with an area under the ROC curve (AUROC) of 0.799 (0.738 – 0.851); the COVID‐GRAM Critical Illness Risk Score (COVID‐GRAM), with AUROC of 0.785 (0.723 – 0.838); the quick COVID‐19 Severity Index (qCSI), with an AUROC of 0.749 (0.685 – 0.806); and the National Early Warning Score (NEWS), with an AUROC of 0.764 (0.700 – 0.819) [14]. Adding to this, the present study reports that the ABC-GOALScl score has an AUROC of 0.73 (0.66 – 0.79), which is very similar to the results of other scores (Table 4); however, the sensitivity and specificity balance was better for ABC-GOALScl with our data, even when inflammatory parameters such as interleukin 6 and neutrophil-to-lymphocyte ratio were not considered. Median age, male prevalence, Charlson comorbidity index, and sample size make our results similar to those reported by Covino and others [2020] [14].Table 4Scoring systems for predicting death of COVID‐19 patients over 60 years of ageScoreAUROC (Min, Max)NCut off valueSensitivity (%)Specificity (%)PPVNPVNEWS0.764 (0.700, 0.819)210 > 466.7693549.2COVID‐GRAM0.785 (0.723, 0.838)210 > 17.788.161.336.395.4ISARIC‐4C0.799 (0.738, 0.851)210 > 888.15.933.394.9qCSI0.749 (0.685,0.806)210 > 56977.443.390.9PSI (65–84)0.85 (0.80, 0.90)4389110038.323.998.5PSI (> 85)0.69 (0.60, 0.79)2019110019.814.994.7CURB-65 (65–84)0.73 (0.65, 0.82)438N/A65.574.716.396.5CURB-65 (> 85)0.60 (0.48, 0.73)201N/A47.465.912.792.3ABC-GOALScl0.73 (0.66, 0.79)2434.59578.683.152.5AUROC Area Under the ROC Curve, NEWS National Early Warning Score, COVID-GRAM COVID-GRAM Critical Illness Risk Score, ISARIC-4C International Severe Acute Respiratory Infection Consortium Clinical Characterization Protocol-Coronavirus Clinical Characterization Consortium, qCSI quick COVID-19 Severity Index, PSI Pneumonia Severity Index, CURB-65 confusion, blood urea, respiratory rate, blood pressure, pneumonia Severity Score [14, 15] Glucose levels are a significant predictor of mortality in our study. None of the other scores included this parameter. Considering that more than $10\%$ of people in Mexico suffers from diabetes mellitus and it is the third cause of death after COVID-19 [16], our score better fits the characteristics of the Mexican population. Nonetheless, the associated number of comorbidities constitutes an important predictor of mortality in ISARIC-4C and COVID-GRAM scores, as in our study [17–19].
The average BMI of 28.7 represents a population with overweight characteristics. In our study, this was not a predictor of mortality, but Bartoletti et al. [ 2020] reported in a similar score (PREDI-CO) that obesity is a stronger condition for the outcome in hospitalized patients with COVID-19. In geriatric people, risk of malnutrition is also a common feature. In contrast with our results, evidence of a relationship between malnutrition and mortality has been reported using the CONUT score. This event could be explained by the differences among populations [20, 21].
García-Gordillo et al. [ 2021] compared a newly developed score named COVID-IRS against ABC-GOALScl and six other scores to predict the risk of invasive mechanical ventilation in infected patients with COVID-19. ABC-GOALScl had a performance with an AUC intermediate to the newly developed and other implemented scores. Respiratory failure represents the principal cause of death in hospitalized COVID-19 patients [22].
Other groups have reported models for in-hospital mortality in the general population. A Chinese group has also reported a model that includes age, history of hypertension, and coronary heart disease, with an area under the curve of 0.88 ($95\%$ CI 0.80 – 0.95), sensitivity of $92.31\%$, and specificity of $77.44\%$; however, this model is for the general population, not for geriatric patients [23]. This same laboratory developed a model to predict in-hospital mortality, but based on age and lab results (high-sensitivity C-reactive protein, peripheral capillary oxygen saturation, neutrophil and lymphocyte count, d-dimer, aspartate aminotransferase, and glomerular filtration), with an AUC of 0.83 (0.68–0.93) [23].
Other studies have described sex, increased fraction of inspired oxygen, and crackles as the best predictors of mortality, with 4, 1, and 2.4 times increased probability of mortality, respectively [24]. ABC-GOALScl includes these factors and agrees with Mendes and colleagues’ results [2020].
Mesas and colleagues [2020] published a systematic review that included 60 studies, in which they investigate predictors of in-hospital mortality by gender, age, and health parameters; they concluded that dyspnea is an important factor as well as obesity and several other comorbidities [25]. However, for geriatric subjects, the most important factors were obesity, albumin, total bilirubin, alanine aminotransferase, serum ferritin, C-reactive protein and LDH. The results published by these authors agrees with our study given that they reported similar factors to the ones included in the ABC-GOALScl score. Differences can be attributed to the size and design of the different studies.
We report a higher mortality rate in subjects over 60 years of age ($59.7\%$) compared to other studies, where it has been reported to be around $32\%$ [25, 26]; this could be explained by the fact that our hospital was designated to concentrate patients with severe COVID-19 disease. Another explanation could be that the patient’ records were considered as completed in a longer period of time compared to other studies, where records included for the study were those completed in a month [18].
This study used a tool that has been widely validated to classify patients at risk of ICU admission and therefore at a higher risk of death [13]. A significant advantage of the ABC-GOALScl is that it allows follow-up of patients within the hospital, where the service for patients over 60 years of age can be personalized.
Due to the retrospective design of the present study, we can list some limitations. For instance, previous severe clinical conditions and treatments were not considered. The mortality rates may not be representative of the Mexican population because data were obtained from a hospital designated as a COVID-19 center during the pandemic period. We suggest performing more prospective studies to validate this model and to identify the key predictors for mortality in the population over 60 years of age.
## Conclusions
We conclude that ABC-GOALScl is a useful tool that could be applied in hospitals to give personalized treatments and interventions that might increase the favorable outcomes for patients over 60 years of age.
## References
1. Hu B, Guo H, Zhou P, Shi ZL. **Characteristics of SARS-CoV-2 and COVID-19**. *Nat Rev Microbiol* (2021.0) **19** 141-54. DOI: 10.1038/s41579-020-00459-7
2. 2.Mohanty SK, Satapathy A, Naidu MM, Mukhopadhyay S, Sharma S, Barton LM, et al. Severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) and coronavirus disease 19 (COVID-19) - anatomic pathology perspective on current knowledge. Diagn Pathol 2020;15(1). Available from: https://pubmed.ncbi.nlm.nih.gov/32799894/. [Cited 2021 Nov 30].
3. 3.WHO COVID-19 Dashboard. Geneva: World Health Organization, 2020. Available online: https://covid19.who.int/. Available from: https://covid19.who.int/info/. [Cited 2021 Nov 30].
4. Wu Z, McGoogan JM. **Characteristics of and Important Lessons From the Coronavirus Disease 2019 (COVID-19) Outbreak in China: Summary of a Report of 72 314 Cases From the Chinese Center for Disease Control and Prevention**. *JAMA* (2020.0) **323** 1239-42. DOI: 10.1001/jama.2020.2648
5. 5.Guenther T, Czech-Sioli M, Indenbirken D, Robitailles A, Tenhaken P, Exner M, et al. Investigation of a superspreading event preceding the largest meat processing plant-related SARS-Coronavirus 2 outbreak in Germany. SSRN Electron J. 2020. Available from: https://papers.ssrn.com/abstract=3654517. [Cited 2021 Nov 30].
6. Clark A, Jit M, Warren-Gash C, Guthrie B, Wang HHX, Mercer SW. **Global, regional, and national estimates of the population at increased risk of severe COVID-19 due to underlying health conditions in 2020: a modelling study**. *Lancet Glob Heal* (2020.0) **8** e1003-17. DOI: 10.1016/S2214-109X(20)30264-3
7. Kass DA, Duggal P, Cingolani O. **Obesity could shift severe COVID-19 disease to younger ages**. *Lancet.* (2020.0) **395** 1544-5. DOI: 10.1016/S0140-6736(20)31024-2
8. 8.Vardavas CI, Nikitara K. COVID-19 and smoking: a systematic review of the evidence. Tob Induc Dis 2020;18. International Society for the Prevention of Tobacco Induced Diseases. Available from: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7083240/. [Cited 2021 Jun 7].
9. Lithander FE, Neumann S, Tenison E, Lloyd K, Welsh TJ, Rodrigues JCL. **COVID-19 in older people: a rapid clinical review**. *Age Ageing* (2020.0) **49** 501-15. DOI: 10.1093/ageing/afaa093
10. Bonafè M, Prattichizzo F, Giuliani A, Storci G, Sabbatinelli J, Olivieri F. **Inflamm-aging: Why older men are the most susceptible to SARS-CoV-2 complicated outcomes**. *Cytokine Growth Factor Rev* (2020.0) **1** 33-37. DOI: 10.1016/j.cytogfr.2020.04.005
11. Wu C, Chen X, Cai Y, Xia J, Zhou X, Xu S. **Risk Factors Associated with Acute Respiratory Distress Syndrome and Death in Patients with Coronavirus Disease 2019 Pneumonia in Wuhan, China**. *JAMA Intern Med* (2020.0) **180** 934-43. DOI: 10.1001/jamainternmed.2020.0994
12. Down B, Kulkarni S, Khan AHA, Barker B, Tang I. **Novel coronavirus (COVID-19) infection: What a doctor on the frontline needs to know**. *Ann Med Surg* (2020.0) **1** 24-29. DOI: 10.1016/j.amsu.2020.05.014
13. Mejía-Vilet JM, Córdova-Sánchez BM, Fernández-Camargo DA, Méndez-Pérez RA, Morales-Buenrostro LE, Hernández-Gilsoul T. **A Risk Score to Predict Admission to the Intensive Care Unit in Patients with COVID-19: the ABC-GOALS score**. *Salud Publica Mex* (2020.0) **63** 1-11. DOI: 10.21149/11684
14. Covino M, De Matteis G, Burzo ML, Russo A, Forte E, Carnicelli A. **Predicting In-Hospital Mortality in COVID-19 Older Patients with Specifically Developed Scores**. *J Am Geriatr Soc* (2021.0) **69** 37-43. DOI: 10.1111/jgs.16956
15. Chen JH, Chang SS, Liu JJ, Chan RC, Wu JY, Wang WC. **Comparison of clinical characteristics and performance of pneumonia severity score and CURB-65 among younger adults, elderly and very old subjects**. *Thorax.* (2010.0) **65** 971-7. DOI: 10.1136/thx.2009.129627
16. 16.Instituto Nacional de Estadística y Geografía. Estadísticas a propósito del día mundial de la diabetes (14 de noviembre). México; 2021. Available from: https://www.inegi.org.mx/contenidos/saladeprensa/aproposito/2021/EAP_Diabetes2021.pdf. [Cited 2022 Feb 3].
17. Liang W, Liang H, Ou L, Chen B, Chen A, Li C. **Development and Validation of a Clinical Risk Score to Predict the Occurrence of Critical Illness in Hospitalized Patients With COVID-19**. *JAMA Intern Med* (2020.0) **180** 1081-9. DOI: 10.1001/jamainternmed.2020.2033
18. Knight SR, Ho A, Pius R, Buchan I, Carson G, Drake TM. **Risk stratification of patients admitted to hospital with covid-19 using the ISARIC WHO Clinical Characterisation Protocol: development and validation of the 4C Mortality Score**. *BMJ* (2020.0) **370** 22
19. Haimovich AD, Ravindra NG, Stoytchev S, Young HP, Wilson FP, van Dijk D. **Development and Validation of the Quick COVID-19 Severity Index: A Prognostic Tool for Early Clinical Decompensation**. *Ann Emerg Med* (2020.0) **76** 442-53. DOI: 10.1016/j.annemergmed.2020.07.022
20. Bartoletti M, Giannella M, Scudeller L, Tedeschi S, Rinaldi M, Bussini L. **Development and validation of a prediction model for severe respiratory failure in hospitalized patients with SARS-CoV-2 infection: a multicentre cohort study (PREDI-CO study)**. *Clin Microbiol Infect* (2020.0) **26** 1545-53. DOI: 10.1016/j.cmi.2020.08.003
21. 21.Song F, Ma H, Wang S, Qin T, Xu Q, Yuan H, et al. Nutritional screening based on objective indices at admission predicts in-hospital mortality in patients with COVID-19. Nutr J 2021;20(1). Available from: https://pubmed.ncbi.nlm.nih.gov/34034769/. [Cited 2022 Jun 23].
22. 22.Garcia-Gordillo JA, Camiro-Zúñiga A, Aguilar-Soto M, Cuenca D, Cadena-Fernández A, Khouri LS, et al. COVID-IRS: A novel predictive score for risk of invasive mechanical ventilation in patients with COVID-19. PLoS One 2021;16(4). Available from: https://pubmed.ncbi.nlm.nih.gov/33819261/. [Cited 2022 Jun 23].
23. Wang K, Zuo P, Liu Y, Zhang M, Zhao X, Xie S. **Clinical and Laboratory Predictors of In-hospital Mortality in Patients With Coronavirus Disease-2019: A Cohort Study in Wuhan, China**. *Clin Infect Dis* (2020.0) **71** 2079-88. DOI: 10.1093/cid/ciaa538
24. Mendes A, Serratrice C, Herrmann FR, Genton L, Périvier S, Scheffler M. **Predictors of In-Hospital Mortality in Older Patients With COVID-19: The COVIDAge Study**. *J Am Med Dir Assoc.* (2020.0) **21** 1546. DOI: 10.1016/j.jamda.2020.09.014
25. 25.Mesas AE, Cavero-Redondo I, Álvarez-Bueno C, Cabrera MAS, de Andrade SM, Sequí-Dominguez I, et al. Predictors of in-hospital COVID-19 mortality: A comprehensive systematic review and meta-analysis exploring differences by age, sex and health conditions. PLoS One 2020;15(11). Available from: https://pubmed.ncbi.nlm.nih.gov/33141836/. [Cited 2021 Dec 3].
26. Bağ Soytaş R, Ünal D, Arman P, Suzan V, Emıroğlu Gedık T, Can G. **Factors affecting mortality in geriatric patients hospitalized with COVID-19**. *Turkish J Med Sci* (2021.0) **51** 454-63. DOI: 10.3906/sag-2008-91
|
---
title: Effect of Low-Level Laser Therapy on Early Wound Healing and Levels of Inflammatory
Mediators in Gingival Crevicular Fluid Following Open Flap Debridement
journal: Cureus
year: 2023
pmcid: PMC9999105
doi: 10.7759/cureus.34755
license: CC BY 3.0
---
# Effect of Low-Level Laser Therapy on Early Wound Healing and Levels of Inflammatory Mediators in Gingival Crevicular Fluid Following Open Flap Debridement
## Abstract
Introduction Low-level laser therapy (LLLT) has a beneficial effect on pain relief and wound healing. This study aims at a clinical evaluation of early wound healing and a biochemical evaluation of inflammatory mediators in gingival crevicular fluid (GCF) following LLLT with an open flap debridement (OFD) in periodontal therapy.
Material and methods This randomized controlled trial included 40 chronic periodontitis patients with bilateral attachment loss, pocket depths of 5 mm affecting at least two quadrants, and radiographic evidence of horizontal bone loss. 120 control sites were randomly selected to receive OFD, and contralateral 120 test sites received bio-stimulation with a diode laser (890 nm) after OFD. The wound healing index was recorded at the 1st and 2nd weeks, and clinical parameters such as the plaque index, gingival index, pocket probing depth, clinical attachment level, and GCF inflammatory mediators were evaluated at baseline, 3, and 6 months.
Results From the start of the study to 6 months later, there was a statistically significant drop in plaque index, gingival index, probing pocket depth, and gain clinical attachment levels in both groups. However, when the two groups were compared, there were no significant differences at any time intervals. GCF inflammatory mediators tumor necrosis factor (TNF) alpha and matrix metalloproteinases (MMP-8) decrease, and osteoprotegerin (OPG) levels increase in both the test group and control group from baseline to 3 months and 6 months. In intergroup comparisons, there was a statistically significant reduction in the test group as compared to the control group at 6 months. There was a decline in gingival crevicular fluid - interleukin-6 (GCF IL-6) levels from baseline to 3 months and 6 months in both the groups but when analysed statistically, the results were not significant on intergroup and intragroup comparison at any time interval. The Landry Wound Healing Index values in the 1st and 2nd weeks were showing statistically significant improved healing in the test group as compared to the control group. There was significantly better wound healing at sites where a diode laser was used.
Conclusion LLLT increases early wound healing after periodontal surgical procedures.
## Introduction
Periodontitis is an infectious bacterial disease characterized by loss of attachment and bone supporting the sockets. If left unattended, it leads to gingival recession and pocket formation, which eventually results in early exfoliation of the teeth. It is primarily caused by pathogenic plaque microflora and is altered by systemic factors, resulting in direct or host-mediated tissue injury. The basic motive of periodontal therapy is to halt the disease's progression by removing and altering the factors causing the disease by either surgical or non-surgical means [1].
Lasers have become an indispensable part of periodontal therapy. Since the last decade, lasers have assisted successfully in almost every treatment procedure in periodontics. Maiman created the first laser apparatus in 1960, and Goldman used it in dentistry in 1964 [2]. Soft tissue surgical procedures using lasers were found to achieve hemostasis and sterilization with minimal post-operative pain in contrast to conventional surgical procedures. The conclusive goal of any periodontal therapy is to attain sustainable regeneration and simultaneously ensure that the healing period is brief and comfortable for the patient [2-4]. Low-level laser therapy (LLLT) has also been examined recently for its effectiveness in the medical and dentistry fields' wound-healing procedures. Low-level laser therapy, or photobiostimulation, was first proposed by Mester and his co-workers in 1967 [3].
LLLT refers to the non-invasive photon bombardment of tissues, which brings about multiple changes at the cellular and molecular levels. Adenosine triphosphate (ATP), ribonucleic acid (RNA), and protein synthesis are said to be increased as a result of the cytochrome c oxidase, a mitochondrial component of the cellular respiratory chain, absorbing high-energy photons. After an open flap debridement (OFD), lasers are better at regenerating periodontal tissue and reducing the number of germs in the mouth.
LLLT, or photobiological stimulation, brings about various biological changes that not only assist in the acceleration of healing wounds but also help in the attenuation of pain and postoperative discomfort. LLLT is also known as "soft laser therapy" and "laser biostimulation" [3]. The bio-stimulatory and inhibitory effects of LLLT are governed by the Arndt-Schulz law, wherein low doses increase physiological processes and strong stimuli inhibit physiological activity. LLLT with a diode laser has stimulatory effects on human gingival fibroblasts, which may have beneficial effects on healing wounds [4]. Wound healing certainly starts at the moment of injury, but the complete healing phase almost lasts up to 1 or 2 years. The tissue repair phase, which begins on the third day and lasts about two weeks, involves fibroblast proliferation and collagen genesis. The supplementation of LLLT at this point in time leads to expedited wound-healing events [4].
This study aims at a clinical evaluation of early wound healing and a biochemical evaluation of inflammatory mediators in gingival crevicular fluid (GCF) following LLLT with an open flap debridement (OFD) in periodontal therapy.
## Materials and methods
After getting approval from the institutional ethical committee (reference number IEC/Perio/$\frac{4}{19}$), 40 patients with Stage II Grade B periodontitis who were willing to participate in the study were recruited from the Department of Periodontology, ITS Dental College and Research Centre, Greater Noida, India as per case definition of American Academy of Periodontology (AAP) classification 2017 [5] with residual pocket depths of ≥ 5 mm and <7 mm post-phase 1 therapy affecting a minimum of three sites each in at least two quadrants and radiographic evidence of horizontal bone loss were included in the study. While patients with diabetes, hypertension, atherosclerosis, and other systemic diseases, including pregnant and lactating women, smokers (current or smoking within the last 5 years), patients allergic to medications, and patients with a recent history of antibiotic use (within the last 3 months) known to affect periodontal health, were excluded for the study.
The randomized controlled trial was carried out by tossing the coin where bio-stimulation with a diode laser (890 nm) after OFD was used to treat the site on one side (test group), while open flap debridement was used to treat the site on the other side (OFD; control group). The surgical site was anesthetized with a $2\%$ lignocaine hydrochloride solution with adrenaline (1:80 000) in the control group (OFD). There were made intra-crevicular incisions. The roots were meticulously scaled and planed, and the granulation tissue was removed from the defects. There was no root surface conditioning carried out. After that, simple interrupted sutures were used to close the control sites. While in test groups, the patient and every member of the operating room staff were required to wear safety laser glasses. Sweeping movements were made with the tip in the pocket apico-coronally (vertically) and mesiodistally (horizontally). Laser bio-stimulation was carried out with a power output of 1.5 watts in non-contact mode at baseline, the third, and the seventh day after flap surgery for 30 seconds twice with a 60-second interval in the test group.
Each recall visit included a clinical evaluation of wound healing using the Landry index. According to redness, the presence of granulation tissues, bleeding, suppuration, and epithelialization, the healing index (HI) assigns a score to the healing process. A scale of 1 to 5 was used, with 1 representing very poor tissue repair and 5 representing good tissue healing. The healing is more effective the higher the score. According to the clinical assessment, this index rates the surgical wound. Landry wound healing index [6] was taken down at the 1st and 2nd weeks, and clinical parameters, viz., plaque index (PI), gingival index (GI), pocket probing depth (PPD), clinical attachment level (CAL), and GCF inflammatory mediators tumor necrosis factor (TNF) alpha, IL-6, matrix metalloproteinases (MMP-8), osteoprotegerin (OPG) were estimated at 0, 3, and 6 months. An enzyme-linked immunoassay (ELISA) kit (Chongqing Biospes Co. Ltd., Chongqing, China) was used to analyze the inflammatory mediators.
After gathering, sorting, and tabulating all the data, descriptive and analytical statistics were used. Since the data did not follow a normal distribution, a nonparametric test called the Shapiro-Wilk test was employed to examine the information. The Mann-Whitney U test and Wilcoxon signed-ranked test were employed to examine differences between the groups. Software called Statistical Package for the Social Sciences version 20.1 was used to conduct all of the statistical analyses (IBM Corp, Armonk, USA).
## Results
Two hundred and forty test and control sites from 40 systemically healthy periodontitis patients of both genders with probing pocket depths between 5 and 7 mm and radiographic evidence of horizontal bone loss were selected. Open flap debridement with and without application of LLLT was done in test and control groups, following which clinical parameters, GCF biomarkers, and the Landry early wound healing index were recorded at subsequent intervals.
Comparisons of clinical parameters (plaque index, gingival index, probing pocket depth, and clinical attachment levels) at baseline, 3 months, and 6 months are shown in Table 1.
**Table 1**
| Clinical Parameters | Time interval | Test Group N=120 sites | Control Group N=120 sites | Intergroup Comparison |
| --- | --- | --- | --- | --- |
| | | Mean ± SD | Mean ± SD | P value |
| Plaque Index | At Baseline | 2.14± 0.39 | 2.32±0.32 | 0.056 |
| Plaque Index | 3 months | 1.35±1.197 | 0.98±0.34 | 0.081 |
| Plaque Index | 6 months | 0.83± 0.30 | 0.74±0.17 | 0.107 |
| Gingival index | At Baseline | 2.44± 0.24 | 2.06±0.48 | <0.001 |
| Gingival index | 3 months | 1.25± 0.42 | 1.41±0.63 | 0.120 |
| Gingival index | 6 months | 0.98±0.41 | 0.91±0.35 | 0.443 |
| Pocket Probing Depth | At Baseline | 6.40± 0.64 | 6.47±0.54 | 0.549 |
| Pocket Probing Depth | 3 months | 4.30± 0.87 | 4.09±0.74 | 0.272 |
| Pocket Probing Depth | 6 months | 3.19±0.60 | 3.21±0.57 | 0.859 |
| Clinical attachment level | At Baseline | 3.22±0.65 | 3.26±0.90 | 0.756 |
| Clinical attachment level | 3 months | 2.38±0.59 | 2.49±0.69 | 0.255 |
| Clinical attachment level | 6 months | 1.42±0.35 | 1.54±0.72 | 0.220 |
From the start of the study to 6 months later, there was a statistically significant decrease in plaque index, gingival index, probing pocket depth, and clinical attachment levels in both groups. However, when the two groups were compared, there were no significant differences at any time.
GCF TNF alpha levels decreased in both the test group and control group from baseline to 3 months and 6 months. In intergroup comparison, there was a statistically significant reduction in the test group as compared to the control group at 6 months (p-value < 0.036). There was a decline in GCF IL-6 levels from baseline to 3 months and 6 months in both the groups but results when analyzed statistically were not significant on intergroup and intragroup comparison at any time interval (p-value < 0.265). GCF MMP 8 levels were decreased in both the test and control groups from baseline to 6 months. In intergroup comparison, the reduction in the test group was more than the control group and the results were statistically significant (p-value < 0.005). There was an increase in GCF OPG levels from baseline to 3 months and 6 months in both groups but the escalation in the test group was significantly higher than that of the control group (p-value < 0.023) (Table 2).
**Table 2**
| Biomarkers | Biomarkers.1 | Test Group N=120 sites | Control Group N=120 sites | Intergroup Comparison |
| --- | --- | --- | --- | --- |
| | | Mean ± SD N=120 sites | Mean ± SD N=120 sites | P value |
| TNF alpha | At Baseline | 27.34±6.39 | 28.10±6.03 | 0.556 |
| TNF alpha | 3 months | 23.08±7.59 | 24.56±7.30 | 0.486 |
| TNF alpha | 6 months | 20.39±4.73 | 21.47±4.22 | 0.036 |
| IL-6 | At Baseline | 20.22±3.45 | 20.31±3.61 | 0.95 |
| IL-6 | 3 months | 17.51±2.8 | 18.78±3.3 | 0.37 |
| IL-6 | 6 months | 16.19±3.14 | 17.86±3.3 | 0.265 |
| MMP 8 | At Baseline | 373.71±68.58 | 365.10±61.85 | 0.203 |
| MMP 8 | 3 months | 211.18±53.46 | 240.80±59.85 | 0.139 |
| MMP 8 | 6 months | 95.20±18.91 | 118.04±22.51 | 0.005 |
| OPG | At Baseline | 84.11±24.23 | 87.22±25.54 | 0.39 |
| OPG | 3 months | 173.30±30.01 | 158.70±28.30 | 0.138 |
| OPG | 6 months | 180.80±31.72 | 153.30±25.32 | 0.023 |
On comparison of the Landry Wound Healing Index, the baseline values for the test and control groups were 3±0.67 and 3.2±0.68, respectively, with no statistical difference in the scores. The values on the 7th day for the test and control groups were 3.5±0.67 and 2.5±0.50, respectively, showing statistically significant better wound healing in the test group (p-value < 0.001, S). The Landry Wound Healing Index values in the 2nd week were 4±0.49 and 3±0.75, showing again statistically significant improved healing in the test group as compared to the control group (p-value < 0.001) (Table 3).
**Table 3**
| Unnamed: 0 | Test sites N=120 | Control sites N=120 | P-value of Intergroup Comparison |
| --- | --- | --- | --- |
| | Mean ± SD | Mean ± SD | |
| At Baseline | 3±0.67 | 3.2±0.68 | Not significant |
| 1 week | 3.5±0.67 | 2.5±0.50 | <0.001, S |
| 2 week | 4±0.49 | 3±0.75 | <0.001, S |
## Discussion
Periodontitis is a prevalent disease all over the world. Environmental conditions, the immune response generated by the host, and pathogenic microflora all play an important role in the progression of the disease. Many recent studies have revealed that diode lasers have been successful in enhancing periodontal wound healing after surgical procedures. This study was done on 40 people with periodontitis to see how well LLLT helped early periodontal wounds heal after OFD and to compare clinical and biochemical parameters after OFD with and without laser bio-stimulation.
The PI reflects the patient's overall hygiene status; it is assessed postoperatively according to the index of Silness and Loe [7] which shows a trend toward progressively improving oral hygiene, whereas the GI records gingival inflammation. Dental plaque is a hub of microbes or biofilm that is embedded in an extracellular gelatinous matrix and contributes to periodontal disease [7]. The local cause of inflammation is the buildup of plaque, which causes frequent micro-ulcerations in the epithelium that lines the soft-tissue wall of a periodontal pocket [8]. To estimate the increasing soft tissue destruction, disease enhancement, and response to periodontal treatment, measurements of PPD and CAL are the most frequently used informative parameters [9]. In the present study, there was a statistically significant reduction in the scores of PI, GI, PPD, and CAL from baseline to 6 months in both groups, but the intergroup comparison failed to show a statistical difference. The explanation of improvement in both groups is possibly due to the efficiency done OFD along with proper oral hygiene practices followed by the participants on a regular basis in both groups.
Similar to our findings, Khan F et al. [ 2021] [10], Gokhale SR et al. [ 2012] [11], and Jonnalagadda BD et al. [ 2018] [12] did studies and reported non-significant results on the comparison of tests and control groups showing no additional benefits of diode laser over conventional flap surgery. Kolamala et al. [ 2022] [13], on the other hand, found that the clinical parameters of treatment with diode laser were better than treatment with OFD alone from the start to 6 months. Different lasers, including the diode, Nd:YAG, CO2, Er:YAG, and Er,Cr:YSGG, have been suggested and are anticipated to act as an alternative or supplemental therapy to traditional, mechanical periodontal therapy. Other benefits include hemostasis, reduced postoperative edema, a decrease in the number of microorganisms at the surgical site, fewer sutures required, quicker healing, and less postoperative discomfort.
Periodontitis is a local inflammatory process, and invasion of the periodontium is mediated through inflammatory markers. It is observed that host monocytes, macrophages, fibroblasts, and endothelial cells respond immediately, secreting chemokines and inflammatory markers against bacteria and their toxins [7]. Due to the complexities of periodontitis, relying on a single biomarker may not yield relevant results. To assess biological changes, the GCF assays of four periodontal disease biomarkers (TNF, IL-6, MMP-8, and OPG) were estimated. A description depicting the source and clinical significance in states of health and disease is done, but the values reflect a wide variation in range, possibly due to methodological disparities in sampling, laboratory settings, storage temperature, and methods of assessment. But as compared to traditional methods of assessment, these biomarkers demonstrate optimum prediction of site-specific prognosis and outcome of therapy. Although precise cutoff values for these biomarkers cannot be determined, the increase and decrease of these biomarkers serve as an important tool for assessing biological events occurring at the molecular level in the pocket lining [4,14-18].
TNF-alpha primary sources are monocytes and macrophages and secondary sources are fibroblasts and endothelial cells. It induces the secretion of collagenase by fibroblast, resorption of cartilage and bone, and has been implicated in the destruction of periodontal tissues in periodontitis leading to the synthesis of IL-1 and prostaglandin E2 (PGE 2), also activating osteoclasts and thus inducing bone resorption. It has synergistic effects with the bone resorptive actions of IL-1b. When assessed, a statistically significant reduction was seen in TNF-alpha levels in the test group as compared to the control group. TNF-alpha plays a definite role in alveolar bone loss. This significant decline in TNF-alpha at 6 months in test sites indicates decreased osteoclastic activity and inflammation, which may lead to a conducive environment for bone regeneration. Literature suggests that prolonged LLLT applications are able to lower the tissue concentrations of TNF-alpha by almost $70\%$ and help reach a level closer to healthy, non-inflamed gingiva [14]. Laser treatment reduces periodontopathic microbes, lowering TNF-alpha levels. This could be corroborated by a study performed by Aimbire F et al. [ 2006] [19], who found a reduction in TNF levels after low-level laser therapy in periodontal surgical procedures. The source of OPG is osteoblasts, which play a vital role in inhibiting the differentiation and activation of osteoclasts. A significant increase in GCF OPG levels is seen at 6 months in the test sites compared to the control sites. Bone regeneration is regulated by the interplay of different cytokines. OPG is an osseoprotective marker, and its increase depicts a reduction in the inflammatory process and signifies that the molecular mechanism of bone repair is upregulated in areas where LLLT was done. Huck et al. reported higher levels of salivary OPG in treated periodontitis patients in the maintenance phase than in untreated periodontitis patients [20].
MMP-8 primary sources are monocytes and macrophages, while secondary sources are the fibroblasts and endothelial cells. MMP-8 is a chief collagenase and depicts the state of health and disease in subgingival periodontal tissues. In our study, a decline in MMP-8 levels was seen in both test and control sites, signifying the transition from a state of disease to health following open flap debridement in both groups. There is a statistically significant decrease in the test group when compared to the control group, showing an additional benefit of LLLT on the healing apparatus. MMP-8 is described as a surrogate indicator of neutrophil density as it is stored in secretory granules and released during inflammation. MMP-8 and MMP-9 are the main collagenases for the degradation of type I and type III collagen [18]. Increased levels of host and microbial-derived proteinases (MMPs) are responsible for extracellular matrix degradation in periodontal disease. As a result, lower MMP-8 counts indicate a stable healing apparatus devoid of disease activity at LASER-treated sites compared to control sites.
Similar studies by Deshmukh et al. [ 2018] [21] and Saglam et al. [ 2014] [22] reported significant improvements in clinical parameters and lower MMP-8 levels in laser-treated periodontal disease sites. Sezen et al. [ 2020] [23], on the other hand, found no statistical difference in MMP-8 reduction between conventional flap surgery and laser-assisted surgery.
IL-6 levels' primary sources are the monocytes and macrophages, and secondary sources are fibroblasts and endothelial cells which are responsible for the production of acute-phase proteins. There was a decline in IL-6 levels from baseline to 6 months in both groups, but intergroup comparison failed to yield any significant changes, though a numerically greater reduction was observed in the sites with LLLT application. IL-6 is produced by different types of cells and expressed by both the host and bacteria. The decrease in IL-6 levels in both groups is indicative of healing events taking place in both groups, whereas insignificant differences in test and control sites are suggestive of the low precision potential of this biomarker in wound healing. Bolyarova et al. [ 2019] [24] discovered that combining SRP with LLLT reduces GCF IL-6 levels. Another study, by Kardeşler et al. [ 2011] [25], found that SRP alone resulted in a significant decrease in the IL-6 level concentration of GCF.
The diversity of the results in different studies strengthens the idea that the production of inflammatory mediators differs from site to site and from subject to subject. One possible reason for this difference is that the release of inflammatory mediators is affected by many things, such as genetics and different types of bacteria. Wound healing is a complex biological process that involves well-organized cellular and biochemical events. On comparison of Landry Wound Healing Index scores, there was a statistically significant improvement in wound healing at the 1st and 2nd week intervals in the test group as compared to the control group. The possible explanation for this could be the absorption of laser light by mitochondrial chromophores, causing photodissociation of nitrous oxide and enhanced enzymatic activity and ATP generation. Thus, LLLT expedites the activation of numerous intracellular signalling pathways, which bring about enhanced cellular proliferation, tissue repair, and regeneration [26]. The right amount of LLLT dose causes the growth of new endothelium and blood vessels, which promotes the creation of granulation tissue and speeds up the healing process. Increased revascularization rate, which is known to have a significant impact on the outcome of wound healing after periodontal surgery, is another effect of LLLT on wound healing. Lingamaneni S et al. [ 2019] [27] discovered that LLLT improves wound healing and keratinization after periodontal surgery. Various histopathological studies have proclaimed improved healing, collagen formation, and homogenization of gingival lamina propria following LLLT [28, 29]. In contrast to this, a study by Damante et al. [ 2004] [30] found no additional benefits of laser therapy. No postoperative problems or poor clinical recovery were associated with this treatment method, suggesting that using laser therapy for pocket therapy may not have any negative effects.
Limitations of the current study are that small sample sizes were used. Patients were monitored for a short duration and blinding was not done. The effect of LLLT was assessed on GCF biomarkers at 3- and 6-month intervals. Future studies with the evaluation of biomarkers at 7 and 14 days should be done to elicit their benefits on early wound healing. Although pain perception was not analyzed in our study, the inclusion of this parameter may have helped us access the pain-attenuating benefits of laser therapy. Our study adds to the evidence that using LLLT to help a wound heal faster is a good idea.
## Conclusions
When it came to the plaque index, gingival index, probing pocket depth, and clinical attachment levels, OFD showed similar results as that of laser treatment. The periodontal flap surgery wound healing was greatly improved by the use of a diode laser following OFD. Establishing efficient laser application techniques will enable the implementation of this innovative treatment in periodontology and increase patient comfort.
## References
1. Sobouti F, Khatami M, Heydari M, Barati M. **The role of low-level laser in periodontal surgeries**. *J Lasers Med Sci* (2015) **6** 45-50. PMID: 25987968
2. Karu T. **Primary and secondary mechanisms of action of visible to near-IR radiation on cells**. *J Photochem Photobiol B* (1999) **49** 1-17. PMID: 10365442
3. Karu T. **Laser biostimulation: a photobiological phenomenon**. *J Photochem Photobiol B* (1989) **3** 638-640. PMID: 2507763
4. Bakshi PV, Setty SB, Kulkarni MR. **Photobiomodulation of human gingival fibroblasts with diode laser - a systematic review**. *J Indian Soc Periodontol* (2022) **26** 5-12. PMID: 35136310
5. Caton JG, Armitage G, Berglundh T. **A new classification scheme for periodontal and peri-implant diseases and conditions - Introduction and key changes from the 1999 classification**. *J Clin Periodontol* (2018) **45 Suppl 20** 0-8
6. Aoki A, Mizutani K, Schwarz F. **Periodontal and peri-implant wound healing following laser therapy**. *Periodontol 2000* (2015) **68** 217-269. PMID: 25867988
7. Carranza F, Newman M, Takei H, Klokkevold P. **Carranza’s Clinical Periodontology**. (2006) **p.** 86
8. Fischman SL. **Current status of indices of plaque**. *J Clin Periodontol* (1986) **13** 371-380. PMID: 3013947
9. Machtei EE, Christersson LA, Grossi SG, Dunford R, Zambon JJ, Genco RJ. **Clinical criteria for the definition of "established periodontitis"**. *J Periodontol* (1992) **63** 206-214. PMID: 1593413
10. Khan F, Chopra R, Sharma N, Agrawal E, Achom M, Sharma P. **Comparative evaluation of the efficacy of diode laser as an adjunct to modified Widman flap surgery for the treatment of chronic periodontitis: a randomized split-mouth clinical trial**. *J Indian Soc Periodontol* (2021) **25** 213-219. PMID: 34158688
11. Gokhale SR, Padhye AM, Byakod G, Jain SA, Padbidri V, Shivaswamy S. **A comparative evaluation of the efficacy of diode laser as an adjunct to mechanical debridement versus conventional mechanical debridement in periodontal flap surgery: a clinical and microbiological study**. *Photomed Laser Surg* (2012) **30** 598-603. PMID: 22889285
12. Jonnalagadda BD, Gottumukkala SN, Dwarakanath CD, Koneru S. **Effect of diode laser-assisted flap surgery on postoperative healing and clinical parameters: a randomized controlled clinical trial**. *Contemp Clin Dent* (2018) **9** 205-212. PMID: 29875562
13. Kolamala N, Nagarakanti S, Chava VK. **Effect of diode laser as an adjunct to open flap debridement in treatment of periodontitis - a randomized clinical trial**. *J Indian Soc Periodontol* (2022) **26** 451-457. PMID: 36339386
14. Noh MK, Jung M, Kim SH. **Assessment of IL-6, IL-8 and TNF-α levels in the gingival tissue of patients with periodontitis**. *Exp Ther Med* (2013) **6** 847-851. PMID: 24137277
15. Hernández M, Baeza M, Contreras J. **MMP-8, TRAP-5, and OPG Levels in GCF diagnostic potential to discriminate between healthy patients', mild and severe periodontitis sites**. *Biomolecules* (2020) **10**
16. Goutoudi P, Diza E, Arvanitidou M. **Effect of periodontal therapy on crevicular fluid interleukin-6 and interleukin-8 levels in chronic periodontitis**. *Int J Dent* (2012) **2012** 362905. PMID: 22114600
17. Hassan SH, El-Refai MI, Ghallab NA, Kasem RF, Shaker OG. **Effect of periodontal surgery on osteoprotegerin levels in gingival crevicular fluid, saliva, and gingival tissues of chronic periodontitis patients**. *Dis Markers* (2015) **2015** 341259. PMID: 25814780
18. Sopi M, Koçani F, Bardhoshi M, Meqa K. **The effect of periodontal therapy on the level of MMP-8 in patients with chronic periodontitis**. *Eur J Dent* (2022)
19. Aimbire F, Albertini R, Pacheco MT. **Low-level laser therapy induces dose-dependent reduction of TNFalpha levels in acute inflammation**. *Photomed Laser Surg* (2006) **24** 33-37. PMID: 16503786
20. Huck O, Buduneli N, Bravo D. **Inflammatory mediators in periodontal pathogenesis**. *Mediators Inflamm* (2019) **2019** 2610184. PMID: 31316297
21. Deshmukh K, Shetty D, Shetty A, Bodduru R, Nayyar AS. **Comparative evaluation of the efficacy of closed pocket debridement with diode laser and periodontal open flap debridement: a clinical and microbiologic study**. *J Clin Sci* (2018) **15** 113-122
22. Saglam M, Kantarci A, Dundar N, Hakki SS. **Clinical and biochemical effects of diode laser as an adjunct to nonsurgical treatment of chronic periodontitis: a randomized, controlled clinical trial**. *Lasers Med Sci* (2014) **29** 37-46. PMID: 23161345
23. Sezen D, Hatipoğlu M, Üstün K. **Evaluation of the clinical and biochemical efficacy of erbium, chromium:ytrium-scandium-gallium-garnet (ER,CR:YSGG) laser treatment in periodontitis**. *Lasers Med Sci* (2020) **35** 1567-1575. PMID: 32215767
24. Bolyarova T, Grozdanova R, Micheva G, Dencheva M, Velikova T. **Effect of diode laser on the level of il-6 in the gingival crevicular fluid in patients with chronic periodontitis**. *Comptes rendus de I’Acad des Sci* (2019) **72** 130-135
25. Kardeşler L, Buduneli N, Çetinkalp S, Lappin D, Kinane DF. **Gingival crevicular fluid IL-6, tPA, PAI-2, albumin levels following initial periodontal treatment in chronic periodontitis patients with or without type 2 diabetes**. *Inflamm Res* (2011) **60** 143-151. PMID: 20845058
26. Akram Z, Al-Shareef SA, Daood U. **Bactericidal efficacy of photodynamic therapy against periodontal pathogens in periodontal disease: a systematic review**. *Photomed Laser Surg* (2016) **34** 137-149. PMID: 26982216
27. Lingamaneni S, Mandadi LR, Pathakota KR. **Assessment of healing following low-level laser irradiation after gingivectomy operations using a novel soft tissue healing index: a randomized, double-blind, split-mouth clinical pilot study**. *J Indian Soc Periodontol* (2019) **23** 53-57. PMID: 30692744
28. Pejcic A, Kojovic D, Kesic L, Obradovic R. **The effects of low level laser irradiation on gingival inflammation**. *Photomed Laser Surg* (2010) **28** 69-74. PMID: 19929224
29. Pesevska S, Nakova M, Gjorgoski I, Angelov N, Ivanovski K, Nares S, Andreana S. **Effect of laser on TNF-alpha expression in inflamed human gingival tissue**. *Lasers Med Sci* (2012) **27** 377-381. PMID: 21380536
30. Damante CA, Greghi SL, Sant'Ana AC, Passanezi E, Taga R. **Histomorphometric study of the healing of human oral mucosa after gingivoplasty and low-level laser therapy**. *Lasers Surg Med* (2004) **35** 377-384. PMID: 15611961
|
---
title: 'An examination of causal associations and shared risk factors for diabetes
and cardiovascular diseases in the East Asian population: A Mendelian randomization
study'
authors:
- Yulin Guo
- Jie Gao
- Yan Liu
- Yanxiong Jia
- Xiangguang An
- Xitao Zhang
- Pixiong Su
journal: Frontiers in Endocrinology
year: 2023
pmcid: PMC9999111
doi: 10.3389/fendo.2023.1132298
license: CC BY 4.0
---
# An examination of causal associations and shared risk factors for diabetes and cardiovascular diseases in the East Asian population: A Mendelian randomization study
## Abstract
### Background
One of the major contributors to disability and mortality among diabetics is cardiovascular disease (CVD), with coronary artery disease (CAD) as the most prevalent type. However, previous studies have provided controversial evidence linking diabetes to other types of CVDs, such as atrial fibrillation (AF). In addition, the risk factors that predispose people to the risk of diabetes and its complications differ across ethnicities, but the disease risk profiles in the East Asian population have been less investigated.
### Methods
The causal association between type 2 diabetes (T2D) and two types of CVDs (i.e., AF and CAD) in the East Asian population was first studied using Mendelian randomization (MR) analyses. Next, we examined the causal effect of 49 traits on T2D and CAD to identify their separate and shared risk factors in East Asians. A causal mediation analysis was performed to examine the role of T2D in mediating the relationship between the identified shared risk factors and CAD.
### Results
T2D was causally associated with CAD, but not AF, in East Asians. A screening of the risk factors indicated that six and 11 traits were causally associated with T2D and CAD, respectively, with suggestive levels of evidence. Alkaline phosphatase (ALP) was the only trait associated with both T2D and CAD, as revealed by the univariable MR analyses. Moreover, the causal association between ALP and CAD no longer existed after adjusting T2D as a covariable in the causal mediation study.
### Conclusion
Our study highlights the risk profiles in the East Asian population, which is important in formulating targeted therapies for T2D and CVDs in East Asians.
## Introduction
Up to $8.8\%$ of the world’s population suffers from diabetes, and International Diabetes Federation projections reveal that by 2040, the number of incidences will have risen to 642 million [1]. One of the main contributors to disability among patients with diabetes is cardiovascular disease (CVD) [2, 3]. The percentage of people with CVD is higher in diabetic patients than in adults without diabetes [4]. CVD leads to the death of roughly $70\%$ of type 2 diabetic patients at and above 65 years old [5]. To elucidate, a systematic review that included 4,549,481 type 2 diabetes (T2D) patients showed an overall CVD prevalence of $32.2\%$ [2]. Coronary artery disease (CAD) ($21.2\%$) was the most common kind of CVD reported [2]. However, previous works have led to controversial conclusion about the association between diabetes and a particular type of CVD, such as atrial fibrillation (AF), the most prevalent type of arrhythmia in the world [6]. For example, a study using a cohort of patients having new-onset AF did not establish the association between the symptoms of AF and diabetes [7].
Ethnic disparities in health conditions are well-recognized [8]. For example, Asian Indians in the US are more likely to have diabetes, although they have lower chance to be obese [9]. In addition, East Asians have more body fat and prone to visceral adiposity at a given body mass index (BMI), which promote the development of diabetes [10]. The risk factors that contribute to the development of diabetes complications also differ across Asian and European populations [11]. Thus, it is important to understand ethnic differences in disease risk profiles to formulate better treatment strategies.
Mendelian randomization (MR) is a method for inferring causation, which reduces the bias owing to reverse causality and residual confounding. In MR analyses, the genetic instruments are used as a proxy for exposures [12]. In causal mediation analyses using a two-step MR design, the direct and indirect effects of exposure on the outcome can also be evaluated [13]. Individual-level data was not applied in MR analyses because these analyses use summary statistics from genome-wide association studies (GWAS), which are normally produced using populations with large sample sizes [14]. In addition, the availability of GWAS datasets makes it easier to screen disease risk factors at the phenome-wide level [15].
In the current study, we first investigated the potential causal association between T2D and two types of CVDs (i.e., AF and CAD) in the East Asian population. Next, we tested the causal effect of 49 traits on T2D and CAD to identify their separate and shared risk factors in East Asians. A causal mediation analysis was also performed to examine the role of T2D in mediating the relationship between identified shared risk factors and CAD.
## Methods
The GWAS dataset for T2D was obtained from the Diabetes Meta-analysis of Trans-ethnic Association Studies (DIAMANTE) Consortium [16], in which GWAS was performed for the East Asian population. For other traits, the method for traits selection (Supplementary Figure 1) was similar to the one used in a recent paper [13]. We only included the GWAS summary statistics datasets generated in the Biobank Japan study [17] to ensure that the MR analyses were conducted using genetic data from East Asians. Detailed information was included in Supplementary Table 1. The causal relationships between 49 traits (Supplementary Figure 1) and T2D/CAD were investigated by univariable MR analyses. For the identified trait (shared risk factor) that can lead to both T2D and CAD, we performed causal mediation analyses, where T2D was deemed as a potential mediator. A reciprocal link between mediator and exposure was not permitted in the mediation studies, so it was necessary to conduct a reverse univariable MR to infer whether these traits could be induced by T2D. The direct effect of trait (shared risk factor) on CAD was estimated using multivariable MR, in which T2D was adjusted as a covariable. The product of the beta coefficient of the effect of trait (shared risk factor) on T2D and the beta coefficient of the association between T2D and CAD (with trait adjusted as covariable) represented the indirect effects of trait (shared risk factor) on CAD.
In the univariable MR studies, the instrumental variables (IVs) used for exposure traits were selected according to various factors. First, the phenotypes should be highly associated with IVs ($P \leq 5$×10−8). Second, a linkage disequilibrium (LD) of R2 < 0.001 and clumping with a 10-Mb window were used to ensure that the IVs were not related to each other. Third, each trait’s IVs should have at least five variants as biallelic single-nucleotide polymorphisms (SNPs). In the univariable MR studies, the inverse-variance weighted (IVW) method, weighted median method, and MR-Egger were used, with the IVW approach being regarded as the primary method. Potential horizontal pleiotropy was examined using the MR-Egger intercept test. A $5\%$ false-discovery rate (FDR) was used to correct multiple comparisons. The code for the MR studies was modified from a recent work [13], in which the R packages TwoSampleMR and MVMR, respectively, were applied to conducted the MR analyses.
## Results
The results of the MR analysis using the IVW approach indicated a significant association between genetically predicted T2D and CAD ($$P \leq 6.63$$×10−5) (Figure 1 and Supplementary Figure 2). However, no causal association between T2D and AF was observed ($$P \leq 0.97$$) (Figure 1 and Supplementary Figure 2). The same relationship trajectory was apparent in the MR sensitivity analyses using the weighted median and MR-Egger methods (Figure 1). Moreover, a leave-one-out sensitivity analysis suggested that not a single SNP was responsible for the causal effect of T2D on CAD (Supplementary Figure 3). The intercept term of the MR-Egger method was applied to examine the horizontal pleiotropy, which revealed that it was not significant ($$P \leq 0.34$$) in the studies.
**Figure 1:** *Scatter plots (A) and forest plots (B) showing the results of Mendelian randomization (MR) analyses studying the causal association between T2D and cardiovascular diseases in the East Asian population.*
After confirming the causal effect of T2D on CAD in the East Asian population, we next examine the separate and shared risk factors of these two diseases by including the GWAS summary datasets of 49 traits from the Biobank Japan study (Supplementary Table 1), in accordance with the criteria indicated in the flowchart (Supplementary Figure 1). Univariable MR analyses indicated that out of the 49 traits, six were associated with T2D at suggestive levels of evidence ($P \leq 0.05$) (Figure 2 and Supplementary Tables 2 - 3). Three of these six traits (i.e., hemoglobin A1c, blood sugar, and red blood cell count) survived $5\%$ FDR correction for multiple comparisons (Supplementary Table 4). Reverse MR analyses suggested that alkaline phosphatase was the only trait that could not be altered by T2D (Figure 3 and Supplementary Tables 2 - 3). Eleven of 49 traits showed causal association with CAD at suggestive levels of evidence ($P \leq 0.05$) (Figure 4 and Supplementary Tables 2 - 3), and six of the 11 traits, namely, total cholesterol (TC), triglycerides, low-density-lipoprotein cholesterol (LDL), high-density lipoprotein cholesterol (HDL), alkaline phosphatase (ALP), and activated partial thromboplastin time (APTT), survived $5\%$ FDR correction (Supplementary Table 4). Thus, the results revealed that ALP was causally associated with both T2D and CAD (Figure 5A), and the following causal mediation analysis based on two-step MR indicated that ALP was no longer associated with CAD after adjusting T2D as a covariable in the multivariable MR (Figure 5B).
**Figure 2:** *Forest plots showing the causal effect of traits on T2D with suggestive levels of evidence (P < 0.05) in the East Asian population.* **Figure 3:** *Forest plots showing the causal effect of T2D on traits with suggestive levels of evidence (P < 0.05) in the East Asian population.* **Figure 4:** *Forest plots showing the causal effect of traits on CAD with suggestive levels of evidence (P < 0.05) in the East Asian population.* **Figure 5:** *Shared and independent risks of T2D and CAD are presented in a Venn diagram (A), and the total, indirect, and direct effects of alkaline phosphatase (ALP) on CAD are studied by causal mediation analyses (B) in the East Asian population.*
## Discussion
In the present study, the results of MR analyses suggested that T2D was causally associated with CAD, but not AF, in East Asians. The screening of the risk factors indicated that six and 11 traits were causally associated with T2D and CAD, respectively, with suggestive levels of evidence. ALP was the only trait associated with both T2D and CAD, as revealed by the univariable MR analyses. The causal association between ALP and CAD no longer existed after adjusting T2D as a covariable in the causal mediation study (direct effect).
T2D can approximately shorten life expectancy by a decade, and CVD is a major cause of death in T2D patients [18]. However, the association of T2D and AF, as well as the exact pathophysiology of AF in diabetes patients, has not been fully established [19]. The Framingham Heart Study correlated elevated glycemic levels with an increased risk of AF [20]. Moreover, diabetes patients with AF had increased rates of overall and cardiovascular mortality, coupled with a decline in life quality compared with patients who only had AF but were not diabetic [21]. However, a correlation between diabetes and non-paroxysmal AF was not observed [22]. Diabetes cannot independently lead to AF after confounder adjustment, according to a survey in China [23]. Thus, it is still unknown whether there is a causative association between diabetes and AF. Our MR study using genetic data from the East Asian population suggested that diabetes could not causally lead to AF.
CVD is a significant contributor to comorbidity and mortality among T2D patients, with CAD having the highest prevalence rate [2]. Research has indicated that patients with diabetes have a higher susceptibility to CAD compared with non-diabetics [24]. We consistently observed a causal association of diabetes with CAD in the East Asian population. Several reasons, such as insulin resistance, dyslipidemia, and hyperglycemia, have been postulated to explain the high vulnerability to CAD among patients with diabetes. These processes can be linked to abnormal functioning of the platelets, causing vascular smooth muscle dysfunction, and irregularity in the functioning of endothelial cells [25]. Indeed, atherosclerotic plaques in diabetic patients are often lipid-laden, making them more prone to rupture compared with those of people without diabetes [26]. In addition, critical to atherosclerosis is the process of inflammation, whose activation in T2D is often linked to insulin resistance and obesity [27]. Hyperglycemia has also been linked with the promotion of epigenetic alterations that initiate the over-expression of genes linked to vascular inflammation, thus establish a basis for atherosclerosis and endothelial dysfunction [28].
The Collaborative Analysis of *Diagnostic criteria* in Europe (DECODE) study indicated that the prevalence of diabetes was higher in urban Chinese and Japanese patients aged 30–69 years than in Europeans [29]. Young patients have a higher chance to experience β-cell failure and long-lasting disease, making them have a higher risk for microvascular and macrovascular problems [10]. For example, patients with T2D from East Asia are more likely than those from Europe to experience renal issues [10]. One of the potential reasons for this interethnic disparity is that Asians, at a given BMI, usually have higher visceral adiposity compared to Caucasians, which is likely to be more harmful and can cause insulin resistance [30]. For example, American Japanese patients have higher level of visceral adiposity than their Caucasian counterparts [31]. For other race, the association between metabolic parameters and CVD can be different in Black and White population, and ethnicity is also responsible for the disparities in the metabolic syndrome associated CVD and T2D [32]. Because of the ethnic differences in risk profiles, we screened and identified the independent and shared risk factors of diabetes and CAD using GWAS summary data generated from the East Asian population.
Red blood cell (RBC) changes are likely to happen in diabetes patients [33]. For example, red blood cell parameters are correlated with glycemic control among adult patients with T2D in Eastern Ethiopia [34]. Consistent with the literature, the causality interference by MR analyses in the present study suggested that RBC count was negatively associated with diabetes risk. Long-term hyperglycemia leads to the production of free oxygen radicals and the irreversible glycation of hemoglobin and RBC membrane proteins, resulting in a relative drop in RBC count [35]. Thus, these processes make RBCs become less deformable and have a reduced chance of survival [36].
A moderate to very significant correlation between triglyceride levels and the risk of coronary heart disease has been observed [37]. The measurement of TC is helpful in estimating CVD risk and making clinical decision for the start of statin therapy [38]. Indeed, the risk of coronary heart disease increases by $24\%$ for males and $20\%$ for females for every 1 mmol/L increase in TC [39]. For LDL cholesterol, angiographic trials confirm the significance of LDL cholesterol reduction in reducing the risk of CAD [40]. Widespread epidemiological research indicates that low levels of HDL are a sign of increased cardiovascular risk [41]. Consistent with these clinical observations, our MR analyses identified the causal association between lipid profile and CAD in the East Asian population. As a common coagulation screening test, the measurement of APTT can be used to estimate intrinsic coagulation pathway activity [42]. The degree and severity of coronary stenosis can be estimated by using APTT in individuals undergoing coronary angiography; notably, the patients who had ST-Segment Elevation Myocardial Infarction (STEMI) had low APTT values [43]. Moreover, a short APTT is correlated with higher thrombin production and an increased risk for thrombosis [44]. Consistently, our study revealed that higher APTT was causally associated with decreased CAD risk in the East Asian population.
ALP is a plasma membrane-anchored enzyme that is widely present in nature [45]. A correlation between baseline serum ALP levels and new-onset diabetes has been established in hypertensive individuals [46]. In an Iranian population, the level of ALP and the risk of coronary heart disease were independently correlated [47]. However, ALP was not associated with diabetes, according to the results of MR research, which only included persons of European ancestry [48]. Our MR analysis in an East Asian population indicated that ALP was negatively associated with both diabetes and CAD, and the association between ALP and CAD was not significant after adjusting T2D in the multivariable MR. Mechanistically, ALP reduce the bioavailability of nitric oxide (NO), leading to an altered endothelial NO synthase activity [46]. Besides serum ALP, intestinal alkaline phosphatase (IAP), as a membrane-bound glycoprotein mainly expressed in proximal small intestine, is also related to T2D [49]. For instance, T2D can be observed in mice lacking IAP [50]. Additionally, oral administration of IAP protects and even reverses high-fat-diet-induced T2D in wild-type mice by reducing metabolic endotoxemia and detoxifying lipopolysaccharides (LPS) [49].
This study has several limitations. First, a relatively high level of multiple comparison burden may exist when many traits are included in the analyses. To address this point, we also presented the results with suggestive levels of evidence. Second, as an inherent drawback, an MR study cannot completely rule out the potential horizontal pleiotropy. Thus, we used multiple MR methods as sensitivity analyses to enhance the credibility of our conclusion.
## Conclusion
T2D is causally associated with CAD, but not AF, in the East Asian population. Multiple traits were identified as separate risk factors of T2D or CAD. A mediating effect of T2D on the association between ALP and CAD was observed. Our study highlights the risk profiles in the East Asian population, which is important for formulating targeted therapies for T2D and CVDs in East Asians.
## Data availability statement
The original contributions presented in the study are included in the article/ Supplementary Material. Further inquiries can be directed to the corresponding authors.
## Author contributions
YG and PS: conception and design, data analysis, and interpretation. JG, YL, YJ, XA, and XZ: collection and assembly of data, and prepared the manuscript. All authors contributed to the article and approved the submitted version.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fendo.2023.1132298/full#supplementary-material
## References
1. Lovic D, Piperidou A, Zografou I, Grassos H, Pittaras A, Manolis A. **The growing epidemic of diabetes mellitus**. *Curr Vasc Pharmacol* (2020) **18**. DOI: 10.2174/1570161117666190405165911
2. Einarson TR, Acs A, Ludwig C, Panton UH. **Prevalence of cardiovascular disease in type 2 diabetes: A systematic literature review of scientific evidence from across the world in 2007-2017**. *Cardiovasc Diabetol* (2018) **17** 83. DOI: 10.1186/s12933-018-0728-6
3. Li S, Wang J, Zhang B, Li X, Liu Y. **Diabetes mellitus and cause-specific mortality: A population-based study**. *Diabetes Metab J* (2019) **43**. DOI: 10.4093/dmj.2018.0060
4. Ceriello A, Prattichizzo F. **Variability of risk factors and diabetes complications**. *Cardiovasc Diabetol* (2021) **20** 101. DOI: 10.1186/s12933-021-01289-4
5. Kannel WB, McGee DL. **Diabetes and cardiovascular disease. the framingham study**. *Jama.* (1979) **241**. DOI: 10.1001/jama.1979.03290450033020
6. Şerban RC, Scridon A. **Data linking diabetes mellitus and atrial fibrillation-how strong is the evidence? from epidemiology and pathophysiology to therapeutic implications**. *Can J Cardiol* (2018) **34**. DOI: 10.1016/j.cjca.2018.08.018
7. Reynolds MR, Lavelle T, Essebag V, Cohen DJ, Zimetbaum P. **Influence of age, sex, and atrial fibrillation recurrence on quality of life outcomes in a population of patients with new-onset atrial fibrillation: The fibrillation registry assessing costs, therapies, adverse events and lifestyle (FRACTAL) study**. *Am Heart J* (2006) **152**. DOI: 10.1016/j.ahj.2006.08.011
8. Lillie-Blanton M, Laveist T. **Race/ethnicity, the social environment, and health**. *Soc Sci Med* (1996) **43** 83-91. DOI: 10.1016/0277-9536(95)00337-1
9. Mohanty SA, Woolhandler S, Himmelstein DU, Bor DH. **Diabetes and cardiovascular disease among Asian indians in the united states**. *J Gen Intern Med* (2005) **20**. DOI: 10.1111/j.1525-1497.2005.40294.x
10. Ma RC, Chan JC. **Type 2 diabetes in East asians: Similarities and differences with populations in Europe and the united states**. *Ann N Y Acad Sci* (2013) **1281** 64-91. DOI: 10.1111/nyas.12098
11. Luk A, Chan JC. **Diabetic nephropathy–what are the unmet needs**. *Diabetes Res Clin Pract* (2008) **82**. DOI: 10.1016/j.diabres.2008.09.033
12. Michaëlsson M, Yuan S, Melhus H, Baron JA, Byberg L, Larsson SC. **The impact and causal directions for the associations between diagnosis of ADHD, socioeconomic status, and intelligence by use of a bi-directional two-sample mendelian randomization design**. *BMC Med* (2022) **20** 106. DOI: 10.1186/s12916-022-02314-3
13. Walker VM, Vujkovic M, Carter AR, Davies NM, Udler MS, Levin MG. **Separating the direct effects of traits on atherosclerotic cardiovascular disease from those mediated by type 2 diabetes**. *Diabetologia.* (2022) **65**. DOI: 10.1007/s00125-022-05653-1
14. Hemani G, Zheng J, Elsworth B, Wade KH, Haberland V, Baird D. **The MR-base platform supports systematic causal inference across the human phenome**. *Elife* (2018) **7**. DOI: 10.7554/eLife.34408
15. Nikpay M, Mohammadzadeh S. **Phenome-wide screening for traits causally associated with the risk of coronary artery disease**. *J Hum Genet* (2020) **65**. DOI: 10.1038/s10038-019-0716-z
16. Mahajan A, Spracklen CN, Zhang W, Ng MCY, Petty LE, Kitajima H. **Multi-ancestry genetic study of type 2 diabetes highlights the power of diverse populations for discovery and translation**. *Nat Genet* (2022) **54**. DOI: 10.1038/s41588-022-01058-3
17. Nagai A, Hirata M, Kamatani Y, Muto K, Matsuda K, Kiyohara Y. **Overview of the BioBank Japan project: Study design and profile**. *J Epidemiol* (2017) **27** S2-s8. DOI: 10.1016/j.je.2016.12.005
18. Cubbon R, Kahn M, Kearney MT. **Secondary prevention of cardiovascular disease in type 2 diabetes and prediabetes: A cardiologist's perspective**. *Int J Clin Pract* (2008) **62**. DOI: 10.1111/j.1742-1241.2007.01646.x
19. Wang A, Green JB, Halperin JL, Piccini JP. **Atrial fibrillation and diabetes mellitus: JACC review topic of the week**. *J Am Coll Cardiol* (2019) **74**. DOI: 10.1016/j.jacc.2019.07.020
20. Benjamin EJ, Levy D, Vaziri SM, D'Agostino RB, Belanger AJ, Wolf PA. **Independent risk factors for atrial fibrillation in a population-based cohort. the framingham heart study**. *Jama.* (1994) **271**. DOI: 10.1001/jama.1994.03510350050036
21. Echouffo-Tcheugui JB, Shrader P, Thomas L, Gersh BJ, Kowey PR, Mahaffey KW. **Care patterns and outcomes in atrial fibrillation patients with and without diabetes: ORBIT-AF registry**. *J Am Coll Cardiol* (2017) **70**. DOI: 10.1016/j.jacc.2017.07.755
22. Bano A, Rodondi N, Beer JH, Moschovitis G, Kobza R, Aeschbacher S. **Association of diabetes with atrial fibrillation phenotype and cardiac and neurological comorbidities: Insights from the Swiss-AF study**. *J Am Heart Assoc* (2021) **10**. DOI: 10.1161/JAHA.121.021800
23. Zhou Z, Hu D. **An epidemiological study on the prevalence of atrial fibrillation in the Chinese population of mainland China**. *J Epidemiol* (2008) **18**. DOI: 10.2188/jea.JE2008021
24. Patsouras A, Farmaki P, Garmpi A, Damaskos C, Garmpis N, Mantas D. **Screening and risk assessment of coronary artery disease in patients with type 2 diabetes: An updated review**. *In Vivo* (2019) **33**. DOI: 10.21873/invivo.11572
25. Bornfeldt KE, Tabas I. **Insulin resistance, hyperglycemia, and atherosclerosis**. *Cell Metab* (2011) **14**. DOI: 10.1016/j.cmet.2011.07.015
26. Marfella R, D'Amico M, Esposito K, Baldi A, Di Filippo C, Siniscalchi M. **The ubiquitin-proteasome system and inflammatory activity in diabetic atherosclerotic plaques: Effects of rosiglitazone treatment**. *Diabetes.* (2006) **55**. DOI: 10.2337/diabetes.55.03.06.db05-0832
27. Black PH. **The inflammatory consequences of psychologic stress: Relationship to insulin resistance, obesity, atherosclerosis and diabetes mellitus, type II**. *Med Hypotheses* (2006) **67**. DOI: 10.1016/j.mehy.2006.04.008
28. De Rosa S, Arcidiacono B, Chiefari E, Brunetti A, Indolfi C, Foti DP. **Type 2 diabetes mellitus and cardiovascular disease: Genetic and epigenetic links**. *Front Endocrinol (Lausanne)* (2018) **9**. DOI: 10.3389/fendo.2018.00002
29. Qiao Q, Hu G, Tuomilehto J, Nakagami T, Balkau B, Borch-Johnsen K. **Age- and sex-specific prevalence of diabetes and impaired glucose regulation in 11 Asian cohorts**. *Diabetes Care* (2003) **26**. DOI: 10.2337/diacare.26.6.1770
30. Lim U, Ernst T, Buchthal SD, Latch M, Albright CL, Wilkens LR. **Asian Women have greater abdominal and visceral adiposity than Caucasian women with similar body mass index**. *Nutr Diabetes* (2011) **1** e6. DOI: 10.1038/nutd.2011.2
31. Fujimoto WY, Boyko EJ, Hayashi T, Kahn SE, Leonetti DL, McNeely MJ. **Risk factors for type 2 diabetes: Lessons learned from Japanese americans in Seattle**. *J Diabetes Investig* (2012) **3**. DOI: 10.1111/j.2040-1124.2012.00195.x
32. Osei K, Gaillard T. **Disparities in cardiovascular disease and type 2 diabetes risk factors in blacks and whites: Dissecting racial paradox of metabolic syndrome**. *Front Endocrinol (Lausanne)* (2017) **8**. DOI: 10.3389/fendo.2017.00204
33. Martínez M, Vayá A, Server R, Gilsanz A, Aznar J. **Alterations in erythrocyte aggregability in diabetics: The influence of plasmatic fibrinogen and phospholipids of the red blood cell membrane**. *Clin Hemorheol Microcirc* (1998) **18**
34. Arkew M, Asmerom H, Tesfa T, Tsegaye S, Gemechu K, Bete T. **Red blood cell parameters and their correlation with glycemic control among type 2 diabetic adult patients in Eastern Ethiopia: A comparative cross-sectional study**. *Diabetes Metab Syndr Obes* (2022) **15**. DOI: 10.2147/DMSO.S386093
35. Cho YI, Mooney MP, Cho DJ. **Hemorheological disorders in diabetes mellitus**. *J Diabetes Sci Technol* (2008) **2**. DOI: 10.1177/193229680800200622
36. Wang Y, Yang P, Yan Z, Liu Z, Ma Q, Zhang Z. **The relationship between erythrocytes and diabetes mellitus**. *J Diabetes Res* (2021) **2021** 6656062. DOI: 10.1155/2021/6656062
37. Tirosh A, Rudich A, Shochat T, Tekes-Manova D, Israeli E, Henkin Y. **Changes in triglyceride levels and risk for coronary heart disease in young men**. *Ann Intern Med* (2007) **147**. DOI: 10.7326/0003-4819-147-6-200709180-00007
38. Keaney JF, Curfman GD, Jarcho JA. **A pragmatic view of the new cholesterol treatment guidelines**. *N Engl J Med* (2014) **370**. DOI: 10.1056/NEJMms1314569
39. Peters SA, Singhateh Y, Mackay D, Huxley RR, Woodward M. **Total cholesterol as a risk factor for coronary heart disease and stroke in women compared with men: A systematic review and meta-analysis**. *Atherosclerosis.* (2016) **248**. DOI: 10.1016/j.atherosclerosis.2016.03.016
40. Kannel WB. **Range of serum cholesterol values in the population developing coronary artery disease**. *Am J Cardiol* (1995) **76** 69c-77c. DOI: 10.1016/S0002-9149(99)80474-3
41. Kwiterovich PO. **The antiatherogenic role of high-density lipoprotein cholesterol**. *Am J Cardiol* (1998) **82** 13q-21q. DOI: 10.1016/S0002-9149(98)00808-X
42. van Ryn J, Stangier J, Haertter S, Liesenfeld KH, Wienen W, Feuring M. **Dabigatran etexilate–a novel, reversible, oral direct thrombin inhibitor: Interpretation of coagulation assays and reversal of anticoagulant activity**. *Thromb Haemost* (2010) **103**. DOI: 10.1160/TH09-11-0758
43. Sotoudeh Anvari M, Tavakoli M, Lotfi-Tokaldany M, Broumand M, Rezahosseini O, Hakki-Kazzazi E. **Coronary artery disease presentation and its association with shortened activated partial thromboplastin time**. *J Tehran Heart Cent* (2018) **13** 1-5. PMID: 29997663
44. Korte W, Clarke S, Lefkowitz JB. **Short activated partial thromboplastin times are related to increased thrombin generation and an increased risk for thromboembolism**. *Am J Clin Pathol* (2000) **113**. DOI: 10.1309/G98J-ANA9-RMNC-XLYU
45. Brichacek AL, Brown CM. **Alkaline phosphatase: A potential biomarker for stroke and implications for treatment**. *Metab Brain Dis* (2019) **34** 3-19. DOI: 10.1007/s11011-018-0322-3
46. Zhang Y, Zhou C, Li J, Zhang Y, Xie D, Liang M. **Serum alkaline phosphatase levels and the risk of new-onset diabetes in hypertensive adults**. *Cardiovasc Diabetol* (2020) **19** 186. DOI: 10.1186/s12933-020-01161-x
47. Kabootari M, Raee MR, Akbarpour S, Asgari S, Azizi F, Hadaegh F. **Serum alkaline phosphatase and the risk of coronary heart disease, stroke and all-cause mortality: Tehran lipid and glucose study**. *BMJ Open* (2018) **8**. DOI: 10.1136/bmjopen-2018-023735
48. Liu J, Au Yeung SL, Lin SL, Leung GM, Schooling CM. **Liver enzymes and risk of ischemic heart disease and type 2 diabetes mellitus: A mendelian randomization study**. *Sci Rep* (2016) **6** 38813. DOI: 10.1038/srep38813
49. Malo MS. **A high level of intestinal alkaline phosphatase is protective against type 2 diabetes mellitus irrespective of obesity**. *EBioMedicine* (2015) **2**. DOI: 10.1016/j.ebiom.2015.11.027
50. Kaliannan K, Hamarneh SR, Economopoulos KP, Nasrin Alam S, Moaven O, Patel P. **Intestinal alkaline phosphatase prevents metabolic syndrome in mice**. *Proc Natl Acad Sci USA* (2013) **110**. DOI: 10.1073/pnas.1220180110
|
---
title: Trends in the Prevalence of Metabolically Healthy Obesity Among US Adults,
1999-2018
authors:
- Jiang-Shui Wang
- Peng-Fei Xia
- Meng-Nan Ma
- Yue Li
- Ting-Ting Geng
- Yan-Bo Zhang
- Zhou-Zheng Tu
- Limiao Jiang
- Li-Rong Zhou
- Bing-Fei Zhang
- Wen-Wei Tong
- Zhilei Shan
- Gang Liu
- Kun Yang
- An Pan
journal: JAMA Network Open
year: 2023
pmcid: PMC9999245
doi: 10.1001/jamanetworkopen.2023.2145
license: CC BY 4.0
---
# Trends in the Prevalence of Metabolically Healthy Obesity Among US Adults, 1999-2018
## Key Points
### Question
Has the prevalence of metabolically healthy obesity (MHO) changed among US adults in the past 20 years?
### Findings
In this survey study of 20 430 adults using data from the 1999-2018 National Health and Nutrition Examination Survey cycles, the age-standardized prevalence of MHO increased significantly from $3\%$ in 1999-2002 to $7\%$ in 2015-2018; the proportion of MHO among adults with obesity also increased significantly from $11\%$ to $15\%$. Disparities existed in trends across sociodemographic subgroups.
### Meaning
The results of this study suggest that the prevalence of MHO among US adults with obesity has increased significantly in the past 2 decades, with variations across sociodemographic subgroups.
## Abstract
This survey study uses data from the 1999-2018 National Health and Nutrition Examination Survey cycles to examine trends in the prevalence of metabolically healthy obesity among US adults.
### Importance
Improved understanding of trends in the proportion of individuals with metabolically healthy obesity (MHO) may facilitate stratification and management of obesity and inform policy efforts.
### Objectives
To characterize trends in the prevalence of MHO among US adults with obesity, overall and by sociodemographic subgroups.
### Design, Setting, and Participants
This survey study included 20 430 adult participants from 10 National Health and Nutrition Examination Survey (NHANES) cycles between 1999-2000 and 2017-2018. The NHANES is a series of cross-sectional and nationally representative surveys of the US population conducted continuously in 2-year cycles. Data were analyzed from November 2021 to August 2022.
### Exposures
National Health and Nutrition Examination Survey cycles from 1999-2000 to 2017-2018.
### Main Outcomes and Measures
Metabolically healthy obesity was defined as a body mass index of 30.0 (calculated as weight in kilograms divided by height in meters squared) without any metabolic disorders in blood pressure, fasting plasma glucose (FPG), high-density lipoprotein cholesterol (HDL-C), or triglycerides based on established cutoffs. Trends in the age-standardized prevalence of MHO were estimated using logistic regression analysis.
### Results
This study included 20 430 participants. Their weighted mean (SE) age was 47.1 (0.2) years; $50.8\%$ were women, and $68.8\%$ self-reported their race and ethnicity as non-Hispanic White. The age-standardized prevalence ($95\%$ CI) of MHO increased from $3.2\%$ ($2.6\%$-$3.8\%$) in the 1999-2002 cycles to $6.6\%$ ($5.3\%$-$7.9\%$) in the 2015-2018 cycles ($P \leq .001$ for trend). There were 7386 adults with obesity. Their weighted mean (SE) age was 48.0 (0.3) years, and $53.5\%$ were women. The age-standardized proportion ($95\%$ CI) of MHO among these 7386 adults increased from $10.6\%$ ($8.8\%$-$12.5\%$) in the 1999-2002 cycles to $15.0\%$ ($12.4\%$-$17.6\%$) in the 2015-2018 cycles ($$P \leq .02$$ for trend). Substantial increases in the proportion of MHO were observed for adults aged 60 years or older, men, non-Hispanic White individuals, and those with higher income, private insurance, or class I obesity. In addition, there were significant decreases in the age-standardized prevalence ($95\%$ CI) of elevated triglycerides (from $44.9\%$ [$40.9\%$-$48.9\%$] to $29.0\%$ [$25.7\%$-$32.4\%$]; $P \leq .001$ for trend) and reduced HDL-C (from $51.1\%$ [$47.6\%$-$54.6\%$] to $39.6\%$ [$36.3\%$-$43.0\%$]; $$P \leq .006$$ for trend). There was also a significant increase in elevated FPG (from $49.7\%$ [$95\%$ CI, $46.3\%$-$53.0\%$] to $58.0\%$ [$54.8\%$-$61.3\%$]; $P \leq .001$ for trend) but no significant change in elevated blood pressure (from $57.3\%$ [$53.9\%$-$60.7\%$] to $54.0\%$ [$50.9\%$-$57.1\%$]; $$P \leq .28$$ for trend).
### Conclusions and Relevance
The findings of this cross-sectional study suggest that the age-standardized proportion of MHO increased among US adults from 1999 to 2018, but differences in trends existed across sociodemographic subgroups. Effective strategies are needed to improve metabolic health status and prevent obesity-related complications in adults with obesity.
## Introduction
The prevalence of obesity has increased substantially in the past 2 decades, reaching an epidemic level in the US.1 *Obesity is* associated with most cardiovascular risk factors, including metabolic syndrome (MetS), hypertension, type 2 diabetes, and dyslipidemia.2 However, large interindividual heterogeneity in the development of obesity-related complications has been suggested.3 Despite increased body fat, a subset of people with obesity do not have obesity-related cardiometabolic abnormalities; this is referred to as metabolically healthy obesity (MHO).4,5,6,7,8,9 Individuals with MHO have favorable metabolic profiles and thus relatively lower risk for adverse cardiovascular consequences of obesity compared with individuals with metabolically unhealthy obesity (MUO).4,10 Evidence suggests that weight management strategies are more effective among individuals with MUO compared with those with MHO,11,12 indicating the potential value of the concept of obesity phenotypes.
Previous studies have reported on the proportion of US adults with MHO; however, the estimated prevalence of MHO varies widely across studies, partly due to large discrepancies in definitions.4,5,13,14,15,16 Most studies have used body mass index (BMI) to define obesity status and MetS components to reflect metabolically healthy status, but the cutoff values and number of parameters vary considerably. In recent years, researchers have proposed a strict definition of MHO as the absence of all MetS components in individuals with obesity, based on the rationale that patients with known cardiometabolic risk factors cannot be regarded as healthy.17,18 Evidence from a meta-analysis10 and prospective studies19,20,21 supports the comparable cardiovascular risk of MHO under this definition to that of metabolically healthy individuals with normal weight. Furthermore, insulin resistance and low-grade chronic inflammation, which provide additional information on metabolic health, have also been suggested as potential markers to assess MHO status.9,10,22 In the context of the obesity epidemic, better understanding of trends in MHO may facilitate the stratification and treatment of patients with obesity and inform policy efforts. However, whether the proportion of MHO, defined by conventional risk factors and other surrogate markers, has changed over the past 2 decades is largely unknown for US adults.
In this study, we aimed to characterize trends in the prevalence of MHO among US adults with obesity from 1999 to 2018, overall and in key sociodemographic subgroups. Our secondary objective was to compare trends in MHO under several commonly used criteria.
## Study Population
The National Health and Nutrition Examination Survey (NHANES) is a serial, cross-sectional, national survey with a complex, stratified, multistage probability design to monitor the health status of the civilian US population. The NHANES has been conducted continuously in 2-year cycles since 1999. Details of the NHANES are described elsewhere.23 The NHANES was approved by the research ethics review board of the US Centers for Disease Control and Prevention (CDC) National Center for Health Statistics, and written informed consent was obtained from all adult participants.23 The Institutional Review Board of Tongji Medical College determined that this study was exempt from review given the use of deidentified data. This study followed the American Association for Public Opinion Research (AAPOR) reporting guideline.
We used data from 10 NHANES cycles between 1999-2000 and 2017-2018. The response rate decreased from $76\%$ in 1999-2000 to $49\%$ in 2017-2018. We included nonpregnant adults aged 20 years or older in the fasting subsample, whose blood samples were obtained after an overnight fast of at least 8 hours (eTable 1 in Supplement 1). The fasting subsample was included because fasting glucose level is a key component of the MHO definition. Individuals who did not fulfill the fasting criteria or had missing values for BMI or metabolic parameters of interest were excluded.
## Data Collection
Information on participant age, sex, race and ethnicity, education, income, insurance status, medical history, and medication use was collected through household questionnaires. Race and ethnicity was not consistently reported in the NHANES (eg, Hispanic participants were not oversampled before 2007 and non-Hispanic Asian participants were not classified until 2011).24 For consistency over time, we categorized participants as self-reported Mexican American, non-Hispanic Black, non-Hispanic White, or other race and ethnicity (eg, non-Hispanic Asian or multiple). The family income-to-poverty ratio reflected annual family income relative to the federal poverty threshold and was used as a measure of income classified into 3 groups (≤$100\%$, $101\%$-$399\%$, and ≥$400\%$).25 Weight, height, waist circumference, and blood pressure (BP) were measured at mobile examination centers by trained staff according to standardized procedures.23 Body mass index was calculated as weight in kilograms divided by height in meters squared. Three BP measurements were assessed, and systolic BP and diastolic BP were calculated as the mean of all available measurements.
Participants were asked to provide blood samples at the mobile examination centers. The samples were stored at −20 °C and sent to central laboratories to determine lipid, plasma glucose, serum insulin, and C-reactive protein levels following standard protocols.23 A subset of participants were randomly selected to attend the morning session after an overnight fast; triglycerides, fasting plasma glucose (FPG), and insulin were measured for those who fasted at least 8 hours. Insulin resistance was assessed with the homeostasis model assessment score.26 Although there were changes in the laboratories, methods, and instruments used to measure lipid levels,27 all laboratories participated in the CDC Lipids Standardization Program,28 thus ensuring the accuracy, precision, and comparability of lipid measurements across cycles. To account for changes in laboratory methods over time, we calibrated FPG and serum insulin measurements to early cycles using the recommended backward equations.23
## MHO and MUO Criteria
Obesity and abdominal obesity were defined as a BMI of 30.0 or more and a waist circumference of 102 cm or more for men and 88 cm or more for women. The ethnicity-specific BMI cutoff for non-Hispanic Asian individuals was not used due to the lack of classification of this subgroup in the NHANES before 2011.24 Metabolic health was defined according to the harmonized definition proposed by Lavie et al17 and Ortega et al.18 Adults with obesity were classified as having MHO if they had 0 of 4 MetS components29,30: [1] elevated BP (systolic BP ≥130 mm Hg, diastolic BP ≥85 mm Hg, or antihypertensive medication use); [2] elevated FPG (≥100 mg/dL [to convert to millimoles per liter, multiply by 0.0555] or antidiabetic medication use); [3] reduced high-density lipoprotein cholesterol (HDL-C) (<40 mg/dL for men and <50 mg/dL for women [to convert to millimoles per liter, multiply by 0.0259]); or [4] elevated triglycerides (≥150 mg/dL [to convert to millimoles per liter, multiply by 0.0113]). Waist circumference was excluded for collinearity with BMI. Since data for cholesterol medication were available only for general use but not for treatment of elevated triglycerides or reduced HDL-C specifically, we did not utilize this information to avoid overestimation of these components, consistent with previous reports on MetS.31 Participants with obesity who met any of the above criteria were classified as having MUO.
## Statistical Analysis
We first evaluated trends in the prevalence of obesity, MUO, and MHO among all study participants from 1999 to 2018. Prevalence estimates were age standardized to the 2000 US Census population, using 3 age groups (20-39, 40-59, and ≥60 years) by the direct method. To calculate the number of individuals with obesity, MUO, or MHO, we next multiplied age-standardized prevalence estimates by the total noninstitutionalized adult population for each NHANES cycle.32 Trends in MHO proportion and individual metabolic indicators among those with obesity were then evaluated overall and by age group, sex, race and ethnicity, education level, income-to-poverty ratio, home ownership, and health insurance type. Proportion estimates were age standardized to all nonpregnant adults with obesity in the 2015-2018 NHANES cycles, using the same 3 age groups. To improve the reliability and precision of weighted estimates, 2 adjacent cycles were combined in consideration of the low prevalence of MHO. Linear trends over time were evaluated using logistic regression after regressing MHO on survey cycles (modeled as a continuous independent variable). Factors associated with metabolic health among adults with obesity were further identified with logistic regression models, adjusting for age group, sex, and race and ethnicity.
The complex survey design factors for the NHANES, including sample weights, clustering, and stratification, were accounted for as specified in the NHANES statistical analysis guideline.24 We used morning fasting subsample weights in all analyses to produce estimates representative of the US population. Standard errors were estimated with Taylor series linearization. Complete case analysis was applied if the missing data level for analyses was $10\%$ or less. Several sensitivity analyses were conducted to evaluate the impact of different criteria on MHO trends. First, information on self-reported cholesterol medication use was also used to define MUO and MHO. Second, individuals with a previous diagnosis of cardiovascular disease (CVD) were regarded as having MUO, regardless of their metabolic status.33 Third, abdominal obesity was used as a surrogate of general obesity in the definitions of MHO and MUO. Finally, other definitions commonly used by previous studies based on MetS components,29,30 insulin resistance,4 or together with inflammation5,6 were used to define metabolic health (eTable 2 in Supplement 1).
All analyses were performed with SAS, version 9.4 (SAS Institute Inc). Two-sided $P \leq .05$ was considered statistically significant. Adjustment for multiple comparisons was not performed as in previous reports,1,34 and the results should be interpreted as exploratory due to the potential for type I error. Statistical analyses were conducted from November 2021 to August 2022.
## Results
This survey study included 20 430 NHANES participants with a weighted mean (SE) age of 47.1 (0.2) years; $50.8\%$ were women and $49.2\%$ were men. In terms of race and ethnicity, $8.2\%$ participants self-identified as Mexican American, $10.8\%$ as non-Hispanic Black, $68.8\%$ as non-Hispanic White, and $12.3\%$ as other race or ethnicity (eTable 3 in Supplement 1). Data on education, income-to-poverty ratio, home ownership, and health insurance were missing for $0.1\%$, $7.3\%$, $1.0\%$, and $0.6\%$ of participants. Analyses of trends in MHO proportion and individual metabolic indicators were restricted to 7386 adults with obesity. Their weighted mean (SE) age was 48.0 (0.3) years; $53.5\%$ were women and $46.5\%$ were men. From the 1999-2002 to 2015-2018 cycles, the proportions of participants with some college education or more, government insurance, or higher-class obesity increased (Table 1).
**Table 1.**
| Characteristic | Percentage of adults by year (95% CI) | Percentage of adults by year (95% CI).1 | Percentage of adults by year (95% CI).2 | Percentage of adults by year (95% CI).3 | Percentage of adults by year (95% CI).4 |
| --- | --- | --- | --- | --- | --- |
| Characteristic | 1999-2002 (n = 1073) | 2003-2006 (n = 1198) | 2007-2010 (n = 1725) | 2011-2014 (n = 1625) | 2015-2018 (n = 1765) |
| Age, mean (SE) [95% CI], y | 46.9 (0.8) [45.3-48.6] | 47.2 (0.5) [46.2-48.3] | 48.2 (0.5) [47.3-49.2] | 48.6 (0.6) [47.3-49.9] | 48.6 (0.7) [47.2-49.9] |
| Age group, y | | | | | |
| 20-39 | 34.0 (29.0-39.1) | 33.4 (30.5-36.4) | 33.0 (30.7-35.4) | 32.1 (28.2-35.9) | 33.6 (30.4-36.8) |
| 40-59 | 43.4 (39.1-47.7) | 44.0 (40.7-47.4) | 40.9 (38.6-43.1) | 40.5 (37.4-43.7) | 37.2 (33.9-40.6) |
| ≥60 | 22.6 (18.8-26.3) | 22.5 (19.3-25.8) | 26.1 (24.0-28.3) | 27.4 (24.6-30.2) | 29.1 (25.1-33.1) |
| Sex | | | | | |
| Men | 45.2 (41.7-48.6) | 48.1 (44.7-51.4) | 47.1 (44.3-49.8) | 45.1 (42.4-47.8) | 46.9 (42.8-51.0) |
| Women | 54.8 (51.4-58.3) | 51.9 (48.6-55.3) | 52.9 (50.2-55.7) | 54.9 (52.2-57.6) | 53.1 (49.0-57.2) |
| Race and ethnicitya | | | | | |
| Mexican American | 7.5 (5.2-9.9) | 8.0 (4.9-11.1) | 9.2 (5.6-12.9) | 11.0 (7.5-14.5) | 10.8 (7.0-14.5) |
| Non-Hispanic Black | 13.4 (9.7-17.1) | 15.1 (12.0-18.2) | 14.7 (11.1-18.3) | 14.6 (10.6-18.5) | 13.2 (9.4-17.1) |
| Non-Hispanic White | 70.6 (66.0-75.2) | 70.1 (65.0-75.1) | 66.1 (59.4-72.8) | 64.9 (58.6-71.2) | 62.5 (56.6-68.4) |
| Other | 8.4 (4.1-12.8) | 6.8 (4.3-9.3) | 10.0 (7.1-12.9) | 9.5 (6.8-12.1) | 13.5 (11.1-15.8) |
| Education levelb | | | | | |
| Less than high school | 21.1 (18.5-23.8) | 18.7 (15.4-21.9) | 22.0 (19.3-24.7) | 18.4 (15.5-21.3) | 13.7 (11.2-16.1) |
| High school or equivalent | 28.7 (24.7-32.8) | 28.5 (25.7-31.4) | 24.4 (21.7-27.0) | 22.2 (19.5-24.8) | 26.6 (23.5-29.7) |
| Some college or more | 50.1 (45.3-54.9) | 52.8 (48.9-56.7) | 53.7 (50.0-57.3) | 59.4 (55.9-62.8) | 59.7 (56.2-63.3) |
| Income-to-poverty ratio, %c | | | | | |
| ≤100 | 13.2 (10.0-16.4) | 10.8 (8.4-13.1) | 13.9 (11.5-16.3) | 17.8 (14.0-21.6) | 14.1 (11.1-17.0) |
| 101-399 | 54.8 (49.8-59.8) | 56.0 (50.6-61.4) | 53.8 (50.5-57.0) | 53.7 (49.0-58.5) | 51.5 (46.1-56.9) |
| ≥400 | 32.0 (26.9-37.1) | 33.2 (28.6-37.9) | 32.4 (28.4-36.3) | 28.5 (24.0-33.0) | 34.5 (28.1-40.8) |
| Home ownershipd | | | | | |
| Owned home | 71.1 (65.8-76.4) | 71.8 (67.3-76.4) | 71.8 (68.1-75.6) | 63.5 (59.5-67.6) | 65.4 (60.4-70.5) |
| Rented home or other arrangement | 28.9 (23.6-34.2) | 28.2 (23.6-32.7) | 28.2 (24.4-31.9) | 36.5 (32.4-40.5) | 34.6 (29.5-39.6) |
| Health insurance typee | | | | | |
| Private | 71.3 (67.5-75.1) | 68.1 (63.5-72.7) | 66.2 (62.5-69.9) | 58.6 (54.9-62.2) | 62.7 (58.5-67.0) |
| Government | 13.9 (10.8-17.0) | 15.1 (12.3-17.8) | 15.9 (13.4-18.4) | 22.0 (18.7-25.4) | 25.3 (21.7-28.8) |
| | 14.8 (11.8-17.8) | 16.8 (13.8-19.8) | 17.9 (15.0-20.8) | 19.4 (16.8-22.0) | 12.0 (9.2-14.8) |
| Weight group by obesity class (BMI range) | | | | | |
| Class I (30.0-34.9) | 59.1 (54.2-63.9) | 56.3 (53.3-59.4) | 57.8 (55.7-59.9) | 56.2 (52.7-59.8) | 53.7 (50.4-56.9) |
| Class II (35.0-39.9) | 26.8 (23.6-30.1) | 25.6 (22.8-28.5) | 26.0 (22.7-29.2) | 23.8 (21.3-26.2) | 26.1 (23.6-28.6) |
| Class III (≥40.0) | 14.1 (10.6-17.6) | 18.0 (15.2-20.9) | 16.2 (14.0-18.4) | 20.0 (17.1-22.9) | 20.2 (16.9-23.5) |
| Abdominal obesityf | | | | | |
| Yes | 96.3 (94.6-98.0) | 97.4 (96.4-98.4) | 96.2 (94.9-97.5) | 96.2 (94.8-97.6) | 97.1 (96.2-98.0) |
| No | 3.7 (2.0-5.4) | 2.6 (1.6-3.6) | 3.8 (2.5-5.1) | 3.8 (2.4-5.2) | 2.9 (2.0-3.8) |
## Trends in MHO Prevalence Among the Population With Obesity
For the whole study population, the age-standardized prevalence ($95\%$ CI) of obesity increased significantly from $28.6\%$ ($26.3\%$-$30.9\%$) in the 1999-2002 cycles to $40.9\%$ ($37.9\%$-$43.8\%$) in the 2015-2018 cycles ($P \leq .001$ for trend). The age-standardized prevalence ($95\%$ CI) of MUO also increased from $25.4\%$ ($23.3\%$-$27.6\%$) in 1999-2002 to $34.3\%$ ($31.6\%$-$36.9\%$) in 2015-2018 ($P \leq .001$ for trend). Finally, the prevalence ($95\%$ CI) of MHO increased from $3.2\%$ ($2.6\%$-$3.8\%$) in 1999-2002 to $6.6\%$ ($5.3\%$-$7.9\%$) in 2015-2018 ($P \leq .001$ for trend; Figure 1A).
**Figure 1.:** *Trends in the Prevalence of Obesity, Metabolically Unhealthy Obesity (MUO), and Metabolically Healthy Obesity (MHO) Among US Adults, 1999-2018A, Trends in the prevalence of obesity, MUO, and MHO among US adults. From 1999-2002 to 2015-2018, P < .001 for trend in obesity, MUO, and MHO prevalence. From 2003-2006 to 2015-2018, P < .001 for trend in obesity and MUO prevalence and P = .02 for trend in MHO prevalence. B, Trends in the proportion of MHO among US adults with obesity. From 1999-2002 to 2015-2018, P = .02 for trend. From 2003-2006 to 2015-2018, P = .51 for trend. Obesity was defined as a body mass index of 30.0 or greater (calculated as weight in kilograms divided by height in meters squared). Among participants with obesity, MUO was defined as having any component of the metabolic syndrome (waist circumference excluded) and MHO was defined as meeting none of the metabolic syndrome criteria. In A, prevalence estimates were age standardized to the 2000 US Census population, using 3 age groups (20-39, 40-59, and ≥60 years). In B, proportion estimates were age standardized to the nonpregnant adult population with obesity in the 2015-2018 National Health and Nutrition Examination Survey cycles, using the same 3 age groups. All estimates were weighted, and error bars indicate 95% CIs. Linear trends over time were evaluated using logistic regression. Specific estimates are shown in Table 2 and eTable 4 in Supplement 1.*
Within racial and ethnic subgroups, more than $40\%$ of Mexican American adults and non-Hispanic Black adults in the 2015-2018 cycles had MUO; however, the prevalence of MHO was low among all racial and ethnic subpopulations (eTable 4 in Supplement 1). In the 2015-2018 cycles, an estimated 81.1 million US adults ($95\%$ CI, 74.7-87.4) had MUO and 15.6 million (12.5-18.6) had MHO (eFigure 1 in Supplement 1).
Among the 7386 participants with obesity, the age-standardized proportion ($95\%$ CI) of MHO increased significantly from $10.6\%$ ($8.8\%$-$12.5\%$) in the 1999-2002 cycles to $15.0\%$ ($12.4\%$-$17.6\%$) in the 2015-2018 cycles ($$P \leq .02$$ for trend; Figure 1B). A substantial increase was observed among individuals aged 60 years or older, men, and non-Hispanic White adults as well as those with higher income, private insurance, or class I obesity (all $P \leq .05$ for trend; Table 2). However, this increase was largely attributable to an increase between the 1999-2002 and 2003-2006 cycles. When trends from the 2003-2006 to 2015-2018 cycles were evaluated, there was no significant increase in the age-standardized proportion of MHO (Table 2).
**Table 2.**
| Characteristic | Adults with MHO, % (95% CI)b | Adults with MHO, % (95% CI)b.1 | Adults with MHO, % (95% CI)b.2 | Adults with MHO, % (95% CI)b.3 | Adults with MHO, % (95% CI)b.4 | P value for trendc | P value for trendc.1 |
| --- | --- | --- | --- | --- | --- | --- | --- |
| Characteristic | 1999-2002 (n = 1073) | 2003-2006 (n = 1198) | 2007-2010 (n = 1725) | 2011-2014 (n = 1625) | 2015-2018 (n = 1765) | 1999-2002 | 2003-2006 |
| Overall % | 10.6 (8.8-12.5) | 14.0 (11.7-16.3) | 14.1 (11.6-16.6) | 14.7 (12.9-16.4) | 15.0 (12.4-17.6) | .02 | .51 |
| Age, y | | | | | | | |
| 20-39 | 18.2 (12.9-23.4) | 26.8 (20.8-32.7) | 25.1 (18.9-31.4) | 24.9 (20.2-29.7) | 27.2 (21.7-32.8) | .09 | .86 |
| 40-59 | 10.4 (6.7-14.1) | 10.5 (6.9-14.1) | 12.6 (9.4-15.8) | 12.2 (8.9-15.5) | 11.4 (7.5-15.3) | .56 | .79 |
| ≥60 | 2.4 (0.5-4.4) | 4.3 (2.0-6.6) | 3.6 (1.5-5.7) | 6.4 (3.8-8.9) | 5.9 (3.0-8.9) | .03 | .18 |
| Sex | | | | | | | |
| Men | 7.9 (5.2-10.6) | 11.9 (8.0-15.8) | 13.2 (9.7-16.7) | 12.7 (9.9-15.4) | 13.9 (10.2-17.5) | .04 | .52 |
| Women | 12.9 (10.1-15.8) | 16.1 (13.0-19.2) | 14.8 (11.8-17.9) | 16.2 (14.0-18.5) | 16.0 (12.5-19.5) | .23 | .82 |
| Race and ethnicityd | | | | | | | |
| Mexican American | 10.3 (7.2-13.5) | 13.5 (8.6-18.5) | 13.0 (9.0-17.1) | 12.2 (8.6-15.8) | 12.8 (9.2-16.5) | .85 | .89 |
| Non-Hispanic Black | 14.7 (10.0-19.4) | 19.1 (14.0-24.2) | 16.7 (12.8-20.6) | 16.1 (13.0-19.3) | 15.5 (12.9-18.0) | .54 | .13 |
| Non-Hispanic White | 7.5 (5.4-9.6) | 12.5 (9.2-15.8) | 13.6 (10.1-17.2) | 14.5 (12.0-16.9) | 15.7 (11.5-20.0) | .002 | .20 |
| Other | 21.8 (12.3-31.4) | 14.2 (6.5-21.9) | 12.0 (7.0-17.0) | 14.3 (7.6-21.0) | 13.4 (10.1-16.8) | .20 | .93 |
| Education levele | | | | | | | |
| Less than high school | 13.1 (8.4-17.9) | 10.4 (5.9-14.9) | 8.8 (5.3-12.3) | 10.4 (6.4-14.4) | 12.3 (6.3-18.2) | .79 | .59 |
| High school or equivalent | 8.6 (4.3-12.8) | 9.5 (5.5-13.6) | 14.9 (9.9-19.9) | 13.8 (10.0-17.6) | 11.5 (7.4-15.6) | .15 | .72 |
| Some college or more | 10.9 (7.8-14.0) | 17.5 (14.5-20.6) | 15.7 (12.3-19.0) | 16.3 (13.9-18.8) | 17.0 (13.0-21.1) | .12 | .97 |
| Income-to-poverty ratio, %f | | | | | | | |
| ≤100 | 12.7 (7.6-17.8) | 13.9 (6.7-21.0) | 9.1 (3.4-14.7) | 12.7 (8.7-16.6) | 12.2 (8.9-15.5) | .94 | .92 |
| 101-399 | 10.5 (7.5-13.6) | 13.5 (10.9-16.2) | 13.3 (9.8-16.9) | 14.8 (11.7-18.0) | 13.7 (10.3-17.1) | .22 | .92 |
| ≥400 | 9.2 (4.6-13.8) | 15.0 (10.8-19.2) | 16.9 (11.7-22.1) | 18.0 (13.1-22.9) | 18.8 (12.8-24.8) | .03 | .38 |
| Home ownershipg | | | | | | | |
| Owned home | 11.3 (8.4-14.1) | 14.8 (12.1-17.6) | 13.4 (10.4-16.5) | 15.7 (13.1-18.4) | 15.8 (11.8-19.9) | .07 | .43 |
| Rented home or other arrangement | 9.5 (5.5-13.5) | 11.9 (8.4-15.4) | 14.4 (10.0-18.7) | 13.3 (10.0-16.6) | 13.6 (10.5-16.7) | .13 | .87 |
| Health insurance typeh | | | | | | | |
| Private | 10.0 (7.9-12.1) | 14.8 (11.7-18.0) | 15.8 (12.9-18.7) | 15.4 (12.2-18.6) | 16.5 (12.8-20.3) | .01 | .54 |
| Government | 9.5 (2.0-16.9) | 9.3 (4.4-14.2) | 6.3 (2.8-9.9) | 13.0 (7.9-18.1) | 13.3 (9.5-17.0) | .05 | .03 |
| | 15.9 (11.1-20.7) | 13.4 (9.0-17.8) | 13.5 (8.7-18.2) | 17.4 (10.4-24.3) | 14.5 (9.1-19.9) | .63 | .44 |
| Weight group by obesity class (BMI range) | | | | | | | |
| Class I (30.0-34.9) | 12.1 (9.4-14.9) | 17.5 (14.1-21.0) | 16.8 (13.7-19.9) | 18.8 (15.9-21.7) | 18.6 (14.5-22.8) | .02 | .51 |
| Class II (35.0-39.9) | 9.0 (5.1-12.8) | 14.4 (9.4-19.3) | 12.5 (9.3-15.6) | 12.3 (8.1-16.5) | 13.0 (7.6-18.4) | .48 | .85 |
| Class III (≥40.0) | 7.5 (3.2-11.8) | 3.8 (1.7-5.8) | 7.8 (4.0-11.6) | 6.4 (3.4-9.4) | 8.2 (4.3-12.2) | .34 | .10 |
## Trends in Individual Metabolic Indicators Among the Population With Obesity
During the past 2 decades, there was a substantial divergence in trends for clinical metabolic indicators among individuals with obesity. From the 1999-2002 to 2015-2018 cycles, significantly decreasing trends in the age-standardized percentage ($95\%$ CI) of elevated triglycerides (from $44.9\%$ [$40.9\%$-$48.9\%$] to $29.0\%$ [$25.7\%$-$32.4\%$]; $P \leq .001$ for trend) and reduced HDL-C (from $51.1\%$ [$47.6\%$-$54.6\%$] to $39.6\%$ [$36.3\%$-$43.0\%$]; $$P \leq .006$$ for trend) were observed. However, no significant trend in the percentage of elevated BP (from $57.3\%$ [$95\%$ CI, $53.9\%$-$60.7\%$] to $54.0\%$ [$50.9\%$-$57.1\%$]; $$P \leq .28$$ for trend) was observed, whereas the percentage of elevated FPG increased significantly (from $49.7\%$ [$46.3\%$-$53.0\%$] to $58.0\%$ [$54.8\%$-$61.3\%$]; $P \leq .001$ for trend; Figure 2).
**Figure 2.:** *Trends in the Percentage of Individual Clinical Metabolic Parameters Among Adults With Obesity, 1999-2018A, Elevated blood pressure (BP; systolic BP ≥130 mm Hg, diastolic BP ≥85 mm Hg, or hypertension medication use). No significant trend was observed from 1999-2002 to 2015-2018 (P = .28 for trend) or from 2003-2006 to 2015-2018 (P = .92 for trend). B, Elevated fasting plasma glucose (FPG; ≥100 mg/dL or antidiabetic medication use). A significant increasing trend was observed from 1999-2002 to 2015-2018 (P < .001 for trend) and from 2003-2006 to 2015-2018 (P = .02 for trend). C, Reduced high-density lipoprotein cholesterol (HDL-C; <40 mg/dL for men and <50 mg/dL for women). A significant decreasing trend was observed from 1999-2002 to 2015-2018 (P = .006 for trend) but not from 2003-2006 to 2015-2018 (P = .47 for trend). D, Elevated triglycerides (TG; ≥150 mg/dL). A significant decreasing trend was observed from 1999-2002 to 2015-2018 and from 2003-2006 to 2015-2018 (both P < .001 for trend). Percentage estimates were age standardized to the nonpregnant adult population with obesity in the 2015-2018 National Health and Nutrition Examination Survey cycles, using 3 age groups (20-39, 40-59, and ≥60 years). All estimates were weighted and the error bars indicate 95% CIs. Linear trends over time were evaluated using logistic regression.*
## Factors Associated With Metabolic Health Among the Population With Obesity
Among all US participants with obesity in the 1999-2018 NHANES cycles, younger adults, women, non-Hispanic Black individuals, and those with some college education or more, higher income, home ownership, or lower obesity class were generally more likely to be metabolically healthy (Table 3). Women with obesity were more likely to have reduced HDL-C but less likely to have elevated BP, FPG, and triglycerides compared with men with obesity. Non-Hispanic Black individuals with obesity were more likely to have elevated BP but less likely to have elevated triglycerides and reduced HDL-C compared with non-Hispanic White adults with obesity.
**Table 3.**
| Characteristic | Adjusted odds ratio (95% CI)a | Adjusted odds ratio (95% CI)a.1 | Adjusted odds ratio (95% CI)a.2 | Adjusted odds ratio (95% CI)a.3 | Adjusted odds ratio (95% CI)a.4 |
| --- | --- | --- | --- | --- | --- |
| Characteristic | MHOb | BP not elevatedc | FPG not elevatedd | HDL-C not reducede | Triglycerides not elevatedf |
| Age, y | | | | | |
| 20-39 | 6.44 (4.86-8.52) | 11.37 (9.52-13.57) | 6.89 (5.73-8.28) | 0.48 (0.42-0.55) | 1.42 (1.19-1.71) |
| 40-59 | 2.54 (1.86-3.45) | 2.94 (2.48-3.49) | 2.25 (1.97-2.57) | 0.68 (0.58-0.79) | 0.97 (0.82-1.14) |
| ≥60 | 1 [Reference] | 1 [Reference] | 1 [Reference] | 1 [Reference] | 1 [Reference] |
| Sex | | | | | |
| Men | 1 [Reference] | 1 [Reference] | 1 [Reference] | 1 [Reference] | 1 [Reference] |
| Women | 1.28 (1.06-1.54) | 1.40 (1.22-1.61) | 1.66 (1.47-1.89) | 0.63 (0.54-0.73) | 1.43 (1.27-1.62) |
| Race and ethnicityg | | | | | |
| Mexican American | 0.92 (0.73-1.16) | 1.38 (1.18-1.62) | 0.74 (0.62-0.89) | 0.99 (0.84-1.15) | 0.93 (0.80-1.08) |
| Non-Hispanic Black | 1.32 (1.10-1.59) | 0.60 (0.52-0.69) | 1.14 (0.98-1.32) | 1.77 (1.52-2.07) | 3.28 (2.78-3.86) |
| Non-Hispanic White | 1 [Reference] | 1 [Reference] | 1 [Reference] | 1 [Reference] | 1 [Reference] |
| Other | 1.15 (0.88-1.50) | 1.17 (0.95-1.43) | 0.85 (0.67-1.07) | 1.16 (0.95-1.42) | 1.12 (0.91-1.38) |
| Education level | | | | | |
| Less than high school | 1 [Reference] | 1 [Reference] | 1 [Reference] | 1 [Reference] | 1 [Reference] |
| High school or equivalent | 1.09 (0.81-1.47) | 0.96 (0.78-1.19) | 1.11 (0.92-1.34) | 1.17 (0.97-1.40) | 1.10 (0.92-1.31) |
| Some college or more | 1.60 (1.21-2.10) | 1.15 (0.98-1.35) | 1.31 (1.11-1.54) | 1.43 (1.20-1.70) | 1.30 (1.11-1.52) |
| Income-to-poverty ratio, % | | | | | |
| ≤100 | 1 [Reference] | 1 [Reference] | 1 [Reference] | 1 [Reference] | 1 [Reference] |
| 101-399 | 1.21 (0.93-1.59) | 1.00 (0.82-1.21) | 1.19 (1.03-1.37) | 1.29 (1.09-1.52) | 1.16 (0.98-1.37) |
| ≥400 | 1.64 (1.20-2.25) | 1.06 (0.86-1.32) | 1.37 (1.14-1.66) | 1.64 (1.34-2.00) | 1.30 (1.06-1.60) |
| Home ownership | | | | | |
| Owned home | 1.22 (1.00-1.48) | 1.01 (0.85-1.18) | 1.08 (0.93-1.26) | 1.24 (1.04-1.47) | 1.08 (0.94-1.24) |
| Rented home or other arrangement | 1 [Reference] | 1 [Reference] | 1 [Reference] | 1 [Reference] | 1 [Reference] |
| Health insurance type | | | | | |
| Private | 1.08 (0.87-1.34) | 0.77 (0.64-0.92) | 1.06 (0.89-1.26) | 1.38 (1.18-1.61) | 1.14 (0.96-1.35) |
| Government | 0.72 (0.54-0.97) | 0.65 (0.52-0.81) | 0.76 (0.62-0.93) | 1.07 (0.90-1.27) | 0.98 (0.80-1.20) |
| | 1 [Reference] | 1 [Reference] | 1 [Reference] | 1 [Reference] | 1 [Reference] |
| Weight group by obesity class (BMI range) | | | | | |
| Class I (30.0-34.9) | 3.22 (2.39-4.33) | 2.34 (1.89-2.89) | 2.79 (2.30-3.40) | 1.71 (1.45-2.02) | 1.20 (1.00-1.42) |
| Class II (35.0-39.9) | 1.99 (1.47-2.70) | 1.78 (1.39-2.27) | 1.91 (1.54-2.38) | 1.20 (1.00-1.44) | 1.09 (0.90-1.32) |
| Class III (≥40.0) | 1 [Reference] | 1 [Reference] | 1 [Reference] | 1 [Reference] | 1 [Reference] |
## Sensitivity Analysis
When individuals who used cholesterol medication or had a previous CVD diagnosis were further classified as having MUO, the proportions of MHO among adults with obesity were slightly smaller because more individuals were classified into the metabolically unhealthy group (eTables 5 and 6 in Supplement 1). Trends in metabolically healthy abdominal obesity generally followed the same patterns as observed for MHO, albeit with more notable changes (eTable 7 in Supplement 1). Sample sizes for some sensitivity analyses under other MHO criteria were somewhat smaller due to missing values for certain variables. In the 2015-2018 NHANES cycles, the age-standardized prevalence ($95\%$ CI) of MHO in the total population varied from $3.5\%$ ($2.5\%$-$4.4\%$) to $18.1\%$ ($16.1\%$-$20.2\%$) when using other MHO definitions, and the proportion of MHO among the population with obesity varied from $8.0\%$ ($6.0\%$-$9.9\%$) to $42.4\%$ ($39.6\%$-$45.1\%$) (eFigure 2 and eTable 8 in Supplement 1). There were increasing trends in the prevalence of MHO under other criteria based on MetS components. However, decreasing trends were observed when insulin resistance was used to define metabolic health. Trends in age-standardized mean concentrations of all individual metabolic parameters among adults with obesity, MUO, and MHO are shown in eTable 9 in Supplement 1.
## Discussion
The results of this nationally representative survey study suggest that the age-standardized prevalence of obesity, MUO, and MHO increased significantly among US adults from 1999 to 2018. The proportion of MHO among adults with obesity and its trends varied across different criteria. When defined as the absence of all MetS components, the proportion of MHO increased significantly from $10.6\%$ in 1999-2002 to $15.0\%$ in 2015-2018. However, this increase was largely due to an increase between 1999-2002 and 2003-2006, and disparities existed among sociodemographic subgroups. Our results suggest that the overall increase in MHO was driven primarily by the decrease in dyslipidemia (ie, elevated triglycerides and reduced HDL-C) among adults with obesity; however, elevated BP remained stable and elevated FPG increased during the past 2 decades.
Different MHO criteria used in previous studies have led to large discrepancies in estimates of MHO prevalence, which precludes direct comparisons among studies. Previous reviews reported that the proportion of MHO among the population with obesity ranged between $6\%$ and $40\%$, depending on the criteria used.7,8,33 *From a* clinical and public health point of view, we used strict criteria based on BMI and MetS components to define MHO in our main analyses.17,18 Our estimates of MHO prevalence among US adults (range, $3.2\%$-$6.6\%$ across years) and MHO proportion among the population with obesity (range, $10.6\%$-$15.0\%$) were consistent with previous reports using the same criteria.14,15 One study based on 2009-2016 NHANES data reported a smaller proportion of MHO ($6.8\%$), mainly because the investigators used $\frac{120}{80}$ mm Hg as the cut point for elevated BP.13 Unsurprisingly, our estimates were lower than those in studies with looser MHO criteria4,5,6; however, research has shown that most studies have overestimated the prevalence of MHO.3,33 Large heterogeneity in MHO prevalence estimates using different definitions underscores the need to establish a standardized definition of this obesity phenotype.
We have reported, to our knowledge, the most recent and comprehensive national trend estimates of MHO. The observation that MHO proportions increased from 1999 to 2018 should be treated with caution, as trends between the 2003-2006 and 2015-2018 cycles were relatively stable. These results may be better interpreted when combined with trends in individual metabolic indicators. For example, the overall increase in MHO may be driven primarily by the decrease in dyslipidemia among the population with obesity, which has also been observed for the population overall.27,35 A plausible explanation may include increased awareness, diagnosis, and treatment of dyslipidemia as well as decreased smoking, removal of trans-fatty acids from foods, and improved diet quality.27,36,37 In contrast, the plateau in the proportion of MHO from 2003-2006 to 2015-2018 may result from a combination of leveling off of reduced HDL-C, no significant change in elevated BP, and the significant increase in elevated FPG over the same period. Previous studies examining trends in cardiovascular health metrics among US adults with obesity have reported the following: decreases for untreated hypertension and untreated dyslipidemia between 1999 and 201038; nonsignificant changes in elevated BP and improvements in mean HDL-C, but deteriorations in mean hemoglobin A1c between 1988 and 201437; and increases in the proportion of individuals without prior cardiovascular events or cardiometabolic diseases between 1999 and 2016.39 Although different time periods may contribute to variations in trend estimates, our results were generally consistent with these findings. Given the complex interplay between obesity and glucose control, greater attention should be paid to the increase in elevated FPG among adults with obesity.40 Beyond conventional risk factors, our study further complemented a recent study on trends in metabolic phenotypes defined by MetS components by incorporating insulin resistance and chronic inflammation to capture a wider breadth of metabolic abnormalities.16 *It is* noteworthy that the use of insulin resistance to define poor metabolic health mitigated or even reversed the overall increasing trends in MHO, which may be linked to an increase in sedentary time, waist circumference, and nonalcoholic fatty liver disease.41,42,43 Although reasons for these trends may be complex and warrant further investigation, these results highlight the importance of reinforcing glucose management and reducing insulin resistance among adults with obesity.
The overall increase in the proportion of MHO should also be treated in the context of existing disparities in subpopulations. Among racial and ethnic subgroups, we observed a significant increase in the proportion of MHO only in non-Hispanic White adults, which may be attributed in part to higher income, wider insurance coverage, more accessible health services, sociocultural differences, and other social determinants.44,45,46 Previous studies have suggested that higher-income groups tend to have improved diet quality,36 increased adherence to physical activity guidelines,43 and decreased smoking prevalence,25 which may contribute to favorable trends in the proportion of MHO. In contrast, adults with lower levels of education or lower income were more likely to be metabolically unhealthy; this is important to note given their already higher prevalence of obesity and lack of weight self-awareness.47,48 The disproportionate prevalence of and trends in metabolic alterations could aggravate obesity disparities, as these are all CVD risk factors; thus, these findings underscore the urgency for more accessible strategies to reach racial and ethnic minority individuals and those residing in low-income communities.
Although there is no consensus on the protective effect of MHO compared with metabolically healthy normal weight,10,49,50 accumulating evidence suggests that individuals with MHO have a better CVD prognosis than their MUO counterparts.12,17,33 Previous studies suggest that mechanisms including visceral and ectopic fat accumulation, adipose dysfunction, insulin resistance, inflammatory dysregulation, and gut microbiota may play a part.33,51 However, MHO has been considered a transitory state for most individuals with obesity, and those whose status converts to MUO would have higher risk.9,22 Therefore, detailed and repeated metabolic phenotyping among adults with obesity should be taken into consideration in clinical risk assessment to improve the inherent shortcomings of BMI assessment and to help those with MHO maintain their status.8 It should also be emphasized that although the proportion of MHO increased in this study, the absolute number of adults with MUO has increased dramatically in the past 2 decades, suggesting that MUO is still a major health concern. Effective strategies to address the double burden of obesity and metabolic disorders and to curb the increase in MUO are important.
## Limitations
This survey study has several limitations. First, there is no universally accepted definition of MHO; thus, we provided estimates under several commonly used criteria. Second, misclassification of MHO was possible because metabolic parameters such as glycemic levels and lipids were measured only once, particularly considering the transient nature of MHO.22 Third, we did not evaluate physical activity, cardiovascular fitness, and body fat distribution due to inconsistent or lacking assessments across survey cycles, which might be important in understanding the metabolic health status of individuals with obesity.9,49 Fourth, the response rate declined across surveys. Finally, although 2 adjacent NHANES cycles were combined, there was a possibility of insufficient power to detect variabilities over time, particularly in some subgroups with limited sample size.
## Conclusions
In this cross-sectional study of US adults, we observed a low prevalence of MHO and a large, increasing burden of MUO. Although the proportion of MHO among adults with obesity increased during the past 2 decades, disparities among sociodemographic subpopulations were observed. These results highlight the need for effective strategies to optimize metabolic status and prevent obesity-related complications among people with obesity, especially among vulnerable subpopulations. Priority should be placed on reinforcing glucose management and reducing insulin resistance among individuals with obesity.
## References
1. Hales CM, Carroll MD, Fryar CD, Ogden CL
2. Kotsis V, Jordan J, Micic D. **Obesity and cardiovascular risk: a call for action from the European Society of Hypertension Working Group of Obesity, Diabetes and the High-risk Patient and European Association for the Study of Obesity: part A: mechanisms of obesity induced hypertension, diabetes and dyslipidemia and practice guidelines for treatment**. *J Hypertens* (2018.0) **36** 1427-1440. DOI: 10.1097/HJH.0000000000001730
3. Neeland IJ, Poirier P, Després JP. **Cardiovascular and metabolic heterogeneity of obesity: clinical challenges and implications for management**. *Circulation* (2018.0) **137** 1391-1406. DOI: 10.1161/CIRCULATIONAHA.117.029617
4. Meigs JB, Wilson PWF, Fox CS. **Body mass index, metabolic syndrome, and risk of type 2 diabetes or cardiovascular disease**. *J Clin Endocrinol Metab* (2006.0) **91** 2906-2912. DOI: 10.1210/jc.2006-0594
5. Wildman RP, Muntner P, Reynolds K. **The obese without cardiometabolic risk factor clustering and the normal weight with cardiometabolic risk factor clustering: prevalence and correlates of 2 phenotypes among the US population (NHANES 1999-2004)**. *Arch Intern Med* (2008.0) **168** 1617-1624. DOI: 10.1001/archinte.168.15.1617
6. Karelis AD, Rabasa-Lhoret R. **Inclusion of C-reactive protein in the identification of metabolically healthy but obese (MHO) individuals**. *Diabetes Metab* (2008.0) **34** 183-184. DOI: 10.1016/j.diabet.2007.11.004
7. Primeau V, Coderre L, Karelis AD. **Characterizing the profile of obese patients who are metabolically healthy**. *Int J Obes (Lond)* (2011.0) **35** 971-981. DOI: 10.1038/ijo.2010.216
8. Phillips CM. **Metabolically healthy obesity: definitions, determinants and clinical implications**. *Rev Endocr Metab Disord* (2013.0) **14** 219-227. DOI: 10.1007/s11154-013-9252-x
9. Stefan N, Häring HU, Hu FB, Schulze MB. **Metabolically healthy obesity: epidemiology, mechanisms, and clinical implications**. *Lancet Diabetes Endocrinol* (2013.0) **1** 152-162. DOI: 10.1016/S2213-8587(13)70062-7
10. Eckel N, Meidtner K, Kalle-Uhlmann T, Stefan N, Schulze MB. **Metabolically healthy obesity and cardiovascular events: a systematic review and meta-analysis**. *Eur J Prev Cardiol* (2016.0) **23** 956-966. DOI: 10.1177/2047487315623884
11. Soll D, Gawron J, Pletsch-Borba L, Spranger J, Mai K. **Long-term impact of the metabolic status on weight loss-induced health benefits**. *Nutr Metab (Lond)* (2022.0) **19** 25. DOI: 10.1186/s12986-022-00660-w
12. Stefan N, Häring HU, Schulze MB. **Metabolically healthy obesity: the low-hanging fruit in obesity treatment?**. *Lancet Diabetes Endocrinol* (2018.0) **6** 249-258. DOI: 10.1016/S2213-8587(17)30292-9
13. Araújo J, Cai J, Stevens J. **Prevalence of optimal metabolic health in American adults: National Health and Nutrition Examination Survey 2009-2016**. *Metab Syndr Relat Disord* (2019.0) **17** 46-52. DOI: 10.1089/met.2018.0105
14. Smith GI, Mittendorfer B, Klein S. **Metabolically healthy obesity: facts and fantasies**. *J Clin Invest* (2019.0) **129** 3978-3989. DOI: 10.1172/JCI129186
15. Adair KE, Bowden RG, Funderburk LK, Forsse JS, Ylitalo KR. **Metabolic health, obesity, and renal function: 2013-2018 National Health and Nutrition Examination Surveys**. *Life (Basel)* (2021.0) **11** 888. DOI: 10.3390/life11090888
16. Liu J, Zhang Y, Lavie CJ, Moran AE. **Trends in metabolic phenotypes according to body mass index among US adults, 1999-2018**. *Mayo Clin Proc* (2022.0) **97** 1664-1679. DOI: 10.1016/j.mayocp.2022.02.013
17. Lavie CJ, Laddu D, Arena R, Ortega FB, Alpert MA, Kushner RF. **Healthy weight and obesity prevention:**. *J Am Coll Cardiol* (2018.0) **72** 1506-1531. DOI: 10.1016/j.jacc.2018.08.1037
18. Ortega FB, Lavie CJ, Blair SN. **Obesity and cardiovascular disease**. *Circ Res* (2016.0) **118** 1752-1770. DOI: 10.1161/CIRCRESAHA.115.306883
19. Lassale C, Tzoulaki I, Moons KGM. **Separate and combined associations of obesity and metabolic health with coronary heart disease: a pan-European case-cohort analysis**. *Eur Heart J* (2018.0) **39** 397-406. DOI: 10.1093/eurheartj/ehx448
20. Zembic A, Eckel N, Stefan N, Baudry J, Schulze MB. **An empirically derived definition of metabolically healthy obesity based on risk of cardiovascular and total mortality**. *JAMA Netw Open* (2021.0) **4**. DOI: 10.1001/jamanetworkopen.2021.8505
21. Al-Khalidi B, Kimball SM, Kuk JL, Ardern CI. **Metabolically healthy obesity, vitamin D, and all-cause and cardiometabolic mortality risk in NHANES III**. *Clin Nutr* (2019.0) **38** 820-828. DOI: 10.1016/j.clnu.2018.02.025
22. Kouvari M, Panagiotakos DB, Yannakoulia M. **Transition from metabolically benign to metabolically unhealthy obesity and 10-year cardiovascular disease incidence: the ATTICA cohort study**. *Metabolism* (2019.0) **93** 18-24. DOI: 10.1016/j.metabol.2019.01.003
23. 23National Center for Health Statistics. National Health and Nutrition Examination Survey. US Centers for Disease Control and Prevention. Accessed April 18, 2022. https://www.cdc.gov/nchs/nhanes/index.htm
24. 24National Center for Health Statistics. NHANES survey methods and analytic guidelines. US Centers for Disease Control and Prevention. Accessed April 1, 2022. https://wwwn.cdc.gov/nchs/nhanes/analyticguidelines.aspx
25. Odutayo A, Gill P, Shepherd S. **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
26. Matthews DR, Hosker JP, Rudenski AS, Naylor BA, Treacher DF, Turner RC. **Homeostasis model assessment: insulin resistance and beta-cell function from fasting plasma glucose and insulin concentrations in man**. *Diabetologia* (1985.0) **28** 412-419. DOI: 10.1007/BF00280883
27. Rosinger A, Carroll MD, Lacher D, Ogden C. **Trends in total cholesterol, triglycerides, and low-density lipoprotein in US adults, 1999-2014**. *JAMA Cardiol* (2017.0) **2** 339-341. DOI: 10.1001/jamacardio.2016.4396
28. 28US Centers for Disease Control and Prevention. LSP: Lipids Standardization Program. Accessed February 1, 2022. https://www.cdc.gov/labstandards/lsp.html
29. Alberti KG, Eckel RH, Grundy SM. **Harmonizing the metabolic syndrome: a joint interim statement of the International Diabetes Federation Task Force on Epidemiology and Prevention; National Heart, Lung, and Blood Institute; American Heart Association; World Heart Federation; International Atherosclerosis Society; and International Association for the Study of Obesity**. *Circulation* (2009.0) **120** 1640-1645. DOI: 10.1161/CIRCULATIONAHA.109.192644
30. **Executive summary of the third report of the National Cholesterol Education Program (NCEP) Expert Panel on Detection, Evaluation, and Treatment of High Blood Cholesterol in Adults (Adult Treatment Panel III)**. *JAMA* (2001.0) **285** 2486-2497. DOI: 10.1001/jama.285.19.2486
31. Aguilar M, Bhuket T, Torres S, Liu B, Wong RJ. **Prevalence of the metabolic syndrome in the United States, 2003-2012**. *JAMA* (2015.0) **313** 1973-1974. DOI: 10.1001/jama.2015.4260
32. 32National Center for Health Statistics. National Health and Nutrition Examination Survey: NHANES response rates and population totals. Accessed August 27, 2022. https://wwwn.cdc.gov/nchs/nhanes/responserates.aspx
33. Blüher M. **Metabolically healthy obesity**. *Endocr Rev* (2020.0) **41** 405-420. DOI: 10.1210/endrev/bnaa004
34. Cao J, Zhang S. **Multiple comparison procedures**. *JAMA* (2014.0) **312** 543-544. DOI: 10.1001/jama.2014.9440
35. O’Hearn M, Lauren BN, Wong JB, Kim DD, Mozaffarian D. **Trends and disparities in cardiometabolic health among U.S. adults, 1999-2018**. *J Am Coll Cardiol* (2022.0) **80** 138-151. DOI: 10.1016/j.jacc.2022.04.046
36. Shan Z, Rehm CD, Rogers G. **Trends in dietary carbohydrate, protein, and fat intake and diet quality among US adults, 1999-2016**. *JAMA* (2019.0) **322** 1178-1187. DOI: 10.1001/jama.2019.13771
37. Guo F, Garvey WT. **Trends in cardiovascular health metrics in obese adults: National Health and Nutrition Examination Survey (NHANES), 1988-2014**. *J Am Heart Assoc* (2016.0) **5**. DOI: 10.1161/JAHA.116.003619
38. Saydah S, Bullard KM, Cheng Y. **Trends in cardiovascular disease risk factors by obesity level in adults in the United States, NHANES 1999-2010**. *Obesity (Silver Spring)* (2014.0) **22** 1888-1895. DOI: 10.1002/oby.20761
39. Chelliah P, Li X, Adams-Huet B, Lingvay I. **Trends in the prevalence of cardiometabolic disease and cardiovascular events by body mass index category in adults from 1999 to 2016**. *Postgrad Med J* (2020.0) **96** 655-659. DOI: 10.1136/postgradmedj-2020-137749
40. Scheen AJ, Van Gaal LF. **Combating the dual burden: therapeutic targeting of common pathways in obesity and type 2 diabetes**. *Lancet Diabetes Endocrinol* (2014.0) **2** 911-922. DOI: 10.1016/S2213-8587(14)70004-X
41. Younossi ZM, Stepanova M, Younossi Y. **Epidemiology of chronic liver diseases in the USA in the past three decades**. *Gut* (2020.0) **69** 564-568. DOI: 10.1136/gutjnl-2019-318813
42. Stefan N, Häring HU. **The metabolically benign and malignant fatty liver**. *Diabetes* (2011.0) **60** 2011-2017. DOI: 10.2337/db11-0231
43. Du Y, Liu B, Sun Y, Snetselaar LG, Wallace RB, Bao W. **Trends in adherence to the physical activity guidelines for Americans for aerobic activity and time spent on sedentary behavior among US adults, 2007 to 2016**. *JAMA Netw Open* (2019.0) **2**. DOI: 10.1001/jamanetworkopen.2019.7597
44. He J, Zhu Z, Bundy JD, Dorans KS, Chen J, Hamm LL. **Trends in cardiovascular risk factors in us adults by race and ethnicity and socioeconomic status, 1999-2018**. *JAMA* (2021.0) **326** 1286-1298. DOI: 10.1001/jama.2021.15187
45. Wang L, Li X, Wang Z. **Trends in prevalence of diabetes and control of risk factors in diabetes among US adults, 1999-2018**. *JAMA* (2021.0) **326** 1-13. DOI: 10.1001/jama.2021.9883
46. Min J, Goodale H, Xue H, Brey R, Wang Y. **Racial-ethnic disparities in obesity and biological, behavioral, and sociocultural influences in the United States: a systematic review**. *Adv Nutr* (2021.0) **12** 1137-1148. DOI: 10.1093/advances/nmaa162
47. Ogden CL, Fakhouri TH, Carroll MD. **Prevalence of obesity among adults, by household income and education—United States, 2011-2014**. *MMWR Morb Mortal Wkly Rep* (2017.0) **66** 1369-1373. DOI: 10.15585/mmwr.mm6650a1
48. Kwak YE, McMillan R, McDonald EK. **Trends in overweight and obesity self-awareness among adults with overweight or obesity in the United States, 1999 to 2016**. *Ann Intern Med* (2021.0) **174** 721-723. DOI: 10.7326/M20-3882
49. Ortega FB, Cadenas-Sanchez C, Migueles JH. **Role of physical activity and fitness in the characterization and prognosis of the metabolically healthy obesity phenotype: a systematic review and meta-analysis**. *Prog Cardiovasc Dis* (2018.0) **61** 190-205. DOI: 10.1016/j.pcad.2018.07.008
50. Opio J, Croker E, Odongo GS, Attia J, Wynne K, McEvoy M. **Metabolically healthy overweight/obesity are associated with increased risk of cardiovascular disease in adults, even in the absence of metabolic risk factors: a systematic review and meta-analysis of prospective cohort studies**. *Obes Rev* (2020.0) **21**. DOI: 10.1111/obr.13127
51. Iacobini C, Pugliese G, Blasetti Fantauzzi C, Federici M, Menini S. **Metabolically healthy versus metabolically unhealthy obesity**. *Metabolism* (2019.0) **92** 51-60. DOI: 10.1016/j.metabol.2018.11.009
|
---
title: 'Artificial Intelligence in Community-Based Diabetic Retinopathy Telemedicine
Screening in Urban China: Cost-effectiveness and Cost-Utility Analyses With Real-world
Data'
journal: JMIR Public Health and Surveillance
year: 2023
pmcid: PMC9999255
doi: 10.2196/41624
license: CC BY 4.0
---
# Artificial Intelligence in Community-Based Diabetic Retinopathy Telemedicine Screening in Urban China: Cost-effectiveness and Cost-Utility Analyses With Real-world Data
## Abstract
### Background
Community-based telemedicine screening for diabetic retinopathy (DR) has been highly recommended worldwide. However, evidence from low- and middle-income countries (LMICs) on the choice between artificial intelligence (AI)–based and manual grading–based telemedicine screening is inadequate for policy making.
### Objective
The aim of this study was to test whether the AI model is more worthwhile than manual grading in community-based telemedicine screening for DR in the context of labor costs in urban China.
### Methods
We conducted cost-effectiveness and cost-utility analyses by using decision-analytic Markov models with 30 one-year cycles from a societal perspective to compare the cost, effectiveness, and utility of 2 scenarios in telemedicine screening for DR: manual grading and an AI model. Sensitivity analyses were performed. Real-world data were obtained mainly from the Shanghai Digital Eye Disease Screening Program. The main outcomes were the incremental cost-effectiveness ratio (ICER) and the incremental cost-utility ratio (ICUR). The ICUR thresholds were set as 1 and 3 times the local gross domestic product per capita.
### Results
The total expected costs for a 65-year-old resident were US $3182.50 and US $3265.40, while the total expected years without blindness were 9.80 years and 9.83 years, and the utilities were 6.748 quality-adjusted life years (QALYs) and 6.753 QALYs in the AI model and manual grading, respectively. The ICER for the AI-assisted model was US $2553.39 per year without blindness, and the ICUR was US $15,216.96 per QALY, which indicated that AI-assisted model was not cost-effective. The sensitivity analysis suggested that if there is an increase in compliance with referrals after the adoption of AI by $7.5\%$, an increase in on-site screening costs in manual grading by $50\%$, or a decrease in on-site screening costs in the AI model by $50\%$, then the AI model could be the dominant strategy.
### Conclusions
Our study may provide a reference for policy making in planning community-based telemedicine screening for DR in LMICs. Our findings indicate that unless the referral compliance of patients with suspected DR increases, the adoption of the AI model may not improve the value of telemedicine screening compared to that of manual grading in LMICs. The main reason is that in the context of the low labor costs in LMICs, the direct health care costs saved by replacing manual grading with AI are less, and the screening effectiveness (QALYs and years without blindness) decreases. Our study suggests that the magnitude of the value generated by this technology replacement depends primarily on 2 aspects. The first is the extent of direct health care costs reduced by AI, and the second is the change in health care service utilization caused by AI. Therefore, our research can also provide analytical ideas for other health care sectors in their decision to use AI.
## Introduction
Diabetic retinopathy (DR) is a leading cause of blindness worldwide. It often develops 10-15 years after the onset of diabetes and can take several forms—all potentially causing vision loss or blindness, including diabetic macular edema (DME) due to increased retinal vascular permeability and central retinal thickening, retinal ischemia resulting in the damage or death of light-sensing retinal photoreceptors, and proliferative DR, where the growth of fragile new blood vessels causes vitreous hemorrhage and retinal detachment [1].
In 2020, the global burden of DR and sight-threatening DR (STDR) was estimated to be 103 million and 29 million people, respectively, which will increase to 161 million and 45 million by 2045 due to the increasing prevalence of diabetes mellitus (DM) [1,2]. From 1990 to 2010, visual impairment due to DR increased by $64\%$ and blindness by $27\%$ [1], both of which were due to the rising DM prevalence in low- and middle-income countries (LMICs). Several studies have confirmed the benefits of telemedicine screening for DR [3-6]. Compared to no screening and traditional face-to-face screening, telemedicine screening is highly cost-effective in the long term. As a result, telemedicine screening will become the main form of community-based eye disease screening [5]. Some recent studies have suggested that using artificial intelligence (AI) can further reduce the costs of telemedicine screening [3,7-10]. Studies in high-income countries such as the United Kingdom and Singapore have shown that when AI is used in DR screening programs, screening costs can be reduced by up to approximately $20\%$ compared with the costs incurred in manual grading [7-9]. This can be easily understood: from the perspective of a health economic evaluation, the main difference between the AI model and manual grading is that technology costs replace labor costs. Therefore, in settings where labor costs are high, such as in high-income countries, using AI instead of manual grading would save a lot of screening costs, making the screening more cost-effective [10]. However, because labor costs are low in low-income countries, conclusions from high-income countries may not be equally suitable, and evidence from LMICs is inadequate. Therefore, the objective of our community-based telemedicine screening for DR was to examine whether the AI model can be more cost-effective than manual grading in LMICs. We conducted a health economic evaluation by using real-world data from a large community-based telemedicine screening program for DR in Shanghai, China. We expect that this study will provide a reference for policy making with regard to DR screening in the context of low labor costs.
## Study Setting
This study was conducted in Shanghai, China, wherein the prevalence of type 2 diabetes among adults was $6.25\%$ between 2016 and 2019 [11]. Since 2010, Shanghai has conducted a teleophthalmology-based DR screening program under which residents can undergo fundus photography at community health service centers. After retinal experts at designated DR diagnosis centers have made a diagnosis based on these images, screening results are fed into the community health service center, where patients are counselled by general practitioners, and medical advice is offered. By 2017, all 250 community health service centers in Shanghai were equipped to participate in this program, and plans had begun to build an AI-assisted DR screening system [12,13]. A convolutional neural network, a type of deep learning model [14], was applied to the problem of diagnosing DR from fundus images, with the aim of replacing retinal experts in DR diagnosis centers with the AI algorithm on a cloud-based server. Since 2020, 56 community health service centers have shifted to an AI-assisted DR screening system. In 2021, these centers screened approximately 40,000 community residents for DR by using AI. To maximize the efficiency of the Shanghai program, community health service centers coordinated voluntary screening for residents of a given community at a particular place and time. Those who were diagnosed with DR at hospitals could still participate in the free annual community screening to monitor disease progression.
## Model Overview
TreeAge Pro (TreeAge Software) was used to build a decision-analytic Markov model to compare the actual cost, effectiveness, and utility of manual grading telemedicine screening and AI-based assessment for DR (Multimedia Appendices 1-4). The incremental cost-effectiveness ratio (ICER) and incremental cost-utility ratio (ICUR) were calculated as the primary results. The effectiveness was defined as years without blindness per 100,000 people with DM, and the utility was evaluated by quality-adjusted life years (QALYs). Although all residents with DM could participate in our community-based screening, the majority were older people [15,16]; therefore, a hypothetical cohort of community residents with DM was followed in the model from the age of 65 years through a total of 30 one-year Markov cycles [5]. The characteristics of the simulated cohort were extracted using the Shanghai Digital Eye Disease Screening Program (Table 1).
**Table 1**
| Unnamed: 0 | Unnamed: 1 | All CHCsa | CHCs using artificial intelligence | CHCs using manual grading | P value |
| --- | --- | --- | --- | --- | --- |
| Community health center characteristicsb, mean (SD) | Community health center characteristicsb, mean (SD) | Community health center characteristicsb, mean (SD) | Community health center characteristicsb, mean (SD) | Community health center characteristicsb, mean (SD) | Community health center characteristicsb, mean (SD) |
| | Number of full-time or part-time ophthalmologists | 1.1 (0.7) | 1.1 (0.5) | 1.2 (0.7) | .62 |
| | Annual numbers of ophthalmology outpatients | 6248.6 (6019.0) | 6774.8 (6238.2) | 6016.2 (5946.4) | .54 |
| Screened residents’ characteristicsc | Screened residents’ characteristicsc | Screened residents’ characteristicsc | Screened residents’ characteristicsc | Screened residents’ characteristicsc | Screened residents’ characteristicsc |
| | Age (years), mean (SD) | 69.3 (7.2) | 69.8 (6.9) | 68.8 (7.4) | <.001 |
| | Sex (male), n (%) | 15,032 (46.0) | 7225 (46.1) | 7807 (45.8) | .60 |
| | Duration of diabetes (years), mean (SD) | 10.3 (7.7) | 9.8 (8.0) | 10.7 (7.3) | <.001 |
| | Visual acuity of right eye (logMAR), mean (SD) | 0.4 (0.4) | 0.4 (0.4) | 0.3 (0.4) | <.001 |
| | Visual acuity of left eye (logMAR), mean (SD) | 0.4 (0.4) | 0.4 (0.4) | 0.3 (0.4) | <.001 |
Individuals were enrolled as healthy (free from DR) or unhealthy (experiencing DR) and could die due to any reason. According to the English National Screening Program for Diabetic Retinopathy, a Markov model was constructed that included non-STDR, STDR, and DME [5,15-17]. The category was assigned based on the DR grade in the worse eye. During each 1-year cycle, an individual had a risk of progressing to the more severe stage or staying in the same stage. However, the model does not allow returning to an earlier stage even with treatment because of the nature of the disease. Moreover, the treatment can only decrease the probability of progression to the next stage. The prevalence of DR, the incidence of DR (including STDR and DME), transition probabilities, characteristics of DR screening tests, referral and treatment compliance, utility, mortality, and other relevant parameters were collected from published studies specific to Shanghai, other cities in China, and other Asian regions, as well as unpublished data sources (eg, Shanghai Digital Eye Disease Screening Program). The costs of screening, ocular examinations, and treatment were all derived from a real-world eye disease screening program in Shanghai and the unified health care service pricing of the Shanghai Municipal Health Commission. The parameters used in the basic analysis and the ranges used in the sensitivity analyses are listed in detail in Multimedia Appendices 1-4.
## Manual Grading–Based Telemedicine Screening
We invited the entire population with DM living in communities to participate in the DR screening program at local community health centers. All the participants underwent a series of screening tests conducted by trained general practitioners, ophthalmic technicians, optometrists, and ophthalmologists. The screening included a vision acuity test, refraction measurement by an autorefractor, and fundus photography using a non–mydriatic fundus camera. The data were transferred to the corresponding designated diagnosis center through a telemedicine platform after the completion of all the tests. After all the participants in 1 community health center completed the annual screening, the community health center contacted the designated diagnosis center, and 2 retinal experts (ophthalmologists) began to make the diagnosis based on retinal photography. In 2 weeks, screening results were provided as feedback to the community health center, where residents could receive medical advice from the general practitioners. Finally, patients with suspected STDR were referred to specialized ophthalmic hospitals or tertiary hospitals for a detailed re-examination to confirm the diagnosis (Multimedia Appendix 5 shows the screening and referral pathway). Those who were confirmed to have STDR were assumed to receive appropriate treatment and routine clinical care according to the severity of DR.
## AI-Assisted DR Telemedicine Screening
We invited the entire population with DM living in the community to participate in the DR screening program at the local community health center. The screening process was the same as that described for the manual grading–based telemedicine screening. However, after all the screening tests were completed, the data were transmitted to the AI algorithm on a cloud-based server center through the telemedicine platform. The screening results were provided as feedback immediately. Further management of patients with suspected STDR was the same as that described for manual grading–based telemedicine screening.
## Prevalence and Transition Probabilities
Data on the prevalence and incidence of DR, DME, and STDR were collected from published studies in Shanghai [15,18]. Because Jin et al’s [18] study only reported the 5-year incidence of STDR and DME, the 1-year incidence was calculated based on the formula: r = −log (1 − p)/t, where r represents the 1-year incidence and p represents the cumulative incidence over time interval t [19]. Other transition probabilities were obtained from published studies specific to China, and if few data were available for Chinese patients, data from other Asian regions were used. We searched PubMed and China National Knowledge Infrastructure by using the following combinations of terms: “diabetic retinopathy” AND “progression” OR “transition” AND “Chinese” OR “China.”
## Screening and Intervention Costs
Our study included both direct and indirect costs and analyzed them from a societal perspective. Direct medical costs comprised the charges of screening, examination, and treatment. Direct nonmedical costs consisted of transportation costs related to hospital visits, and indirect costs consisted of family members’ time associated with the visits and their wage loss. All costs were collected in Chinese yuan and then converted into US dollars at an exchange rate of CNY 6.90 per dollar [20]. All cost data are listed in Multimedia Appendices 4 and 6-8.
The screening costs were determined based on the Shanghai Digital Eye Disease Screening Program. The screening costs consisted of the purchase and maintenance costs of equipment, labor costs of medical personnel, transportation fees, and income loss for residents. We calculate the annualized cost for fixed assets by assuming a life span of 5 years and no salvage value. The construction and maintenance costs of the telemedicine platform were based on the Shanghai Digital Eye Disease Screening Program. Based on our field observations, it took 6.2, 3, 3.3, and 4.8 minutes on average for 1 participant to complete registration, visual acuity test autorefraction, and retinal photography, respectively. Theoretically, a team with 4 optometrists could screen approximately 100 participants per day, but under real-world working conditions, this is nearly 30 per day. As the participants in our model were older than 65 years, we assumed that they did not incur wage loss. Moreover, we did not include wage loss for the accompanying family members in the screening costs. Therefore, the total costs per person for manual grading–based and AI-based telemedicine screening were US $10.10 and US $9.60, respectively. Multimedia Appendix 6 shows the detailed composition of the screening costs.
To calculate the costs of the detailed re-examinations after referral, direct medical costs consisted of the costs of ocular examinations and equipment and wages for medical personnel; direct nonmedical costs comprised transportation fees related to the visits; and indirect costs included 1 accompanying family member’s wage loss for time spent and per capita daily income in Shanghai in 2020. The examination costs were the unified pricing of the Shanghai Municipal Health Commission. Because public hospitals are nonprofit institutions, the money from these fees is mainly used to subsidize the cost of health care services. Hence, prices in public hospitals can be used to estimate the direct medical costs. Detailed information on the hospital-based examination costs is provided in Multimedia Appendix 7. It was assumed that the wage loss of the accompanying family member for referral was 0 because the majority of them were older than 65 years.
For treatment costs, direct medical costs included the costs of treatment, equipment, and wages for medical personnel; direct nonmedical costs consisted of costs of transportation related to the visits; and indirect costs included 1 accompanying family member’s wage loss based on time spent and per capita daily income in Shanghai in 2020. In the first year, patients with DME were assumed to have received 3 antivascular endothelial growth factor injections. Photocoagulation or vitrectomy was administered to patients with severe nonproliferative DR or proliferative DR. In the follow-up years, an average of 1 antivascular endothelial growth factor injection was administered, and an annual outpatient review was required for patients with STDR. Direct medical costs were estimated using the prices of health care services in public hospitals. The total economic burden for blind patients in the first year was estimated to be US $8920, which included $53.2\%$ direct medical costs, $6.4\%$ direct nonmedical costs, and $40.4\%$ indirect costs (loss of labor resources for family members and low-vision services costs), and there were only indirect costs in the follow-up years [5,21]. Detailed information on the treatment costs is provided in Multimedia Appendix 8.
## Utility and QALYs
We estimated the utility values for each DR stage (seen in Multimedia Appendix 9) to calculate QALYs. Utility values were based on published studies from China and other Asian countries [21,22]. Because the residents who participated in the screening should have diabetes, utility was assumed to be 0.87 but not 1.0 for people without DR, 0.79 for those with non-STDR, and 0.7 for those with STDR (including severe nonproliferative DR, proliferative DR, and DME). The utility value for people with blindness was assumed to be 0.55 [22]. All the values for the base case and sensitivity analyses are listed in Multimedia Appendix 3.
## Compliance
Compliance with referral to specialized ophthalmic hospitals or tertiary hospitals for a full examination among patients screened for signs of STDR was assumed to be $50.4\%$ for manual grading–based telemedicine screening, according to our investigation in Shanghai [23]. However, compliance with AI-based telemedicine screening was unclear. Because only 1 published study suggested that adopting an AI-assisted diagnosis model in DR screening may impact the participants’ adherence to ophthalmic care [24], the evidence is insufficient. Therefore, we assumed that compliance with referral in AI-based telemedicine screening was the same as that in manual grading–based telemedicine screening, while we set a wide range (±$25\%$) for sensitivity analysis (Multimedia Appendix 3).
## Screening Accuracy
The accuracy of AI-based telemedicine screening was extracted from published studies specific to the AI-assisted screening model conducted in Shanghai based on the current dominant architecture of convolutional neural networks (Multimedia Appendix 10) [25]. Briefly, the sensitivity was $80.47\%$ ($95\%$ CI $75.07\%$-$85.14\%$) and the specificity was $97.96\%$ ($95\%$ CI $96.75\%$-$98.81\%$) for STDR [25]. In our screening program, 2 experienced ophthalmologists were employed to make the diagnoses based on the retinal images. Furthermore, the accuracy of the manual grading–based telemedicine screening was assumed to be $100\%$, which was in accordance with the DR diagnosis criteria [8,9,26,27]. However, as described in some other studies, since trained graders instead of ophthalmologists performed the grading and diagnosis [5,7] in the sensitivity analysis, we adopted the accuracy range of the manual grading based on the Singaporean study (Multimedia Appendix 3) [7].
## Other Parameters
The natural age-specific mortality rates of the general Chinese population reported by Zhang and Wei [28] were used in this study. Increased odds of mortality were assumed for people without DR but with DM, non-STDR, STDR, and blindness (Multimedia Appendix 3) [29,30]. Both costs and health state utilities were discounted at a $3.5\%$ annual rate in the base analysis, following the National Institute for Health and Care Excellence recommendations [31]. For the cost-effectiveness threshold, 2 thresholds representing cost-effectiveness and high cost-effectiveness were used according to the World Health Organization recommendations [5,21]. Among the interventions improving the patients’ utilities, those that cost less than the gross domestic product (GDP) per capita are defined as highly cost-effective, those that cost 1-3 times the GDP per capita are defined as cost-effective, and those that cost more than 3 times the GDP per capita are determined as not cost-effective [32]. On the contrary, among the interventions reducing the participants’ utilities, among those saving costs higher than 3 times, the GDP per capita was defined as highly cost-effective; among those saving between 1 and 3 times, the GDP per capita was defined as cost-effective; and those costing less than the GDP per capita were determined as not cost-effective [32]. As the GDP per capita in Shanghai in 2020 was reported to be US $22,600, the thresholds in this study were defined as US $22,600 and US $67,800 [33].
## Outcomes
The ICER and ICUR were calculated as the difference in the total costs between the AI-assisted and manual grading telemedicine screening divided by the difference in the total years without blindness and the QALYs between the 2 conditions, respectively. Values for the AI-assisted screening cohort minus those for the manual grading screening cohort, which were set as the baseline, were calculated as the differences.
## Sensitivity Analysis
Extensive 1-way deterministic and probabilistic sensitivity analyses were performed to calculate the uncertainties of the base-case results. A variation of $10\%$ was adopted because probability-related statistics (ie, utility, prevalence, sensitivity, specificity, transition probability, and compliance) were mainly derived from previously published studies. For the influence of AI use on compliance with referral, a range of $25\%$ was used. A large floating range of $50\%$ was adopted for these costs. In addition, we adopted the accuracy range for manual grading according to the Singaporean study (Multimedia Appendix 3) to account for the influence of trained graders performing the grading and diagnosis instead of retinal experts [7]. A probabilistic sensitivity analysis was conducted using Monte Carlo simulation for 10,000 simulations to assess the robustness of the base case analysis. Beta distributions were adopted for probability-related data and utility values, gamma distributions were used for costs, and log-normal distributions were used for odds ratios. The methods and results conformed to the Consolidated Health Economic Evaluation Reporting Standards [2022] (Multimedia Appendix 11).
## Ethics Approval
This study was mainly based on the secondary analyses of published data. Written informed consent was obtained from all the participants. All the study data were anonymous. There were no compensation fees for the participants. This study was approved by the Institutional Review Board of the Shanghai General Hospital (2022SQ272) and Shanghai Eye Diseases Prevention and Treatment Center (2022SQ007).
## Results
The cost-effectiveness and cost-utility analyses showed that AI-based telemedicine screening was dominated by manual grading–based telemedicine screening in Shanghai (Table 2). In the manual grading–based telemedicine screening, a community resident with DM would incur a total cost of US $3265.40, including screening, hospital referral for confirmation, and treatment as needed, with 9.83 years without blindness and 6.753 QALYs. In the AI-based telemedicine screening, a community resident with DM would incur a total cost of US $3182.50, with 9.80 years without blindness and 6.748 QALYs. Therefore, compared with the cost of manual grading–based telemedicine screening, that of the AI-based telemedicine screening model was $2.5\%$ lower, while the years without blindness was $0.3\%$ less, and the QALYs were $0.1\%$ less.
**Table 2**
| Unnamed: 0 | Costs per person (USD) | Incremental costs per 100,000 people screened (USD) | Years without blindness per person | Incremental years without blindness per 100,000 people screened | QALYb per person | Incremental quality-adjusted life years per 100,000 people screened | ICERc (USD) | ICURd (USD) |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| AIe-assisted model | 3182.47 | –8,289,840.65 | 9.8 | –3121.32 | 6.748 | –544.78 | 2553.39 | 15216.96 |
| Manual grading | 3265.37 | N/Af | 9.83 | | 6.753 | | | |
Our results showed that by replacing manual grading–based telemedicine screening with AI-based telemedicine screening, 1 participant could save US $15,216.96 but needed to lose 1 more QALY (ICUR=US $15,216.96), indicating that AI-based telemedicine screening was not cost-effective as in Shanghai in 2020; at least US $22,600 (GDP per capita) should be saved if 1 more QALY is lost due to the shift in interventions. A 1-way deterministic sensitivity analysis of cost-effectiveness and cost-utility analyses indicated that the impact of the adoption of AI on compliance with referral, costs of on-site screening in manual grading–based telemedicine screening, costs of on-site screening in AI-based telemedicine screening, treatment costs for the follow-up of patients with DME, and treatment costs for the follow-up of patients with severe nonproliferative DR and proliferative DR were the 5 most influential variables. In particular, according to the cost-utility analysis, if the adoption of AI could improve compliance with referrals by $7.5\%$, the AI-assisted model might be cost-effective; if compliance was improved by $17.5\%$, the AI-assisted model might be highly cost-effective; and if compliance was improved by $25\%$, the AI-assisted model might be the absolutely dominant strategy, as it could save costs and increase the years without blindness and QALYs (Multimedia Appendix 12). Moreover, the increase in the costs of on-site screening in manual grading–based telemedicine screening and the decrease in the costs of on-site screening in AI-based telemedicine screening might help the AI-based telemedicine screening to be cost-effective (Figure 1). The detailed sensitivity analysis results of the other parameters are shown in Multimedia Appendices 13 and 14.
**Figure 1:** *One-way deterministic sensitivity analysis (Tornado diagram). A. One-way sensitivity analysis of cost-effectiveness. B. One-way sensitivity analysis of cost-utility. Since negative values of incremental cost-effectiveness ratio or incremental cost-utility ratio might occur due to the change of compliance with referral after the adoption of artificial intelligence (multiplier), detailed results have been shown in Multimedia Appendix 12 separately. Therefore, in this Tornado diagram, the impact of the change of compliance with referral after the adoption of artificial intelligence (multiplier) is not shown. AI: artificial intelligence; DME: diabetic macular edema; ICER: incremental cost-effectiveness ratio; ICUR: incremental cost-utility ratio; NPDR: nonproliferative diabetic retinopathy; PDR: proliferative diabetic retinopathy.*
Probabilistic sensitivity analysis showed that the base-case ICER and ICUR were robust to randomly distributed parameters (Figure 2). We obtained cost-effectiveness acceptability curves by taking 10,000 random draws (Figure 3). This means that when both AI-based and manual grading–based telemedicine screening were available, manual grading–based telemedicine screening was the dominant strategy in $60.6\%$ of the simulations under the threshold of GDP per capita (US $22,600) and in $84.5\%$ of the simulations under the threshold of 3 times the GDP per capita (US $67,800).
**Figure 2:** *Probabilistic sensitivity analysis. A. Probabilistic sensitivity analysis of cost-effectiveness. B. Probabilistic sensitivity analysis of cost-utility. GDP: gross domestic product.* **Figure 3:** *Cost-effectiveness acceptability curves. AI: artificial intelligence; GDP: gross domestic product.*
## Principal Findings
This study presents one of the first health economic evaluations in the context of low labor costs in an LMIC setting of competing telemedicine models for community-based DR screening using manual grading and AI models. In line with previous studies [7,8], our analysis was based on an established telemedicine screening program. We showed that, in this context, the value of AI-based telemedicine DR screening depended heavily on the referral compliance of patients with suspected STDR. If this compliance did not increase, AI-based telemedicine DR screening would not be more cost-effective than manual grading–based telemedicine DR screening because it would decrease the long-term screening effectiveness and individuals’ health utility and not save enough costs.
Prior studies in Singapore and the United Kingdom [7,8] showed that replacing manual grading with an AI model for DR screening led to a $12\%$-$20\%$ cost reduction. A study in Scotland reported an even greater cost reduction of $46.7\%$ [9]. These studies were conducted in high-income countries in the context of high labor costs. However, China’s national conditions differ from those in high-income countries. One of the most important differences is that the labor costs of the medical staff are much lower in China. For example, in Singapore, the labor cost for DR grading is US $26 per participant [7], which is over 20 times that of the labor cost for manual grading in Shanghai. Consequently, in the context of low labor costs, a reduction in screening costs resulting from the use of AI solutions is limited in China.
In the Shanghai program, the on-site screening cost for 1 participant, including screening examinations and diagnosis, was US $10.10 in the manual grading model and US $9.60 in the AI-assisted model—a reduction of only $5\%$. Moreover, the labor costs of the medical staff in our screening program were among the highest in China. For example, according to the Wenzhou ophthalmologic screening program, the labor cost for on-site screening examinations and diagnosis of glaucoma was US $1.70 per participant, which was about a quarter of the labor costs in our Shanghai program (US $7.20 per participant for manual grading–based telemedicine screening, Multimedia Appendix 6) [21], and according to the Finance Department and Procurement Center of Beijing Tongren Hospital, Beijing Tongren Eye Centre Ocular Reading Centre, and China Intelligent Ophthalmology Big Data Research Center, the labor costs for on-site screening examinations and diagnosis were US $1.75 per participant for traditional face-to-face screening and US $0.80 per participant for manual grading–based telemedicine screening [5], which is only one-ninth of the labor costs in our Shanghai program (US $7.20 per participant for manual grading–based telemedicine screening, Multimedia Appendix 6). Because the main difference between the cost components of the AI-assisted model and those of the manual grading model is that equipment and telemedicine platform costs replace labor costs, in settings where labor costs are extremely low, cost reduction via the adoption of AI is expected to be even lesser. Therefore, AI-assisted DR screening is less cost-effective in other urban areas of China.
Our sensitivity analysis confirmed these results. The on-site screening costs of both manual grading and AI-assisted models are among the most influential variables. An increase in the on-site screening costs of manual grading by $50\%$ or a decrease in the on-site screening costs of the AI-assisted model by $50\%$ may help AI-based telemedicine screening be cost-effective. In other words, the gap between the on-site screening costs of the manual grading model and the AI-assisted model is the key point. Moreover, the cost of AI software was only $7\%$ of the on-site costs in AI-based telemedicine screening (Multimedia Appendix 6). Therefore, even if the AI software were completely free, a $50\%$ reduction in the on-site screening costs of the AI-assisted model would not be achieved. As a result, unless the labor costs of medical staff increase dramatically in the future, the AI-assisted model will be hardly cost-effective in Shanghai, holding all the other conditions constant.
However, there was one exception to this. Our sensitivity analysis shows that if the referral compliance of patients with suspected STDR increased after the adoption of AI even to a small extent, then the AI-assisted model would be cost-effective. A study in Missouri [24] suggested that after the adoption of the automated retinal image assessment system, which is based on AI, the rates of completed referral eye examinations at 3, 6, and 12 months after screening increased from $9.4\%$ to $32.6\%$, from $13.4\%$ to $46.7\%$, and from $18.7\%$ to $55.4\%$, respectively [24]. However, relevant evidence is still inadequate; therefore, it is difficult to determine whether this improvement is an isolated case. Previously, we implemented a discrete choice experiment in Shanghai to measure individuals’ preferences for AI-based screening [34]. The results suggested that the impact of the adoption of AI on individuals’ preferences may be bidirectional. On the one hand, algorithm aversion should be noted, which means that compared to manual grading, the residents were in disfavor of the AI-assisted screening technology [34,35]. On the other hand, the immediate feedback of retinal screening results by the adoption of AI could increase the individuals’ preferences and have profound effects on participants’ follow-up behavior [24,34]. Nevertheless, there is still a lack of empirical studies on the association between the results of feedback efficiency and residents’ referral compliance.
This study has several strengths. This study may provide a reference for policy making in planning community-based DR screening in LMICs by modelling 2 practical telemedicine screening models for DR by using real-world data from an ongoing program in urban China. In addition, we conducted a sensitivity analysis of our models within wide ranges and identified the most influential variables affecting the decision to use AI and manual grading in telemedicine screening. Therefore, our conclusions provide practical value in the policy-making process regarding when to deploy AI-assisted diagnostic technology.
## Limitations
Our study had several limitations despite its numerous strengths. Most notably, we only compared the models for centralized screening. Other models must also be considered going forward. For example, in Shanghai, some community health service centers are beginning to provide DR screening as part of their outpatient services for patients with diabetes. This change in the model may impact both the costs and patients’ compliance, thus altering the results of health economic evaluations such as ours. Second, our comparison is based on the premise that both human- and AI-based models are available and affordable. However, in some remote regions, due to the lack of human resources, manual screening for eye disease may be impractical, and AI-based screening, if available, may be the only option. Third, our study is mainly based on empirical data from Shanghai; therefore, it cannot be representative of the whole of China because of the huge regional and medical care differences between urban and rural areas. Therefore, there is an urgent need for more extensive and in-depth studies. However, as we have discussed above, the labor costs of medical staff in Shanghai are among the highest in China, and AI-based telemedicine screening will become even more less cost-effective if the labor cost of medical staff is further reduced. Therefore, our findings can be extrapolated within the Chinese context.
## Conclusion
Our study may provide a reference for policy making in planning community-based telemedicine screening for DR in LMICs. Our findings indicate that unless the referral compliance of patients with suspected STDR increases, the adoption of the AI model may not further improve the value of telemedicine screening compared to that of manual grading in LMICs. The main reason is that in the context of low labor costs, the direct health care costs saved by replacing manual grading with AI are limited, and screening effectiveness will decrease. In conclusion, our study suggests that the magnitude of the value generated by this technology replacement depends mainly on 2 aspects. The first is the extent of direct health care costs reduced by using AI, and the second is the change in health care service utilization caused by using AI. Therefore, our research also provides analytical ideas for other health care sectors in addition to eye care when deciding whether to use AI.
## References
1. Jampol LM, Glassman AR, Sun J. **Evaluation and Care of Patients with Diabetic Retinopathy**. *N Engl J Med* (2020.0) **382** 1629-1637. DOI: 10.1056/nejmra1909637
2. Teo ZL, Tham Y, Yu M, Chee ML, Rim TH, Cheung N, Bikbov MM, Wang YX, Tang Y, Lu Y, Wong IY, Ting DSW, Tan GSW, Jonas JB, Sabanayagam C, Wong TY, Cheng C. **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
3. Vujosevic S, Aldington SJ, Silva P, Hernández Cristina, Scanlon P, Peto T, Simó Rafael. **Screening for diabetic retinopathy: new perspectives and challenges**. *Lancet Diabetes Endocrinol* (2020.0) **8** 337-347. DOI: 10.1016/S2213-8587(19)30411-5
4. Siegel KR, Ali MK, Zhou X, Ng BP, Jawanda S, Proia K, Zhang X, Gregg EW, Albright AL, Zhang P. **Cost-effectiveness of Interventions to Manage Diabetes: Has the Evidence Changed Since 2008?**. *Diabetes Care* (2020.0) **43** 1557-1592. DOI: 10.2337/dci20-0017
5. Li R, Yang Z, Zhang Y, Bai W, Du Y, Sun R, Tang J, Wang N, Liu H. **Cost-effectiveness and cost-utility of traditional and telemedicine combined population-based age-related macular degeneration and diabetic retinopathy screening in rural and urban China**. *Lancet Reg Health West Pac* (2022.0) **23** 100435. DOI: 10.1016/j.lanwpc.2022.100435
6. Avidor D, Loewenstein A, Waisbourd M, Nutman A. **Cost-effectiveness of diabetic retinopathy screening programs using telemedicine: a systematic review**. *Cost Eff Resour Alloc* (2020.0) **18** 16. DOI: 10.1186/s12962-020-00211-1
7. Xie Y, Nguyen QD, Hamzah H, Lim G, Bellemo V, Gunasekeran DV, Yip MYT, Qi Lee X, Hsu W, Li Lee M, Tan CS, Tym Wong H, Lamoureux EL, Tan GSW, Wong TY, Finkelstein EA, Ting DSW. **Artificial intelligence for teleophthalmology-based diabetic retinopathy screening in a national programme: an economic analysis modelling study**. *The Lancet Digital Health* (2020.0) **2** e240-e249. DOI: 10.1016/s2589-7500(20)30060-1
8. Tufail A, Rudisill C, Egan C, Kapetanakis VV, Salas-Vega S, Owen CG, Lee A, Louw V, Anderson J, Liew G, Bolter L, Srinivas S, Nittala M, Sadda S, Taylor P, Rudnicka AR. **Automated Diabetic Retinopathy Image Assessment Software: Diagnostic Accuracy and Cost-Effectiveness Compared with Human Graders**. *Ophthalmology* (2017.0) **124** 343-351. DOI: 10.1016/j.ophtha.2016.11.014
9. Scotland GS, McNamee P, Philip S, Fleming AD, Goatman KA, Prescott GJ, Fonseca S, Sharp PF, Olson JA. **Cost-effectiveness of implementing automated grading within the national screening programme for diabetic retinopathy in Scotland**. *Br J Ophthalmol* (2007.0) **91** 1518-23. DOI: 10.1136/bjo.2007.120972
10. Dismuke C. **Progress in examining cost-effectiveness of AI in diabetic retinopathy screening**. *The Lancet Digital Health* (2020.0) **2** e212-e213. DOI: 10.1016/s2589-7500(20)30077-7
11. Ping L, Yuanyuan W. **Investigation on the prevalence and influencing factors of type 2 diabetes among adults in Shanghai and research on prevention and treatment strategies**. *Chinese Primary Health Care* (2020.0)
12. Xu Y, Wang Y, Liu B, Tang L, Lv L, Ke X, Ling S, Lu L, Zou H. **The diagnostic accuracy of an intelligent and automated fundus disease image assessment system with lesion quantitative function (SmartEye) in diabetic patients**. *BMC Ophthalmol* (2019.0) **19** 184. DOI: 10.1186/s12886-019-1196-9
13. Dai L, Wu L, Li H, Cai C, Wu Q, Kong H, Liu R, Wang X, Hou X, Liu Y, Long X, Wen Y, Lu L, Shen Y, Chen Y, Shen D, Yang X, Zou H, Sheng B, Jia W. **A deep learning system for detecting diabetic retinopathy across the disease spectrum**. *Nat Commun* (2021.0) **12** 3242. DOI: 10.1038/s41467-021-23458-5
14. Li B, Chen H, Zhang B, Yuan M, Jin X, Lei B, Xu J, Gu W, Wong DCS, He X, Wang H, Ding D, Li X, Chen Y, Yu W. **Development and evaluation of a deep learning model for the detection of multiple fundus diseases based on colour fundus photography**. *Br J Ophthalmol* (2022.0) **106** 1079-1086. DOI: 10.1136/bjophthalmol-2020-316290
15. He J, Xu X, Zhu J, Zhu B, Zhang B, Lu L, He X, Bai X, Xu X, Zou H. **Lens Power, Axial Length-to-Corneal Radius Ratio, and Association with Diabetic Retinopathy in the Adult Population with Type 2 Diabetes**. *Ophthalmology* (2017.0) **124** 326-335. DOI: 10.1016/j.ophtha.2016.10.041
16. Lian JX, Gangwani RA, McGhee SM, Chan CKW, Lam CLK, Wong David Sai Hung. **Systematic screening for diabetic retinopathy (DR) in Hong Kong: prevalence of DR and visual impairment among diabetic population**. *Br J Ophthalmol* (2016.0) **100** 151-5. DOI: 10.1136/bjophthalmol-2015-307382
17. Harding S, Greenwood R, Aldington S, Gibson J, Owens D, Taylor R, Kohner E, Scanlon P, Leese G. **Grading and disease management in national screening for diabetic retinopathy in England and Wales**. *Diabet Med* (2003.0) **20** 965-71. DOI: 10.1111/j.1464-5491.2003.01077.x
18. Jin P, Peng J, Zou H, wang W, Fu J, Bai X, Xu X, Zhang X. **A 5-year prospective study of type 2 diabetes patients in Shanghai Xinjing Community, the incidence and risk factors of diabetic retinopathy and diabetic macular edema in Chinese type 2 diabetes residents**. *Chin J Exp Ophthalmol (in Chinese)* (2016.0)
19. Rothman K. *Epidemiology: An Introduction, Second Edition* (2012.0)
20. **Central parity rate of CNY (historical data)**. *State Administration of Foreign Exchange*
21. Tang J, Liang Y, O'Neill C, Kee F, Jiang J, Congdon N. **Cost-effectiveness and cost-utility of population-based glaucoma screening in China: a decision-analytic Markov model**. *The Lancet Global Health* (2019.0) **7** e968-e978. DOI: 10.1016/s2214-109x(19)30201-3
22. Nguyen HV, Tan GSW, Tapp RJ, Mital S, Ting DSW, Wong HT, Tan CS, Laude A, Tai ES, Tan NC, Finkelstein EA, Wong TY, Lamoureux EL. **Cost-effectiveness of a National Telemedicine Diabetic Retinopathy Screening Program in Singapore**. *Ophthalmology* (2016.0) **123** 2571-2580. DOI: 10.1016/j.ophtha.2016.08.021
23. Zhu X, Xu Y, Lu L, Zou H. **Patients' perspectives on the barriers to referral after telescreening for diabetic retinopathy in communities**. *BMJ Open Diabetes Res Care* (2020.0) **8** e000970. DOI: 10.1136/bmjdrc-2019-000970
24. 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
25. Zhang W, Li D, Wei Q, Ding D, Meng L, Wang Y, Zhao X, Chen Y. **The Validation of Deep Learning-Based Grading Model for Diabetic Retinopathy**. *Front Med (Lausanne)* (2022.0) **9** 839088. DOI: 10.3389/fmed.2022.839088
26. Lee Aaron Y, Yanagihara Ryan T, Lee Cecilia S, Blazes Marian, Jung Hoon C, Chee Yewlin E, Gencarella Michael D, Gee Harry, Maa April Y, Cockerham Glenn C, Lynch Mary, Boyko Edward J. **Multicenter, Head-to-Head, Real-World Validation Study of Seven Automated Artificial Intelligence Diabetic Retinopathy Screening Systems**. *Diabetes Care* (2021.0) **44** 1168-1175. DOI: 10.2337/dc20-1877
27. Lin D, Xiong J, Liu C, Zhao L, Li Z, Yu S, Wu X, Ge Z, Hu X, Wang B, Fu M, Zhao X, Wang X, Zhu Y, Chen C, Li T, Li Y, Wei W, Zhao M, Li J, Xu F, Ding L, Tan G, Xiang Y, Hu Y, Zhang P, Han Y, Li JO, Wei L, Zhu P, Liu Y, Chen W, Ting DSW, Wong TY, Chen Y, Lin H. **Application of Comprehensive Artificial intelligence Retinal Expert (CARE) system: a national real-world evidence study**. *The Lancet Digital Health* (2021.0) **3** e486-e495. DOI: 10.1016/s2589-7500(21)00086-8
28. Zhang W, Wei M. **The Evaluation of the Mortality and Life Expectancy of Chinese Population**. *Population Journal. (in Chinese)* (2016.0) **38** 18-28. DOI: 10.16405/j.cnki.1004-129X.2016.03.002
29. Vijan S, Hofer T P, Hayward R A. **Cost-utility analysis of screening intervals for diabetic retinopathy in patients with type 2 diabetes mellitus**. *JAMA* (2000.0) **283** 889-96. DOI: 10.1001/jama.283.7.889
30. Bragg F, Holmes MV, Iona A, Guo Y, Du H, Chen Y, Bian Z, Yang L, Herrington W, Bennett D, Turnbull I, Liu Y, Feng S, Chen J, Clarke R, Collins R, Peto R, Li L, Chen Z. **Association Between Diabetes and Cause-Specific Mortality in Rural and Urban Areas of China**. *JAMA* (2017.0) **317** 280-289. DOI: 10.1001/jama.2016.19720
31. **Guide to the methods of technology appraisal 2013**. *National Institute for Health and Care Excellence*
32. Al MJ. **Cost-effectiveness acceptability curves revisited**. *Pharmacoeconomics* (2013.0) **31** 93-100. DOI: 10.1007/s40273-012-0011-8
33. **GDP per capita in Shanghai (1978-2020)**. *Shanghai Municipal Statistics Bureau*
34. Lin S, Li L, Zou H, Xu Y, Lu L. **Medical Staff and Resident Preferences for Using Deep Learning in Eye Disease Screening: Discrete Choice Experiment**. *J Med Internet Res* (2022.0) **24** e40249. DOI: 10.2196/40249
35. Dietvorst BJ, Simmons JP, Massey C. **Algorithm aversion: people erroneously avoid algorithms after seeing them err**. *J Exp Psychol Gen* (2015.0) **144** 114-26. DOI: 10.1037/xge0000033
|
---
title: 'Toward Individualized Prediction of Binge-Eating Episodes Based on Ecological
Momentary Assessment Data: Item Development and Pilot Study in Patients With Bulimia
Nervosa and Binge-Eating Disorder'
journal: JMIR Medical Informatics
year: 2023
pmcid: PMC9999257
doi: 10.2196/41513
license: CC BY 4.0
---
# Toward Individualized Prediction of Binge-Eating Episodes Based on Ecological Momentary Assessment Data: Item Development and Pilot Study in Patients With Bulimia Nervosa and Binge-Eating Disorder
## Abstract
### Background
Prevention of binge eating through just-in-time mobile interventions requires the prediction of respective high-risk times, for example, through preceding affective states or associated contexts. However, these factors and states are highly idiographic; thus, prediction models based on averages across individuals often fail.
### Objective
We developed an idiographic, within-individual binge-eating prediction approach based on ecological momentary assessment (EMA) data.
### Methods
We first derived a novel EMA-item set that covers a broad set of potential idiographic binge-eating antecedents from literature and an eating disorder focus group ($$n = 11$$). The final EMA-item set (6 prompts per day for 14 days) was assessed in female patients with bulimia nervosa or binge-eating disorder. We used a correlation-based machine learning approach (Best Items Scale that is Cross-validated, Unit-weighted, Informative, and Transparent) to select parsimonious, idiographic item subsets and predict binge-eating occurrence from EMA data (32 items assessing antecedent contextual and affective states and 12 time-derived predictors).
### Results
On average 67.3 (SD 13.4; range 43-84) EMA observations were analyzed within participants ($$n = 13$$). The derived item subsets predicted binge-eating episodes with high accuracy on average (mean area under the curve 0.80, SD 0.15; mean $95\%$ CI 0.63-0.95; mean specificity 0.87, SD 0.08; mean sensitivity 0.79, SD 0.19; mean maximum reliability of rD 0.40, SD 0.13; and mean rCV 0.13, SD 0.31). Across patients, highly heterogeneous predictor sets of varying sizes (mean 7.31, SD 1.49; range 5-9 predictors) were chosen for the respective best prediction models.
### Conclusions
Predicting binge-eating episodes from psychological and contextual states seems feasible and accurate, but the predictor sets are highly idiographic. This has practical implications for mobile health and just-in-time adaptive interventions. Furthermore, current theories around binge eating need to account for this high between-person variability and broaden the scope of potential antecedent factors. Ultimately, a radical shift from purely nomothetic models to idiographic prediction models and theories is required.
## Binge Eating
Binge eating (objectively excessive food intake accompanied by feelings of loss of control) represents a core symptom of bulimia nervosa (BN), binge-eating disorder (BED), and the binge-purge subtype of anorexia nervosa. It is also the most debilitating symptom in most eating disorders (alongside the associated compensatory behavior in BN and binge-purge subtype of anorexia nervosa), accounting for gastrointestinal comorbidities, along with psychological consequences (eg, shame, secrecy, and social isolation [1]). Thus, interventions have focused on binge eating to ameliorate psychological consequences and subsequent purging behavior, which further contributes to oral and dental harms. However, treatment as usual–cognitive behavioral therapy for eating disorders (EDs) is effective for only about $65\%$ of individuals with an ED [2] and has high relapse rates ($26.8\%$ across EDs) [3].
## Nomothetic Binge-Eating Models
To predict binge eating, researchers typically rely on nomothetic theories—theories that are based on the average characteristics of multiple individuals in groups. Some nomothetic findings hold that individuals with BN and BED overeat in response to negative emotions, whereas healthy controls do not [4]. However, although particular efforts have been directed at predicting high-risk states for binge eating based on a variety of measures (eg, negative emotions or irregular eating patterns) [5,6], nomothetic binge-eating models often fail to translate to an idiographic–individual–level [7-10]. To illustrate, nomothetic theories claiming that emotional eating underlies binge eating [11] imply that emotion regulation interventions provide causative help [12,13]. However, this reasoning might not be applicable to patients who are prone to binge eating when impulsive, after extensive fasting periods, or experiencing dissociative states [6,13,14]. Correspondingly, various nomothetic theories of binge eating have proliferated. They differ substantially in the assumed causal mechanisms, which include, but are not limited to, emotional eating, impulsivity, restrained eating, food addiction, ego depletion, associative learning, and emotion-regulation or coping with emotions [4,15-21].
## Idiographic Binge-Eating Models and Interventions
As binge eating can be highly impulsive, automatic, and difficult to resist, interventions that target binge eating based on its antecedents are promising as they attempt to stop the process as soon as possible, before the binge-eating pressure builds up. As such states can fluctuate quickly, they need to be assessed and evaluated with a high timely resolution to inform about the appropriate timing for interventions for high-risk states. Recently, the methodologies of just-in-time adaptive interventions (JITAIs) [22] and high-frequency ecological momentary assessment (EMA) have merged into a methodological framework that can be applied to binge-eating prediction and prevention. JITAIs have been shown to enhance cognitive behavioral therapy in BED and BN [23] and have been successfully implemented in other domains of eating behavior (eg, in weight loss) [24].
In the “OnTrack” weight-loss intervention, Forman et al [24] sampled emotions and stress, next to eating history and context conditions such as watching television or alcohol consumption. By investigating a wide range of antecedents for dietary lapses, they go well beyond what emotional-eating theory suggests as predictors (eg, negative emotions). Similarly, with their “Think Slim” app, Spanakis et al [25] showed that a sample of participants with normal weight and overweight can be clustered into multiple groups according to the different momentary states in which they tend to eat unhealthily. Therefore, applying a single nomothetic binge-eating theory might be insufficient to identify a broad spectrum of individually varying antecedents and would yield inaccurate predictions of binge eating in most individuals [26]. Instead, to cover all relevant antecedents for many patients, a broad set of EMA items is required. Notably, items that serve this purpose in weight disorders (eg, “OnTrack” JITAI-enhanced weight-loss intervention by Forman et al [24]) may not cover all antecedent states that arise in patients with clinical binge eating. Furthermore, despite being often disregarded within nomothetic frameworks, protective factors (eg, positive emotions or healthy coping [27,28]) have the potential to improve prediction accuracies in idiographic machine learning frameworks because of their negative associations with binge-eating likelihood. However, to balance participant burden with broad sampling, a baseline phase with the full item set could be followed by a phase with a reduced EMA-item set, based on a prediction model that identifies the ideographic subset of items that best predict binge eating for a given individual.
## Aims and Hypothesis
This study examined the feasibility of the first part of this approach, that is, whether subsets of items could be found with good prediction accuracies for binge eating.
Furthermore, 2 studies were conducted to establish a conceptual and empirical foundation for JITAIs on binge eating. In study 1, we collected a comprehensive set of binge-eating antecedents in the form of EMA items. We combined a literature review with qualitative and quantitative interviews (focus group with 11 inpatients) following Soyster and Fisher [29]. In study 2, an algorithm was used to select idiographic subsets of binge-eating predictors based on Elleman et al [30], Kaiser et al [10], and Soyster et al [31]. We hypothesized that these idiographic binge-eating antecedents would predict binge eating with high accuracy. This selection and prediction were tested in 13 patients with BN or BED.
## Ethics Approval
All participants signed an informed consent form (stating which data were stored, where and for how long, who the investigator was, and the purpose of the study) approved by the ethics committee of the University of Salzburg (EK-GZ: $\frac{37}{2018}$).
## Literature Research
A PhD-level researcher systematically searched Google Scholar, PsycINFO, and PubMed databases for articles with the word “binge” in their title and the terms “ecological momentary” or “experience sampling” to find risk state descriptors with relevance to binge eating in the literature. The search resulted in 509 articles that were deduplicated and scanned for relevance. Only empirical articles reporting the results of EMA studies on binge eating were retained. A total of 262 articles were subsequently analyzed (see Multimedia Appendix 1, Figure S1 for an attrition diagram).
## Text Analysis Using Word Embedding
Abstracts of all articles in the literature were retrieved. The R package text2vec [32] was used to perform global vector word-embedding analysis on these abstracts. Word embedding is an “unsupervised” learning algorithm that maps words to a vector space based on their similarity. It is unsupervised as no labeling of training data is needed because training is performed on aggregated global word-word cooccurrence statistics from a corpus. A matrix is calculated where each element Xij represents how often wordi appears in the context of wordj (ie, in the same sentence). Thus, words can be represented numerically and their similarities can be compared [32].
The following parameters were set for training the word vectors (vector dimensions=100, window size=15, and minimum word count to be included in the model=5). The English stop words were removed. Single words (eg, “sadness”), as well as combinations of 2 words (eg, “negative affect”), were allowed in the model. The cosine similarity between word vectors was used to quantify the similarity between word embeddings. This metric computes the angle between 2 vectors to quantify the similarity in the vector space they inhabit. The interpretation of cosine similarity resembles that of the correlation coefficients. Perfectly similar word vectors have a cosine similarity of 1, whereas perfectly dissimilar vectors have a similarity of −1. We calculated the cosine similarity of all retained words with the words “binge” or “binges” retaining only words that had at least a cosine similarity of +.10 or −.10 (resembling a small effect according to the criteria of Cohen [33] for the interpretation of correlation coefficients). In this way, we intended to find words that were conceptually similar to “binge eating” while covering a wide range of binge-eating antecedents.
## Integration Into a Preliminary Item List
In the next step, 2 authors independently rated whether a given retained word was quantifiable with a psychometric item (ie, the words “dissociation” or “dissociative” were rated as quantifiable with the item “I feel detached from myself.” [ 0=not at all to very much=100]) and in terms of usefulness for an EMA survey. Items were only retained if they were rated as quantifiable and useful by both authors. Overlapping constructs were organized into categories to reduce redundancy. Finally, a preliminary list of 47 items was compiled from the empirical and theoretical constructs and complemented by constructs derived from previous EMA studies (Multimedia Appendix 2, Table S1).
## Patient Focus Group
A focus group of inpatients (11 female adolescents and young adults in treatment for regular binge-eating episodes at the Schoen Clinic Roseneck, Germany) complemented this literature-based approach. It was conducted to tap into antecedents that nomothetic EMA research might have overlooked so far. After an individual written brainstorming session on “triggers and circumstances associated with binge eating,” the inpatients rated the preliminary list of EMA items on relevance to their binge-eating episodes (“happens before/during/after binge eating…”: 1=[almost] never, 3=might or might not, 5=[almost] always). A moderated discussion of the brainstormed and provided items concluded the sessions.
Next, 2 researchers analyzed the rating data and integrated patient-generated items. This led to the following changes: several constructs missing in the preliminary item list were identified and items were added to cover these gaps (eg, eating based on internal opposed to external motivation: “Did you eat on your own accord?”; ( not) following a regular meal structure: “How much did you follow a regular meal structure today?”; and restricting specific foods: “Are you restricting on certain foods right now?”).
The focus group participants further rated 27 of the provided items as positively associated with their binge-eating episodes (mean >3.5), 11 items as negatively associated (mean <2.5), and 9 items as unrelated to their binge-eating episodes (mean 2.5-3.5; Multimedia Appendix 2, Table S1). Some items were scored as unrelated (eg, “Right now I feel: tired” and “I engaged in increased levels of sport.”), and items with large SDs (SD >1.00; eg, “Right now I feel: relived,” “Right now I am shopping for groceries.” and “I acted upon my plans regarding my eating behavior.”) were disregarded, merged (eg, “I am in company.” with “I am on my own.”), or exchanged (eg, “I feel strained due to...work / university / school; close social network; wider social network; everyday stressors” with “Do you feel like you can handle all upcoming tasks and problems?”). As the patients expressed concerns over the redundancy of emotional states, 4 more items were disregarded (“Right now I feel: calm/ashamed/guilty/frustrated”). Finally, 4 items regarding eating behaviors such as “resistance to food craving” or “restriction” were rephrased to map more accurately on constructs introduced by the focus group (see Multimedia Appendix 3, Figure S1 for all item iterations)
## Feedback of Clinicians
Finally, clinicians with experience in ED treatment ($$n = 4$$) provided feedback on the gaps in the included constructs. This feedback was integrated by adding concepts such as accessibility to tasty food, day structure (ie, regular sleep and eating patterns), self-regulation intentions, and eating alone. This feedback further led us to include the autoregressive effect of binge-eating episodes on subsequent binge-eating risk in our models [34].
## First Pilot
The EMA items were then piloted by 2 authors and 1 female patient with BN (consistent with the Diagnostic Statistical Manual-5 [DSM-5] [1]) to evaluate content, coverage, wording, and participant burden. Piloting revealed that some items needed further changes to map more accurately on the intended constructs: 1 item about adaptive coping strategies was added (“How much did you try to distract yourself from a possible urge to overeat by healthy strategies [e.g., relaxation, social activity, mindfulness, etc.]?”) to complement the items on dysfunctional coping and distraction strategies, which were merged into one item (“how much did you try to distract yourself from a possible urge to overeat by unhealthy strategies [eg, alcohol, cigarettes, drugs, self-harm, etc]?”). Two items were rephrased, and 1 item assessing food craving was split up and rephrased to differentiate food craving, overeating, and objective binge-eating episodes (food craving: “how strong is your craving for certain foods right now?”; overeating: “how strong is your urge to overeat right now?”; and binge-eating episodes: “how high would you rate your risk for a binge-eating episode right now?”).
The highly compliant participant with BN (all 84 EMA signals answered) reported that the participant burden was too high. Thus, 6 more items were disregarded to shorten the extensive list of items assessing different forms of self-licensing [35,36] and restrictions. Finally, the authors integrated the information gathered in the previous steps (ie, literature review, feedback of the focus group, feedback from clinicians, and feedback of the pilot patient) to make final iterations to the EMA-item set (see Multimedia Appendix 3, Figure S1 for all item iterations, and Multimedia Appendix 4, Tables S1 and S2 for the final EMA-item set).
## Participants
Female patients with current BN ($$n = 12$$) or BED ($$n = 1$$) were recruited via mail from the waiting list for inpatient treatment of the Schoen Clinic Roseneck, Germany ($$n = 10$$), and from web-based forums on eating disorders and psychology ($$n = 3$$; see Multimedia Appendix 5, Figure S1 for a CONSORT [Consolidated Standards of Reporting Trials] flowchart). This study was advertised as a pilot study for a smartphone-based binge-eating intervention. The data were collected between April 2020 and April 2021.
## Procedure
All participants completed the following study protocol. First, the BN and BED research diagnoses according to DSM-5 [1] were determined via telephone using the Eating Disorder Examination interview [37] and the Structured Clinical Interview for DSM-IV [38]. Both interviews were adapted to the diagnostic criteria of the DSM-5 (eg, 1 binge-eating episode per week for 3 months instead of 2 binges per week for 3 months).
The participants were then introduced to the EMA items and logged into the customized smartphone app SmartEater. SmartEater was used during the subsequent EMA phase, in which signal-based EMA questionnaires were inquired up to 84 times per participant (6 signal-contingent prompts per day, in intervals of 2.5 hours for 2 weeks; questionnaires expired 1 hour after the initial prompt). In addition, an event-contingent EMA questionnaire on overeating, loss of control, and binge-eating episodes was accessible. Participants were instructed to fill in this event-contingent questionnaire whenever they felt like they overate or felt a sense of loss of control over food intake or both. The event-contingent questionnaire included questions to differentiate between subjective and objective binge eating and objective overeating (Multimedia Appendix 4, Table S2). EMA items assessing emotions were presented in a randomized order. However, the other items were presented in a fixed order to prevent carryover effects. The participants were able to review and change their answers through a “back” button. Answering all items (except branched items) was mandatory for submission of the questionnaires.
After the EMA phase of 2 weeks, a JITAI phase of 2 weeks started, in which the participants received short intervention suggestions from the app to prevent binge-eating episodes at ideographically predicted high-risk times. Every study stage was accompanied by web-based questionnaires that assessed current eating behavior pathology, demographic data, perceived acceptability, feasibility, and so on. Data from the intervention phase were not covered in the present article. For reimbursement, the participants received €30 (US $32.80) and personalized feedback on their EMA data and psychometric web-based questionnaires.
## Data Preparation and Measures
To avoid the violation of the assumption of equally spaced time series [39], empty rows were inserted in the data set after every last signal for a given day. This prevented the prediction algorithm from regressing on data from the previous day.
## Binge-Eating Episodes–Criterion
Objective binge-eating episodes, characterized by [1] “feelings of loss of control over eating behavior” and [2] “consumption of objectively large, inappropriate amounts of food” [1,37,40], were identified from eating episodes reported over the signal-based (1 item: “Was your meal a main meal, snack, or binge?”) and event-based EMA questionnaires (2 items: “Would other people rate the amount of food as excessive under similar circumstances?” and “Did you feel like you are losing control of your eating behavior?”). The signal-based and the 2 event-based items were recoded into a binary variable indicating the occurrence of an objective binge-eating episode (binge-eating episode reported=1, no binge-eating episode reported=0). As the algorithm was supposed to predict future binge-eating episodes, this variable was shifted backward in time by one signal (approximately 2.5 hour).
## EMA Predictors
An unshifted version of the binge-eating variable was included as a possible predictor of the autoregressive effects of binge eating. Furthermore, additional EMA items ($$n = 31$$) were used to model possible binge-eating antecedents. Thus, only items that were assessed with every signal-based questionnaire were included (aside from the binge-eating classifier), as each variable needed to have a sufficient percentage of data points within a person (see Multimedia Appendix 4, Table S1 for the wording of each item).
## Time Predictors
Time variables, especially in the form of circles and distinct times of day, have been shown to be highly predictive in everyday life [41]. EMA studies have even found peak times for certain binge-eating antecedents (ie, food cravings or hunger; [42]) and binge eating itself [43-45]. Thus, as temporal data are passively collected in the EMA setting via timestamps, without additional participant input, we decided to include different temporal predictors that could detect a single high-risk time per day (24-hour oscillation) or several times per day (sub-24 hour oscillation).
Variables representing 8-, 12- and 24-hour sinusoidal and cosinusoidal cycles were computed based on the cumulative sum of time differences between assessments (eg, 10:30 AM-8 AM, 1 PM-10:30 AM=2.5, 5, 7.5...). For example, a 24-hour sinusoid cycle was calculated using the following formula: sin24h = sin(2π: 24 * Δt), where ∆t is the difference between assessment points in hours (here: 2.5). Finally, dummy-coded variables representing the time of day were calculated for each signal (morning, late morning, early afternoon, afternoon, evening, and late evening). This allows for identifying a daytime when binge eating is particularly likely for a given participant (eg, when returning from work) that could not be well captured by the cyclical predictors.
## Application of the Best Items Scale that is Cross-validated, Unit-weighted, Informative, and Transparent Algorithm to EMA Data (for Idiographic Predictor Selection and Prediction of Binge Eating)
The machine learning algorithm Best Item Scales that are Cross-validated, Unit-weighted, Informative and Transparent (BISCUIT) [30] of the bestScales function from the R package psych [46] was applied separately to the EMA data of each patient to select the best idiographic predictors of binge-eating episodes. This method was chosen because of its [1] robustness to missing data; [2] use of unit-weighted scoring of predictors, which was found to be more generalizable, especially in the context of prediction; and [3] tendency to select more parsimonious predictor sets compared with other approaches such as Elastic Net regression [30,47,48]. BISCUIT is a simple algorithm that correlates a set of predictors (here all EMA and time variables) with a criterion (here, the binary time-shifted binge-eating variable at t+1) and retains the predictors with the highest correlation to form a unit-weighted scale [10,30,46]. This scale was then used to estimate the out-of-sample predictive performance using 10-fold cross-validation. The average correlation of the scale with the criterion across 10 cross-validation splits was then computed, and the set of items with the highest cross-validated correlation was retained [30,46]. The output of BISCUIT is the selection of items showing maximum predictive validity, as the cutoff values that lead to the highest combination of sensitivity and specificity are retained [30,46].
Thus, multiple Rs (pairwise Pearson correlations) of all predictors with time-shifted binge-eating episodes (at t+1) were calculated for each participant separately to select the idiographic predictor sets. Furthermore, the area under the curve (AUC) with a bootstrapped $95\%$ CI, specificity, sensitivity, and within- and out-of-sample reliability were calculated as prediction accuracy measures of the idiographic predictor sets and their prediction of binge eating in the next 2.5 hours (t+1).
## Study 1—EMA-Item Set
The final signal-contingent EMA questionnaire included 36 EMA items (momentary emotions, stress, exhaustion, and context; eg, being alone, social interactions, dissociations, eating behavior, resistance to food craving, distraction, and coping), which were designed to be assessed 6 times per day. In addition, an optional event-contingent EMA questionnaire on overeating, loss-of-control eating, and binge eating was self-initialized and included 20 items. See Multimedia Appendix 4, Tables S1 and S2 for all interval- and event-contingent items and their wording. A flowchart of all iterations applied to the EMA-item set can be found in Multimedia Appendix 3, Figure S1).
This study used a mixed methods approach to develop a conceptual and statistical basis for an idiographic JITAI for binge eating. The EMA-item development in study 1 followed a replicable procedure similar to Soyster and Fisher [29] while considering nomothetic theories on binge-eating antecedents (ie, emotional eating) and underwent several qualitative (literature research and focus group brainstorming and discussion) and quantitative (focus group ratings) iterations and piloting.
This resulted in a broad EMA-item set (Multimedia Appendix 4), including several constructs underrepresented in the nomothetic literature (eg, “I feel detached from myself.”, and “specific” restrictions “did you restrict yourself [eg, by eating less, avoiding certain foods]?” [ 12,50,51]). This approach also helped us shorten the extensive lists of emotional states (eg, “right now I feel...calm/relived/ashamed/guilty/frustrated.”) because within-person ratings for similar emotions were often identical, and concerns about redundancy were expressed during the moderated discussion. Furthermore, we did not only incorporate risk factors into the EMA-item set but also protective factors that could potentially decrease the likelihood of binge eating (ie, healthy coping strategies to keep oneself from binge eating or positive emotions [27,28]). The role of protective factors is often overlooked in nomothetic binge-eating theories but is crucial to idiographic binge-eating prediction and intervention models.
## Selection of Idiographic Predictor Subsets
The patients ($$n = 13$$) answered on an average 67.3 out of 84 EMA prompts (SD 13.4; range 43-84; see Multimedia Appendix 6, Table S1 for EMA compliance and occurrences of binge-eating episodes per patient). Across participants, the algorithm selected highly heterogeneous predictor sets of varying sizes (mean 7.31, SD 1.49; range 5-9 predictors) for the prediction of binge-eating episodes.
Figure 1 shows the idiographic predictor selection that showed maximum predictive validity for each participant. Thus, the predictors (at t) with the highest multiple Rs (pairwise Pearson correlations) with time-shifted binge-eating episodes (at t+1) were selected. All listed items were selected as idiographic predictors of binge eating, independent of their significance. However, we additionally calculated the significance of the correlations for the context. The exact P values, codes, and data can be found in the corresponding project in the Open Science Framework [49]. Note that the results might vary slightly, as the R function set.seed does not apply to the cross tables.
**Figure 1:** *Idiographic predictor subsets for binge eating with Pairwise Pearson Correlations (Multiple Rs) of each selected predictor of binge eating in the next 2.5 hours (t+1). *, **, and *** indicate that the correlations are significant at a level of .05, .01, and .001, respectively; 2-tailed. EMA: ecologic momentary assessment.*
## Prediction of Binge Eating by Idiographic Predictor Subsets
The selection of idiographic predictor sets resulted in good average prediction accuracy (mean AUC 0.80, SD 0.15; mean $95\%$ CI 0.63-0.95; mean specificity 0.87, SD 0.08; mean sensitivity 0.79, SD 0.19; mean maximum reliability of rD 0.40, SD 0.13; mean rCV 0.13, SD 0.31). The mean AUC of 0.80 indicates that there is on average an $80\%$ chance that the idiographic models predict binge and nonbinge episodes accurately. The mean specificity of 0.87 indicates that the idiographic models mistakenly classified 13 of 100 episodes as binge-eating episodes. The mean sensitivity of 0.79 indicates that the idiographic models mistakenly classified 21 out of 100 binge-eating episodes as nonbinge episodes. Table 1 shows the prediction accuracies of the idiographic predictor subsets for binge-eating episodes per participant. R code and data are available from the Open Science Framework [49].
**Table 1**
| Unnamed: 0 | Unnamed: 1 | Participants | Participants.1 | Participants.2 | Participants.3 | Participants.4 | Participants.5 | Participants.6 | Participants.7 | Participants.8 | Participants.9 | Participants.10 | Participants.11 | Participants.12 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| | | 01 | 02 | 03 | 04 | 05 | 06 | 07 | 08 | 09 | 10 | 11 | 12 | 13 |
| Model fit indices | Model fit indices | Model fit indices | Model fit indices | Model fit indices | Model fit indices | Model fit indices | Model fit indices | Model fit indices | Model fit indices | Model fit indices | Model fit indices | Model fit indices | Model fit indices | Model fit indices |
| | AUCa (95% CI)b | 0.92 (0.75-1.00) | 0.97 (0.92-1.00) | 0.84 (0.70-0.98) | 0.51 (0.23-0.80) | 0.73 (0.45-1.00) | 0.93 (0.77-1.00) | 0.93 (0.83-1.00) | 0.85 (0.75-0.95) | 0.63 (0.41-0.85) | 0.75 (0.58-0.93) | 0.72 (0.53-0.92) | 0.60 (0.29-0.90) | 0.98 (0.94-1.00) |
| | Specificity | 0.84 | 0.89 | 0.74 | 1.00 | 0.86 | 0.85 | 0.90 | 0.81 | 1.00 | 0.74 | 0.87 | 0.91 | 0.96 |
| | Sensitivity | 1.00 | 1.00 | .86 | .45 | .67 | 1.00 | 1.00 | .80 | .56 | .73 | .69 | .56 | 1.00 |
| Derivation step (within-sample performance) | Derivation step (within-sample performance) | Derivation step (within-sample performance) | Derivation step (within-sample performance) | Derivation step (within-sample performance) | Derivation step (within-sample performance) | Derivation step (within-sample performance) | Derivation step (within-sample performance) | Derivation step (within-sample performance) | Derivation step (within-sample performance) | Derivation step (within-sample performance) | Derivation step (within-sample performance) | Derivation step (within-sample performance) | Derivation step (within-sample performance) | Derivation step (within-sample performance) |
| | rDc,d (SD) | 0.48 (0.04) | 0.42 (0.05) | 0.53 (0.04) | 0.33 (0.05) | 0.34 (0.03) | 0.47 (0.06) | 0.53 (0.06) | 0.51 (0.10) | 0.10 (0.22) | 0.46 (0.06) | 0.50 (0.06) | 0.18 (0.10) | 0.32 (0.23) |
| Validation step (out of sample performance) | Validation step (out of sample performance) | Validation step (out of sample performance) | Validation step (out of sample performance) | Validation step (out of sample performance) | Validation step (out of sample performance) | Validation step (out of sample performance) | Validation step (out of sample performance) | Validation step (out of sample performance) | Validation step (out of sample performance) | Validation step (out of sample performance) | Validation step (out of sample performance) | Validation step (out of sample performance) | Validation step (out of sample performance) | Validation step (out of sample performance) |
| | rCVd,e (SD) | 0.41 (0.34) | 0.10 (0.56) | –0.02 (0.72) | 0.36 (0.38) | 0.38 (0.62) | –0.36 (0.28) | 0.29 (0.30) | 0.54 (0.42) | –0.11 (0.46) | 0.34 (0.27) | 0.27 (0.64) | 0.08 (0.78) | –0.56 (0.57) |
## Study 2—Prediction Based on Idiographic Predictor Subsets
Regularly completing extensive EMA-item sets (such as the present one with 36 interval-contingent and 20 event-contingent items) becomes increasingly burdensome over prolonged study periods. Thus, we applied a machine learning algorithm to the EMA data of patients with BN and BED to select parsimonious idiographic subsets of EMA items. This data-driven selection optimizes the predictive power within participants and decreases potential researcher bias.
The idiographic item subsets predicted binge-eating episodes with a high average accuracy (mean AUC 0.80) across 13 patients. Notably, the sensitivity approached $100\%$ (successful prediction of every reported binge) in several patients, without forfeiting much specificity (predicting no binge when none occurred). This is noteworthy as outcome frequency was not extremely high (mean 10.4, SD 7.4; range 2-28 binge-eating episodes; see also Multimedia Appendix 6, Table S1 “Number of binge-eating episodes and total data points per participant”).
## Secondary Findings
Regarding the composition of the selected item sets, a high selection rate of items with high proximity to the binge-eating construct was evident (ie, hunger, food craving, urge to overeat, subjective binge-eating risk, and preceding binge-eating episodes). This suggests that some patients may accurately predict upcoming binge-eating episodes. This reveals a relatively high level of insight into the temporal evolution of the symptoms in some patients. Surprisingly, hunger and food craving were negatively correlated with binge eating in 3 patients. One could speculate that the negative correlations between hunger and binge eating in patients 06 and 11 point to disinhibition, that is, because of the temporary abandonment of rigid diet rules after eating in the absence of hunger [52,53].
In addition to items with conceptual similarity to binge eating, emotional items were selected in 9 patients. This supports the relevance of emotional eating in binge-eating predictions [4,11,19]. However, the selected emotion sets were highly heterogeneous across the 9 patients. In fact, no single emotion item (or specific set of emotion items) was consistently selected across all patients. This speaks against a singular and generalizable emotional eating theory of binge eating. Similarly, because no other nonemotion–related predictor was consistently selected across all patients, our pilot data provide no evidence for other generalizable nomothetic theories of binge eating. Thus, several nomothetic theories are needed to explain present heterogeneity, which may in turn explain the multitude of competing nomothetic binge-eating theories. Clearly, nomothetic theories must model individual differences more explicitly to account for these findings. These findings also support the use of a broad EMA-item set that covers a large range of possible binge-eating antecedents in the context of idiographic prediction [4-6,14-21].
Interestingly, time-derived predictors were selected only in 7 patients. In 6 of these patients, discrete time predictors were chosen that were consistent with the literature on the timing of binge-eating peaks (ie, afternoon to late evening) [43-45]. Time cycles were only selected in 3 patients. This is surprising given the observation of cyclic symptoms (eg, in depression [41]). However, time-based predictors may be more powerful if EMA items with conceptual similarity to binge eating are omitted. In the case of binge eating, time cycles could represent a rising and falling urge to overeat (eg, due to prolonged restriction between meals) [41,54]. Discrete time variables could represent the time of the day where a patient usually binges (eg, due to contexts such as being alone at home every afternoon) [41]. Assessing such time-derived variables does not require user input and thus does not contribute to the participant burden. This makes them valuable for the predictions in the JITAI framework.
## Limitations and Strengths
Compared with the high average within-sample performance (mean rD 0.40), the average out-of-sample performance (mean rCV 0.13) was lower, suggesting limited out-of-sample generalizability. This might be because of 10-fold cross-validation, which does not account for the serial correlation and potential nonstationarity of time-series data [55]. Future studies could resort to alternative time-series–specific techniques (ie, roll forward cross-validation and out-of-sample evaluation) that ensure that training data always precede test data. However, X-fold cross-validation has been shown to outperform these techniques [55]. Furthermore, the number of observations was limited (max 84 per participant), leading to relatively small splits in the 10-fold cross-validation. Thus, there was a high possibility of randomly drawn training sets that were unrepresentative of the data set. The results from the validation step might also vary slightly, as the R function set.seed does not apply to the cross tables.
*Another* general drawback of the BISCUIT algorithm is that nonlinear trends and interaction effects among predictors are not considered. In addition, when applied under optimal conditions (ie, big data sets and no missing data), gold standard machine learning approaches, such as random forests [56] and XGBoost [57] in combination with super learners [58], calibrate better to the data. However, for typical EMA data sets, the conditions are rarely optimal for these algorithms. Missing data and a limited number of observations are typical features of high-burden EMA sampling schemes. However, BISCUIT was created to handle these problematic properties. BISCUIT outperformed random forest and elastic net approaches in other studies with smaller idiographic data sets and more missing data (Beck et al [59]: mean 57.4, SD 16.3; range 40-109 EMA observations; present data: mean 67.3, SD 13.4; range 43-84 EMA observations).
Finally, the idiographic approach used in this study precludes mechanistic and theoretical inferences about binge eating. Generally, machine learning algorithms are silent about the underlying mechanisms; instead, they tailor models are as close as possible to the given data and conditions. Thus, the present results are highly specific, for example, to the used “prediction interval” of 2.5 hours between predictors and outcome. This could be problematic as it has been shown that emotions and eating can influence each other at different time intervals [60].
## Implications and Future Directions
In addition to emphasizing the importance of a broad predictor set, the results have a direct implication for the JITAI and EMA methodology: participant burden in longer EMA sampling periods precludes the use of large EMA-item sets. Thus, such EMA studies might prune their large EMA-item sets after a “calibration period” by applying the described predictor-selection approach. Therefore, the participant burden is reduced, whereas accurate idiographic binge-eating predictions are retained. Such predictions can then be used to trigger JITAIs, as done by Forman et al [24,61] in a JITAI on dietary lapses.
Future studies may transfer the present work to a range of disordered and maladaptive eating behaviors (eg, purging behaviors or food restrictions) to develop low-threshold JITAIs. EMA-item–based prediction should be compared with predictions generated from passive data sources (ie, smartphone sensors, use data, and wearable data) that do not inflict much user burden [10,62-64]. In the long term, acceptance, dropout rates, and effectiveness of JITAI protocols on binge eating need to be tested in microrandomized trials [65] and classic randomized controlled trials against nonadaptive, non–real-time interventions before the ultimate recommendation as the gold standard.
Finally, feeding back personal binge-eating predictors can serve as a psychoeducational intervention and raise awareness of personal risk and protective factors. Such personal binge-eating predictors can also inform conventional face-to-face psychotherapy. Patients with a clear dominance of emotion-related predictors might profit from emotion-focused interventions [66] more than patients with a dominance of impulsive or craving-related predictors, who might profit more from impulse control intervention [67].
## References
1. 1American Psychiatric AssociationDiagnostisches und Statistisches Manual Psychischer Störungen DSM-5®2015Göttingen, GermanyHogrefe. *Diagnostisches und Statistisches Manual Psychischer Störungen DSM-5®* (2015)
2. Fairburn CG, Bailey-Straebler S, Basden S, Doll HA, Jones R, Murphy R, O'Connor ME, Cooper Z. **A transdiagnostic comparison of enhanced cognitive behaviour therapy (CBT-E) and interpersonal psychotherapy in the treatment of eating disorders**. *Behav Res Ther* (2015) **70** 64-71. DOI: 10.1016/j.brat.2015.04.010
3. Helverskov JL, Clausen L, Mors O, Frydenberg M, Thomsen PH, Rokkedal K. **Trans-diagnostic outcome of eating disorders: a 30-month follow-up study of 629 patients**. *Eur Eat Disord Rev* (2010) **18** 453-63. DOI: 10.1002/erv.1025
4. Cardi V, Leppanen J, Treasure J. **The effects of negative and positive mood induction on eating behaviour: a meta-analysis of laboratory studies in the healthy population and eating and weight disorders**. *Neurosci Biobehav Rev* (2015) **57** 299-309. DOI: 10.1016/j.neubiorev.2015.08.011
5. Haedt-Matt AA, Keel PK. **Revisiting the affect regulation model of binge eating: a meta-analysis of studies using ecological momentary assessment**. *Psychol Bull* (2011) **137** 660-81. DOI: 10.1037/a0023660
6. Zunker C, Peterson CB, Crosby RD, Cao L, Engel SG, Mitchell JE, Wonderlich SA. **Ecological momentary assessment of bulimia nervosa: does dietary restriction predict binge eating?**. *Behav Res Ther* (2011) **49** 714-7. DOI: 10.1016/j.brat.2011.06.006
7. Adolf J, Schuurman NK, Borkenau P, Borsboom D, Dolan CV. **Measurement invariance within and between individuals: a distinct problem in testing the equivalence of intra- and inter-individual model structures**. *Front Psychol* (2014) **5** 883. DOI: 10.3389/fpsyg.2014.00883
8. Beltz AM, Wright AG, Sprague BN, Molenaar PC. **Bridging the nomothetic and idiographic approaches to the analysis of clinical data**. *Assessment* (2016) **23** 447-58. DOI: 10.1177/1073191116648209
9. Fisher AJ, Medaglia JD, Jeronimus BF. **Lack of group-to-individual generalizability is a threat to human subjects research**. *Proc Natl Acad Sci U S A* (2018) **115** E6106-15. DOI: 10.1073/pnas.1711978115
10. Kaiser T, Butter B, Arzt S, Pannicke B, Reichenberger J, Ginzinger S, Blechert J. **Time-lagged prediction of food craving with qualitative distinct predictor types: an application of BISCWIT**. *Front Digit Health* (2021) **3** 694233. DOI: 10.3389/fdgth.2021.694233
11. Reichenberger J, Schnepper R, Arend AK, Richard A, Voderholzer U, Naab S, Blechert J. **Emotional eating across different eating disorders and the role of body mass, restriction, and binge eating**. *Int J Eat Disord* (2021) **54** 773-84. DOI: 10.1002/eat.23477
12. Engelberg MJ, Steiger H, Gauvin L, Wonderlich SA. **Binge antecedents in bulimic syndromes: an examination of dissociation and negative affect**. *Int J Eat Disord* (2007) **40** 531-6. DOI: 10.1002/eat.20399
13. Hilbert A, Tuschen-Caffier B. **Maintenance of binge eating through negative mood: a naturalistic comparison of binge eating disorder and bulimia nervosa**. *Int J Eat Disord* (2007) **40** 521-30. DOI: 10.1002/eat.20401
14. McShane JM, Zirkel S. **Dissociation in the binge-purge cycle of bulimia nervosa**. *J Trauma Dissociation* (2008) **9** 463-79. DOI: 10.1080/15299730802225680
15. Evers C, Dingemans A, Junghans AF, Boevé A. **Feeling bad or feeling good, does emotion affect your consumption of food? A meta-analysis of the experimental evidence**. *Neurosci Biobehav Rev* (2018) **92** 195-208. DOI: 10.1016/j.neubiorev.2018.05.028
16. Schag K, Schönleber J, Teufel M, Zipfel S, Giel KE. **Food-related impulsivity in obesity and binge eating disorder--a systematic review**. *Obes Rev* (2013) **14** 477-95. DOI: 10.1111/obr.12017
17. Gearhardt AN, White MA, Potenza MN. **Binge eating disorder and food addiction**. *Curr Drug Abuse Rev* (2011) **4** 201-7. DOI: 10.2174/1874473711104030201
18. Loth KA, Goldschmidt AB, Wonderlich SA, Lavender JM, Neumark-Sztainer D, Vohs KD. **Could the resource depletion model of self-control help the field to better understand momentary processes that lead to binge eating?**. *Int J Eat Disord* (2016) **49** 998-1001. DOI: 10.1002/eat.22641
19. Dingemans A, Danner U, Parks M. **Emotion regulation in binge eating disorder: a review**. *Nutrients* (2017) **9** 1274. DOI: 10.3390/nu9111274
20. Jansen A. **A learning model of binge eating: cue reactivity and cue exposure**. *Behav Res Ther* (1998) **36** 257-72. DOI: 10.1016/s0005-7967(98)00055-2
21. Bongers P, Jansen A. **Emotional eating and Pavlovian learning: evidence for conditioned appetitive responding to negative emotional states**. *Cogn Emot* (2017) **31** 284-97. DOI: 10.1080/02699931.2015.1108903
22. Smith KE, Juarascio A. **From Ecological Momentary Assessment (EMA) to Ecological Momentary Intervention (EMI): past and future directions for ambulatory assessment and interventions in eating disorders**. *Curr Psychiatry Rep* (2019) **21** 53. DOI: 10.1007/s11920-019-1046-8
23. Juarascio AS, Parker MN, Lagacey MA, Godfrey KM. **Just-in-time adaptive interventions: a novel approach for enhancing skill utilization and acquisition in cognitive behavioral therapy for eating disorders**. *Int J Eat Disord* (2018) **51** 826-30. DOI: 10.1002/eat.22924
24. Forman EM, Goldstein SP, Crochiere RJ, Butryn ML, Juarascio AS, Zhang F, Foster GD. **Randomized controlled trial of OnTrack, a just-in-time adaptive intervention designed to enhance weight loss**. *Transl Behav Med* (2019) **9** 989-1001. DOI: 10.1093/tbm/ibz137
25. Spanakis G, Weiss G, Boh B, Lemmens L, Roefs A. **Machine learning techniques in eating behavior e-coaching**. *Pers Ubiquit Comput* (2017) **21** 645-59. DOI: 10.1007/s00779-017-1022-4
26. Levinson CA, Vanzhula I, Brosof LC. **Longitudinal and personalized networks of eating disorder cognitions and behaviors: targets for precision intervention a proof of concept study**. *Int J Eat Disord* (2018) **51** 1233-43. DOI: 10.1002/eat.22952
27. Cardi V, Leppanen J, Leslie M, Esposito M, Treasure J. **The use of a positive mood induction video-clip to target eating behaviour in people with bulimia nervosa or binge eating disorder: an experimental study**. *Appetite* (2019) **133** 400-4. DOI: 10.1016/j.appet.2018.12.001
28. Kelly NR, Lydecker JA, Mazzeo SE. **Positive cognitive coping strategies and binge eating in college women**. *Eat Behav* (2012) **13** 289-92. DOI: 10.1016/j.eatbeh.2012.03.012
29. Soyster PD, Fisher AJ. **Involving stakeholders in the design of ecological momentary assessment research: an example from smoking cessation**. *PLoS One* (2019) **14** e0217150. DOI: 10.1371/journal.pone.0217150
30. Elleman LG, McDougald SK, Condon DM, Revelle W. **That takes the BISCUIT: predictive accuracy and parsimony of four statistical learning techniques in personality data, with data missingness conditions**. *Eur J Psychol Assess* (2020) **36** 948-58. DOI: 10.1027/1015-5759/a000590
31. Soyster PD, Ashlock L, Fisher AJ. **Pooled and person-specific machine learning models for predicting future alcohol consumption, craving, and wanting to drink: a demonstration of parallel utility**. *Psychol Addict Behav* (2022) **36** 296-306. DOI: 10.1037/adb0000666
32. Selivanov D, Bickel M, Wang Q. **text2vec: Modern Text Mining Framework for R**. *The Comprehensive R Archive Network* (2022)
33. Cohen J. *Statistical Power Analysis for the Behavioral Sciences. 2nd edition* (1988)
34. Lavender JM, Utzinger LM, Cao L, Wonderlich SA, Engel SG, Mitchell JE, Crosby RD. **Reciprocal associations between negative affect, binge eating, and purging in the natural environment in women with bulimia nervosa**. *J Abnorm Psychol* (2016) **125** 381-6. DOI: 10.1037/abn0000135
35. Prinsen S, Evers C, Wijngaards L, van Vliet R, de Ridder D. **Does self-licensing benefit self-regulation over time? An ecological momentary assessment study of food temptations**. *Pers Soc Psychol Bull* (2018) **44** 914-27. DOI: 10.1177/0146167218754509
36. Witt Huberts JC, Evers C, De Ridder DT. **License to sin: self-licensing as a mechanism underlying hedonic consumption**. *Eur J Soc Psychol* (2012) **42** 490-6. DOI: 10.1002/ejsp.861
37. Hilbert A, Tuschen-Caffier B, Ohms M. **Eating disorder examination: deutschsprachige version des strukturierten essstörungsinterviews**. *Diagnostica* (2004) **50** 98-106. DOI: 10.1026/0012-1924.50.2.98
38. Wittchen HU, Zaudig M, Fydrich T. *SKID. Strukturiertes Klinisches Interview für DSM-IV. Achse I und II* (1997)
39. Jebb AT, Tay L, Wang W, Huang Q. **Time series analysis for psychological research: examining and forecasting change**. *Front Psychol* (2015) **6** 727. DOI: 10.3389/fpsyg.2015.00727
40. **International Statistical Classification of Diseases and Related Health Problems (11th revision)**. *World Health Organization* (2019)
41. Fisher AJ, Bosley HG. **Identifying the presence and timing of discrete mood states prior to therapy**. *Behav Res Ther* (2020) **128** 103596. DOI: 10.1016/j.brat.2020.103596
42. Reichenberger J, Richard A, Smyth JM, Fischer D, Pollatos O, Blechert J. **It's craving time: time of day effects on momentary hunger and food craving in daily life**. *Nutrition* (2018) **55-56** 15-20. DOI: 10.1016/j.nut.2018.03.048
43. Smyth JM, Wonderlich SA, Sliwinski MJ, Crosby RD, Engel SG, Mitchell JE, Calogero RM. **Ecological momentary assessment of affect, stress, and binge-purge behaviors: day of week and time of day effects in the natural environment**. *Int J Eat Disord* (2009) **42** 429-36. DOI: 10.1002/eat.20623
44. Stein RI, Kenardy J, Wiseman CV, Dounchis JZ, Arnow BA, Wilfley DE. **What's driving the binge in binge eating disorder?: a prospective examination of precursors and consequences**. *Int J Eat Disord* (2007) **40** 195-203. DOI: 10.1002/eat.20352
45. Schreiber-Gregory DN, Lavender JM, Engel SG, Wonderlich SA, Crosby RD, Peterson CB, Simonich H, Crow S, Durkin N, Mitchell JE. **Examining duration of binge eating episodes in binge eating disorder**. *Int J Eat Disord* (2013) **46** 810-4. DOI: 10.1002/eat.22164
46. Revelle W. **psych: Procedures for Psychological, Psychometric, and Personality Research. version 2.2.5**. *The Comprehensive R Archive Network* (2022)
47. Dana J, Dawes RM. **The superiority of simple alternatives to regression for social science predictions**. *J Educ Behav Stat* (2016) **29** 317-31. DOI: 10.3102/10769986029003317
48. Lichtenberg JM, Şimşek Ö. **Simple regression models**. *Proceedings of the NIPS 2016 Workshop on Imperfect Decision Makers* (2016) 13-25
49. Arend AK, Kaiser T, Pannicke B, Reichenberger J, Naab S, Voderholzer U, Blechert J. **Code and Data for 'Toward Individualized Prediction of Binge-Eating Episodes Based on Ecological Momentary Assessment Data: Item Development and Pilot Study in Patients with Bulimia Nervosa and Binge-Eating Disorder'**. *Open Science Framework* (2022)
50. Dakanalis A, Carrà G, Calogero R, Fida R, Clerici M, Zanetti MA, Riva G. **The developmental effects of media-ideal internalization and self-objectification processes on adolescents' negative body-feelings, dietary restraint, and binge eating**. *Eur Child Adolesc Psychiatry* (2015) **24** 997-1010. DOI: 10.1007/s00787-014-0649-1
51. Guertin TL, Conger AJ. **Mood and forbidden foods' influence on perceptions of binge eating**. *Addict Behav* (1999) **24** 175-93. DOI: 10.1016/s0306-4603(98)00049-5
52. Heatherton TF, Polivy J, Herman CP. **Dietary restraint: some current findings and speculations**. *Psychol Addict Behav* (1990) **4** 100-6. DOI: 10.1037/h0080580
53. Polivy J, Herman CP. **Dieting and binging. A causal analysis**. *Am Psychol* (1985) **40** 193-201. DOI: 10.1037//0003-066x.40.2.193
54. Waterhouse J, Fukuda Y, Morita T. **Daily rhythms of the sleep-wake cycle**. *J Physiol Anthropol* (2012) **31** 5. DOI: 10.1186/1880-6805-31-5
55. Bergmeir C, Hyndman RJ, Koo B. **A note on the validity of cross-validation for evaluating autoregressive time series prediction**. *Comput Stat Data Anal* (2018) **120** 70-83. DOI: 10.1016/j.csda.2017.11.003
56. Breiman L. **Random forests**. *Mach Learn* (2001) **45** 5-32. DOI: 10.1023/A:1010933404324
57. Chen T, Guestrin C. **XGBoost: a scalable tree boosting system**. *Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining* (2016) 785-94. DOI: 10.1145/2939672.2939785
58. van der Laan MJ, Polley EC, Hubbard AE. **Super learner**. *Stat Appl Genet Mol Biol* (2007) **6** Article25. DOI: 10.2202/1544-6115.1309
59. Beck ED, Jackson JJ. **Personalized prediction of behaviors and experiences: an idiographic person-situation test**. *Psychol Sci* (2022) **33** 1767-82. DOI: 10.1177/09567976221093307
60. Pannicke B, Blechert J, Reichenberger J, Kaiser T. **Clustering individuals' temporal patterns of affective states, hunger, and food craving by latent class vector-autoregression**. *Int J Behav Nutr Phys Act* (2022) **19** 57. DOI: 10.1186/s12966-022-01293-1
61. Forman EM, Goldstein SP, Zhang F, Evans BC, Manasse SM, Butryn ML, Juarascio AS, Abichandani P, Martin GJ, Foster GD. **OnTrack: development and feasibility of a smartphone app designed to predict and prevent dietary lapses**. *Transl Behav Med* (2019) **9** 236-45. DOI: 10.1093/tbm/iby016
62. Crochiere RJ, Zhang FZ, Juarascio AS, Goldstein SP, Thomas JG, Forman EM. **Comparing ecological momentary assessment to sensor-based approaches in predicting dietary lapse**. *Transl Behav Med* (2021) **11** 2099-109. DOI: 10.1093/tbm/ibab123
63. Juarascio AS, Crochiere RJ, Tapera TM, Palermo M, Zhang F. **Momentary changes in heart rate variability can detect risk for emotional eating episodes**. *Appetite* (2020) **152** 104698. DOI: 10.1016/j.appet.2020.104698
64. Crochiere RJ, Kerrigan SG, Lampe EW, Manasse SM, Crosby RD, Butryn ML, Forman EM. **Is physical activity a risk or protective factor for subsequent dietary lapses among behavioral weight loss participants?**. *Health Psychol* (2020) **39** 240-4. DOI: 10.1037/hea0000839
65. Walton A, Nahum-Shani I, Crosby L, Klasnja P, Murphy S. **Optimizing digital integrated care via micro-randomized trials**. *Clin Pharmacol Ther* (2018) **104** 53-8. DOI: 10.1002/cpt.1079
66. Clyne C, Blampied NM. **Training in emotion regulation as a treatment for binge eating: a preliminary study**. *Behav Change* (2004) **21** 269-81. DOI: 10.1375/bech.21.4.269.66105
67. Schag K, Rennhak SK, Leehr EJ, Skoda EM, Becker S, Bethge W, Martus P, Zipfel S, Giel KE. **IMPULS: impulsivity-focused group intervention to reduce binge eating episodes in patients with binge eating disorder - a randomised controlled trial**. *Psychother Psychosom* (2019) **88** 141-53. DOI: 10.1159/000499696
|
---
title: 'Evaluation of a Secure Messaging System in the Care of Children With Medical
Complexity: Mixed Methods Study'
journal: JMIR Formative Research
year: 2023
pmcid: PMC9999262
doi: 10.2196/42881
license: CC BY 4.0
---
# Evaluation of a Secure Messaging System in the Care of Children With Medical Complexity: Mixed Methods Study
## Abstract
### Background
The Connecting2gether (C2) platform is a web and mobile–based information-sharing tool that aims to improve care for children with medical complexity and their families. A key feature of C2 is secure messaging, which enables parental caregivers (PCs) to communicate with their child’s care team members (CTMs) in a timely manner.
### Objective
The objectives of this study were to [1] evaluate the use of a secure messaging system, [2] examine and compare the content of messages to email and phone calls, and [3] explore PCs’ and CTMs’ perceptions and experiences using secure messaging as a method of communication.
### Methods
This is a substudy of a larger feasibility evaluation of the C2 platform. PCs of children with medical complexity were recruited from a tertiary-level complex care program to use the C2 platform for 6 months. PCs could invite CTMs involved in their child’s care to register on the platform. Messages were extracted from C2, and phone and email data were extracted from electronic medical records. Quantitative data from the use of C2 were analyzed using descriptive statistics. Messaging content codes were iteratively developed through a review of the C2 messages and phone and email communication. Semistructured interviews were completed with PCs and CTMs. Communication and interview data were analyzed using thematic analysis.
### Results
A total of 36 PCs and 66 CTMs registered on the C2 platform. A total of 1861 messages were sent on C2, with PCs and nurse practitioners sending a median of 30 and 74 messages, respectively. Of all the C2 messages, $85.45\%$ ($\frac{1257}{1471}$) were responded to within 24 hours. Email and phone calls focused primarily on clinical concerns and medications, whereas C2 messaging focused more on parent education, proactive check-ins, and nonmedical aspects of the child’s life. Four themes emerged from the platform user interviews related to C2 messaging: [1] connection to the care team, [2] efficient communication, [3] clinical uses of secure messaging, and [4] barriers to use.
### Conclusions
Overall, our study provides valuable insight into the benefits of secure messaging in the care of children with medical complexity. Secure messaging provided the opportunity for continued family teaching, proactive check-ins from health care providers, and casual conversations about family and child life, which contributed to PCs feeling an improved sense of connection with their child’s health care team. Secure messaging can be a beneficial additional communication method to improve communication between PCs and their care team, reducing the associated burden of care coordination and ultimately enhancing the experience of care delivery. Future directions include the evaluation of secure messaging when integrated into electronic medical records, as this has the potential to work well with CTM workflow, reduce redundancy, and allow for new features of secure messaging.
## Introduction
Children with medical complexity are characterized by medical fragility and chronic conditions, and they often require the involvement of multiple specialists across multiple care settings and medical technologies to support activities of daily living [1]. Parental caregivers (PCs) of children with medical complexity often bear the complex burden of coordinating their child’s care, performing tasks such as liaising with multiple care team members (CTMs), coordinating appointments, and merging medical recommendations [2,3]. Some key principles for improving care coordination for children with medical complexity have been suggested, including shared goals, mutual respect, and real-time communication [4].
PCs and CTMs of children with medical complexity have noted several barriers that impede their ability to communicate promptly with members of the child’s health care team, subsequently leading to increased care coordination requirements [5,6]. One such barrier is organizational policies that limit the use of email due to security concerns and promote the use of telephone communication and faxing [6]. A solution to improve these communication challenges and ensuing negative impacts on care coordination is the implementation of real-time communication including email, texting, or secure messaging [6]. Real-time messaging can offer children with medical complexity and their families additional support with the many moving parts of the child’s care, especially outside of medical appointments or admissions to hospitals. Key workers involved in the care of children with medical complexity aim to provide individualized information, support, and direction to families through comprehensive knowledge of a child’s condition and assist with coordination, communication, and follow-through with plans of care [7,8]. As described by the Complex Care for Kids Ontario program standard [9], nurse practitioners (NPs) often play the role of the key clinical worker and are needed as the lead for care coordination for a child with medical complexity. We suggest that real-time communication with key workers for children with medical complexity and their families could enable streamlined care provision, efficient utilization of resources, and improved patient-centered care.
The use of real-time communication methods has been explored in several different pediatric populations and populations of adult patients that require caregivers [10-15]. For instance, telephone-based messaging interventions for caregivers of people with dementia have demonstrated positive impacts on the quality of life for patients, reduced caregiver burden, and reduced emergency department visits [15]. Notably, in pediatric populations, PCs of children with 1 chronic illness (eg, cystic fibrosis, diabetes mellitus, juvenile arthritis, and sickle cell disease) have reported that real-time communication methods helped to reduce barriers in communication, facilitated early intervention in screening, aided referral for treatments, and allowed for the formation of an ongoing relationship with health care providers (HCPs) that was not previously possible [14,16]. However, to our knowledge, there are no studies that examine the use of real-time communication in the care of children with multiple or complex chronic conditions. Therefore, the objectives of this study were 3-fold: [1] to evaluate and compare the use of a real-time secure messaging system by CTMs of children with medical complexity, [2] to examine and compare the content of the messages to that of emails and phone calls, and [3] to explore PCs and CTMs’ perceptions and experiences using secure messaging as a method of communication.
## Study Design
This mixed methods evaluation of a web and mobile–based patient-facing platform used messaging, phone call, and email data as well as semistructured interviews to investigate the experiences of PCs and CTMs of children with medical complexity when using a secure messaging system. A mixed methods evaluation enabled us to assess the objective usage of the platform among both CTMs and PCs. It also helped us understand the subjective experiences of users and how they were able to use and incorporate secure messaging into the care of children with medical complexity. This study is a component of a larger study investigating the feasibility of a web and mobile–based patient-facing platform to improve communication and care coordination for children with medical complexity [6].
## Platform Overview
Connecting2gether (C2) is a web and mobile–based patient-facing platform developed for children with medical complexity that supports various functions, including secure messaging, health tracking, educational resources, and a shared medical summary. This platform was developed by using adult literature on secure messaging, referring to caregiver models, and customizing features of the online platform to the health care needs of children with medical complexity. The platform was accessible through desktop, tablet, and mobile devices. The secure messaging feature enabled one-to-one communication through an instant messaging interface. PCs could use secure messaging with CTMs that they invited to C2. All platform users received a push notification to their mobile devices and an email alerting them that they had received a message on C2. Attachments (ie, pictures, videos, and documents) could also be sent in C2 secure messaging. Messages could be responded to immediately, within seconds after a message was received, or days after the message was received.
## Ethics Approval
Institutional research ethics approval was obtained at The Hospital for Sick Children (SickKids; 1000060804), Royal Victoria Regional Health Centre (RVH; R18-013), and Credit Valley Hospital (CVH; 973) [17]. Informed consent was obtained from all participants, and all methods were carried out in accordance with relevant guidelines and regulations. All study data were deidentified before analysis.
## Participant Recruitment
PCs of children with medical complexity were recruited from Complex Care Programs at SickKids, RVH, and CVH. To be eligible for the Complex Care Program, children must meet at least 1 criterion from each of the following conditions: technology dependence and/or users of high-intensity care (eg, mechanical ventilator, constant medical/nursing supervision), fragility (eg, severe/life-threatening condition, an intercurrent illness causing immediate serious health risk), chronicity (condition expected to last at least 6 more months or life expectancy less than 6 months), and complexity (involvement of at least 5 health care practitioners/teams at 3 different locations or family circumstances that impede their ability to provide day-to-day care of decision-making for a child with medical complexity) [18]. Children with medical complexity were also between 0 and 18 years of age at the time of study initiation. Purposive sampling guided parental participant selection to ensure diversity in role, communication experience, age, ethnicity, and location [19,20].
PCs were eligible to participate if they were English-speaking, had access to the internet and a computer, and were the primary caregiver of a child with medical complexity. CTMs were approached prior to recruitment to ensure it was an appropriate time to engage in research for the families (eg, hospitalization, end-of-life, or PC physical/mental health concerns).
In this study, “NPs” refers to the nurse practitioners of children with medical complexity in the Complex Care Program, and “HCPs” refers to other hospital and community–based health care providers. CTMs comprise both NPs and HCPs together.
Every PC had their assigned Complex Care Program NP on the platform. PCs were also able to invite other members of their child’s care team (eg, CTMs like social workers, patient information coordinators, pediatricians, etc) to use C2. CTMs that registered on C2 were presented with the terms of use of the platform and the study information letter. If interested, they were approached by the study research coordinator (RC) and presented with information about the research study and the opportunity to participate. CTMs that declined to participate in the research study were still able to use C2. PCs and NPs received training before registering on C2 (duration of 30 to 60 minutes), and the training presentation was later made available on C2. In addition, CTMs could set up a disclaimer on C2 if they were away or designate time slots in which they would respond to messages (eg, 8 AM to 4 PM) to aid in setting expectations with PCs.
All research study participants received remuneration for participating in the research study. PCs were given CAD $60 (US $44.59) in gift cards (CAD $20 at baseline and CAD $40 after completing the study), and HCPs that completed the end-of-study questionnaire were entered into a draw for a CAD $100 (US $74.32) gift card. Participants that completed the end-of-study semistructured interview received an additional gift card worth CAD $20 (US $14.86). C2 also had a built-in points system where PCs received a specified number of points when completing a platform activity (ie, accessing educational material). As a usage incentive, PCs received a gift card worth CAD $5 (US $3.72) when they reached predetermined point milestones. NPs also received a CAD $5 (US $3.72) gift card for every 50 messages that they sent through C2.
## Quantitative Data
Platform users used C2 for 6 months between September 2019 and June 2020. Secure messaging data were extracted from C2. Electronic medical record (EMR) documentation, including email and phone call data during the study period, was extracted from each patient’s EMR. Phone calls were documented by the HCP or NP with a summary of what was discussed and the participants, and emails were summarized or transcribed verbatim and included the number of people on the email thread. The phone call and email data were inputted into the EMR by the HCP anytime after the actual phone calls were completed and email messages were sent and received. Therefore, the time that the phone call and email data were added was not equivalent to when these were completed.
All secure messaging, email, and phone call data were deidentified and analyzed using descriptive statistics. “ Messaging response” was defined as a secure message that required a response and received a response. The messaging response was categorized subjectively by a single reviewer (author CP) by identifying the messages that required a response (eg, questions) and whether they received a response.
The time and date that each secure message was sent were categorized as evening, weekend, or weekday. Weekday messages were sent between 9 AM EST and 5 PM EST from Monday to Friday. Evening messages were sent after 5 PM EST or before 9 AM EST from Monday to Friday, and weekend messages were sent anytime on Saturday or Sunday.
Content analysis methods and inductive category development [21] were used to analyze the content of secure messages on C2, email messages, and phone calls. First, team members performed an initial review of all secure messaging conversations, email messages, and phone call communication. This led to the development of codes (by authors CP and SM) that were used to categorize each message, for example, medication-related messages, appointment-related messages, inpatient hospital stay–related messages, or messages related to the C2 platform. Following this, discussions took place between members of the research team (authors CP, SM, MB, and JO), and codes were grouped together based on subjective similarities defined by the research team. For instance, appointment and inpatient hospital stay messages were combined under the code “clinical encounter.” A finalized coding tree was developed after analysis of the terms and discussion of disagreements, which led to a total of 9 codes (Multimedia Appendix 2). Disagreements in codes were resolved through discussions between the research team members (authors CP, SM, MB, and JO).
## Qualitative Data
Platform users were asked to participate in a semistructured interview at the end of the study period. The research team purposefully sampled platform users to include those with high and low platform usage. Informed consent was obtained, and 2 members of the research team (authors CM and MB) led interviews over the phone or via online video conference. Semistructured interview guides explored platform experiences using C2, with specific questions surrounding each platform feature. The qualitative data used in this study included all questions related to the secure messaging function (Multimedia Appendix 1), such as: Can you tell me about your experience using C2 for communication? How did the messaging system compare to your usual methods for communicating with your child’s care team? Were there any barriers to using secure messaging?
Semistructured interviews were audio recorded and transcribed verbatim. Transcribed data were managed using NVivo 12 software (QSR International) [22]. We did not have any preconceived themes prior to starting data analysis; instead, themes were developed through an inductive process as a research team. Thematic analysis was used to analyze all interview data, which was completed by 3 team members (authors CP, MB, and CM). Thematic analysis began following the first interview. Braun and Clark’s [23] 6 steps of thematic analysis were adapted. First, the analysis process began with familiarization of the data related to secure messaging, where authors CM and MB transcribed the interview verbatim, read the transcripts, and reread the transcripts while listening to the audio recordings. Following this, through inductive coding, author CP generated initial codes by using an iterative process of reviewing all interview data and discussing potential codes that could be applied to the interview data for both PCs and CTMs with other authors (CM, MB, and JO). For example, the initial code of “fast responses” was deducted from quotes discussing the speed of responses on secure messaging. CP then grouped the codes into potential themes and reviewed the themes using constant comparative analysis by comparing themes across and within participant groups. One example of this was the combination of the initial codes of “informal and simple communication” and “fast responses” into the higher-level theme of “efficient communication.” Three research team members (authors CP, MB, and CM) worked together to refine, define, and name the final 4 themes.
## Overview
A total of 36 PCs of children with medical complexity registered on C2. Other platform members in the study included 7 NPs and 59 HCPs, $35\%$ ($$n = 21$$) of whom were pediatricians.
## Quantitative Evaluation of Secure Messaging
During the study period, the RC sent 1818 secure messages to all platform users. PCs sent 1133 (median 30; IQR 13-46) messages, NPs sent 615 (median 74; IQR 37-119.5) messages, and HCPs sent 113 (median 2; IQR 1-4.5) messages. Moreover, $85.45\%$ ($\frac{1257}{1471}$) of secure messages sent during the study period were responded to within 24 hours, and $92.45\%$ ($\frac{1360}{1471}$) within 72 hours. PCs replied fastest to HCPs and slowest to the RC. The RC had the fastest response times (16 minutes) compared to all other platform users (Table 1).
**Table 1**
| Platform user type | Median response time (range) (hh:mm) | Response time <24 hours, n (%) | Response time <72 hours, n (%) |
| --- | --- | --- | --- |
| PCa | 0:36 (0:00-547:18) | 635 (85.1) | 691 (92.6) |
| RCb | 0:16 (0:00-304:49) | 342 (93.4) | 358 (97.8) |
| NPc | 2:36 (0:00-290:06) | 209 (74.6) | 238 (85) |
| HCPd | 0:56 (0:01-330:29) | 64 (87.6) | 67 (91.8) |
The messaging response was $84.53\%$ ($\frac{1295}{1532}$) for all secure messages over the study period. Most secure messages from PCs ($\frac{261}{280}$, $93.2\%$) that required a response from the NP received a response. In messages from PCs to HCPs that required a response, $73\%$ ($\frac{45}{62}$) received a response.
The majority ($\frac{2987}{3679}$, $81.19\%$) of secure messages sent on C2 were sent on weekdays, while $17.09\%$ ($$n = 629$$) of all messages were sent during evenings, and $1.71\%$ ($$n = 63$$) were sent on weekends.
Among the PCs who had phone calls with their care team during the study period ($\frac{21}{36}$, $58\%$), there were, on average, 4 (range 1-10) phone calls with their care team (Figure 1). PCs that sent emails to their care team during the study period sent an average of 5 (range 1-19) emails. On average, PCs sent 17 secure messages to both NPs and HCPs on C2.
**Figure 1:** *Comparison of the number of Connecting2gether (C2) secure messages, emails, and phone calls by parental caregivers (PCs) to their child’s nurse practitioners (NPs) and health care providers (HCPs) during the study period.*
Due to technical limitations of the platform, conversations on C2 could only include 2 people (ie, the parent and 1 other user). Email allowed individuals to have any number of people within each email thread. Accordingly, $73\%$ ($\frac{100}{137}$) of the emails contained 2 participants, $19\%$ ($$n = 26$$) contained 3, $5\%$ ($$n = 7$$) contained 4, and $3\%$ ($$n = 4$$) contained 5 participants.
## Content Analysis and Comparison for Secure Messaging, Phone Calls, and Emails
Approximately one-third ($\frac{1126}{3679}$, $30.60\%$) of the secure messages on C2 were coded as “C2,” indicating that these messages were related to how to use C2 and its features. These messages were unique to secure messaging on C2 and therefore are not included in Multimedia Appendix 3. The 2 most frequent codes across all 3 forms of communication (secure messaging, email, and phone) were “clinical concern and encounter” and “medications and medical equipment” (Multimedia Appendix 3). There was a higher frequency of “education and resources,” “child and PC life, nonclinical,” and “check-in” codes for secure messaging compared to email and phone communication.
## Themes From Semistructured Interviews
Four themes emerged from the platform user interview data related to secure messaging on C2: [1] connection to the care team, [2] efficient communication, [3] clinical uses of secure messaging, and [4] barriers to use (Table 2). The first theme (connection to the care team) focuses on how secure messaging impacted access to CTMs and the personal relationships and connections PCs developed with their CTMs. For instance, PCs discussed how beneficial it was to be able to communicate directly with their care team. The second theme (efficient communication) focuses on the convenience and ease of using C2, which both CTMs and PCs endorsed. The third theme (clinical use of secure messaging) describes how PCs and CTMs used secure messaging to improve their clinical experiences. For example, secure messaging was used to supplement in-person visits, discuss medication changes, and communicate with community-based HCPs. The fourth theme (barriers to use) discusses the limitations of secure messaging, how it could be improved, and why some platform users prefer to use email and phone. Many PCs and CTMs mentioned that the extra steps to log into C2 to read and respond to messages acted as a deterrent and described that at times they were unsure of when to use each communication platform. It is important to note, however, that in interviews with CTMs, none expressed that there was an increase in their workload due to the volume of messages being sent.
**Table 2**
| Themes | PCb quotes | CTMc quotes |
| --- | --- | --- |
| Connection to the care team | “Over time it became a partnership between the care team and [us]. There’s an invitation there to suggest treatments or alternatives. We [were] able to be taken seriously and to start this cooperation in the care of our daughter.” PC#8 “I feel heard, and more connected because of the [NP’sd] quick response. We feel that we’re genuinely being taken care of by her, and that she’s concerned, not just about my son and the medical side, but just our family’s wellbeing in general. That definitely came through, because of [C2].” PC#23 | “[With messaging on C2], it feels less formal than email so you can just send like a quick, one or two lines to the family.” CTM#3 “[With] email, the amount of detail [parents] go into might be a little bit more. With [C2], they were more direct, [showed] their concerns, and [asked] their questions.” CTM#35 |
| Efficient communication | “It’s easier to pull up previous conversations in a message, as opposed to different emails all over the place. If I kept scrolling up [on C2] then I could see what we had talked about two months ago.” PC#26 “I found [C2 to be] an easy way to send off quick messages. I think it was more friendly or laid-back… Maybe more casual.” PC#22 “I think my favourite [C2 feature] was the messenger. To be able to contact the pediatrician directly, instead of trying to call an office, and then trying to book a time for us to be able to chat with him.” PC#54 | “I would say [email and C2] are similar. And perhaps [C2] was collated, better. [Emails] can get buried with other emails but [on C2 conversations are] specific to that patient so [there is] a bit of a trail.” CTM#16 “[With] email the amount of detail they go into might be a little bit more. With [C2] I felt like [PCs] were able to be more direct, show their concerns, and ask their questions.” CTM#35 “I found that it was much quicker to get a hold of them [via C2], rather than email or calling [PCs].” CTM#27 |
| Clinical uses of secure messaging | “I think [C2 is] a great supplement. For example, our Complex Care meetings will be twice a year. Stuff happens in the six months in between and that is where [C2] was very helpful.” PC#8 “[The community-based social worker and I] had a lot of back-and-forth based on accessibility equipment that we were having ordered and measured. [We were] able to communicate with her when she had a quick question, or if we were relaying information to her, as well as ask questions and prepare things between visits. [C2] made that piece of it easier for us.” PC#19 “It was really nice that [our NP] would [message] and say, “Hey, just checking in, how’s [son] doing?”. Had I not had this platform, I wouldn’t have told her that [for example] he’s got a cough. We would deal with it at home and if need be, we would go to the Emergency in our town. She’s more up-to-date with what was going on with him.” PC#26 | “If it’s not something that’s a new onset of a certain symptom or a certain issue, then the messaging platform is fine, especially dealing with ongoing issues. When something new has happened, and you need more detail, then I can see the email or phone might be more useful.” CTM#35 “I mainly used [C2] if there was a follow-up from an [previous] issue, as a check-in but also, if they had an upcoming visit, reaching out to them just to see if there’s anything in particular that they wanted to discuss at the visit.” CTM#03 “I think [C2] allows continuity between [visits] for things that we’re trying to work on. I think this [C2] messaging] would allow [for] communication between the clinical team and the family [between appointments]. It would mean that you’re using the time between appointments.” CTM#40 |
| Barriers to use | “The helpfulness [of C2] was limited in this study period primarily because the health care providers on my son’s [care] team didn’t accept the invitations to participate.” PC#36 “If there was a way [on C2] to, communicate with all three [HCPs] at the same time, to get the same message to all of them, that would be a benefit” PC#47 “I normally use email because it’s the most convenient. Requires me to take the fewest extra steps in order to send the message. I am usually [already] logged into my email anyway.” PC#53 | “It would be nice if [C2] were, linked to [the EMRe]. It’s hard to keep track of all of the different ways that people can try to get in touch with you and it’s easy to miss [a message] if you don’t respond right away. Adding another system makes me worry that I wouldn’t be able to handle everything.” CTM#42 “My issue with using all three [forms of communication is that parent] wasn’t sure which one was the “right” way to message me. So she wouldn’t consistently use the same [form of communication].” CTM#25 |
## Principal Findings
The use of secure messaging as a means of communication in health care is a relatively novel concept in the pediatric literature. To our knowledge, this is the first mixed methods study investigating the use and perceptions of a secure messaging system housed in a web and mobile–based patient-facing platform for PCs and CTMs of children with medical complexity. Our findings demonstrate that secure messaging was highly used, allowed for diverse topics of conversation, and enhanced the PC-CTM relationship. This paper also demonstrated the utility of secure messaging as a significant collaborative tool between CTMs and parental caregivers, as demonstrated by the large volume of messages exchanged, as an addition to existing forms of communication. In addition to medical information and administrative conversations, secure messaging has surrounded the sharing of educational resources, proactive check-ins initiated by CTMs, and general conversation about the child/family life. Our discussion of this paper will focus on 3 key areas: the clinical utility of secure messaging, the enhancement of the role of the PC with secure messaging, and the workflow and risks of secure messaging.
There is limited pediatric literature to date showing the clinical utility and role of SM; however, there has been noted enthusiasm about the exploration of this topic [11,24,25]. There is mixed evidence in the literature of how secure messaging could be used clinically. Only 1 study with children and adolescents with sickle cell disease found that messaging was used for proactive check-ins and sharing of educational information [14], whereas a study with pediatric surgical patients did not [26]. Considering the current literature, our findings suggest that children with higher care needs and chronic conditions, such as those with medical complexity and those with sickle cell disease, benefit from the use of secure messaging as it allows for proactive check-ins from their CTMs as well as continued education surrounding their medical conditions. For children with medical complexity, there is a need for constant check-ins from CTMs, and secure messaging enables this. In particular, for these children and their families, there is a clinical utility in accessible secure messaging, as it enables prompt clinical response to the high acuity issues we frequently see in this population. Our finding suggesting that secure messaging could be a modality to support parental education is especially salient as previous research has shown that PCs of children with medical complexity often wish to learn more about their child’s condition(s) and how to provide care at home [27,28].
Taken together, we suggest that all the noted benefits of secure messaging can contribute to reducing the burden of care coordination for PCs of children with medical complexity. The four defining characteristics of care coordination include [1] family-centeredness; [2] planned, proactive, and comprehensive focus; [3] promotion of self-care skills and independence; and [4] emphasis on cross-organizational relationships [4]. Indeed, improving the connection and ease of communication between PCs and CTMs increases the opportunity for partnership, family-centered care, and independence, while the finding of proactive check-ins by CTMs aligns well with a planned, proactive, and comprehensive focus.
A barrier to communication between PCs of children with medical complexity and their CTMs is a perceived lack of partnership, as PCs do not feel acknowledged for their expertise and cannot contribute to their child’s care plan [6]. PCs in our study noted that secure messaging allowed them to play a more active role in their child’s care, allowed for more informal conversations with their child’s health care team, and ultimately improved their perceived partnership with their child’s health care team. This finding is supported in adult literature; a study of women with breast cancer demonstrated that the use of an interactive communication tool allowed participants to feel more empowered to participate in their health care and improved their relationship with their health care team [29].
A common concern in the literature surrounding real-time communication and secure messaging is that patients may overwhelm CTMs by sending many messages, significantly increasing their workload [30,31]. In this study, PCs and NPs received training on the appropriate use of C2 prior to registering. Research on the use of patient portals in primary care clinics demonstrates that our efforts to provide clear and structured training addressed any concerns on how to communicate with CTMs appropriately [32]. Our research adds to the literature demonstrating that patients do not overuse secure messaging systems [33]. The findings of this paper dispel the concerns of secure messaging overuse because secure messaging did not add significantly to the workload for CTMs, and it can be constricted to weekday daytime hours.
Both PCs and CTMs noted limitations that affected their use of the secure messaging system on C2. These limitations, including additional log-in steps associated with secure messaging and difficulty identifying which communication method to use, are similar to those reported by Hsiao et al [34], who implemented a secure messaging platform in the care of pediatric patients with respiratory diseases. CTMs in our study felt that having the secure messaging system integrated with their EMR would greatly reduce any barriers to access and the additional log-in steps required. Indeed, previous research has demonstrated a high uptake of secure messaging systems when integrated into a patient portal and EMR [35]. In this study, CTMs were required to both communicate by secure messaging and copy the conversation into the patient’s chart. Secure messaging as a part of the EMR would thus lead to reduced redundancy in separate health information platforms, as all information would be housed in a single system. In addition, if they are separate systems, other people who are not using the separate platform would not have access to the secure messages that are important aspects of their child’s care. The integration of secure messaging into the EMR also enables messaging with patients to become a part of a CTM’s standard workflow and includes the messaging as a permanent record in the EMR [36]. Moreover, having secure messaging as a core function within an existing EMR would enhance health care delivery. Specifically, this would allow the provider to respond to secure messages with access to accurate and real-time information including up-to-date medication lists, upcoming appointment times, and hospitalizations or emergency department encounters. Ultimately, including secure messaging in the EMR blends well into the workflow of CTMs and leads to family-centered care, wherein all care providers and PCs have access to the same information. Future research should focus on health-related outcomes (for instance, hospitalizations and emergency department visits) associated with integrating messaging into the EMR, as well as perceptions of HCPs on how integration affected their clinical workflow.
## Limitations
This study has several limitations. First, most of our relatively small PC sample who used secure messaging on the web-based patient-facing platform were highly educated and had high household incomes. Therefore, our results may not be representative of families with lower education, health literacy, or income. However, previous research has shown that economically disadvantaged families still access the internet regularly and are interested in receiving hospital communication through electronic means [12]. Further, the platform was only available in English, and we did not collect PC immigration status. Using secure messaging may present different benefits and challenges to families who are new to Canada and those who do not speak English fluently. There was a possible participation bias, as PCs who chose to use the platform for 6 months and participate in the research study likely already had positive perceptions of and were comfortable using technology such as smartphones, the internet, and computers. However, we attempted to mitigate this bias in the semistructured interviews by recruiting high and low platform users. Finally, this study’s primary focus was on the evaluation of platform usage. Further analysis will include understanding the individual-level characteristics and utilization patterns of secure messaging.
## Conclusions
Our study demonstrates that secure messaging was an effective and additive form of communication for PCs of children with medical complexity and their CTMs that had a high uptake among users. Through a web and mobile–based patient-facing platform, secure messaging provided the opportunity for continued patient education, proactive check-ins from CTMs, and casual conversations about family and child life, all of which contributed to PCs feeling improved connections with their child’s health care team. Therefore, we suggest that secure messaging can improve communication between PCs of children with medical complexity and their care team to reduce the burden of care coordination and ultimately improve the experience of care delivery. Future research should focus on a more rigorous evaluation of secure messaging in the care of children with medical complexity and how it can affect health-related outcomes, especially when integrated into the EMR.
## Data Availability
The data sets generated and analyzed during this study are available from the corresponding author upon reasonable request.
## References
1. Cohen E, Kuo DZ, Agrawal R, Berry JG, Bhagat SKM, Simon TD, Srivastava R. **Children with medical complexity: an emerging population for clinical and research initiatives**. *Pediatrics* (2011.0) **127** 529-38. DOI: 10.1542/peds.2010-0910
2. Cady R, Belew J. **Parent perspective on care coordination services for their child with medical complexity**. *Children (Basel)* (2017.0) **4** 45. DOI: 10.3390/children4060045
3. Kuo DZ, Cohen E, Agrawal R, Berry JG, Casey PH. **A national profile of caregiver challenges among more medically complex children with special health care needs**. *Arch Pediatr Adolesc Med* (2011.0) **165** 1020-6. DOI: 10.1001/archpediatrics.2011.172
4. Kuo D, McAllister J, Rossignol L, Turchi R, Stille C. **Care coordination for children with medical complexity: whose care is it, anyway?**. *Pediatrics* (2018.0) **141** S224-S232. DOI: 10.1542/peds.2017-1284g
5. Foster CC, Mangione-Smith R, Simon TD. **Caring for children with medical complexity: perspectives of primary care providers**. *J Pediatr* (2017.0) **182** 275-282.e4. DOI: 10.1016/j.jpeds.2016.11.017
6. Adams S, Beatty M, Moore C, Desai A, Bartlett L, Culbert E, Cohen E, Stinson J, Orkin J. **Perspectives on team communication challenges in caring for children with medical complexity**. *BMC Health Serv Res* (2021.0) **21** 300. DOI: 10.1186/s12913-021-06304-8
7. Kingsnorth S, Lacombe‐Duncan A, Keilty K, Bruce‐Barrett C, Cohen E. **Inter‐organizational partnership for children with medical complexity: the integrated complex care model**. *Child Care Health Dev* (2013.0) **41** 57-66. DOI: 10.1111/cch.12122
8. Orkin J, Chan CY, Fayed N, Lin JLL, Major N, Lim A, Peebles ER, Moretti ME, Soscia J, Sultan R, Willan AR, Offringa M, Guttmann A, Bartlett L, Kanani R, Culbert E, Hardy-Brown K, Gordon M, Perlmutar M, Cohen E. **Complex care for kids Ontario: protocol for a mixed-methods randomised controlled trial of a population-level care coordination initiative for children with medical complexity**. *BMJ Open* (2019.0) **9** e028121. DOI: 10.1136/bmjopen-2018-028121
9. **CCKO functions of a complex care clinic and program standard**. *Provincial Council for Maternal and Child Health* (2017.0)
10. Dudas RA, Crocetti M. **Pediatric caregiver attitudes toward email communication: survey in an urban primary care setting**. *J Med Internet Res* (2013.0) **15** e228. DOI: 10.2196/jmir.2738
11. Masterman M, Cronin RM, Davis SE, Shenson JA, Jackson GP. **Adoption of secure messaging in a patient portal across pediatric specialties**. *AMIA Annu Symp Proc* (2016.0) **2016** 1930-1939. PMID: 28269952
12. Saidinejad M, Teach SJ, Chamberlain JM. **Internet access and electronic communication among families in an urban pediatric emergency department**. *Pediatr Emerg Care* (2012.0) **28** 553-7. DOI: 10.1097/PEC.0b013e318258ad76
13. Rosen P, Kwoh C. **Patient-physician e-mail: an opportunity to transform pediatric health care delivery**. *Pediatrics* (2007.0) **120** 701-6. DOI: 10.1542/peds.2007-1094
14. Jacob E, Pavlish C, Duran J, Stinson J, Lewis MA, Zeltzer L. **Facilitating pediatric patient-provider communications using wireless technology in children and adolescents with sickle cell disease**. *J Pediatr Health Care* (2013.0) **27** 284-92. DOI: 10.1016/j.pedhc.2012.02.004
15. Possin KL, Merrilees JJ, Dulaney S, Bonasera SJ, Chiong W, Lee K, Hooper SM, Allen IE, Braley T, Bernstein A, Rosa TD, Harrison K, Begert-Hellings H, Kornak J, Kahn JG, Naasan G, Lanata S, Clark AM, Chodos A, Gearhart R, Ritchie C, Miller BL. **Effect of collaborative dementia care via telephone and internet on quality of life, caregiver well-being, and health care use: the Care Ecosystem randomized clinical trial**. *JAMA Intern Med* (2019.0) **179** 1658-1667. DOI: 10.1001/jamainternmed.2019.4101
16. Britto MT, Hesse EA, Kamdar OJ, Munafo JK. **Parents' perceptions of a patient portal for managing their child's chronic illness**. *J Pediatr* (2013.0) **163** 280-1.e1. DOI: 10.1016/j.jpeds.2013.02.041
17. Cohen E, Lacombe-Duncan A, Spalding K, MacInnis J, Nicholas D, Narayanan UG, Gordon M, Margolis I, Friedman JN. **Integrated complex care coordination for children with medical complexity: a mixed-methods evaluation of tertiary care-community collaboration**. *BMC Health Serv Res* (2012.0) **12** 366. DOI: 10.1186/1472-6963-12-366
18. **The standard operational definition for children with medical complexity**. *Complex Care Kids Ontario*
19. Palinkas LA, Horwitz SM, Green CA, Wisdom JP, Duan N, Hoagwood K. **Purposeful sampling for qualitative data collection and analysis in mixed method implementation research**. *Adm Policy Ment Health* (2015.0) **42** 533-44. DOI: 10.1007/s10488-013-0528-y
20. Moser A, Korstjens I. **Series: Practical guidance to qualitative research. Part 3: Sampling, data collection and analysis**. *Eur J Gen Pract* (2018.0) **24** 9-18. DOI: 10.1080/13814788.2017.1375091
21. Hsieh H, Shannon SE. **Three approaches to qualitative content analysis**. *Qual Health Res* (2005.0) **15** 1277-88. DOI: 10.1177/1049732305276687
22. **NVivo (version 12)**. *QSR International* (2018.0)
23. Braun V, Clarke V. **Using thematic analysis in psychology**. *Qual Res Psychol* (2006.0) **3** 77-101. DOI: 10.1191/1478088706qp063oa
24. Cronin RM, Davis SE, Shenson JA, Chen Q, Rosenbloom ST, Jackson GP. **Growth of secure messaging through a patient portal as a form of outpatient interaction across clinical specialties**. *Appl Clin Inform* (2015.0) **6** 288-304. DOI: 10.4338/ACI-2014-12-RA-0117
25. Steitz B, Cronin R, Davis S, Yan E, Jackson G. **Long-term patterns of patient portal use for pediatric patients at an academic medical center**. *Appl Clin Inform* (2017.0) **08** 779-793. DOI: 10.4338/aci-2017-01-ra-0005
26. Riera KM, Robinson JR, Van Arendonk KJ, Jackson GP. **Care delivered by pediatric surgical specialties through patient portal messaging**. *J Surg Res* (2019.0) **234** 231-239. DOI: 10.1016/j.jss.2018.09.013
27. Nelson BB, Coller RJ, Saenz AA, Chung PJ, Kaplan A, Lerner CF, Klitzner TS. **How avoidable are hospitalizations for children with medical complexity? Understanding parent perspectives**. *Acad Pediatr* (2016.0) **16** 579-86. DOI: 10.1016/j.acap.2016.04.009
28. Nageswaran S, Golden SL. **Improving the quality of home health care for children with medical complexity**. *Acad Pediatr* (2017.0) **17** 665-671. DOI: 10.1016/j.acap.2017.04.019
29. Lu H, Shaw BR, Gustafson DH. **Online health consultation: examining uses of an interactive cancer communication tool by low-income women with breast cancer**. *Int J Med Inform* (2011.0) **80** 518-28. DOI: 10.1016/j.ijmedinf.2011.03.011
30. Kittler A, Wald J, Volk L, Pizziferri L, Jagannath Y, Harris C, Lippincott M, Yu T, Hobbs J, Bates D. **The role of primary care non-physician clinic staff in e-mail communication with patients**. *Int J Med Inform* (2004.0) **73** 333-40. DOI: 10.1016/j.ijmedinf.2004.02.004
31. Miller DP, Latulipe C, Melius KA, Quandt SA, Arcury TA. **Primary care providers' views of patient portals: interview study of perceived benefits and consequences**. *J Med Internet Res* (2016.0) **18** e8. DOI: 10.2196/jmir.4953
32. Sieck CJ, Hefner JL, Schnierle J, Florian H, Agarwal A, Rundell K, McAlearney AS. **The rules of engagement: perspectives on secure messaging from experienced ambulatory patient portal users**. *JMIR Med Inform* (2017.0) **5** e13. DOI: 10.2196/medinform.7516
33. Stamenova V, Agarwal P, Kelley L, Fujioka J, Nguyen M, Phung M, Wong I, Onabajo N, Bhatia RS, Bhattacharyya O. **Uptake and patient and provider communication modality preferences of virtual visits in primary care: a retrospective cohort study in Canada**. *BMJ Open* (2020.0) **10** e037064. DOI: 10.1136/bmjopen-2020-037064
34. Hsiao AL, Bazzy-Asaad A, Tolomeo C, Edmonds D, Belton B, Benin AL. **Secure web messaging in a pediatric chronic care clinic: a slow takeoff of "kids' airmail"**. *Pediatrics* (2011.0) **127** e406-13. DOI: 10.1542/peds.2010-1086
35. Zhou YY, Garrido T, Chin HL, Wiesenthal AM, Liang LL. **Patient access to an electronic health record with secure messaging: impact on primary care utilization**. *Am J Manag Care* (2007.0) **13** 418-24. PMID: 17620037
36. Hess R, Bryce CL, Paone S, Fischer G, McTigue KM, Olshansky E, Zickmund S, Fitzgerald K, Siminerio L. **Exploring challenges and potentials of personal health records in diabetes self-management: implementation and initial assessment**. *Telemed J E Health* (2007.0) **13** 509-17. DOI: 10.1089/tmj.2006.0089
|
---
title: Incidence and Progression of Diabetic Retinopathy in American Indian and Alaska
Native Individuals Served by the Indian Health Service, 2015-2019
authors:
- Stephanie J. Fonda
- Sven-Erik Bursell
- Drew G. Lewis
- Dawn Clary
- Dara Shahon
- Jerry Cavallerano
journal: JAMA Ophthalmology
year: 2023
pmcid: PMC9999279
doi: 10.1001/jamaophthalmol.2023.0167
license: CC BY 4.0
---
# Incidence and Progression of Diabetic Retinopathy in American Indian and Alaska Native Individuals Served by the Indian Health Service, 2015-2019
## Abstract
This cohort study investigates the rates of incidence and progression for diabetic retinopathy in American Indian and Alaska Native individuals.
## Key Points
### Question
What are current rates of incidence and progression for diabetic retinopathy (DR) in American Indian and Alaska Native individuals?
### Findings
In this cohort study of 8374 American Indian and Alaska Native individuals evaluated in 2015 and again in 2016 to 2019 by the Indian Health Service (IHS) teleophthalmology program, cumulative incidence of DR (any level) was $18\%$, and $27\%$ of patients with mild nonproliferative DR (NPDR) progressed to moderate NPDR or worse, lower than previous reports.
### Meaning
These estimates suggest the viability of extending follow-up intervals for retinopathy assessment in IHS patients with no DR or mild NPDR if follow-up compliance is not jeopardized.
### Importance
Estimates of diabetic retinopathy (DR) incidence and progression in American Indian and Alaska Native individuals are based on data from before 1992 and may not be informative for strategizing resources and practice patterns.
### Objective
To examine incidence and progression of DR in American Indian and Alaska Native individuals.
### Design, Setting, and Participants
This was a retrospective cohort study conducted from January 1, 2015, to December 31, 2019, and included adults with diabetes and no evidence of DR or mild nonproliferative DR (NPDR) in 2015 who were reexamined at least 1 time during the 2016 to 2019 period. The study setting was the Indian Health Service (IHS) teleophthalmology program for diabetic eye disease.
### Exposure
Development of new DR or worsening of mild NPDR in American Indian and Alaska Native individuals with diabetes.
### Main Outcomes and Measures
Outcomes were any increase in DR, 2 or more (2+) step increases, and overall change in DR severity. Patients were evaluated with nonmydriatic ultra-widefield imaging (UWFI) or nonmydriatic fundus photography (NMFP). Standard risk factors were included.
### Results
The total cohort of 8374 individuals had a mean (SD) age of 53.2 (12.2) years and a mean (SD) hemoglobin A1c level of $8.3\%$ ($2.2\%$) in 2015, and 4775 were female ($57.0\%$). Of patients with no DR in 2015, $18.0\%$ (1280 of 7097) had mild NPDR or worse in 2016 to 2019, and $0.1\%$ (10 of 7097) had PDR. The incidence rate from no DR to any DR was 69.6 cases per 1000 person-years at risk. A total of $6.2\%$ of participants (441 of 7097) progressed from no DR to moderate NPDR or worse (ie, 2+ step increase; 24.0 cases per 1000 person-years at risk). Of patients with mild NPDR in 2015, $27.2\%$ (347 of 1277) progressed to moderate NPDR or worse in 2016 to 2019, and $2.3\%$ (30 of 1277) progressed to severe NPDR or worse (ie, 2+ step progression). Incidence and progression were associated with expected risk factors and evaluation with UWFI.
### Conclusions and Relevance
In this cohort study, the estimates of DR incidence and progression were lower than those previously reported for American Indian and Alaska Native individuals. The results suggest extending the time between DR re-evaluations for certain patients in this population, if follow-up compliance and visual acuity outcomes are not jeopardized.
## Introduction
Although the rate of increase has declined recently,1 diabetes prevalence in American Indian and Alaska Native individuals remains higher than in other race and ethnic groups in the US, with $14.7\%$ of American Indian and Alaska Native individuals having been diagnosed with diabetes.2 The Centers for Disease Control and Prevention predicts that one-half of American Indian and Alaska Native individuals born in 2000 will develop diabetes some time in their lives.3 Additionally, an analysis of 1990 to 1998 Indian Health Service (IHS) outpatient data found that diabetes is being diagnosed at younger ages in American Indian and Alaska Native individuals.4 Longer duration could mean greater likelihood of diabetes-associated complications. In this vein, between 2015 and 2050, the number of people who are blind is projected to double,5 and diabetic retinopathy (DR) will likely be a major contributor because DR remains a leading cause of preventable blindness in US adults.6 *In this* context of disease burden, it is important to understand the prevalence and incidence of diabetes complications to allocate surveillance programs and specialty services appropriately. Recent publications suggest that diabetic eye disease prevalence has declined in American Indian and Alaska Native patients of the US IHS.7,8 However, estimates of DR incidence in American Indian and Alaska Native patients are not current; instead, they based on data from before 1992.9,10,11,12 This study estimates recent cumulative incidence, incidence rates, and progression of DR in American Indian and Alaska Native patients served by this IHS primary care–based teleophthalmology program.
## Setting and Study Population
This was a retrospective cohort study using deidentified medical record data obtained during routine clinical operations of the IHS teleophthalmology program at 75 primary care clinics distributed among 20 states. The IHS serves enrolled members of federally recognized tribes. The study was reviewed and approved by the IHS institutional review board at Phoenix Indian Medical Center under the exempt process. Written informed consent from participants was not required or obtained.
Details regarding the teleophthalmology program’s origins, protocols, distribution, and outcomes have been previously described.13 Briefly, the program evaluates patients from participating primary care clinics. It is a validated American Telemedicine Association Category 3 program and its graders identify the Early Treatment Diabetic Retinopathy Study (ETDRS)–defined clinical levels of DR and diabetic macular edema (DME) severity.13,14,15 Graders are certified and licensed optometrists who render a diagnosis using standardized protocols. The program currently recommends that patients receive annual DR examinations.
Before selecting the analytic cohort for this study, we defined a baseline period of January 1, 2015, to December 31, 2015, and a follow-up period of January 1, 2016, to December 31, 2019. Eligible patients had at least 1 IHS teleophthalmology examination with the program in both periods. Additionally, eligible patients were 20 years or older and had no evidence of DR or had mild nonproliferative DR (NPDR; ETDRS levels 10, 14, 15, 20) in the baseline period. Patients with severe/very severe NPDR (ETDRS levels 53 a-e), proliferative DR (PDR; ETDRS levels 61, 65, 71, 75, 81, 85), and/or any level of DME are referred out of the teleophthalmology program to specialty eye care; therefore, these patients were excluded. Referral recommendations of patients with moderate NPDR (ETDRS levels 35, 43, 47) are dependent on risk factors; therefore, these patients were also excluded. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines.
## Outcomes in the Follow-up Period
The IHS teleophthalmology program uses 2 configurations of commercially available technology for image acquisition to assess DR severity level. The first configuration uses a low-illumination, nonmydriatic fundus photography (NMFP) digital imaging system (Topcon NW6S [Topcon Medical Systems]) with a custom digital camera back (Megavision Retinal Image Capture).16 *Three nonsimultaneous* stereo-pair 45° images from different retinal regions and 2 nonsimultaneous stereo-pair 30° digital images of the optic disc and the macula from the retina of each eye are obtained, for an approximate total retinal coverage of 90° to 135°. One external image of each eye is also obtained. NMFP showed substantial agreement with ETDRS controls for diagnosis of DR severity level (unweighted κ = 0.81; $95\%$ CI, 0.73-0.89).16 The second configuration is nonmydriatic ultra-widefield imaging (UWFI) scanning laser ophthalmoscopy (SLO) (Daytona [Optos]). The UWFI protocol includes nonsimultaneous stereo-pair 200° images from each eye, centered on the macula. Previous research has shown that UWFI agrees perfectly with ETDRS photography in $84\%$ of cases and agrees within 1 level of severity in $91\%$ of cases (unweighted κ = 0.79).17 UWFI is the dominant configuration this program uses.
The grading outcomes were no evidence of DR, mild NPDR, moderate NPDR, severe/very severe NPDR, PDR, or ungradable. Level of DR at any 1 imaging encounter was defined by the more severely affected eye. If 1 eye was ungradable, the diagnosis for the other eye was used. If a patient received more than 1 teleophthalmology examination during the follow-up period, their maximum diagnosis was used in this analysis.
This study measured incidence and progression as follows: [1] any increase in level of DR; [2] occurrence of a 2 or more (2+) step increase; and [3] DR severity level. For patients with no evidence of DR at baseline, any increase in level meant mild NPDR or worse was found at follow-up, and a 2+ step increase meant moderate NPDR or worse was found. For patients with mild NPDR at baseline, any increase in level meant moderate NPDR or worse was found, and a 2+ step increase meant severe NPDR or worse was found. Severity levels at follow-up included patients who regressed from mild NPDR to no DR, but regression was not explored.
## Background Variables
The IHS teleophthalmology program records patient demographics (age, sex [self-reported]) and known DR risk factors in templates used by the imagers and graders, taking data from the IHS electronic medical record patient summary. Risk factors recorded include glycosylated hemoglobin A1c (HbA1c) level, diabetes therapy, duration of diabetes (but not diabetes type), hypertension, cardiovascular disease, hypercholesterolemia, peripheral neuropathy, and nephropathy. The program also records the clinic where the imaging occurred and whether UWFI or NMFP was used. We created measures indicating whether UWFI (vs NMFP) was used at baseline only, follow-up only, both examinations, or never. IHS administrative areas (derived from clinic addresses) are shown to describe the geographical distribution of the cohort.
## Statistical Analysis
The analysis calculated descriptive statistics of the selected cohort’s baseline characteristics and the imaging modalities patients were examined with over time. The analysis also calculated descriptive statistics for the patients who did not have a follow-up examination for comparison with the selected cohort. We compared the groups with t tests and χ2 tests.
The analysis next calculated cumulative incidence, cumulative progression, incidence rate, and progression rate. The rates equaled the number of new or worse cases of DR identified during the 2016 to 2019 period divided by total person-years (PY) at risk.18 PY contributed by a patient were truncated at the date of their examination that identified new or worsening DR. If the patient did not develop new DR or have worsening or progression of their DR, the PY they contributed was truncated at the date of their last examination in the follow-up period (≤4 years).
To estimate net associations between background characteristics and outcomes, the analyses conducted separate multivariable robust Poisson regressions. The dependent variables were as follows: [1] any new DR in patients with no DR at baseline, [2] occurrence of a 2+ step increase in DR for patients with no DR at baseline, and [3] any progression of DR in patients with mild NPDR at baseline. Variables representing imaging modality assessed whether UWFI increased detection of worsening disease net of other factors. Analyses obtained the robust SEs to calculate the CIs and 2-sided P values.19 P values <.05 were considered statistically significant. A model for a 2+ step increase in DR from baseline mild NPDR was not estimated due to the small number of patients in this category.
Descriptive statistics were performed using SAS software, version 9.4 (SAS Institute). Incidence rates and robust Poisson regressions were calculated using R software, version 4.1.2 (R Core Team).20
## Patient Characteristics and Imaging Modality
The number of patients evaluated by the program in the baseline year who were 20 years or older and had no evidence of DR or mild NPDR at that examination was 13 694 (Table 1). Of these patients, 8374 ($61.2\%$; mean [SD] age of 53.2 [12.2] years; 4775 females [$57.0\%$]; 3599 males [$43.0\%$]) had at least 1 examination during the follow-up period. Mean (SD) time from the baseline examination to the first follow-up examination was 20.7 (9.5) months. A total of 4581 of 8374 patients ($54.7\%$) who were followed up had 2 or more examinations during the follow-up period.
**Table 1.**
| Characteristic | No. (%)a | No. (%)a.1 | P value |
| --- | --- | --- | --- |
| Characteristic | Selected (n = 8374) | No follow-up (n = 5320) | P value |
| No DR at baseline | 7097 (84.8) | 4493 (84.5) | .64 |
| Mild NPDR at baseline | 1277 (15.3) | 827 (15.6) | |
| Age, mean (SD), y | 53.2 (12.2) | 52.8 (13.7) | .08 |
| Sex (self-report) | | | |
| Male | 3599 (43.0) | 2260 (42.5) | .57 |
| Female | 4775 (57.0) | 3060 (57.5) | |
| Diabetes duration, mean (SD),b y | 8.6 (7.4) | 8.8 (8.1) | .27 |
| Diabetes duration categories, y | | | |
| <1 | 689 (8.2) | 496 (9.3) | <.001 |
| 1-5 | 2396 (28.6) | 1381 (26.0) | <.001 |
| 6-10 | 1791 (21.4) | 904 (17.0) | <.001 |
| 11-15 | 1236 (14.8) | 655 (12.3) | <.001 |
| >15 | 1143 (13.7) | 752 (14.1) | <.001 |
| Not specified | 1119 (13.4) | 1132 (21.3) | <.001 |
| HbA1c, mean (SD),b % | 8.3 (2.2) | 8.2 (2.3) | .01 |
| Missing a recent HbA1c value | | | |
| No | 6789 (81.1) | 4265 (80.2) | .19 |
| Yes | 1585 (18.9) | 1055 (19.8) | .19 |
| Diabetes therapy | | | |
| Diet only | 618 (7.4) | 498 (9.4) | <.001 |
| Oral medications only | 4401 (52.6) | 2483 (46.7) | <.001 |
| Insulin only | 722 (8.6) | 585 (11.0) | <.001 |
| Oral medications and insulin | 1557 (18.6) | 795 (14.9) | <.001 |
| Not specified | 1076 (12.8) | 959 (18.0) | <.001 |
| Other patient diagnoses | Other patient diagnoses | Other patient diagnoses | Other patient diagnoses |
| Hypercholesterolemia | | | |
| Not present | 5643 (67.4) | 3906 (73.4) | <.001 |
| Present | 2731 (32.6) | 1414 (26.6) | <.001 |
| Cardiovascular disease | | | |
| Not present | 7837 (93.6) | 4963 (93.3) | .49 |
| Present | 537 (6.4) | 357 (6.7) | .49 |
| Hypertension | | | |
| Not present | 3815 (45.6) | 2903 (55.6) | <.001 |
| Present | 4559 (54.4) | 2417 (45.4) | <.001 |
| Peripheral neuropathy | | | |
| Not present | 7866 (93.9) | 4971 (93.4) | .24 |
| Present | 508 (6.1) | 349 (6.6) | .24 |
| Nephropathy | | | |
| Not present | 7982 (95.3) | 5108 (96.0) | .05 |
| Present | 392 (4.7) | 212 (4.0) | .05 |
| IHS office/region | | | |
| Alaska (5 clinics) | 12 (0.1) | 72 (1.3) | <.001 |
| Albuquerque (5 clinics) | 342 (4.1) | 574 (10.8) | <.001 |
| Bemidji (5 clinics) | 141 (1.7) | 272 (5.1) | <.001 |
| Billings (7 clinics) | 177 (2.1) | 216 (4.1) | <.001 |
| Great Plains (6 clinics) | 331 (4.0) | 438 (8.2) | <.001 |
| Nashville (5 clinics) | 155 (1.9) | 45 (0.9) | <.001 |
| Navajo (9 clinics) | 2402 (28.7) | 1479 (27.8) | <.001 |
| Oklahoma City (13 clinics) | 1437 (17.2) | 793 (14.9) | <.001 |
| Phoenix (10 clinics) | 2594 (31.0) | 855 (16.1) | <.001 |
| Portland (12 clinics) | 443 (5.3) | 390 (7.3) | <.001 |
| Tucson (3 clinics) | 340 (4.1) | 186 (3.5) | <.001 |
| Imaging technology | | | |
| NMFP | | | |
| Both examinations | 734 (8.8) | | |
| First examination, UWFI second examination | 1839 (22.0) | | |
| UWFI | | | |
| First examination, NMFP second examination | 476 (5.7) | | |
| Both examinations | 5325 (63.6) | | |
In 2015, the mean (SD) HbA1c level of the analyzed cohort was $8.3\%$ ($2.2\%$; to convert HbA1c to proportion of total Hb, multiply by 0.01) (Table 1). The mean (SD) duration of diabetes was 8.6 (7.4) years, and 4401 of 8374 patients ($52.6\%$) were managing their diabetes with oral medications only. Hypercholesterolemia (2731 [$32.6\%$]) and hypertension (4559 [$54.4\%$]) were the most common risk factors. A total of $53.4\%$ of patients (2436 of 4559) in the analytic cohort with diagnosed hypertension were taking blood pressure–lowering medications (percentage not shown in Table 1). The Phoenix, Navajo, and Oklahoma City IHS areas imaged the most patients, with 2594 ($31.0\%$), 2402 ($28.7\%$), and 1437 ($17.2\%$) being imaged, respectively. A total of 5325 of 8374 patients ($63.6\%$) were imaged with UWFI both times, 1839 ($22.0\%$) were imaged with UWFI at follow-up only, and 734 ($8.8\%$) were imaged with NMFP for both examinations.
The selected cohort characteristics were not significantly different from patients who did not have a follow-up examination, except that proportionally fewer of them were missing information about their diabetes duration and diabetes therapy, their mean HbA1c level was slightly higher (mean [SD], $8.3\%$ [$2.2\%$] vs $8.2\%$ [$2.3\%$]), and proportionally more of them had hypercholesterolemia (2731 of 8374 [$32.6\%$] vs 1414 of 5320 [$26.6\%$]), hypertension (4559 [$54.4\%$] vs 2417 [$45.4\%$]), or nephropathy (392 [$4.7\%$] vs 212 [$4.0\%$]) (Table 1).
## Incidence and Progression
Of patients with no evidence of DR at baseline, 1280 of 7097 ($18.0\%$) had some level of DR at follow-up, for an incidence rate of 69.6 cases per 1000 PY (Table 2). Of the new DR found, 839 of 1280 cases ($65.5\%$) were mild NPDR. Cumulative incidence of PDR was $0.1\%$ (10 of 7097), for an incidence rate of 0.5 cases per 1000 PY. Of patients with no evidence of DR at baseline, 441 of 7097 ($6.2\%$) had a 2+ step increase in DR over time (24.0 cases per 1000 PY).
**Table 2.**
| Baseline status | Follow-up status | Follow-up status.1 | Follow-up status.2 | Follow-up status.3 | Follow-up status.4 | Follow-up status.5 | Follow-up status.6 | Follow-up status.7 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Baseline status | Any increase | 2+ Step increase | Severity level | Severity level | Severity level | Severity level | Severity level | Severity level |
| Baseline status | Any increase | 2+ Step increase | No evidence | NPDR | NPDR | NPDR | PDR | Ungradable |
| Baseline status | Any increase | 2+ Step increase | No evidence | Mild | Moderate | Severe/very severe | PDR | Ungradable |
| No DR (n = 7097) a | No DR (n = 7097) a | No DR (n = 7097) a | No DR (n = 7097) a | No DR (n = 7097) a | No DR (n = 7097) a | No DR (n = 7097) a | No DR (n = 7097) a | No DR (n = 7097) a |
| No. (%) | 1280 (18.0) | 441 (6.2) | 5582 (78.7) | 839 (11.8) | 431 (6.1) | 0 (0) | 10 (.14) | 235 (3.3) |
| Cases/1000 PY | 69.6 | 24.0 | | 45.6 | 23.4 | 0 | .5 | |
| Cases/1000 PY (95% CI) | (65.8-73.5) | (21.8-26.3) | | (42.6-48.8) | (21.3-25.7) | 0 | (.3-1.0) | |
| Mild NPDR (n = 1277) b | Mild NPDR (n = 1277) b | Mild NPDR (n = 1277) b | Mild NPDR (n = 1277) b | Mild NPDR (n = 1277) b | Mild NPDR (n = 1277) b | Mild NPDR (n = 1277) b | Mild NPDR (n = 1277) b | Mild NPDR (n = 1277) b |
| No. (%) | 347 (27.2) | 30 (2.3) | 427 (33.4) | 437 (34.2) | 317 (24.8) | 2 (0.16) | 28 (2.2) | 66 (5.2) |
| Cases/1000 PY | 111.7 | 9.7 | | | 102.0 | .6 | 9.0 | |
| Cases/1000 PY (95% CI) | (100.2-124.0) | (6.5-13.8) | | | (91.1-113.9) | (0.1-2.3) | (6.0-13.0) | |
A total of 347 of 1277 patients ($27.2\%$) with mild NPDR at baseline developed a more severe DR level in the follow-up period, for an incidence rate of 111.7 cases per 1000 PY. A 2+ step increase in DR occurred for $2.3\%$ of these patients (30 of 1277). Regarding DR severity level, cumulative incidences of severe/very severe NPDR and PDR were $0.2\%$ (2 of 1277) and $2.2\%$ (28 of 1277), respectively, for incidence rates of 0.6 and 9.0 cases per 1000 PY, respectively.
## Patient Characteristics and DR Outcomes
Characteristics associated with any DR incidence as well as occurrence of a 2+ step increase were longer diabetes duration (>15 y, any DR: risk ratio [RR], 2.0, $95\%$ CI, 1.7-2.4; $P \leq .001$; 2+ step: RR, 3.2; $95\%$ CI, 2.3-4.4; $P \leq .001$), higher HbA1c level (any DR: RR, 1.1; $95\%$ CI, 1.1-1.2; $P \leq .001$; 2+ step: RR, 1.3; $95\%$ CI, 1.2-1.3; $P \leq .001$), and diabetes therapy, particularly insulin use alone (any DR: RR, 2.1; $95\%$ CI, 1.5-2.9; $P \leq .001$; 2+ step: RR, 4.5; $95\%$ CI, 1.8-11.2; $$P \leq .001$$) or with oral medications (any DR: RR, 2.2; $95\%$ CI, 1.6-3.0; $P \leq .001$; 2+ step: RR, 4.5; $95\%$ CI, 1.8-11.1; $$P \leq .001$$) (Table 3). For example, compared with patients receiving diet therapy alone, patients taking both oral medications and insulin had 4.5 times the rate of a 2+ step increase in DR. Notable characteristics associated with any progression from mild NPDR were longer duration of diabetes (>15 y, RR, 1.8; $95\%$ CI, 1.2-2.5; $$P \leq .002$$), higher HbA1c level (RR, 1.1; $95\%$ CI, 1.0-1.1; $P \leq .001$), and presence of peripheral neuropathy (RR, 1.5; $95\%$ CI, 1.2-2.0; $$P \leq .001$$).
**Table 3.**
| Measures | Incident DR for patients with no DR at baseline (n = 7097) | Incident DR for patients with no DR at baseline (n = 7097).1 | Incident DR for patients with no DR at baseline (n = 7097).2 | Incident DR for patients with no DR at baseline (n = 7097).3 | Mild NPDR at baseline (n = 1277) | Mild NPDR at baseline (n = 1277).1 |
| --- | --- | --- | --- | --- | --- | --- |
| Measures | Any increase (n = 1280) | Any increase (n = 1280) | 2+ step increase (n = 441) | 2+ step increase (n = 441) | Any increase (n = 347) | Any increase (n = 347) |
| Measures | RR (95% CI) | P value | RR (95% CI) | P value | RR (95% CI) | P value |
| Intercept | 0 (0) | <.001 | 0 (0) | <.001 | 22.3 (0-0.1) | <.001 |
| Age, y | 1.0 (1.0-1.0) | .09 | 1.0 (1.0-1.0) | .02 | 1.0 (1.0-1.0) | . <.001 |
| Sex (self-report) | | | | | | |
| Male | 1.2 (1.0-1.3) | .004 | 1.3 (1.1-1.6) | .003 | 1.1 (0.9-1.3) | .41 |
| Female | 1 [Reference] | | 1 [Reference] | | 1 [Reference] | .41 |
| Diabetes duration, y | | | | | | |
| <1 | 0.8 (0.6-1.0) | .03 | 0.5 (0.3-0.9) | .03 | 0.9 (0.5-1.8) | .74 |
| 1-5 | 1 [Reference] | | 1 [Reference] | | 1 [Reference] | |
| 6-10 | 1.4 (1.2-1.6) | <.001 | 2.3 (1.7-3.0) | <.001 | 1.7 (1.2-2.5) | .004 |
| 11-15 | 1.5 (1.3-1.8) | <.001 | 2.5 (1.8-3.4) | <.001 | 2.0 (1.4-2.8) | <.001 |
| >15 | 2.0 (1.7-2.4) | <.001 | 3.2 (2.3-4.4) | <.001 | 1.8 (1.2-2.5) | .002 |
| Not specified | 1.4 (1.2-1.7) | <.001 | 1.7 (1.2-2.6) | .004 | 1.4 (0.9-2.1) | .13 |
| HbA1c, %a | 1.1 (1.1-1.2) | <.001 | 1.3 (1.2-1.3) | <.001 | 1.1 (1.0-1.1) | <.001 |
| Diabetes therapy | | | | | | |
| Diet only | 1 [Reference] | | 1 [Reference] | | 1 [Reference] | |
| Oral medications only | 1.4 (1.1-1.9) | .02 | 3.0 (1.2-7.3) | .014 | 1.8 (0.7-4.3) | .20 |
| Insulin only | 2.1 (1.5-2.9) | <.001 | 4.5 (1.8-11.2) | .001 | 2.3 (0.9-5.5) | .07 |
| Oral medications and insulin | 2.2 (1.6-3.0) | <.001 | 4.5 (1.8-11.1) | .001 | 2.0 (0.8-4.9) | .12 |
| Not specified | 1.5 (1.1-2.0) | .02 | 3.5 (1.4-8.7) | .008 | 1.9 (0.8-4.8) | .15 |
| Other patient diagnoses | | | | | | |
| Hypercholesterolemia | 1.0 (0.9-1.1) | .43 | 0.7 (0.6-0.9) | .009 | 0.8 (0.7-1.0) | .03 |
| Cardiovascular disease | 0.9 (0.8-1.1) | .51 | 0.8 (0.6-1.2) | .37 | 1.1 (0.8-1.5) | .58 |
| Hypertension | 1.1 (1.0-1.2) | .13 | 1.1 (0.9-1.3) | .59 | 0.9 (0.8-1.1) | .53 |
| Peripheral neuropathy | 1.3 (1.1-1.5) | .01 | 1.1 (0.8-1.6) | .56 | 1.5 (1.2-2.0) | .001 |
| Nephropathy | 1.3 (1.0-1.5) | .03 | 1.1 (0.8-1.7) | .50 | 1.2 (0.9-1.7) | .14 |
| Imaging technology | | | | | | |
| NMFP | | | | | | |
| Both examinations | 1 [Reference] | | 1 [Reference] | | 1 [Reference] | |
| First examination, UWFI second | 1.5 (1.2-1.9) | <.001 | 2.2 (1.4-3.5) | .001 | 1.5 (1.0-2.2) | .05 |
| UWFI | | | | | | |
| First examination, NMFP second | 0.9 (0.7-1.3) | .74 | 1.0 (0.5-2.1) | .94 | 1.1 (0.7-1.7) | .82 |
| Both examinations | 1.2 (1.0-1.5) | .04 | 1.9 (1.2-3.0) | .006 | 1.1 (0.8-1.6) | .63 |
For comparison with other studies (Table 4),9,10,11,12,21,22,23,24,25,26,27,28,29,30,31 we conducted several post hoc analyses. These found that of American Indian and Alaska Native patients diagnosed with diabetes before age 30 years and taking insulin alone or with oral medications, $36.9\%$ (90 of 244) developed any new DR within 4 years. Of American Indian and Alaska Native patients diagnosed at 30 years or older and taking insulin, $28.5\%$ (378 of 1324) and $0.1\%$ (1 of 1324) developed any new DR and PDR, respectively. Additionally, a separate regression model with only hypertension and blood pressure medication (yes/no) found that patients with hypertension were $14\%$ more likely to develop new DR and no more or less likely to progress than patients without hypertension.
**Table 4.**
| Sourcea | Study design | Setting/population | Time frame | Results |
| --- | --- | --- | --- | --- |
| Studies of American Indian or Alaska Native individuals | Studies of American Indian or Alaska Native individuals | Studies of American Indian or Alaska Native individuals | Studies of American Indian or Alaska Native individuals | Studies of American Indian or Alaska Native individuals |
| Knowler et al,9 1980 | Prospective cohort study (emphasis on blood pressure) | Pima Indians of Arizona, Gila River Indian Community (n = 188 people with diabetes); 77.0% follow-up | 1965+, 2 Examinations given, 6 y apart (to within 2 y) | Participants not taking insulin: 24% developed hemorrhages, 24% developed exudatesParticipants taking insulin: 56% developed hemorrhages, 64% developed exudates |
| Nelson et al,10 1989 | Longitudinal community-based study with medical examinations | Residents of Gila River, Arizona, who were at least 50% Pima and/or Papagado Indian (n = 953); denominator not reported | 1983-1987, Biennial examinations | 2.6% Incident PDR |
| Lee et al,11 1992 | Cohort follow-up study | IHS clinics serving American Indian individuals within Oklahoma (n = 380); follow-up rate was 41% (total)/73.8% (survivors only) | 1972-1980 Baseline, 1987-1991 follow-up | 72.3% Incident DR15.4% Developed PDR |
| Rith-Najarian et al,12 1993 | Analysis of IHS diabetes registry data | Chippewa Indian individuals and related tribes in northern Minnesota, served by an IHS clinic (n = 429); denominator not reported | 1986-1988 | 2.8% Incident PDR (12 cases/1000 PY) |
| Studies of Hispanic American individuals | Studies of Hispanic American individuals | Studies of Hispanic American individuals | Studies of Hispanic American individuals | Studies of Hispanic American individuals |
| Varma et al,21 2010Population-based, prospective, cohort study (Los Angeles Latino Eye Study) | 6 census tracts in La Puente, California; Latino individuals aged ≥40 (n = 775); follow-up rate for people with 2 fundus examinations not reported;2000-2003 baseline, 2004-2008 follow-up | 34.0% Incident DR38.9% Progressed | 14.0% Improved | If NPDR at first examination: 5.3% had PDR at follow-up and 1.9% had PDR with high-risk characteristics |
| Tudor et al,22 1998 | Geographically based case-control study (San Luis Valley Diabetes Study) | Hispanic and non-Hispanic White individuals residing in San Luis Valley of Colorado (n = 244); 62.7% of people examined originally had at least 1 follow-up | 1984-1988 Baseline, 1988-1992 follow-up | Incident DR: 26.2% For non-Hispanic White individuals (76.1 cases/1000 PY);20.8% For Hispanic individuals (58.3 cases/1000 PY)Progression:27.2% For non-Hispanic White individuals (79.3 cases/1000 PY);23.0% For Hispanic individuals (65.2 cases/1000 PY) |
| Studies predominantly composed of White and Black individuals in the US | Studies predominantly composed of White and Black individuals in the US | Studies predominantly composed of White and Black individuals in the US | Studies predominantly composed of White and Black individuals in the US | Studies predominantly composed of White and Black individuals in the US |
| Klein et al,23 1989 | Population-based study (Wisconsin Epidemiologic Study of Diabetic Retinopathy) | People with diabetes who were diagnosed before age 30 y, taking insulin, in southern Wisconsin (n = 891); follow-up rate was 89.5% | 1980-1982 Baseline with 1984-1986 follow-up | No DR at baseline: 59.0% Incident DR0.4% Developed PDRDR at baseline:10.5% Developed PDR |
| Klein et al,24 1989 | Population-based study (Wisconsin Epidemiologic Study of Diabetic Retinopathy) | People diagnosed with diabetes at age ≥30 y, in southern Wisconsin (n = 987); follow-up rate was 72.0% | July 1, 1979-June 30, 1980, baseline with 1984-1986 follow-up | Not taking insulin: 34.4% Incident DR2.3% New PDRTaking insulin:47.4% Incident DR7.4% New PDR |
| Wong et al,252007 | Population-based, prospective cohort study (Atherosclerosis Risk in Communities Study) | 4 communities in the US, dispersed (n = 981 African American and White participants, with and without diabetes, aged 45-64 y); follow-up rate was 90.5% | 1993-1995 Baseline with 1996 follow-up | Of those with diabetes: 10.0% incident DR |
| International studies | International studies | International studies | International studies | International studies |
| Song et al,26 2011 | Retrospective study | Hong Kong Chinese individuals enrolled in a diabetic retinopathy screening program, aged ≥30 y (n = 5160); denominator not reported | 2005-2009 | 15.2%, 14.5%, 0.7%, and 0.03% Incident any DR, mild NPDR, moderate NPDR, and sight-threatening retinopathy, respectively6.6% With baseline DR progressed and 45.5% regressed |
| Kanjee et al,27 2017 | Retrospective chart analysis | 49 Communities in Northern Manitoba, Canada (n = 4676; 1976 with no evidence of DR at initial examination and examined at least 2 times); 55.2% follow-up for total sample | May 2007-July 2013 | 17.1% Incident DR |
| Burgess et al,28 2017 | Cohort study; hospital-based, primary care diabetes clinic | Southern region of Malawi (n = 135); follow-up rate was 53% | 2007-2012 | 48.4% Incident DR% With PDR at follow-up was: 0% Of those with baseline level 104.5% Of those with baseline level 2022% Of those with baseline level 3040% Of those with baseline level 40 |
| Shani et al,29 2018 | Retrospective, longitudinal, community-based study | Israel (n = 516); 67.8% had a first follow-up examination in specified time frame | 2000-2002 Baseline, at least 1 follow-up examination by end of 2007 | 18.8% and 2.7% Incident NPDR and PDR, respectively |
| Shani et al,29 2018 | Retrospective, longitudinal, community-based study | Israel (n = 516); 67.8% had a first follow-up examination in specified time frame | 2000-2002 Baseline, at least 1 follow-up examination by end of 2007 | 30% Cumulative NPDR (baseline + new incidence) |
| Li et al,30 2020 | Meta-analysis | Europe; 4 studies selected, 2 English and 2 Spanish (n = 71 307 people with type 2 diabetes) | Studies published 1996-2017 | 4.6% Pooled annual DR incidence calculated by meta-regressions0.5% Pooled annual sight-threatening DR and/or maculopathy calculated by meta-regressions |
| Sabanayagam et al,31 2019 | Systematic review | International; 8 studies selected, 5 from Asia, and 1 each from North America, Caribbean, and Sub-Saharan Africa | Studies published after 2000 | Annual incidence of DR ranged from 2.2% to 12.7% and progression to PDR from 3.4% to 12.3% |
## UWFI and DR Outcomes
UWFI for follow-up or both examinations was associated with any DR incidence (RR, 1.2; $95\%$ CI, 1.0-1.5; $$P \leq .04$$) and a 2+ step increase (RR, 1.9; $95\%$ CI, 1.2-3.0; $$P \leq .006$$) in patients with no DR at baseline. For example, compared with patients imaged with NMFP at both examinations, patients imaged with UWFI at follow-up had 2.2 times ($95\%$ CI, 1.4-3.5; $$P \leq .001$$) the rate of a 2+ step increase in DR (Table 3). UWFI was associated with DR progression when used for the follow-up examination in patients with mild NPDR at baseline.
## Discussion
To prevent or reduce the damaging effects of diabetes complications in American Indian and Alaska Native individuals, the IHS implemented programs such as the Special Diabetes Program for Indians (SPDI) and this American Telemedicine Association Category 3 teleophthalmology program. The SDPI has increased access to diabetes treatment services and reduced hyperglycemia, blood lipid levels, and kidney failure. HbA1c level decreased from $9.0\%$ in 1996 to $8.0\%$ in 2020, low-density lipoprotein cholesterol level decreased from 118 mg/dL in 1998 to 89 mg/dL in 2020 (to convert cholesterol to millimole per liter, multiply by 0.0259), and kidney failure decreased $54\%$ between 1996 and 2013.32 The teleophthalmology program itself conducted 264 437 examinations of 120 075 patients between January 1, 2000, and October 31, 2021. Substantial changes in diabetes medications have occurred in the past 25 years as well.33 Coinciding with this expansion of diabetes programs and medications, the prevalence of diabetic eye disease in American Indian and Alaska Native individuals served by the IHS teleophthalmology program appears to have declined.7,8 This article updates estimates of incidence and progression in this population. Eighteen percent of patients (1280 of 7097) with no evidence of DR in 2015 developed some level of DR during 2016 to 2019, $6.2\%$ (441 of 7097) had a 2+ step increase, and $0.1\%$ (10 of 7097) developed PDR. The estimates reported here are lower than previous estimates in American Indian and Alaska Native patients.9,10,11,12 Estimates from this study are also lower than or similar to estimates of DR incidence in Hispanic American individuals.21,22 The target populations and methods used in other US-based studies make comparisons to this study problematic.23,24,25 For comparison with the benchmark Wisconsin Epidemiologic Study of Diabetic Retinopathy (WESDR),23,24 we conducted an additional analysis of DR incidence by age of diabetes diagnosis (before or after 30 years) plus whether the American Indian and Alaska Native patients were taking insulin. The percentages of incident any DR and PDR that we found were lower than those reported in WESDR articles with a similar follow-up period of approximately 4 years.
The estimates from this study are more comparable with studies conducted after 2000 and outside the US.26,27,29,31 For example, in Hong Kong, incident any DR was $15.2\%$ and sight-threatening DR was $0.03\%$ in Chinese people surveilled with digital fundus photography.26 Approximately $2.2\%$ of patients (28 of 1277) in this study who had mild NPDR at baseline progressed to PDR in the examined time frame compared with $0.1\%$ of patients (10 of 7097) who had no DR at baseline, which is consistent with a recent systematic analysis of international studies.31 The influence of most of the risk factors examined on the development of DR was as expected, with duration of diabetes, hyperglycemia, and therapeutic regimen being significant contributors. Additionally, UWFI at the follow-up examination detected more incident DR and more progression of DR, perhaps reflecting the identification of predominantly peripheral lesions (PPL).34 The lack of a net effect for hypertension in this study was inconsistent with prior research.6,9 A more refined measure of hypertension (such as actual blood pressure measurements) may be needed to understand its association with DR changes in this cohort; however, that information is not currently in the program’s database.
In this national, primary care–based program, UWF fluorescein angiography (UWF-FA) is not available to identify PPL or extent of retinal capillary nonperfusion on FA. Newly published results from the DRCR Retina Network found that, over 4 years, greater initial retinal nonperfusion and FA PPL on UWF-FA were statistically associated with worsening DR as measured by 2+ step progressions or DR treatment.35 Based on those findings, more incident DR and DR progression may have been found if UWF-FA was used in this program.
## Strengths and Limitations
A strength of this study includes the geographical distribution of the American Indian and Alaska Native cohort and the large sample size. Previous studies focused on specific areas and had smaller sample sizes. However, the data are exclusively from the IHS, which has a user population of 2.56 million,36 representing approximately $25\%$ of the total estimated 9.7 million American Indian and Alaska Native individuals in the US.37 Generalizations from this report should be restricted to American Indian and Alaska Native individuals served by the IHS.
There are study limitations to acknowledge. This study focused on DR, omitting DME, another leading cause of vision loss in people with diabetes. Thus, this study likely underestimated the overall burden of diabetic eye disease incidence for American Indian and Alaska Native patients. However, we believe that the underestimate is modest. First, a recent previous report found that DME prevalence in American Indian and Alaska Native patients was $3.0\%$,8 using clinical data from UWFI. Aiello and colleagues38 found that UWFI has a low sensitivity for the detection of DME compared with spectral-domain optical coherence tomography, which suggests that DME prevalence estimates derived from UWFI may contain false positives. Second, a recent meta-analysis30 of *European data* showed that annual incidence of DME was $0.4\%$. Omission of DME in this study may have underestimated the incidence of diabetic eye diseases overall by approximately $1.6\%$ to $2.0\%$. The percentage might be lower still because some patients with incident DME may have also had incident DR and were already counted in the estimate.
Another potential limitation of this study is that $61.2\%$ of the total patients evaluated by the program in the baseline period had an examination during the follow-up period; ie, $38.8\%$ were not reexamined in 2016 to 2019. This follow-up rate is lower than several studies,9,23,24,25 but similar to22 or higher than11,28 other studies. Some retrospective studies do not report a denominator, precluding comparison of this study’s follow-up rate with theirs. To understand the implications of $38.8\%$ attrition for the results, we compared the baseline characteristics of patients who were not reexamined with those who were and found that the groups were similar except that the followed cohort was slightly less healthy. The estimates reported here likely are reasonable even with the attrition rate.
## Conclusions
The results of this cohort study suggest that recent DR incidence and progression among American Indian and Alaska Native individuals served by the IHS are substantially lower than they were 30 or more years ago and are now comparable with estimates from non–American Indian and Alaska Native populations examined in the last 20 years. Further, these low rates support the viability of safely extending the follow-up interval for retinopathy assessment in IHS patients who have no evidence of DR or mild NPDR. This may be possible if the IHS patients also have no DME, have minimal risk factors, will be examined with UWFI, and their follow-up adherence is not jeopardized. Currently, the IHS teleophthalmology program recommends annual DR examinations, consistent with the American Academy of Ophthalmology Preferred Practice Patterns39 and those of other professional organizations, but biennial frequency as recommended by the American Diabetes Association40 is well documented and might be an appropriate frequency for the IHS as well. Such a practice change, however, requires examination of adherence to the current recommendations. If a practice change extending follow-up were implemented, further research would be needed to determine if the change affected vision outcomes and adherence rates.
## References
1. Bullock A, Sheff K, Hora I. **Prevalence of diagnosed diabetes in American Indian and Alaska Native adults, 2006-2017**. *BMJ Open Diabetes Res Care* (2020.0) **8**. DOI: 10.1136/bmjdrc-2020-001218
2. 2Centers for Disease Control and Prevention. National Diabetes Statistics Report 2020. Accessed January 5, 2023. https://www.cdc.gov/diabetes/pdfs/data/statistics/national-diabetes-statistics-report.pdf
3. 3National Indian Council on Aging. Diabetes still highest among AI/AN. Accessed January 5, 2023. https://www.nicoa.org/diabetes-still-highest-among-ai-an/
4. Acton KJ, Burrows NR, Moore K, Querec L, Geiss LS, Engelgau MM. **Trends in diabetes prevalence among American Indian and Alaska native children, adolescents, and young adults**. *Am J Public Health* (2002.0) **92** 1485-1490. DOI: 10.2105/AJPH.92.9.1485
5. Varma R, Vajaranant TS, Burkemper B. **Visual impairment and blindness in adults in the US: demographic and geographic variations from 2015 to 2050**. *JAMA Ophthalmol* (2016.0) **134** 802-809. DOI: 10.1001/jamaophthalmol.2016.1284
6. Fong DS, Aiello L, Gardner TW. **Retinopathy in diabetes**. *Diabetes Care* (2004.0) **27** S84-S87. DOI: 10.2337/diacare.27.2007.S84
7. Bursell SE, Fonda SJ, Lewis DG, Horton MB. **Prevalence of diabetic retinopathy and diabetic macular edema in a primary care-based teleophthalmology program for American Indians and Alaskan Natives**. *PLoS One* (2018.0) **13**. DOI: 10.1371/journal.pone.0198551
8. Fonda SJ, Bursell SE, Lewis DG, Clary D, Shahon D, Silva PS. **Prevalence of diabetic eye diseases in American Indians and Alaska Natives (AI/AN) as identified by the Indian Health Service’s national teleophthalmology program using ultra-widefield imaging (UWFI)**. *Ophthalmic Epidemiol* (2022.0) **29** 672-680. DOI: 10.1080/09286586.2021.1996611
9. Knowler WC, Bennett PH, Ballintine EJ. **Increased incidence of retinopathy in diabetics with elevated blood pressure—a 6-year follow-up study in Pima Indians**. *N Engl J Med* (1980.0) **302** 645-650. DOI: 10.1056/NEJM198003203021201
10. Nelson RG, Wolfe JA, Horton MB, Pettitt DJ, Bennett PH, Knowler WC. **Proliferative retinopathy in NIDDM. Incidence and risk factors in Pima Indians**. *Diabetes* (1989.0) **38** 435-440. DOI: 10.2337/diab.38.4.435
11. Lee ET, Lee VS, Kingsley RM. **Diabetic retinopathy in Oklahoma Indians with NIDDM. Incidence and risk factors**. *Diabetes Care* (1992.0) **15** 1620-1627. DOI: 10.2337/diacare.15.11.1620
12. Rith-Najarian SJ, Valway SE, Gohdes DM. **Diabetes in a northern Minnesota Chippewa Tribe—prevalence and incidence of diabetes and incidence of major complications, 1986-1988**. *Diabetes Care* (1993.0) **16** 266-270. DOI: 10.2337/diacare.16.1.266
13. Fonda SJ, Bursell SE, Lewis DG, Clary D, Shahon D, Horton MB. **The Indian Health Service primary care-based teleophthalmology program for diabetic eye disease surveillance and management**. *Telemed J E Health* (2020.0) **26** 1466-1474. DOI: 10.1089/tmj.2019.0281
14. Horton MB, Brady CJ, Cavallerano J. **Practice guidelines for ocular telehealth-diabetic retinopathy, 3rd edition**. *Telemed J E Health* (2020.0) **26** 495-543. PMID: 32209018
15. **Grading diabetic retinopathy from stereoscopic color fundus photographs—an extension of the modified Airlie House classification: ETDRS report number 10**. *Ophthalmology* (1991.0) **98** 786-806. DOI: 10.1016/S0161-6420(13)38012-9
16. Silva PS, Walia S, Cavallerano JD. **Comparison of low-light nonmydriatic digital imaging with 35-mm ETDRS 7 standard-field stereo color fundus photographs and clinical examination**. *Telemed J E Health* (2012.0) **18** 492-499. DOI: 10.1089/tmj.2011.0232
17. Silva PS, Cavallerano JD, Sun JK, Noble J, Aiello LM, Aiello LP. **Nonmydriatic ultra-widefield retinal imaging compared with dilated standard 7-field 35-mm photography and retinal specialist examination for evaluation of diabetic retinopathy**. *Am J Ophthalmol* (2012.0) **154** 549-559.e2. DOI: 10.1016/j.ajo.2012.03.019
18. Porta M. *A Dictionary of Epidemiology, 5th ed* (2008.0)
19. Chen W, Qian L, Shi J, Franklin M. **Comparing performance between log-binomial and robust Poisson regression models for estimating risk ratios under model misspecification**. *BMC Med Res Methodol* (2018.0) **18** 63. DOI: 10.1186/s12874-018-0519-5
20. 20R Core Team. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing; 2021.. *R: A Language and Environment for Statistical Computing* (2021.0)
21. Varma R, Choudhury F, Klein R, Chung J, Torres M, Azen SP. **Four-year incidence and progression of diabetic retinopathy and macular edema: the Los Angeles Latino Eye Study**. *Am J Ophthalmol* (2010.0) **149** 752-61.e1, 3. DOI: 10.1016/j.ajo.2009.11.014
22. Tudor SM, Hamman RF, Baron A, Johnson DW, Shetterly SM. **Incidence and progression of diabetic retinopathy in Hispanics and non-Hispanic Whites with type 2 diabetes—San Luis Valley Diabetes Study, Colorado**. *Diabetes Care* (1998.0) **21** 53-61. DOI: 10.2337/diacare.21.1.53
23. Klein R, Klein BE, Moss SE, Davis MD, DeMets DL. **The Wisconsin Epidemiologic Study of Diabetic Retinopathy, IX; 4-year incidence and progression of diabetic retinopathy when age at diagnosis is less than 30 years**. *Arch Ophthalmol* (1989.0) **107** 237-243. DOI: 10.1001/archopht.1989.01070010243030
24. Klein R, Klein BE, Moss SE, Davis MD, DeMets DL. **The Wisconsin Epidemiologic Study of Diabetic Retinopathy, X: 4-year incidence and progression of diabetic retinopathy when age at diagnosis is 30 years or more**. *Arch Ophthalmol* (1989.0) **107** 244-249. DOI: 10.1001/archopht.1989.01070010250031
25. Wong TY, Klein R, Amirul Islam FM. **Three-year incidence and cumulative prevalence of retinopathy: the Atherosclerosis Risk in Communities study**. *Am J Ophthalmol* (2007.0) **143** 970-976. DOI: 10.1016/j.ajo.2007.02.020
26. Song H, Liu L, Sum R, Fung M, Yap MK. **Incidence of diabetic retinopathy in a Hong Kong Chinese population**. *Clin Exp Optom* (2011.0) **94** 563-567. DOI: 10.1111/j.1444-0938.2011.00628.x
27. Kanjee R, Dookeran RI, Mathen MK, Stockl FA, Leicht R. **Six-year prevalence and incidence of diabetic retinopathy and cost-effectiveness of teleophthalmology in Manitoba**. *Can J Ophthalmol* (2017.0) **52** S15-S18. DOI: 10.1016/j.jcjo.2017.09.022
28. Burgess PI, Harding SP, García-Fiñana M. **Incidence and progression of diabetic retinopathy in Sub-Saharan Africa: a 5-year cohort study**. *PLoS One* (2017.0) **12**. DOI: 10.1371/journal.pone.0181359
29. Shani M, Eviatar T, Komaneshter D, Vinker S. **Diabetic retinopathy—incidence and risk factors in a community setting: a longitudinal study**. *Scand J Prim Health Care* (2018.0) **36** 237-241. DOI: 10.1080/02813432.2018.1487524
30. Li JQ, Welchowski T, Schmid M. **Prevalence, incidence, and future projection of diabetic eye disease in Europe: a systematic review and meta-analysis**. *Eur J Epidemiol* (2020.0) **35** 11-23. DOI: 10.1007/s10654-019-00560-z
31. Sabanayagam C, Banu R, Chee ML. **Incidence and progression of diabetic retinopathy: a systematic review**. *Lancet Diabetes Endocrinol* (2019.0) **7** 140-149. PMID: 30005958
32. 32Department of Health and Human Services. Fiscal year 2022—Indian Health Service: justification of estimates for appropriations committees. Accessed January 4, 2023. https://www.ihs.gov/sites/budgetformulation/themes/responsive2017/display_objects/documents/FY_2022.pdf
33. White JR. **A brief history of the development of diabetes medications**. *Diabetes Spectr* (2014.0) **27** 82-86. DOI: 10.2337/diaspect.27.2.82
34. Marcus DM, Silva PS, Liu D. **Association of predominantly peripheral lesions on ultra-widefield imaging and the risk of diabetic retinopathy worsening over time**. *JAMA Ophthalmol* (2022.0) **140** 946-954. DOI: 10.1001/jamaophthalmol.2022.3131
35. Silva PS, Marcus DM, Liu D. **Association of ultra-widefield fluorescein angiography-identified retinal nonperfusion and the risk of diabetic retinopathy worsening over time**. *JAMA Ophthalmol* (2022.0) **140** 936-945. DOI: 10.1001/jamaophthalmol.2022.3130
36. 36Indian Health Service. IHS Profile. Accessed January 4, 2023. https://www.ihs.gov/newsroom/factsheets/ihsprofile/
37. Jones N, Marks R, Ramirez R, Rios-Vargas M
38. Aiello LP, Jacoba CMP, Sun JK, Silva PS. **Integrating macular optical coherence tomography with ultra-widefield imaging in a diabetic retinopathy telemedicine program using a single device**. *Invest Ophthalmol Vis Sci* (2021.0) **62** 1941
39. Flaxel CJ, Adelman RA, Bailey ST. **Diabetic retinopathy preferred practice pattern**. *Ophthalmology* (2020.0) **127** 66-P145. DOI: 10.1016/j.ophtha.2019.09.025
40. Solomon SD, Chew E, Duh EJ. **Diabetic retinopathy: a position statement by the American Diabetes Association**. *Diabetes Care* (2017.0) **40** 412-418. DOI: 10.2337/dc16-2641
|
---
title: Patient, carer and healthcare professional perspectives on increasing calorie
intake in Amyotrophic Lateral Sclerosis
authors:
- Elizabeth Coates
- Nicolò Zarotti
- Isobel Williams
- Sean White
- Vanessa Halliday
- Daniel Beever
- Gemma Hackney
- Theocharis Stavroulakis
- David White
- Paul Norman
- Christopher McDermott
journal: Chronic Illness
year: 2021
pmcid: PMC9999280
doi: 10.1177/17423953211069090
license: CC BY 4.0
---
# Patient, carer and healthcare professional perspectives on increasing calorie intake in Amyotrophic Lateral Sclerosis
## Abstract
### Objectives
Research suggests that higher Body Mass *Index is* associated with improved survival in people with Amyotrophic Lateral Sclerosis (pwALS). Yet, understanding of the barriers and enablers to increasing calorie intake is limited. This study sought to explore these issues from the perspective of pwALS, informal carers, and healthcare professionals.
### Methods
Interviews with 18 pwALS and 16 informal carers, and focus groups with 51 healthcare professionals. Data were analysed using template analysis and mapped to the COM-B model and Theoretical Domains Framework (TDF).
### Results
All three COM-B components (Capability, Opportunity and Motivation) are important to achieving high calorie diets in pwALS. Eleven TDF domains were identified: Physical skills (ALS symptoms); Knowledge (about high calorie diets and healthy eating); Memory, attention, and decision processes (reflecting cognitive difficulties); Environmental context/resources (availability of informal and formal carers); Social influences (social aspects of eating); Beliefs about consequences (healthy eating vs. high calorie diets); Identity (interest in health lifestyles); Goals (sense of control); Reinforcement (eating habits); and Optimism and Emotion (low mood, poor appetite).
### Discussion
To promote high calorie diets for pwALS, greater clarity around the rationale and content of recommended diets is needed. Interventions should be tailored to patient symptoms, preferences, motivations, and opportunities.
## Introduction
Amyotrophic lateral sclerosis (ALS), also commonly known as motor neuron disease (MND)1, is a devastating neurodegenerative disorder characterised by the loss of motor neurons, which causes progressive paralysis and eventually death.1 *Onset is* typically focal (limb, bulbar, or respiratory), later spreading elsewhere in the body. Since there is no cure for ALS, treatment focuses on slowing progression and managing symptoms, including malnutrition and weight loss.2 When the muscles involved in swallowing are affected, it can become difficult for people with ALS (pwALS) to eat and drink enough to sustain adequate nutritional intake, thereby placing them at high risk of developing malnutrition.2 These issues are also exacerbated by hyper-metabolism,3 with resting energy expenditure being on average $20\%$ higher than in healthy individuals.4 Weight loss at diagnosis is a negative prognostic indicator in ALS, and nutritional parameters typically worsen with disease progression.5 A population based study found that two thirds of pwALS presented with weight loss at diagnosis, with the risk of death increasing by $23\%$ for every $10\%$ increase in weight loss.6 Similarly, a recent systematic review has shown that having a higher Body Mass Index (BMI) at diagnosis is associated with greater long-term survival in pwALS.7 Another study showed that, among pwALS whose BMI had decreased by more than 2 points per year after diagnosis, only $10\%$ were still alive after two years, while $60\%$ of those whose weight had decreased by 2 points or less (or remained unchanged or increased) were still alive. These findings suggest that weight loss before or after diagnosis with ALS, is associated with shorter survival.
Therefore, increasing weight through modifying behaviours that maximise calorie intake is potentially a key therapeutic goal for healthcare professionals (HCPs) treating pwALS, particularly in the early stages of the disease. With regard to this, the use of high-calorie fatty diets in pwALS was recently tested in a placebo-controlled RCT,8 but there was no significant effect on weight loss or survival. However, post hoc analysis and further analysis of neurofilament light chains levels in study blood samples have suggested a survival and biological effect of a high-caloric fatty diet on pwALS with fast progressing disease.9 Another study evaluating support from a dietician or an mHealth app did not result in any significant differences in weight changes compared to standard care.10 These mixed results may be due to the oral nutritional interventions not being tailored to patients, as well as low adherence, high drop-out and study design. Thus, further research examining the psychosocial and physical barriers and enablers to increasing calorie intake in pwALS is needed to develop more effective interventions.
In addition, previous research on nutrition in ALS has mostly focussed on problems resulting from dysphagia11 or enteral feeding,12 as opposed to addressing issues with oral nutritional interventions (‘food first approach’) which typically comes before enteral feeding.13 One qualitative study has investigated psychosocial issues related to dysphagia for pwALS from the caregiver perspective,14 reporting how dysphagia can change mealtime experiences of caregivers and pwALS due to fear of choking, frustration with being unable to prevent weight loss, use of avoidance as a coping strategy, and a desire to maintain normality. Another qualitative study explored the psychological factors influencing nutritional management of pwALS from the HCP perspective.15 This emphasised the importance of ALS-specific knowledge of nutrition, psychosocial aspects of eating and drinking, early engagement and psychological adjustment, as well as promoting perceived control over decision-making. However, these previous studies have not considered the views of patients and how they may differ to those of carers and HCPs.
Consequently, the present study aimed to provide a more comprehensive understanding of the barriers and enablers to high calorie diets in ALS, from the perspectives of pwALS, caregivers and HCPs. In addition, the present study sought to draw on the COM-B (Capability, Opportunity, Motivation, Behaviour) model of behaviour change16 as an overarching theoretical framework to interpret the key barriers and enablers. The COM-B model proposes that behaviour change is driven by a person's Capability (psychological or physical capability to enact a behaviour), Opportunity (the physical or social environment that enables or inhibits behaviour), and Motivation (reflective or automatic mechanisms that guide behaviour). The model has been used to identify barriers and enablers to a wide range of health-related behaviours, including physical activity in overweight and obese pregnant women17 and various studies looking at nutritional behaviours.18–20 The Theoretical Domains Framework (TDF)21 represents an elaboration of the COM-B components into 14 domains, each reflecting a different influence on behaviour: [1] knowledge, [2] skills, [3] social/professional role and identity, [4] beliefs about capabilities, [5] optimism, [6] beliefs about consequences, [7] reinforcement, [8] intentions, [9] goals, [10] memory, attention, and decision processes, [11] environmental context and resources, [12] social influences, [13] emotion, and [14] behavioural regulation. Together, the COM-B and the TDF provide a theoretical framework to understanding factors that could be targeted in oral nutritional interventions to promote increased calorie intake in pwALS. The adoption of these frameworks was deemed important in light of their successful implementation in previous research around nutritional behaviours in several populations,18–20 as well as to facilitate the development of a new intervention to support increased calorie intake by people with ALS.
## Setting and design
A multi-centre qualitative study was carried out in eight MND centres in secondary care National Health Service (NHS) hospitals in the UK. HCPs from all centres took part in focus groups, while patient and carers were recruited from five of these centres for individual or joint interviews.
## Eligibility criteria
HCPs involved in nutritional management of pwALS were eligible to take part, including medics, specialist nurses, dietitians, speech and language therapists (SLTs), physiotherapists, occupational therapists (OTs), and psychologists. Patients were eligible to participate if they were aged 18 years or over and had received a diagnosis of ALS. Additional inclusion criteria included the progressive muscular atrophy variant where appropriate investigation excluded mimics of ALS, clinician judgement indicating suitability of patient to take part, and capacity to give informed consent and fluency in English. Patient exclusion criteria were: co-morbidity that would affect survival or metabolic state (e.g. unstable thyroid disease or diabetes mellitus) and BMI ≥ 35 kg/m2. Anyone acting as one of the main carers for pwALS was eligible to participate.
## Sampling technique
The eight NHS MND centres were sampled purposively based on geographical location, hospital size, and service configuration. Within each centre, individual staff were recruited for focus groups (FGs) using convenience sampling to reflect team composition, staff availability, and willingness to participate. Where feasible, the objective was to achieve representation from all clinical disciplines present in local teams, with up to a maximum of eight participants in each FG. Patients were sampled purposively from five of these centres to capture variation in terms of age, gender, time since diagnosis, carer presence, stage of the disease, and presence of eating and drinking difficulties.
## Recruitment
Local gatekeepers in each centre made the initial approach to HCPs by introducing the study during a staff meeting or via email. Staff members could opt-in by contacting the research team directly or providing consent for the gatekeeper to pass on contact details. The research team then contacted all volunteers to provide them with more study information and FG logistics.
Clinical staff from participating centres identified and recruited patients via current caseloads, clinics, and local advertising. Patients were asked to opt-in by contacting the research team directly or giving consent for the clinical team to pass on contact details. Members of the research team then contacted potential participants to provide more information and confirm eligibility. Following confirmation of eligibility and based on the purposive sampling strategy, patients and carers were contacted to organise a mutually convenient time and place for the interview. Where participants had significant communication difficulties, they were sent a communication support plan to complete, to ensure that they were adequately supported during interviews. A simplified version of the interview schedule was also sent to all participants beforehand.22
## Ethics
Ethical approval for this study was granted by the North West – Greater Manchester East NHS Research Ethics Committee (ref: 18/NW/0638), and governance approval was granted by the Health Research Authority (ref: 250732). Access to research sites was granted via local NHS Research and Development departments. All participants provided informed consent to take part.
## Participants and procedure
In total, 51 HCPs took part in eight FGs that were conducted across the eight MND centres (see Table 1). The number of participants in each FG varied from five to nine, and the composition of each varied in line with local team membership and their availability to participate. Discussions were completed face-to-face on NHS property sites convenient to participants. The median length of the focus groups was 62 min (58–75 min). See details of FG discussion guide in Supplementary Material 1.
**Table 1.**
| Focus group | Participants | Specialism attended | Specialism attended.1 | Specialism attended.2 | Specialism attended.3 | Specialism attended.4 | Specialism attended.5 | Specialism attended.6 | Specialism attended.7 | Specialism attended.8 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Focus group | Participants | Doctor | Nurse | Dietitian | Speech and Language Therapist | Occupational Therapist | Physiotherapist | Community Outreach | Psychology | Care Co-ordinator |
| 1 | 7 | 1 | 1 | 1 | 1 | 1 | 0 | 1 | 1 | 1 |
| 2 | 7 | 1 | 3 | 1 | 1 | 0 | 1 | 0 | 0 | 0 |
| 3 | 9 | 1 | 1 | 3 | 2 | 1 | 0 | 0 | 0 | 1 |
| 4 | 6 | 0 | 1 | 4 | 1 | 0 | 0 | 0 | 0 | 0 |
| 5 | 5 | 3 | 1 | 1 | 0 | 0 | 0 | 0 | 0 | 0 |
| 6 | 6 | 0 | 2 | 2 | 0 | 0 | 1 | 0 | 0 | 1 |
| 7 | 5 | 0 | 0 | 1 | 1 | 1 | 0 | 0 | 0 | 1 |
| 8 | 6 | 2 | 1 | 2 | 0 | 0 | 0 | 0 | 0 | 1 |
| Total | 51 | 8 | 10 | 15 | 6 | 3 | 2 | 1 | 1 | 5 |
Thirty-four pwALS and carers were interviewed, comprising 20 individual interviews (eleven patients and nine carers) and seven joint interviews. These took place face-to-face at their homes or on hospital premises. The median length was 40 min (25–60 min) for individual interviews and 71 min (59–81 min) for joint ones. Details of the interview schedule are in Supplementary Material 2. The recruitment flow chart is provided in Supplementary Material 3.
Table 2 outlines the demographic and nutritional characteristics of pwALS ($$n = 18$$). Sixty-one percent of participants were male, median 67 years old, and generally in their second year post-diagnosis. Two-thirds ($67\%$) of patients reported limb onset ALS, $61\%$ reported dysphagia and $28\%$ reported placement of an enteral feeding (PEG: percutaneous endoscopic gastronomy) tube, although all patients were still able to take food by mouth. The majority reported reasonably good mobility ($61\%$). Participants’ median BMI was 25 kg/m2, and there was a fairly even split between those with stable weight and those who experienced weight loss or gain. Most participants had received some nutritional advice, could still eat until they were satisfied, and felt that they took longer to finish their meals since diagnosis. Around half of participants reported modifying their food texture and fewer followed an enriched diet.
**Table 2.**
| Characteristic | Value |
| --- | --- |
| Median age (min-max) | 67.0 (40.0-75.0) |
| Female N (%) | 7 (38.9) |
| Median months since diagnosis (min-max) | 17.0 (11.0-84.0) |
| ALS onset N (%) | |
| Limb | 12 (66.7%) |
| Bulbar | 6 (33.3%) |
| Reported dysphagia N (%) | 11 (61.1%) |
| PEG placement N (%) | 5 (27.8%) |
| Reported comorbidities N (%) | 3 (16.7) |
| Reported mobility N (%) | |
| Good | 11 (61.1) |
| Average | 3 (16.6) |
| Poor | 4 (22.2) |
| Median Body Mass Index (min-max) | 24.7 (15.70-28.0) |
| Weight change (%) | |
| Reduced | 6 (33.3) |
| Stable | 7 (38.9) |
| Increased | 5 (27.8) |
| Nutritional intervention (%) | |
| Has received support/advice | 14 (77.8) |
| Follows an enriched diet | 7 (38.8) |
| Modifies food texture | 8 (44.4) |
| Takes longer to finish meals | 11 (61.1) |
All data were collected between 11 December 2018 and 7 March 2019. Data collection stopped when data saturation was reached. All interviews and focus groups were conducted by NZ, IW and EC, were recorded using a digitally encrypted recording device, and transcribed verbatim.
Data analysis Data were analysed with a template analysis approach,23 using NVivo® software to support data processing. For the purposes of analysis, the dataset was divided into [1] FGs with HCPs and [2] patient and carer individual and dyad interviews. The template analysis was completed by EC, NZ and IW, following the six key stages, i.e., [1] familiarisation with the datasets by reviewing all the transcripts; [2] identifying preliminary codes in the data and [3] organising them into a hierarchy or structure, and then using this to [4] develop an initial coding template, which was then [5] applied to the dataset and developed iteratively to lead to the last stage [6] where the template was finalised and applied across the dataset.
Following completion of the template analysis and identification of codes for both datasets, the codes were deductively mapped to the TDF21 and COM-B model16 by PN (and double mapped by EC) to help structure the interpretation of the identified barriers and enablers.
## Results
Fifty-three codes regarding barriers and enablers to increasing calorie intake in pwALS emerged from both datasets (Table 3). These are presented below, structured around the COM-B model components and TDF domains. Specific codes are highlighted in italics and illustrated with quotes from HCPs, pwALS (P) and carers (C) given in Supplementary Material 4.
**Table 3.**
| COM-B Component | TDF Domain | Code | Barrier or enabler | P/Cs | HCPs |
| --- | --- | --- | --- | --- | --- |
| Capability - Physical | Physical (skill) | Swallow/dysphagia | B | ✓ | ✓ |
| Capability - Physical | Physical (skill) | Chewing | B | ✓ | ✓ |
| Capability - Physical | Physical (skill) | Weakness and fatigue | B | ✓ | ✓ |
| Capability - Physical | Physical (skill) | Capacity to eat | B | ✓ | ✓ |
| Capability - Physical | Physical (skill) | Capacity to cook or shop | B | ✓ | ✓ |
| Capability - Physical | Physical (skill) | Choking or aspiration | B | ✓ | ✓ |
| Capability - Physical | N/A (Disease characteristic) | Weight | B | ✓ | ✓ |
| Capability - Physical | N/A (Disease characteristic) | Body shape changes | B | ✓ | ✓ |
| Capability - Physical | N/A (Disease characteristic) | Taste | B | ✓ | ✓ |
| Capability - Physical | N/A (Disease characteristic) | Salivation or secretions | B | ✓ | ✓ |
| Capability - Physical | N/A (Disease characteristic) | Heterogeneity of changes | B | ✓ | ✓ |
| Capability - Physical | N/A (Disease characteristic) | Mobility | B | ✓ | ✓ |
| Capability - Physical | N/A (Disease characteristic) | Breathing | B | ✓ | ✓ |
| Capability - Physical | N/A (Disease characteristic) | Lip seal | B | ✓ | |
| Capability - Physical | N/A (Disease characteristic) | Other health needs | B | | ✓ |
| Capability - Psychological | Knowledge | Knowledge about ‘healthy eating’ | B&E | ✓ | ✓ |
| Capability - Psychological | Knowledge | Knowledge about high calorie diets | B&E | ✓ | ✓ |
| Capability - Psychological | Knowledge | Lack of guidance | B | | ✓ |
| Capability - Psychological | Memory, attention and decision processes | Cognitive difficulties | B | ✓ | ✓ |
| Capability - Psychological | Memory, attention and decision processes | Comprehension of healthcare professional advice or support | B&E | ✓ | |
| Capability - Psychological | Memory, attention and decision processes | Overwhelmed at diagnosis | B | | ✓ |
| Opportunity – Physical | Environmental context/resources | Availability of informal carers or support | B&E | ✓ | ✓ |
| Opportunity – Physical | Environmental context/resources | Availability of formal carers or support | B&E | ✓ | ✓ |
| Opportunity – Physical | Environmental context/resources | Availability of peer support network | B&E | ✓ | |
| Opportunity – Physical | Environmental context/resources | Availability of food | B&E | ✓ | |
| Opportunity – Physical | Environmental context/resources | Availability of healthcare professional advice or support | B&E | ✓ | ✓ |
| Opportunity – Physical | Environmental context/resources | Lack of care continuity and geographical variation | B | | ✓ |
| Opportunity - Social | Social influences | Social aspects of eating | B | ✓ | ✓ |
| Opportunity - Social | Social influences | Influence of informal carers or support | B&E | ✓ | ✓ |
| Opportunity - Social | Social influences | Influence of formal carers or support | B&E | ✓ | ✓ |
| Opportunity - Social | Social influences | Influence of peer support network | B&E | ✓ | |
| Opportunity - Social | Social influences | Influence of healthcare professional advice or support | B&E | ✓ | ✓ |
| Opportunity - Social | Social influences | Delivery of person centred care | E | | ✓ |
| Opportunity - Social | Social influences | Building relationships with patients | E | | ✓ |
| Motivation – Reflective | Beliefs about consequences | Beliefs about healthy eating | B&E | ✓ | ✓ |
| | | Beliefs about high calorie diets | B&E | ✓ | ✓ |
| | Identity | Interest in healthy lifestyles | B | ✓ | ✓ |
| | | Acceptance of and adjustment to diagnosis | B&E | ✓ | ✓ |
| | Goals | Body weight goals | B | ✓ | |
| | | Adherence / receptivity to healthcare professional advice | B&E | ✓ | ✓ |
| | | Sense of control | B&E | ✓ | ✓ |
| | | Independence | B&E | ✓ | ✓ |
| | | Patient priorities (i.e. not food) | B | | ✓ |
| | | Importance of patient choice and compromise | B&E | | ✓ |
| | Optimism | Living in the present | E | ✓ | ✓ |
| | | Uncertainty about future | B | ✓ | ✓ |
| Motivation – Automatic | Reinforcement | Eating habits and routines | B&E | ✓ | ✓ |
| | Emotion | Appetite and thirst | B | ✓ | ✓ |
| | | Food enjoyment | B&E | ✓ | ✓ |
| | | Resistance and denial | B | ✓ | ✓ |
| | | Low mood | B | ✓ | ✓ |
| | | Embarrassment | B | ✓ | ✓ |
| | | Carer burden | B | ✓ | ✓ |
## Physical skills and disease characteristics
A number of physical changes were identified by pwALS, carers, and HCPs as barriers to increasing calorie intake. These included swallowing difficulties (dysphagia) causing difficulties with particular food textures; breathing difficulties which may increase the risk of aspiration or choking; general muscle weakness, fatigue and reduced mobility, which can affect the conduct of everyday nutritional behaviours, such as shopping, cooking, handling cutlery, eating, or sitting upright. Other physical changes included excessive salivation and secretions, chewing difficulties, issues with lip seal, loss of sense of taste, weight loss and body shape changes.
HCPs consistently highlighted the heterogeneous nature of the physical changes caused by ALS. This reflects the characteristics of different disease onsets, but also the heterogeneity of symptoms and disease course, which can exacerbate the challenge of initiating nutritional interventions in a timely fashion.
## Knowledge
Patients’ and carers’ knowledge about food and nutritional behaviours served as both a barrier and enabler to increasing calorie intake of pwALS, with the contrast between individual knowledge of healthy eating and ALS recommended diets being difficult for many to reconcile.
This was more of an issue for those participants with a greater interest in healthy lifestyles (see ‘Reflective Motivation’). For others, there was a willingness to increase the calorie value of their diets in line with HCP advice, albeit amidst confusion about the appropriateness of this approach. HCPs consistently demonstrated their knowledge of the potential benefit of high-calorie diets for pwALS, and explained the rationale for this approach, relating it to their knowledge of current research.
In contrast, some HCPs critiqued the evidence and, despite their support for increasing calories, they were keen to convey both its strengths and weaknesses during patient consultations. Related to this, HCPs reported a general lack of guidance specific to the nutritional management of pwALS. Although NICE guidelines (both MND care and nutrition support) and MND Association standards for care were mentioned, many HCPs still relied on professional experience to inform practice. Whilst HCPs were typically confident about their level of expertise, not directly identifying this as a barrier, the differences in care caused by a lack of clear guidance were acknowledged (see ‘Physical Opportunity’), especially in terms of addressing potential unhelpful beliefs.
## Memory, attention and decision processes
Another challenge identified by HCPs related to cognitive difficulties affecting some patients’ capacity to comprehend dietetic advice alongside all of the other healthcare advice imparted to them, particularly if they are overwhelmed at diagnosis. HCPs were conscious of how overwhelming the time post-diagnosis can be, and the challenges of educating patients about high-calorie diets when their priorities may lie elsewhere.
## Environmental context/resources
The presence of informal carers (partners, family members, friends) was another important enabler of nutritional behaviours identified by HCPs, pwALS, and carers themselves. Where available, support from informal carers was important to the availability of food and conduct of shopping, cooking and eating, although this sometimes came at the cost of carer burden (see ‘Automatic Motivation’). As a potential solution for some patients, formal carers (including visiting care assistants, nurses, or live-in carers) provided other essential sources of support.
However, for pwALS who did not have access to family or peer support, and for whom formal care or living in a care home are the only options, the caring experience may vary considerably based on income or financial availability, especially in terms of control over nutritional choices.
*More* generally, some pwALS and their carers also highlighted the importance of a wider peer support network and how this can help facilitate oral nutritional behaviours.
Notwithstanding the difficulties posed by the contrast between healthy eating and ALS dietary advice, the availability of healthcare professional advice and support was key to promoting or impeding behaviours to increase calorie intake. Some patients reported that they had not received any advice or were unable to recall this, and that they felt unsupported. This issue was corroborated by HCPs, who recognised the challenges of delivering timely, person-centred nutritional care within a ‘window of opportunity’ (FG3, FG4), given the speed of deterioration of ALS: an issue which was compounded in those centres with inadequate funding for staff time. This issue was exacerbated by the time-consuming nature of ALS care, for which a typical appointment duration was described as insufficient.
HCPs also identified several other issues that impede continuity of care and create geographical variation in services offered. These included inconsistent team composition (absence of dietitian), variation in funding available for nutritional management interventions (adaptive cutlery, oral nutritional supplements), limited specialist knowledge beyond the multi-disciplinary team (e.g. non-specialist neurological departments, GPs or community services), or the presence of special interest in ALS by staff, which could act as both a strength and limitation of teams. The passion and commitment to ALS of some HCPs was duly recognised, but so was the vulnerability of the staffing arrangement.
## Social influences
Corresponding with their availability, the influence of informal or formal carers was another major influence (both positive and negative) on nutritional behaviours identified by all participants. Specifically, the wider peer support network was identified as a facilitator of social opportunities for eating and drinking by pwALS and carers.
*More* generally, the social aspects of eating and drinking, such as socialising, eating together as families or visiting restaurants and cafes were highlighted during all FGs. HCPs spoke about the loss they see for pwALS and their families as the disease progresses. Carers and pwALS also explained how disease progression can hamper the social aspects of eating, such as meal sharing or eating out.
The influence of health professional advice and support was also key for pwALS, despite varied levels of receptivity and acceptance of the diagnosis (see ‘Reflective Motivation’). Adopting a proactive approach to nutrition was often described by HCPs as another enabling aspect of patient-centred care for pwALS. This included the chance to offer dietetic and speech and language therapy advice before developing problems with eating and drinking, as well as undertaking frequent home visits and sharing insights across the Multi-Disciplinary Team. Introducing ideas proactively and sensitively was believed to provide a platform for future care discussions when patients might be more ready to make changes to nutritional behaviours. Correspondingly, most HCPs spoke about the facilitative value of building relationships with patients and their families to help understand priorities as well as promoting patient choice and compromise.
## Beliefs about consequences
A major aspect which could enable or inhibit changes to nutritional behaviours was differing beliefs about healthy eating versus high calorie diets to promote good health. These were largely in line with the knowledge of pwALS, their carers, and HCPs (see ‘Psychological Capability’). However, when in conflict, reconciling them could prove challenging, making it difficult for pwALS and their carers to adopt higher calorie diets.
## Identity and goals
Increasing calorie intake was often seen as problematic for those pwALS who used to be very active and for whom living a ‘healthy lifestyle’ and meeting desired body weight goals was previously an important part of their identity.
HCPs also reported a range of patient responses to being encouraged to consume more calories, from the positive, to the confused or resistant, and they related this to patients’ perspectives on healthy lifestyles.
Despite some positive feedback of promoting high calorie diets, HCPs in each FG also spoke about the challenges of encouraging adherence to high-calorie diets by pwALS who have strong beliefs about healthy lifestyles. They also spoke about the facilitative potential of patient choice and compromise by working closely with them to understand their priorities, knowledge, and beliefs about food.
Many patients and carers were keen to express their receptivity to advice and support from HCPs, but this varied considerably among participants. The contrast between ‘healthy eating’ and high calorie dietary advice for pwALS, or the absence of regular or timely support, were influential factors.
Another theme related to pwALS’ acceptance of the ALS diagnosis and lack of psychological adjustment to the need to change nutritional behaviours, which presented a potential barrier to achieving a high calorie diet. Conversely, others talked about feelings of control and independence, and how exercising agency or decisional input over some aspects of nutrition may be an empowering experience, at least whilst they were still able to maintain their independence.
## Optimism
Levels of optimism about the diagnosis and the future also influenced how pwALS approached their nutritional management. For example, where pwALS approached their illness in a more positive way or were more accepting of their condition, they were typically more keen to make changes to their diets, and vice versa.
## Reinforcement
Eating habits and routines relating to food choices and household roles in shopping and cooking were other influences on the nutritional behaviours of pwALS. Food preferences can be very difficult to change in line with dietary advice as the disease progresses, as this can affect the whole household. Many participants described changes to roles in the maintenance of food-related activities which could prove challenging to both the pwALS and carer, regardless of their domestic roles pre-diagnosis. Relinquishing responsibility for shopping and cooking could be upsetting to pwALS regardless of gender, as well as a source of tension in relationships.
## Emotion
The physical changes that pwALS experience can also impact on appetite and thirst, and impede efforts to maintain a high calorie intake. Physical symptoms also impacted on the enjoyment of food and drinks. Although many pwALS were able to continue eating and drinking as usual, others found this less enjoyable over time, and this was a source of loss.
Some participants also talked about feelings of resistance and denial, from both patients and carers, which could interfere with increasing calorie intake. HCPs also noted how challenging it can be to engage more resistant individuals.
Low mood, coupled with the embarrassment of challenges with eating or drinking in public or social settings was identified as another barrier to achieving high calorie diets.
Finally, emotional responses from spouses, family members, and friends – and the carer burden of providing long-term support to pwALS – were also mentioned as barriers to changing eating behaviours. More specifically, asking pwALS to change their dietary intake and seeing the person deteriorate can place a substantial burden on informal carers, triggering low mood or anxiety, which in turn could limit their capacity to support changes to food and drink.
## Summary of main findings
This study identified a complex range of factors that influence the uptake of high calorie diets among pwALS. To our knowledge, this is the first qualitative investigation to specifically triangulate the views of pwALS, carers, and HCPs on high calorie diets in ALS. The use of the COM-B model16 and TDF21 was helpful in structuring the analysis and interpretation of barriers and enablers. To achieve high-calorie diets for pwALS, the analysis highlighted the need to address all three COM-B components and 11 TDF domains.
In relation to the ‘Capability - Physical’ component, many physical symptoms associated with the progression of ALS affect people's eating and drinking capabilities over time. In line with evidence documenting the risks of weight loss in ALS,7 most current research has focussed upon clinical management of these symptoms, including dysphagia11 or enteral feeding.12 Whilst it is important not to diminish the impact of the physical progression of the disease, our analysis also highlighted the importance of social and psychological influences on eating and drinking behaviours in ALS.
The ‘Capability – Psychological’ component captured the importance of knowledge about healthy eating and high calorie diets. Whilst some pwALS were keen to increase their intake as they enjoyed calorie-dense foods, others struggled to make those changes due to concerns about the impact on general physical health, influenced by pre-existing views about healthy eating. Despite some evidence that high calorie diets may be beneficial, along with considerable support for this approach amongst HCPs in this study, the limits of the evidence base and a lack of specific supporting guidance represents a challenge for current practice. This finding is consistent with recent evidence charting the provision of ALS nutritional management with HCPs.15, 24 Within the ‘Opportunity – Physical’ and ‘Opportunity – Social’ components, the availability and influence of both formal healthcare provision and informal caring networks were highlighted. A lack of care continuity and geographical variation, and the facilitative power of person-centred care, have also previously been documented in a survey of ALS HCPs.24 The importance of case management in ALS has also recently been highlighted.25, 26 Similarly, the significant role played by informal carers in ALS is well documented,27, 28 but our study sheds light on how they contribute in relation to eating and drinking.
Considering the ‘Motivation – Reflective’ and ‘Motivation – Automatic’ components, this study also identified a number of psychological factors. These included beliefs about healthy eating and high-calorie diets, the varied importance of pre-diagnosis identity and lifestyle, as well as adjustment to diagnosis and engagement with healthcare interventions. Moreover, pwALS, their carers, and HCPs drew attention to the impact of different goals, levels of optimism, existing habits, and routines on behaviour. Finally, a number of emotional barriers to changing behaviour, such as emotional burden upon carers and low mood, resistance, denial, and embarrassment amongst pwALS were documented. Whilst new within the context of psychosocial issues related to eating and drinking, these findings are consistent with previous evidence on the general impact of dysphagia in pwALS.14, 15 Likewise, there is a body of work which considers the psychological burden of ALS on both people with the disease and carers.28
## Strengths and limitations
This is the first study to provide an in-depth exploration of the barriers and enablers to increasing calorie intake in pwALS within a relatively large qualitative sample of patients, carers, and HCPs in the UK. This will aid the development of effective interventions to prevent weight loss and potentially increase survival in pwALS.
Nonetheless, this study has some limitations. In particular, the self-selecting sample may have introduced some biases. It is possible that we have not accurately captured the impact of some aspects, such as patient shock at diagnosis and disengagement from healthcare provision, as those pwALS are unlikely to participate in research. Nonetheless, participants highlighted a wide range of factors affecting dietary behaviour. Moreover, the insights of HCPs, who spoke at length about the challenges of changing eating behaviour in resistant or disengaged patients, may also give confidence that a full range of important barriers were identified (notwithstanding the possibility that disengaged patients may have different views about these issues to HCPs). Finally, while it is possible that we mainly captured the perspectives of the HCPs most engaged with ALS nutritional care, as the HCPs were self-selecting, we nonetheless recruited a broadly representative sample of staff involved in the nutritional management of pwALS.
## Implications for clinical practice
Our results clearly outline the need for nutritional management interventions for pwALS which can be tailored in order to accommodate differing symptoms, knowledge, beliefs, and nutritional preferences, as well as the environmental contexts/resources and social influences at play. In tailoring interventions, HCPs must also consider where pwALS are psychologically – in terms of acceptance of the diagnosis, adjustment to symptoms, and other emotional responses29 – because, without this wider perspective, our findings suggest that the effectiveness of interventions will be restricted. Future research should therefore focus on developing and evaluating complex interventions that are able to address the wide range of barriers in the oral nutritional management of pwALS. We believe our findings provide a valuable starting point in this direction.
## References
1. McDermott CJ, Shaw PJ. **Diagnosis and management of motor neurone disease**. *Br Med J* (2008.0) **336** 658-662. PMID: 18356234
2. Leigh PN. **THE MANAGEMENT OF MOTOR NEURONE DISEASE**. *J Neurol Neurosurg Psychiatry* (2003.0) **74** 32-47
3. He J. **Hypermetabolism associated with worse prognosis of amyotrophic lateral sclerosis**. *J. Neurol* (2021.0) **2021**
4. Genton L, Viatte V, Janssens JP. **Nutritional state, energy intakes and energy expenditure of amyotrophic lateral sclerosis (ALS) patients**. *Clin Nutr* (2011.0) **30** 553-559. PMID: 21798636
5. Marin B. **Alteration of nutritional status at diagnosis is a prognostic factor for survival of amyotrophic lateral sclerosis patients**. *J Neurol Neurosurg Psychiatry* (2011.0) **82** 628-634. PMID: 21097551
6. Janse Van Mantgem MR. **Prognostic value of weight loss in patients with amyotrophic lateral sclerosis: a population-based study**. *J Neurol Neurosurg Psychiatry* (2020.0) **91** 867-875. PMID: 32576612
7. Ning P. **Systematic review of the prognostic role of body mass index in amyotrophic lateral sclerosis**. *Amyotroph. Lateral Scler. Front. Degener* (2019.0) **20** 356-367
8. Ludolph AC. **Effect of high-caloric nutrition on survival in amyotrophic lateral sclerosis**. *Ann. Neurol* (2020.0) **87** 206-216. PMID: 31849093
9. Dorst J. **Effect of high-caloric nutrition on serum neurofilament light chain levels in amyotrophic lateral sclerosis**. *J Neurol Neurosurg Psychiatry* (2020.0) **91** 1007-1009. PMID: 32788256
10. Wills AM. **Nutritional counseling with or without mobile health technology: a randomized open-label standard-of-care-controlled trial in ALS**. *BMC Neurol* (2019.0) **19**
11. Onesti E. **Dysphagia in amyotrophic lateral sclerosis: impact on patient behavior, diet adaptation, and riluzole management**. *Front. Neurol* (2017.0) **8**
12. Stavroulakis T. **The impact of gastrostomy in motor neurone disease: challenges and benefits from a patient and carer perspective**. *BMJ Support. Palliat. Care* (2016.0) **6** 52-59
13. White S. **The nutritional management of people living with amyotrophic lateral sclerosis: a national survey of dietitians**. *J. Hum. Nutr. Diet* (2021.0). DOI: 10.1111/JHN.12900
14. Lisiecka D, Kelly H, Jackson J. **‘This is your golden time. You enjoy it and you’ve plenty time for crying after’: how dysphagia impacts family caregivers of people with amyotrophic lateral sclerosis – A qualitative study**. *Palliat. Med* (2020.0) **34** 1097-1107. PMID: 32552499
15. Zarotti N. **Health care professionals’ views on psychological factors affecting nutritional behaviour in people with motor neuron disease: a thematic analysis**. *Br. J. Health Psychol* (2019.0) **24** 953-969. PMID: 31449739
16. Michie S, van Stralen MM, West R. **The behaviour change wheel: a new method for characterising and designing behaviour change interventions**. *Implement. Sci* (2011.0) **6** 42. PMID: 21513547
17. Flannery C. **Enablers and barriers to physical activity in overweight and obese pregnant women: an analysis informed by the theoretical domains framework and COM-B model**. *BMC Pregnancy Childbirth* (2018.0) **18** 178. PMID: 29783933
18. Bentley MRN, Mitchell N, Sutton L. **Sports nutritionists’ perspectives on enablers and barriers to nutritional adherence in high performance sport: a qualitative analysis informed by the COM-B model and theoretical domains framework**. *J. Sports Sci* (2019.0) **37** 2075-2085. PMID: 31124393
19. Grady A.. **Barriers and enablers to implementation of dietary guidelines in early childhood education centers in Australia: application of the theoretical domains framework**. *J. Nutr. Educ. Behav* (2018.0) **50** 229-237.e1. PMID: 29170057
20. Timlin D, McCormack JM, Simpson EEA. **Using the COM-B model to identify barriers and facilitators towards adoption of a diet associated with cognitive function (MIND diet)**. *Public Health Nutr* (2021.0) **24** 1657-1670. PMID: 32799963
21. Cane J, O’Connor D, Michie S. **Validation of the theoretical domains framework for use in behaviour change and implementation research**. *Implement. Sci* (2012.0) **7** 37. PMID: 22530986
22. Stavroulakis T. **Factors influencing decision-making in relation to timing of gastrostomy insertion in patients with motor neurone disease**. *BMJ Support. Palliat. Care* (2014.0) **4** 57-63
23. Brooks J, McCluskey S, Turley E. **The utility of template analysis in qualitative psychology research**. *Qual. Res. Psychol* (2015.0) **12** 202-222. PMID: 27499705
24. Halliday V. **Delivery of nutritional management services to people with amyotrophic lateral sclerosis (ALS)**. *Amyotroph. Lateral Scler. Front. Degener* (2021.0) 350-359. DOI: 10.1080/21678421.2021.1874991
25. Bakker M. **Need and value of case management in multidisciplinary ALS care: a qualitative study on the perspectives of patients, spousal caregivers and professionals**. *Amyotroph. Lateral Scler. Front. Degener* (2015.0) **16** 180-186
26. 26NICE. Motor Neurone Disease: Assessment and Management NICE Guidelines (NG42). NICE 2019. Accessed at: https://www.nice.org.uk/guidance/ng42/resources/motor-neurone-disease-assessment-and-management-pdf-1837449470149 .
27. Galvin M. **Caregiving in ALS - A mixed methods approach to the study of burden**. *BMC Palliat. Care* (2016.0) **15** 81. PMID: 27596749
28. Harris M. **Supporting wellbeing in motor neurone disease for patients, carers, social networks, and health professionals: a scoping review and synthesis**. *Palliative and Supportive Care* (2018.0) **16** 228-237. PMID: 28829002
29. Zarotti N, Mayberry E, Ovaska-Stafford N. **Psychological interventions for people with motor neuron disease: a scoping review**. *Amyotroph. Lateral Scler. Front. Degener* (2021.0) 1-11. DOI: 10.1080/21678421.2020.1788094
|
---
title: Alterations in circulating lipidomic profile in patients with type 2 diabetes
with or without non-alcoholic fatty liver disease
authors:
- Assim A. Alfadda
- Abdulrahman M. Almaghamsi
- Suphia Murad Sherbeeni
- Adel N. Alqutub
- Abdullah S. Aldosary
- Arthur C. Isnani
- Nasser Al-Daghri
- Simon D. Taylor-Robinson
- Rukhsana Gul
journal: Frontiers in Molecular Biosciences
year: 2023
pmcid: PMC9999296
doi: 10.3389/fmolb.2023.1030661
license: CC BY 4.0
---
# Alterations in circulating lipidomic profile in patients with type 2 diabetes with or without non-alcoholic fatty liver disease
## Abstract
Objective: Non-alcoholic fatty liver disease (NAFLD) and Type 2 diabetes mellitus (T2DM) often coexist and drive detrimental effects in a synergistic manner. This study was designed to understand the changes in circulating lipid and lipoprotein metabolism in patients with T2DM with or without NAFLD.
Methods: Four hundred thirty-four T2DM patients aged 18–60 years were included in this study. Fatty liver was assessed by FibroScan. The comprehensive metabolic lipid profiling of serum samples was assessed by using high-throughput proton NMR metabolomics.
Results: *Our data* revealed a significant association between steatosis and serum total lipids in VLDL and LDL lipoprotein subclasses, while total lipids in HDL subclasses were negatively associated. A significant positive association was found between steatosis and concentration of lipids, phospholipids, cholesterol, and triglycerides in VLDL and LDL subclasses, while HDL subclasses were negatively associated. Furthermore, a significant, association was observed between fibrosis and concentrations of lipids, phospholipids, cholesterol, and triglycerides in very small VLDL, large, and very large HDL subclasses. Subgroup analysis revealed a decrease in the concentrations of lipids, phospholipids, cholesterol, and other lipid biomolecules in patients using antilipemic medications.
Conclusion: The metabolomics results provide evidence that patients with T2DM with higher steatosis grades have altered lipid metabolomics compared to patients without steatosis. Increased lipid, phospholipids, cholesterol, and triglycerides concentration of VLDL and LDL subclasses are associated with steatosis in patients with T2DM.
## Introduction
Non-alcoholic fatty liver disease (NAFLD) is the most common hepatic condition detected in patients with Type 2 diabetes (T2DM) with an estimated prevalence of $65\%$ among the Saudi population (Alfadda et al., 2022). The fatty liver primarily known as steatosis is defined by the presence of >$5\%$ of fat infiltration in hepatocytes (Contos and Sanyal, 2002). Fatty liver can progress to a more severe condition non-alcoholic steatohepatitis (NASH) causing hepatocyte inflammation which can finally progress to permanent end stage liver disease such as cirrhosis and hepatocellular carcinoma (Angulo et al., 1999).
Steatosis or fatty liver is linked with dyslipidemia due to the increased synthesis and deposition of lipids in the hepatocytes which attribute to the formation of lipid droplets (Cohen et al., 2011). This increased intrahepatic fat accumulation causes lipotoxicity, comorbidity associated with fatty liver resulting from an increase in the levels of fatty acids in the serum that flow from peripheral adipose tissues to the liver (Wasilewska and Lebensztejn, 2021). The increased fat accumulation in fatty liver is a result of the imbalance in fatty acid uptake, increased production of lipids via hepatic de novo lipogenesis, inadequate hepatic fat export through very‐low‐density lipoproteins (VLDLs) and its oxidation of fatty acids (Dowman et al., 2010). An increase in lipotoxicity in NAFLD due to impaired fatty acid oxidation activate inflammatory signaling that attributes to hepatocyte injury and are accountable for the progression to NASH (Trovato et al., 2014).
We have recently revealed the prevalence NAFLD in a Saudi cohort of patients with T2DM and demonstrated the associations between fatty liver and dyslipidemia (Alfadda et al., 2022). Our data showed that controlled attenuation parameter (CAP) values are positively correlated with triglycerides, and negatively with high-density lipoprotein (HDL). Given the importance of serum lipoproteins in NAFLD, here we aimed to investigate the alterations in serum lipids and lipoprotein subclasses using high-throughput proton nuclear magnetic resonance (NMR) spectroscopy metabolomics in patients with T2DM with or without steatosis. Recent studies have expanded knowledge regarding the lipid profiling in fatty liver and revealed the associations between serum lipidome and NAFLD (Kaikkonen et al., 2017). However, to the best of our knowledge no studies have been conducted using NMR-metabolomic approach to investigate the changes in circulating lipids and lipoproteins in Saudi patients with T2DM with or without NAFLD.
## Study population
This study included patients who participated in the Cohort of Non-alcoholic Fatty Liver Disease in Saudis with T2DM (the CORDIAL Study). This prospective cohort study started in 2015 and recruited patients from King Fahad Medical City (KFMC) and affiliated Primary Care Centers in Riyadh, Saudi Arabia. The cohort was approved by the Institutional Review Board at KFMC (study number: 12–344), and all patients provided written, informed consent prior to recruitment. The study was conducted in accordance with the ethical principles for medical research on human subjects adopted by the 18th World Medical Association General Assembly, and the Declaration of Helsinki 1964 and its subsequent amendments. The inclusion criteria included Saudi patients aged 18–60 years who were diagnosed with T2DM and followed up regularly in the diabetes or primary care clinics. Patients were excluded if they tested positive for hepatitis B surface antigen or had antibodies against hepatitis C virus, were diagnosed with other chronic liver diseases (e.g., hemochromatosis, primary biliary cholangitis, or autoimmune hepatitis), known to have pre-existing hepatic or extrahepatic malignancy, or were consuming >20 g of alcohol per day. The patients will be prospectively followed for 10 years and assessed for hepatic, metabolic, renal, and cardiovascular complications.
## Sample size calculation
Using sampling formula for a single cross-sectional survey: Sample size = Z1−α/2 2 p (1−p)/d2, where Z1−a/2 = is the standard normal variate (at $5\%$ type 1 error ($p \leq 0.05$) or $1\%$ type 1 error ($p \leq 0.01$). As in majority of studies p values are considered significant below 0.05 hence 1.96 is used in formula, p = expected proportion in population and d = the absolute error or precision. Based on $95\%$ confidence interval (1.96) and absolute precision of $5\%$. Using a previously reported estimated all ages prevalence of NAFLD of $24.8\%$ in Saudi Arabia (Alswat et al., 2018), the calculated sample size was 287. To account for loss of cases, it was decided to include 434 patients.
## Clinical and laboratory data collection
The participants’ characteristics and anthropometric indices, including age, sex, body weight, height, body mass index (BMI), and blood pressure, were obtained. BMI was calculated as body weight (kg) divided by body height (m2). Blood was sampled for laboratory assays after the patients had fasted for ten to 12 h overnight. Fasting blood glucose and serum lipids were measured using Abbott—Architect Plus, a clinical chemistry autoanalyzer (Abbott, Abbott Park, IL, United States). Glycated hemoglobin (HbA1c) determination was performed using D-100®, a high-performance liquid chromatography analyzer (Bio-Rad Laboratories, Hercules, CA, United States).
## Liver FibroScan examination
FibroScan® 502 and FibroScan® 530 Compact, with two probes - Medium (M+) and Extra-large (XL+) (Echosens Ltd., Paris, France) were used for measuring CAP—as surrogate measure of liver fat content, and liver stiffness measurement (LSM)—as a surrogate measure of hepatic fibrosis. The device estimates liver steatosis in decibel/meter (dB/m) and liver stiffness in kilopascal (kPa). CAP and LSM were obtained simultaneously in each examination. The type of probe required for each participant was selected by an automatic probe selection tool embedded within the FibroScan® operating software. A successful vibration controlled transient elastography (VCTE) exam was defined by the acquisition of ten successful measurements, where the interquartile range of the LSM did not exceed $30\%$ of the median LSM. Therefore, an “uninterpretable” VCTE examination encompassed failures on one or both accounts. Each patient underwent VCTE examination after 3 h of fasting. All VCTE examinations were performed by two experienced physicians. The optimal cut-off values for classifying steatosis grades were as follows: S0 (CAP <248 dB/m), no steatosis; S1 (CAP 248 to <268 dB/m), mild steatosis; S2 (CAP 268 to <280 dB/m), moderate steatosis; and (S3 CAP ≥280 dB/m), severe steatosis (Karlas et al., 2017), and the optimal cut-off values for classifying fibrosis grades were: F0-F1 (<7.9 kPa), no fibrosis; F2 (7.9 to <8.8 kPa), moderate fibrosis; F3 (8.8 to <11.7 kPa), severe fibrosis; and F4 (≥11.7 kPa), liver cirrhosis (Abeysekera et al., 2020).
## Quantitative NMR metabolic profiling
Metabolic biomarkers were quantified from serum samples using untargeted high-throughput proton nuclear magnetic resonance (NMR) spectroscopy metabolomics platform (Nightingale Health Plc, Helsinki, Finland). The details of the methodology used have been described previously (Soininen et al., 2015). The samples were barcoded for sample identification and kept frozen at −80°C for analysis. Metabolites were measured by a quantitative high-throughput NMR experimental set up for the simultaneous quantification of lipids and lipoprotein subclass profiling in 350 µL of serum. All liquid handling procedures were completed prior to the NMR studies, and the SampleJet robotic sample charger was set up at a cooled temperature to prevent sample deterioration. Every single metabolic measurement was subjected to a number of statistical quality control procedures and cross-referenced with a sizable collection of quantitative molecular data.
## Statistical analysis
Data were analyzed using the Statistical Package for Social Sciences (SPSS) version 23.0 (SPSS Inc., IBM, Armonk, New York, United States). Test of normality of distribution was carried out using the Shapiro-Wilk test. The results were expressed as numbers and percentages (categorical variables) and as mean, standard deviation, and minimum and maximum for continuous variables. Measurement of the strength and direction of the relationship/correlation between two continuous and categorical variables in normally distributed data was performed using the Pearson correlation test and the chi-square (X2) test, respectively. An independent student t-test was performed to determine the difference between two means. One-way analysis of variance (ANOVA) with Tukey’s post hoc analysis was used to determine significant differences in the means of the laboratory results according to grades of steatosis. All p-values were two-tailed, and statistical significance was set at $p \leq 0.05.$
## Demographic and baseline laboratory characteristic of the patients
In total, 434 patients aged 18–60 years diagnosed with T2DM (227, $52.3\%$ males and 207, $47.7\%$ females) were included in this analysis. The mean age of all patients was 50.1 ± 7.6 years. The mean BMI was 32.64 ± 5.7 kg/m2, where 287 patients ($66.1\%$) were diagnosed with obesity (BMI ≥30 kg/m2). The duration of diabetes ranged from 1 to 41 years (mean = 10.66 ± 7.5 years). Figure 1 shows the box and whisker plots of baseline anthropometrics and laboratory tests.
**FIGURE 1:** *Box and whisker plots of baseline anthropometrics and laboratory tests for 434 patients with T2DM.*
In terms of medication use 217 ($50\%$) were on oral hypoglycemic agents mainly Metformin, 80 ($18.4\%$) were on insulin, 161 ($37.1\%$) were on antihypertensive and 203 ($46.7\%$) were on antilipemic medication.
FibroScan showed out of 434 patients the proportion of patients with no steatosis (S0), mild (S1), moderate (S2) and severe steatosis (S3) were 89 ($20.5\%$) (CAP <248 dB/m), 41 ($9.4\%$) (CAP≥248 to <268 dB/m), 24 ($5.5\%$) (CAP≥268 to <280 dB/m), 280 ($64.5\%$) (CAP≥280 dB/m) respectively. Conversely, the proportion of patients without (F0-F1) or with (F2-F4) fibrosis were 398 ($91.7\%$) (LSM<7.9 kPa) and 36 ($8.2\%$) (LSM≥7.9 kPa) respectively. Supplementary Table S1 shows the lipidomic profile for all 434 patients with T2DM. A total of 81 lipid types, species and biomolecules were identified and analyzed. These included total lipids in lipoprotein particles, total phospholipids, total cholesterol and triglycerides in lipoprotein particles, phosphoglycerides, total cholines, phosphatidylcholines, sphingomyelins, apolipoproteins B and A1, and fatty acids. Lipidomic profile for all patients are shown in Supplementary Table S1.
## Association between liver steatosis and lipid biomolecules
Table 1 shows the association between lipid biomolecules and steatosis using the CAP values measured in dB/m. Steatosis was positively associated with total lipids in lipoprotein particles ($r = 0.113$, $$p \leq 0.018$$), including total lipids in chylomicrons and extremely large VLDL, and total lipids in very large, large, medium and small size VLDL particles ($r = 0.211$, $p \leq 0.001$, $r = 0.225$, $p \leq 0.001$, $r = 0.224$, $p \leq 0.001$, $r = 0.183$, $p \leq 0.001$, and $r = 0.152$, $$p \leq 0.001$$, respectively). Total lipids in medium LDL was also positively associated with CAP ($r = 0.094$, $$p \leq 0.050$$). Total lipids in very large HDL and large HDL were negatively associated with CAP (r = −0.239, $p \leq 0.001$ and r = −0.207, $p \leq 0.001$, respectively), while total lipids in small HDL was positively associated with CAP ($r = 0.189$, p = ˂0.001). For phospholipids in lipoprotein particles, CAP was positively associated with phospholipids in chylomicrons and extremely large VLDL, and large, medium, and small VLDL ($r = 0.204$, $p \leq 0.001$, $r = 0.221$, $p \leq 0.001$, $r = 0.222$, $p \leq 0.001$, $r = 0.150$, $$p \leq 0.002$$, $r = 0.117$, $$p \leq 0.014$$). On the other hand, phospholipids in very large, large and small HDL were negatively associated with CAP (r = −0.232, $p \leq 0.001$, r = −0.191, $p \leq 0.001$, r = −0.175, $p \leq 0.001$, respectively). Positive association was observed between CAP and cholesterol in chylomicrons and extremely large VLDL, very large VLDL, and large VLDL ($r = 0.213$, $p \leq 0.001$, $r = 0.209$, $p \leq 0.001$, and $r = 0.208$, $p \leq 0.001$, respectively). On the other hand, negative associations were observed between CAP and cholesterol in HDL, very large HDL, large HDL and small HDL (r = −0.124, $$p \leq 0.010$$, r = −0.256, $p \leq 0.001$, r = −0.229, $p \leq 0.001$, r = −0.160, $$p \leq 0.001$$, respectively). Triglycerides and triglycerides in chylomicrons and extremely large to small subclasses of VLDL, and medium to small subclasses of LDL were positively associated with CAP ($p \leq 0.05$). The ratio of triglycerides to phosphoglycerides was positively associated with CAP ($r = 0.255$, $p \leq 0.001$). Other lipid biomolecules that were found to be associated with CAP were total fatty acids, omega 3 and omega 6 fatty acids, polyunsaturated fatty acids (PUFA), monounsaturated fatty acids (MUFA) and saturated fatty acids (SFA) ($r = 0.185$, $p \leq 0.001$, $r = 0.119$, $$p \leq 0.014$$, $r = 0.100$, $$p \leq 0.038$$, $r = 0.111$, $$p \leq 0.021$$, $r = 0.196$, $p \leq 0.001$ and $r = 0.211$, $p \leq 0.001$, respectively).
**TABLE 1**
| Lipid biomolecules | Association with steatosis CAP (dB/m) r (p-value) | S0 CAP <248 dB/m | S1 CAP 248 to <268 dB/m | S2 CAP 268 to <280 dB/m | S3 CAP ≥280 dB/m | ANOVA p values | Post-hoc analysis (p-value) |
| --- | --- | --- | --- | --- | --- | --- | --- |
| Lipid biomolecules | Association with steatosis CAP (dB/m) r (p-value) | N = 89 | N = 41 | N = 24 | N = 280 | ANOVA p values | |
| Total lipids in lipoprotein particles | 0.113 (0.018) | 9.36 (2.1) | 9.09 (1.9) | 9.30 (1.7) | 9.60 (2.2) | 0.432 | S0-S3 (0.001) |
| Total lipids in chylomicrons extremely large VLDL | 0.211 (<0.001) | 0.200 (0.23) | 0.093 (0.13) | 0.389 (0.50) | 0.233 (0.20) | 0.001 | S0-S3 (0.001) |
| Total lipids in very large VLDL | 0.225 (<0.001) | 0.21 (0.14) | 0.22 (0.18) | 0.27 (0.21) | 0.31 (0.23) | 0.001 | S0-S3 (0.001) |
| Total lipids in large VLDL | 0.224 (<0.001) | 0.36 (0.2) | 0.37 (0.2) | 0.43 (0.3) | 0.49 (0.3) | 0.001 | S0-S3 (0.002) |
| Total lipids in large VLDL | 0.224 (<0.001) | 0.36 (0.2) | 0.37 (0.2) | 0.43 (0.3) | 0.49 (0.3) | 0.001 | S1/S3 (0.044) |
| Total lipids in medium VLDL | 0.183 (<0.001) | 0.63 (0.2) | 0.57 (0.3) | 0.64 (0.2) | 0.71 (0.2) | 0.006 | S1-S3 (0.016) |
| Total lipids in small VLDL | 0.152 (0.001) | 0.439 (0.2) | 0.416 (0.1) | 0.428 (0.1) | 0.473 (0.1) | 0.046 | S1-S3 (0.038) |
| Total lipids in very small VLDL | 0.017 (0.731) | 0.369 (0.12) | 0.350 (0.09) | 0.333 (0.07) | 0.362 (0.10) | 0.470 | NS |
| Total lipids in IDL | −0.052 (0.278) | 1.26 (0.4) | 1.17 (0.3) | 1.15 (0.3) | 1.17 (0.3) | 0.141 | NS |
| Total lipids in large LDL | 0.019 (0.698) | 1.78 (0.4) | 1.66 (0.4) | 1.67 (0.4) | 1.72 (0.4) | 0.392 | NS |
| Total lipids in medium LDL | 0.094 (0.050) | 0.778 (0.2) | 0.723 (0.2) | 0.747 (0.2) | 0.780 (0.2) | 0.329 | NS |
| Total lipids in small LDL | 0.082 (0.088) | 0.349 (0.1) | 0.325 (0.1) | 0.338 (0.1) | 0.349 (0.1) | 0.309 | NS |
| Total lipids in HDL | −0.067 (0.166) | 3.01 (0.6) | 3.08 (0.5) | 3.04 (0.6) | 2.93 (0.5) | 0.196 | NS |
| Total lipids in very large HDL | −0.239 (<0.001) | 0.132 (0.07) | 0.125 (0.04) | 0.118 (0.05) | 0.105 (0.05) | 0.001 | S0-S3 (0.001) |
| Total lipids in large HDL | −0.207 (<0.001) | 0.55 (0.2) | 0.56 (0.1) | 0.53 (0.3) | 0.46 (0.2) | 0.001 | S0-S3 (0.004) |
| Total lipids in large HDL | −0.207 (<0.001) | 0.55 (0.2) | 0.56 (0.1) | 0.53 (0.3) | 0.46 (0.2) | 0.001 | S1-S3 (0.041) |
| Total lipids in medium HDL | −0.002 (0.964) | 1.05 (0.2) | 1.09 (0.2) | 1.08 (0.2) | 1.05 (0.2) | 0.436 | NS |
| Total lipids in small HDL | 0.189 (<0.001) | 1.27 (0.2) | 1.29 (0.2) | 1.31 (0.2) | 1.32 (0.2) | 0.128 | NS |
| Total phospholipids in lipoprotein particles | 0.066 (0.172) | 2.95 (0.6) | 2.92 (0.5) | 2.93 (0.5) | 2.96 (0.5) | 0.95 | NS |
| Phospholipids in chylomicrons and extremely large VLDL | 0.204 (<0.001) | 0.023 (0.02) | 0.010 (0.02) | 0.048 (0.06) | 0.027 (0.025) | 0.001 | S0-S3 (0.002) |
| Phospholipids in very large VLDL | 0.221 (<0.001) | 0.036 (0.03) | 0.037 (0.03) | 0.045 (0.04) | 0.053 (0.04) | 0.001 | S0-S3 (0.002) |
| Phospholipids in large VLDL | 0.222 (<0.001) | 0.067 (0.04) | 0.067 (0.05) | 0.079 (0.05) | 0.092 (0.06) | 0.001 | S0-S3 (0.002) |
| Phospholipids in large VLDL | 0.222 (<0.001) | 0.067 (0.04) | 0.067 (0.05) | 0.079 (0.05) | 0.092 (0.06) | 0.001 | S1-S3 (0.041) |
| Phospholipids in medium VLDL | 0.150 (0.002) | 0.133 (0.05) | 0.119 (0.05) | 0.131 (0.04) | 0.143 (0.05) | 0.033 | S1-S3 (0.037) |
| Phospholipids in small VLDL | 0.117 (0.014) | 0.101 (0.03) | 0.094 (0.03) | 0.096 (0.03) | 0.105 (0.03) | 0.147 | NS |
| Phospholipids in very small VLDL | 0.028 (0.560) | 0.100 (0.03) | 0.095 (0.02) | 0.091 (0.02) | 0.099 (0.03) | 0.46 | NS |
| Phospholipids in IDL | −0.049 (0.308) | 0.303 (0.08) | 0.285 (0.06) | 0.278 (0.08) | 0.284 (0.07) | 0.156 | NS |
| Phospholipids in large LDL | 0.013 (0.780) | 0.40 (0.1) | 0.38 (0.1) | 0.38 (0.1) | 0.38 (0.1) | 0.388 | NS |
| Phospholipids in medium LDL | 0.085 (0.075) | 0.191 (0.05) | 0.178 (0.04) | 0.83 (0.04) | 0.190 (0.04) | 0.415 | NS |
| Phospholipids in small LDL | 0.053 (0.268) | 0.097 (0.02) | 0.090 (0.02) | 0.093 (0.02) | 0.096 (0.02) | 0.314 | NS |
| Phospholipids in HDL | −0.040 (0.407) | 1.50 (0.3) | 1.55 (0.2) | 1.52 (0.3) | 1.48 (0.3) | 0.355 | NS |
| Phospholipids in very large HDL | −0.232 (<0.001) | 0.054 (0.04) | 0.051 (0.02) | 0.046 (0.03) | 0.039 (0.03) | 0.002 | S0-S3 (0.002) |
| Phospholipids in large HDL | −0.191 (<0.001) | 0.28 (0.1) | 0.28 (0.1) | 0.26 (0.1) | 0.22 (0.1) | 0.003 | S0-S3 (0.016) |
| Phospholipids in large HDL | −0.191 (<0.001) | 0.28 (0.1) | 0.28 (0.1) | 0.26 (0.1) | 0.22 (0.1) | 0.003 | S1-S3 (0.039) |
| Phospholipids in medium HDL | −0.031 (0.518) | 0.48 (0.1) | 0.50 (0.1) | 0.49 (0.1) | 0.48 (0.1) | 0.497 | NS |
| Phospholipids in small HDL | −0.175 (<0.001) | 0.70 (0.1) | 0.72 (0.1) | 0.72 (0.1) | 0.73 (0.1) | 0.151 | NS |
| Total cholesterol | 0.019 (0.688) | 5.14 (1.3) | 4.87 (1.1) | 4.91 (1.1) | 5.0 (1.1) | 0.555 | NS |
| Cholesterol in chylomicrons and extremely large VLDL | 0.213 (<0.001) | 0.045 (0.05) | 0.019 (0.03) | 0.07 (0.09) | 0.05 (0.04) | 0.001 | S0-S3 (0.001) |
| Cholesterol in very large VLDL | 0.209 (<0.001) | 0.061 (0.03) | 0.060 (0.04) | 0.07 (0.04) | 0.08 (0.05) | 0.002 | S0-S3 (0.006) |
| Cholesterol in very large VLDL | 0.209 (<0.001) | 0.061 (0.03) | 0.060 (0.04) | 0.07 (0.04) | 0.08 (0.05) | 0.002 | S1-S3 (0.043) |
| Cholesterol in large VLDL | 0.208 (<0.001) | 0.102 (0.06) | 0.108 (0.06) | 0.120 (0.07) | 0.140 (0.08) | 0.002 | S0-S3 (0.007) |
| Cholesterol in large VLDL | 0.208 (<0.001) | 0.102 (0.06) | 0.108 (0.06) | 0.120 (0.07) | 0.140 (0.08) | 0.002 | S1-S3 (0.037) |
| Cholesterol in medium VLDL | 0.037 (0.436) | 0.203 (0.08) | 0.177 (0.08) | 0.183 (0.07) | 0.196 (0.08) | 0.297 | NS |
| Cholesterol in small VLDL | 0.084 (0.082) | 0.190 (0.07) | 0.174 (0.06) | 0.173 (0.05) | 0.192 (0.06) | 0.217 | NS |
| Cholesterol in very small VLDL | −0.034 (0.476) | 0.203 (0.06) | 0.190 (0.06) | 0.179 (0.05) | 0.192 (0.06) | 0.242 | NS |
| Cholesterol in IDL | −0.066 (0.168) | 0.86 (0.2) | 0.79 (0.2) | 0.77 (0.2) | 0.77 (0.2) | 0.089 | NS |
| Cholesterol in large LDL | 0.014 (0.768) | 1.29 (0.4) | 1.19 (0.3) | 1.20 (0.3) | 1.24 (0.3) | 0.371 | NS |
| Cholesterol in medium LDL | 0.091 (0.057) | 0.555 (0.2) | 0.513 (0.1) | 0.531 (0.1) | 0.556 (0.1) | 0.325 | NS |
| Cholesterol in small LDL | 0.078 (0.106) | 0.238 (0.06) | 0.221 (0.06) | 0.229 (0.05) | 0.237 (0.06) | 0.323 | NS |
| Cholesterol in HDL | −0.124 (0.010) | 1.39 (0.3) | 1.40 (0.2) | 1.39 (0.3) | 1.32 (0.2) | 0.017 | NS |
| Cholesterol in very large HDL | −0.256 (<0.001) | 0.07 (0.03) | 0.07 (0.02) | 0.06 (0.02) | 0.05 (0.02) | <0.0001 | S0-S3 (<0.001) |
| Cholesterol in large HDL | −0.229 (<0.001) | 0.26 (0.1) | 0.26 (0.1) | 0.24 (0.1) | 0.20 (0.1) | <0.0001 | S0-S3 (<0.001) |
| Cholesterol in large HDL | −0.229 (<0.001) | 0.26 (0.1) | 0.26 (0.1) | 0.24 (0.1) | 0.20 (0.1) | <0.0001 | S1-S3 (0.037) |
| Cholesterol in medium HDL | −0.059 (0.219) | 0.53 (0.1) | 0.54 (0.1) | 0.54 (0.1) | 0.51 (0.1) | 0.124 | NS |
| Cholesterol in small HDL | −0.160 (0.001) | 0.53 (0.1) | 0.53 (0.1) | 0.54 (0.1) | 0.54 (0.1) | 0.452 | NS |
| Triglycerides | 0.218 (<0.001) | 1.27 (0.6) | 1.30 (0.7) | 1.46 (0.8) | 1.64 (0.9) | 0.001 | S0-S3 (0.001) |
| Triglycerides in chylomicrons and extremely large VLDL | 0.212 (<0.001) | 0.132 (0.15) | 0.063 (0.08) | 0.265 (0.345) | 0.155 (0.135) | 0.001 | S0-S3 (0.001) |
| Triglycerides in very large VLDL | 0.230 (<0.001) | 0.117 (0.08) | 0.124 (0.12) | 0.153 (0.13) | 0.177 (0.14) | <0.0001 | S0-S3 (0.001) |
| Triglycerides in large VLDL | 0.229 (<0.001) | 0.186 (0.11) | 0.193 (0.14) | 0.226 (0.15) | 0.260 (0.17) | <0.0001 | S0-S3 (0.001) |
| Triglycerides in medium VLDL | 0.227 (<0.001) | 0.293 (0.13) | 0.273 (0.13) | 0.324 (0.16) | 0.368 (0.19) | <0.0001 | S0-S3 (0.002) |
| Triglycerides in small VLDL | 0.203 (<0.001) | 0.148 (0.06) | 0.148 (0.06) | 0.158 (0.06) | 0.177 (0.08) | 0.002 | S0-S3 (0.005) |
| Triglycerides in very small VLDL | 0.134 (0.005) | 0.065 (0.02) | 0.064 (0.02) | 0.064 (0.02) | 0.070 (0.02) | 0.079 | NS |
| Triglycerides in IDL | 0.089 (0.065) | 0.97 (0.02) | 0.095 (0.02) | 0.094 (0.02) | 0.101 (0.02) | 0.35 | NS |
| Triglycerides in large LDL | 0.087 (0.069) | 0.097 (0.02) | 0.094 (0.02) | 0.094 (0.02) | 0.100 (0.02) | 0.373 | NS |
| Triglycerides in medium LDL | 0.138 (0.004) | 0.032 (0.01) | 0.031 (0.01) | 0.032 (0.01) | 0.034 (0.01) | 0.094 | NS |
| Triglycerides in small LDL | 0.184 (<0.001) | 0.013 (0.01) | 0.013 (0.01) | 0.015 (0.01) | 0.016 (0.01) | 0.009 | S0-S3 (0.021) |
| Triglycerides in HDL | 0.153 (0.001) | 0.12 (0.04) | 0.12 (0.04) | 0.12 (0.03) | 0.13 (0.05) | 0.037 | S0-S3 (0.030) |
| Triglycerides in very large HDL | −0.211 (<0.001) | 0.005 (0.01) | 0.005 (0.01) | 0.006 (0.01) | 0.006 (0.01) | 0.775 | NS |
| Triglycerides in large HDL | −0.010 (0.842) | 0.02 (0.01) | 0.02 (0.01) | 0.02 (0.01) | 0.02 (0.01) | 0.916 | NS |
| Triglycerides in medium HDL | 0.166 (<0.001) | 0.043 (0.01) | 0.047 (0.02) | 0.046 (0.01) | 0.050 (0.01) | 0.019 | S0-S3 p = 0.012 |
| Triglycerides in small HDL | 0.227 (<0.001) | 0.048 (0.01) | 0.050 (0.01) | 0.051 (0.01) | 0.056 (0.01) | 0.001 | S0-S3 (0.001) |
| Phosphoglycerides | 0.111 (0.021) | 2.55 (0.53) | 2.53 (0.49) | 2.52 (0.42) | 2.60 (0.51) | 0.643 | NS |
| Ratio of triglycerides to phosphoglycerides | 0.255 (<0.001) | 0.49 (0.17) | 0.50 (0.21) | 0.58 (0.30) | 0.60 (0.22) | <0.0001 | S0-S3 (<0.001) |
| Ratio of triglycerides to phosphoglycerides | 0.255 (<0.001) | 0.49 (0.17) | 0.50 (0.21) | 0.58 (0.30) | 0.60 (0.22) | <0.0001 | S1-S3 (0.020) |
| Total cholines | 0.087 (0.070) | 2.85 (0.55) | 2.81 (0.48) | 2.81 (0.44) | 2.88 (0.50) | 0.736 | NS |
| Phosphatidylcholines | −0.104 (0.031) | 2.34 (0.50) | 2.32 (0.48) | 2.30 (0.41) | 2.39 (0.41) | 0.583 | NS |
| Sphingomyelins | −0.035 (0.466) | 0.54 (0.1) | 0.51 (0.07) | 0.51 (0.09) | 0.52 (0.08) | 0.172 | NS |
| Apolipoprotein B | 0.052 (0.281) | 0.98 (0.31) | 0.90 (0.27) | 0.91 (0.25) | 0.96 (0.27) | 0.404 | NS |
| Apolipoprotein A1 | −0.027 (0.573) | 1.38 (0.26) | 1.42 (0.22) | 1.40 (0.24) | 1.36 (0.21) | 0.45 | NS |
| Ratio of apolipoprotein B to apolipoprotein A1 | 0.067 (0.162) | 0.72 (0.22) | 0.64 (0.20) | 0.70 (0.16) | 0.71 (0.20) | 0.109 | NS |
| Total fatty acids | 0.185 (<0.001) | 13.14 (2.3) | 13.04 (2.4) | 13.36 (2.1) | 13.87 (2.7) | 0.044 | NS |
| Degree of unsaturation | −0.035 (0.470) | 1.32 (0.08) | 1.29 (0.09) | 1.29 (0.10) | 1.30 (0.09) | 0.270 | NS |
| Omega-3 fatty acids | 0.119 (0.014) | 0.52 (0.14) | 0.52 (0.19) | 0.47 (0.14) | 0.54 (0.16) | 0.168 | NS |
| Omega-6 fatty acids | 0.100 (0.038) | 4.80 (0.72) | 4.70 (0.66) | 4.78 (0.59) | 4.86 (0.71) | 0.556 | NS |
| Polyunsaturated fatty acids | 0.111 (0.021) | 5.32 (0.8) | 5.22 (0.8) | 5.25 (0.7) | 5.39 (0.8) | 0.492 | NS |
| Monounsaturated fatty acids | 0.196 (<0.001) | 3.33 (0.7) | 3.29 (0.8) | 3.52 (0.8) | 3.64 (0.9) | 0.008 | S0-S3 (0.021) |
| Saturated fatty acids | 0.211 (<0.001) | 4.49 (0.9) | 4.53 (0.9) | 4.59 (0.8) | 4.83 (1.0) | 0.016 | S0-S3 (0.024) |
| Linoleic acid | 0.092 (0.057) | 3.70 (0.8) | 3.62 (0.7) | 3.72 (0.7) | 3.76 (0.8) | 0.759 | NS |
| Docosahexaenoic acid | 0.006 (0.895) | 0.24 (0.06) | 0.24 (0.07) | 0.22 (0.05) | 0.24 (0.05) | 0.208 | NS |
The analysis of variance (ANOVA) with post hoc analysis showed significant differences between stage 0 and stage 3 steatosis in several lipid biomolecules. Patients with stage 3 steatosis had significantly higher total lipids in lipoprotein particles, total lipids in chylomicrons and extremely large VLDL, total lipids in very large, large, and medium size VLDL particles, and total lipids in very large and large HDL particles. Furthermore, patients with stage 3 steatosis had higher phospholipids, cholesterol and triglycerides in several lipoprotein particles. Their ratio of triglycerides to phosphoglycerides, and their mean MUFA and SFA were significantly higher when compared to patients with no steatosis (Table 1).
## Association between liver fibrosis and lipid biomolecules
We observed several significant but weak positive associations between liver fibrosis, assessed by measuring liver stiffness in kPa, and lipid biomolecules including total lipids in very small VLDL ($r = 0.135$, $$p \leq 0.005$$), total lipids in very large ($r = 0.278$, $p \leq 0.001$), and large ($r = 0.199$, $p \leq 0.001$) HDL, phospholipids in very small VLDL ($r = 0.143$, $$p \leq 0.003$$), and phospholipids in HDL ($r = 0.104$, $$p \leq 0.031$$), very large ($r = 0.288$, $p \leq 0.001$), and large ($r = 0.200$, $p \leq 0.001$) HDL. Positive association was observed between liver stiffness and cholesterol in very small VLDL ($r = 0.129$, $$p \leq 0.007$$), and very large ($r = 0.253$, $p \leq 0.001$), and large ($r = 0.186$, $p \leq 0.001$) HDL, and triglycerides in very small VLDL ($r = 0.101$, $$p \leq 0.035$$), IDL ($r = 0.134$, $$p \leq 0.005$$), large LDL ($r = 0.126$, $$p \leq 0.008$$), HDL ($r = 0.100$, $$p \leq 0.037$$), very large HDL ($r = 0.146$, $$p \leq 0.002$$) and large HDL ($r = 0.217$, $p \leq 0.001$). On the other hand, we observed a significantly negative correlation between liver stiffness and total lipids in small HDL (r = −0.109, $$p \leq 0.023$$), and cholesterol in small HDL (r = -0.165, $$p \leq 0.001$$). Compared to patients without fibrosis, patients with fibrosis had significantly higher mean total lipids in very small VLDL ($$p \leq 0.011$$), phospholipids in very small VLDL ($$p \leq 0.007$$) and cholesterol in very small VLDL ($$p \leq 0.016$$). Furthermore, patients with fibrosis also had significantly higher mean total lipids in very large HDL and large HDL ($p \leq 0.001$ and $$p \leq 0.027$$), phospholipids in very large and large HDL ($p \leq 0.001$ and $$p \leq 0.027$$), cholesterol in very large and large HDL ($p \leq 0.001$ and $$p \leq 0.041$$), and triglycerides in very large and large HDL ($$p \leq 0.017$$ and $$p \leq 0.003$$). Mean triglycerides in IDL and large IDL was also significantly higher among patients with fibrosis ($$p \leq 0.018$$ and $$p \leq 0.026$$). On the other hand, mean total lipids in small HDL and mean cholesterol in small HDL were significantly lower among patients with fibrosis ($$p \leq 0.024$$ and $$p \leq 0.002$$) as shown in Table 2.
**TABLE 2**
| Biomolecules | Fibrosis E (Kpa) correlation coefficient (p-value) | Without fibrosis N = 398 | With fibrosis N = 36 | p-value |
| --- | --- | --- | --- | --- |
| Total lipids in lipoprotein particles | 0.010 (0.828) | 9.48 (2.1) | 9.55 (2.5) | 0.853 |
| Total lipids in chylomicrons extremely large VLDL | −0.050 (0.303) | 0.26 (0.3) | 0.22 (0.2) | 0.489 |
| Total lipids in very large VLDL | −0.050 (0.299) | 0.28 (0.2) | 0.26 (0.2) | 0.583 |
| Total lipids in large VLDL | −0.046 (0.335) | 0.45 (0.2) | 0.43 (0.3) | 0.613 |
| Total lipids in medium VLDL | −0.033 (0.496) | 0.68 (0.2) | 0.67 (0.3) | 0.869 |
| Total lipids in small VLDL | 0.016 (0.739) | 0.46 (0.1) | 0.47 (0.1) | 0.530 |
| Total lipids in very small VLDL | 0.135 (0.005) | 0.36 (0.1) | 0.40 (0.1) | 0.011 |
| Total lipids in IDL | 0.069 (0.150) | 1.18 (0.3) | 1.25 (0.3) | 0.212 |
| Total lipids in large LDL | −0.013 (0.790) | 1.73 (0.4) | 1.73 (0.5) | 0.988 |
| Total lipids in medium LDL | −0.052 (0.284) | 0.77 (0.2) | 0.76 (0.2) | 0.602 |
| Total lipids in small LDL | −0.038 (0.433) | 0.35 (0.1) | 0.34 (0.1) | 0.843 |
| Total lipids in HDL | 0.101 (0.036) | 2.96 (0.5) | 3.02 (0.6) | 0.561 |
| Total lipids in very large HDL | 0.278 (<0.001) | 0.11 (0.1) | 0.15 (0.1) | <0.001 |
| Total lipids in large HDL | 0.199 (<0.001) | 0.48 (0.2) | 0.57 (0.3) | 0.027 |
| Total lipids in medium HDL | 0.025 (0.601) | 1.06 (0.2) | 1.04 (0.2) | 0.581 |
| Total lipids in small HDL | −0.109 (0.023) | 1.31 (0.2) | 1.25 (0.1) | 0.024 |
| Total phospholipids in lipoprotein particles | 0.048 (0.320) | 2.95 (0.5) | 2.99 (0.6) | 0.642 |
| Phospholipids in chylomicrons and extremely large VLDL | −0.043 (0.374) | 0.03 (0.01) | 0.03 (0.02) | 0.526 |
| Phospholipids in very large VLDL | −0.038 (0.430) | 0.05 (0.03) | 0.05 (0.04) | 0.737 |
| Phospholipids in large VLDL | −0.042 (0.379) | 0.08 (0.05) | 0.08 (0.05) | 0.678 |
| Phospholipids in medium VLDL | −0.026 (0.590) | 0.14 (0.05) | 0.14 (0.06) | 0.976 |
| Phospholipids in small VLDL | 0.011 (0.820) | 0.10 (0.03) | 0.11 (0.04) | 0.571 |
| Phospholipids in very small VLDL | 0.143 (0.003) | 0.10 (0.02) | 0.11 (0.03) | 0.007 |
| Phospholipids in IDL | 0.067 (0.166) | 0.29 (0.07) | 0.30 (0.08) | 0.228 |
| Phospholipids in large LDL | −0.022 (0.649) | 0.39 (0.08) | 0.38 (0.09) | 0.841 |
| Phospholipids in medium LDL | −0.056 (0.241) | 0.19 (0.04) | 0.18 (0.05) | 0.529 |
| Phospholipids in small LDL | −0.027 (0.576) | 0.10 (0.02) | 0.09 (0.02) | 0.868 |
| Phospholipids in HDL | 0.104 (0.031) | 1.49 (0.3) | 1.52 (0.3) | 0.535 |
| Phospholipids in very large HDL | 0.288 (<0.001) | 0.04 (0.02) | 0.06 (0.03) | <0.001 |
| Phospholipids in large HDL | 0.200 (<0.001) | 0.24 (0.1) | 0.28 (0.1) | 0.027 |
| Phospholipids in medium HDL | 0.026 (0.585) | 0.48 (0.1) | 0.48 (0.1) | 0.612 |
| Phospholipids in small HDL | −0.078 (0.103) | 0.72 (0.1) | 0.69 (0.1) | 0.065 |
| Total cholesterol | 0.017 (0.724) | 5.00 (1.1) | 5.06 (1.2) | 0.755 |
| Cholesterol in chylomicrons and extremely large VLDL | −0.034 (0.483) | 0.06 (0.05) | 0.05 (0.04) | 0.630 |
| Cholesterol in very large VLDL | −0.035 (0.463) | 0.07 (0.04) | 0.07 (0.04) | 0.776 |
| Cholesterol in large VLDL | −0.033 (0.494) | 0.13 (0.07) | 0.13 (0.05) | 0.811 |
| Cholesterol in medium VLDL | −0.005 (0.919) | 0.19 (0.07) | 0.19 (0.08) | 0.791 |
| Cholesterol in small VLDL | 0.030 (0.528) | 0.19 (0.06) | 0.20 (0.07) | 0.339 |
| Cholesterol in very small VLDL | 0.129 (0.007) | 0.19 (0.06) | 0.21 (0.08) | 0.016 |
| Cholesterol in IDL | 0.059 (0.220) | 0.80 (0.2) | 0.84 (0.3) | 0.286 |
| Cholesterol in large LDL | −0.021 (0.667) | 1.24 (0.3) | 1.24 (0.3) | 0.923 |
| Cholesterol in medium LDL | −0.058 (0.230) | 0.55 (0.1) | 0.54 (0.2) | 0.544 |
| Cholesterol in small LDL | −0.046 (0.343) | 0.24 (0.1) | 0.23 (0.1) | 0.702 |
| Cholesterol in HDL | 0.082 (0.089) | 1.34 (0.2) | 1.36 (0.3) | 0.756 |
| Cholesterol in very large HDL | 0.253 (<0.001) | 0.06 (0.02) | 0.08 (0.03) | <0.001 |
| Cholesterol in large HDL | 0.186 (<0.001) | 0.22 (0.1) | 0.26 (0.2) | 0.041 |
| Cholesterol in medium HDL | 0.011 (0.811) | 0.52 (0.1) | 0.51 (0.1) | 0.438 |
| Cholesterol in small HDL | −0.165 (0.001) | 0.54 (0.1) | 0.51 (0.1) | 0.002 |
| Triglycerides | −0.025 (0.597) | 1.52 (0.1) | 1.49 (0.1) | 0.820 |
| Triglycerides in chylomicrons and extremely large VLDL | −0.055 (0.252) | 0.17 (0.2) | 0.14 (0.1) | 0.448 |
| Triglycerides in very large VLDL | −0.058 (0.227) | 0.16 (0.1) | 0.14 (0.1) | 0.484 |
| Triglycerides in large VLDL | −0.053 (0.268) | 0.34 (0.1) | 0.33 (0.2) | 0.515 |
| Triglycerides in medium VLDL | −0.042 (0.383) | 0.34 (0.2) | 0.33 (0.2) | 0.710 |
| Triglycerides in small VLDL | 0.003 (0.945) | 0.17 (0.1) | 0.17 (0.1) | 0.787 |
| Triglycerides in very small VLDL | 0.101 (0.035) | 0.07 (0.02) | 0.07 (0.03) | 0.066 |
| Triglycerides in IDL | 0.134 (0.005) | 0.10 (0.02) | 0.11 (0.03) | 0.018 |
| Triglycerides in large LDL | 0.126 (0.008) | 0.10 (0.02) | 0.11 (0.03) | 0.026 |
| Triglycerides in medium LDL | 0.075 (0.120) | 0.03 (0.01) | 0.04 (0.01) | 0.173 |
| Triglycerides in small LDL | 0.015 (0.754) | 0.02 (0.01) | 0.02 (0.01) | 0.712 |
| Triglycerides in HDL | 0.100 (0.037) | 0.13 (0.04) | 0.14 (0.05) | 0.172 |
| Triglycerides in very large HDL | 0.146 (0.002) | 0.01 (0.002) | 0.07 (0.003) | 0.017 |
| Triglycerides in large HDL | 0.217 (<0.001) | 0.02 (0.01) | 0.03 (0.02) | 0.003 |
| Triglycerides in medium HDL | 0.081 (0.092) | 0.05 (0.02) | 0.05 (0.02) | 0.307 |
| Triglycerides in small HDL | 0.011 (0.816) | 0.05 (0.02) | 0.05 (0.02) | 0.792 |
| Phosphoglycerides | 0.075 (0.120) | 2.57 (0.5) | 2.64 (0.6) | 0.438 |
| Ratio of triglycerides to phosphoglycerides | −0.056 (0.246) | 0.57 (0.2) | 0.54 (0.2) | 0.446 |
| Total cholines | 0.073 (0.128) | 2.86 (0.5) | 2.93 (0.6) | 0.450 |
| Phosphatidylcholines | 0.073 (0.130) | 2.37 (0.5) | 2.43 (0.6) | 0.489 |
| Sphingomyelins | 0.029 (0.545) | 0.52 (0.08) | 0.52 (0.09) | 0.813 |
| Apolipoprotein B | 0.004 (0.929) | 0.95 (0.3) | 0.97 (0.3) | 0.705 |
| Apolipoprotein A1 | 0.091 (0.058) | 1.37 (0.2) | 1.38 (0.3) | 0.698 |
| Ratio of apolipoprotein B to apolipoprotein A1 | −0.042 (0.378) | 0.70 (0.2) | 0.71 (0.2) | 0.839 |
| Total fatty acids | 0.030 (0.533) | 13.60 (2.5) | 13.74 (2.9) | 0.767 |
| Degree of unsaturation | −0.032 (0.511) | 1.30 (0.1) | 1.29 (0.1) | 0.286 |
| Omega-3 fatty acids | 0.029 (0.554) | 0.53 (0.1) | 0.52 (0.1) | 0.842 |
| Omega-6 fatty acids | 0.004 (0.942) | 4.83 (0.7) | 4.79 (0.7) | 0.815 |
| Polyunsaturated fatty acids | 0.009 (0.857) | 5.36 (0.8) | 5.32 (0.8) | 0.806 |
| Monounsaturated fatty acids | 0.036 (0.460) | 3.53 (0.9) | 3.62 (1.0) | 0.582 |
| Saturated fatty acids | 0.039 (0.424) | 4.72 (0.9) | 4.80 (1.2) | 0.642 |
| Linoleic acid | 0.006 (0.894) | 3.73 (0.8) | 3.71 (0.8) | 0.862 |
| Docosahexaenoic acid | 0.037 (0.439) | 0.24 (0.05) | 0.23 (0.05) | 0.384 |
## Subgroup analysis showing the effect of lipid lowering medications on lipidomic profile
To demonstrate the impact of anti-lipemic medications on lipid biomolecules, we performed a subgroup analysis between the two groups of patients with T2DM taking versus not taking the lipid lowering agent Supplementary Table S2. Our data revealed that total lipids in lipoprotein particles ($$p \leq 0.001$$), total lipids in medium, small, and very small VLDL particles ($$p \leq 0.021$$, $$p \leq 0.045$$ and $$p \leq 0.006$$, respectively), total lipids in large, medium, small LDL particles ($p \leq 0.001$, $$p \leq 0.002$$, and $$p \leq 0.004$$, respectively), and total lipids in HDL ($p \leq 0.001$), very large, large, medium, small HDL particles ($$p \leq 0.004$$, $$p \leq 0.001$$, $$p \leq 0.001$$, and $$p \leq 0.012$$ respectively) were significantly lower among patients who were on anti-lipemic medication compared to those who were not on anti-lipemic medication. For total phospholipids in lipoprotein particles ($p \leq 0.001$), phospholipids in medium, small, very small VLDL particles ($$p \leq 0.009$$, $$p \leq 0.016$$, and $$p \leq 0.023$$, respectively), phospholipids in large, medium, small LDL particles ($p \leq 0.001$, $$p \leq 0.002$$, and $$p \leq 0.007$$, respectively), phospholipids in HDL ($p \leq 0.001$), very large, large, medium, small HDL particles ($$p \leq 0.005$$, $$p \leq 0.001$$, $$p \leq 0.003$$ and $$p \leq 0.010$$, respectively) were also significantly lower in patients who were on anti-lipemic medication compared to those who were not on anti-lipemic medication. Additionally, total cholesterol ($p \leq 0.001$), cholesterol in medium, small, very small VLDL ($$p \leq 0.001$$, $$p \leq 0.012$$ and $$p \leq 0.001$$, respectively), cholesterol in large, medium, and small LDL ($p \leq 0.001$, $$p \leq 0.002$$ and $$p \leq 0.003$$, respectively), cholesterol in HDL ($p \leq 0.001$), very large, large, medium, small HDL ($$p \leq 0.006$$, $$p \leq 0.001$$, $$p \leq 0.001$$ and $$p \leq 0.020$$, respectively) were also lower in patients taking anti-lipemic medications. Other lipid biomolecules that were found to be higher in patients not taking anti-lipemic medications compared to those taking these medications were phosphoglycerides ($$p \leq 0.001$$), total cholines ($p \leq 0.001$), phosphatidylcholines ($p \leq 0.001$), sphingomyelins ($p \leq 0.001$), Apo B ($$p \leq 0.001$$), Apo A ($p \leq 0.001$), omega 6 fatty acids ($$p \leq 0.030$$), PUFA ($$p \leq 0.007$$), and linoleic acid ($$p \leq 0.003$$).
## Discussion
We have recently demonstrated the prevalence of steatosis and fibrosis, and their associated risk factors in our cohort of patients with T2DM (Alfadda et al., 2022). In our current study of 434 patients with T2DM, CAP values were obtained by FibroScan and alterations in lipidomic profile were demonstrated by using high-throughput proton NMR metabolomics approach. We identified and analyzed a total of 81 lipid biomolecules. Our data highlight an association between steatosis and circulating concentration of lipids, phospholipids, cholesterol and triglycerides in VLDL and LDL subclasses in patients with T2DM. In particular, patients with S3 grade steatosis have higher concentration of lipids, phospholipids, cholesterol and triglycerides in VLDL and LDL subclasses compared to patients with no steatosis. On contrary, a negative association was observed between steatosis and circulating concentration of lipids, phospholipids, cholesterol and triglycerides in HDL subclasses. Moreover, ratio of triglycerides to phosphoglycerides, MUFA and SFA were also significantly higher in patients with S3 grade steatosis compared to patients with no steatosis. Furthermore, an association was observed between fibrosis and concentration of lipids, phospholipids, cholesterol and triglycerides in very small VLDL, large and very large HDL subclasses.
In patients with T2DM, insulin resistance upsurges fatty acid buildup in hepatocytes as a consequence of increased flux of non-esterified fatty acids released during adipose tissue lipolysis and de novo lipogenesis (Birkenfeld and Shulman, 2014). These fatty acids are esterified with glycerol to form triglycerides. Increase in the intrahepatic triglyceride accumulation intensifies the formation of lipid droplets in liver (Ress and Kaser, 2016). The triglyceride rich lipids droplets are packaged and secreted as VLDL into circulation and transported to peripheral tissues such as adipose tissue for storage or other metabolic organs like heart, skeletal muscles where they are hydrolyzed to release free fatty acids (FFA) to be consumed for energy (Dowman et al., 2010). The increase in triglyceride buildup and reduction in VLDL secretion and oxidation of fatty acids initiate fat build up in hepatocytes and lead to progression of steatosis (Kawano and Cohen, 2013). In agreement with previous studies, we observed a positive association between steatosis and serum total lipid in VLDL and LDL lipoprotein subclasses, while total lipids HDL subclasses were negatively associated (Kaikkonen et al., 2017). Phospholipids in very large to small subclasses of VLDL were positively associated while very large to large HDL subclasses were negatively associated with steatosis. Furthermore, ANOVA with post hoc analysis demonstrated significant differences between total lipids in very large to medium VLDL and very large to large HDL and S0 and S3 grade. Total cholesterol in chylomicrons in very large to large subclasses of VLDL were also significantly different between different grades of steatosis. Additionally, triglycerides and triglycerides in very large to small subclasses of VLDL, medium to small subclasses of LDL and small to medium subclasses of HDL and were significantly different between different grades of steatosis. In consistent with previous studies, cholesterol in very to large subclasses, triglycerides and triglycerides to phosphoglycerides ratio were positively associated with the risk of steatosis (Fukuda et al., 2016). The ratio of triglycerides to phosphoglycerides were significantly higher in S3 grade compared to S0 grade.
The prospective associations of metabolic abnormalities in lipoprotein subclass profile with progression to fatty liver and consequent fibrosis has been revealed previously (Heeren and Scheja, 2021). Recently, an increase in VLDL particle size was linked to steatohepatitis, whereas decrease in the concentration of small VLDL particles was associated with fibrosis (Jiang et al., 2016). In our cohort we observed an association between fibrosis and concentration of lipids and phospholipids of extremely small subclasses of VLDL, small, large and very large subclasses of HDL. A correlation was observed between fibrosis and total cholesterol concentration of very small VLDL and small, large and very large HDL. Additionally, an association was observed between fibrosis and concentration of triglycerides in very small VLDL, IDL, large LDL, HDL, large and very large HDL. A negative correlation but insignificant association was observed between fibrosis and concentration of lipids, phospholipids and cholesterol in small HDL.
The fatty acid composition of the serum reflects the risk of steatosis (Puri et al., 2007). Alteration in the fatty acid composition and dyslipidemia are implicated in the development of early steatosis and NASH (Bjermo et al., 2012; Zheng et al., 2012; Rosqvist et al., 2014; Walle et al., 2016; Luukkonen et al., 2018; Hajduch et al., 2021; Ooi et al., 2021). Serum total MUFA and SFA proportion are reported to be higher in NASH compared to NAFL. Additionally, liver biopsies of patients with steatosis and NASH show increased MUFA and SFA (Puri et al., 2007; Allard et al., 2008; Chiappini et al., 2017). In T2DM patients with NAFLD Increase in circulatory MUFA and SFA are associated with risk of cardiovascular disease (CVD) and strongly with steatosis (Petit et al., 2012; Würtz et al., 2015). In line with these findings, we also observed a positive association between steatosis and serum MUFA and SFA proportion. Our data revealed significant differences in serum MUFA and SFA between patients with no steatosis (S0) and patients with severe steatosis (S3). On the contrary, it has been shown that n‐3 and n‐6 PUFA exhibit a protective role and are inversely related to the steatosis in patients with insulin resistance (López-Vicario et al., 2014). In our study, we observed an association between n‐3 and n‐6 PUFA and steatosis, but no significant differences were observed between different grades of steatosis.
Dyslipidemia or changes in the lipid profile is a common risk factor observed in patients with liver steatosis (Chatrath et al., 2012). Dyslipidemia mainly refers to disorders in lipid metabolism and is characterized by an increase in triglycerides and cholesterol and decrease in HDL in these patients. Thus, use of antilipemic-medications improves lipid profile and decrease the risk of cardiovascular diseases in these patients. For the management of dyslipidemia nearly $47\%$ of the patients with T2DM were on anti-lipemic medication. We looked at the influence of the antilipemic medications in our cohort by performing a sub-analysis of data comparing patients who were on anti-lipemic versus those who were not on anti-lipemic medications. Our data showed a significant decrease in lipids and phospholipids in lipoprotein particles, cholesterol, phosphoglycerides, cholines, phosphatidylcholines, sphingomyelins, ApoB and ApoA1, PUFA and linoleic acid among patients who were on these medications compared to those who were not.
## Conclusion
With the increase in the prevalence of NAFLD in patients with T2DM, there is an urgent need to identify the potential circulation biomarkers to develop an effective screening and therapeutic strategies. In the present study we have utilized a metabolomics platform, and have identified for the first time, the association of circulating adverse lipids and lipoprotein subclasses with NAFLD in Saudi patients with T2DM. Our data demonstrated a significant positive association between steatosis (S3 grade) and dyslipidemia in patients with T2DM. Furthermore, a significant association was observed between liver fibrosis (F2-F4 grade) and concentration of lipids, phospholipids, cholesterol and triglycerides in very small VLDL, large and very large HDL subclasses. On contrary, a negative but weak association was observed between fibrosis and concentration of lipids, phospholipids and cholesterol in small HDL. Furthermore, use of antilipemic medications markedly decreased the concentration of several lipid biomolecules including lipids, phospholipids, and cholesterol in our cohort. In conclusion, the potential serum biomarkers for dyslipidemia identified in current study are important in exploring the association between NAFLD and its associated comorbidities in patients with T2DM.
## Data availability statement
The raw data supporting the conclusion of this article will be made available by the authors, without undue reservation.
## Ethics statement
The studies involving human participants were reviewed and approved by the Institutional Review Board, King Fahad Medical City, Riyadh, Saudi Arabia IRB log number 12-344. The patients/participants provided their written informed consent to participate in this study.
## Author contributions
AAA designed and supervised the research. Together with AAA, AMA, SS, ANA, and ASA planned and carried out the implementation of the research from recruitment and selection of patients, data and sample collection, and provided critical feedback. NA-D and ST-R participated in planning the study. AI and RG performed data analysis, interpretation of the results and drafting of the findings of this work. All authors discussed the results and contributed to the final manuscript.
## Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
## Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
## Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmolb.2023.1030661/full#supplementary-material
## References
1. Abeysekera K. W. M., Fernandes G. S., Hammerton G., Portal A. J., Gordon F. H., Heron J.. **Prevalence of steatosis and fibrosis in young adults in the UK: a population-based study**. *Lancet Gastroenterol. Hepatol.* (2020) **5** 295-305. DOI: 10.1016/S2468-1253(19)30419-4
2. Alfadda A. A., Sherbeeni S. M., Alqutub A. N., Aldosary A. S., Aldaghri N. M., Taylor-Robinson S. D.. **Transient elastography for the prevalence of non-alcoholic fatty liver disease in patients with type 2 diabetes: Evidence from the CORDIAL cohort study**. *Saudi J. Gastroenterol.* (2022) **28** 426-433. DOI: 10.4103/sjg.sjg_73_22
3. Allard J. P., Aghdassi E., Mohammed S., Raman M., Avand G., Arendt B. M.. **Nutritional assessment and hepatic fatty acid composition in non-alcoholic fatty liver disease (NAFLD): a cross-sectional study**. *J. Hepatol.* (2008) **48** 300-307. DOI: 10.1016/j.jhep.2007.09.009
4. Alswat K., Aljumah A. A., Sanai F. M., Abaalkhail F., Alghamdi M., Al Hamoudi W. K.. **Nonalcoholic fatty liver disease burden - Saudi Arabia and United Arab Emirates, 2017-2030**. *Saudi J. Gastroenterol.* (2018) **24** 211-219. DOI: 10.4103/sjg.SJG_122_18
5. Angulo P., Keach J. C., Batts K. P., Lindor K. D.. **Independent predictors of liver fibrosis in patients with nonalcoholic steatohepatitis**. *Hepatology* (1999) **30** 1356-1362. DOI: 10.1002/hep.510300604
6. Birkenfeld A. L., Shulman G. I.. **Nonalcoholic fatty liver disease, hepatic insulin resistance, and type 2 diabetes**. *Hepatology* (2014) **59** 713-723. DOI: 10.1002/hep.26672
7. Bjermo H., Iggman D., Kullberg J., Dahlman I., Johansson L., Persson L.. **Effects of n-6 PUFAs compared with SFAs on liver fat, lipoproteins, and inflammation in abdominal obesity: a randomized controlled trial**. *Am. J. Clin. Nutr.* (2012) **95** 1003-1012. DOI: 10.3945/ajcn.111.030114
8. Chatrath H., Vuppalanchi R., Chalasani N.. **Dyslipidemia in patients with nonalcoholic fatty liver disease**. *Semin. Liver Dis.* (2012) **32** 22-29. DOI: 10.1055/s-0032-1306423
9. Chiappini F., Coilly A., Kadar H., Gual P., Tran A., Desterke C.. **Metabolism dysregulation induces a specific lipid signature of nonalcoholic steatohepatitis in patients**. *Sci. Rep.* (2017) **7** 46658. DOI: 10.1038/srep46658
10. Cohen J. C., Horton J. D., Hobbs H. H.. **Human fatty liver disease: old questions and new insights**. *Science* (2011) **332** 1519-1523. DOI: 10.1126/science.1204265
11. Contos M. J., Sanyal A. J.. **The clinicopathologic spectrum and management of nonalcoholic fatty liver disease**. *Adv. Anat. Pathol.* (2002) **9** 37-51. DOI: 10.1097/00125480-200201000-00005
12. Dowman J. K., Tomlinson J. W., Newsome P. N.. **Pathogenesis of non-alcoholic fatty liver disease**. *Qjm* (2010) **103** 71-83. DOI: 10.1093/qjmed/hcp158
13. Fukuda Y., Hashimoto Y., Hamaguchi M., Fukuda T., Nakamura N., Ohbora A.. **Triglycerides to high-density lipoprotein cholesterol ratio is an independent predictor of incident fatty liver; a population-based cohort study**. *Liver Int.* (2016) **36** 713-720. DOI: 10.1111/liv.12977
14. Hajduch E., Lachkar F., Ferré P., Foufelle F.. **Roles of ceramides in non-alcoholic fatty liver disease**. *J. Clin. Med.* (2021) **10** 792. DOI: 10.3390/jcm10040792
15. Heeren J., Scheja L.. **Metabolic-associated fatty liver disease and lipoprotein metabolism**. *Mol. Metab.* (2021) **50** 101238. DOI: 10.1016/j.molmet.2021.101238
16. Jiang Z. G., Tapper E. B., Connelly M. A., Pimentel C. F., Feldbrügge L., Kim M.. **Steatohepatitis and liver fibrosis are predicted by the characteristics of very low density lipoprotein in nonalcoholic fatty liver disease**. *Liver Int.* (2016) **36** 1213-1220. DOI: 10.1111/liv.13076
17. Kaikkonen J. E., Würtz P., Suomela E., Lehtovirta M., Kangas A. J., Jula A.. **Metabolic profiling of fatty liver in young and middle-aged adults: Cross-sectional and prospective analyses of the Young Finns Study**. *Hepatology* (2017) **65** 491-500. DOI: 10.1002/hep.28899
18. Karlas T., Petroff D., Sasso M., Fan J. G., Mi Y. Q., de Ledinghen V.. **Individual patient data meta-analysis of controlled attenuation parameter (CAP) technology for assessing steatosis**. *J. Hepatol.* (2017) **66** 1022-1030. DOI: 10.1016/j.jhep.2016.12.022
19. Kawano Y., Cohen D. E.. **Mechanisms of hepatic triglyceride accumulation in non-alcoholic fatty liver disease**. *J. Gastroenterol.* (2013) **48** 434-441. DOI: 10.1007/s00535-013-0758-5
20. López-Vicario C., González-Périz A., Rius B., Morán-Salvador E., García-Alonso V., Lozano J. J.. **Molecular interplay between Δ5/Δ6 desaturases and long-chain fatty acids in the pathogenesis of non-alcoholic steatohepatitis**. *Gut* (2014) **63** 344-355. DOI: 10.1136/gutjnl-2012-303179
21. Luukkonen P. K., Sädevirta S., Zhou Y., Kayser B., Ali A., Ahonen L.. **Saturated fat is more metabolically harmful for the human liver than unsaturated fat or simple sugars**. *Diabetes Care* (2018) **41** 1732-1739. DOI: 10.2337/dc18-0071
22. Ooi G. J., Meikle P. J., Huynh K., Earnest A., Roberts S. K., Kemp W.. **Hepatic lipidomic remodeling in severe obesity manifests with steatosis and does not evolve with non-alcoholic steatohepatitis**. *J. Hepatol.* (2021) **75** 524-535. DOI: 10.1016/j.jhep.2021.04.013
23. Petit J. M., Guiu B., Duvillard L., Jooste V., Brindisi M. C., Athias A.. **Increased erythrocytes n-3 and n-6 polyunsaturated fatty acids is significantly associated with a lower prevalence of steatosis in patients with type 2 diabetes**. *Clin. Nutr.* (2012) **31** 520-525. DOI: 10.1016/j.clnu.2011.12.007
24. Puri P., Baillie R. A., Wiest M. M., Mirshahi F., Choudhury J., Cheung O.. **A lipidomic analysis of nonalcoholic fatty liver disease**. *Hepatology* (2007) **46** 1081-1090. DOI: 10.1002/hep.21763
25. Ress C., Kaser S.. **Mechanisms of intrahepatic triglyceride accumulation**. *World J. Gastroenterol.* (2016) **22** 1664-1673. DOI: 10.3748/wjg.v22.i4.1664
26. Rosqvist F., Iggman D., Kullberg J., Cedernaes J., Johansson H. E., Larsson A.. **Overfeeding polyunsaturated and saturated fat causes distinct effects on liver and visceral fat accumulation in humans**. *Diabetes* (2014) **63** 2356-2368. DOI: 10.2337/db13-1622
27. Soininen P., Kangas A. J., Wurtz P., Suna T., Ala-Korpela M.. **Quantitative serum nuclear magnetic resonance metabolomics in cardiovascular epidemiology and genetics**. *Circ. Cardiovasc Genet.* (2015) **8** 192-206. DOI: 10.1161/CIRCGENETICS.114.000216
28. Trovato F. M., Catalano D., Musumeci G., Trovato G. M.. **4Ps medicine of the fatty liver: The research model of predictive, preventive, personalized and participatory medicine-recommendations for facing obesity, fatty liver and fibrosis epidemics**. *Epma J.* (2014) **5** 21. DOI: 10.1186/1878-5085-5-21
29. Walle P., Takkunen M., Männistö V., Vaittinen M., Lankinen M., Kärjä V.. **Fatty acid metabolism is altered in non-alcoholic steatohepatitis independent of obesity**. *Metabolism* (2016) **65** 655-666. DOI: 10.1016/j.metabol.2016.01.011
30. Wasilewska N., Lebensztejn D. M.. **Non-alcoholic fatty liver disease and lipotoxicity**. *Clin. Exp. Hepatol.* (2021) **7** 1-6. DOI: 10.5114/ceh.2021.104441
31. Würtz P., Havulinna A. S., Soininen P., Tynkkynen T., Prieto-Merino D., Tillin T.. **Metabolite profiling and cardiovascular event risk: a prospective study of 3 population-based cohorts**. *Circulation* (2015) **131** 774-785. DOI: 10.1161/CIRCULATIONAHA.114.013116
32. Zheng J. S., Xu A., Huang T., Yu X., Li D.. **Low docosahexaenoic acid content in plasma phospholipids is associated with increased non-alcoholic fatty liver disease in China**. *Lipids* (2012) **47** 549-556. DOI: 10.1007/s11745-012-3671-4
|
---
title: 'Food insecurity status and its contributing factors in slums’ dwellers of
southwest Iran, 2021: a cross-sectional study'
authors:
- Hassan Joulaei
- Parisa Keshani
- Zohre Foroozanfar
- Sima Afrashteh
- Zahra Hosseinkhani
- Mohammad Ali Mohsenpour
- Ghasem Moghimi
- Arash Homayouni Meymandi
journal: Archives of Public Health
year: 2023
pmcid: PMC9999310
doi: 10.1186/s13690-023-01049-8
license: CC BY 4.0
---
# Food insecurity status and its contributing factors in slums’ dwellers of southwest Iran, 2021: a cross-sectional study
## Abstract
### Background
One major factor causing food insecurity is believed to be poverty. Approximately 20 million Iranians live in slums with a vulnerable socioeconomic context. The outbreak of COVID-19, on top of the economic sanctions against Iran, has increased this vulnerability and made its inhabitants prone to food insecurity. The current study investigates food insecurity and its associated socioeconomic factors among slum residents of Shiraz, southwest Iran.
### Methods
Random cluster sampling was used to select the participants in this cross-sectional study. The heads of the households completed the validated Household Food Insecurity Access Scale questionnaire to assess food insecurity. Univariate analysis was utilized to calculate the unadjusted associations between the study variables. Moreover, a multiple logistic regression model was employed to determine the adjusted association of each independent variable with the food insecurity risk.
### Results
Among the 1227 households, the prevalence of food insecurity was $87.20\%$, with $53.87\%$ experiencing moderate and $33.33\%$ experiencing severe food insecurity. A significant relationship was observed between socioeconomic status and food insecurity, indicating that people with low socioeconomic status are more prone to food insecurity ($P \leq 0.001$).
### Conclusions
The current study revealed that food insecurity is highly prevalent in slum areas of southwest Iran. The socioeconomic status of households was the most important determinant of food insecurity among them. Noticeably, the coincidence of the COVID-19 pandemic with the economic crisis in Iran has amplified the poverty and food insecurity cycle. Hence, the government should consider equity-based interventions to reduce poverty and its related outcomes on food security. Furthermore, NGOs, charities, and governmental organizations should focus on local community-oriented programs to make basic food baskets available for the most vulnerable households.
## Introduction
Food insecurity, a condition under which households cannot provide adequate, healthy, and nutritious food, can deleteriously affect their health [1]. This phenomenon can lead to nutrient inadequacy [2], an increased risk of hypertension, diabetes [3], coronary heart disease, and heart attack [4], as well as all-cause and cardiovascular-related mortalities [5]. Based on the report jointly released in 2022, from 2019 to 2021, in southeast Asia, southern Asia, northern Africa, and Sub-Saharan Africa, food insecurity has affected $18.8\%$, $39.4\%$, $31.1\%$, and $60.9\%$ of the population, respectively [6]. Additionally, this complication has affected $42.2\%$ of the population in the Middle East and North Africa (MENA) region, as well as $49\%$ of the households in Iran [7]. Since nutrient deficiency has endangered more than one billion lives, food insecurity has attracted particular attention worldwide [8].
Food insecurity is a multidimensional phenomenon. Poverty in low-income populations is a significant factor causing food insecurity [9]. As household-related factors, household type, female-headed households, single-parent families, the householder’s education, ethnicity, the household’s size, the residential area, household income, reliance on financial assistance, household expenditure, and unemployment can all directly affect food insecurity risk [10, 11]. In addition, the above-mentioned household-related factors indirectly lead to a higher food insecurity risk due to lower socioeconomic status and degrees of poverty [12]. In European cities and suburbs, $10\%$ of the households were at risk of food insecurity due to low income and cannot follow a healthy diet [13]. In Iran, evidence showed that household economic status was the most important factor influencing food insecurity in rural households [14].
Another factor causing food insecurity is international political issues. With the reimposition of economic sanctions on Iran since 2018, food markets in Iran have seen a steep rise in prices while incomes have remained stagnant. As a result, maintaining a healthy and nutritious diet has become more challenging for Iranian households, negatively affecting food security among Iranians [15]. Sanctions have caused unemployment, inflation, and loss of the right to health [16].
On the other hand, the outbreak of COVID-19 has exacerbated poverty and, as a result, food insecurity [17]. It has been estimated that approximately 49 million people would face poverty in 2020 due to the pandemic [18] and that 820 million people would face hunger globally by the end of 2020 [19]. Furthermore, in developing countries, a large proportion of the population’s income depends on informal labor, and these populations have low savings. Thus, during the COVID-19 outbreak, lockdown policies reduced the income of a large population in developing countries, thereby increasing food insecurity. As a result, rural–urban migration has increased [20]. Health-related policies during COVID-19 have impaired Iranian rural farmers’ food security by raising unemployment and lowering income [21]. It was shown that the Iranian rural population faced deteriorated food security and changes in consumption of some food groups [22].
According to official statistics, approximately 20 million Iranians live in slums, with 11 million living in informal settlements and the remainder in the forewarn districts. Shiraz, the capital of Fars province and a metropolitan city in the southwest of Iran, is experiencing an increase in the number of people living in the slums, similar to other major cities in Iran [23, 24]. The socioeconomic context of these areas, on the one hand, and the coincidence of the COVID-19 pandemic with economic sanctions against Iran, on the other, can probably exacerbate the vulnerability of its residents to food insecurity. Therefore, the current study aimed to investigate the prevalence of food insecurity and its contributing factors (such as demographic and socioeconomic factors and covid-19 induced economic shortage) among slum residents of Shiraz.
## Participants
Data collection was conducted between January and June 2021. Random cluster sampling was used to select the participants in this cross-sectional study in Shiraz, the capital of Fars province in southwest Iran. Each slum area was treated as a cluster, with three out of the eleven suburban areas randomly chosen. All the regions were considered equal in terms of socioeconomic status. The sample size was determined using a $50\%$ prevalence of food insecurity in the *Shiraz slum* area population, comprising approximately 200,000 people, and a $95\%$ confidence level. Since the households were selected randomly for the sampling, $25\%$ more households were selected in each block so that in case of reluctance to cooperate, the sampling would not face problems. The sample size was calculated to be 1250 households overall and 415–420 households from each of the three slum areas, with ten blocks randomly selected from each. Eventually, 40–45 households in each block were interviewed.
## Measurements
Face-to-face and door-to-door interviews were conducted with household heads after completing informed consent forms. *Some* general information within each household was collected, such as the head of the household’s age, education level (illiterate, primary, secondary, or high school, diploma, and college), employment status (unemployed, employed, self-employed, pensioner), the number of family members. Their socioeconomic status was calculated considering possession of 9 specific items, including home, personal vehicle, washing machine, LCD TV, dishwasher, refrigerator, handmade rug, laptop, and microwave. Based on the number of items possessed by households, the socioeconomic status was categorized into three groups, low (3 items or less), moderate (4 to 6 items), and high (more than 7 items) [25]. In addition, they were asked whether they had chronic diseases (at least one of the non-communicable diseases, such as diabetes, cardiovascular disease, kidney disease, and cancer), a vulnerable group member in the household (child under 6, adolescent, disabled member, pregnant, handicapped, and elderly), receive financial help from the charity, the portion of income allocated to food purchase, covid-19-induced poverty (including job loss, reduced income, and reduced food purchase), and marital status. The heads of households completed the validated HFIAS (Household Food Insecurity Access Scale) questionnaire to assess food insecurity [26]. The FAO Indicator Guide was used to score a nine-item HFIAS questionnaire [27]. The results were categorized into mild/moderate and severe to make the results more understandable and more appropriate for interventions for policymakers.
## Statistical analysis
Percentages were used to report the distribution of categorized variables. The chi-square test was used to measure the association between the grouped variables in this study. In addition, multiple logistic regression was used to examine the adjusted relationships (odds ratios and $95\%$ CI) between the study variables (household heads, parents’ education, parents’ job, household size, having a chronic disease, having a member of vulnerable groups, consumption of vegetables, fruits, meat, nuts and legumes, socioeconomic status, cost, food cost to income ratio, government financial assistance, covid-19-induced poverty) and food insecurity.
Variables were included in the model (p-value under 0.2 in the univariable analysis) if they significantly contributed to the model’s fitness using the stepwise selection method, and the final model was reported. P-value < 0.05 was considered significant. STATA14.0 software (Stata, College Station, TX, USA) was used for all the statistical analyses.
## Results
In 1227 households, the prevalence of food insecurity was $87.20\%$, with $53.87\%$ experiencing mild and moderate hunger and $33.33\%$ experiencing severe hunger. According to Table 1, 1059 ($86.3\%$) of the household heads were male, and approximately 763 ($62.5\%$) households had four family members. Most mothers ($49.2\%$) and fathers ($50.1\%$) had completed primary or secondary school. Most mothers ($95.3\%$) were housewives, while nearly half the fathers ($48.3\%$) worked. Over half of the households had a chronic disease ($66.3\%$). The main reason behind migration among the participants in this study was the lack of job opportunities in their hometown ($92.6\%$). Only $17\%$ of the households were not included as members of the vulnerable groups. In the present study, $8.7\%$, $9.6\%$, $3.4\%$, $5.7\%$, $0.8\%$, and $10.7\%$ of the households consumed vegetables, fruits, red meat, legumes, nuts, and dairy products per day, respectively, which were significantly different between secure and insecure households ($P \leq 0.001$). In the group with food insecurity, the portion of households headed by mothers was significantly higher than father-headed ones ($P \leq 0.001$). Confirming that the father’s job closely correlates with food insecurity, the proportion of unemployed fathers in this group was higher than that of the secure group ($P \leq 0.001$). According to our results, the prevalence of chronic diseases was significantly higher in the insecure group ($$P \leq 0.012$$). In addition, the number of vulnerable people in food-insecure families was significantly higher. ( $$P \leq 0.032$$) (Table 1).Table 1Characteristics of slums’ dwellers household of South-west Iran, 2021VariablesTotalSecureInsecurep-valueMild/ModerateSevereHead of household1227157661409Father1059(86.3)144(91.7)593(89.7)322(78.7)0.001Mother156(12.7)9(5.7)62(9.4)85(20.8)Other12(1.0)4(2.5)6(0.9)2(0.5)Father’s education1072148593331Illiterate86(8.0)8(5.4)45(7.6)33(10.0)0.085Primary or secondary537(50.1)71(48.0)290(48.9)176(53.2)High school & diploma370(34.5)56(37.8)206(34.7)108(32.6)College79(7.4)13(8.8)52(8.8)14(4.2)Mother’s education1194153643398Illiterate134(11.2)18(11.8)56(8.7)60(15.1)0.003Primary or secondary588(49.2)64(41.8)317(49.3)207(52.0)High school & diploma400(33.5)60(39.2)225(35.0)115(28.9)College72(6.0)11(7.2)45(7.0)16(4.0)Father’s job1072150594328Workless132(12.3)6(4.0)65(10.9)61(18.6)0.001Worker518(48.3)62(41.3)287(48.3)169(51.5)Self-employed230(21.5)39(26.0)133(22.4)58(17.7)Employed & pensioner192(17.9)43(28.7)109(18.4)40(12.2)Mother’s job1171150635386Housewife1116(95.3)145(96.7)607(95.6)364(94.3)0.448Employment55(4.7)5(3.3)28(4.4)22(5.7)Household size1220158658404 < 4457(37.5)57(36.1)248(37.7)152(37.6)0.928 ≥ 4763(62.5)101(63.9)410(62.3)252(62.4)Having chronic disease1209158648403Yes802(66.3)118(74.7)435(67.1)249(61.8)0.012No407(33.7)40(25.3)213(32.9)154(38.2)Cause of Migration51757266194Lack of agriculture land & water23(4.4)3(5.3)9(3.4)11(5.7)0.515Lack of jobs479(92.6)53(93.0)251(94.4)175(90.2)Other15(2.9)1(1.8)6(2.3)8(4.1)Having vulnerable group1213158650405Child under 6 years200(16.5)28(17.7)102(15.7)70(17.3)0.032Teenager391(32.2)58(36.7)210(32.3)123(30.4)Both child under 6&Teenager256(21.1)27(17.1)145(22.3)84(20.7)Pregnant, handicap or disabled78(6.4)5(3.2)32(4.9)41(10.1)Elderly82(6.8)10(6.3)50(7.7)22(5.4)No206(17.0)30(19.0)111(17.1)65(16.0)Vegetables consumption1207156649402Daily105(8.7)39(25.0)52(8.0)14(3.5)0.001Weekly743(61.6)103(66.0)426(65.6)214(53.2)Monthly332(27.5)12(7.7)162(25.0)158(39.3)Rarely27(2.2)2(1.3)9(1.4)16(4.0)Fruit consumption1222155660407Daily117(9.6)55(35.5)52(7.9)10(2.5)0.001Weekly476(39.0)77(49.7)285(43.2)114(28.0)Monthly532(43.5)22(14.2)289(43.8)221(54.3)Rarely97(7.9)1(0.6)34(5.2)62(15.2)Meat consumption1211155652404Daily41(3.4)28(18.1)11(1.7)2(0.5)0.001Weekly326(26.9)69(44.5)187(28.7)70(17.3)Monthly652(53.8)52(33.5)376(57.7)224(55.4)Rarely192(15.9)6(3.9)78(12.0)108(26.7)Legumes consumption1202150652400Daily69(5.7)32(21.3)27(4.1)10(2.5)0.001Weekly713(59.3)104(69.3)409(62.7)200(50.0)Monthly387(32.2)12(8.0)201(30.8)174(43.5)Rarely33(2.7)2(1.3)15(2.3)16(4.0)Nuts consumption1216155653408Daily10(0.8)5(3.2)3(0.5)2(0.5)0.001Weekly63(5.2)26(16.8)25(3.8)12(2.9)Monthly166(13.7)45(29.0)85(13.0)36(8.8)Rarely977(80.3)79(51.0)540(82.7)358(87.7)Dairy consumption1213155654404Daily130(10.7)57(36.8)50(7.6)23(5.7)0.001Weekly726(59.9)91(58.7)432(66.1)203(50.2)Monthly311(25.6)5(3.2)151(23.1)155(38.4)Rarely46(3.8)2(1.3)21(3.2)23(5.7)Data reported as number (%) Table 2 represents a significant relationship between socioeconomic status and food insecurity, suggesting that people with low socioeconomic status are more prone to food insecurity ($P \leq 0.001$). The share of income allocated to food purchases was higher in the food-insecure group ($P \leq 0.001$). Over $90\%$ of the households did not receive any charitable assistance in this study, which was significantly associated with food insecurity ($P \leq 0.001$). During the COVID-19 epidemic, the insecure group had a higher rate of job loss and lower food purchases ($P \leq 0.05$) (Table 2).Table 2Association between household food security status and socioeconomic variables, in slums’ dwellers of South-west Iran, 2021VariablesTotalSecureInsecurep-valueMild/ModerateSevereSocioeconomic status1172150634388Low952(81.2)95(63.3)500(78.9)357(92.0)0.001*Moderate220(18.8)55(36.7)134(21.1)31(8.0)Costs of Living1222157657408 < 1 Million197(16.1)15(9.6)69(10.5)113(27.7)0.001*1–2 Million751(61.5)68(43.3)448(68.2)235(57.6) ≥ 3Million274(22.4)74(47.1)140(21.3)60(14.7)Share of income allocated to food purchase1207156648403More than half536(44.4)75(48.1)285(44.0)176(43.7)0.001*One third to Half429(35.5)30(19.2)264(40.7)135(33.5)less than One third242(20.0)51(32.7)99(15.3)92(22.8)Receiving governmental assistance (4$ per months per person)1225158658409Yes1172(95.7)147(93.0)633(96.2)392(95.8)0.210No53(4.3)11(7.0)25(3.8)17(4.2)Receive financial help from charity1225158659408Yes115(9.4)8(5.1)44(6.7)63(15.4)0.001*No1110(90.6)150(94.9)615(93.3)345(84.6)Covid-19 & job loss12141556564030.031*Yes795(65.5)87(56.1)440(67.1)268(66.5)No419(34.5)68(43.9)216(32.9)135(33.5)Covid-19 epidemic reduced their income1219156657406Yes872(71.5)101(64.7)469(71.4)302(74.4)0.076No347(28.5)55(35.3)188(28.6)104(25.6)Covid-19 epidemic reduced their food purchasing power1222157659406Yes1098(89.9)132(84.1)596(90.4)370(91.1)0.035*No124(10.1)25(15.9)63(9.6)36(8.9)*Significant at 0.05 level
## Socioeconomic factors associated with food insecurity
Table 3 illustrates the multiple logistic regression used for evaluating the relationship between food insecurity and socioeconomic factors. According to our findings, allocating “one-third to half” (OR = 5.62, $95\%$CI: 3.21–9.83, $P \leq 0.001$) and “more than half” (OR = 1.95, $95\%$CI: 1.21–3.13, $$P \leq 0.005$$) of the income to food purchases were significantly more affected by food insecurity compared to the group with the income to food purchase ratio of less than one third. Father’s self-employment (OR self-employed/ jobless = 0.29, $95\%$CI: 0.11–0.79, $$P \leq 0.016$$), employment, and retirement (OR employed and pensioner / workless = 0.26, $95\%$CI: 0.09–0.73, $$P \leq 0.009$$), having moderate socioeconomic status (OR moderate/low = 0.0.58, $95\%$CI: 0.37–0.90, $$P \leq 0.016$$), and spending more than 30 million Rials per month (Iran’ currency equal to 110 US$) on living costs (OR ≥ 30 million/ < 10 Million = 0.28, $95\%$CI: 0.12–0.66, $$P \leq 0.004$$) were reported as food insecurity-associated determinants (Table 3).Table 3Factors associated with food insecurity: Univariable and multivariable analysisVariablesCrudeOdds Ratio($95\%$ CI)AdjustedOdds Ratio($95\%$ CI)Head of household Father11 ($$P \leq 0.201$$)** Mother2.57(1.28–5.15)0.80 (0.20–3.18) Other0.31(0.093–1.05)0.18 (0.02–1.19)Father’s Job workless11 ($$P \leq 0.042$$)** Worker0.35(0.14–0.82)0.41 (0.16–1.08) Self-employed0.23(0.09–0.56)0.29 (0.11–0.79) Employed & pensioner0.16(0.06–0.40)0.26 (0.11–0.79)*Having a* chronic disease No11 Yes1.58(1.08–2.31)1.58 (0.99–2.52)*Socioeconomic status* Low11 Moderate0.33(0.22–0.48)0.58 (0.37–0.90)Cost < 1Million11 ($P \leq 0.001$)** 1-2Million0.82(0.46–1.48)1.06 (0.46–2.41) ≥ 3Million0.22(0.12–0.40)0.28 (0.12–0.66)Share of income allocated to food purchase less than One third11 ($P \leq 0.001$)** One third to Half3.55(2.19–5.75)5.62 (3.21–9.83) More than half1.64(1.10–2.43)1.95 (1.21–3.13)Receive financial help from charity No11 Yes2.08(0.99–4.37)2.89 (0.66–12.64)Covid-19 & job loss No11 Yes1.57(1.11–2.21)1.40 (0.85–2.31)Covid-19 & reduced food purchases No11 Yes1.84(1.14–2.97)1.82 (0.97–3.39)**Global p-valueCI Confidence interval
## Discussion
This cross-sectional study described food insecurity and its associated factors among Shiraz suburban households in southwest Iran. The findings suggest that the food insecurity prevalence was $87.2\%$ ($53.87\%$ were moderately and $33.3\%$ were severely hungry). According to the obtained findings, the portion of more than half of one’s income to food purchases increases the chances for food insecurity. Statistically, self-employment, employment by others, retirement, moderate socioeconomic status, and spending more than 30 million Rials (Iran’s currency equal to 110 US$) per month on family expenses were protective factors against food insecurity.
Our findings highlighted that food insecurity is more prevalent among the slum area residents in the southwest of Iran. Based on previous studies, the prevalence of food insecurity in different regions of Iran ranges from $27.8\%$ (urban residents of Shiraz) to $82\%$ (slums of Kerman) [14, 28], which is due to a recent increase in the dust because of wetlands drying up in the southwest and the shortage of water, agricultural mismanagement and low quality of living standards in the southeast of Iran that has led people to migrate to better-off places with job opportunities, such as Shiraz, and staying in suburban areas [29].
A meta-analysis conducted in Iran in 2004 revealed that the prevalence of mild, moderate, and severe food insecurity is $9.3\%$, $5.6\%$, and $3.7\%$, respectively. However, a similar study conducted in 2015 revealed a $49\%$ prevalence [7, 30]. It shows that the trend of food insecurity has increased recently in Iran. Meanwhile, food security has improved significantly in other developing countries, such as India [31]. Evidently, the prevalence of food insecurity was higher in rural and slum than the urban areas [14, 32–34].
Furthermore, studies imply that the prevalence of food insecurity in developing countries is remarkably higher in developed countries. For instance, the prevalence of food insecurity in Lebanon, Nigeria, Nairobi, and Kampala was $50\%$, $81\%$, $87\%$, and $93\%$, respectively, while it was $16.5\%$ and $15.9\%$ in Portugal and Canada, respectively [35–40]. This indicates that food insecurity is a concerning issue in developing countries. Noticeably, the tools used in various studies for measuring food insecurity were different; thus, comparing different studies should be done cautiously.
Omidvar et al. clustered MENA countries based on GDP and political stability, showing that the frequency of severe food insecurity was significantly different amongst the clusters. Severe food insecurity was $5\%$, $13.6\%$, and $26.7\%$ in rich, stable countries, middle-low income with less political stability, and middle-low income politically unstable countries [18].
The socioeconomic status of the households was the most substantial factor associated with food insecurity in the current study; this finding is consistent with those of previous papers in both developed and developing countries [31, 37, 39, 41–43]. The World Bank reported Iran’s poverty rate for 2019 at $17.80\%$, showing a $3.8\%$ increase from 2018 [44]. Notably, $81.2\%$ of the studied families were categorized as having low socioeconomic status, and $44.4\%$ spent more than half of their monthly income on food. Food insecurity has been discovered to be a significant issue among low-income households. In slum areas, unlike rural ones, most households were immigrants, and due to insufficient natural resources and land for agriculture, their heads worked as manual workers with low wages. [ 45]. On the other hand, in this study, some active people were jobless, which could worsen the economic status of their households. The overall unemployment rate for 2021 was $11.46\%$ in Iran, based on World Bank reports [46]. In the slum areas of Ahvaz, southern Iran, $16\%$ of households’ heads were unemployed or job-seekers [45]. Furthermore, the present study showed that female-headed households are more likely to have severe food insecurity than male-headed households, consistent with studies conducted in the United States and Kenya [47, 48]. This gender difference revealed that female-headed households would be the top priority for promoting food security programs. As other countries’ experiences show, governmental equity-oriented interventions are required to combat the socioeconomic roots of food insecurity [47–49].
According to our results, the proportion of the members with chronic diseases in the food-insecure group was significantly higher than that of the secure group, consistent with previous studies [35, 50]. In India, households with a physically disabled family member were twice prone to food insecurity [35]. Dean et al. showed that chronic diseases are strenuously linked to food insecurity and higher healthcare expenditures [51]. Evidently, each 1.0 percentage point enhancement in insurance coverage was associated with a food insecurity reduction of 0.4 percentage points [52]. Assisting people in getting health insurance could start a virtuous cycle of improving consequences for health and healthcare as well as food insecurity [52]. A role for such developments in comprehensive anti-poverty strategies may be recommended by food security improvement following the development of health insurance [53].
Similar to studies from Canada and urban areas of Iran, food insecurity was more prevalent in households headed by females than males [33, 43, 50]. Although these households require additional assistance, our findings showed that more than $90\%$ did not receive enough assistance from charities, which was significantly associated with food insecurity in these people. Charities, non-governmental organizations (NGOs), and governmental organizations could have notable roles in empowering female-headed households. Targeted programs for increasing their knowledge and skills, supporting the production and sale of household products (such as sewing, cooking, baking, handicrafts, and jewelry making), and other income-generating skills, such as child and elderly care, can be suggested for planning and implementation regarding cultural, social, and environmental potentials, via these NGOs [54, 55]. Female-headed families in Iran were the subject of a qualitative investigation revealing numerous difficulties that could pose a serious threat [18]. To reduce household food insecurity, governmental organizations should prioritize training them to adapt their new and multidimensional functions, supplying more financial assistance, and supporting them in raising their social standing.
Previous studies in Iran showed that the sanctions and poor domestic policies had significant negative impacts on Iran’s economy. However, among all households, some vulnerable groups suffered more, and sanctions drastically decreased their welfare [18].
Notably, the coincidence of Iran’s economic sanctions and the COVID-19 epidemic has reduced Iranians’ food-purchasing power, exacerbating food insecurity among the middle and low-socioeconomic populations [15]. On the other hand, it has caused an economic recession in Iran, reducing the government’s ability to support low-income families [15, 56]. According to the obtained results, there was no statistically significant relationship between COVID-19 and unemployment or decreased income; meanwhile, it did have a close to significant association ($$P \leq 0.058$$) with the decrease in food-purchasing power. Although mismanagement of agricultural sectors in the supply and distribution of food at reasonable prices despite climate change and lack of water resources could not be ignored, Economic sanctions starting over one year before the onset of the COVID-19 epidemic were effective in disturbing the Iranian’s economic status and reducing their purchasing power. In other MENA region countries such as Lebanon, results showed that after the COVID-19 pandemic, food insecurity was estimated at $36\%$ to $39\%$, with a 50–$70\%$ reduction in the population’s income [57]. Moreover, after the COVID-19 pandemic, half the studied population had a low food consumption score, more than half of the households ate less than two meals per day, and about $70\%$ of them missed their meals to spare food [58].
The most important strengths of the present study are as follows: first of all, the results were representative of the general population due to the sampling method. Second, sample size of the study was considered appropriate to guarantee the study’s power. Third, data collection was done with validated tools and a face-to-face method. Finally, an adjusted odds ratio was calculated to determine the association between food insecurity and its contributing factors, resulting in most confounding effects being adjusted.
One of the most significant limitations of this study was its cross-sectional design, limiting the researchers’ ability to investigate the causal pathways between food insecurity and various risk factors. Furthermore, information on the recent trend of food insecurity could not be provided due to the nature of the study. There was a possibility of selection bias by the people not participating in the study due to personal issues and shame. Furthermore, information was possible bias due to their consideration of drawing attention to themselves or shame about their situation.
## Conclusion
The current study revealed that household food insecurity is highly prevalent in slum areas of southwest Iran. The socioeconomic status of households was the most important determinant of food insecurity among them. Hence, mid-term equity-based governmental interventions should be considered to reduce poverty and its related outcomes on food security. In the short-term, governmental and non-governmental organizations should focus on programs that make basic food baskets available for the most vulnerable households, including female-headed and jobless ones. Given the role of charities in reducing the severity of food insecurity, involving them in local community-oriented programs to combat this issue is recommended.
## References
1. Gundersen C, Ziliak JP. **Food insecurity and health outcomes**. *Health Aff* (2015.0) **34** 1830-1839. DOI: 10.1377/hlthaff.2015.0645
2. Christian AK, Marquis GS, Colecraft EK, Lartey A, Soueida R. **Household food insecurity but not dietary diversity is associated with children's mean micronutrient density adequacy in rural communities across Ghana**. *Nutrition* (2019.0) **65** 97-102. DOI: 10.1016/j.nut.2019.03.006
3. Mendy VL, Vargas R, Cannon-Smith G, Payton M, Enkhmaa B, Zhang L. **Food insecurity and cardiovascular disease risk factors among Mississippi adults**. *Int J Environ Res Public Health* (2018.0) **15** 2016. DOI: 10.3390/ijerph15092016
4. Liu Y, Eicher-Miller HA. **Food Insecurity and Cardiovascular Disease Risk**. *Curr Atheroscler Rep* (2021.0) **23** 1-12. DOI: 10.1007/s11883-021-00923-6
5. Sun Y, Liu B, Rong S, Du Y, Xu G, Snetselaar LG. **Food Insecurity Is Associated With Cardiovascular and All-Cause Mortality Among Adults in the United States**. *J Am Heart Assoc* (2020.0) **9** e014629. DOI: 10.1161/JAHA.119.014629
6. 6.FAO, IFAD, UNICEF, WFP, WHOThe State of Food Security and Nutrition in the World 2022. Repurposing food and agricultural policies to make healthy diets more affordable2022. *The State of Food Security and Nutrition in the World 2022. Repurposing food and agricultural policies to make healthy diets more affordable* (2022.0)
7. Behzadifar M, Behzadifar M, Abdi S, Malekzadeh R, Salmani MA, Ghoreishinia G. **Prevalence of food insecurity in Iran: a systematic review and meta-analysis**. *Arch Iran Med* (2016.0) **19** 288-294. PMID: 27041526
8. Barrett CB. **Measuring food insecurity**. *Science* (2010.0) **327** 825-828. DOI: 10.1126/science.1182768
9. Loopstra R, Reeves A, Tarasuk V. **The rise of hunger among low-income households: an analysis of the risks of food insecurity between 2004 and 2016 in a population-based study of UK adults**. *J Epidemiol Community Health* (2019.0) **73** 668-673. DOI: 10.1136/jech-2018-211194
10. Sulaiman N, Yeatman H, Russell J, Law LS. **A food insecurity systematic review: experience from Malaysia**. *Nutrients* (2021.0) **13** 945. DOI: 10.3390/nu13030945
11. Zaçe D, Di Pietro ML, Caprini F, De Waure C, Ricciardi W. **Prevalence and correlates of food insecurity among children in high-income European countries. A systematic review**. *Ann lst Super Sanità.* (2020.0) **56** 90-98
12. 12.Mortazavi Z, Dorosty AR, Eshraghian MR, Ghaffari M, Ansari-Moghaddam A, Mohammadi M. Household food insecurity in Southeastern Iran: severity and related factors. Int J Food Sci. 2017;2017. 10.1155/2017/7536024.
13. Penne T, Goedemé T. **Can low-income households afford a healthy diet? Insufficient income as a driver of food insecurity in Europe**. *Food Policy* (2021.0) **99** 101978. DOI: 10.1016/j.foodpol.2020.101978
14. 14.Amiresmaeili M, Yazdi‐Feyzabadi V, Heidarijamebozorgi M. Prevalence of food insecurity and related factors among slum households in Kerman, south of Iran. The International Journal of Health Planning and Management. 2021. 10.1002/hpm.3242.
15. 15.Hejazi J, Emamgholipour S. The effects of the re-imposition of US sanctions on food security in Iran. Int J Health Policy Manag. 2022;11(5):651–7. 10.34172/ijhpm.2020.207.
16. Kokabisaghi F. **Assessment of the Effects of Economic Sanctions on Iranians' Right to Health by Using Human Rights Impact Assessment Tool: A Systematic Review**. *Int J Health Policy Manag* (2018.0) **7** 374-393. DOI: 10.15171/ijhpm.2017.147
17. Pereira M, Oliveira AM. **Poverty and food insecurity may increase as the threat of COVID-19 spreads**. *Public Health Nutr* (2020.0) **23** 3236-3240. DOI: 10.1017/S1368980020003493
18. 18.Hunger Map [Available from: https://hungermap.wfp.org/.
19. 19.Food & Agriculture Organization of the United Nations. The state of food security and nutrition in the world 2019: Safeguarding against economic slowdowns and downturns: Food & Agriculture Organization; 2019. [Available from: https://www.fao.org/3/ca5162en/ca5162en.pdf].
20. Smith MD, Wesselbaum D. **COVID-19, food insecurity, and migration**. *J Nutr* (2020.0) **150** 2855-2858. DOI: 10.1093/jn/nxaa270
21. Yazdanpanah M, Tajeri Moghadam M, Savari M, Zobeidi T, Sieber S, Löhr K. **The Impact of Livelihood Assets on the Food Security of Farmers in Southern Iran during the COVID-19 Pandemic**. *Int J Environ Res Public Health* (2021.0) **18** 5310. DOI: 10.3390/ijerph18105310
22. Ghanbari Movahed R, Maleki Fard F, Gholamrezai S, Pakravan-Charvadeh MR. **The Impact of COVID-19 Pandemic on Food Security and Food Diversity of Iranian Rural Households**. *Front Public Health* (2022.0) **10** 862043. DOI: 10.3389/fpubh.2022.862043
23. 23.Mahdnejad H, PARHIZ F. Spatial Determination of Urban Poverty Zones (Case Study: Tehran Metropolitan District 12). 2021;53(1):307–21. 10.22059/JHGR.2020.286492.1007986.
24. Joulaei H, Bhuiyan AR, Sayadi M, Morady F, Kazerooni PA. **Slums’ access to and coverage of primary health care services: a cross-sectional study in Shiraz, a Metropolis in Southern Iran**. *Iran J Med Sci* (2014.0) **39** 184. PMID: 24753641
25. Safarpour M, Dorosty Motlagh A, Hosseini SM, Ranjbar Noshari F, Safarpour M, Daneshi Maskooni M. **Prevalence and outcomes of food insecurity and its relationship with some socioeconomic factors**. *Knowledge Health* (2014.0) **8** 193-198
26. Mohammadi F, Omidvar N, Houshiar-Rad A, Khoshfetrat M-R, Abdollahi M, Mehrabi Y. **Validity of an adapted Household Food Insecurity Access Scale in urban households in Iran**. *Public Health Nutr* (2012.0) **15** 149-157. DOI: 10.1017/S1368980011001376
27. 27.Jennifer Coates, Anne Swindale, Paula Bilinsky. Household Food Insecurity Access Scale (HFIAS) for Measurement of Food Access: Indicator Guide 2007 [Available from: https://www.fao.org/fileadmin/user_upload/eufao-fsi4dm/doc-training/hfias.pdf.
28. Ramesh ADM T, Abdollahi M. **Prevalence of household food insecurity in the City of Shiraz and its association with socio-economic and demographic factors**. *Iran J Nutr Sci Food Technol* (2010.0) **4** 53-64
29. Anwar A, Anwar S, Muhammad A, Nawaz F, Hyder S, Noman K. **Climate change and infectious diseases: evidence from highly vulnerable countries**. *Iran J Public Health* (2019.0) **48** 2187. PMID: 31993386
30. Mohammadi Nasrabadi F, Omidvar N, Khoshfetrat M, Kolahdooz F. **Household food insecurity in the Islamic Republic of Iran: a systematic review and meta-analysis**. *EMHJ-East Mediterr Health J* (2014.0) **20** 698-706. DOI: 10.26719/2014.20.11.698
31. Dharmaraju N, Mauleshbhai SS, Arulappan N, Thomas B, Marconi DS, Paul SS. **Household food security in an urban slum: Determinants and trends**. *J Family Med Prim Care* (2018.0) **7** 819. DOI: 10.4103/jfmpc.jfmpc_185_17
32. Gholami A, Foroozanfar Z. **Household food security status in the Northeast of Iran: a cross-sectional study**. *Med J Islam Repub Iran* (2015.0) **29** 227. PMID: 26478885
33. 33.Shakiba M, Salari A, Mahdavi-Roshan M. Food insecurity status and associated factors among rural households in the north of Iran. Nutr Health. 2021;27(3):301–7. 10.1177/0260106021996840.
34. Asadi-Lari M, Jahromi LM, Montazeri A, Rezaee N, Mehrizi AAH, Shams-Beyranvand M. **Socio-economic risk factors of household food insecurity and their population attributable risk: A population-based study**. *Med J Islam Repub Iran* (2019.0) **33** 119. PMID: 32002392
35. Joshi A, Arora A, Amadi-Mgbenka C, Mittal N, Sharma S, Malhotra B. **Burden of household food insecurity in urban slum settings**. *PLoS ONE* (2019.0) **14** e0214461. DOI: 10.1371/journal.pone.0214461
36. Wanyama R, Gödecke T, Qaim M. **Food security and dietary quality in African slums**. *Sustainability* (2019.0) **11** 5999. DOI: 10.3390/su11215999
37. 37.Obayelu OA, Oyekola T. Food insecurity in urban slums: evidence from ibadan metropolis, southwest Nigeria. Journal for the Advancement of Developing Economies. 2018;7(1). 10.32873/unl.dc.jade7.1.2.
38. Jomaa L, Naja F, Cheaib R, Hwalla N. **Household food insecurity is associated with a higher burden of obesity and risk of dietary inadequacies among mothers in Beirut**. *Lebanon BMC public health* (2017.0) **17** 1-14
39. Álvares L, Amaral TF. **Food insecurity and associated factors in the Portuguese population**. *Food Nutr Bull* (2014.0) **35** S395-S402. DOI: 10.1177/156482651403500401
40. 40.PROOF. Household Food Insecurity in Canada, 2021 Agust, 2022 [Available from: https://proof.utoronto.ca/resource/household-food-insecurity-in-canada-2021/.
41. Mokari-Yamchi A, Faramarzi A, Salehi-Sahlabadi A, Barati M, Ghodsi D, Jabbari M. **Food Security and Its Association with Social Support in the Rural Households: A Cross-Sectional Study**. *Prev Nutr Food Sci* (2020.0) **25** 146. DOI: 10.3746/pnf.2020.25.2.146
42. Kleve S, Booth S, Davidson ZE, Palermo C. **Walking the food security tightrope—exploring the experiences of low-to-middle income melbourne households**. *Int J Environ Res Public Health* (2018.0) **15** 2206. DOI: 10.3390/ijerph15102206
43. McIntyre L, Bartoo AC, Emery JH. **When working is not enough: food insecurity in the Canadian labour force**. *Public Health Nutr* (2014.0) **17** 49-57. DOI: 10.1017/S1368980012004053
44. 44.Macrotrends. ran Poverty Rate 1986–2023 2019 Available from: https://www.macrotrends.net/countries/IRN/iran/poverty-rate.
45. Bagheri M. **The challenge of slums: socio-economic disparities**. *Int J Soc Sci Hum* (2012.0) **2** 410
46. 46.Macrotrends. Iran Unemployment Rate 1991–2023 2021 Available from: https://www.macrotrends.net/countries/IRN/iran/unemployment-rate#:~:text=Unemployment%20refers%20to%20the%20share,a%201.43%25%20increase%20from%202019.
47. Ma C, Ho SK, Singh S, Choi MY. **Gender disparities in food security, dietary intake, and nutritional health in the United States**. *Am J Gastroenterol* (2021.0) **116** 584-592. DOI: 10.14309/ajg.0000000000001118
48. Lutomia CK, Obare GA, Kariuki IM, Muricho GS. **Determinants of gender differences in household food security perceptions in the Western and Eastern regions of Kenya**. *Cogent Food Agric* (2019.0) **5** 1694755. DOI: 10.1080/23311932.2019.1694755
49. Yii V, Palermo C, Kleve S. **Population-based interventions addressing food insecurity in Australia: a systematic scoping review**. *Nutr Diet* (2020.0) **77** 6-18. DOI: 10.1111/1747-0080.12580
50. Gholami A, Sani TR, Askari M, Jahromi ZM, Dehghan A. **Food insecurity status and associated factors among rural households in North-East of Iran**. *Int J Prev Med* (2013.0) **4** 1018. PMID: 24130942
51. Dean EB, French MT, Mortensen K. **Food insecurity, health care utilization, and health care expenditures**. *Health Serv Res* (2020.0) **55** 883-893. DOI: 10.1111/1475-6773.13283
52. 52.Sonik RA. Health insurance and food insecurity: sparking a potential virtuous cycle. Am J Public Health. 2019;109(9):1163–5. 10.2105/AJPH.2019.305252.
53. 53.Himmelstein, G. Effect of the Affordable Care Act’s Medicaid Expansions on Food Security, 2010–2016. Am J Public Health. 2019;109(9):1243–8. 10.2105/AJPH.2019.305168.
54. 54.SEYED HOSSEIN ZARHANI. Empowerment of Female headed households, Case Study: "Sedighin" charity institution in Iran. 2011 Available from: http://www.socialsciences.in/article/empowerment-female-headed-households.
55. Yoosefi Lebni J, Mohammadi Gharehghani MA, Soofizad G, Irandoost SF. **Challenges and opportunities confronting female-headed households in Iran: a qualitative study**. *BMC Womens Health* (2020.0) **20** 1-11. DOI: 10.1186/s12905-020-01046-x
56. 56.Saeed Iranmanesh M, Norallah Salehi A, Seyyed Abdolmajid Jalaee B. Using Fuzzy Logic Method to Investigate the Effect of Economic Sanctions on Business Cycles in the Islamic Republic of Iran. Applied Computational Intelligence and Soft Computing. 2021;2021. 10.1155/2021/8833474.
57. Kharroubi S, Naja F, Diab-El-Harake M, Jomaa L. **Food insecurity pre-and post the COVID-19 pandemic and economic crisis in Lebanon: prevalence and projections**. *Nutrients* (2021.0) **13** 2976. DOI: 10.3390/nu13092976
58. Hoteit M, Al-Atat Y, Joumaa H, Ghali SE, Mansour R, Mhanna R. **Exploring the impact of crises on food security in lebanon: results from a national cross-sectional study**. *Sustainability* (2021.0) **13** 8753. DOI: 10.3390/su13168753
|
---
title: 'Association between self-compassion and cyber aggression in the COVID-19 context:
roles of attribution and public stigma'
authors:
- Qinglu Wu
- Tian-Ming Zhang
journal: BMC Psychology
year: 2023
pmcid: PMC9999325
doi: 10.1186/s40359-023-01100-x
license: CC BY 4.0
---
# Association between self-compassion and cyber aggression in the COVID-19 context: roles of attribution and public stigma
## Abstract
Self-compassion is negatively associated with aggressive behaviors. However, the association between self-compassion and cyber aggression toward stigmatized people (e.g., people infected with COVID-19) has not been investigated in the COVID-19 context and the mechanism underlying this association remains underexplored. On the basis of emotion regulation theory and attribution theory, this study examined the indirect effects of self-compassion on cyber aggression toward people infected with COVID-19 through attribution and public stigma of COVID-19. Data were collected from 1162 Chinese college students (415 male, mean age = 21.61 years). Participants completed an online questionnaire including measurement of the key variables and basic demographic information. Results indicated that self-compassion was negatively associated with cyber aggression through the lower attribution of COVID-19 and lower public stigma of COVID-19. A sequential pathway from the attribution of COVID-19 to public stigma of COVID-19 was identified in the relationship between self-compassion and cyber aggression. Our findings are consistent with emotion regulation theory and attribution theory, which posit that emotion regulation strategies are associated with interpersonal mistreatment through cognitive pathways. These findings suggest that emotional self-regulation strategies can be used to reduce cyber aggression toward stigmatized people by reducing attribution and public stigma in the COVID-19 context. Self-compassion improvement could be target for the interventions aiming at alleviating public stigma and interpersonal mistreatment toward stigmatized people.
## Introduction
The outbreak of coronavirus disease 2019 (COVID-19) in 2019 caused great threat to the general public [1]. Individuals with a high risk of COVID-19 exhibited negative attitudes and interpersonal mistreatment (e.g., aggression and stigmatization) in daily life [2, 3], especially demonstrating aggression toward people infected with COVID-19 or accusing them of spreading the disease [4]. Self-compassion, as an effective emotional self-regulation strategy and self-related resource, is helpful in alleviating the negative effects of COVID-19 (e.g., negative emotions, psychological distress, and fear of COVID-19) and aggressive behaviors in daily life [5–8]. Therefore, self-compassion might be associated with aggression toward stigmatized groups, such as individuals infected with COVID-19. To investigate this association and its underlying mechanism, the present study applied emotion regulation theory and attribution theory to examine the indirect effects of self-compassion on cyber aggression through attribution and public stigma of COVID-19.
## Relationship between self-compassion and aggression
Because of the negative effects of the COVID-19 pandemic (e.g., perceived threat, information overload, loneliness, and powerlessness), individuals feel stressed and are more likely to exhibit hostile attitudes and treat others aggressively [2, 9, 10], especially toward stigmatized groups (e.g., individuals infected with or accused of spreading COVID-19) [4, 11]. Cyber aggression has emerged as a common form of aggression due to isolation and decreased social participation caused by the COVID-19 pandemic [3, 4]. Because cyber aggression negatively affected the mental health of both its perpetrators and victims during the COVID-19 pandemic [11, 12], identifying protective factors and psychological resources to reduce its occurrence is necessary.
According to emotion regulation theory [13], self-compassion is an emotional self-regulation strategy, and it may effectively alleviate cyber aggression in the COVID-19 context. Self-compassion is an individuals’ self-related resource and positive response towards themselves when encountering adverse experiences such as inadequacies or difficulties [14, 15]. Moreover, self-compassion is a holistic system including six components that are grouped into compassionate self-responding (self-kindness, common humanity, and mindfulness) and reduced uncompassionate self-responding (decreased self-judgment, isolation, and over-identification) [14, 16]. Instead of responding in an uncompassionate manner, individuals with high levels of self-compassion usually treat themselves with kindness, view adversity as a common shared experience, and maintain a stable mood and a balanced perspective when life difficulties emerge.
The beneficial role of self-compassion in dealing with emotional distress and maintaining positive emotional function has been highly emphasized in studies examining the relationship between self-compassion and aggressive behaviors. A negative association between self-compassion and aggressive behaviors (e.g., verbal and physical aggressive expression) and inclinations has been identified in different populations (e.g., undergraduates and individuals diagnosed with personality disorder) [5, 17, 18]. Self-compassion is helpful in alleviating angry rumination, borderline personality disorder features, and moral disengagement, which subsequently reduce anger and aggressive behavior [5, 18, 19]. As an effective emotional regulation strategy, self-compassion is beneficial for smoothing and regulating negative emotions (e.g., anger) and ruminative thoughts [6, 20]. Therefore, self-compassion can prevent individuals from immersion in negative emotions and prevent their subsequent impulsive and aggressive behaviors toward others [17]. Self-compassionate individuals may be less likely to exhibit hostility and aggressiveness toward people with COVID-19 online.
## Potential indirect effects through attribution and public stigma of COVID-19
Attribution theory provides a perspective for investigating the mechanism underlying the relationship between self-compassion and cyber aggression. Weiner and colleagues [1988] developed a conceptual framework demonstrating that individuals’ attribution of a disease affects their affective responses and behavioral judgment toward groups with the disease. The model posits that people who consider the cause of the disease to be associated with the responsibility and control of those with the disease tend to blame them and endorse increased stigmatization [21]. According to the attribution model of stigma, two common risk factors for aggressive behaviors, namely attribution and public stigma, may play indirect roles in the relationship between self-compassion and cyber aggression toward individuals infected with COVID-19.
Attribution of COVID-19 refers to the cognitive process in which the cause of COVID-19 infection is assigned [22, 23]. Internal attribution, or assigning individual responsibility to those infected with COVID-19, may cause enhanced contempt and aggressive behaviors [24]. A study demonstrated that individuals’ attribution of COVID-19 was linked to their violent behaviors toward those perceived as associated with the disease, such as Asian adults in Western countries [25]. As an adaptive emotional self-regulation strategy and personal resource [15, 26], self-compassion may reduce internal attribution. Studies have identified the negative associations between self-compassion and individuals’ cognitive processes (e.g., rumination and negative cognitive reactions) [20, 27]. Individuals with a high level of self-compassion tend to view suffering as a part of human experience and assume a balanced perspective when facing negative thoughts and emotions instead of isolating themselves and overidentifying during negative experiences. Thus, they are more likely to exhibit positive cognitive reactions and adaptive cognitive processes (e.g., optimism, perspective taking, and positive reframing) and less likely to maintain negative cognitive reactions (e.g., rumination and revenge motivation) [27–29]. Similarly, individuals with a high level of self-compassion may view the negative consequences (e.g., social distancing and quarantine) of the COVID-19 pandemic as an experience shared with others and not be overcome by these adverse experiences. Therefore, they may be less likely to exhibit the causal attribution of COVID-19 (e.g., blame, controllability and responsibility) toward stigmatized people, which further weakens their intentions of aggressive behaviors.
Public stigma of COVID-19 refers to the devaluation of and discriminatory attitudes and beliefs regarding individuals infected with COVID-19 as endorsed by the general population [30]. Social stigmatization toward people associated with COVID-19 increases the likelihood of aggressive behaviors [31]. Because self-compassion may facilitate resilience to stigmatization [32], higher levels of self-compassion might weaken the effect of stigma on aggressive behaviors. Individuals with high levels of self-compassion are less likely to perceive public stigma [33] and more likely to exhibit a prosocial attitude toward others [15, 34]. Individuals with high levels of self-compassion are more likely to demonstrate compassion toward others, which weakens their negative outgroup attitudes [35]. In the context of COVID-19, all Chinese individuals are faced with the negative consequences of the COVID-19 pandemic (e.g., perceived threat and vulnerability to disease, lockdown, social distancing, and quarantine). Thus, when individuals in the COVID-19 context treat themselves kindly, recognize that their experiences are shared by many, and maintain a peaceful mood and balanced perspective of the situation, they are less likely to exhibit hostile or stigmatizing attitudes toward people with COVID-19, which may cause decreased aggressive behaviors.
## Relationship between attribution and public stigma of COVID-19
Studies have provided empirical evidence related to the attribution model in stigma attached to mental illness and infectious disease [36, 37], demonstrating that attribution factors are associated with public stigma. Attribution of the causes of mental illness, such as personal responsibility, predicts negative affective responses and discrimination [38]. The more the individuals perceive those with a disease to be responsible for having that disease, the greater the public stigma is attributed [39]. Several studies have indicated that the more the general population believes that individuals’ infection with COVID-19 is their own responsibility, the more stigmatization they endorse [40, 41]. Negative affect arousal (e.g., fear of contact with COVID-19 or irritation toward individuals infected with COVID-19) is positively associated with elevated levels of stigma [22, 42]. Thus, the association between the attribution and public stigma of COVID-19 was considered in the self-compassion—cyber aggression relationship in this study.
## The present study
Although several studies have addressed aggression during the COVID-19 pandemic, research on the prevention and reduction of aggression has been largely neglected. Identifying psychological factors and intrapersonal resources that may alleviate aggressive behaviors, especially aggression toward individuals infected with COVID-19, is crucial. In addition, delivering interventions on the basis of these factors may help to prevent future mistreatment and stigmatization. According to emotion regulation theory, self-compassion is an effective emotional self-regulation strategy and self-related resource for reducing aggressive behaviors. Moreover, the attribution model of stigma provides a perspective for further examination of the underlying mechanism and potential indirect roles of attribution and public stigma of COVID-19 in the association between self-compassion and cyber aggression. During the COVID-19 pandemic, college students became more reliant on online learning because of lockdown and social distancing policies, which might increase the likelihood of cyberbullying [43, 44]. Thus, in this study, we examined the mechanism underlying the association between self-compassion and cyber aggression toward people infected with COVID-19 among college students. The hypothesized model included three potential indirect effects: self-compassion → attribution of COVID-19 → cyber aggression,self-compassion → public stigma of COVID-19 → cyber aggression; and self-compassion → attribution of COVID-19 → public stigma of COVID-19 → cyber aggression.
## Participants and procedure
Data were sourced from a project conducted to investigate the effect of the pandemic fatigue and public stigma of COVID-19 on psychosocial adjustment. Chinese undergraduates and postgraduates aged ≥ 18 years at Beijing Normal University were recruited to participate in the survey. Participants were informed of the research objectives, procedures, and confidentiality policy. Participation was voluntary, and participants were informed that they had the right to withdraw from the survey without any penalty. Participants provided informed consent before responding the online questionnaire. Ethical approval was provided by the research ethics committee of the School of Social Development and Public Policy at Beijing Normal University. Participants who provided a valid response received a small monetary reward (30 RMB, approximately US$ 4.5).
In total, 1317 students provided valid survey responses. After removing surveys with duplicate responses ($$n = 37$$) and those that failed attention checking ($$n = 118$$), we obtained 1162 valid surveys (415 men, 747 women). The average age of the participants was 21.61 (SD = 2.81) years. The percentages of undergraduate and postgraduate participants were $65.2\%$ and $34.8\%$, respectively. The participants’ household monthly income was 19,811.49 RMB (median = 12,000, approximately US $ 1,769.73). Participant characteristics are listed in Table 1.Table 1Sample characteristicsCharacteristicN%SexMale$41535.7\%$Female$74764.3\%$Educational level = High school diploma (undergraduates)$75865.2\%$ = Bachelor’s degree (postgraduates)$40434.8\%$Household monthly income < 10,$00038733.3\%$10,000–20,$00057649.6\%$ > 20,$00019917.1\%$Physical healthPoor and very poor$363.1\%$Fair$27623.8\%$Good and excellent$85073.1\%$Whether participants or people they know have been infected with COVID-19Yes$13811.9\%$No$102488.1\%$
## Self-compassion
Self-compassion was measured using the 12-item Chinese version of the Self-Compassion Scale-Short Form (SCS-SF) [45, 46]. Acceptable reliability of the SCS-SF has been demonstrated in different Chinese populations (e.g., caregivers, adolescents, and college students) [45, 47, 48]. This scale includes six subscales: self-kindness, common humanity, mindfulness, self-judgment, isolation, and over-identification. An example item was “When I feel inadequate in some way, I try to remind myself that feelings of inadequacy are shared by most people.” Responses were obtained on a 5-point Likert scale ranging from 1 (almost never) to 5 (almost always). Higher mean scores indicated higher levels of self-compassion. The Cronbach’s α for the SCS-SF in the present study was 0.80.
## Attribution of COVID-19
Attribution of COVID-19 was assessed using a three-item scale developed by our research team. Three scale items measuring blame, controllability, and responsibility were extracted from the Chinese version of the scale of public stigma toward various infectious diseases [49]. Each item was rated on a 7-point Likert scale (1 = not at all, 7 = very much), with higher mean scores indicating agreement with the three internal attributing factors. For example, the item “People with COVID-19 are responsible for their own infection” was used to measure responsibility. The Cronbach’s α for this scale in the present study was 0.76.
## Public stigma of COVID-19
Public stigma of COVID-19 was measured using an 11-item scale modified from the Chinese version of an existing scale assessing public stigma of mental illness [37]. The original scale contained two subscales, and we extracted the public stigma dimension with 12 items. Because people infected with COVID-19 were forcibly quarantined in China during the period of data collection, one item, “People with COVID-19 should be quarantined.” was removed [41]. The final scale included 11 items, such as “I am worried that people with COVID-19 will cause harm to others.” Items were rated on a five-point Likert scale (1 = strongly disagree, 5 = strongly agree), with higher mean scores indicating greater public stigma. The Cronbach’s α for this scale in the present study was 0.88.
## Cyber aggression
Cyber aggression toward individuals infected with COVID-19 was measured using three items, which were modified from three original items in Chinese used to measure aggressive online behaviors toward Chinese people because of COVID-19 [4]. Responses were rated on a 7-point Likert scale ranging from 1 (not at all) to 7 (very much). An example item was “I would remind people around me to avoid those people who have been infected with COVID-19 online.” Higher mean scores indicated higher levels of cyber aggression toward individuals with COVID-19. The Cronbach’s α for this scale in the present study was 0.74.
## Covariates
Covariates in the present study included demographic variables (i.e., age, sex, educational level, and household monthly income) and variables related to health (i.e., physical health) and COVID-19 infection experience of participants or people they knew. Physical health was assessed using one item (“*How is* your general physical health?”) that participants rated on a 5-point scale ranging from 1 (very poor) to 5 (excellent). COVID-19 infection experience was assessed using one item that asked participants whether they or people they knew had been infected with COVID-19.
## Data analysis
Initial analyses (i.e., descriptive statistics and bivariate correlation) were conducted using SPSS version 24. The SPSS macro PROCESS (Model 6) [50] was used to investigate the direct and indirect effects of self-compassion on cyber aggression through (a) attribution of COVID-19; (b) public stigma of COVID-19; and (c) the pathway from attribution of COVID-19 to public stigma of COVID-19. Age, sex, educational background, household monthly income, physical health, and whether participants or people they knew had been infected with COVID-19 were controlled when model was examined. A resampling approach of bias-corrected bootstrapping (5000 times) with $95\%$ confidence intervals (CIs) was used to estimate the direct and indirect effects of self-compassion on cyber aggression. The effects were considered significant if their CIs did not include 0.
## Results
The descriptive statistics and bivariate correlations of key variables are displayed in Table 2. Self-compassion was negatively associated with the attribution of COVID-19, public stigma of COVID-19, and cyber aggression. Attribution of COVID-19, public stigma of COVID-19, and cyber aggression were positively associated with each other. Harman’s single-factor test was conducted to test for common method variance [51, 52]. Exploratory factor analysis was performed using the items from key variables to verify whether one general factor accounted for the majority of the covariance among measures. *The* generated principal component analysis revealed that six factors accounted for $59.20\%$ of the total variance. The first unrotated factor explained only $11.55\%$ of the variance in data, suggesting that the common method bias did not substantially affect the model. Table 2Descriptive statistics and correlations among key variablesSelf-compassionAttribution of COVID-19Public stigma of COVID-19Cyber aggressionSelf-compassion–Attribution of COVID-19−.08**–*Public stigma* of COVID-19−.28***.53***–Cyber aggression−.16***.54***.53***–Mean3.352.942.682.36Standard deviation0.541.380.801.25Range1.5–51–71–51–7* $p \leq .05$, ** $p \leq .01$, *** $p \leq .001$ The specific direct and indirect effects of self-compassion on cyber aggression are displayed in Table 3. Self-compassion was indirectly associated with cyber aggression through the attribution of COVID-19 (b = − 0.121, SE = 0.029, $95\%$ CI = − 0.180 to − 0.066), public stigma of COVID-19 (b = − 0.180, SE = 0.027, $95\%$ CI = − 0.236 to − 0.129), and attribution and public stigma of COVID-19 in sequence (b = − 0.052, SE = 0.013, $95\%$ CI = − 0.079 to − 0.029]). Details of the specific paths and unstandardized path coefficients are presented in Fig. 1.Table 3Direct and indirect effects of self-compassion on cyber aggressionUnstandardized parameter estimateS.EBias-corrected CI ($95\%$)LowerUpperDirect effect−.126.059−.242−.010Indirect effects SC → AT → CA−.121.029−.180−.066 SC → PS → CA−.180.027−.236−.129 SC → AT → PS → CA−.052.013−.079−.029SC Self-compassion, AT Attribution, PS Public stigma, CA Cyber aggressionFig. 1Paths and standardized path coefficients for hypothesized model
## Discussion
Our findings are in agreement with those of previous studies indicating that self-compassion is beneficial for individuals’ well-being and that of others [15, 53]. Compassionate people alleviate their own emotional difficulties when facing COVID-19-related stress and are less likely to develop internal attribution, stigmatize others, or exhibit aggressive behaviors toward individuals infected with COVID-19. Our findings support emotion regulation theory and attribution theory by demonstrating that self-compassion is an effective emotional self-regulation strategy that associated with reduced cyber aggression toward stigmatized people in the COVID-19 context, and decreased causal attribution and stigma of COVID-19 are critical in this mechanism. Our findings regarding the indirect effect of self-compassion on cyber aggression suggest that self-compassion is useful for modifying individuals’ maladaptation to life stress and challenges caused by the COVID-19 pandemic (e.g., cyber aggression toward stigmatized people).
## Indirect effects of self-compassion on cyber aggression
Indirect effects of self-compassion on cyber aggression are through attribution and public stigma of COVID-19. This finding supports that self-compassion is associated with individuals’ cognitive processes, especially how they view and interpret their negative experiences and their attitudes [20, 27]. Self-compassion was indicated to be negatively linked to attribution of COVID-19 in the present study. Individuals with a high level of self-compassion are less likely to experience negative affect (e.g., irritation, fear, and sadness) or adverse cognitive reaction under challenging circumstances [27, 54]. Thus, if the general population experiencing COVID-19 exhibits greater self-compassion, it is less likely to attribute the control of COVID-19 infection internally by assigning responsibility to those with COVID-19.
Consistently with prior studies, the results of the current study demonstrated that self-compassion was negatively associated with stigma [55–57]. Self-compassionate individuals are kind toward themselves, view individual experiences as part of collective experience, and are less likely to judge the shortcomings of other people [58]. Therefore, self-compassionate individuals may view the inconvenience and emotional distress caused by the COVID-19 pandemic as a common experience (including of the general public and stigmatized individuals) or collective trauma. Moreover, these individuals can maintain a balanced perspective when experiencing negative thoughts and emotions and are less likely to isolate themselves from others or overidentify [20, 58]. Thus, self-compassion might provide psychological resilience to endorsing stigmatizing attitudes [33, 59], and self-compassionate people are less likely to exhibit hostile attitudes and stigmatizing notions toward people infected with COVID-19, which may alleviate cyber aggression toward this stigmatized group.
Self-compassion is associated with cyber aggression through a sequential pathway from the attribution and public stigma of COVID-19, which supports attribution theory [21, 23]. This theory emphasizes that biased responses are decided by a cognitive process: individuals assign casual factors (e.g., controllability, responsibility, and blame) to an individual’s illness that affect their likelihood of exhibiting hostile attitudes and behaviors [60]. Those who exhibit more agreement with attributing factors have higher levels of public stigma and exhibit higher levels of cyber aggression [22, 40, 43]. This relationship is also exhibited by the general population as a whole. Because self-compassion is helpful to alleviates internal attribution [27], it may reduce the negative impact of attribution of COVID-19, which further weakens public stigma and aggressive behaviors.
## Limitations
This study has some limitations. First, because the present study has a cross-sectional design, it could not examine causal relationships among self-compassion, attribution, public stigma, and cyber aggression. Future studies could employ a longitudinal design to examine the mediating roles of attribution and public stigma of COVID-19 in the relationship between self-compassion and aggression. Second, the sample of Chinese college students in the present study limits the generalizability of the findings to other populations. To improve research diversity, the framework of the present study could be applied in other populations (e.g., individuals losing loved ones or encountering severe economic hardship caused by the COVID-19 pandemic) in different developmental stages and cultures. Third, the present study examined only cognition-based mechanisms (attribution and public stigma) between self-compassion and cyber aggression. Self-compassion affects cyber aggression through other cognitive, emotional, or moral pathways (e.g., angry rumination and moral disengagement) [5, 19]. Future research could comprehensively examine pathways linking self-compassion and cyber aggression. Fourth, factors investigating the indirect effect of self-compassion on cyber aggression are not completely identified. To comprehensively investigate the mechanism underlying the relationship between self-compassion and cyber aggression, ecological systems and influential factors at other levels (e.g., parent–child communication and aggression at school) could be considered [61, 62]. Fifth, Internet-related experiences (e.g., problematic Internet use and average time online) related to cyber aggression were not measured in this study [63, 64]. Future research should control for the potential effect of this covariate on cyber aggression.
## Implications
Our findings have theoretical and practical implications. They contribute to the understanding of how self-compassion is associated with interpersonal mistreatment toward stigmatized people in the COVID-19 context. On the basis of emotion regulation theory and the attribution model of public stigma, our study demonstrated the importance of cognitive pathways (attribution and public stigma) when applying emotion regulation strategies for dealing with cyber aggression. These pathways shed light on the cognitive-based mechanism underlying the emotional regulation/dysregulation–interpersonal mistreatment in the aggression (prevention) studies [65, 66]. This research provides insights into the function of self-compassion in stigma (especially public stigma) and aggression reduction and prevention from a perpetrator’s perspective. Studies have verified that self-compassion is helpful in reducing internalized stigma from the perspective of victims [56, 57]. Some researchers argue that self-compassion weakens the effects of public stigma on self-stigma and other negative outcomes [59]. Our findings broaden the application of research on the relationship between self-compassion and stigma, which mainly focuses on the benefits of self-compassion for stigmatized populations [56, 57]. Our study expands previous research that self-compassion is beneficial to perpetrators and can reduce general public stigmatization. Self-compassion promotion is helpful in providing relief and reducing threat perception [67], and individuals’ ability to care for and comfort themselves is beneficial to interpersonal interaction. Our findings also provide practical implications. Because self-compassion is helpful in alleviating negative cognition (e.g., attribution), attitudes (e.g., public stigma), and behaviors (e.g., verbal aggression) related to environmental events (e.g., the COVID-19 pandemic), self-compassion and psychoeducation programs should be used to improve self-compassion and reduce public stigma of COVID-19 [68, 69].
## References
1. Stanley BL, Zanin AC, Avalos BL, Tracy SJ, Town S. **Collective emotion during collective trauma: a metaphor analysis of the COVID-19 pandemic**. *Qual Health Res* (2021.0) **31** 1890-1903. DOI: 10.1177/10497323211011589
2. Karaoglan Yilmaz FG, Avci U, Yilmaz R. **The role of loneliness and aggression on smartphone addiction among university students**. *Curr Psychol* (2022.0). DOI: 10.1007/s12144-022-03018-w
3. Wang Q, Luo X, Tu R, Xiao T, Hu W. **COVID-19 information overload and cyber aggression during the pandemic lockdown: the mediating role of depression/anxiety and the moderating role of confucian responsibility thinking**. *Int J Environ Res Public Health* (2022.0) **19** 1540. DOI: 10.3390/ijerph19031540
4. Teng F, Wang X, Shi J, Chen Z, Huang Q, Cheng W. **Psychosomatic symptoms and neuroticism following COVID-19: the role of online aggression toward a stigmatized group**. *Int J Environ Res Public Health* (2021.0) **18** 8672. DOI: 10.3390/ijerph18168672
5. Fresnics A, Borders A. **Angry rumination mediates the unique associations between self-compassion and anger and aggression**. *Mindfulness* (2017.0) **8** 554-564. DOI: 10.1007/s12671-016-0629-2
6. Guan F, Wu Y, Ren W, Zhang P, Jing B, Xu Z, Wu S-T, Peng K-P, He J-B. **Self-compassion and the mitigation of negative affect in the era of social distancing**. *Mindfulness* (2021.0) **12** 2184-2195. DOI: 10.1007/s12671-021-01674-w
7. Hatun O, Kurtça TT. **Self-compassion, resilience, fear of COVID-19, psychological distress, and psychological well-being among Turkish adults**. *Curr Psychol* (2022.0). DOI: 10.1007/s12144-022-02824-6
8. Solmazer G, Cebeci EÖ, Demircan Y, Karasakal SG, Yakın N. **Self-compassion and driving anger expression among young drivers: the mediating role of anger rumination**. *Curr Psychol* (2021.0). DOI: 10.1007/s12144-021-02379-y
9. Barlett CP, Rinker A, Roth B. **Cyberbullying perpetration in the COVID-19 era: an application of general strain theory**. *J Soc Psychol* (2021.0) **161** 466-476. DOI: 10.1080/00224545.2021.1883503
10. Deng S, Feng X. **How perceived threat of COVID-19 related to aggressive tendencies during the pandemic in Hubei Province and other regions of China: mediators and moderators**. *Curr Psychol* (2022.0) **41** 3349-3362. DOI: 10.1007/s12144-021-01792-7
11. Yang F. **Coping strategies, cyberbullying behaviors, and depression among Chinese netizens during the COVID-19 pandemic: a web-based nationwide survey**. *J Affect Disorders* (2021.0) **281** 138-144. DOI: 10.1016/j.jad.2020.12.023
12. Varela JJ, Hernández C, Miranda R, Barlett CP, Rodríguez-Rivas ME. **Victims of cyberbullying: feeling loneliness and depression among youth and adult Chileans during the pandemic**. *Int J Environ Res Public Health* (2022.0) **19** 5886. DOI: 10.3390/ijerph19105886
13. Gross JJ. **Emotion regulation: taking stock and moving forward**. *Emotion* (2013.0) **13** 359-365. DOI: 10.1037/a0032135
14. Neff KD, Tóth-Király I, Yarnell LM, Arimitsu K, Castilho P, Ghorbani N, Guo HX, Hirsch JK, Hupfeld J, Hutz CS. **Examining the factor structure of the self-compassion scale in 20 diverse samples: support for use of a total score and six subscale scores**. *Psychol Assess* (2019.0) **31** 27-45. DOI: 10.1037/pas0000629
15. Wu Q, Cao H, Lin X, Zhou N, Chi P. **Child maltreatment and subjective well-being in Chinese emerging adults: a process model involving self-esteem and self-compassion**. *J Interpers Violence* (2021.0). DOI: 10.1177/0886260521993924
16. Neff KD, Long P, Knox MC, Davidson O, Kuchar A, Costigan A, Williamson Z, Rohleder N, Tóth-Király I, Breines JG. **The forest and the trees: Examining the association of self-compassion and its positive and negative components with psychological functioning**. *Self Identity* (2018.0) **17** 627-645. DOI: 10.1080/15298868.2018.1436587
17. Miyagawa Y, Taniguchi J. **Sticking fewer (or more) pins into a doll? The role of self-compassion in the relations between interpersonal goals and aggression**. *Motiv Emot* (2022.0) **46** 1-15. DOI: 10.1007/s11031-021-09913-2
18. Sommerfeld E, Shechory Bitton M. **Rejection sensitivity, self-compassion, and aggressive behavior: the role of borderline features as a mediator**. *Front Psychol* (2020.0). DOI: 10.3389/fpsyg.2020.00044
19. Zeng P, Wang P, Nie J, Ouyang M, Lei L. **Gratitude and cyberbullying perpetration: the mediating role of self-compassion and moral disengagement**. *Child Youth Serv Rev* (2020.0) **119** 105608. DOI: 10.1016/j.childyouth.2020.105608
20. Wu Q, Chi P, Zeng X, Lin X, Du H. **Roles of anger and rumination in the relationship between self-compassion and forgiveness**. *Mindfulness* (2019.0) **10** 272-278. DOI: 10.1007/s12671-018-0971-7
21. Weiner B. **On sin versus sickness: a theory of perceived responsibility and social motivation**. *Am Psychol* (1993.0) **48** 957-965. DOI: 10.1037/0003-066X.48.9.957
22. Nguyen T, Croucher SM, Diers-Lawson A, Maydell E. **Who’s to blame for the spread of COVID-19 in New Zealand? Applying attribution theory to understand public stigma**. *Commun Res Pract* (2021.0) **7** 379-396. DOI: 10.1080/22041451.2021.1958635
23. Weiner B, Perry RP, Magnusson J. **An attributional analysis of reactions to stigmas**. *J Pers Soc Psychol* (1988.0) **55** 738-748. DOI: 10.1037/0022-3514.55.5.738
24. Zuo B, Ye H, Wen F, Ke W, Xiao H, Wang J. **Effect of attribution on the emotions and behavioral intentions of third-party observers toward intergroup discriminators during the COVID-19 pandemic**. *Group Process Intergroup Relat* (2022.0). DOI: 10.1177/13684302211062367
25. 25.Hwang E Z N (2022) Role of COVID-19 attributions in inciting violence toward older Asian individuals Yale University.
26. Neff KD. **The development and validation of a scale to measure self-compassion**. *Self and Identity* (2003.0) **2** 223-250. DOI: 10.1080/15298860309027
27. Muris P, Otgaar H, Meesters C, Heutz A, van den Hombergh M. **Self-compassion and adolescents’ positive and negative cognitive reactions to daily life problems**. *J Child Fam Stud* (2019.0) **28** 1433-1444. DOI: 10.1007/s10826-019-01353-4
28. Miyagawa Y, Taniguchi J. **Self-compassion helps people forgive transgressors: cognitive pathways of interpersonal transgressions**. *Self Identity* (2022.0) **21** 244-256. DOI: 10.1080/15298868.2020.1862904
29. Wong CCY, Yeung NCY. **Self-compassion and posttraumatic growth: cognitive processes as mediators**. *Mindfulness* (2017.0) **8** 1078-1087. DOI: 10.1007/s12671-017-0683-4
30. Nochaiwong S, Ruengorn C, Awiphan R, Kanjanarat P, Ruanta Y, Phosuya C, Boonchieng W, Nanta S, Chongruksut W, Thavorn K, Wongpakaran N, Wongpakaran T. **COVID-19 public stigma scale (COVID-PSS): development, validation, psychometric analysis and interpretation**. *BMJ Open* (2021.0) **11** e048241. DOI: 10.1136/bmjopen-2020-048241
31. Akour A, AlMuhaissen SA, Nusair MB, Al-Tammemi A. **The untold story of the COVID-19 pandemic: perceptions and views towards social stigma and bullying in the shadow of COVID-19 illness in Jordan**. *SN Soc Sci* (2021.0) **1** 240. DOI: 10.1007/s43545-021-00252-0
32. Vigna AJ, Poehlmann-Tynan J, Koenig BW. **Does self-compassion facilitate resilience to stigma? A school-based study of sexual and gender minority youth**. *Mindfulness* (2018.0) **9** 914-924. DOI: 10.1007/s12671-017-0831-x
33. Heath PJ, Brenner RE, Lannin DG, Vogel DL. **Self-compassion moderates the relationship of perceived public and anticipated self-stigma of seeking help**. *Stigma Health* (2018.0) **3** 65-68. DOI: 10.1037/sah0000072
34. Yang Y, Kong X, Guo Z, Kou Y. **Can self-compassion promote gratitude and prosocial behavior in adolescents? A 3-year longitudinal study from China**. *Mindfulness* (2021.0) **12** 1377-1386. DOI: 10.1007/s12671-021-01605-9
35. Vu HA, Rivera LM. **Self-compassion and negative outgroup attitudes: the mediating role of compassion for others**. *Self Identity* (2022.0). DOI: 10.1080/15298868.2022.2117241
36. Corrigan PW, River LP, Lundin RK, Wasowski KU, Campion J, Mathisen J, Goldstein H, Bergman M, Gagnon C, Kubiak MA. **Stigmatizing attributions about mental illness**. *J Community Psychol* (2000.0) **28** 91-102. DOI: 10.1002/(SICI)1520-6629(200001)28:1<91::AID-JCOP9>3.0.CO;2-M
37. Mak WWS, Chong ESK, Wong CCY. **Beyond attributions: Understanding public stigma of mental illness with the common sense model**. *Am J Orthopsychiatry* (2014.0) **84** 173-181. DOI: 10.1037/h0099373
38. Johnson-Kwochka A, Minor KS, Ashburn-Nardo L, Wu W, Stull LG, Salyers MP. **A new look at the attribution model: considerations for the measurement of public mental illness stigma**. *Stigma Health* (2021.0). DOI: 10.1037/sah0000288
39. Ling CYM, Mak WWS, Cheng JNS. **Attribution model of stigma towards children with Autism in Hong Kong**. *J Appl Res Intellect Disabil* (2010.0) **23** 237-249. DOI: 10.1111/j.1468-3148.2008.00456.x
40. Kang S, Kang J. **Age differences in psychological antecedents and behavioral consequences of stigmatization associated with COVID-19 among Koreans**. *Int J Environ Res Publ Health* (2022.0) **19** 8594. DOI: 10.3390/ijerph19148594
41. Zhang T-M, Fang Q, Yao H, Ran M-S. **Public stigma of COVID-19 and its correlates in the general population of China**. *Int J Environ Res Publ Health* (2021.0) **18** 11718. DOI: 10.3390/ijerph182111718
42. Chen X, Huang C, Wang H, Wang W, Ni X, Li Y. **Negative emotion arousal and altruism promoting of online public stigmatization on COVID-19 pandemic**. *Front Psychol* (2021.0). DOI: 10.3389/fpsyg.2021.652140
43. Alsawalqa RO. **Cyberbullying, social stigma, and self-esteem: the impact of COVID-19 on students from East and Southeast Asia at the University of Jordan**. *Heliyon* (2021.0) **7** e06711. DOI: 10.1016/j.heliyon.2021.e06711
44. Yang F, Sun J, Li J, Lyu S. **Coping strategies, stigmatizing attitude, and cyberbullying among Chinese college students during the COVID-19 lockdown**. *Curr Psychol* (2022.0). DOI: 10.1007/s12144-022-02874-w
45. Hu Y, Wang Y, Sun Y, Arteta-Garcia J, Purol S. **Diary study: the protective role of self-compassion on stress-related poor sleep quality**. *Mindfulness* (2018.0) **9** 1931-1940. DOI: 10.1007/s12671-018-0939-7
46. Raes F, Pommier E, Neff KD, Van Gucht D. **Construction and factorial validation of a short form of the self-compassion scale**. *Clin Psychol Psychother* (2011.0) **18** 250-255. DOI: 10.1002/cpp.702
47. Chan BSM, Deng J, Li Y, Li T, Shen Y, Wang Y, Yi L. **The role of self-compassion in the relationship between post-traumatic growth and psychological distress in caregivers of children with autism**. *J Child Fam Stud* (2020.0) **29** 1692-1700. DOI: 10.1007/s10826-019-01694-0
48. Chen G, He J, Cai Z, Fan X. **Perceived parenting styles and body appreciation among Chinese adolescents: exploring the mediating roles of dispositional mindfulness and self-compassion**. *Child Youth Serv Rev* (2020.0) **119** 105698. DOI: 10.1016/j.childyouth.2020.105698
49. Mak WWS, Mo PKH, Cheung RYM, Woo J, Cheung FM, Lee D. **Comparative stigma of HIV/AIDS, SARS, and tuberculosis in Hong Kong**. *Soc Sci Med* (2006.0) **63** 1912-1922. DOI: 10.1016/j.socscimed.2006.04.016
50. Hayes AF. *Introduction to mediation, moderation, and conditional process analysis: a regression-based approach* (2018.0)
51. Podsakoff PM, MacKenzie SB, Lee J-Y, Podsakoff NP. **Common method biases in behavioral research: a critical review of the literature and recommended remedies**. *J Appl Psychol* (2003.0) **88** 879-903. DOI: 10.1037/0021-9010.88.5.879
52. Tehseen S, Ramayah T, Sajilan S. **Testing and controlling for common method variance: a review of available methods**. *J Manage Sci* (2017.0) **4** 142-168
53. Yarnell LM, Neff KD. **Self-compassion, interpersonal conflict resolutions, and well-being**. *Self and Identity* (2013.0) **12** 146-159. DOI: 10.1080/15298868.2011.649545
54. Barnard LK, Curry JF. **Self-compassion: conceptualizations, correlates, & interventions**. *Rev Gen Psychol* (2011.0) **15** 289-303. DOI: 10.1037/a0025754
55. Chan KKS, Lee CWL, Mak WWS. **Mindfulness model of stigma resistance among individuals with psychiatric disorders**. *Mindfulness* (2018.0) **9** 1433-1442. DOI: 10.1007/s12671-018-0887-2
56. Fekete EM, Herndier RE, Sander AC. **Self-compassion, internalized weight stigma, psychological well-being, and eating behaviors in women**. *Mindfulness* (2021.0) **12** 1262-1271. DOI: 10.1007/s12671-021-01597-6
57. Huellemann KL, Calogero RM. **Self-compassion and body checking among women: the mediating role of stigmatizing self-perceptions**. *Mindfulness* (2020.0) **11** 2121-2130. DOI: 10.1007/s12671-020-01420-8
58. Neff KD. **Self-compassion: an alternative conceptualization of a halthy attitude toward oneself**. *Self Identity* (2003.0) **2** 85-101. DOI: 10.1080/15298860309032
59. Wong CCY, Knee CR, Neighbors C, Zvolensky MJ. **Hacking stigma by loving yourself: a mediated-moderation model of self-compassion and stigma**. *Mindfulness* (2019.0) **10** 415-433. DOI: 10.1007/s12671-018-0984-2
60. Corrigan P, Markowitz FE, Watson A, Rowan D, Kubiak MA. **An attribution model of public discrimination towards persons with mental illness**. *J Health Soc Behav* (2003.0) **44** 162-179. DOI: 10.2307/1519806
61. Álvarez-García D, Núñez JC, García T, Barreiro-Collazo A. **Individual, family, and community predictors of cyber-aggression among adolescents**. *European J Psychol Appl Legal Context* (2018.0) **10** 79-88. DOI: 10.5093/ejpalc2018a8
62. Bullo A, Schulz PJ. **Parent-child communication, social norms, and the development of cyber aggression in early adolescence**. *J Youth Adolescence* (2022.0). DOI: 10.1007/s10964-022-01625-1
63. Cebollero-Salinas A, Orejudo S, Cano-Escoriaza J, Iniguez-Berrozpe T. **Cybergossip and problematic internet use in cyberaggression and cybervictimisation among adolescents**. *Comput Human Behav* (2022.0). DOI: 10.1016/j.chb.2022.107230
64. Zych I, Kaakinen M, Savolainen I, Sirola A, Paek H-J, Oksanen A. **The role of impulsivity, social relations online and offline, and compulsive internet use in cyberaggression: a four-country study**. *New Media Soc* (2023.0) **25** 181-198. DOI: 10.1177/14614448211009459
65. Garofalo C, Gillespie SM, Velotti P. **Emotion regulation mediates relationships between mindfulness facets and aggression dimensions**. *Aggressive Behav* (2020.0) **46** 60-71. DOI: 10.1002/ab.21868
66. Garofalo C, Velotti P, Zavattini GC. **Emotion regulation and aggression: the incremental contribution of alexithymia, impulsivity, and emotion dysregulation facets**. *Psychol Violence* (2018.0) **8** 470-483. DOI: 10.1037/vio0000141
67. Palmeira L, Cunha M, Pinto-Gouveia J. **Processes of change in quality of life, weight self-stigma, body mass index and emotional eating after an acceptance-, mindfulness- and compassion-based group intervention (Kg-Free) for women with overweight and obesity**. *J Health Psychol* (2019.0) **24** 1056-1069. DOI: 10.1177/1359105316686668
68. González-García M, Álvarez JC, Pérez EZ, Fernandez-Carriba S, López JG. **Feasibility of a brief online mindfulness and compassion-based intervention to promote mental health among university students during the COVID-19 pandemic**. *Mindfulness* (2021.0) **12** 1685-1695. DOI: 10.1007/s12671-021-01632-6
69. Valeri L, Amsalem D, Jankowski S, Susser E, Dixon L. **Effectiveness of a video-based intervention on reducing perceptions of fear, loneliness, and public stigma related to COVID-19: a randomized controlled trial**. *Int J Public Health* (2021.0) **66** 1604164. DOI: 10.3389/ijph.2021.1604164
|
---
title: Social Isolation, Healthy Habits, Inequality and Mental Health in the United
States
authors:
- Ignacio Amate-Fortes
- Almudena Guarnido-Rueda
- Diego Martínez-Navarro
- Francisco J. Oliver-Márquez
journal: Applied Research in Quality of Life
year: 2023
pmcid: PMC9999329
doi: 10.1007/s11482-023-10155-2
license: CC BY 4.0
---
# Social Isolation, Healthy Habits, Inequality and Mental Health in the United States
## Abstract
The objective of this work is to deepen the analysis of the socioeconomic determinants of mental health, paying special attention to the impact of inequality, not only in income distribution but also in gender, racial, health and education inequality, social isolation, including new variables to measure loneliness, and healthy habits, on the mental health status. For this purpose, a cross-sectional model for a sample of 2735 counties in the United *States is* estimated using Ordinary Least Squares in its robust version to solve the detected heteroscedasticity problems. The results obtained show that inequality, social isolation and certain lifestyles, such as smoking or insomnia, are detrimental to mental health, while sexual activity prevents mental distress. On the other hand, poor counties suffer more cases of suicide, with food insecurity being the main problem for mental health. Finally, we found detrimental effects of pollution on mental health.
## Introduction
In recent decades, the role of mental health has grown in importance, not only for the scientific community, but also for policy makers as reflected by the fact that it has been incorporated into the Sustainable Development Goals. It is important to note that depression is one of the leading causes of disability and that suicide is the leading cause of death in the population between 15 and 29 years of age. In fact, according to the World Health Organization (WHO), more than 700,000 people die by suicide each year.
The United *States is* not immune to this problem and with a mortality rate of 16.1 per 100,000 inhabitants, it is one of the countries with the highest suicide rates. For all these reasons, in this paper we propose to analyze the socioeconomic determinants of mental health in the United States. To this end, we base our analysis on three pillars: firstly, inequality, understood in a broad sense, i.e., inequality in income distribution, gender, race, health, education and the labor market. This in-depth analysis of the incidence of inequality on mental health is the main novelty of this work. Second, we use several variables as proxies for social isolation to test how they affect mental health. In this sense, the use of new variables such as teleworking or driving alone every day to work is another important novelty of this article. Finally, the main lifestyle habits are analyzed to contribute to the analysis of the effect of these variables on mental health.
For this purpose, a robust cross-sectional model was estimated for a sample ranging from 1790 to 2735 U.S. counties (depending on the availability of data for certain variables). The results obtained show that inequality in all its aspects is indeed a risk factor for mental disorders, although social isolation is perhaps more important as an explanatory variable. Finally, tobacco addiction and insomnia are shown to be the habits most detrimental to mental health.
The second empirical analysis establishes the theoretical framework and then, in the third section, explains the model and discusses the results. Finally, in the fourth section, the conclusions are developed.
## Theoretical Framework
The economic literature has extensively studied the effects of inequality on different health outcomes (Pickett & Wilkinson, 2015; Matthew & Brodersen, 2018), and among these, some authors have addressed the relationship between inequality and mental health. Thus, works such as Burns et al. [ 2017], have analyzed the relationship between inequality in income distribution and certain mental disorders. However, as pointed out by Patel et al. [ 2018], a review of the papers published on the relationship between income inequality and mental health shows inconsistent results, with only one third of them concluding that inequality in income distribution is a risk factor for mental health.
Less studied is the case of the association between other forms of inequality and the prevalence of mental disorders. In this sense, there is a lack of work addressing the incidence of gender inequality on mental health (Yu, 2018). Even so, we can highlight the works of Hopcroft and Bradley [2007], and Van de Velde et al. [ 2013], who perform a macro-level analysis of the effects of gender inequality on mental health. As with inequality in income distribution, research on the effects of gender inequality on mental health reflects inconclusive results (Hopcroft & Bradley, 2007; Seedat et al., 2009; Van de Velde et al., 2013; Hagen & Rosenstrôm, 2016).
Most papers that have studied racial inequality as a risk factor for mental health have measured this racial inequality through discrimination (Brown et al., 2000; Lewis et al., 2015; Wallace et al., 2016; Mouzon et al., 2017; Williams, 2018). In this case, the results are indeed conclusive and point out that racial discrimination negatively affects mental well-being. Our work aims to delve deeper into the impact of racial inequality on mental health, measuring this inequality through the unequal distribution of poverty across races.
Regarding social isolation as a determinant of mental health, there is a broad consensus from researchers about the positive impact of interpersonal relationships on mental well-being (Almedom, 2005; Bassett & Moore, 2013), However, an associated problem encountered by researchers is that it is unclear how social isolation, loneliness, and other related concepts should be measured when analyzing their effect on mental health (Windle et al., 2011; Courtin & Knapp, 2015; Rhode et al., 2016; Chirstiansen et al., 2021). Therefore, we propose new measures of social isolation such as teleworking and driving alone to work. In doing so, we aim to give robustness to the results obtained by the already published works.
Finally, the economic literature has also paid attention to the association between healthy habits and mental health. Thus, authors such as Reid et al. [ 2009], Taylor et al. [ 2011], Milojevich and Lukowaki [2016], Chattu et al. [ 2018], Sullivan and Ordiah [2018], and Merikanto and Partonen [2021] among others warn of the adverse effects of insomnia on mental health. Also, the impact of tobacco and alcohol addiction on mental health has aroused the interest of researchers, highlighting the works on the adolescent population by Mason et al. [ 2008], Balogun et al. [ 2014], Skogen et al. [ 2014], and Ferreira et al. [ 2019]. Likewise, the relationship between obesity, physical activity, and mental health has been analyzed (Kivimâki et al., 2009). However, the mechanisms linking obesity and mental illness are unclear (Avila et al., 2015). Thus, there are authors who point out that mental disorders are the cause of obesity (Nicholson, 1946), others speak of a bidirectional relationship (Cameron et al., 2012) and others point to obesity as a risk factor for mental health (De Hert et al., 2011).
Therefore, as mental disorders cannot be explained solely through genetic factors (Sanders et al., 1999; Sullivan et al., 2000; Fava & Kendler, 2000), and given the importance of socioeconomic determinants, we propose, from here on, to continue to deepen the analysis of the incidence of these factors on mental health, with emphasis on inequality, social isolation and healthy living habits.
## Empirical Analysis
A cross-sectional linear model has been estimated to analyze whether social isolation, lifestyle, and inequality, broadly understood, observed in each North American county have any effect on mental health in the United States. The mental health data were obtained from the County Health Rankings & Roadmaps, University of Wisconsin Health Institute, and refer to 2019. In this sense, we have worked with a database of 3,218 U.S. counties, which is almost $100\%$ of all counties in the U.S. Even so, the inequality and mental health data by counties have only allowed us to use a sample between 1,790 and 2,735 counties, depending on the inequality and mental health measure used. In any case, the sample used is representative of the overall U.S. situation.
This study adapts the classic model of Dalghren and Whitehead [1991] for a comparative analysis between counties in the United States. The model of these two economists has been widely used and shows the determinants of health in concentric layers, from structural determinants (external layer) to individual lifestyles (internal layer), placing at the center the characteristics of individuals that cannot be modified, such as sex, age or constitutional factors (Fig. 1).
Fig. 1The Dalghren-Whitehead model of determinants in health. Source: Dalghren and Whitehead [1991] According to these authors, individuals are endowed with risk factors such as age, sex and other genetic factors that influence their potential for ultimate health. Likewise, personal behaviors and lifestyles also play a role. People who are economically disadvantaged tend to exhibit behaviors that depart from healthy living, such as smoking, alcohol and drug abuse, and poor diet. On the other hand, labor and environmental conditions, and access to basic services constitute another set of determinants of health status. Differences in housing conditions, occupational risks, whether one has a job, and the possibility of having free, quality education, basic health services, and infrastructure access to drinking water, sewage systems, paved roads, are key factors in the differences in health shown by different social groups. Finally, the economic, cultural and environmental conditions prevailing in society as a whole, as well as the economic situation of the country, will also affect the health outcomes of the population as a whole.
In our case, we adapt this model to analyze the socioeconomic determinants of mental health.
A. Data The variables used in this work are summarized in the following table (Table 1): Table 1Variable definitions and summary statisticsVariableDescriptionObs. MeanStd. Dev. MinMaxPoor mental health daysIt measures the average number of mentally unhealthy days reported in past 30 days during 2019. Sources: US State and Local Health Agencies. Source: County Health Rankings & Roadmaps 2022 (University of Wisconsin Population Health Institute). https://www.countyhealthrankings.org Accessed on July 28, 2022.31434.890.693.247.46Frequent mental distressPercentage of adults reporting 14 or more days of poor mental health per month (age-adjusted). Frequent mental distress is a derived measure of poor mental health days. It provides a slightly different picture that emphasizes those experiencing more chronic and likely severe mental health problems. Source: County Health Rankings & Roadmaps 2022 (University of Wisconsin Population Health Institute). https://www.countyhealthrankings.org Accessed on August 19, 2022.31410.160.030.0970.26SuicideNumber of deaths by suicide per 100,000 population (age-adjusted). Suicide serves as an important measure of the mental health of a county’s population. Outside of the impact on the emotional and mental health of surviving friends, family members, and loved ones, suicide also has an economic impact and costs the United States an estimated $70 billion per year (Centers for Disease Control and Prevention. Suicide Prevention – Facts About Suicide. Last August 30, 2021. Accessed on August 22, 2022. https://www.cdc.gov/suicide/facts/index.htm Source: County Health Rankings & Roadmaps 2022 (University of Wisconsin Population Health Institute). https://www.countyhealthrankings.org Accessed on August 19, 2022.243318.947.884.98155.37Mean relative incomeIt is defined as the county inflation-adjusted mean household income relative to State mean household income. The data refers to 2020. It is a measure of inequality that aims to study whether poorer populations within a State are more vulnerable to the effects of the pandemic. Source: Prepared by the authors based on data from the United States Census Bureau. https://www.census.gov Accessed on July 21, 2022.30811.160.250.582.74InequalityInequality index by county. The following 5 inequality indexes have been used:• Gini Index (Income inequality). Measure of inequality in the distribution of the county´s households’ income. The value of the index varies between 0 and 1. Source: American Community Survey 2020 (5-Year Estimates). United States Census Bureau. https://www.census.gov Accessed on July 18, 2022.32210.450.040.080.70• $\frac{80}{20}$ Index (Income inequality). A measure of inequality that relates the percentage of the county’s households´ income obtained by the top $20\%$ of income to the bottom $20\%$. Source: Own elaboration based on American Community Survey 2020 (5-Year Estimates). United States Census Bureau. https://www.census.gov Accessed on July 18, 2022.321014.106.624.72190.22• Gender Pay Gap (Gender inequality). It is expressed as “cents on the dollar,“ or women’s median earnings in cents compared to every dollar (100 cents) of men’s median earnings. Women’s median earnings are the level of earnings where half of full-time, year-round, female workers are earning more than this value, and half are earning less. Men’s median earnings follow the same definition for male workers. The wage gap still exists despite the fact that women make up the majority of college-educated adults in the U.S. and that decades after the Equal Pay Act of 1963 outlawed paying men and women different wages for similar work. Source: County Health Rankings & Roadmaps 2022 (University of Wisconsin Population Health Institute). https://www.countyhealthrankings.org Accessed on July 28, 2022.31370.770.0980.421.57• Female poverty (Gender inequality). It measures the percentage of women below the poverty thresholds out of the total number of poor people. As noted by Kawachi et al. [ 1999], and Milner et al. [ 2021], gender inequality is associated with worse mortality outcomes, poorer self-rated health, and greater disability. Source: Own elaboration based on American Community Survey 2020 (5-Year Estimates). United States Census Bureau. https://www.census.gov Accessed on July 19, 2022.32200.560.050.200.81• Black poverty (racial inequality). It is the percentage of black population below the poverty threshold out of the total poor population. Source: Own elaboration based on American Community Survey 2020 (5-Year Estimates). United States Census Bureau. https://www.census.gov Accessed on July 19, 2022.27530.150.2100.98• Racial segregation (racial inequality). It refers to the degree to which black and white residents live separately from one another in a county. The residential segregation index ranges from 0 (complete integration) to 100 (complete segregation). The elimination of discriminatory policies and practices has had an impact on acts of racism, but has had little effect on structural racism, such as residential segregation, resulting in persistent structural inequalities. Residential segregation is a key determinant of racial differences in socioeconomic mobility and, in addition, can create social and physical risks in residential settings that negatively affect health. Although this research field is gaining interest, structural forms of racism and its relationship to health inequalities remain understudied (Gee & Ford, 2011). These authors and Kramer and Hogue [2009] assert that residential segregation is considered a root cause of health disparities in the U.S. and has been linked to poor health outcomes, such as mortality, a wide variety of reproductive, infectious, and chronic diseases, and other adverse conditions. Source: County Health Rankings & Roadmaps 2022 (University of Wisconsin Population Health Institute). https://www.countyhealthrankings.org Accessed on July 28, 2022.207849.116.620.5197.01Mental health providersRefers to the number of mental health providers per 1,000 population. Access to care requires not only financial coverage, but also access to providers. Source: County Health Rankings & Roadmaps 2022 (University of Wisconsin Population Health Institute). https://www.countyhealthrankings.org Accessed on July 28, 2022.29463.0417.290478.57UninsuredIt is the percentage of the population under age 65 without health insurance coverage. A person is uninsured if they are currently not covered by insurance through a current/former employer or union, purchased from an insurance company, Medicare, Medicaid, Medical Assistance, any kind of government-assistance plan for those with low incomes or disability, TRICARE or other military health care, Indian Health Services, VA, or any other health insurance or health coverage plan. More than $8\%$ of the population remains uninsured in the U.S. Lack of health insurance coverage is a major barrier to accessing needed health care and maintaining economic security and using it is intended to test whether it has had any effect on the mental health status. Source: County Health Rankings & Roadmaps 2022 (University of Wisconsin Population Health Intistute). https://www.countyhealthrankings.org Accessed on July 28, 202232210.120.0500.36UniversityIt measures the percentage of the population with university studies. The use of this variable is intended to test whether more education influences mental health. As mentioned above, the aim is to test whether education improves health equity, as argued by Amate-Fortes et al. [ 2020]. The data refer to 2020. Source: United States Census Bureau. https://www.census.gov Accessed on July 21, 2022.32210.160.0700.57UnemploymentUnemployment rate in 2020 by county. The goal is to verify if higher unemployment rate has any effect on mental health. Source: United States Census Bureau. https://www.census.gov Accessed on July 19, 2022.32210.050.0300.35Bad healthIt measures the percentage of adults in a county who consider themselves to be in poor or fair health (age adjusted) during 2019. Source: County Health Rankings & Roadmaps 2022 (University of Wisconsin Population Health Institute). https://www.countyhealthrankings.org Accessed on July 28, 2022.32210.200.0600.45SleepPercentage of adults who report fewer than 7 h of sleep on average per day(age-adjusted) during 2018. Sleep is an important part of a healthy lifestyle, and a lack of sleep can cause psychiatric disorders such as depression and anxiety, risky behavior, and even suicide. Source: County Health Rankings & Roadmaps 2022 (University of Wisconsin Population Health Institute). https://www.countyhealthrankings.org Accessed on July 28, 2022.31430.370.0400.49SmokeAdult *Smoking is* the percentage of the adult population in a county who both report that they currently smoke every day or some days and have smoked at least 100 cigarettes in their lifetime. Each year, according to US Department of Health and Human Services, approximately 480,000 premature deaths can be attributed to smoking. Source: County Health Rankings & Roadmaps 2022 (University of Wisconsin Population Health Institute). https://www.countyhealthrankings.org Accessed on July 28, 2022.32210.200.0500.43ObesityIt is based on responses to the Behavioral Risk Factor Surveillance Survey (BRFSS) and is the percentage of the adult population (ages 18 and older) that reports a body mass index (BMI) greater than or equal to 30 kg/m2. Participants are asked to self-report their height and weight. From these reported values, BMIs for the participants are calculated. Source: County Health Rankings & Roadmaps 2022 (University of Wisconsin Population Health Institute). https://www.countyhealthrankings.org Accessed on July 28, 2022.32210.350.0700.51InactivityPhysical *Inactivity is* based on responses to the Behavioral Risk Factor Surveillance Survey and is the percentage of adults ages 18 and over reporting no leisure-time physical activity in the past month. Physical inactivity is not only associated with individual behavior, but also with community conditions, such as spending on recreational activities, access to infrastructure, and poverty (Lee et al., 2017). Source: County Health Rankings & Roadmaps 2022 (University of Wisconsin Population Health Institute). https://www.countyhealthrankings.org Accessed on July 28, 2022.32210.300.0700.52AlcoholExcessive Drinking measures the percentage of a county’s adult population that reports binge or heavy drinking in the past 30 days. The Centers for Disease Control and Prevention [2009] warn of significant adverse health effects. Source: County Health Rankings & Roadmaps 2022 (University of Wisconsin Population Health Institute). https://www.countyhealthrankings.org Accessed on July 28, 2022.32210.190.0400.30Sexual transmitted infections (STI)It the number of newly diagnosed chlamydia cases per 100,000 population in a county. We use this indicator as a proxy variable for the population’s sexual activity. This will allow us to analyze whether those counties with a greater sexual activity also have a better mental health. Source: County Health Rankings & Roadmaps 2022 (University of Wisconsin Population Health Institute). https://www.countyhealthrankings.org Accessed on July 28, 2022.3025418.3293.603848.9DensityVariable that measures the population per square mile of land area by county. The aim is to verify whether the larger counties, which tend to have a higher population density, have a better mental health. Source: United States Census Bureau. https://www.census.gov Accessed on July 22, 2022.3182258.41706.3069468.4AssociationsNumber of membership associations per 10,000 population in 2019. Source: County Health Rankings & Roadmaps 2022 (University of Wisconsin Population Health Intistute). https://www.countyhealthrankings.org Accessed on August 19, 2022.314311.445.91055.4Work from home (WFH)Percentage of workers who teleworked in 2018. The objective of using this variable is to test whether social isolation has an impact on mental health. Source: National Association of Realtors. https://www.nar.realtor/research-and-statistics/research-reports/work-from-home-counties Accessed on August 3, 2022.30924.752.83027.3DrivingPercentage of the workforce that drives alone to work. Source: County Health Rankings & Roadmaps 2022 (University of Wisconsin Population Health Institute). https://www.countyhealthrankings.org Accessed on August 19, 2022.31430.790.0800.99BroadbandPercentage of households with broadband internet connection. Source: County Health Rankings & Roadmaps 2022 (University of Wisconsin Population Health Instute). https://www.countyhealthrankings.org Accessed on August 19, 2022.31410.780.080.330.97Severe housing problemsPercentage of households with at least 1 of 4 housing problems: overcrowding, high housing costs, lack of kitchen facilities, or lack of plumbing facilities. Source: County Health Rankings & Roadmaps 2022 (University of Wisconsin Population Health Institute). https://www.countyhealthrankings.org Accessed on August 19, 2022.31430.130.0400.70FoodPercentage of the population lacking adequate access to food in 2019. Source: County Health Rankings & Roadmaps 2022 (University of Wisconsin Population Health Institute). https://www.countyhealthrankings.org Accessed on August 19, 2022.31410.130.040.030.29PollutionAir Pollution - Particulate *Matter is* a measure of the fine particulate matter in the air. It is reported as the average daily density of fine particulate matter in micrograms per cubic meter. Fine particulate matter is defined as particles of air pollutants with an aerodynamic diameter less than 2.5 micrometers (PM2.5). Fuente: County Health Rankings & Roadmaps 2022 (University of Wisconsin Population Health Institute). https://www.countyhealthrankings.org Accessed on July 28, 2022.31158.021.652.520.9Source: Own elaboration B. The model
A linear model was developed and estimated through Ordinary Least Squares in its robust version of variances and covariances, since when the Breusch-Pagan test was performed, the p-value obtained showed the presence of heteroscedasticity. The model was estimated without a constant term. Although the decision to use a constant term or not is a problem that generates much discussion (Casella, 1983), nevertheless, there are circumstances in which it is appropriate or even necessary not to use the error term. As Eisenhauer [2003] points out, in the case where the dependent variable is zero if the vector of independent variables is also zero, the error term can be omitted. This is the case of the estimated model where variables such as population density are used. If this variable had a value equal to zero, the variables measuring mental health status would also have a value equal to zero.
The model used is as follows:1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{array}{lllll}MENTALHEALTH={\beta }_{1}INCOME+{\beta }_{2}INEQUALITY+{\beta }_{3}MHP+{\beta }_{4}UNINSURED+ \\ {\beta }_{5}UNIVERSITY+{\beta }_{6}UNEMPLOYMENT+{\beta }_{7}BADHEALTH+{\beta }_{8}SLEEP+ \\ {\beta }_{9}SMOKE+{\beta }_{10}OBESITY+{\beta }_{11}INACTIVITY+{\beta }_{12}ALCOHOL{+\beta }_{13}STI+ \\ {\beta }_{14}DENSITY+{\beta }_{15}ASSOCIATIONS+{\beta }_{16}WFH+{\beta }_{17}DRIVING+ \\ {\beta }_{18}BROADBAND+{\beta }_{19}SHP+{\beta }_{20}FOOD+{\beta }_{21}POLLUTION+{\mu }_{I}\end{array}$$\end{document}MENTALHEALTH=β1INCOME+β2INEQUALITY+β3MHP+β4UNINSURED+β5UNIVERSITY+β6UNEMPLOYMENT+β7BADHEALTH+β8SLEEP+β9SMOKE+β10OBESITY+β11INACTIVITY+β12ALCOHOL+β13STI+β14DENSITY+β15ASSOCIATIONS+β16WFH+β17DRIVING+β18BROADBAND+β19SHP+β20FOOD+β21POLLUTION+μI Where, *Mentalhealth is* the dependent variable. In this sense, three variables that reflect the mental health status have been used, each of them implying an aggravation of mental disorders. Thus, first, the variable “Poor mental health days” has been used, which measures the average number of days of mental unhealthiness reported in the last 30 days during 2019. The second variable used is “Frequent mental distress” which reflects the percentage of adults reporting 14 or more poor mental health days per month (age-adjusted), therefore, it emphasizes the population experiencing more chronic and probably more severe mental health problems. Finally, the variable “Suicide” was used, which measures the number of suicide deaths per 100,000 population (age-adjusted). This variable reflects the extreme case of a mental health problem. The objective, therefore, is to analyze how the independent variables used affect mental health and how these effects change as mental illness worsens.
Income measures the average real income per household in the county in relation to the average real income per household in the state. It is, therefore, a first variable that measures inequality, in this case, between counties.
Inequality is one of the explanatory variables on which we have focused the objective of this work, i.e., the aim is to analyze how inequality within each county affects mental health. In this sense, we have tried to analyze inequality in a broad sense, that is, not only inequality in income distribution, but also gender inequality and racial inequality. For this purpose, six measures of inequality were used: The Gini index and the $\frac{80}{20}$ ratio have been used to measure inequality in income distribution. In terms of gender inequality, the Gender Pay Gap, which measures the average earnings of women in relation to men, and the female poverty variable, which refers to the percentage of poor women in relation to the total poor population, have been used. Racial inequality is measured by the percentage of the African American poor population out of the total population below the poverty line, and by racial segregation, i.e., the degree to which black and white residents live separately from each other in a county.
MHP refers to the number of mental health providers per 1,000 population. It is therefore a proxy variable for access to mental health care, since access to care requires not only financial coverage, but also access to providers.
Uninsured refers to the percentage of people under age 65 who did not have health insurance in 2019. This is a proxy variable for health inequality in terms of health coverage. Therefore, its use is intended to strengthen the analysis of the effects of inequality on mental health.
University measures the percentage of the population with university studies. As with the previous variable, this is a proxy variable for educational inequality and will allow us to delve deeper into inequality as a determinant of mental health.
Unemployment measures the unemployment rate in 2020. This variable shows, on the one hand, the inequality in the labor market between counties and, on the other hand, the lack of income.
Badhealth measures the percentage of adults in a county who consider themselves to be in poor or fair (age-adjusted) health during 2019. The purpose of using this variable is to test whether physical health has a relationship to mental health.
Sleep refers to the percentage of adults reporting having slept less than 7 h on average per day (age-adjusted) during 2018. Sleep is an important part of a healthy lifestyle, and by employing this variable we try to analyze whether lack of sleep can cause psychiatric disorders.
Smoke refers to the percentage of a county’s adult population reporting smoking every day or some days, and having smoked at least 100 cigarettes in their lifetime, in 2019. We use “Smoke” as a proxy variable for addiction.
Obesity is the percentage of the adult population with a body mass index (BMI) equal to or greater than 30 kg/m2. The objective is to see if obesity is a cause of poor mental health.
Inactivity is the percentage of adults aged 18 years and older reporting no leisure-time physical activity in the last month during 2019.
Alcohol measures the percentage of a county’s adult population reporting binge drinking in the past 30 days, during 2019.
STI refers to sexually transmitted diseases measured through the number of newly diagnosed chlamydia cases per 100,000 population in 2019. We use “STI” as a proxy variable for sexual activity.
Density is the population density of the county in 2020, i.e., number of inhabitants per km2. This is the first variable that will allow us to analyze the effects of social isolation on mental health.
Associations measures the number of membership associations per 10,000 inhabitants in 2019.
WFH refers to “work from home,“ i.e., the percentage of people who teleworked in 2018. Teleworking prevents physical contact with coworkers and is therefore a good proxy for loneliness.
Driving measures the percentage of the labor force that drives alone to work.
Broadband is the percentage of households with a broadband Internet connection. We use this variable as a proxy for home Internet use.
SHP is “Severe housing problems” and refers to the percentage of households with at least 1 of these 4 housing problems: overcrowding, high housing costs, lack of kitchen facilities or lack of plumbing facilities. With this, we want to check whether these types of problems lead to poorer mental health.
Food measures the percentage of the population lacking adequate access to food in 2019.
Pollution refers to air pollution - particulate matter and is a measure of fine particles in the air. It is presented as the daily average density of fine particles in micrograms per cubic meter. Fine particulate matter is defined as air pollutant particles with an aerodynamic diameter of less than 2.5 micrometers (PM2.5).
C. Results and Discussion As mentioned above, the model was estimated by OLS in its robust version to solve the problem of heteroscedasticity detected. Eighteen estimates have been made, resulting from the use of 6 different measures of inequality and the three dependent variables used to characterize mental health. The results are reflected in the following tables (Tables 2, 3, and 4): Table 2Results of the estimations (income inequality)Poor mental health daysFrequent mental distressSuicideGini$\frac{80}{20}$Gini$\frac{80}{20}$Gini$\frac{80}{20}$Mean relative income0.19***(5.41)0.20***(5.79)0.003***(2.77)0.003***(3.17)-2.45***(-2.94)-2.63***(-3.29)Inequality0.79***(4.32)0.001(0.58)0.02***(4.09)0.00004(0.57)-1.00(-0.20)0.10(0.56)Mental health provider-0.0003*(-1.66)-0.0003(-1.47)-0.00007(-0.97)-0.00005(-0.80)-0.006(-0.51)-0.006(-0.49)Uninsured-1.001***(-5.90)-0.99***(-5.72)-0.03***(-6.33)-0.03***(-6.18)23.82***(5.16)24.11***(4.94)University-0.31**(-1.99)-0.17(-1.10)-0.007(-1.40)-0.003(-0.59)-10.76**(-2.36)-11.92**(-2.31)Unemployment0.08(0.21)0.17(0.45)-0.009(-0.71)-0.006(-0.48)36.66*(1.65)35.88**(1.94)Bad health3.91***(10.37)4.09***(10.69)0.14***(11.63)0.15***(11.99)-13.49(-1.18)-14.72(-1.23)Sleep5.19***(19.05)5.30***(19.19)0.14***(16.15)0.14***(16.25)19.41*(1.80)20.50*(1.81)Smoke8.38***(24.74)8.34***(24.22)0.33***(29.78)0.33***(29.19)51.12***(3.01)51.52***(3.05)Obesity-1.52***(-6.23)-1.53***(-6.18)-0.03***(-4.20)-0.03***(-4.19)-18.09***(-3.16)-18.32***(-3.19)Inactivity-2.77***(-9.54)-2.70***(-9.27)-0.09***(-9.53)-0.08***(-9.28)-14.32**(-2.17)-14.84**(-2.41)Alcohol-1.48***(-6.47)-1.27***(-5.62)-0.06***(-7.91)-0.05***(-7.15)4.40(0.79)2.48(0.42)Sexual transmitted infections-0.0002***(-4.98)-0.0002***(-4.92)-0.0002***(-4.98)-0.00004***(-3.83)0.0003(0.24)-0.00002(-0.03)Density-0.00002**(-2.31)-0.00002**(-2.09)-0.00004***(-2.76)-0.00007**(-2.55)-0.0003(-1.16)-0.0003(-1.18)Associations-0.0002(-0.17)0.0006(0.49)-0.00004(-1.00)-0.00002(-0.41)-0.03(-0.60)-0.03(-0.63)Work from home0.008***(2.57)0.009***(3.12)0.0003***(2.60)0.0003***(3.08)1.16***(11.41)1.16***(11.58)Driving0.94***(7.47)1.03***(8.29)0.03***(6.45)0.03***(7.22)16.61***(4.84)16.93***(4.82)Broadband0.40***(4.33)0.42***(4.45)0.01***(3.67)0.01***(3.81)13.46***(5.59)13.99***(4.53)Severe housing problems-0.46*(-1.90)-0.30(-1.21)-0.01**(-1.97)-0.01(-1.34)-20.86***(-2.64)-23.45***(-4.16)Food4.40***(13.22)4.66***(14.09)0.16***(15.85)0.17***(16.63)66.98***(6.71)63.81***(4.74)Pollution0.02***(6.70)0.02***(6.62)0.0006***(5.09)0.0006***(5.01)-0.72***(-6.19)-0.71***(-6.38)Number of observations273627332736273323072305R20.99720.99720.99730.99730.90570.9060*Significant at $10\%$ **Significant at $5\%$ ***Significant at $1\%$ Table 3Results of the estimations (gender inequality)Poor mental health daysFrequent mental distressSuicideGPGFemale povertyGPGFemale povertyGPGFemale povertyMean relative income0.21***(5.97)0.20***(5.70)0.003***(3.16)0.003***(3.05)-2.84***(-3.50)-2.55***(-3.08)Inequality0.07(1.11)0.43***(4.02)-0.0001(-0.07)0.01***(3.50)-4.01(-1.39)7.32**(2.18)Mental health provider-0.0003(-1.48)-0.0002(-1.11)-0.00005(-0.82)-0.00004(-0.55)-0.006(-0.48)-0.005(-0.40)Uninsured-0.99***(-5.80)-0.98***(-5.78)-0.03***(-6.22)-0.03***(-6.22)24.01***(5.14)23.96***(5.21)University-0.20(-1.31)-0.17(-1.10)-0.002(-0.49)-0.003(-0.57)-8.80*(-1.93)-11.27***(-2.61)Unemployment0.19(0.52)-0.17(-1.11)-0.006(-0.46)-0.005(-0.44)35.09*(1.67)36.98**(1.69)Bad health4.06***(10.69)4.07***(10.75)0.15***(12.05)0.15***(12.05)-11.59(-1.06)-14.73(-1.25)Sleep5.29***(19.26)5.24***(19.17)0.14***(16.35)0.14***(16.25)18.92*(1.77)20.67*(1.87)Smoke8.32***(24.27)8.38***(24.65)0.33***(29.29)0.33***(29.67)53.17***(2.94)51.94***(3.04)Obesity-1.52***(-6.17)-1.59***(-6.50)-0.03***(-4.19)-0.03***(-4.44)-18.09***(-3.16)-18.90***(-3.32)Inactivity-2.69***(-9.27)-2.76***(-9.53)-0.08***(-9.28)-0.08***(-9.51)-14.58**(-2.26)-15.49**(-2.37)Alcohol-1.28***(-5.56)-1.32***(-5.87)-0.05***(-6.98)-0.06***(-7.41)5.65(1.00)2.94(0.52)Sexual transmitted infections-0.0002***(-4.83)-0.0002***(-4.78)-0.0002***(-3.71)-0.00004***(-3.70)0.0004(0.34)0.0003(0.28)Density-0.00002**(-2.10)-0.00002**(-2.26)-0.00007**(-2.50)-0.00007***(-2.70)-0.0003(-1.10)-0.0003(-1.32)Associations0.0005(0.44)0.0001(0.10)-0.00002(-0.43)-0.00003(-0.74)-0.03(-0.57)-0.04(-0.74)Work from home0.009***(2.95)0.008***(2.71)0.0003***(3.10)0.0003***(2.76)1.18***(11.28)1.14***(11.35)Driving1.01***(7.99)0.93***(7.44)0.03***(7.10)0.03***(6.47)17.42***(5.23)14.63***(4.11)Broadband0.40***(4.22)0.34***(3.57)0.01***(3.75)0.009***(3.03)14.48***(5.50)12.26***(5.06)Severe housing problems-0.29(-1.22)-0.29(-1.23)-0.009(-1.19)-0.01(-1.32)-19.06**(-2.32)-21.43***(-2.91)Food4.71***(14.53)4.67***(14.51)0.17***(17.05)0.17***(17.10)65.96***(6.17)66.45***(6.36)Pollution0.02***(6.61)0.02***(6.64)0.0006***(4.97)0.0006***(5.03)-0.72***(-6.18)-0.72***(-6.20)Number of observations273527362735273623072307R20.99720.99720.99730.99730.90590.9059*Significant at $10\%$ **Significant at $5\%$ ***Significant at $1\%$
Table 4Results of the estimations (racial inequality)Poor mental health daysFrequent mental distressSuicideBlack povertySegregationBlack povertySegregationBlack povertySegregationMean relative income0.20***(5.38)0.17***(4.36)0.003***(2.88)0.002(1.56)-1.97***(-2.67)-1.68**(-2.23)Inequality0.15***(3.18)0.001***(3.21)0.004***(2.98)0.00003**(2.52)-10.32***(-9.88)0.03***(2.87)Mental health provider-0.0002(-1.15)-0.0001(-0.94)-0.00003(-0.59)-0.00002(-0.42)-0.007(-0.65)-0.001(-0.15)Uninsured-1.12***(-6.28)-1.33***(-6.93)-0.04***(-6.59)-0.04***(-7.36)21.59***(5.62)18.68***(4.66)University-0.21(-1.26)0.11(0.62)-0.002(-0.47)0.009*(1.69)-5.44(-1.40)-9.82**(-2.38)Unemployment0.21(0.58)0.43(1.14)-0.003(-0.29)0.0002(0.02)30.63***(3.13)1.55(0.18)Bad health3.77***(9.36)4.02***(9.16)0.14***(10.79)0.15***(11.17)-15.77*(-1.74)-6.17(-0.67)Sleep4.70***(16.66)4.21***(13.86)0.12***(14.15)0.10***(11.40)0.47(0.08)-4.54(-0.82)Smoke8.67***(23.57)8.52***(21.13)0.35***(28.81)0.35***(26.23)25.62***(2.80)30.29***(3.66)Obesity-1.66***(-6.31)-2.15***(-7.69)-0.04***(-4.51)-0.05***(-6.14)-20.09***(-3.83)-21.11***(-3.99)Inactivity-2.05***(-6.58)-1.18***(-3.51)-0.06***(-6.77)-0.04***(-4.26)-11.13**(-2.08)-14.41**(-2.49)Alcohol-1.15***(-4.78)-1.21***(-4.65)-0.05***(-6.08)-0.05***(-5.21)-0.12(-0.02)-4.76(-0.95)Sexual transmitted infections-0.0002***(-4.89)-0.0001***(-3.29)-0.00005***(-3.70)-0.00002(-1.50)0.003***(2.69)-0.002**(-2.30)Density-0.00002*(-1.78)-0.00001(-1.60)-0.00006**(-2.31)-0.00006**(-2.26)-0.00009(-0.49)-0.0002(-1.15)Associations0.0007(0.51)-0.001(-0.68)-0.00003(-0.60)-0.0001**(-2.42)0.05(1.21)-0.11**(-2.42)Work from home0.01***(3.92)0.009**(2.48)0.0004***(4.13)0.0004***(2.98)0.91***(10.75)0.68***(6.88)Driving1.03***(7.74)1.29***(9.21)0.03***(6.62)0.03***(7.89)23.21***(7.51)17.78***(5.25)Broadband0.50***(4.58)0.34***(3.02)0.01***(3.86)0.01***(2.74)2.39(1.04)9.49***(3.99)Severe housing problems-0.09(-0.37)-0.18(-0.63)-0.004(-0.54)-0.01(-1.07)-21.31***(-4.22)-13.90**(-2.47)Food4.70(13.25)4.62***(12.64)0.17***(15.82)0.17***(15.38)63.55***(8.02)55.61***(7.33)Pollution0.02(5.40)0.02***(3.79)0.0005***(4.02)0.0003***(2.59)-0.49***(-5.50)-0.38***(-4.06)Number of observations244919412449194121581790R20.99730.99740.99740.99750.93050.9338*Significant at $10\%$ **Significant at $5\%$ ***Significant at $1\%$ The first conclusion that can be drawn from the 18 estimates is that the model is robust since there are hardly any significant changes in either the estimated regressors or their significance. Likewise, the quality of the fit is good since the R2 ranges between 0.906 and 0.997.
As for the values obtained, in most cases they are those expected a priori. Starting with the variables measuring inequality, the first measure used is the real mean income of the county in relation to that of the state in which it is located. The parameter obtained is highly significant in almost all the estimates made, although the sign changes depending on the measure of mental health. Thus, it is observed that those counties that are richer in relative terms are the ones that suffer more days of poor mental health. However, when these mental health problems become more severe (“frequent mental distress”), the incidence of this variable decreases in value and significance. In the extreme case, i.e., suicide, the sign changes and it is the poorest counties that suffer the most from this problem. This changing result is consistent with what is happening in the economic literature. As Ridley et al. [ 2020] and Shah et al. [ 2021] point out, the results obtained in published work on the subject do not allow inferences to be drawn about causality between income and mental health, which hampers opportunities to inform public policy. Thus, for example, while Gresenz et al. [ 2001] find a strong correlation between individual income level and mental health, although not at the State level, Araya et al. [ 2003] find no association between income and prevalence of common mental disorders. In any case, the results obtained in our work agree with those obtained by Thomson et al. [ 2022], who demonstrate the existence of a deterioration in mental health as a result of lower income. It should be taken into account that in richer and therefore more economically dynamic counties, competition is greater and this can lead to some mental distress (Colantone et al., 2019).
In order to study in depth, the effects of inequality on mental health, six measures of inequality have been used for each county, with the aim of analyzing not only inequality in income distribution, commonly used in the literature, but also gender inequality and racial inequality. This allows us to study the effects of inequality, understood in a global way, on the dependent variables, and to test whether greater inequality within the county has effects on mental health. According to the results obtained, the positive (with one exception) and significant sign in 11 of the 18 estimates allows us to conclude that inequality is a determinant of mental health. The greater the inequality within the county, the worse the mental health. In this sense, precarious working conditions, the stress this entails and the comorbidities associated with poverty that may characterize these counties with higher inequality could explain these results (Llosa et al., 2018; Rönnblad et al., 2019). However, analyzing the inequality measures used one by one, these results can be nuanced. The first thing to note is that racial inequality is the most significant (Table 4). In all estimates, the estimated sign is significant. Moreover, in five of the six estimates made, the sign is positive, so that racial inequality leads to worse mental health, as Wallace et al. [ 2016] also show for the case of the United Kingdom. What is striking is the negative sign obtained when we use suicide as the dependent variable and the percentage of the black population below the poverty line as a measure of inequality. This result invites us to affirm that the black population is less prone to suicide in situations of poverty as Early and Akers [1993], Goldsmith et al. [ 2002] and more recently Riddell et al. [ 2018] have already shown. Regarding income inequality (Table 2), although the sign is always positive, it is only significant in 2 of the 6 estimates made. In this sense, the greater the income inequality within the county, the greater the mental problems, although we did not find any significant result for the extreme case of suicide. This inconclusive result is consistent with that reached by Gresenz et al. [ 2001], who find no relationship between inequality in income distribution and mental health, or that of Yu [2018] who does find a relationship between income inequality and mental health for men, but not for women. Something similar occurs with gender inequality since the estimated regressor is only significant in three of the six estimates made, although in this case we can indeed state that gender inequality harms mental health and is even a determinant of suicide. These results confirm the findings of McAllister et al. [ 2018], according to which better mental health is related to lower gender inequality. Mar et al. [ 2022] also find strong evidence for the relationship between gender inequality, mental health, and suicide.
There are other variants of inequality that can also affect mental health. One of them is health. In this regard, two variables have been used. First, we have employed access to mental health care measured through mental health providers. It is important to note that about $30\%$ of the population lives in a county designated as a mental health professional shortage area (HRSA, 2022). However, the results we obtain are not significant, so we cannot draw conclusions about whether more mental health care facilities lead to a reduction in potential mental disorders. Therefore, we use a second variable, “uninsured”, to analyze whether health coverage, or lack thereof, influences mental health. The results obtained are significant, although different for the extreme case of suicide. Thus, in the case of poor mental health or frequent mental disorders, the parameter obtained is negative, which implies that the higher the percentage of the population that is not insured, the fewer the mental health problems. This result, a priori surprising, can be explained by the high health costs in the USA, which can lead the population without health coverage not to see a specialist when suffering from some type of disorder (Carter et al., 2020). However, when the disease worsens and leads individuals to the extreme solution of suicide, the sign changes and becomes positive, with the uninsured population suffering more from the most severe mental problems, as also shown by Johnson and Brookover [2020], and Ong et al. [ 2021].
Another form of inequality that can affect mental health is related to education. In this work we have used the percentage of the population with university studies since in the USA there is great inequality in access to higher education (Jerrim et al., 2015). The results obtained show a negative sign, although only significant in 7 of the 18 estimates made. This would show that higher education promotes better mental health (Jiang et al., 2020), since there is a certain correlation between the level of education attained and a better job, better life habits and a better health status. It is worth noting that when we use suicide as a dependent variable, education is significant in five of the six estimates made, making education a key factor in the fight against very serious mental disorders (Lorant et al., 2018).
As mentioned above, more education implies a higher probability of finding a job and thus covering all those material needs whose lack can lead to a deterioration of mental health. This is why we use the “unemployment” variable, which also reflects the inequalities between counties in terms of labor markets. However, the parameter obtained is only significant when suicide is used as the dependent variable. The positive sign allows us to conclude that a higher unemployment rate implies a higher suicide rate, as also shown by Amiri [2021].
From here, the next group of variables that have been studied refer to the health status and life habits. First, we analyzed whether there was any relationship between physical and mental health status. The results obtained, significant in all cases except for suicide, show that there is a direct relationship between both variables, i.e., poor physical health leads to poor mental health (Ohrnberger et al., 2017; Luo et al., 2020). However, as we have discussed above, when we use suicide as the dependent variable, the parameter ceases to be significant, contrary to what most of the economic literature says (Fairweather et al., 2006; Phillips & Hempstead, 2022; Qin et al., 2022). Even so, as Fiske et al. [ 2008] argue, the relationship between poor health and suicide tends to occur in older population groups. Furthermore, according to Ahmedani et al. [ 2017] it is important to nuance which determinants of physical health status most affect mental health and suicide. These authors find that lack of sleep is a key factor. This is why we included the variable “sleep” in our analysis. The results obtained show a positive and significant relationship between the percentage of people with sleep problems and the three variables used to measure mental health problems. Therefore, we can affirm that insomnia is a risk factor for mental health as also shown by Chattu et al. [ 2018], Sullivan and Ordiah [2018], and Merikanto and Partonen [2021] among others.
Other variables that reflect the lifestyle of the population have been included in this analysis. Thus, the estimated parameter for the variable “smoke” is always positive and highly significant, whereby the higher the percentage of smokers the worse the mental health, as also argued by Ferreira et al. [ 2019]. There is a common perception that smoking generally helps people to manage stress and can be a form of “self-medication” in people with mental health problems, although this addiction, like others, generates withdrawal symptoms that worsen mental health (Taylor et al., 2021). However, when we estimate the relationship between excessive alcohol consumption and mental health, the result obtained is surprising. The sign is negative and significant in all estimates except when we use suicide as the dependent variable. In this case, the estimated parameter is not significant. Therefore, we cannot affirm that excessive alcohol consumption is a risk factor for mental health. In this regard, Li et al. [ 2022] also conclude that, for certain population groups, alcohol consumption is a protective factor against mental disorders. On the other hand, the economic literature has also analyzed the relationship between obesity and physical activity on mental health (Avila et al., 2015). The results obtained in our work show a negative and significant sign for these two variables. Therefore, we cannot affirm that those counties with a higher percentage of obese people and those who report not doing any physical activity have greater mental problems. In this sense, Biddle et al. [ 2019] also do not obtain evidence of a causal association between physical activity and mental health. To our knowledge, obesity and lack of physical activity would not be a cause but a consequence of mental disorders, as Van der Valk et al. [ 2018] also demonstrate. In fact, Rajan and Menon [2017] point to a bidirectional relationship between obesity and mental health. Finally, the variable “STI” was used as a proxy for sexual activity. The results obtained, significant in 12 of the 18 estimates made, show that sexual activity reduces mental disorders (Brody, 2010; Mollaioli et al., 2021; Gianotten, 2021).
The economic literature has discussed in depth the relationship between social isolation and mental disorders (Wang et al., 2017). This paper aims to delve into the relationship between the two concepts with the use of five variables. Thus, first, we employ the variable measuring population density to test whether denser counties have lower mental problems. Epidemiological studies show that the risk of serious mental illness is higher in cities than in rural areas, where population density is lower (Gruebner et al., 2017) since higher density is associated with lower social contacts (Giacco et al., 2022). However, the negative and significant sign in 11 of the 18 estimates made allow us to affirm that in those counties with higher population per km2 makes personal relationships closer and fosters better mental health. This result, contrary to that shown by other authors, perhaps requires a more specific analysis of this relationship, as Lai et al. [ 2021] have done. These authors conclude that rather than population density per se, it is urban design that determines the relationship between density and mental health. In fact, the evidence of the impacts of increasing urban densification on loneliness and social isolation in humans is still inconclusive. For this reason, we use other variables that reflect social isolation. Thus, for the case of the variable “associations”, we did not find a significant relationship with respect to mental health, making it unclear whether these types of social associations improve mental health (Wakefield et al., 2019). Regarding teleworking, a variable that reflects the lack of social contact in the workplace, the positive and highly significant sign obtained in all the estimates made shows that, indeed, a higher percentage of teleworkers and, therefore, greater social isolation, leads to greater mental disorders and even the extreme case of suicide. Authors such as Mann and Holdsworth [2003] and De Sio et al. [ 2021] have already demonstrated the harmful effects of teleworking on mental health. This result is confirmed when we use the variable “driving” which reflects the percentage of people who drive alone every day to work. Again, this variable is used as a proxy for social isolation, and the results obtained (positive and highly significant sign in all the estimates made) allow us to conclude that social isolation is a key risk factor for mental disorders. Finally, Internet access was used as a proxy variable for Internet use at home and, therefore, less social contact. The positive and highly significant sign shows that the higher the use of the Internet and social networks, the higher the probability of suffering from mental disorders. These results agree with those obtained by Grant et al. [ 2019], Arzani-Birgani et al. [ 2021], and Golin [2022].
Other problems that can cause mental distress are material problems in a household and food insecurity. Now, which of the two generates more distress? *Our analysis* reveals that food insecurity is a risk factor for mental health, whereas we did not find a clear result for the variable reflecting severe household problems. Probably, as suggested by Singh et al. [ 2019], it would be necessary to analyze each of the household problems to see how they individually impact mental health. The harmful effects of food insecurity on mental health have also been shown by authors such as Pourmotabbed et al. [ 2020] among others.
Finally, the effect of pollution on mental health has been analyzed. The results obtained allow us to affirm that pollution is a risk factor for mental health, although in the extreme case of suicide, the sign changes, so we cannot affirm that those more polluting counties suffer from a higher suicide rate (Heo et al., 2021). As Ventriglio et al. [ 2021] argue, the impact of pollution on public health is well known, but the association between environmental pollutants and mental health has been little analyzed, and most of these yield inconclusive results. In any case, our results confirm the theses of Chen et al. [ 2018] and Yang et al. [ 2021].
## Conclusion
Is inequality a risk factor for mental health? What lifestyle habits worsen mental health the most? What are the effects of the implementation of teleworking on mental disorders? To these and other questions we have tried to answer in this paper. Using a sample of 2,735 U.S. counties, a cross-sectional linear model has been estimated. The results obtained allow us to conclude that income is a key factor determining mental health status. While wealthier counties are more likely to suffer from mild mental disorders, when these worsen to the extreme case of suicide, it is the poorer counties that suffer the most. However, it is not severe housing problems that lead to the extreme deterioration of mental health, but rather the food insecurity suffered by poor families. For this reason, public assistance programs to meet the most basic needs are necessary in this country, since there are many households that suffer as a result of the strong inequality that exists. In fact, inequality, understood in a broad sense, is a key determinant of mental health. For this reason, public policies must be implemented to mitigate income differences, and to fight against gender inequality and all types of racial discrimination. In addition, the health coverage network should continue to be extended to the entire American population, and a scholarship plan should be promoted to allow greater access to higher education. All of this will result in fewer mental health problems.
On the other hand, because of the COVID-19 pandemic, many companies have adapted to increased teleworking. This, although it has served to prevent the virus, is a problem for mental health, as our results show. Thus, we show that the social isolation produced by teleworking, driving alone to work and the increased use of the Internet at home are seriously damaging to mental health. Therefore, a more detailed analysis of the pros and cons of promoting teleworking should be carried out by policy makers and companies.
The third pillar on which we have based our analysis of mental disorders is healthy lifestyle habits. In this regard, we emphasize that lack of sleep and addiction to tobacco are risk factors for mental health, while sexual activity is a good medicine against this type of disorder.
Finally, pollution also harms mental health. Thus, we find yet another argument for policymakers to step up measures to combat climate change.
However, this work is not without limitations. The first limitation is the lack of post-COVID-19 mental health data, which would have allowed us to draw stronger conclusions about the incidence of social isolation and mental health. The data we have been able to work with refer to 2019, so future updates of the data will allow the effect of COVID-19 to be included and will undoubtedly improve the analysis performed. Likewise, it would have been desirable to have more measures of both gender and racial inequality to make the analysis more robust. Thus, for example, it would have been very interesting to have available measures of gender inequality by US counties such as the Global Gender Gap, Gender Inequality Index or the Social Watch Gender Equity Index. Likewise, having a database on multidimensional racial inequality, as calculated by Rohde and Guest [2013] might have solved some of the problems of non-significance that we have encountered.
## References
1. Ahmedani, B. K., Peterson, E. L., Hu, Y., Rossom, R. C., Lynch, F., Lu, C. Y., … Simon, G. E. (2017). Major physical health conditions and risk of suicide. American Journal of Preventive Medicine,53(3), 308–315.
2. Almedom AM. **Social capital and mental health: an interdisciplinary review of primary evidence**. *Social Science & Medicine* (2005.0) **61** 943-964. DOI: 10.1016/j.socscimed.2004.12.025
3. Amate-Fortes I, Guarnido-Rueda A, Molina-Morales A. **World analysis of the determinants of the inequality in health. Is the measurement of Inequality important?**. *Revija za Socijalnu Politiku* (2020.0) **27** 83-98. DOI: 10.3935/rsp.v27i1.1600
4. Amiri S. **Unemployment and suicide mortality, suicide attempts, and suicide ideation: a meta-analysis**. *International Journal of Mental Health* (2021.0). DOI: 10.1080/00207411.2020.1859347
5. Araya R, Lewis G, Rojas G, Fritsch R. **Education and income: which is more important for mental health?**. *Journal of Epidemiology & Community Health* (2003.0) **57** 501-505. DOI: 10.1136/jech.57.7.501
6. Arzani-Birgani A, Zarei J, Favaregh L, Ghanaatiyan E. **Internet addiction, mental health, and sleep quality in students of medical sciences, Iran: a cross-sectional study**. *Journal of Education and Health Promotion* (2021.0). DOI: 10.4103/jehp.jehp_1506_20
7. Avila C, Holloway AC, Hahn MK, Morrison KM, Restivo M, Anglin R, Taylor VH. **An overview of links between obesity and mental health**. *Current Obesity Reports* (2015.0) **4** 303-310. DOI: 10.1007/s13679-015-0164-9
8. Balogun O, Koyanagi A, Stickley A, Gilmour S, Shibuya K. **Alcohol consumption and psychological distress in adolescents: a multi-country study**. *Journal of Adolescent Health* (2014.0) **54** 228-234. DOI: 10.1016/j.jadohealth.2013.07.034
9. Bassett, E., & Moore, S. (2013). Mental health and social capital: Social capital as a promising initiative to improving the mental health of communities. In A.J. Rodriguez-Morales (Ed.), Current Topics in Public Health (Ch. 28) InTech.
10. Biddle SJ, Ciaccioni S, Thomas G, Vergeer I. **Physical activity and mental health in children and adolescents: an updated review of reviews and an analysis of causality**. *Psychology of Sport and Exercise* (2019.0) **42** 146-155. DOI: 10.1016/j.psychsport.2018.08.011
11. Burns JK, Tomita A, Lund C. **Income inequality widens the existing income-related disparity in depression risk in post-apartheid South Africa: Evidence from a nationally representative panel study**. *Health & place* (2017.0) **45** 10-16. DOI: 10.1016/j.healthplace.2017.02.005
12. Brody S. **The relative health benefits of different sexual activities**. *The Journal of Sexual Medicine* (2010.0) **7** 1336-1361. DOI: 10.1111/j.1743-6109.2009.01677.x
13. Brown TN, Williams DR, Jackson JS, Neighbors HW, Torres M, Sellers SL, Brown KT. **Being black and feeling blue”: the mental health consequences of racial discrimination**. *Race and Society* (2000.0) **2** 117-131. DOI: 10.1016/S1090-9524(00)00010-3
14. Cameron AJ, Magliano DJ, Dunstan DW, Zimmet PZ, Hesketh K, Peeters A, Shaw JE. **A bi-directional relationship between obesity and health-related quality of life: evidence from the longitudinal AusDiab study**. *International Journal of Obesity* (2012.0) **36** 295-303. DOI: 10.1038/ijo.2011.103
15. Carter SP, Cowan T, Snow A, Cerel J, Tucker R. **Health insurance and mental health care utilization among adults who identify as transgender and gender diverse**. *Psychiatric Services* (2020.0) **71** 151-157. DOI: 10.1176/appi.ps.201900289
16. Casella G. **Leverage and regression through the origin**. *The American Statistician* (1983.0) **37** 147-152. DOI: 10.1080/00031305.1983.10482728
17. **Sociodemographic differences in binge drinking among adults–14 states, 2004**. *MMWR: Morbidity and Mortality Weekly Report* (2009.0) **58** 301-304. PMID: 19343009
18. Chattu, V. K., Manzar, M. D., Kumary, S., Burman, D., Spence, D. W., & Pandi-Perumal, S. R. (2018, December). The global problem of insufficient sleep and its serious public health implications. In Healthcare (Vol.7, No. 1, p.1). MDPI.
19. Chen, S., Oliva, P., & Zhang, P. (2018). Air pollution and mental health: evidence from China. NBER Working Paper w24686.
20. Christiansen, J., Qualter, P., Friis, K., Pedersen, S. S., Lund, R., Andersen, C. M., … Lasgaard, M. (2021). Associations of loneliness and social isolation with physical and mental health among adolescents and young adults. Perspectives in Public Health,141(4), 226–236.
21. Colantone I, Crino R, Ogliari L. **Globalization and mental distress**. *Journal of International Economics* (2019.0) **119** 181-207. DOI: 10.1016/j.jinteco.2019.04.008
22. Courtin E, Knapp M. *Health and wellbeing consequences of social isolation and loneliness in old age* (2015.0)
23. Dalghren G, y Whitehead M. *Policies and strategies to promote social equity in health* (1991.0)
24. De Hert, M., Correll, C. U., Bobes, J., Cetkovich-Bakmas, M., Cohen, D. A. N., Asai, I., … Leucht, S. (2011). Physical illness in patients with severe mental disorders. I. Prevalence, impact of medications and disparities in health care. World Psychiatry,10(1), 52.
25. De Sio, S., Cedrone, F., Nieto, H. A., Lapteva, E., Perri, R., Greco, E., … Buomprisco, G. (2021). Telework and its effects on mental health during the COVID-19 lockdown. European Review for Medical And Pharmacological Sciences,25(10), 3914–3922.
26. Early KE, Akers RL. **“It’s a white thing”: an exploration of beliefs about suicide in the african-american community**. *Deviant Behavior* (1993.0) **14** 277-296. DOI: 10.1080/01639625.1993.9967947
27. Eisenhauer JG. **Regression through the origin**. *Teaching Statistics* (2003.0) **25** 76-80. DOI: 10.1111/1467-9639.00136
28. Fairweather AK, Anstey KJ, Rodgers B, Butterworth P. **Factors distinguishing suicide attempters from suicide ideators in a community sample: social issues and physical health problems**. *Psychological Medicine* (2006.0) **36** 1235-1245. DOI: 10.1017/S0033291706007823
29. Fava M, Kendler KS. **Major depressive disorder**. *Neuron* (2000.0) **28** 335-341. DOI: 10.1016/S0896-6273(00)00112-4
30. Ferreira VR, Jardim TV, Sousa ALL, Rosa BMC, Jardim PCV. **Smoking, alcohol consumption and mental health: data from the Brazilian study of cardiovascular risks in adolescents (ERICA)**. *Addictive Behaviors Reports* (2019.0) **9** 100147. DOI: 10.1016/j.abrep.2018.100147
31. Fiske A, O’Riley AA, Widoe RK. **Physical health and suicide in late life: an evaluative review**. *Clinical Gerontologist* (2008.0) **31** 31-50. DOI: 10.1080/07317110801947151
32. Gee GC, Ford CL. **Structural racism and health inequities: old issues, New Directions1**. *Du Bois Review: Social Science Research on Race* (2011.0) **8** 115-132. DOI: 10.1017/S1742058X11000130
33. Giacco D, Kirkbride JB, Ermakova AO, Webber M, Xanthopoulou P, Priebe S. **Neighbourhood characteristics and social isolation of people with psychosis: a multi-site cross-sectional study**. *Social Psychiatry and Psychiatric Epidemiology* (2022.0) **57** 1907-1915. DOI: 10.1007/s00127-021-02190-x
34. Gianotten, W. L. (2021). The (mental) health benefits of sexual expresión. In M. Lew-Starowicz, A. Giraldi, T. H. C. Krüger (Eds.), Psychiatry and sexual medicine: A comprehensive guide for clinical practitioners. Springer International Publishing. 10.1007/978-3-030-52298-8
35. Goldsmith SK, Pellmar TC, Kleinman AM, Bunney WE. *Reducing suicide: a national imperative* (2002.0)
36. Golin M. **The effect of broadband internet on the gender gap in mental health: Evidence from Germany**. *Health Economics* (2022.0) **31** 6-21. DOI: 10.1002/hec.4570
37. Grant JE, Lust K, Chamberlain SR. **Problematic smartphone use associated with greater alcohol consumption, mental health issues, poorer academic performance, and impulsivity**. *Journal of Behavioral Addictions* (2019.0) **8** 335-342. DOI: 10.1556/2006.8.2019.32
38. Gresenz CR, Sturm R, Tang L. **Income and mental health: unraveling community and individual level relationships**. *Journal of Mental Health Policy and Economics* (2001.0) **4** 197-204. PMID: 12119429
39. Gruebner O, Rapp MA, Adli M, Kluge U, Galea S, Heinz A. **Cities and mental health**. *Deutsches Ärzteblatt International* (2017.0) **114** 121. PMID: 28302261
40. Hagen EH, Rosenström T. **Explaining the sex difference in depression with a unified bargaining model of anger and depression**. *Evolution Medicine and Public Health* (2016.0) **2016** 117-132. PMID: 26884416
41. Heo S, Lee W, Bell ML. **Suicide and associations with air pollution and ambient temperature: a systematic review and meta-analysis**. *International Journal of Environmental Research and Public Health* (2021.0) **18** 7699. DOI: 10.3390/ijerph18147699
42. Hopcroft RL, Bradley DB. **The sex difference in depression across 29 countries**. *Social Forces* (2007.0) **85** 1483-1507. DOI: 10.1353/sof.2007.0071
43. HRSA (Health Resources and Services Administration) (2022). Designated health professional shortage areas statistics. US Department of Health & Human Services. Available at https://data.hrsa.gov. Accessed 29 July 2022.
44. Jerrim J, Chmielewski AK, Parker P. **Socioeconomic inequality in access to high-status colleges: a cross-country comparison**. *Research in Social Stratification and Mobility* (2015.0) **42** 20-32. DOI: 10.1016/j.rssm.2015.06.003
45. Jiang W, Lu Y, Xie H. **Education and mental health: evidence and mechanisms**. *Journal of Economic Behavior & Organization* (2020.0) **180** 407-437. DOI: 10.1016/j.jebo.2020.09.032
46. Johnson KF, Brookover DL. **Counselors’ role in decreasing suicide in mental health professional shortage areas in the United States**. *Journal of Mental Health Counseling* (2020.0) **42** 170-186. DOI: 10.17744/mehc.42.2.06
47. Kawachi I, Kennedy BP, Gupta V, Prothrow-Stith D. **Women’s status and the health of women and men: a view from the States**. *Social Science & Medicine* (1999.0) **48** 21-32. DOI: 10.1016/S0277-9536(98)00286-X
48. Kivimäki, M., Batty, G. D., Singh-Manoux, A., Nabi, H., Sabia, S., Tabak, A. G., … Jokela, M. (2009). Association between common mental disorder and obesity over the adult life course. The British Journal of Psychiatry,195(2), 149–155.
49. Kramer MR, Hogue CR. **Is segregation bad for your health?**. *Epidemiologic Reviews* (2009.0) **31** 178-194. DOI: 10.1093/epirev/mxp001
50. Lai, K. Y., Sarkar, C., Kumari, S., Ni, M. Y., Gallacher, J., & Webster, C. (2021). Calculating a national anomie density ratio: Measuring the patterns of loneliness and social isolation across the UK’s residential density gradient using results from the UK Biobank study. Landscape and Urban Planning, 215, 104194.
51. Lee KH, Dvorak RG, Schuett MA, van Riper CJ. **Understanding spatial variation of physical inactivity across the continental United States**. *Landscape and Urban Planning* (2017.0) **168** 61-71. DOI: 10.1016/j.landurbplan.2017.09.020
52. Lewis TT, Cogburn CD, Williams DR. **Self-reported experiences of discrimination and health: scientific advances, ongoing controversies, and emerging issues**. *Annual Review of Clinical Psychology* (2015.0) **11** 407. DOI: 10.1146/annurev-clinpsy-032814-112728
53. Li, Y., Zhang, C., Ding, S., Li, J., Li, L., Kang, Y., … Duan, Y. (2022). Physical activity, smoking, alcohol consumption and depressive symptoms among young, early mature and late mature people: a cross-sectional study of 76,223 in China. Journal of Affective Disorders,299, 60–66.
54. Llosa JA, Menéndez-Espina S, Agulló-Tomás E, Rodríguez-Suárez J. **Job insecurity and mental health: a meta-analytical review of the consequences of precarious work in clinical disorders**. *Anales de Psicología* (2018.0) **34** 211-221. DOI: 10.6018/analesps.34.2.281651
55. Lorant, V., de Gelder, R., Kapadia, D., Borrell, C., Kalediene, R., Kovács, K., … Mackenbach, J. P. (2018). Socioeconomic inequalities in suicide in Europe: the widening gap. The British Journal of Psychiatry,212(6), 356–361.
56. Luo MS, Chui EWT, Li LW. **The longitudinal associations between physical health and mental health among older adults**. *Aging & Mental Health* (2020.0) **24** 1990-1998. DOI: 10.1080/13607863.2019.1655706
57. Mann S, Holdsworth L. **The psychological impact of teleworking: stress, emotions and health**. *New Technology Work and Employment* (2003.0) **18** 196-211. DOI: 10.1111/1468-005X.00121
58. Mar, J., Larrañaga, I., Ibarrondo, O., González-Pinto, A., las, Hayas, C., Fullaondo, A., …. UPRIGHT Consortium. (2022). Socioeconomic and gender inequalities in mental disorders among adolescents and young adults. Revista de Psiquiatría y Salud Mental. 10.1016/j.rpsm.2022.07.001
59. Mason WA, Kosterman R, Haggerty KP, Hawkins JD, Redmond C, Spoth RL, Shin C. **Dimensions of adolescent alcohol involvement as predictors of young-adult major depression**. *Journal of Studies on Alcohol and Drugs* (2008.0) **69** 275-285. DOI: 10.15288/jsad.2008.69.275
60. Matthew P, Brodersen DM. **Income inequality and health outcomes in the United States: an empirical analysis**. *The Social Science Journal* (2018.0) **55** 432-442. DOI: 10.1016/j.soscij.2018.05.001
61. McAllister A, Fritzell S, Almroth M, Harber-Aschan L, Larsson S, Burström B. **How do macro-level structural determinants affect inequalities in mental health?–a systematic review of the literature**. *International Journal for Equity in Health* (2018.0) **17** 1-14. DOI: 10.1186/s12939-018-0879-9
62. Merikanto I, Partonen T. **Eveningness increases risks for depressive and anxiety symptoms and hospital treatments mediated by insufficient sleep in a population-based study of 18,039 adults**. *Depression and Anxiety* (2021.0) **38** 1066-1077. DOI: 10.1002/da.23189
63. Milner A, Kavanagh A, Scovelle AJ, O’Neil A, Kalb G, Hewitt B, King TL. **Gender equality and health in high-income countries: a systematic review of within-country indicators of gender equality in relation to health outcomes**. *Women’s Health Reports* (2021.0) **2** 113-123. DOI: 10.1089/whr.2020.0114
64. Milojevich, H. M., & Lukowski, A. F. (2016). Sleep and mental health in undergraduate students with generally healthy sleep habits. PLoS One, 11(6), e0156372.
65. Mollaioli D, Sansone A, Ciocca G, Limoncin E, Colonnello E, Di Lorenzo G, Jannini EA. **Benefits of sexual activity on psychological, relational, and sexual health during the COVID-19 breakout**. *The Journal of Sexual Medicine* (2021.0) **18** 35-49. DOI: 10.1016/j.jsxm.2020.10.008
66. Mouzon DM, Taylor RJ, Keith VM, Nicklett EJ, Chatters LM. **Discrimination and psychiatric disorders among older African Americans**. *International Journal of Geriatric Psychiatry* (2017.0) **32** 175-182. DOI: 10.1002/gps.4454
67. Nicholson WM. **Emotional factors in obesity**. *The American Journal of the Medical Sciences* (1946.0) **211** 443-447. DOI: 10.1097/00000441-194604000-00007
68. Ohrnberger J, Fichera E, Sutton M. **The relationship between physical and mental health: a mediation analysis**. *Social Science & Medicine* (2017.0) **195** 42-49. DOI: 10.1016/j.socscimed.2017.11.008
69. Ong, M. S., Lakoma, M., Gees Bhosrekar, S., Hickok, J., McLean, L., Murphy, M., … Ross-Degnan, D. (2021). Risk factors for suicide attempt in children, adolescents, and young adults hospitalized for mental health disorders. Child and Adolescent Mental Health,26(2), 134–142.
70. Patel V, Burns JK, Dhingra M, Tarver L, Kohrt BA, Lund C. **Income inequality and depression: a systematic review and meta-analysis of the association and a scoping review of mechanisms**. *World Psychiatry* (2018.0) **17** 76-89. DOI: 10.1002/wps.20492
71. Phillips JA, Hempstead K. **The role of context in shaping the relationship between physical health and suicide over the life course**. *SSM-Population Health* (2022.0) **17** 101059. DOI: 10.1016/j.ssmph.2022.101059
72. Pickett KE, Wilkinson RG. **Income inequality and health: a causal review**. *Social Science & Medicine* (2015.0) **128** 316-326. DOI: 10.1016/j.socscimed.2014.12.031
73. Pourmotabbed, A., Moradi, S., Babaei, A., Ghavami, A., Mohammadi, H., Jalili, C., … Miraghajani, M. (2020). Food insecurity and mental health: a systematic review and meta-analysis. Public Health Nutrition,23(10), 1778–1790.
74. Qin P, Syeda S, Canetto SS, Arya V, Liu B, Menon V, Gunnell D. **Midlife suicide: A systematic review and meta-analysis of socioeconomic, psychiatric and physical health risk factors**. *Journal of Psychiatric Research* (2022.0) **154** 233-241. DOI: 10.1016/j.jpsychires.2022.07.037
75. Rajan TM, Menon V. **Psychiatric disorders and obesity: a review of association studies**. *Journal of Postgraduate Medicine* (2017.0) **63** 182. DOI: 10.4103/jpgm.JPGM_712_16
76. Reid GJ, Hong RY, Wade TJ. **The relation between common sleep problems and emotional and behavioral problems among 2-and 3‐year‐olds in the context of known risk factors for psychopathology**. *Journal of Sleep Research* (2009.0) **18** 49-59. DOI: 10.1111/j.1365-2869.2008.00692.x
77. Riddell CA, Harper S, Cerdá M, Kaufman JS. **Comparison of rates of firearm and nonfirearm homicide and suicide in black and white non-hispanic men, by US state**. *Annals of Internal Medicine* (2018.0) **168** 712-720. DOI: 10.7326/M17-2976
78. Ridley M, Rao G, Schilbach F, Patel V. **Poverty, depression, and anxiety: causal evidence and mechanisms**. *Science* (2020.0) **370** eaay0214. DOI: 10.1126/science.aay0214
79. Rohde N, Guest R. **Multidimensional racial inequality in the United States**. *Social Indicators Research* (2013.0) **114** 591-605. DOI: 10.1007/s11205-012-0163-0
80. Rohde N, D’Ambrosio C, Tang KK, Rao P. **Estimating the mental health effects of social isolation**. *Applied Research in Quality of Life* (2016.0) **11** 853-869. DOI: 10.1007/s11482-015-9401-3
81. Rönnblad, T., Grönholm, E., Jonsson, J., Koranyi, I., Orellana, C., Kreshpaj, B., … Bodin, T. (2019). Precarious employment and mental health. Scandinavian Journal of Work Environment & Health,45(5), 429–443.
82. Sanders, A. R., Detera-Wadleigh, S. D., & Gershon, E. S. (1999). Molecular genetics of mood disorders. In D. S. Charney, E. J. Nestler, & B. S. Bunney (Eds.), Neurobiology of Mental Illness (299–316). Oxford.
83. Seedat, S., Scott, K. M., Angermeyer, M. C., Berglund, P., Bromet, E. J., Brugha, T. S., … Kessler, R. C. (2009). Cross-national associations between gender and mental disorders in the World Health Organization World Mental Health surveys. Archives of General Psychiatry,66(7), 785–795.
84. Shah, N., Walker, I. F., Naik, Y., Rajan, S., O’Hagan, K., Black, M., … Stansfield, J. (2021). National or population level interventions addressing the social determinants of mental health–an umbrella review. BMC Public Health,21(1), 1–14.
85. Singh A, Daniel L, Baker E, Bentley R. **Housing disadvantage and poor mental health: a systematic review**. *American Journal of Preventive Medicine* (2019.0) **57** 262-272. DOI: 10.1016/j.amepre.2019.03.018
86. Skogen, J. C., Sivertsen, B., Lundervold, A. J., Stormark, K. M., Jakobsen, R., & Hysing, M. (2014). Alcohol and drug use among adolescents: and the co-occurrence of mental health problems. Ung@ hordaland, a population-based study. BMJ Open, 4(9), e005357.
87. Sullivan PF, Neale MC, Kendler KS. **Genetic epidemiology of major depression: review and meta-analysis**. *American Journal of Psychiatry* (2000.0) **157** 1552-1562. DOI: 10.1176/appi.ajp.157.10.1552
88. Sullivan K, Ordiah C. **Association of mildly insufficient sleep with symptoms of anxiety and depression**. *Neurology Psychiatry and Brain Research* (2018.0) **30** 1-4. DOI: 10.1016/j.npbr.2018.03.001
89. Taylor DJ, Gardner CE, Bramoweth AD, Williams JM, Roane BM, Grieser EA, Tatum JI. **Insomnia and mental health in college students**. *Behavioral Sleep Medicine* (2011.0) **9** 107-116. DOI: 10.1080/15402002.2011.557992
90. Taylor, G. M., Lindson, N., Farley, A., Leinberger-Jabari, A., Sawyer, K., te Water Naudé, R., … Aveyard, P. (2021). Smoking cessation for improving mental health. Cochrane Database of Systematic Reviews, (3), CDO13522. 10.1002/14651858.CD013522.pub2
91. Thomson, R. M., Igelström, E., Purba, A. K., Shimonovich, M., Thomson, H., McCartney, G., … Katikireddi, S. V. (2022). How do income changes impact on mental health and wellbeing for working-age adults? A systematic review and meta-analysis. The Lancet Public Health,7(6), e515–e528.
92. Van der Valk ES, Savas M, van Rossum EF. **Stress and obesity: are there more susceptible individuals?**. *Current Obesity Reports* (2018.0) **7** 193-203. DOI: 10.1007/s13679-018-0306-y
93. Van de Velde S, Huijts T, Bracke P, Bambra C. **Macro-level gender equality and depression in men and women in Europe**. *Sociology of Health & Illness* (2013.0) **35** 682-698. DOI: 10.1111/j.1467-9566.2012.01521.x
94. Ventriglio A, Bellomo A, di Gioia I, Di Sabatino D, Favale D, De Berardis D, Cianconi P. **Environmental pollution and mental health: a narrative review of literature**. *CNS Spectrums* (2021.0) **26** 51-61. DOI: 10.1017/S1092852920001303
95. Wakefield, J. R., Bowe, M., Kellezi, B., McNamara, N., & Stevenson, C. (2019). When groups help and when groups harm: Origins, developments, and future directions of the “Social Cure” perspective of group dynamics. Social and Personality Psychology Compass, 13(3), e12440.
96. Wallace S, Nazroo J, Bécares L. **Cumulative effect of racial discrimination on the mental health of ethnic minorities in the United Kingdom**. *American Journal of Public Health* (2016.0) **106** 1294-1300. DOI: 10.2105/AJPH.2016.303121
97. Wang J, Lloyd-Evans B, Giacco D, Forsyth R, Nebo C, Mann F, Johnson S. **Social isolation in mental health: a conceptual and methodological review**. *Social Psychiatry and Psychiatric Epidemiology* (2017.0) **52** 1451-1461. DOI: 10.1007/s00127-017-1446-1
98. Williams DR. **Stress and the mental health of populations of color: advancing our understanding of race-related stressors**. *Journal of Health and Social Behavior* (2018.0) **59** 466-485. DOI: 10.1177/0022146518814251
99. Windle, K., Francis, J., & Coomber, C. (2011). Preventing loneliness and social isolation: interventions and outcomes (pp. 1–16). Social Care Institute for Excellence. Available from: http://www.scie.org.uk/publications/briefings/files/briefing39.pdf. Accessed 5 Aug 2022.
100. Yang Z, Song Q, Li J, Zhang Y, Yuan XC, Wang W, Yu Q. **Air pollution and mental health: the moderator effect of health behaviors**. *Environmental Research Letters* (2021.0) **16** 044005. DOI: 10.1088/1748-9326/abe88f
101. Yu S. **Uncovering the hidden impacts of inequality on mental health: a global study**. *Translational Psychiatry* (2018.0) **8** 1-10. DOI: 10.1038/s41398-018-0148-0
|
---
title: 'Northern Ireland Cohort for the Longitudinal Study of Ageing (NICOLA): health
assessment protocol, participant profile and patterns of participation'
authors:
- Charlotte E Neville
- Ian S Young
- Frank Kee
- Ruth E Hogg
- Angela Scott
- Frances Burns
- Jayne V Woodside
- Bernadette McGuinness
journal: BMC Public Health
year: 2023
pmcid: PMC9999338
doi: 10.1186/s12889-023-15355-x
license: CC BY 4.0
---
# Northern Ireland Cohort for the Longitudinal Study of Ageing (NICOLA): health assessment protocol, participant profile and patterns of participation
## Abstract
### Background
The Northern Ireland Cohort for the Longitudinal Study of Ageing (NICOLA) is a prospective, longitudinal study of a representative cohort of older adults living in Northern Ireland, United Kingdom. Its aim is to explore the social, behavioural, economic and biological factors of ageing and how these factors change as people age. The study has been designed to maximize comparability with other international studies of ageing thereby facilitating cross-country comparisons. This paper provides an overview of the design and methodology of the health assessment which was carried out as part of Wave 1.
### Methods
Three thousand, six hundred and fifty five community dwelling adults, aged 50 years and over participated in the health assessment as part of Wave 1 of NICOLA. The health assessment included a battery of measurements across various domains that addressed key indicators of ageing namely: physical function, vision and hearing, cognitive function, and cardiovascular health. This manuscript describes the scientific rationale for the choice of assessments, provides an overview of the core objective measures carried out in the health assessment and describes the differences in characteristics of participants who took part in the health assessment compared to those who did not take part.
### Results
The manuscript highlights the importance of incorporating objective measures of health in population based studies as a means of complementing subjective measures and as a way to advance our understanding of the ageing process. The findings contextualize NICOLA as a data resource within Dementias Platform UK (DPUK), the Gateway to Global Ageing (G2G) and other existing networks of population based longitudinal studies of ageing.
### Conclusion
This manuscript can help inform design considerations for other population based studies of ageing and facilitate cross-country comparative analysis of key life-course factors affecting healthy ageing such as educational attainment, diet, the accumulation of chronic conditions (including Alzheimer’s disease, dementia and cardiovascular disease) as well as welfare and retirement policies.
## Background to NICOLA
The Northern Ireland Cohort for the Longitudinal Study of Ageing (NICOLA) [1] is a population-based, nationally representative ageing cohort study of adults aged 50 years and over and living in Northern Ireland [2, 3]. It is the first large-scale, in-depth, longitudinal study of ageing in Northern Ireland, and is one of a family of similar ageing studies across the globe aiming to gain a better understanding of factors affecting health and social outcomes in our rapidly increasing ageing populations. Sampling procedures and study design have been described fully elsewhere [3]. In brief, 8283 community dwelling adults, aged 50 years or older and living in private households across Northern Ireland were recruited from a randomised, stratified sample of Northern Ireland addresses obtained from the Business Service Organization General Practitioner Register and stratified by geographical location and postcode, thus ensuring a representative sample. Those who were institutionalized or who lacked capacity to provide informed consent were not eligible to participate. Wave 1 of the study had three components: a computer assisted personal interview (CAPI) conducted in the participant’s home, a pen and paper self-completion questionnaire (SCQ) including a dietary questionnaire and a health assessment. The Wave 1 CAPI interviews, conducted between December 2013 and July 2016, included questions on pensions, employment, living standards, health aspects including service needs and usage, as well as social contact and formal and informal care. The self-completion questionnaire included questions on relationship quality, loneliness, stressful and traumatic life events, worry and alcohol intake. The current paper focuses on the protocol used within the Wave 1 health assessment which took place between January 2014 and August 2018. Although spouses aged under 50 years were interviewed as part of the CAPI mainly to provide couple- or household-level data, they are not included in the current analysis. A follow-up (Wave 2) of the cohort took place between May 2017 and March 2022 and included a CAPI, SCQ and a COVID-19 questionnaire. A health assessment was not conducted in Wave 2. The results from Wave 2 will be presented in a separate manuscript.
## Scientific Rationale/Importance of NICOLA health assessment
International comparability was a key consideration in the design of NICOLA in order to ensure adoption of best practice and allow cross-national comparisons of results. As such, many of the methods employed in NICOLA are closely harmonized with those of other longitudinal studies of ageing, including The Irish Longitudinal Study of Ageing (TILDA) [4], the English Longitudinal Study of Ageing (ELSA) [5], the EPIC Norfolk Eye study [6], the UK Biobank Eye Study [7] and the US-based Health and Retirement Study (HRS) [8]. Further to this, the data and/or meta-data from the health assessment (along with the CAPI and SCQ data) is deposited in various data repositories including the Dementias Platform UK (DPUK), UK Data Service (UKDS), UK Longitudinal Linkage Collaboration (UK LLC) and Gateway to Global Ageing (G2G) which hold similar data from other cohort studies; this helps to maximise data sharing and foster research collaborations. Such data repositories help facilitate pooled analyses of core topics and allow comparability of the NICOLA cohort to other studies of ageing. More locally, NICOLA provides a strong, ongoing evidence base which will be used to inform local policy by helping researchers and policy makers understand better the social, health and demographic challenges of our ageing society.
The health assessment conducted as part of NICOLA was a fundamental component of the study, specifically designed to complement the self-reported data obtained in the CAPI, providing a range of objective measures of the health of the older population of Northern Ireland. The information typically obtained from objective measures of health and function can often be very different to information obtained from subjective self-reports. Integrating both objective and subjective measures therefore enables us to validate the self-reported data, identify previously undiagnosed illnesses such as hypertension or diabetes and act as useful indicators of the early signs of decline in health or physical function prior to symptomatic disease.
## Research direction on health assessment content
The content of the health assessment crossed a range of clinical domains, also drawing on the expertise of a wide range of research disciplines. Priority was given to health domains that were known to be of most relevance to the ageing process and which could reliably be measured in a population based study such as NICOLA. A unique methodological feature of the NICOLA health assessment was the detailed assessment of ophthalmic health.
## Objective of this overview
The purpose of this paper is to present an overview of the design and methodology of the NICOLA Wave 1 health assessment. The findings presented encapsulate the core objective measures of health and wellbeing of older adults who took part in the NICOLA Wave 1 health assessment. The information presented can inform design considerations for other population based studies of ageing and overall will add to the global body of evidence regarding harmonization of health measurements in older adults.
## Design of health assessment protocol
All NICOLA participants who completed the baseline home CAPI as detailed previously [3] were sent a letter inviting them to attend a health assessment. Based on the relatively small geographical area of Northern Ireland (14,130 km2),the population distribution and accessible transport network, the Wellcome Trust-Wolfson Clinical Research Facility (CRF) located at the Belfast City Hospital was deemed a suitable location to perform the health assessments. The duration of the CRF-based health assessment was approximately two to three hours. All assessments were undertaken by research nurses and research assistants who received comprehensive training in the methodologies and provided clear step-by-step instructions to all participants. To encourage participation, travelling expenses to and from the CRF were provided to participants. A more condensed nurse-led home assessment lasting approximately 2 h was offered to respondents who were unable or unwilling to attend the CRF. Participants were phoned prior to the nurse attending their home.
## Health assessment methods
A robust battery of standardised assessments of cardiovascular function, respiratory function, physical function including hand grip strength, balance, walk speed, visual health, hearing and cognitive health were used, all of which are comparable to those used in other longitudinal studies internationally. Other standard clinical measures including blood pressure, height, weight, and hip and waist circumference were also collected. Non-fasting blood and urine samples were also obtained as part of the health assessment. If glucose or lipid results were outside the normal expected range, then both the participant and participant’s General Practitioner were informed in writing. The assessment methods and their rationale for inclusion in the health assessment are detailed below.
Table 1 provides an overview of the physical, cognitive health, mental health, dietary assessment measures and biological samples measures included in the health assessment and compares the measures to other comparative longitudinal studies of ageing. While many of these measures are described in detail, a comprehensive description of the protocols used is beyond the remit of this article. Further manuscripts detailing specific strands of research being conducted within NICOLA that are not included in this manuscript will be forthcoming including the results from the analysis of the Wave 1 dietary questionnaire.
Table 1Measures used in the NICOLA health assessment compared to other similar longitudinal studies of ageingOutcome MeasureType of assessmentMeasuresComparative studyPhysical HealthAnthropometricWeightHeightWaist and hip circumferenceTILDA, ELSABody compositionBodystat (% body fat)NoneCardiovascularBlood pressureOrthostatic blood pressureTILDA, ELSARespiratorySpirometryELSAMobility and strengthStep testTimed up and goGrip strength (dynamometry)TILDA, ELSAVisionVisual acuityMulti-modal retinal imagingTILDA (visual acuity)Facial photographPhysical attractiveness / signs of ageingNoneCognitive HealthCognitive functionMMSEMOCAColour trails 2Animal recallTILDADietary Intake*Food frequency questionnaireDietary intake (EPIC-FFQ)Special dietsCooking and food shoppingFood supplements / vitaminsELSA (Wave 9 only, online FFQ (Oxford-WebQ))Mental HealthMental well-beingDepressionWarwick-Edinburgh Mental Well-Being Scale (WEMWBS)Centre for Epidemiologic Studies Depression Scale (CES-D)NoneELSABiological SamplesBlood and urine sample (non-fasting)Lipid profileGenomic biomarkersDietary biomarkersBone markersInflammatory markersOther biomarkersTILDA, ELSA* Not detailed in this paper
## Body composition: height, weight, BMI, waist, hip, body fat
Changes in body composition are a normal part of ageing and often occur simultaneously with declines in physical function. Anthropometric measurements were made to provide a quantitative measure of body composition, obesity and body fat distribution that is related to overall health status and can be tracked over time. Standing height and weight were measured using standard techniques, BMI was computed as weight/height2 (kg/m2).
Waist and hip measurements were recorded using a SECA measuring tape. The waist was measured midway between the iliac crest and the costal margin (lower rib) while the hip circumference was measured at the widest circumference over the buttocks and below the iliac crest. Measurements were repeated twice. Waist-to-hip ratio was calculated as a measure of body fat distribution which is an important indicator of risk of cardiovascular disease [9]. Percentage body fat was also measured using the Bodystat 1500 MDD body composition analyser. This measures the amount of lean and fat mass that makes up total body weight.
## Physical function – step test, timed up and go, grip strength
Physical function is one of the most important indicators of health status in older adults and is closely related to quality of life. Ageing is associated with numerous anatomical and physiological changes which can adversely affect physical function, thus contributing to an increased risk of falls, fractures and disability. At a population level, impaired physical function is known to be associated with frailty [10], increased mortality [11] and greater utilisation of health services [12].
In NICOLA, objective measures of strength, mobility and balance were used to capture overall physical function as they are robust early indicators of decline in physical function. These biomarkers can help provide an indication of future risk of many health conditions and loss of independence. They are therefore useful indicators of healthy ageing as well as being a sign that early intervention is required.
The ‘step up’ test was used to measure dynamic standing balance, combining a measure of balance and lower-extremity motor control [13]. It was recorded as the number of times the participant fully stepped on and off a 7.5 cm block step in 15 s. Measurements were taken for each leg and the number of times the participant stepped up was counted and averaged across the right and left feet. The greatest number of steps completed corresponded to better dynamic standing balance.
The timed up-and-go (TUG) test is a test of mobility commonly used in clinical practice to measure mobility and risk of falling [14, 15]. Impaired mobility often precedes the onset of physical disability, falls, frailty and cognitive impairment. Slower test speeds have been shown to be related to increased risk of health conditions and mortality in older adults [16]. The test measures the time taken by the participant to stand up from a standard arm chair, walk three meters at their usual pace, turn, walk back to the chair and sit down again [15]. It is a robust test of functional mobility as it assesses proximal muscle strength, balance, executive function and gait speed. Typical values range from 8 to 11.5 s with a faster time indicating greater mobility. A time greater than 12–15 s is often used as an indicator of a high risk of falling [17] and greater than 10 s an indicator of frailty [18]. Participants were permitted to use their usual assistive device such as a cane or walking aid, and. were also permitted to stop and rest (but not sit down) during the test, if required..
Grip strength affects everyday function, such as the ability to hold heavy objects, and declines with age. A higher grip strength is associated with a reduced risk of early mortality, cardiovascular disease and disability [19]. It is also a good indicator of biological ageing [20]. Hand-grip strength was assessed using a Baseline hydraulic hand-held dynamometer. This method has previously been shown to be a reliable and valid instrument for assessing muscle strength and function [21–23] and is an indicator of frailty in older adults [24]. The participant stood with their forearm flexed at 90 degrees and squeezed the handle of the dynamometer with maximum force. Measurements were repeated twice with each hand, alternating between the dominant and non-dominant hand. The data presented represents the average of two tests using the dominant hand.
## Cardiovascular function – blood pressure
Blood pressure is a modifiable risk factor for adverse cardiovascular events such as coronary heart disease and stroke. Hypertension is recognised as one of the most preventable causes of premature morbidity and mortality. The prevalence of both diabetes and hypertension increases sharply with age but can only be dealt with properly at a population level if we know how many go undiagnosed with these conditions. Evidence suggests that many older adults are unaware that they have hypertension. In the UK, 1 in 3 adults suffer from hypertension(a reading of $\frac{140}{90}$ mm Hg or higher; [25] rising to at least 1 in 2 in those aged 65 years and over [26]. In addition, as a person ages, the tendency for postural hypotension (BP drop on standing) increases. This can result in dizziness, light headedness and increases the risk of falls. Systolic (SBP) and diastolic blood pressure (DBP) was measured using the OMRON TM digital automatic blood pressure monitor (Model M10-IT). Blood pressure and heart rate was measured three times (one minute apart) on either arm. The one-minute gap between blood pressure measurements was based on the 2005 AHA position statement [27] which recommended at that time, that at least two blood pressure readings should be taken at intervals of at least one minute and an average calculated. Given the pragmatic approach used in the design of the health assessment, a one-minute gap was also deemed more logistically feasible, in order to keep each assessment as short as possible for the participant. Two of the measurements were taken with the participant seated, while the third was recorded immediately upon standing (postural blood pressure). Hypertension was defined as SBP ≥ 140 mmHg or DBP ≥ 90 mmHg or current blood pressure-lowering treatment [25].
## Respiratory function
The respiratory system undergoes various anatomical, physiological and immunological changes with age. Ageing is associated with a progressive decline in respiratory function that accompanies changes in the structure of the chest wall due to loss of supporting tissue, increased air trapping and decreased respiratory muscle strength [28]. Respiratory function was measured using the CareFusion Microlab Spirometer with the participant seated. Measurements included forced expiratory volume in one second (FEV1, l), forced vital capacity (FVC, l) and forced expiratory flow (FEF) 25–$75\%$. Measures of lung function (FEV1 and FVC) are associated with all-cause and cardiovascular mortality [29, 30]. Low FEV1 is also recognised as an independent predictor of non-cardiopulmonary comorbidities including diabetes, chronic kidney disease, osteoporosis and dementia [31–34]. For the purposes of this manuscript the highest FEV1 and FVC reading was used. A maximum of five attempts were undertaken to obtain three satisfactory readings. Analyses are only based on participants who obtained at least three satisfactory readings.
## Vision – visual acuity
Significant losses in visual function are known to occur with normal ageing. With increasing age, the incidence of eye diseases such as cataract, age-related macular degeneration (AMD), glaucoma, and diabetic retinopathy increases significantly. Globally, cataracts, glaucoma and AMD are leading causes of adult-onset blindness [35]. Although age is a major risk factor for visual loss, other risk factors include smoking, genetic tendency, pigmentation, arterial hypertension, ultra violet light exposure and consumption of an unbalanced diet. Even in the presence of relatively good visual acuity, decreases in visual function with age are related to a decreased quality of life, mobility and independence in older adults [36]. A unique strength of NICOLA compared to our comparative studies is our ability to exploit research areas such as eye health where we have core research expertise. To maximise capacity in this area of research and capitalize on our in-house expertise, the health assessment included an in-depth ophthalmic assessment comprised of two sections: i)the Optometric assessment which evaluated visual function using distance visual acuity, refractive status using auto refractor (Shin Nippon Accuref K-900) and intra ocular pressure using the Ocular Response Analyser. Distance visual acuity measurements were performed in each eye. Habitual visual acuity was measured using the Early Treatment Diabetic Retinopathy Study (ETDRS) classification chart [37] and was recorded as the number of letters correctly identified from either the 4 m chart or the 1 m chart with and without a pinhole occlude. Participants wore their own glasses or contact lenses during the measurement. The ETDRS classification system is considered to be the gold standard for the measurement of visual acuity in clinical research and practice [38].ii)Multi-modal retinal imaging using the Canon CX-1 Color Fundus digital camera (Canon USA, Inc.), Optus P200Tx wide field retinal imaging camera (Optos plc, Dunfermline, UK) and spectral domain optical coherence tomography (Spectalis HRA + OCT) (Heidelberg Engineering, Heidelberg, Germany).
Prior to the ophthalmic imaging, tropicamide $1\%$ eye drops were applied to the pupil of each eye (or the non-dominant eye if preferred by the participant or if the participant was driving within 4 h or neither eye) of the participant in order to enlarge the pupil and thus achieve good quality retinal images. 626 ($17\%$) participants did not give consent to have eye drops administered. While reduced pupil size impacts the quality of the colour fundus photographs (CFP) the most [39], OCT is more robust to pupil size. Obtaining multiple imaging types was therefore a strength of NICOLA compared to most epidemiological studies which usually only capture CFP. Images were acquired using stereo colour fundus photography centred on the disc and macula, a single non-stereo unsteered pseudocolour ultrawide field image, Autofluorescence (AF), MultiColor (MC), Macular OCT scan centred on the fovea and a circle scan of the optic disc.
Standardized multi-modal retinal grading by the Network of Ophthalmic Reading Centres UK [40] was used to identify features of common eye conditions such as AMD, glaucoma, diabetic retinopathy, vitreous interface changes and macular holes [39]. Features of AMD such as drusen type, size and location, the presence of hyper pigmentation, presence of focal or geographic atrophy or signs of retinal neovascularization were identified [41]. Participants were then classified into AMD grades, based on the Beckman Clinical Classification System which provides a severity scale for AMD spanning from no AMD to the most severe clinical manifestations which are accompanied by vision loss. [ 42] The AMD grades are: (i) No ageing changes (no drusen present and no AMD pigmentary abnormalities); (ii) Normal ageing (small drusen ≤ 63 μm present and no AMD pigmentary abnormalities); (iii) Early AMD (medium drusen > 63 μm and ≤ 125 μm and no AMD pigmentary abnormalities); (iv) Intermediate AMD (large drusen > 125 μm and/or any AMD pigmentary abnormalities); (v) Advanced AMD (neovascular AMD and/or any geographic atrophy) [42]. Diabetic retinopathy and maculopathy were also identified and graded from all retinal imaging modalities. The English Classification system was used to categorise participants according to level of severity [43].
A subset of participants who were suspected of having glaucoma due to optic disc appearance or raised intraocular pressure were also invited to a follow up health assessment for further evaluation of glaucoma by a glaucoma expert. Tests performed at this visit included (i) visual field testing using Humphrey’s Matrix frequency doubling technology (FDT) perimetry (Carl Zeiss Meditec Inc., Dublin, CA, USA) in low illumination, (ii) Gonioscopy and (iii) pupil dilation and biomicroscopy including optic disc examination [44, 45].
## Hearing
Hearing loss is highly prevalent in older populations and is the most common sensory impairment in older adults [46]. If left untreated, hearing loss can have a profound impact on overall quality of life and everyday life through its effect on the ability to communicate and remain independent [47]. Untreated hearing loss also has indirect health, psychosocial, and economic effects thus resulting in increased feelings of loneliness, emotional distress, social isolation and withdrawal from social situations [48–51]. Those experiencing hearing loss are also likely to have other age-related conditions and are at greater risk of falls and frailty [52], as well as higher rates of cognitive decline [53–56]. Although not successful in everyone, hearing aids can improve several aspects of life that have been compromised by hearing loss. However, despite the availability of hearing aids and major technical progress in the last decade, uptake of hearing aids is poor and only a relatively small proportion of adults with hearing impairment seek help for their hearing problems and use hearing aids. In NICOLA,hearing was not measured objectively, but rather by self-report which assessed participant’s hearing ability, their use of hearing aids and coping with hearing problems including the impact of hearing loss on following conversation or using a telephone. A validation study of the self-report methods was also carried out separately in a subsample of NICOLA participants ($$n = 120$$) to examine the association between self-reported measures of hearing loss and measured hearing loss using pure-tone audiometry, the gold standard method of hearing loss assessment [57]. Low but significant correlation, and fair agreement using weighted kappa was found between self-reported measures of hearing loss and measured hearing loss by pure-tone audiometry [58].
## Facial photograph
It has previously been suggested that life experiences are reflected in your face. For example, some people look younger in a photograph than they actually are. Participants were informed in advance, via the participant information sheet, that a photograph would be taken of them sitting in a chair and that the purpose of the photo was to see how appearance changes as people get older. Two facial photos in portrait format (one face-on, one side profile) were taken of each participant using a Nikon Coolpix L610 digital camera, in order to enable comparisons with other indicators of ageing. The photo captured the participant’s face, hair and part of the neck. The participant was asked to not smile in the photo and to remove glasses and headwear. Make-up and other items such as jewellery or hearing aids were permitted. The camera lighting was set in order to capture facial texture.
## Cognitive health
Preventing dementia and cognitive decline is a global health priority. In 2010, it was estimated that there were 35.6 million people with dementia worldwide [59]. It has been predicted that this figure will approximately double every 20 years [59]. Cognitive function outcomes were determined using a cognitive battery comprising four standardized measures which assessed memory, planning, attention and reasoning. These measures included a combination of pen and paper based tests or verbal tests, with responses being recorded by the research nurse. All cognitive tests were conducted in a quiet room and in a fixed order.
## MMSE
The Mini-Mental State Examination (MMSE) was used to assess global cognition [60]. It consists of 30 brief questions (verbal and pen/paper based) which are designed to measure a range of cognitive domains including attention and concentration, memory, language, visuo-constructional skills, calculations and orientation. The MMSE took approximately 5 min to administer. A score (out of 30) based on performance across the 11 components of the test (orientation to time, orientation to place, registration, attention and calculation, recall, naming, repetition, comprehension, reading, writing and drawing) was calculated for each participant. Total scores ranged between 0 and 30, with lower scores indicative of more severe cognitive impairment and scores of 25 or over indicating no cognitive impairment.
## MOCA
The Montreal Cognitive Assessment (MoCA) [61] is typically used as a rapid screening instrument for mild cognitive impairment. It is more sensitive than the MMSE to mild cognitive impairment [62]. It assesses different cognitive domains including attention and concentration, executive functions, memory, language, visuoconstructional skills, conceptual thinking, calculation and orientation. The test which included a combination of verbal and pen/paper based tests took approximately 5–10 min to complete. A score (out of 30) based on performance was calculated for each participant with lower scores indicating greater cognitive impairment and scores of 26 or over indicating normal cognitive functioning.
## Colour trails 2
The Colour Trails 2 test was used to measure executive function and visual scanning. Participants were instructed to draw a line as quick as possible between consecutively numbered circles, but alternating between pink and yellow colours [63]. The length of time taken to complete the test was recorded in minutes, seconds and centi-seconds. The number of near misses, prompts, colour sequence errors and number sequence errors made by the participant was also recorded.
## Animal Recall
Animal recall is a measure of executive function (e.g. strategic search and set-shifting) and semantic memory. Participants were asked to verbally name as many animals as possible within 60 s [64]. One point was given for each animal named. The number of animals named was recorded by the research nurse. Different species, genders or generations of animals were counted separately (e.g. dog, spaniel, bull, calf) but redundancies were not (e.g. brown cow, white cow). One point was allocated for each animal named by the participant with the total number reflecting verbal fluency score.
## Warwick Edinburgh Mental Well-being Scale
The Warwick-Edinburgh Mental Wellbeing Scale (WEMWBS) is a scale of 14 positively worded items such as “I’ve been feeling interested in other people” and “I’ve been feeling good about myself” and is used to assess the mental health of the general population. For each statement, participants were asked to rate on a Likert scale of 1 to 5 how often they had felt like that, with one being “none of the time” and five being “all of the time”. Scores on the WEMWBS ranged from 14 to 70, with higher scores indicating higher levels of wellbeing. Participants self-completed the questionnaire, using pen-paper, in private and then returned the completed questionnaire to the research nurse in a sealed envelope. The WEMWBS has been validated for use in the UK in those aged 16 years and above [65] and specifically for the general population in Northern Ireland [66].
## Centre for Epidemiological Studies Depression Scale (CES-D)
Depressive symptoms are known to influence cognition and it is important to control for mood when analysing cognitive results. However, many studies of ageing have excluded patients with depression from cognitive trials and vice versa in depression trials. It is important to be able to track the stability of a person’s mood over time and how changes in mood relate to future health status. The CES-D consists of 20 items phrased as statements, each one assessing symptoms associated with depression [67]. Participants had to verbally respondon a scale of 0 to 3 how often they had experienced that symptom over the previous 7-day period ranging from 0 (“rarely or none of the time”) to 3 (“most or all of the time”). Four of the items are positive statements which are inversely scored. Responses to each item were summed to generate a total score ranging from 0 to 60 with a higher score indicating a higher degree of depressive symptoms [68]. *In* general, a score of ≥ 16 is indicative of moderate or potentially clinically relevant depressive symptoms while a score of 8–15 indicates mild or sub-threshold depressive symptoms [67, 68]. The CES-D scale has been shown to be reliable at measuring the number, types and duration of depressive symptoms [68]. However, it is important to note that while the CESD is widely used across large scale population based epidemiological studies, it only assesses symptoms over the previous 7-day period rather than over a longer period of time. It is also considered to be a psychometric screening tool for depression and not a diagnostic tool [68].
## Dietary intake
A healthy diet is an integral part of healthy ageing and plays a key role in chronic disease prevention and in reducing the risk of cognitive decline [69–71]. Indeed, the importance of eating a healthy balanced diet as we get older cannot be underestimated as it protects against illness, helps to speed recovery from illness and importantly, maximizes the chances of living longer and independently in good health [72, 73]. However, the ageing process results in many physiological, social and psychological changes that can affect nutritional intake and status thus increasing the risk of malnutrition [73–76]. NICOLA is unique in that it is one of the few longitudinal studies of ageing which includes a detailed dietary assessment [77].
Exploring the effects of diet on the ageing process is a core focus of NICOLA. Dietary intake was assessed using the validated 130-item food frequency questionnaire (FFQ) (EPIC-Norfolk) (CAMB/PQ/$\frac{6}{1205}$) [78]. Participants were asked to record the frequency of consumption (never or less than once a month, 1–3 times per month, once a week, 2–4 times per week, 5–6 times per week, once a day, 2–3 times per day, 4–5 times per day, 6 + times per day) of standard portions of foods over the previous 12 month period. Additional components of the FFQ included questions relating to special diets, supplement use, eating outside the home and the cooking, preparation and shopping for food.
Given the uncertainty over the utility of a FFQ to determine dietary intake in older people, a validation study of the NICOLA FFQ was also incorporated into the design of the dietary assessment [79]. In addition to completing a FFQ and providing a blood sample, a subsample of the NICOLA cohort ($$n = 44$$ men and $$n = 51$$ women) also completed two food diaries as a reference method (6 months apart) and provided additional blood, urine and saliva samples for measurement of nutritional biomarkers. Of these 95, 23 participants also took part in multiple 24 h recalls.
Findings from the in-depth dietary analysis of the FFQ including energy and nutrient intakes, dietary patterns, dietary supplement use and the dietary validation study are beyond the remit of this paper and will become available in due course. This work will allow us to address the lack of dietary validation studies in older people to date and will allow us to test numerous hypotheses around diet-disease and diet-function relationships in older people.
## Biological samples
Analysing biological samples enables us to objectively evaluate biomarkers that act as an indicator of a person’s health. Biomarkers can also provide an early indication of disease before symptoms arise, provide us with information on disease progression and/or suggest therapies. In NICOLA, non-fasting venous blood samples were obtained from consenting participants. These included blood serum, plasma (EDTA/clot activator), glucose (potassium oxalate/sodium fluoride) and RNA (PAXgene). A spot urine sample was also obtained from all participants. All biological samples were transported in temperature controlled containers to a central laboratory and processed within 4 h. Aliquoted samples were subsequently frozen at − 80 °C until analysis. A dedicated courier service was used for transporting samples collected at the home-based assessments. As described previously, detailed laboratory analysis was conducted on all of the samples which included multi-omic biomarkers, lipid profiling, dietary biomarkers, inflammatory biomarkers and hormones [3]. All laboratory assays were standardised against available international standards, and quality control samples were included in every run. Participants consented separately for the collection of blood, DNA, urine, retinal images, facial photograph and the administration of the eye drops including consent for analysis, storage and future contact. Data are currently available for 28 biochemical biomarkers from 3082 participants within the NICOLA cohort. Participants were also offered rapid testing and feedback from blood glucose and lipid levels. NICOLA has a strong focus on molecular biomarkers and complementary genetic, epigenetic and transcriptomic data is available for a subset of participants. There is also 551,830 directly genotyped and 18,148,478 imputed SNPs currently available for 2969 participants.
## Study management
NICOLA is managed under the ethics and governance approval processes of Queen’s University Belfast. Ethical approval for the health assessment was granted by the School of Medicine, Dentistry and Biomedical Sciences Ethics Committee, Queen’s University Belfast. Written consent was obtained from all participants prior to participation in the health assessment.
The NICOLA data and sample resource is governed by the NICOLA Steering Committee, Data Access Committee and Research Support Team. The Steering Committee provides oversight on all research carried out on study participants and on data and advises on the best ways of optimising scientific potential. This interdisciplinary team includes experts across various research areas including chronic illness, physical activity, the built environment, nutrition, eye health, cognitive health, mental health, frailty, social environment, and multi-omics. Approved researchers, members of academia, and others from the third sector, practitioner, government and policy communities who wish to access the anonymised dataset can do so by making an application using the designated Research Proposal Form available on the study website. Research proposals to access data from the NICOLA resource must be in accordance with the NICOLA Data Access Policy and follow a standardised review and approval process by the NICOLA Data Access Committee. The approval structure includes regular operation of the Data Access Committee which oversees data access, review proposals, and tracks published papers and public engagements. Each separate research project is assessed against data governance criteria and a determination is made as to whether the outcomes meet the remit of the NICOLA study governance objectives. The Research Support *Team is* responsible for maintaining the security of the data and ensuring confidential access and for managing and curating the research data generated from NICOLA. All outputs generated from the data are subject to a disclosure control assessment. The NICOLA research support team currently manage the integration of study data with linked routine records, integrate the research application process and provide secure data access to research users.
All data is collected, stored and disseminated in accordance with the QUB Research Management Policy as well as in line with UK General Data Protection Regulations (GDPR), Data Protection Act [2018], Human Tissue Authority Codes of Practice and in accordance with the NICOLA Data and Sample Access Policy https://www.qub.ac.uk/sites/NICOLA/InformationforResearchers/#requesting-access-to-nicola-data-or-biological-samples-910951-1. The Data and Sample Access Policy describes in detail the general processes and procedures involved in accessing the NICOLA data resource (defined as data already collected and the participants themselves for the purposes of new data collection) and NICOLA samples (biological, clinical, and multi-omic). Within NICOLA, we aim to encourage and facilitate data access with all ‘bona fide’ researchers and research organisations as defined by UK Research and Innovation (UKRI) (https://www.ukri.org/) and welcome proposals from researchers, either for collaborative projects or for other forms of data access to help advance research knowledge.
## Statistical analysis
Descriptive statistics were obtained for all selected baseline variables of interest. Continuous and categorical variables were summarized as mean (SD) and n (%) respectively. Data where distributions were positively skewed are presented as median (interquartile range). Chi-square tests were used to compare categorical data. The main statistical analysis offered in this paper is designed for descriptive purposes [80]. For all comparisons, study participants have been classified according to type of physical health assessment: clinic based, home based or none. Characteristics of respondents were compared across visit type using analysis of variance for continuous measures and chi-square tests for categorical variables. Binary logistic regression was used to compare participants who attended the health assessment versus non-attendees. Values for the logistic regression analysis are presented as Odds ratio ($95\%$ CI). For all analyses, $p \leq 0.05$ was considered statistically significant. Statistical analyses were performed with SPSS v24.0 for Windows (SPSS Inc, Chicago, IL).
## Wave 1 Health Assessment Response Rates
Response rates are presented in Table 2. Of the participants who completed the Wave 1 CAPI, $44\%$ ($$n = 3655$$) also took part in the health assessment. The majority of participants attended the Clinical Facility for the health assessment ($95\%$, $$n = 3462$$), with the remainder taking place in the participants’ home ($5\%$, $$n = 193$$). The majority ($96\%$, $$n = 3514$$) of participants who attended the health assessment also consented to providing a venous blood and urine sample.
Table 2Completion rates of the NICOLA Wave 1 health assessmentHealth AssessmentAge group (yrs)All (n)Clinic Based (n)Home Based (n)50–64197719374065–741187112562≥ 7549140091Total36553462193
## Characteristics of health assessment attendees
Table 3 describes selected baseline demographic, anthropometric and biological characteristics of participants who attended the health assessment (either at home or at the clinic). The majority of participants who attended the health assessment were in the youngest age category (i.e. aged 50–64 years); had reached secondary level education; were married; retired; and were a non-smoker. Participants in the older age category (i.e. age ≥ 75 years), with a lower level of education, and single were less likely to take part in the health assessment. Almost all participants ($99\%$) were of white ethnicity.
Table 3Selected sociodemographic and physical characteristics of participants who completed the NICOLA health assessment, by genderAll(nmax=3655)Men(nmax=1761)Women(nmax=1894)P value1 Health Assessment location ClinicHome3462 (94.7)193 (5.3)1678 (95.3)83 (4.7)1784 (94.2)110 (5.8)0.14 Age group (yrs) 50–6465–74≥ 751977 (54.1)1187 (32.5)491 (13.4)884 (50.2)613 (34.8)264 (15.0)1093 (57.7)574 (30.3)227 (12.0)< 0.001 *Marital status* Married / cohabitingNever marriedSeparated / divorced / widowed2643 (72.3)244 (6.7)768 (21.0)1372 (77.9)103 (5.8)286 (16.2)1271 (67.2)141 (7.4)482 (25.4)< 0.001 Employment RetiredEmployedSelf-employed (incl farming)UnemployedPermanently sick / disabledLooking after home / family / other21857 (50.9)1059 (29.0)369 (10.1)86 (2.4)141 (3.9)137 (3.7)908 (51.7)439 (25.0)254 (14.5)59 (3.4)73 (4.1)23 (1.3)949 (50.1)620 (32.8)115 (6.1)27 (1.4)68 (3.6)114 (6.0)< 0.001 Education Primary / noneSecondaryHigher577 (15.8)1602 (43.8)1476 (40.4)353 (20.0)738 (41.9)670 (38.0)224 (11.8)864 (45.6)806 (42.6)< 0.001 Region UrbanIntermediateRural635 (17.4)2048 (56.2)964 (26.4)283 (16.1)1013 (57.7)461 (26.2)352 (18.6)1035 (54.8)503 (26.6)0.09 Multiple Deprivation Measure 3 1 (least deprived)2345 (most deprived)991 (27.1)763 (20.9)733 (20.1)658 (18.0)510 (14.0)498 (28.3)360 (20.4)353 (20.0)296 (16.8)254 (14.4)493 (26.0)403 (21.3)380 (20.1)362 (19.1)256 (13.5)0.26 Anthropometric and body composition Height (cm)Weight (kg)Body mass index (kg/m2)Body fat (%)Waist circumference (cm)Hip circumference (cm)Waist:hip ratio165.6 (9.3)79.7 (16.6)29.0 (5.2)44.2 (6.5)95.8 (14.1)104.7 (9.8)0.91 (0.1)172.6 (6.7)87.2 (14.8)29.2 (4.5)44.7 (5.7)101.9 (11.9)104.5 (7.6)0.97 (0.1)159.1 (6.1)72.7 (15.2)28.7 (5.8)43.8 (7.2)90.2 (13.7)105.0 (11.4)0.86 (0.1)< 0.001< 0.0010.002< 0.001< 0.0010.06< 0.001 Blood pressure SBP (mmHg)DBP (mmHg)SBP, postural (mmHg)DBP, postural (mmHg)132.7 (18.8)81.3 (10.9)131.6 (20.4)83.8 (11.2)136.6 (17.8)82.4 (10.9)136.3 (20.1)85.1 (11.1)129.1 (19.1)80.2 (10.9)127.2 (19.7)82.5 (11.1)< 0.001< 0.001< 0.001< 0.001 Lipid and diabetes profile Total cholesterol (mmol/l)5.1 (1.2)4.8 (1.1)5.5 (1.1)< 0.001HDL cholesterol (mmol/l)Triglycerides (mmol/l)HbA1c (%)1.5 (0.4)1.5 (1.1, 2.1)5.8 (0.9)1.3 (0.3)1.6 (1.1, 2.2)5.8 (1.0)1.6 (0.4)1.4 (1.0, 2.0)5.8 (0.8)< 0.001< 0.0010.07Respiratory function:Forced vital capacity (FVC) (l)Forced expiratory flow (FEV1) (l)3.5 (0.9)2.6 (0.7)4.1 (0.9)3.0 (0.7)3.0 (0.6)2.2 (0.5)< 0.001< 0.001Cognition:MMSE (score)MOCA (score)Animal recall (number)Colour trails 2 (time in secs)28.4 (1.8)25.3 (3.3)19.0 (5.6)118.5 (41.5)28.2 (1.8)25.0 (3.2)19.2 (5.7)124.4 (45.2)28.6 (1.8)25.6 (3.3)18.9 (5.5)113.4 (37.4)< 0.001< 0.0010.11< 0.001Mobility and strength:Step testTimed Up and Go (seconds)Grip strength (kg)16.4 (4.3)10.1 (2.8)31.2 (11.8)16.7 (4.3)10.0 (2.6)39.8 (9.8)16.2 (4.2)10.1 (3.0)23.0 (6.4)0.0010.27< 0.001 Hearing Use of hearing aid (self-reported)All of the timeSome of the time176 (4.8)171 (4.7)91 (5.2)104 (5.9)85 (4.5)67 (3.5)0.004Quality of hearing (self-reported)Excellent / very good / goodFair / poor2765 (76.0)875 (24.0)1222 (69.7)530 (30.3)1543 (81.7)345 (18.3)< 0.001 Mental Health CES-D (score)WEMWBS (score)5 [2, 11]55 [49, 61]4 [2, 10]55 [49, 61]6 [2, 12]55 [48, 61]< 0.0010.27 Ophthalmic Wear glasses or contact lensesVisited optometrist in last 12 moCataract (self-reported)Glaucoma (self-reported)AMD (self-reported)Visual acuity, distance, pin hole, better eye (number of letters)Refractive error, most severe eye (spherical equivalent from auto refractor)3488 (95.8)2242 (61.5)775 (21.3)92 (2.5)90 (2.5)82 [6]0.83 (2.9)1672 (95.2)1026 (58.4)344 (19.6)45 (2.6)51 (2.9)83 [6]0.93 (2.69)1816 (96.4)1216 (64.5)431 (22.9)47 (2.5)39 (2.1)82 [6]0.74 (3.09)0.11< 0.0010.010.890.11< 0.0010.06Age-related macular degeneration4No AMDNormal ageingEarly AMDIntermediate AMDAdvanced AMD1637 (51.4)761 (23.9)525 (16.5)235 (7.4)26 (0.8)789 (51.3)374 (24.3)232 (15.0)129 (8.4)15 (1.0)848 (51.6)387 (23.5)293 (17.8)106 (6.4)11 (0.7)0.07Glaucoma5Absent (both eyes)3202 (96.9)1549 (96.8)1653 (97.0)0.75Present (either eye)103 (3.1)52 (3.2)51 (3.0)Values are unweighted mean (SD) or median (IQR) for continuous variables or n (%) for categorical variables1 p value represents difference between men and women2 Other includes those in education/training (categories were combined due to low cell counts)3 Based on the Northern Ireland Multiple Deprivation Measure 2010 [81]4 Measurement based on retinal image and Beckman Clinical Classification5 The International Society of Geographical and Epidemiological Ophthalmology (ISGEO) Classification As presented in Table 3, significant differences were evident between men and women who attended the health assessment. Women tended to be in the youngest age group category (50–64 years); were more likely to be separated/divorced/widowed; to be employed or looking after home/family; and tended to have a higher level of education compared to men.
In terms of physical characteristics, blood pressure and lung function measurements (FVC and FEV1) were significantly higher in men who attended the health assessment than in women. Lung function measurements are typically higher in men than women although as well as sex, lung function also depends on age and height [82]. This will be examined in more detail at a later stage. Anthropometric measurements which included height, weight, BMI, body fat, waist circumference and waist:hip ratio were also higher in men than women. On average, NICOLA participants were overweight, with a mean BMI of 29.2 kg/m2 in males and 28.7 kg/m2 in females. Waist circumference measurements were also high in both men and women (101.9 cm and 90.2 cm, respectively). A waist circumference of ≥ 94 cm in men and ≥ 80 cm in women is associated with increased risk of developing obesity-related health problems. As well as measuring waist circumference, the ratio of waist to hip circumference is also used to indicate health risk. Mean waist:hip ratio was higher than the recommended level for both men and women (0.97 and 0.86, respectively). A waist-hip ratio > 0.90 and 0.85 for men and women respectively is associated with increased risk of a number of diseases including heart disease and Type 2 diabetes and is a better predictor of early mortality than BMI or waist circumference in older adults [9]. Grip strength was also higher in male attendees than women (39.8 kg and 23.0 kg, respectively). When examined by age category, age related decline in grip strength was greater for men than women (mean grip strength at age 50–64 years: men 43.8 kg, women 24.7 kg; age 65–74 years: 37.7 kg and 21.4 kg, respectively; age 75 + years: 31.2 kg and 18.3 kg, respectively) (data not shown). These values are within the expected normative values. For example, grip strength for a 65–75 year old is between 42.3 kg and 35.6 kg for men and between 25.3 kg and 21.4 kg for women [83]. This is consistent with the findings of TILDA [4, 84]. In terms of hearing and vision, a higher proportion of men ($11\%$) reported using a hearing aid compared to women ($8\%$) and reported that their hearing was fair/poor compared to women ($30\%$ and $18\%$, respectively). Likewise, TILDA similarly reported that men are more likely than women to use a hearing aid and to also report their hearing as fair/poor [84]. Visits to the optometrist in the previous 12 month period were higher in women than men ($65\%$ and $58\%$, respectively) and a higher proportion of women reported that they had been diagnosed with cataracts than men ($23\%$ and $20\%$, respectively). Just under $3\%$ of both men and women reported a previous diagnosis of glaucoma and AMD. Based on the retinal image measurements rather than self-reported history, approximately three quarters of participants showed retinal changes consistent with normal ageing or no AMD (Class 0 or 1) while $16\%$ had early AMD, $7.4\%$ had intermediate AMD and $0.8\%$ had advanced AMD in their worst eye. This is higher than the prevalence of AMD reported in the TILDA study [84] which had an estimated overall prevalence of $7.2\%$, with early / intermediate AMD accounting for $6.6\%$ and late AMD accounting for $0.6\%$. Of the participants in the current study who were found to have advanced AMD, based on retinal imaging, approximately $40\%$ did not report a positive history of having AMD (data not shown). Glaucoma was prevalent in approximately $3\%$ of participants. These findings are comparable to pooled estimations of glaucoma prevalence in other European populations (based on age range 40–80 years) ($2.93\%$, $95\%$CI 1.85, 4.40) [85]. Only $30\%$ of participants who were found to have glaucoma during the health assessment had reported a positive history (data not shown).
In terms of cognition, the mean MMSE and MOCA score was 28.4 and 25.3, respectively. Similar to the findings from TILDA [83], the majority of participants (approximately $95\%$) showed normal levels of cognition (i.e. MMSE score 25–30) (data not shown). Performance in the colour trails 2 test showed the greatest difference between males and females, with females completing the test significantly faster than males (113 and 124 s, respectively). Similar to TILDA, symptoms of depression as reflected by the CES-D score were higher in women than men [84].
In terms of lipid profile, females had higher mean total cholesterol (5.5 mmol/l), HDL cholesterol (1.6 mmol/l) and lower triglycerides (1.4 mmol/l) compared to males (4.8 mmol/l, 1.3 mmol/l and 1.6 mmol/l, respectively). Mean cholesterol levels in females were higher than current recommendations which suggest that total cholesterol levels should be < 5 mmol/l in both men and women. HDL cholesterol levels should be ≥ 1.1 mmol/l in men and ≥ 1.2 mmol/l in women while non-fasting triglyceride should ideally be < 2.3 mmol/l [86].
Table 4 presents the difference in selected characteristics among those who had a home based health assessment, clinic based health assessment or no health assessment. When comparing characteristics across categories, those who attended the clinic for the health assessment tended to be 50–64 years old, women, married or cohabiting, living with others, retired, had secondary level of education, lived in an intermediate area in terms of urban/rural divide, were least deprived, were more likely to be a non-smoker, and a current consumer of alcohol. A greater proportion ($44\%$) of those who attended the clinic based health assessment reported excellent or very good levels of health compared to those who opted for a home based assessment ($23\%$) or who did not have a health assessment ($32\%$).
Table 4Selected characteristics of attendees (clinic-based versus home-based assessment) and non-attendees of the health assessmentOdds ratio for attending health assessment (home or clinic)1Clinic based health assessment($$n = 3462$$)Home based health assessment($$n = 193$$)No health assessment($$n = 4628$$)P value2Exp (B) ($95\%$ CI)n (%)n (%)n (%) Age group (yrs) 50–6465–74≥ 75Ref***0.97 (0.85, 1.11)0.53 (0.45, 0.63)1937 (56.0)1125 (32.5)400 (11.6)40 (20.7)62 (32.1)91 (47.2)2208 (47.8)1309 (28.3)1111 (24.0)< 0.001 Gender MaleWomenRef***0.80 (0.73, 0.88)1678 (48.5)1784 (51.5)83 (43.0)110 (57.0)1990 (43.0)2638 (57.0)< 0.001 *Marital status* Married / cohabitingNever marriedSeparated / divorced / widowedRef1.72 (0.70, 4.21)1.93 (0.80, 4.65)2557 (73.9)227 (6.6)678 (19.5)86 (44.6)17 (8.8)90 (46.6)2772 (59.9)420 (9.1)1436 (31.0)< 0.001 *Living status* Living aloneLiving with othersRef2.27 (0.94, 5.45)908 (26.2)2554 (73.8)107 (55.4)86 (44.6)1876 (40.5)2752 (59.5)< 0.001 Employment 3 RetiredEmployed/Self-employedPermanently sick/ disabledLooking after home / familyIn education / training / otherRef***0.83 (0.72, 0.95)0.74 (0.54, 1.01)0.44 (0.34, 0.53)0.50 (0.38, 0.64)1712 (49.6)1413 (40.9)122 (3.5)174 (5.0)35 (1.0)145 (75.1)15 (7.8)19 (9.8)14 (7.3)0 [0]2327 (51.2)1303 (28.6)463 (10.2)420 (9.2)35 (0.8)< 0.001 Education Primary / noneSecondaryHigherRef***1.71 (1.51, 1.94)2.54 (2.21, 2.92)511 (14.8)1504 (43.5)1444 (41.7)63 (32.6)98 (50.8)32 (16.6)1529 (33.6)2034 (44.6)993 (21.8)< 0.001 Region UrbanIntermediateRuralRef***0.78 (0.68, 0.89)0.67 (0.58, 0.78)612 (17.7)1754 (50.7)1096 (31.6)26 (13.5)99 (51.3)68 (35.2)733 (15.8)2304 (49.8)1591 (34.4)0.04 Multiple Deprivation Measure 4 1 (least deprived)2345 (most deprived)Ref***0.98 (0.85, 1.14)0.77 (0.66, 0.88)0.82 (0.70, 0.94)0.62 (0.53, 0.72)962 (27.8)733 (21.2)696 (20.1)600 (17.3)471 (13.6)29 (15.0)30 (15.5)37 (19.2)58 (30.1)39 (20.2)825 (17.8)756 (16.3)1017 (22.0)963 (20.8)1067 (23.1)< 0.001 Self-reported health ExcellentVery goodGoodFairPoorRef***0.94 (0.80, 1.10)0.89 (0.76, 1.05)0.76 (0.64, 0.90)0.60 (0.48, 0.74)522 (15.1)997 (28.8)1057 (30.6)648 (18.7)235 (6.8)12 (6.2)32 (16.6)39 (20.2)63 (32.6)47 (24.4)428 (9.4)1020 (22.3)1245 (27.3)1153 (25.3)719 (15.7)< 0.001 *Smoking status* CurrentFormerNeverRef***2.01 (1.73, 2.34)1.97 (1.70, 2.29)349 (10.1)1294 (37.4)1816 (52.5)32 (16.6)76 (39.4)85 (44.0)980 (21.5)1498 (33.0)2066 (45.5)<0.001 *Alcohol status* CurrentFormerNeverRef***0.74 (0.66, 0.85)0.81 (0.71, 0.92)2408 (69.6)491 (14.2)561 (16.2)80 (41.4)53 (27.5)60 (31.1)2455 (54.1)1045 (23.0)1040 (22.9)< 0.001***$p \leq 0.001$ for difference between socio-demographic and health characteristics and odds of health assessment attendance1 Each odds ratio controls for the other variables included in the table2 P value is the statistical difference between groups3 Some categories were combined due to low cell counts4 Based on the Northern Ireland Multiple Deprivation Measure [81] In contrast to the clinic based assessment, those who opted for a home based health assessment tended to be older (aged 75 years or over), separated/divorced/widowed, living alone, were more socially deprived, and had fair or poor self-reported health. Compared to those who attended the clinic, a higher proportion who opted for the home assessment were women, retired and with secondary education. Thedifferences in characteristics of participants depending on the location of the health assessment are somewhat consistent with those observed by TILDA with the exception of smoking status. In TILDA, respondents who chose to have a home based assessment were more likely to be a current smoker [4]. Just over a third ($35\%$) of those who opted for a home based health assessment were from a rural area versus $13.5\%$ of those who resided in an urban area. In comparison, TILDA reported that $13\%$ of rural participants opted for a home assessment compared to $7.8\%$ of those from an urban area [4].
Those who did not undertake a health assessment were more likely to be 50–64 years old, women, married/cohabiting, living with others, retired, had secondary education, living in an “intermediate” area i.e. other city or town outside the city of Belfast, were more deprived, had good self-reported health, were a non-smoker and currently consumed alcohol.
Table 4 also shows the association between population characteristics and the odds of the participant taking part in the health assessment, regardless of whether it was conducted at home or at the clinic. The likelihood of attending the health assessment (either home or clinic) was significantly higher in the youngest age category (i.e. 50–64 years), in males, retired, in those with a higher level of education and who rated their health as excellent. Participation rates in the health assessment were higher in those who were less deprived and lived in an urban area. Respondents were also twice as likely to be a former smoker (or non-smoker) and more likely to consume alcohol.
Table 5 shows the differences in selected physical and biological characteristics of participants who opted for a home based health assessment compared to a clinic based health assessment. Those who opted for a home based health assessment were shorter in height, had a higher percentage body fat, waist and hip circumference and waist:hip ratio compared to those who attended the clinic for the health assessment. Similar to TILDA, BMI and SBP was also higher in those who opted for a home based health assessment. Lipid profiles differed between home-assessed participants compared to clinic-assessed with lower levels of total cholesterol and HDL cholesterol and higher levels of HbA1c in those who had a home based health assessment. TILDA also reported lower levels of total cholesterol in home-assessed participants [4]. Similar to TILDA [4], levels of cognition, psychological health, and physical function were also lower in those who had a home based health assessment.
Table 5Differences in selected physical characteristics of participants according to type of health assessment i.e. clinic-based versus home-basedClinic based health assessment(nmax=3462)Home based health assessment(nmax=186)P value1 Anthropometric and body composition Height (cm)Weight (kg)Body mass index (kg/m2)Body fat (%)Waist circumference (cm)Hip circumference (cm)Waist:hip ratio165.8 (9.2)79.7 (16.6)28.9 (5.1)44.0 (6.4)95.6 (14.0)104.6 (9.6))0.91 (0.09)161.8 (9.3)78.8 (18.1)30.1 (6.6)47.9 (7.8)100.0 (15.6)107.1 (13.0)0.93 (0.09)< 0.0010.46< 0.01< 0.001< 0.0010.02< 0.01 Blood pressure SBP (mmHg)DBP (mmHg)SBP, postural (mmHg)DBP, postural (mmHg)132.5 (18.7)81.4 (10.9)131.4 (20.2)80.4 (12.4)137.2 (20.9)78.1 (11.8)134.6 (23.0)84.0 (11.1)< 0.01< 0.0010.08< 0.001 Lipid and diabetes profile Total cholesterol (mmol/l)HDL cholesterol (mmol/l)Triglycerides (mmol/l)HbA1c (%)5.1 (1.2)1.5 (0.4)1.5 (1.1, 2.1)5.8 (0.9)4.6 (1.3)1.3 (0.4)1.5 (1.1, 2.2)6.1 (1.1)< 0.001< 0.0010.05< 0.001Cognition:MMSE (score)MOCA (score)Animal recall (number)Colour trails 2 (time in secs)28.5 (1.7)25.5 (3.1)19.2 (5.5)116.5 (39.4)26.8 (2.7)22.2 (4.2)15.3 (5.2)165.0 (60.4)< 0.001< 0.001< 0.001< 0.001Mobility and strength:Step testTimed Up and Go (seconds)Grip Strength (kg)16.6 (4.2)10.0 (2.7)31.5 (11.8)13.0 (4.1)13.1 (4.8)25.0 (10.6)< 0.001< 0.001< 0.001 Mental Health CES-D (score)WEMWBS (score)5 [2, 11]55 [49, 61]8 [3, 16]52 [43, 59]< 0.001< 0.001Values are unweighted mean (SD) or median (IQR).1P value for statistical difference between groups
## Discussion
NICOLA is the first large scale longitudinal study of ageing in Northern Ireland, providing a basis for future government policy by following the trajectories of ageing in 8,500 men and women aged 50 years and over. The study adopts a conceptual framework [87], approach and methods that are closely aligned to other large scale longitudinal studies of ageing across the world including the HRS, ELSA, and TILDA [4, 5, 8], thus allowing cross-national comparisons of the NICOLA findings with those from other studies. Enabling comparative studies and learning from best practice is important for identifying local population needs and informing the modernisation of health and socio-economic policies and public services for older adults [88, 89].
The work presented within this paper demonstrates the multi-disciplinary nature of NICOLA and describes the scientific rationale for the choice of health assessments as well as providing an overview of the design and methodologies used in conducting the health assessment component of the study. The information presented can inform other population based studies of ageing in relation to study design and incorporation of objective measures of health into methodologies. The scope and wealth of data obtained will help contribute to the global body of evidence regarding harmonization of health measurements in older adults.
This paper also highlights the marked differences in characteristics of participants who attended the clinicbased health assessment compared to those who had a home based health assessment or no health assessment. Attendance at the health assessment clearly depended on the demographic characteristics, health and wellbeing of the respondents. Indeed there were marked differences in characteristics of participants who opted for a home based health assessment compared to those participants who travelled to Belfast for the clinic based health assessment. Significant differences were also evident between those who took part in the health assessment (either at home or clinic) compared to those who declined to take part in the health assessment in terms of demographic characteristics, behavioural factors, physical function and health status.
Respondents who attended the health assessment (either at home or clinic) were more likely to be younger (i.e. 50–64 years), male, retired or self-employed, have a higher level of education, and rate their health as excellent. Participation rates in the health assessment were higher in those who lived in an urban area with low levels of deprivation. Health assessment participants were also twice as likely to be a non-smoker (or former) and more likely to consume alcohol. The differences observed are consistent with those reported in other longitudinal cohort studies. Indeed, in cohort studies of older adults, age and cognition have been identified as the two main contributing factors to non-participation [90].
The findings presented highlights the importance of offering participants a home based assessment or a clinic based assessment. Offering participants the option of a home based health assessment helps to boost participation rates and helps to avoid potential under representation of older and more frail participants, particularly those who have mobility problems. However, while including a home based option might help optimize participation in the health assessment it nonetheless has limitations in terms of the breadth and scope of measures that can be undertaken. Clinic based health assessments can help facilitate a much broader and detailed physical health assessment.
Those who chose not to participate in the health assessment also represent a distinct group of older adults. Despite the robust sampling strategy within NICOLA, the difference in characteristics of those who took part in the health assessment versus those who did not take part highlights a need to target future recruitment strategies at certain demographic groups in order to ensure better representation of the population. Weights have subsequently been derived within the dataset to allow for these systematic differences in response and to ensure that estimates derived from the sample in different analyses remain representative of the Northern Ireland older population. These weights are based on factors which were shown to affect the likelihood of attending the health assessment including: age, sex, education, marital status, self-reported health, smoking status, alcohol status, location (Belfast; city or town; rural) and income domain score.
## Strengths of the NICOLA data and bioresource
The data from the NICOLA health assessment will provide a more comprehensive picture and understanding of the health challenges faced by today’s older adults and provide a discovery platform for researchers to try to unravel and address these challenges. The combination of objective and subjective data can shed light on the underlying mechanisms and pathways to sustained health as we age, and tell us more about the relationships between our biology, our lifestyle and our health outcomes. The findings will also undoubtedly provide a key knowledge base for decision makers developing and prioritising policy initiatives that are core to the health and wellbeing of older populations.
The value of NICOLA lies in its longitudinal design and large sample size. Without this, it is impossible to understand the crucial drivers of trajectories of ageing in Northern Ireland. The longitudinal design of NICOLA makes it well placed to continuously monitor changes in the trajectory of ageing and health status of older adults and review the impact of health policies on outcomes in Northern Ireland. As we follow up the NICOLA participants into old age, the insights will be further enriched, therefore the full potential of the data resource has yet to be exploited. Further in-depth research on various health domains and identification of novel biomarkers of ageing is ongoing.
*The* generation of molecular biomarkers and availability of rich multi-omic data within NICOLA’s bioresource provides a powerful resource. *The* generation of genetic-epigenetic-transcriptomic data, linked to biochemical biomarkers and extensive phenotype information, will help facilitate a broad spectrum of research. To date, NICOLA’s bioresource has helped identify multiple biological markers associated with more than 30 different phenotypes [91]. NICOLA has also contributed to developing innovative new approaches for multi-omic analyses, critically highlighting the importance of careful DNA and RNA storage for robust experimental studies. Early detection of declining health, particularly in the asymptomatic stages, is very important to facilitate early interventions that promote health and minimise loss of function and NICOLA is already identifying novel biomarkers for cardiovascular, eye, and kidney-related outcomes [91–95]. The combination of psychosocial phenotypes derived from our CAPI and the bioresource is also facilitating exploration of how social experiences and life adversity, for example social disadvantage, stressful exposures or traumatic events which is captured within the CAPI impacts the epigenome and health outcomes. This work is exploring how life circumstances in both childhood and adulthood affect epigenetic change and how different historical and life-course events and experiences influence health outcomes and the rate at which we age. Biological markers identified through this work could then be used to promote and maximise healthy ageing. Through this work we are also examining whether epigenetic changes are a cause or a consequence of particular ageing trajectories [95]. This research provides the opportunity for NICOLA to harmonise data with other international cohort studies and generate new molecular data.
## Future work
The NICOLA study will continue to expand its global impact and breadth of research. One such example is its current involvement in a new global collaboration to support cross-national research into dementia as part of a US National Institute of Health (NIH) grant for Harmonizing Cognitive Assessments in Irish, English and American Longitudinal Studies. An additional remit of this research involves exploring mechanistic pathways of cognitive health in relation to the built environment. This is being conducted as part of an ESRC funded Social Behavioural Design Research programme entitled Supportive environments for Physical and social Activity, healthy ageing and CognitivE health (SPACE). Overall, this work will help expand research into the epidemiology of cognitive decline and dementia and will contribute to global harmonisation of cognitive data thus providing new approaches towards prevention and potential treatment of Alzheimer’s disease and related dementias.
In-depth research on various other health domains is ongoing and the identification of biomarkers of ageing continues to be a major avenue of ongoing work with growing partnerships and joint funding. Data linkage and data harmonisation is also a focus of current work (details will follow in a separate manuscript). Further reports from the health assessment, in particular, more in-depth findings from the analysis of the retinal images [39], FFQ [784] and dietary validation study [73], will be forthcoming as well as bespoke reports on other specific age-related topics. Anonymised data from Wave 1 (CAPI, SCQ and Health Assessment), Wave 2 (CAPI, SCQ and COVID questionnaire) are now available for researchers to access. Further information regarding the application process for accessing data (Waves 1 and 2) and/or biological samples (Wave 1) is available on the researcher section of the NICOLA website [1].
Wave 3 of the study will commence later in 2023 which will continue the trajectory of longitudinal data collection and development of this data resource. This third wave will involve a follow-up of the current cohort of NICOLA participants (i.e. those who participated in Wave 2) and will involve a repeat health assessment, CAPI and SCQ with a focus on COVID immune response, microbiome, digital inclusion, food insecurity and eye health. However, for Wave 3, it is our intention to conduct a home based health assessment rather than participants attending the hospital facility. Due to the increasing age of the cohort, home visits have been deemed more acceptable and feasible and will help to reduce the burden on the participant. This awareness has come through informal feedback from current NICOLA participants with many indicating a preference for a home visit. Alternatively, participants will still be given the option of attending a clinicsetting in Wave 3 if they do not wish to have a home visit. Based on Waves 1 and 2, we have also identified a specific need to collect more bespoke data and samples as part of the Wave 3 health assessment including the analysis of the microbiome and COVID antibodies to further enhance the value of the study. The continued focus on COVID into Wave 3 will allow us to uniquely contribute to the path to post COVID recovery and to the rich and developing suite of Longitudinal Population Studies across the UK.
## Conclusion
In summary, this manuscript documents the scientific and methodological processes involved in the development and conduct of the health assessment component of NICOLA Wave 1 and highlights the difference in characteristics of participants taking part. The objective measures of the NICOLA health assessment allow innovative exploration of ageing including greater understanding of the ageing process and its determinants. Data from future waves of NICOLA will further enrich this data resource and will provide information relating to trajectories of health related to ageing.
## Sponsor’s role
The funding sources had no role in the design, methodology, subject recruitment, data collection, analysis or preparation of this manuscript.
## References
1. 1.Northern Ireland Cohort for the Longitudinal Study of Ageing (NICOLA). [https://www.qub.ac.uk/sites/NICOLA/]. Accessed 04 Jan 2023.
2. 2.Neville C, Burns F, Cruise S, Scott A, O’Reilly D, Kee F, Young I. (in press). Cohort profile: the Northern Ireland Cohort for the Longitudinal Study of Ageing (NICOLA). Int J Epid 2023, In press.
3. 3.Neville CE, Cruise SM, Burns F. The Northern Ireland Cohort for the Longitudinal Study of Aging (NICOLA). In: Gu D, Dupre M, editors Encyclopedia of Gerontology and Population Aging 2019.Springer, Cham.
4. Cronin H, O’Regan C, Finucane C, Kearney P, Kenny RA. **Health and Ageing: development of the irish longitudinal study on Ageing Health Assessment**. *J Am Geriatr Soc* (2013.0) **61** 269-S278
5. Steptoe A, Breeze E, Banks J, Nazroo J. **Cohort profile: the English Longitudinal Study of Ageing**. *Int J Epidemiology* (2013.0) **42** 1640-8
6. Khawaja AP, Chan MP, Hayat S, Broadway DC, Luben R, Garway-Heath DF, Sherwin JC, Yip JL, Dalzell N, Wareham NJ, Khaw KT, Foster PJ. **The EPIC-Norfolk Eye Study: rationale, methods and a cross-sectional analysis of visual impairment in a population-based cohort**. *BMJ Open* (2013.0) **3** e002684. PMID: 23516272
7. Chua SYL, Thomas D, Allen N, Lotery A, Desai P, Patel P, Muthy Z, Sudlow C, Peto T, Khaw PT, Foster PJ. **Cohort profile: design and methods in the eye and vision consortium of UK Biobank**. *BMJ Open* (2019.0) **9** e025077. PMID: 30796124
8. Sonnega A, Faul JD, Ofstedal MB, Langa KM, Phillips JW, Weir DR. **Cohort profile: the Health and Retirement Study (HRS)**. *Int J Epidemiol* (2014.0) **43** 576-85. PMID: 24671021
9. Price GM, Uauy R, Breeze E, Bulpitt CJ, Fletcher AE. **Weight, shape, and mortality risk in older persons: elevated waist-hip ratio, not high body mass index, is associated with a greater risk of death**. *Am J Clin Nutr* (2006.0) **84** 449-60. PMID: 16895897
10. Nyunt MSZ, Soh CY, Gao Q, Gwee X, Ling ASL, Lim WS, Lee TS, Yap PLK, Yap KB, Ng TP. **Characterisation of physical frailty and associated physical and functional impairments in mild cognitive impairment**. *Front Med* (2017.0) **4** 230
11. Pavasini R, Guralnik J, Brown JC, di Bari M, Cesari M, Landi F, Vaes B, Legrand D, Verghese J, Wang C, Stenholm S, Ferrucci L, Lai JC, Bartes AA, Espaulella J, Ferrer M, Lim JY, Ensrud KE, Cawthon P, Turusheva A, Frolova E, Rolland Y, Lauwers V, Corsonello A, Kirk GD, Ferrari R, Volpato S, Campo G. **Short physical performance battery and all-cause mortality: systematic review and meta-analysis**. *BMC Med* (2016.0) **14** 215. PMID: 28003033
12. Fried LP, Guralnik JM. **Disability in older adults: evidence regarding significance, etiology, and risk**. *J Am Geriatr Soc* (1997.0) **45** 92-100. PMID: 8994496
13. Mercer VS, Freburger JK, Chang SH, Purser JL. **Step test scores are related to measures of activity and participation in the first 6 months after stroke**. *Phys Ther* (2009.0) **89** 1061-71. PMID: 19661158
14. Mathias S, Nayak US, Isaacs B. **Balance in elderly patients: the “get-up and go” test**. *Arch Phys Med Rehabil* (1986.0) **67** 387-9. PMID: 3487300
15. Podsiadlo D, Richardson S. **The timed “Up & Go”: a test of basic functional mobility for frail elderly persons**. *J Am Geriatr Soc* (1991.0) **39** 142-8. PMID: 1991946
16. Chun S, Shin DW, Han K, Jung JH, Kim B, Jung HW, Son KY, Lee SP, Lee SC. **The timed up and go test and the ageing heart: findings from a national health screening of 1,084,875 community-dwelling older adults**. *Eur J Prev Cardiol* (2021.0) **28** 213-9. PMID: 33838038
17. 17.NICE. Falls – risk assessment. 2019. [https://cks.nice.org.uk/falls-risk-assessment#!scenario]. Accessed 18 Oct 2022.
18. 18.British Geriatric Society. Fit for Frailty [https://www.bgs.org.uk/resources/resource-series/fit-for-frailty]. Accessed 18 Oct 2022.
19. Soysal P, Hurst C, Demurtas J, Firth J, Howden R, Yang L, Tully MA, Koyanagi A, Ilie PC, López-Sánchez GF, Schwingshackl L, Veronese N, Smith L. **Handgrip strength and health outcomes: Umbrella review of systematic reviews with meta-analyses of observational studies**. *J Sport Health Sci* (2021.0) **10** 290-5. PMID: 32565244
20. Simpkin AJ, Cooper R, Howe LD, Relton CL, Davey Smith G, Teschendorff A, Widschwendter M, Wong A, Kuh D, Hardy R. **Are objective measures of physical capability related to accelerated epigenetic age? Findings from a british birth cohort**. *BMJ Open* (2017.0) **7** e016708. PMID: 29092899
21. Abizanda P, Navarro JL, Garcia-Tomas MI, López-Jiménez E, Martínez-Sánchez E, Paterna G. **Validity and usefulness of hand-held dynamometry for measuring muscle strength in community dwelling older persons**. *Arch Gerontol Geriatr* (2012.0) **54** 21-7. PMID: 21371760
22. Stark T, Walker B, Phillips JK, Fejer R, Beck R. **Hand-held dynamometry correlation with the gold standard isokinetic dynamometry: a systematic review**. *PM R* (2011.0) **3** 472-9. PMID: 21570036
23. Arnold CM, Warkentin KD, Chilibeck PD, Magnus CR. **The reliability and validity of handheld dynamometry for the measurement of lower-extremity muscle strength in older adults**. *J Strength Cond Res* (2010.0) **24** 815-24. PMID: 19661831
24. Syddall H, Cooper C, Martin F, Briggs R, Aihie Sayer A. **Is grip strength a useful single marker of frailty?**. *Age Ageing* (2003.0) **32** 650-6. PMID: 14600007
25. 25.Piepoli MF, Hoes AW, Agewall S, Albus C, Brotons C, Catapano AL, Cooney MT, Corrà U, Cosyns B, Deaton C, Graham I, Hall MS, Hobbs FD, Løchen ML, Löllgen H, Marques-Vidal P, Perk J, Prescott E, Redon J, Richter DJ, Sattar N, Smulders Y, Tiberi M, van der Worp HB, van Dis I, Verschuren WM; Additional Contributor: Simone Binno (Italy); Document Reviewers:, De Backer G, Roffi M, Aboyans V, Bachl N, Bueno H, Carerj S, Cho L, Cox J, De Sutter J, Egidi G, Fisher M, Fitzsimons D, Franco OH, Guenoun M, Jennings C, Jug B, Kirchhof P, Kotseva K, Lip GY, Mach F, Mancia G, Bermudo FM, Mezzani A, Niessner A, Ponikowski P, Rauch B, Rydén L, Stauder A, Turc G, Wiklund O, Windecker S, Zamorano JL. 2016 European Guidelines on cardiovascular disease prevention in clinical practice: The Sixth Joint Task Force of the European Society of Cardiology and Other Societies on Cardiovascular Disease Prevention in Clinical Practice (constituted by representatives of 10 societies and by invited experts) Developed with the special contribution of the European Association for Cardiovascular Prevention & Rehabilitation (EACPR). Eur Heart J. 2016, 37(29):2315–2381.
26. 26.Age UK. 2020 [https://www.ageuk.org.uk/information-advice/health-wellbeing/conditions-illnesses/.]. Accessed 14 Sept 2022.
27. Pickering TG, Hall JE, Appel LJ, Falkner BE, Graves JW, Hill MN, Jones DH, Kurtz T, Sheps SG, Roccella EJ. **Recommendations for blood pressure measurement in humans: an AHA Scientific Statement from the Council on high blood pressure Research Professional and Public Education Subcommittee**. *J Clin Hypertens (Greenwich)* (2005.0) **7** 102-9. DOI: 10.1111/j.1524-6175.2005.04377.x
28. Sharma G, Goodwin J. **Effect of ageing on respiratory system physiology and immunology**. *Clin Interv Ageing* (2006.0) **1** 253-60
29. 29.Magnussen C, Ojeda FM, Rzayeva N, Zeller T, Sinning CR, Pfeiffer N et al. FEV1 and FVC predict all-cause mortality independent of cardiac function- results from the population-based Gutenberg Health Study. Int J Cardiol. 2017, 234:64–8. doi: 10.1016/j.ijcard.2017.02.012
30. 30.Duong M, Islam S, Rangarajan S, Leong D, Kurmi O, Teo KK et al. on behalf of the PURE Investigators. Mortality and cardiovascular and respiratory morbidity in individuals with impaired FEV1 (PURE): an international, community-based cohort study. Lancet Glob. Health. 2019, 7:e613–23. doi: 10.1016/S2214-109X(19)30070-1
31. 31.Kinney G, Baker E, Klein OL, Black-Shinn JL, Wan ES, Make B et al. Pulmonary predictors of incident diabetes in smokers. Chr Obstr Pulm Dis. 2016, 3:739–47. doi: 10.15326/jcopdf.3.4.2016.0137
32. 32.Lee I, Leem A, Lee S, Rhee Y, Ha Y, Kim Y. Relationship between pulmonary function and bone mineral density in Korean Health and Nutrition Examination Survey. Korean J Intern Med. 2016, 31:899–909. doi: 10.3904/kjim.2015.127
33. 33.Navaneethan S, Mandayam S, Arrigan S, Winkelmayer W, Schold J. Obstructive and restrictive lung function measures and CKD: National Health and Nutrition Examination Survey (NHANES) 2007–2012. Am J Kidney Dis. 2016, 68:414–21. doi: 10.1053/j.ajkd.2016.03.415
34. 34.Russ T, Starr J, Stamatakis E, Kivimaki M, Batty G. Pulmonary function as a risk factor for dementia death: an individual participants meta-analysis of six UK general population cohort studies. J Epidemiol Community Health. 2015, 69:550–6. doi: 10.1136/jech-2014-204959
35. 35.World Health Organisation (WHO). Blindness and vision impairment prevention. Priority eye diseases;. 2020 [https://www.who.int/blindness/causes/priority/en/index7.html]. Accessed 04 Sept 2022.
36. Salive ME, Guralnik J, Glynn RJ, Christen W, Wallace RB, Ostfeld AM. **Association of visual impairment with mobility and physical function**. *J Am Geriatr Soc* (1994.0) **42** 287-92. PMID: 8120313
37. **Early photocoagulation for diabetic retinopathy. ETDRS report number 9. Early Treatment Diabetic Retinopathy Study Research Group**. *Ophthalmology* (1991.0) **98** 766-85. PMID: 2062512
38. 38.Tiraset N, Poonyathalang A, Padungkiatsagul T, Deeyai M, Vichitkunakorn P, Vanikieti K. Comparison of Visual Acuity Measurement Using Three Methods: Standard ETDRS Chart, Near Chart and a Smartphone-Based Eye Chart Application. Clin Ophthalmol 2021, 26(15):859–869. doi: 10.2147/OPTH.S304272. PMID: 33664563.
39. 39.Hogg RE, Wright DM, Quinn NB, Muldrew KA, Hamill B, Smyth L, McKnight AJ, Woodside J, Tully MA, Cruise S, McGuinness B, Young IS, Kee F, Peto T, Chakravarthy U. Prevalence and risk factors for age-related macular degeneration in a population-based cohort study of older adults in Northern Ireland using multimodal imaging: NICOLA Study. Br J Ophthalmol. 2022 Oct;10:bjophthalmol–2021. 10.1136/bjo-2021-320469. Epub ahead of print.
40. 40.Network of Ophthalmic. Reading Centres UK [www.networcuk.com]. Last accessed 10/02/2023.
41. Quinn NB, Steel DH, Chakravarthy U, Peto T, Hamill B, Muldrew A, Graham K, Elliott D, Hennessy R, Cruise S, McGuinness B, Young IS, Kee F, Hogg RE. **Assessment of the vitreomacular interface using high-resolution OCT in a population-based cohort study of older adults**. *Ophthalmol Retina* (2020.0) **4** 801-13. PMID: 32335034
42. Ferris FL, Wilkinson CP, Bird A, Chakravarthy U, Chew E, Csaky K, Sadda SR. **Beckman Initiative for Macular Research classification Committee. Clinical classification of age-related macular degeneration**. *Ophthalmology* (2013.0) **120** 844-51. PMID: 23332590
43. 43.Public Health England. NHS Diabetic Eye Screening Programme: grading definitions for referable disease. August 2021. https://www.gov.uk/government/publications/diabetic-eye-screening-retinal-image-grading-criteria/nhs-diabetic-eye-screening-programme-grading-definitions-for-referable-disease. Last accessed 10 Jan 2023.
44. McCann P, Hogg R, Wright DM, Pose-Bazarra S, Chakravarthy U, Peto T, Cruise S, McGuinness B, Young IS, Kee F, Azuara-Blanco A. **Glaucoma in the Northern Ireland Cohort for the Longitudinal Study of Ageing (NICOLA): cohort profile, prevalence, awareness and associations**. *Br J Ophthalmol* (2020.0) **104** 1492-9. PMID: 32034006
45. McCann P, Hogg R, Wright DM, McCann P, Hogg R, Wright DM, Chakravarthy U, Peto T, Cruise S, McGuinness B, Young IS, Kee F, Azuara-Blanco A. **Intraocular pressure and circumpapillary retinal nerve fibre layer thickness in the Northern Ireland Cohort for the Longitudinal Study of Ageing (NICOLA): distributions and associations**. *Br J Ophthalmol* (2021.0) **105** 948-56. PMID: 32732345
46. Roth TN, Hanebuth D, Probst R. **Prevalence of age-related hearing loss in Europe: a review**. *Eur Arch Otorhinolaryngol* (2011.0) **268** 1101-7. PMID: 21499871
47. Tseng YC, Liu SH, Lou MF, Huang GS. **Quality of life in older adults with sensory impairments: a systematic review**. *Qual Life Res* (2018.0) **27** 1957-71. PMID: 29404924
48. Gopinath B, Hickson L, Schneider J, McMahon CM, Burlutsky G, Leeder SR, Mitchell P. **Hearing-impaired adults are at increased risk of experiencing emotional distress and social engagement restrictions five years later**. *Age Ageing* (2012.0) **41** 618-23. PMID: 22591986
49. Pronk M, Deeg DJH, Smits C, van Tilburg TG, Kuik DJ, Festen JM, Kramer SE. **Prospective effects of hearing status on loneliness and depression in older persons: identification of subgroups**. *Int J Audiol* (2011.0) **50** 887-96. PMID: 21929374
50. Monzani D, Galeazzi GM, Genovese E, Marrara A, Martini A. **Psychological profile and social behaviour of working adults with mild or moderate hearing loss**. *Acta Otorhinolaryngol Ital* (2008.0) **28** 61-6. PMID: 18669069
51. Li CM, Zhang X, Hoffman HJ, Cotch MF, Themann CL, Wilson MR. **Hearing impairment associated with depression in US adults, National Health and Nutrition Examination Survey 2005–2010**. *JAMA Otolaryngol Head Neck Surg* (2014.0) **140** 293-302. PMID: 24604103
52. 52.Canney M, McNicholas T, Scarlett S, Briggs B. Prevalence and Impact of Chronic Debilitating Disorders. In: Health and Wellbeing: Active Ageing for Older Adults in Ireland. Evidence from the Irish Longitudinal Study on Ageing. 2017 [https://tilda.tcd.ie/publications/reports/pdf/w3-key-findings-report/Chapter%207.pdf]. Last accessed 10 Feb 2023.
53. Viljanen A, Kaprio J, Pyykkö I, Sorri M, Koskenvuo M, Rantanen T. **Hearing acuity as a predictor of walking difficulties in older women**. *J Am Geriatr Soc* (2009.0) **57** 2282-6. PMID: 19874410
54. Dawes P, Emsley R, Cruickshanks KJ, Moore DR, Fortnum H, Edmondson-Jones M, McCormack A, Munro KJ. **Hearing loss and cognition: the role of hearing AIDS, social isolation and depression**. *PLoS ONE* (2015.0) **10** e0119616. PMID: 25760329
55. Moore DR, Edmondson-Jones M, Dawes P, Fortnum H, McCormack A, Pierzycki RH, Munro KJ. **Relation between speech-in-noise threshold, hearing loss and cognition from 40–69 years of age**. *PLoS ONE* (2014.0) **9** 1-10
56. 56.Lin FR, Yaffe K, Xia J, Xue QL, Harris TB, Purchase-Helzner E, Satterfield S, Ayonayon HN, Ferrucci L, Simonsick EM, Health ABC Study Group. ;. Hearing loss and cognitive decline in older adults. JAMA Intern Med 2013, 173(4):293–299.
57. 57.Sindhusake D, Mitchell P, Smith W, Golding M, Newall P, Hartley D, Rubin G. Validation of self-reported hearing loss. The Blue Mountains Hearing Study. Int J Epidemiol 2001, 30(6) 1371–8.
58. 58.Gallagher NE. Diet, hearing loss and healthy ageing. (Doctoral dissertation, Queen’s University Belfast, Faculty of Medicine, Health and Life Sciences, July 2017).
59. Prince M, Bryce R, Albanese E, Wimo A, Ribeiro W, Ferri CP. **The global prevalence of dementia: a systematic review and metaanalysis**. *Alzheimers Dement* (2013.0) **9** 63-75. PMID: 23305823
60. 60.Psychological Assessment Resources., Inc. (PAR), 16204 North Florida Avenue, Lutz, Florida 33549, from the Mini Mental State Examination, by Marshal Folstein and Susan Folstein, Copyright 1975, 1998, 2001 by Mini Mental LLC, Inc. Published 2001 by PAR. [[email protected]].
61. Nasreddine ZS, Phillips NA, Bedirian V, Charbonneau S, Whitehead V, Collin I, Cummings JL, Chertkow H. **The Montreal Cognitive Assessment, MoCA: a brief screening tool for mild cognitive impairment**. *J Am Geriatr Soc* (2005.0) **53** 695-9. PMID: 15817019
62. Pinto TCC, Machado L, Bulgacov TM, Rodrigues-Júnior AL, Costa MLG, Ximenes RCC, Sougey EB. **Is the Montreal Cognitive Assessment (MoCA) screening superior to the Mini-Mental State Examination (MMSE) in the detection of mild cognitive impairment (MCI) and Alzheimer’s Disease (AD) in the elderly?**. *Int Psychogeriatr* (2019.0) **31** 491-504. PMID: 30426911
63. Radanovic M, Diniz BS, Mirandez RM, Novaretti TMDS, Flacks MK, Yassuda MS, Forlenza OV. **Verbal fluency in the detection of mild cognitive impairment and Alzheimer’s disease among brazilian portuguese speakers: the influence of education**. *Int Psychogeriatr* (2009.0) **21** 1081-7. PMID: 19619390
64. Reas ET, Laughlin GA, Bergstrom J, Kritz-Silverstein D, Barrett-Connor E, McEvoy LK. **Effects of sex and education on cognitive change over a 27-year period in older adults: the Rancho Bernardo Study**. *Am J Geriatr Psychiatry* (2017.0) **25** 889-99. PMID: 28433548
65. Tennant R, Hiller L, Fishwick R, Platt S, Joseph S, Weich S, Parkinson J, Secker J, Stewart-Brown S. **The Warwick-Edinburgh mental wellbeing scale (WEMWBS): development and UK validation**. *Health Qual Life Outcomes* (2007.0) **5** 63. PMID: 18042300
66. Lloyd K, Devine P. **Psychometric Properties of the Warwick–Edinburgh Mental Wellbeing scale (WEMWBS) in Northern Ireland**. *J Ment Health* (2012.0) **21** 257-63. PMID: 22574955
67. Radloff LS. **The CES-D scale: a self-report depression scale for research in the general population**. *Appl Psychol Meas* (1977.0) **1** 385-401
68. Lewinsohn PM, Seeley JR, Roberts RE, Allen NB. **Center for Epidemiological Studies-Depression Scale (CES-D) as a screening instrument for depression among community-residing older adults**. *Psychol Ageing* (1997.0) **12** 277-87
69. McGrattan AM, McGuinness B, McKinley MC, Kee F, Passmore P, Woodside JV, McEvoy CT. **Diet and inflammation in cognitive ageing and Alzheimer’s Disease**. *Curr Nutr Rep* (2019.0) **8** 53-65. DOI: 10.1007/s13668-019-0271-4
70. Yeung SSY, Kwan M, Woo J. **Healthy Diet for healthy ageing**. *Nutrients* (2021.0) **13** 4310. DOI: 10.3390/nu13124310
71. Wickramasinghe K, Mathers JC, Wopereis S, Marsman DS, Griffiths JC. **From lifespan to healthspan: the role of nutrition in healthy ageing**. *J Nutr Sci* (2020.0) **9** e33. DOI: 10.1017/jns.2020.26
72. 72.World Health Organization. Ageing and Health, 1. October 2022. https://www.who.int/news-room/fact-sheets/detail/ageing-and-health. Last accessed 10/02/2023.
73. Dean M, Raats M, Grunert K, Lumbers M. **Factors influencing eating a varied diet in old age**. *Public Health Nutr* (2009.0) **12** 2421-7. DOI: 10.1017/S1368980009005448
74. Ni Lochlainn M, Cox NJ, Wilson T, Hayhoe RPG, Ramsay SE, Granic A, Isanejad M, Roberts HC, Wilson D, Welch C, Hurst C, Atkins JL, Mendonça N, Horner K, Tuttiett ER, Morgan Y, Heslop P, Williams EA, Steves CJ, Greig C, Draper J, Corish CA, Welch A, Witham MD, Sayer AA, Robinson S. **Nutrition and Frailty: Opportunities for Prevention and Treatment**. *Nutrients* (2021.0) **13** 2349. DOI: 10.3390/nu13072349
75. 75.Roberts SB, Silver RE, Das SK, Fielding RA, Gilhooly CH, Jacques PF, Kelly JM, Mason JB, McKeown NM, Reardon MA, Rowan S, Saltzman E, Shukitt-Hale B, Smith CE, Taylor AA, Wu D, Zhang FF, Panetta K, Booth S. Healthy Ageing-Nutrition Matters: Start Early and Screen Often. Adv Nutr 2021, 12(4):1438–1448. doi: 10.1093/advances/nmab032. Erratum in: Adv Nutr. 2021 Jul 30;12(4):1597–1598. PMID: 33838032; PMCID: PMC8994693.
76. Landi F, Calvani R, Tosato M, Martone AM, Ortolani E. **Anorexia of ageing: risk factors, consequences, and potential treatments**. *Nutrients* (2016.0) **8** 69. PMID: 26828516
77. Shim JS, Oh K, Kim HC. **Dietary assessment methods in epidemiologic studies**. *Epidemiol Health* (2014.0) **36** e2014009. PMID: 25078382
78. 78.. EPIC Norfolk Food Frequency Questionnaire. [https://www.epic-norfolk.org.uk/wp-content/uploads/2020/11/CAMB-PQ-6-1205a_front.pdf]. Last accessed 07 Oct 2022.
79. Neville C, McKinley M, Kee F, Young I, Cardwell C, Woodside J. **Relative validity of a food frequency questionnaire to assess fruit and vegetable intake in healthy older adults**. *Proc Nutr Soc* (2020.0) **79** E513
80. 80.Lesko CR, Fox MP, Edwards JK. A framework for descriptive epidemiology.Am J Epidemiol 2022 Jul1:kwac115. doi: 10.1093/aje/kwac115. Epub ahead of print.
81. 81.Northern Ireland Statistics and Research AgencyNorthern Ireland multiple deprivation measure2010BelfastNISRA. *Northern Ireland multiple deprivation measure* (2010.0)
82. 82.Stanojevic S, Kaminsky DA, Miller MR, Thompson B, Aliverti A, Barjaktarevic I, Cooper BG, Culver B, Derom E, Hall GL, Hallstrand TS, Leuppi JD, MacIntyre N, McCormack M, Rosenfeld M, Swenson ER. ERS/ATS technical standard on interpretive strategies for routine lung function tests. Eur Respir J 2022, 60(1):2101499. doi: 10.1183/13993003.01499-2021. PMID: 34949706.
83. Dodds RM, Syddall HE, Cooper R, Benzeval M, Deary IJ. **Grip strength across the life course: normative data from twelve british studies**. *PLoS ONE* (2014.0) **9** e113637. PMID: 25474696
84. 84.Nolan J, Newman L, Donoghue O. Objective indicators of health and function. In: Health and Wellbeing: Active Ageing for Older Adults in Ireland. Evidence from the Irish Longitudinal Study on Ageing. 2017, [https://tilda.tcd.ie/publications/reports/pdf/w3-key-findings-report/Chapter%205.pdf ]. Last accessed 10 Feb 2023.
85. Tham YC, Li X, Wong TY, Quigley HA, Aung T. **Global prevalence of glaucoma and projections of glaucoma burden through 2040: a systematic review and meta-analysis**. *Ophthalmology* (2014.0) **121** 2081-90. PMID: 24974815
86. **Treatment of high blood cholesterol in A. Third Report of the national cholesterol Education Program (NCEP) Expert Panel on detection, evaluation, and treatment of high blood cholesterol in adults (Adult Treatment Panel III) final report**. *Circulation* (2002.0) **106** 3143-421. PMID: 12485966
87. Kelly MP, Stewart E, Morgan A, Killoran A, Fischer A, Threlfall A, Bonnefoy J. **A conceptual framework for public health: NICE’s emerging approach**. *Public Health* (2009.0) **123** e14-e20. PMID: 19100588
88. Edwards N, Plotnikoff RC. **Mobilizing an underused resource: cohort studies for population health intervention research**. *Int J Epidemiol* (2018.0) **47** 1730-3. DOI: 10.1093/ije/dyy191
89. 89.MRC Population Health Sciences GroupMaximising the value of UK Population cohorts: MRC Strategic Review of the largest UK Population Cohort Studies2014London/Swindon, UKMedical Research Council (MRC). *Maximising the value of UK Population cohorts: MRC Strategic Review of the largest UK Population Cohort Studies* (2014.0)
90. Chatfield MD, Brayne CE, Matthews FE. **A systematic literature review of attrition between waves in longitudinal studies in the elderly shows a consistent patterns of dropout between differing studies**. *J Clin Epidemiol* (2005.0) **58** 13-9. PMID: 15649666
91. 91.Cañadas-Garre M, Smyth LJ, Neville CE, Kee F, Woodside JV, McKnight AJ. Chapter 8, Molecular Biomarkers. In: NICOLA Health Assessment Report 2021 [https://www.qub.ac.uk/sites/NICOLA/FileStore/Filetoupload,1287506,en.pdf ]
92. Paterson EN, Maxwell AP, Kee F, Cruise S, Young IS, McGuinness B, McKay GJ. **Association of renal impairment with cognitive dysfunction in the Northern Ireland Cohort for the Longitudinal Study of Ageing (NICOLA)**. *Nephrol Dial Transplant* (2021.0) **36** 1492-9. PMID: 34038557
93. Paterson EN, Neville CE, Wallace SM, Woodside JV, Kee F, Young IS, Cruise S, McGuinness B, Maxwell AP, McKay GJ. **Dietary patterns associated with renal impairment in the Northern Ireland Cohort for the Longitudinal Study of Ageing (NICOLA)**. *Eur J Nutr* (2021.0) **60** 4045-54. PMID: 33959803
94. O’Neill RA, Maxwell AP, Kee F, Young I, Hogg RE, Cruise S, McGuinness B, McKay GJ. **Association of reduced retinal arteriolar tortuosity with depression in older participants from the Northern Ireland Cohort for the Longitudinal Study of Ageing**. *BMC Geriatr* (2021.0) **21** 62. PMID: 33446119
95. 95.Smyth LJ, Cruise SM, Tang J, Young I, McGuinness B, Kee F, McKnight AJ. An investigation into DNA methylation patterns associated with risk preference in older individuals.Epigenetics 2021, Oct30:1–14. doi: 10.1080/15592294.2021.1992910. Epub ahead of print.
|
---
title: Corticosteroids Increase the Risk of Invasive Fungal Infections More Than Tumor
Necrosis Factor-Alpha Inhibitors in Patients With Inflammatory Bowel Disease
authors:
- Martin H Gregory
- Andrej Spec
- Dustin Stwalley
- Anas Gremida
- Carlos Mejia-Chew
- Katelin B Nickel
- Matthew A Ciorba
- Richard P Rood
- Margaret A Olsen
- Parakkal Deepak
journal: Crohn's & Colitis 360
year: 2023
pmcid: PMC9999356
doi: 10.1093/crocol/otad010
license: CC BY 4.0
---
# Corticosteroids Increase the Risk of Invasive Fungal Infections More Than Tumor Necrosis Factor-Alpha Inhibitors in Patients With Inflammatory Bowel Disease
## Abstract
### Background
Invasive fungal infections are a devastating complication of inflammatory bowel disease (IBD) treatment. We aimed to determine the incidence of fungal infections in IBD patients and examine the risk with tumor necrosis factor-alpha inhibitors (anti-TNF) compared with corticosteroids.
### Methods
In a retrospective cohort study using the IBM MarketScan Commercial Database we identified US patients with IBD and at least 6 months enrollment from 2006 to 2018. The primary outcome was a composite of invasive fungal infections, identified by ICD-$\frac{9}{10}$-CM codes plus antifungal treatment. Tuberculosis (TB) infections were a secondary outcome, with infections presented as cases/100 000 person-years (PY). A proportional hazards model was used to determine the association of IBD medications (as time-dependent variables) and invasive fungal infections, controlling for comorbidities and IBD severity.
### Results
Among 652 920 patients with IBD, the rate of invasive fungal infections was 47.9 cases per 100 000 PY ($95\%$ CI 44.7–51.4), which was more than double the TB rate (22 cases [CI 20–24], per 100 000 PY). Histoplasmosis was the most common invasive fungal infection (12.0 cases [CI 10.4–13.8] per 100 000 PY). After controlling for comorbidities and IBD severity, corticosteroids (hazard ratio [HR] 5.4; CI 4.6–6.2) and anti-TNFs (HR 1.6; CI 1.3–2.1) were associated with invasive fungal infections.
### Conclusions
Invasive fungal infections are more common than TB in patients with IBD. The risk of invasive fungal infections with corticosteroids is more than double that of anti-TNFs. Minimizing corticosteroid use in IBD patients may decrease the risk of fungal infections.
## Introduction
Inflammatory bowel disease (IBD) is characterized by pathologic inflammation in the gastrointestinal tract, so logically immunosuppression is the primary treatment modality.1 *Tumor necrosis* factor-alpha inhibitors (anti-TNFs) are more selective immunosuppressants than corticosteroids and are associated with a lower risk of bacterial infection.2,3 However, the anti-TNF mechanism of action may make patients especially susceptible to invasive fungal infections, a potentially devastating complications of IBD treatment.4 TNF is essential for granuloma formation, which is needed for defending against fungal infections and tuberculosis (TB).4 Reactivation of latent TB is a well-described side effect of anti-TNF therapy.5 In addition, reports of invasive fungal infections, when anti-TNFs became available, led to a black box warning for anti-TNFs.4 However, there have been no large studies examining the risk of specific invasive fungal infections in patients with IBD, particularly among those treated with anti-TNFs compared with corticosteroids. With the rise in the incidence of IBD and increasing use of anti-TNFs,6 clarifying the risks of complications with different IBD treatments is essential to appropriately weigh the risks of treatment for individual patients.
Our aims were to identify how common invasive fungal infections are among IBD patients in the United States and to identify risk factors for invasive fungal infections. For reference, we compared the incidence to both latent TB and TB disease, which all IBD patients should be screened for prior to initiating biologic therapy. We hypothesized that invasive fungal infections are much more common than TB in IBD patients in the United States. Our second aim was to identify risk factors for invasive fungal infections, specifically focused on IBD medications. Despite a mechanism of action that may predispose patients on anti-TNFs to fungal infections, we hypothesized that corticosteroids are associated with higher risk of invasive fungal infections than anti-TNFs.
## Materials and Methods
This was a retrospective cohort study using the IBM MarketScan Commercial Database from 2006 to 2018. This database includes inpatient, outpatient, and pharmacy claims from participating private insurance health plans and large employers throughout the United States. It contains data from more than 150 million employees and their dependents. It includes a variety of payment models, including fee-for-service, preferred provider organizations and capitated health plans.
We used a previously validated algorithm designed to identify patients with IBD by International Classification of Diseases (ICD) 9 coding.7 Patients were required to have at least 2 ICD-$\frac{9}{10}$-CM diagnosis codes for Crohn’s disease (CD) or ulcerative colitis (UC), with at least 1 coded on an outpatient visit. Patients were required to have at least 6 months of health insurance enrollment before infection to identify comorbidities. We did not exclude patients by age, but since the IBM MarketScan Commercial Database does not include Medicare plans, all patients were younger than 65 years.
The primary outcome was a composite of invasive fungal infections, including invasive candidiasis, histoplasmosis, coccidiomycosis, aspergillus, cryptococcus, Pneumocystis jiroveci, blastomycosis, mucormycosis, paracoccidiomycosis, fungal pneumonia, and fungal meningitis. Since there has been no validated definition for fungal infection using administrative data, we required all patients coded for one of the invasive fungal infections to fill a prescription for an antifungal medication within 30 days of a fungal diagnosis code to reduce the likelihood of coding error (eg, history of previously treated infection, unconfirmed infection). Diagnosis codes for P. jiroveci were required to have a prescription for one of the following within 30 days (trimethoprim–sulfamethoxazole, atovaquone, clindamycin and primaquine, trimethoprim and dapsone, or pentamidine). Only the first instance of fungal infection was counted to avoid identifying subsequent coding for the same infection as a new infection (see Supplementary Table S1 for list of diagnosis and procedure codes used). Since esophageal candidiasis is less severe than invasive candida infections, it was not included in the primary analysis.
TB infection was a secondary outcome. We defined TB disease as an ICD-$\frac{9}{10}$-CM diagnosis code for active TB and prescriptions for isoniazid plus 2 other medications (one of rifampin, rifapentine, or rifabutin, and one of ethambutol, pyrazinamide, or moxifloxacin) within 30 days. Latent TB was defined as an ICD-$\frac{9}{10}$-CM diagnosis code for a positive TB skin or blood test with a prescription for isoniazid or rifampin alone for at least 90 days. Patients with a code for TB disease and a prescription for only isoniazid or rifampin were considered as latent TB. Additional secondary outcomes were individual fungal infections.
Baseline comorbidities, IBD medications, emergency department visits, hospitalizations, IBD-related surgeries, and opioid prescriptions were captured in the first 6 months of enrollment (see Supplementary Table S1 for list of diagnosis and procedure codes used). The severity of CD was captured using codes for complicated CD, either CD-related fistula or stricture during follow-up. For UC, we captured disease location (proctitis, left-sided colitis, and pancolitis).
## Statistical Analysis
The incidence rate for the primary outcome was calculated by counting the number of persons with invasive fungal infections divided by the sum of patient follow-up time, expressed as a rate per 100 000 person-years (PY). CIs were estimated with Poisson regression. The incidence rate of TB disease and latent TB infection was calculated similarly.
A Cox proportional hazards model using time-dependent variables for IBD medications was used to determine the association of IBD medications with invasive fungal infection. Patients were censored when they lost health insurance coverage, died, or at the end of the study period ($\frac{12}{31}$/2018). Follow-up time for each patient was divided into 90-day intervals. We determined whether a patient had a prescription for an IBD medication during each interval. Thus, the association between an IBD medication and a fungal infection was only assessed while the patient was on that medication. We chose a 90-day interval because several medications of interest are administered every 60 days, and a 90-day window would account for scheduling delays (eg, insurance authorization). We tested 30- and 60-day treatment intervals in sensitivity analyses. The primary medications of interest were anti-TNFs (adalimumab, infliximab, certolizumab pegol, and golimumab) and oral corticosteroids. In a separate analysis we examined immunomodulators (azathioprine and methotrexate) as time-dependent variables, both alone and in combination with anti-TNFs and corticosteroids. For each 90-day interval, we determined if a patient had a prescription for a corticosteroid, an immunomodulator or an anti-TNF and whether an invasive fungal infection occurred. Mutually exclusive variables were created for each combination of the 3 medications. Using no medication as the reference, we assessed the hazard ratio (HR) for time to invasive fungal infection for each combination of the time-dependent variables. Since vedolizumab ($\frac{5}{20}$/2014) and ustekinumab ($\frac{9}{23}$/2016) were approved later in the follow-up period, there were too few fungal infections to assess the risk with these medications. Since there is evidence that even low-dose or short-term corticosteroid use is associated with increased risk of infection, we included any oral steroid prescription (prednisone, dexamethasone, prednisolone, and budesonide) in the 90-day treatment period as being on corticosteroids.8,9 In sensitivity analysis we included only moderate or high-dose steroids, defined as 20 mg of prednisone or more for at least 14 days. For the sensitivity analysis, budesonide was not included as steroid treatment. Since esophageal candidiasis is less severe than other invasive candida infections, we excluded esophageal candidiasis from the primary analysis but did a separate sensitivity analysis including it in the primary fungal infection outcome.
In the multivariate Cox proportional hazards model, we controlled for IBD disease severity by including variables found to be associated with IBD-related hospitalization in administrative research.10 This publication used separate models for CD and UC. Since we combined UC and CD, we included variables that were found to be associated with IBD-related hospitalization in each of the models, including age, sex, anemia, opioid prescription, emergency room visit, hospitalization, or IBD-related surgery in the baseline period (see Supplementary Table S1 for list of diagnosis and procedure codes used).10 For CD severity we included fistulizing or stricturing CD and for UC we included disease location (pancolitis, left-sided colitis, and proctitis). Prescription for total parenteral nutrition (TPN) was included as a time-dependent variable in the model as TPN is a risk factor for fungemia.11 We assessed whether a patient had a procedure code for TPN during each 90-day interval (see Supplementary Table S1 for list of diagnosis and procedure codes used). We also controlled for the presence of diabetes mellitus, leukemia, and lymphoma during the baseline period, as these are known risk factors for fungal infection.12,13 SAS version 9.4 was used for all statistical analyses.
## Ethical Considerations
The Washington University Human Resources Protection Office exempted this study from oversight by the Institutional Review Board.
## Results
We identified 652 920 patients with IBD, including 353 165 with UC and 291 506 with CD (Table 1). There were 8225 patients with codes for both UC and Crohn’s. The median follow-up was 1.6 years (interquartile range: 2.8 years). Among Crohn’s patients, 13 075 ($4.5\%$) had stricturing disease and 5817 ($2.0\%$) had a fistula. The mean age was 42.2 years (range 0–65). The population was $54.6\%$ female.
**Table 1.**
| Characteristic | N (%) |
| --- | --- |
| Total IBD | 652 920 (100) |
| Ulcerative colitis | 353 165 (54.09) |
| Pancolitis | 274 190 (77.64) |
| Left-sided | 37 346 (5.72%) |
| Proctitis | 41 629 (6.38%) |
| Crohn’s disease | 291 506 (44.65) |
| Stricturing | 13 075 (4.49) |
| Penetrating | 5817 (2.00) |
| IBD-unclassified | 8225 (1.26) |
| Medications | Medications |
| Aminosalicylates | 304 722 (46.67) |
| Immunomodulatora | 116 400 (17.83) |
| Corticosteroids | Corticosteroids |
| Systemic | 313 781 (48.06) |
| Budesonide | 70 895 (10.86) |
| Anti-TNF | 103 755 (15.89) |
| Ustekinumab | 3597 (0.55) |
| Vedolizumab | 8149 (1.25) |
| Female | 356 769 (54.64) |
| Age | Age |
| <18 | 35 249 (5.40) |
| 18–24 | 56 640 (8.67) |
| 25–29 | 51 965 (7.96) |
| 30–34 | 57 837 (8.86) |
| 35–39 | 62 340 (9.55) |
| 40–44 | 67 545 (10.35) |
| 45–49 | 74 526 (11.41) |
| 50–54 | 88 082 (13.49) |
| 55–59 | 83 774 (12.83) |
| 60+ | 74 962 (11.48) |
| Comorbidities | Comorbidities |
| Diabetes mellitus | 23 737 (3.64) |
| Chronic kidney disease | 2748 (0.42) |
| Hypertension | 45 635 (6.99) |
| Anemia | 22 779 (3.49) |
| Opioid prescription | 140 496 (21.52) |
| Leukemia | 485 (0.07) |
| Lymphoma | 970 (0.15) |
| ED visit, hospitalization, or IBD surgeryb | 178 660 (27.36) |
| Total parenteral nutrition | 3727 (0.57) |
| Follow-up in years after first code for IBD | Follow-up in years after first code for IBD |
| Mean | 2.48 |
| Minimum | 0.003 |
| 25th percentile | 0.64 |
| Median | 1.59 |
| 75th percentile | 3.47 |
| Maximum | 12.50 |
Almost half ($48.1\%$) of patients had a prescription for oral corticosteroids during follow-up. There were 103 755 patients ($15.89\%$) who received a prescription for an anti-TNF at some point during follow-up. An immunomodulator was prescribed in 116 400 ($17.8\%$) of patients. Ustekinumab and vedolizumab were prescribed in 3597 ($0.55\%$) and 8149 ($1.25\%$), respectively. There were 3727 ($0.6\%$) patients who received TPN during follow-up.
There were 775 fungal infections during 1 616 941 years of follow-up for a rate of 47.9 per 100 000 PY ($95\%$ CI 44.7–51.4) (Figure 1 and Table 2). Histoplasmosis was the most common invasive fungal infection (12 per 100 000 PY [$95\%$ CI 10.4–13.8]), followed by candidiasis (9.3 per 100 000 PY [$95\%$ CI 7.9–10.9]) and coccidiomycosis (8.8 per 100 000 PY [$95\%$ CI 7.4–10.3] (Table 2) Other fungal infections were rare. There were 305 cases of latent TB (18.5 cases per 100 000 PY [$95\%$ CI 16.6–20.8]) and 60 cases of TB disease (3.5 cases per 100 000 PY [$95\%$ CI 2.7–4.6]) (Figure 1).
After controlling for comorbidities, IBD severity and TPN use, both corticosteroids (HR 5.4, CI 4.6–6.2) and anti-TNF use (HR 1.6, CI 1.3–2.1) were associated with increased risk of invasive fungal infections (Figure 2 and Supplementary Table S2). TPN was associated with a 16-fold increase in the risk of invasive fungal infection (CI 11.4–23.4). Anemia, diabetes mellitus, leukemia, and lymphoma were also associated with significantly increased risk of fungal infection. In sensitivity analysis, using 30- or 60-day intervals for medication exposure instead of 90-day intervals, had minimal effect on the HRs for anti-TNFs and corticosteroids (see Supplementary Tables S3 and S4). In an additional sensitivity analysis limiting corticosteroid exposure to only moderate or high dose corticosteroids (see definition in methods), the results were similar (see Supplementary Table S5). Results were similar if esophageal candidiasis was included in the primary outcome (see Supplementary Table S6). Immunomodulators alone were not associated with an increased risk of invasive fungal infection (HR 1.1, CI 0.8–1.6) and the risk of invasive fungal infection with immunomodulator and anti-TNF combination therapy (HR 1.6, CI 0.7–3.7) was similar to anti-TNF monotherapy (HR 2.5, CI 1.7–3.7) (Table 3). The addition of corticosteroids to any regimen significantly increased the risk of invasive fungal infection. For example, the addition of corticosteroids to anti-TNF increased the HR from 2.5 (CI 1.7–3.7) to 8.1 (CI 5.7–11.6).
## Discussion
In this study, invasive fungal infections were 2 times more common than TB in patients with IBD. While both corticosteroids and anti-TNFs were associated with increased risk of invasive fungal infection, the risk with corticosteroids was more than 3-fold higher than with anti-TNFs alone or in combination with an immunomodulator. Additionally, TPN was associated with an 16-fold increased risk of invasive fungal infection.
To our knowledge, no large studies have examined the risk of fungal infections specifically in IBD patients. A smaller prospective cohort study of patients on anti-TNFs for multiple indications in France found an age- and sex-adjusted incidence rate for all non-TB opportunistic infections of 151.6 per 100 000 PY.14 Similar to our study, both anti-TNFs and corticosteroids were associated with increased risk of opportunistic infection.14 Another study of non-viral opportunistic infections found patients on anti-TNFs had an infection rate of 270 per 100 000 PY.15 This rate was higher than in our study but only $40\%$ of the opportunistic infections were fungal, and the study included older patients and those on anti-TNFs for other immune-mediated inflammatory diseases.15 Similar to our study, both anti-TNF and corticosteroids were associated with increased risk of opportunistic infection, with corticosteroids associated with higher risk.15 Anti-TNFs have a black box warning for invasive fungal infection based on case reports.16 Indeed, we found that IBD patients on anti-TNFs were at increased risk for invasive fungal infections. However, in our study corticosteroid exposure was associated with a much greater risk of fungal infection than anti-TNF. Thus, our results suggest anti-TNFs should be preferred over continued exposure to corticosteroids, with respect to the risk of fungal infections. It is possible that ustekinumab or vedolizumab may be a good alternative for IBD patients at a high risk for invasive fungal infection, such as those on TPN or hematologic malignancy. Unfortunately, since these agents were approved at the end of our follow-up period, we had too few patients on these therapies to assess the risk of invasive fungal infections with these newer therapies. Future studies should examine whether these agents are associated with a lower risk of invasive fungal infections than anti-TNFs.
Given the rarity of TB, it may be reasonable to forego annual TB testing in patients on biologics who have previously tested negative and do not have new risk factors, as the American Rheumatology Society has recommended.17 Indeed, a recent study of repeat testing for latent TB in IBD patients showed the few new abnormal results were false positives.18 Another recent study found that indeterminate results were associated with delays in treatment and subsequent hospitalization.19 None of the patients with indeterminate results were diagnosed with latent TB.19 Professional GI societies in the United States might consider updating guidance of latent TB screening in IBD patients who previously tested negative and without new risk factors for exposure to avoid therapy interruption.
The major strength of our study is the use of a large dataset that captures patients in both academic and community settings, including information from inpatient and outpatient encounters as well as pharmacy data. One advantage of this claims data is the ability to capture complications even for patients readmitted to a different health system.
We attempted to limit the risk of miscoding by using a validated algorithm for IBD and requiring an antifungal prescription with a diagnosis code for fungal infection. Our dataset only included patients with commercial insurance in the United States, so the results may not be generalizable to patients with Medicare or Medicaid, the elderly, those without insurance, and to patients with IBD in non-US geographical locations where TB is endemic. While we are confident that medications administered as infusions were received because they have an associated Healthcare Common Procedure Coding System code, we cannot be certain that a patient was compliant in their usage of an injectable or oral medication prescribed for the treatment of IBD.
We used time-dependent variables to assess the association of medications with fungal infections. We chose a 90-day interval because many of the medications of interest are typically administered every 8 weeks. Some biologics are given more frequently than every 8 weeks, such as adalimumab, which can be given as often as every week. However, it is unclear how long these medications impact susceptibility to infection. They likely act longer than their last dose. Our estimates may be conservative because if a patient developed a fungal infection a few days into a subsequent 90-day interval, the association with the medication in the previous interval would not be reflected in the HR. We tested 30- and 60-day time intervals in sensitivity analysis and the HR for anti-TNFs and corticosteroids were similar.
## Conclusions
We found that invasive fungal infections were more common than TB in patients with IBD in a US commercial claims database. Anti-TNFs were associated with 1.6-fold increased risk of invasive fungal infections, but corticosteroids were associated with much higher risk (5.4-fold). Future studies should examine whether newer corticosteroid-sparing biologics such as vedolizumab and ustekinumab may be safer than anti-TNFs, particularly in patients at increased risk for fungal infections, such as those on TPN.
## Funding
This publication was supported by The Foundation for Barnes-Jewish Hospital and their generous donors; and the Washington University Institute of Clinical and Translational Sciences (Just-In-Time Core Usage Funding Program grant #JIT687) which is, in part, supported by the NIH/National Center for Advancing Translational Sciences (NCATS), CTSA grant #UL1 TR002345. Dr. Gregory is supported by Washington University Institute of Clinical and Translational Sciences grant (UL1 TR002345 and T32DK007130) and the Eckert Fellowship Clinical Research Fund. Dr. Deepak is supported by a Junior Faculty Development Award from the American College of Gastroenterology and IBD Plexus of the Crohn’s & Colitis Foundation. Dr. Ciorba is supported by R01DK109384 and by the Lawrence C. Pakula MD IBD Education and Innovation Fund and Pfizer (IIS #61798927).
## Authors’ Contributions
M.H.G.: Conceived and designed the study, obtained funding, analyzed and interpreted the data, drafted the manuscript, critically revised the manuscript for important intellectual content, and approved the final version of the manuscript for submission. A.S.: Designed study, analyzed and interpreted the data, critically revised the manuscript for important intellectual content, and approved the final version of the manuscript for submission. D.S.: Acquired the data, analyzed and interpreted the data, critically revised the manuscript for important intellectual content, and approved the final version of the manuscript for submission. A.G.: Conceived the study, critically revised the manuscript for important intellectual content, and approved the final version of the manuscript for submission. C.M.-C.: Designed study, critically revised the manuscript for important intellectual content, and approved the final version of the manuscript for submission. K.B.N.: Designed study, data acquisition, critically revised the manuscript for important intellectual content, and approved the final version of the manuscript for submission. M.A.C.: Conceived the study, critically revised the manuscript for important intellectual content, and approved the final version of the manuscript for submission. R.P.R.: Conceived study, critically revised the manuscript for important intellectual content, and approved the final version of the manuscript for submission. M.A.O.: Designed study, analyzed and interpreted the data, critically revised the manuscript for important intellectual content, and approved the final version of the manuscript for submission. P.D.: Conceived and designed the study, analyzed and interpreted the data, critically revised the manuscript for important intellectual content, and approved the final version of the manuscript for submission.
## Conflicts of Interest
M.H.G., A.S., D.S., A.G., C.M.-C., and K.B.N.: None declared. M.A.C.: Consulting or advisory boards for AbbVie, Pfizer, Bristol Myers Squibb, Janssen, and Takeda. Educational grants or sponsored research agreement unrelated to the data in the paper from Pfizer, Janssen, BMS, Takeda, AbbVie, and the Crohn’s and Colitis Foundation. R.P.R.: Research support from Janssen, Gilead, Eli Lilly, and Protagonist. Consultant, Coloplast. M.A.O. received grant support from Pfizer for a study unrelated to this project, and consulting fees from Pfizer for work unrelated to this study. P.D.: Research support under a sponsored research agreement unrelated to the data in the paper and/or consulting from AbbVie, Arena Pharmaceuticals, Boehringer Ingelheim, Bristol Myers Squibb, Janssen, Pfizer, Prometheus Biosciences, Takeda Pharmaceuticals, Scipher Medicine, and CorEvitas, LLC. P.D. holds the position of Associate Editor for Crohn’s & Colitis 360 and has been recused from reviewing or making decisions for the manuscript.
## Data Availability
SAS code used for analysis is available upon request. Access to marketscan is restricted by institutional and contractual obligations and cannot be requested.
## References
1. Chang JT.. **Pathophysiology of inflammatory bowel diseases**. *N Engl J Med.* (2020.0) **383** 2652-2664. PMID: 33382932
2. Gregory MH, McKinnon A, Stwalley D. **Anti-tumour necrosis factor therapy for inflammatory bowel diseases do not impact serious infections after arthroplasty**. *J Crohns Colitis.* (2018.0) **13** 182-188
3. Lichtenstein GR, Feagan BG, Cohen RD. **Serious infection and mortality in patients with Crohn’s disease: more than 5 years of follow-up in the TREAT™ registry**. *Am J Gastroenterol.* (2012.0) **107** 1409-1422. PMID: 22890223
4. Wallis RS, Broder MS, Wong JY, Hanson ME, Beenhouwer DO.. **Granulomatous infectious diseases associated with tumor necrosis factor antagonists**. *Clin Infect Dis.* (2004.0) **38** 1261-1265. PMID: 15127338
5. Lichtenstein GR, Loftus EV, Isaacs KL, Regueiro MD, Gerson LB, Sands BE.. **ACG Clinical Guideline: management of Crohn’s disease in adults**. *Am J Gastroenterol.* (2018.0) **113** 481-517. PMID: 29610508
6. Ng SC, Shi HY, Hamidi N. **Worldwide incidence and prevalence of inflammatory bowel disease in the 21st century: a systematic review of population-based studies**. *Lancet.* (2017.0) **390** 2769-2778. PMID: 29050646
7. Hou JK, Tan M, Stidham RW. **Accuracy of diagnostic codes for identifying patients with ulcerative colitis and Crohn’s disease in the Veterans Affairs Health Care System**. *Dig Dis Sci.* (2014.0) **59** 2406-2410. PMID: 24817338
8. Waljee AK, Rogers MAM, Lin P. **Short term use of oral corticosteroids and related harms among adults in the United States: population based cohort study**. *BMJ.* (2017.0) **357** j1415. PMID: 28404617
9. George MD, Baker JF, Winthrop K. **Risk for serious infection with low-dose glucocorticoids in patients with rheumatoid arthritis: a cohort study**. *Ann Intern Med.* (2020.0) **173** 870-878. PMID: 32956604
10. Chen G, Lissoos T, Dieyi C, Null KD.. **Development and validation of an inflammatory bowel disease severity index using US administrative claims data: a retrospective cohort study**. *Inflamm Bowel Dis.* (2020.0) **27** 1177-1183
11. Blumberg HM, Jarvis WR, Soucie JM. **Risk factors for candidal bloodstream infections in surgical intensive care unit patients: the NEMIS Prospective Multicenter Study**. *Clin Infect Dis.* (2001.0) **33** 177-186. PMID: 11418877
12. Carey IM, Critchley JA, DeWilde S, Harris T, Hosking FJ, Cook DG.. **Risk of infection in type 1 and type 2 diabetes compared with the general population: a matched cohort study**. *Diabetes Care.* (2018.0) **41** 513-521. PMID: 29330152
13. Thomas-Rüddel DO, Schlattmann P, Pletz M, Kurzai O, Bloos F.. **Risk factors for invasive**. *Chest.* (2022.0) **161** 345-355. PMID: 34673022
14. Salmon-Ceron D, Tubach F, Lortholary O. **Drug-specific risk of non-tuberculosis opportunistic infections in patients receiving anti-TNF therapy reported to the 3-year prospective French RATIO registry**. *Ann Rheum Dis.* (2011.0) **70** 616-623. PMID: 21177290
15. Baddley JW, Winthrop KL, Chen L. **Non-viral opportunistic infections in new users of tumour necrosis factor inhibitor therapy: results of the SAfety Assessment of Biologic ThERapy (SABER) study**. *Ann Rheum Dis.* (2014.0) **73** 1942-1948. PMID: 23852763
16. 16.
United States Food and Drug Administration. Accessed August 16, 2021. https://www.accessdata.fda.gov/drugsatfda_docs/label/2013/103772s5359lbl.pdf
17. Singh JA, Furst DE, Bharat A. **2012 update of the 2008 American College of Rheumatology recommendations for the use of disease-modifying antirheumatic drugs and biologic agents in the treatment of rheumatoid arthritis**. *Arthritis Care Res.* (2012.0) **64** 625-639
18. Fine S, Vecchio M, Filipe Goncalves Monteiro J. **Overuse of tuberculosis surveillance testing in patients with inflammatory bowel disease compared to non-IBD patients on biologic therapy**. *Crohns Colitis 360.* (2021.0) **3** 1-7
19. Vajravelu RK, Osterman MT, Aberra FN. **Indeterminate QuantiFERON-TB Gold increases likelihood of inflammatory bowel disease treatment delay and hospitalization**. *Inflamm Bowel Dis.* (2017.0) **24** 217-226. PMID: 29272482
|
---
title: Differences in Inflammatory Pathways Between Dutch South Asians vs Dutch Europids
With Type 2 Diabetes
authors:
- Maaike E Straat
- Borja Martinez-Tellez
- Huub J van Eyk
- Maurice B Bizino
- Suzanne van Veen
- Eleonora Vianello
- Rinke Stienstra
- Tom H M Ottenhoff
- Hildo J Lamb
- Johannes W A Smit
- Ingrid M Jazet
- Patrick C N Rensen
- Mariëtte R Boon
journal: The Journal of Clinical Endocrinology and Metabolism
year: 2022
pmcid: PMC9999357
doi: 10.1210/clinem/dgac598
license: CC BY 4.0
---
# Differences in Inflammatory Pathways Between Dutch South Asians vs Dutch Europids With Type 2 Diabetes
## Body
South Asian individuals, originating from the Indian subcontinent and encompassing $20\%$ of the world population, are at particularly high risk of developing type 2 diabetes (T2D). In high-income countries such as the Netherlands, South Asian people have a 2- to 4-times higher risk of developing T2D compared to people of European origin, in the manuscript further called “Europid” [1]. Notably, at the moment of diagnosis, a high proportion of South Asian patients with T2D have a lower body mass index (BMI) and are at a younger age as compared to Europids. Additionally, in South Asians microvascular and macrovascular complications of T2D evolve within a shorter duration after start of the disease [2, 3]. Several factors have been proposed to underlie the increased susceptibility of South Asians to develop T2D compared to Europids, such as genetic predisposition, differences in lifestyle, central adiposity, low lean mass, low brown adipose tissue volume, and insulin resistance [4, 5]. However, the high rate of T2D in South Asian individuals cannot be fully explained by these factors alone.
Over the last decades, inflammation has been increasingly acknowledged to play an important role in the pathogenesis of T2D, at least partly by accelerating insulin resistance (6–8). Accordingly, clinical studies have shown that anti-inflammatory treatments, such as the interleukin-1-receptor antagonist Anakinra and the anti-inflammatory compound salsalate, improve glycemic control in patients with T2D (9–12). Although anti-inflammatory therapy is thus regarded as a promising strategy to improve T2D regulation, these studies have so far been conducted only in patients of Europid origin. Interestingly, previous studies support the presence of a more proinflammatory phenotype in South Asians compared with Europids. Concentrations of the nonspecific inflammatory marker C-reactive protein are higher in healthy middle-aged South Asian compared with Europid men and women [13, 14]. Furthermore, interleukin-6 levels are higher in healthy young South Asian men [15] and healthy middle-aged South Asian women than in matched Europids [16]. These data thus support a possible pathophysiological role for inflammation in explaining the increased risk of the South Asian population to develop T2D. However, which aspects of the immune system may be differentially regulated in South Asians with T2D is currently unknown.
To further pinpoint such a role and to elucidate the possible clinical benefit of anti-inflammatory therapy to reduce T2D burden in the South Asian population, a detailed overview of the inflammatory state of South Asian patients with T2D is urgently needed. Therefore, the aim of this study was to investigate whether circulating messenger RNA (mRNA) transcripts of a broad range of immune related genes (using an untargeted approach by measuring, eg, markers of T cells, B cells, natural killer cells, and interleukins) are different between patients with T2D from Dutch South Asian vs Dutch Europid descent.
## Abstract
### Context
South Asian individuals are more prone to develop type 2 diabetes (T2D) coinciding with earlier complications than Europids. While inflammation plays a central role in the development and progression of T2D, this factor is still underexplored in South Asians.
### Objective
This work aimed to study whether circulating messenger RNA (mRNA) transcripts of immune genes are different between South Asian compared with Europid patients with T2D.
### Methods
A secondary analysis was conducted of 2 randomized controlled trials of Dutch South Asian ($$n = 45$$; age: 55 ± 10 years, body mass index [BMI]: 29 ± 4 kg/m2) and Dutch Europid ($$n = 44$$; age: 60 ± 7 years, BMI: 32 ± 4 kg/m2) patients with T2D. Main outcome measures included mRNA transcripts of 182 immune genes (microfluidic quantitative polymerase chain reaction; Fluidigm Inc) in fasted whole-blood, ingenuity pathway analyses (Qiagen).
### Results
South Asians, compared to Europids, had higher mRNA levels of B-cell markers (CD19, CD79A, CD79B, CR2, CXCR5, IGHD, MS4A1, PAX5; all fold change > 1.3, false discovery rate [FDR] < 0.008) and interferon (IFN)-signaling genes (CD274, GBP1, GBP2, GBP5, FCGR1A/B/CP, IFI16, IFIT3, IFITM1, IFITM3, TAP1; all FC > 1.2, FDR < 0.05). In South Asians, the IFN signaling pathway was the top canonical pathway (z score 2.6; $P \leq .001$) and this was accompanied by higher plasma IFN-γ levels (FC = 1.5, FDR = 0.01). Notably, the ethnic difference in gene expression was larger for women ($\frac{20}{182}$ [$11\%$]) than men ($\frac{2}{182}$ [$1\%$]).
### Conclusion
South Asian patients with T2D show a more activated IFN-signaling pathway compared to Europid patients with T2D, which is more pronounced in women than men. We speculate that a more activated IFN-signaling pathway may contribute to the more rapid progression of T2D in South Asian compared with Europid individuals.
## Study Design and Participants
The present study is a secondary analysis of 2 previously performed double-blind, placebo-controlled, randomized clinical trials that were both designed to investigate the effect of 26-week liraglutide treatment on cardiovascular end points in overweight and obese patients with T2D [17, 18]. In the first trial (performed 2013-2016), 49 Dutch Europid patients with T2D were included [17]. In the second trial (performed 2015-2018), 47 Dutch South Asian patients with T2D were included, of whom South Asian ethnicity was based on being born and raised in the Netherlands and having 4 grandparents from South Asian descent [18]. Due to missing samples, in the present study 44 Dutch Europid (19 women; $43\%$) and 45 Dutch South Asian (27 women; $60\%$) patients with T2D were included. For both trials, inclusion criteria were BMI greater than or equal to 23; age 18 to 74 years; and glycated hemoglobin A1c (HbA1c) greater than or equal to $6.5\%$ and less than or equal to $11.0\%$ (≥ 47.5 and ≤ 96.4 mmol/mol). Patients were allowed to be treated with glucose-lowering medication (exclusively metformin, sulfonylurea derivatives, and insulin), although with a stable dosage for at least 3 months before participation in the study. Patients were allowed to use antihypertensives and statins. Exclusion criteria were use of glucose-lowering medication other than those mentioned earlier; presence of renal disease; congestive heart failure according to New York Heart Association classification III to IV; uncontrolled hypertension (systolic blood pressure > 180 mm Hg and/or diastolic blood pressure > 110 mm Hg); or an acute coronary or cerebrovascular accident within 30 days before study inclusion. The trials were conducted at the Leiden University Medical Center, the Netherlands, and were approved by the local ethics committee. Written informed consent was obtained from all individuals before inclusion. The trials were conducted in accordance with the principles of the revised Declaration of Helsinki and were registered at clinicaltrials.gov (NCT01761318 and NCT02660047, respectively).
## Study Procedures
This is a cross-sectional, study. The data used for the present secondary analyses were obtained at baseline before patients started their liraglutide treatment. During this visit, participants fasted for at least 6 hours. Their medical history was consulted to, among other things, obtain information about their diabetes duration and medication use. Body weight and total fat mass were assessed using bioelectrical impedance analysis (Bodystat 1500, Bodystat Ltd). Visceral adipose tissue (VAT) mass and abdominal subcutaneous adipose tissue (SAT) mass were assessed by magnetic resonance imaging (MRI).
## Magnetic Resonance Imaging Protocol
The MRI protocol has been described in detail previously [19]. In short, all participants underwent an MRI using a clinical 3 Tesla Ingenia whole-body MR system (Philips Medical Systems) at baseline and after 26 weeks of liraglutide treatment. Participants were scanned in the supine position after at least 6 hours fasting. Semiautomated segmentation of VAT and abdominal SAT was depicted by threshold-based inclusion of fat, with manual correction. VAT and SAT were calculated as mean area of fat in 3 slices.
## Blood Samples
Venous blood samples were drawn from the antecubital vein. To obtain plasma, blood samples were centrifuged, aliquoted, and stored at −80 °C until batch-wise analyses. Plasma total cholesterol, high-density lipoprotein cholesterol, triglycerides, and C-reactive protein concentrations were measured on a Roche Modular analyzer (Roche Diagnostics). low-density lipoprotein cholesterol was calculated according to the Friedewald formula. HbA1c was measured with ion-exchange high-performance liquid chromatography (HPLC; Tosoh G8, Sysmex Nederland B.V.). The commercially available protein biomarker panel “Target 96 Inflammation” from Olink proteomics (Olink Bioscience) was used to measure interferon (IFN)-γ [20]. Blood samples for RNA isolation were collected in PAXgene Blood RNA tubes (BD Biosciences) and stored at −80 °C following instructions from the manufacturer until batch-wise analyses.
## RNA isolation
Total RNA was extracted from whole-blood samples in PAXgene Blood RNA tubes using the automated PAXgene Blood miRNA Kit (PreAnalitiX) procedure, according to the manufacturer's protocol. Briefly, cells were pelleted and lysed. Cell contents were treated with proteinase K and silica-based column extraction was performed, including on-column DNAse I treatment. Total RNA quantity was determined using the Qubit RNA BR Assay Kit (Thermo Fisher Scientific).
## Complementary DNA synthesis and preamplification
Complementary DNA (cDNA) was synthesized by performing reverse transcription of 50 ng RNA (incubation at 25 °C for 5 minutes, 42 °C for 30 minutes and 85 °C for 5 minutes). Reverse Transcription Master Mix (Fluidigm), containing M-MLV reverse transcriptase, random hexamer, and oligo dT primers, was used. cDNA was preamplified to increase the amount of input material needed for our high-throughput quantitative polymerase chain reaction (qPCR) technique. For preamplification, we used a pool of the target TaqMan assays (Thermo Fisher Scientific, 0.2× each in TE buffer: 10 mM Tris-HCl, 0.1 mM EDTA, pH 8.0) and Preamp Master Mix (Fluidigm) according to the manufacturers’ instructions. Thermal cycling conditions were 95 °C for 2 minutes followed by 14 cycles at 95 °C for 15 seconds and 60 °C for 4 minutes. Preamplified cDNA was diluted 1:5 in TE buffer and stored at −20 °C before analysis.
## High-throughput quantitative polymerase chain reaction gene expression analysis
mRNA transcripts of 182 genes were measured by high-throughput microfluidic qPCR using 96.96 IFC chips on the Biomark HD system (Fluidigm), as described by the manufacturer. Each TaqMan Assay (20X, FAM-MGB; Supplementary Table S1) [21] was diluted in Assay Loading Reagent (Fluidigm) to a 10× assay mix. Sample mixes were prepared containing 1× TaqMan Universal PCR Master Mix (Thermo Fisher Scientific), 1× Sample Loading Reagent (Fluidigm), and 2.25 μL of preamplified cDNA. The 96.96 IFC chip was primed with Control Line Fluid (Fluidigm) and assay and sample mixes were loaded into the chip using the IFC Controller HX (Fluidigm). qPCR was performed with the Biomark HD using the following thermal cycling protocol: 95 °C for 10 minutes, followed by 40 cycles at 95 °C for 15 seconds and 60 °C for 1 minute. Data were analyzed using Fluidigm Real-Time PCR Analysis Software (version 4.1.3). A cycle threshold (Ct) value less than or equal to 35 was determined as the cutoff for reliable detection. Relative target gene expression was determined by calculating ΔCt using GAPDH as the reference gene.
## Ingenuity Pathway Analysis
Ingenuity pathway analysis (IPA; Qiagen) was performed to assess transcriptional regulators and canonical pathways involved in observed differences in mRNA levels of immune genes between ethnicities. For each potential transcriptional regulator, the program calculates an overlap P value and activation z score. The overlap P value is calculated using the Fisher exact test and indicates whether overlap between the genes in the data set and genes regulated by the transcriptional regulator is statistically significant. The activating z score quantifies the predicted activation state of a transcriptional regulator. Overlap P values less than.01 and z scores greater than 2 are considered statistically significant.
## Statistical Analyses
Statistical analyses were conducted using R (version 3.6.2, Team, 2019) and Prism 9 for Windows (version 9.0.1, 2021, GraphPad Software LLC). Normal distribution of the baseline characteristics was tested using the Shapiro-Wilk test. Dependent on whether the data followed normal Gaussian distribution, baseline characteristics between ethnicities and between sexes were compared using a 2-tailed unpaired t test or the nonparametric Mann-Whitney U test in R. mRNA levels of immune genes were expressed as ΔCT values on logarithmic scale with a base of 2 (log2). Detected proteins were generated as normalized protein expression values on a log2 scale, with larger numbers representing higher protein levels in the sample. mRNA levels of immune genes and protein levels were compared between ethnicities and between sexes using an analysis of variance model (aov) in R with “ethnicity” and/or “sex” as between-subjects factor. For comparisons between ethnicities and/or sex, P values were corrected for multiple comparisons using the Benjamini-Hochberg false discovery rate (FDR) in R. An FDR-adjusted P value of less than.05 was considered statistically significant. Pearson correlation and simple linear regression were performed in Prism with mRNA levels as the dependent outcome, and IFN-γ or fat mass percentage as independent outcomes. Here, P value of less than.05 was considered statistically significant.
Since mRNA levels of immune genes were expressed as ΔCT values on a log2 scale, larger values represent lower mRNA levels in the sample. For the calculation of the log2 fold changes (log2FC) for volcano plots, the mean ΔCT value of Europid patients was subtracted from the mean ΔCT value of Dutch South Asian patients and multiplied by −1 to obtain the correct direction ([Dutch South Asian – Europid] × –1). Relative expression values (= fold changes) used in bar graphs were calculated using the ΔΔCT method, with the Europid patients as the control group. Figures represent geometric mean and geometric standard deviation and all figures were prepared with Prism 9 for Windows (version 9.0.1, 2021, GraphPad Software LLC). All supplemental figures are located in a digital research materials repository [21].
## Baseline Characteristics
Baseline characteristics are shown in Table 1 and in Supplementary Table S2 [21] for men and women separately. As reported in a previous publication in which the primary end point of the present study was described [22], compared to Dutch Europids, Dutch South Asian participants had a lower BMI (29.4 ± 4.0 vs 32.3 ± 4.0) and lower low-density lipoprotein cholesterol concentration (2.1 ± 0.8 vs 2.6 ± 0.9 mmol/L). In addition, despite being younger (54.7 ± 10.3 vs 59.6 ± 6.5 years), Dutch South Asians had longer diabetes duration (17.4 ± 9.9 vs 10.7 ± 6.4 years) and higher rates of retinopathy (51 vs $9\%$) and macrovascular diseases (27 vs $5\%$) than Dutch Europids [22].
**Table 1.**
| Unnamed: 0 | Dutch Europid(n = 44) | Dutch South Asian(n = 45) |
| --- | --- | --- |
| Demographics | Demographics | Demographics |
| ȃWomen, n, % | 19, 43% | 27, 60% |
| ȃAge, y | 59.6 ± 6.5 | 54.6 ± 10.3a |
| ȃDiabetes duration, y | 10.7 ± 6.4 | 17.4 ± 9.9b |
| Clinical parameters | Clinical parameters | Clinical parameters |
| ȃWeight, kg | 97.0 ± 14.0 | 79.6 ± 11.9c |
| ȃLength, cm | 173.2 ± 8.9 | 165.6 ± 8.8c |
| ȃBMI | 32.3 ± 4.0 | 29.4 ± 4.0b |
| ȃWaist-height ratio | 0.64 ± 0.06 | 0.61 ± 0.06a |
| ȃBody fat, % | 37.1 ± 9.4 | 37.1 ± 9.2 |
| ȃVAT/SAT ratio | 0.6 ± 0.3 | 0.6 ± 0.3 |
| ȃHbA1c, mmol/mol | 66.1 ± 11.0 | 67.7 ± 11.5 |
| ȃHbA1c, % | 8.2 ± 1.0 | 8.3 ± 1.1 |
| ȃTotal cholesterol, mmol/L | 4.8 ± 1.0 | 4.2 ± 1.0b |
| ȃHDL-C, mmol/L | 1.2 ± 0.3 | 1.2 ± 0.3 |
| ȃLDL-C, mmol/L | 2.6 ± 0.9 | 2.1 ± 0.8b |
| ȃCRP, mmol/L | 3.1 ± 3.3 | 3.6 ± 4.1 |
| Diabetic complications/comorbidity | Diabetic complications/comorbidity | Diabetic complications/comorbidity |
| ȃDiabetic retinopathy, n, % | 4, 9% | 23, 51%c |
| ȃDiabetic nephropathy, n, % | 11, 25% | 10, 22% |
| ȃDiabetic neuropathy, n, % | 15, 34% | 12, 27% |
| ȃMacrovascular disease, n, % | 2, 5% | 12, 27%b |
| Concomitant medication use | Concomitant medication use | Concomitant medication use |
| ȃMetformin, mg/d | 2047 ± 569 | 1750 ± 659a |
| ȃSulfonylurea, n, % | 13, 30% | 8, 18% |
| ȃInsulin, n, % | 28, 64% | 34, 76% |
| ȃStatin, n, % | 36, 82% | 34, 76% |
| ȃAntihypertensive drug, n, % | 34, 77% | 32, 71% |
## Circulating Messenger RNA Levels of B-Cell Markers and Interferon-Signaling Genes Are Higher in Dutch South Asians Compared With Dutch Europids With Type 2 Diabetes
mRNA levels of 30 of 182 ($16\%$) immune-related genes were significantly different (FDR < 0.05) between Dutch South Asian and Dutch Europid participants (Fig. 1). Among those, mRNA levels of 3 genes were significantly lower in Dutch South Asians (scavenger receptor MARCO, anti-inflammatory cytokine IL10, and inflammasome component NLRP2; all FC < 0.7, FDR < 0.04), whereas mRNA levels of 27 genes were higher in Dutch South Asians compared to Dutch Europids. Specifically, mRNA levels of the apoptosis involved CASP8, oncogene AKT1, T-cell subset marker CD3E, natural killer cell marker KLRC$\frac{2}{3}$, cytotoxicity marker GZMA, and pattern recognition receptors TLR7, TLR10, NOD1, and NOD2 were higher in Dutch South Asians than in Dutch Europids (all FC > 1.2, FDR < 0.05). Additionally, we found a clear pattern in which mRNA levels of 8 of 10 ($80\%$) measured B-cell markers (CD19, CD79A, CD79B, CR2, CXCR5, IGHD, MS4A1, PAX5; all FC > 1.4, FDR < 0.008; Fig. 2A) and 10 of 25 ($40\%$) measured IFN-signaling genes (CD274, FCGR1A/B/CP, GBP1, GBP2, GBP5, IFI16, IFITM1, IFITM3, IFIT3, TAP1; all FC > 1.2, FDR < 0.05; Fig. 2B) were higher in Dutch South Asians than in Dutch Europids.
**Figure 1.:** *Differences in circulating messenger RNA (mRNA) levels of immune genes between Dutch South Asian vs Dutch Europid patients with type 2 diabetes (T2D). Volcano plot showing the differences in circulating mRNA levels of immune-related genes between Dutch South Asian and Dutch Europid patients with T2D. The x-axis shows the log2FC between Dutch South Asian and Dutch Europid patients, the y-axis shows the P value. P values were obtained from one-way analysis of variance and thereafter corrected using Benjamini-Hochberg's FDR procedure. The top dashed line represents FDR = 0.05. EU, Dutch Europid patients; FDR, false discovery rate–adjusted P value; Log2FC, log2 fold change; SA, Dutch South Asian patients.* **Figure 2.:** *Differences in messenger RNA (mRNA) levels of B-cell markers and interferon (IFN)-signaling genes in Dutch South Asian compared to Dutch Europid patients with type 2 diabetes (T2D). Bar charts showing mRNA levels of A, B-cell markers and B, IFN-signaling genes that are different between Dutch South Asian vs Dutch Europid T2D patients (false discovery rate [FDR] < 0.05). mRNA levels are represented relative to levels of Europids, with geometric mean and geometric SD. *FDR < 0.05, **FDR < 0.01, ***FDR < 0.001 (one-way analysis of variance, thereafter corrected using the Benjamini-Hochberg FDR procedure).*
## Interferon-Signaling Pathway Is the top Canonical Pathway That Could Explain Observed Ethnic Difference in Messenger RNA Levels of Immune-Related Genes
IPA was performed to identify transcriptional regulators and canonical pathways that could explain the observed ethnic differences in mRNA levels of immune genes. In total, 469 potential transcriptional regulators were identified with a statistically significant overlap ($P \leq .01$) between the genes in our data set and genes known to be regulated by these transcriptional regulators. Among those, 60 transcriptional regulators had a z score greater than 2 or less than −2. The top 20 hits with the highest z score are shown in Fig. 3A. IFN-γ had the highest z score (z score: 4.1; $P \leq .001$), followed by IFN-α (z score: 3.7; $P \leq .001$), IFN regulatory factor 7 (IRF7; z score: 3.4; $P \leq .001$), IRF1 (z score: 3.4; $P \leq .001$), IRF3 (z score: 3.3; $P \leq .001$), and IFN-α/β receptor (IFNAR; z score: 3.2; $P \leq .001$). In addition, using IPA, 68 significant canonical pathways ($P \leq .01$) were identified, of which only 4 pathways had a z score greater than 2 (Fig. 3B). These top canonical pathways were IFN signaling (z score: 2.6; $P \leq .001$), role of pattern recognition receptors in recognition of bacteria and viruses (z score: 2.6; $P \leq .001$), Th1 pathway (z score: 2.4; $P \leq .001$), and systemic lupus erythematosus in B-cell signaling pathway (z score: 2.3; $P \leq .001$). Thus, both in the transcriptional regulators and canonical pathways, IFN-signaling pathways appeared as a top hit.
**Figure 3.:** *Transcriptional regulators and top canonical pathways that could explain the observed ethnic difference in messenger RNA (mRNA) levels of immune genes. A, The top 10 upregulating and top 10 downregulating transcriptional regulators with a statistically significant overlap (P < .01) and a z score greater than 2 or less than −2. The y-axis shows the name of the transcriptional regulators; the x-axis shows the z score. B, Top canonical pathways with a statistically significant overlap (P < .01) and a z score greater than or equal to 2. Those with a z score greater than 2 appear in red. The y-axis shows the name of the pathway; the x-axis shows the z score. Transcriptional regulators and top canonical pathways are obtained from ingenuity pathway analyses, performed on the list of genes with a differential expression (P < .05) between Dutch South Asians and Dutch Europids, with as input the log2 fold change differences in mRNA levels.*
Moreover, plasma protein levels of IFN-γ were higher in Dutch South Asians compared to Dutch Europids (FC = 1.5, FDR = 0.01; Supplementary Fig. S1A) [21]. IFN-γ protein levels were positively associated with mRNA levels of several IFN-signaling genes that were different between Dutch South Asians and Dutch Europids (Supplementary Fig. S1B) [21]. In both ethnicities, IFN-γ protein levels were positively associated with mRNA levels of CD274, GBP1, GBP5, and IFIT3. Additionally, in Dutch South Asians only, IFN-γ protein levels were positively associated with mRNA levels of GBP2 and FCGR1A/B/CP, and in Dutch Europids only, with mRNA levels of IFITM3 (Supplementary Fig. S1B) [21].
## Ethnic Differences in Immune Messenger RNA Levels Are Greater in Women Than in Men
Interestingly, although we did not observe an interaction between sex and ethnic difference in mRNA levels ($P \leq .05$), the ethnic difference in mRNA levels of immune genes was more pronounced in women ($\frac{20}{182}$, $11\%$) than men ($\frac{2}{182}$ genes, $1\%$; Fig. 4A and 4B). In Dutch South Asian males, the mRNA level of ETV7 was lower than in Europid males (FC = 0.4, FDR = 0.05), and B-cell marker IGHD was higher (FC = 1.7, FDR = 0.05; see Fig. 4A). In Dutch South Asian women, the mRNA level of MARCO was lower than in Europid women (FC = 0.4, FDR = 0.01), whereas levels of CASP8, NOD2, ZNF532, LAG3, and AKT1 were higher (all FC > 1.3, FDR < 0.05; see Fig. 4B). Moreover, 7 of 10 B-cell markers (CD19, CD79A, CD79B, CXCR5, IGHD, MS4A1, PAX5; all FC > 1.5, FDR < 0.04) and 7 of 25 IFN-signaling genes (GBP1, GBP2, GBP5, IFI16, IFITM1, IFITM3, TAP1; all FC > 1.5, FDR < 0.05) were higher in Dutch South Asian females (see Fig. 4B).
**Figure 4.:** *Differences in messenger RNA (mRNA) levels of immune genes between Dutch South Asian vs Dutch Europid males and females with type 2 diabetes (T2D). A and B, Volcano plot showing the differences in mRNA levels of immune-related genes between Dutch South Asian and Dutch Europid T2D patients, in A, men and B, women. The x-axes show the log2FC between Dutch South Asian and Dutch Europid patients; the y-axes show the P value. P values were obtained from one-way analysis of variance and thereafter corrected using the Benjamini-Hochberg FDR procedure. The top dashed line represents FDR = 0.05. EU, Dutch Europid patients; FDR, false discovery rate–adjusted P value; Log2FC, log2 fold change; SA, Dutch South Asian patients.*
Next, we performed correlation analyses between baseline characteristics and mRNA levels of B-cell markers, IFN-signaling genes, and IFN-γ protein level, for men and women separately (Table 2). Only a few genes showed an association with baseline characteristics ($P \leq .05$); however, none of the correlations remained significant after FDR correction for multiple comparisons.
**Table 2.**
| Unnamed: 0 | Unnamed: 1 | Unnamed: 2 | Unnamed: 3 | Unnamed: 4 | Males | Males.1 | Males.2 | Males.3 | Males.4 | Males.5 | Males.6 | Males.7 | Males.8 | Males.9 | Females | Females.1 | Females.2 | Females.3 | Females.4 | Females.5 | Females.6 | Females.7 | Females.8 | Females.9 | Females.10 | Females.11 | Females.12 | Females.13 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| | Dutch Europid | Dutch Europid | Dutch Europid | Dutch Europid | Dutch Europid | Dutch Europid | Dutch Europid | Dutch South Asian | Dutch South Asian | Dutch South Asian | Dutch South Asian | Dutch South Asian | Dutch South Asian | Dutch South Asian | Dutch Europid | Dutch Europid | Dutch Europid | Dutch Europid | Dutch Europid | Dutch Europid | Dutch Europid | Dutch Europid | Dutch South Asian | Dutch South Asian | Dutch South Asian | Dutch South Asian | Dutch South Asian | Dutch South Asian |
| | Age | DD | BMI | Bodyfat % | Waist-heightratio | HbA1c | Metf.dose | Age | DD | BMI | Bodyfat % | Waist-heightratio | HbA1c | Metf.dose | Age | DD | BMI | Bodyfat % | Waist-heightratio | HbA1c | Metf.dose | Age | DD | BMI | Bodyfat % | Waist-heightratio | HbA1c | Metf.dose |
| B-cell markers | B-cell markers | B-cell markers | B-cell markers | B-cell markers | B-cell markers | B-cell markers | B-cell markers | B-cell markers | B-cell markers | B-cell markers | B-cell markers | B-cell markers | B-cell markers | B-cell markers | B-cell markers | B-cell markers | B-cell markers | B-cell markers | B-cell markers | B-cell markers | B-cell markers | B-cell markers | B-cell markers | B-cell markers | B-cell markers | B-cell markers | B-cell markers | B-cell markers |
| CD19 | −0.14 | 0.39 | −0.03 | −0.21 | 0.07 | 0.18 | 0.13 | −0.12 | 0.17 | 0.12 | −0.04 | 0.10 | −0.14 | 0.17 | −0.10 | 0.27 | 0.13 | 0.10 | 0.19 | −0.41 | 0.11 | −0.08 | −0.01 | 0.16 | 0.11 | 0.20 | −0.25 | 0.47a |
| MS4A1 (CD20) | −0.12 | 0.23 | 0.20 | −0.12 | 0.21 | 0.03 | −0.09 | −0.22 | −0.03 | 0.25 | 0.18 | 0.14 | 0.00 | 0.06 | 0.13 | −0.27 | 0.30 | 0.37 | 0.43 | −0.31 | 0.05 | −0.09 | −0.05 | 0.04 | −0.06 | 0.10 | −0.26 | 0.34 |
| CR2 (CD21) | −0.35 | 0.17 | 0.01 | −0.17 | −0.02 | 0.26 | −0.05 | −0.13 | −0.25 | 0.19 | 0.19 | 0.19 | 0.05 | −0.16 | −0.58** | 0.17 | 0.59** | 0.22 | 0.38 | 0.13 | −0.15 | −0.26 | −0.15 | 0.15 | 0.01 | 0.16 | −0.19 | 0.33 |
| CD79A | −0.24 | 0.24 | 0.07 | −0.21 | 0.11 | 0.17 | −0.02 | −0.12 | 0.02 | 0.16 | 0.12 | −0.22 | −0.17 | 0.49 | 0.12 | 0.24 | 0.33 | 0.28 | 0.30 | −0.52a | 0.15 | −0.11 | −0.06 | 0.11 | 0.04 | 0.16 | −0.21 | 0.43 |
| CD79B | −0.14 | 0.28 | −0.01 | −0.20 | 0.06 | 0.11 | 0.00 | −0.07 | 0.26 | −0.23 | −0.21 | −0.07 | −0.09 | 0.25a | −0.02 | 0.24 | 0.19 | 0.03 | 0.42 | −0.44 | 0.01 | −0.09 | 0.00 | 0.17 | 0.11 | 0.21 | −0.11 | 0.41a |
| CXCR5 | −0.19 | 0.33 | 0.01 | −0.15 | 0.06 | 0.08 | 0.04 | −0.07 | 0.15 | −0.01 | −0.19 | 0.22 | −0.23 | −0.09 | −0.15 | 0.17 | 0.33 | 0.25 | 0.24 | −0.35 | 0.05 | −0.24 | −0.09 | 0.27 | 0.15 | 0.09 | −0.21 | 0.37a |
| IGHD | −0.39 | 0.24 | 0.15 | −0.18 | 0.09 | 0.04 | −0.05 | 0.11 | 0.18 | 0.12 | −0.06 | 0.21 | −0.01 | 0.09 | −0.05 | −0.02 | 0.34 | 0.25 | 0.50a | −0.41 | 0.01 | −0.19 | −0.09 | 0.28 | 0.17 | 0.18 | −0.21 | 0.37 |
| PAX5 | −0.23 | 0.24 | 0.12 | −0.15 | 0.12 | 0.14 | −0.03 | 0.04 | 0.25 | 0.29 | 0.24 | 0.36 | 0 | 0.19 | −0.09 | −0.12 | 0.4 | 0.37 | 0.33 | −0.32 | 0.07 | −0.1 | 0.01 | 0.07 | −0.03 | 0.14 | −0.24 | 0.34 |
| IFN-signaling genes + IFN-γ | IFN-signaling genes + IFN-γ | IFN-signaling genes + IFN-γ | IFN-signaling genes + IFN-γ | IFN-signaling genes + IFN-γ | IFN-signaling genes + IFN-γ | IFN-signaling genes + IFN-γ | IFN-signaling genes + IFN-γ | IFN-signaling genes + IFN-γ | IFN-signaling genes + IFN-γ | IFN-signaling genes + IFN-γ | IFN-signaling genes + IFN-γ | IFN-signaling genes + IFN-γ | IFN-signaling genes + IFN-γ | IFN-signaling genes + IFN-γ | IFN-signaling genes + IFN-γ | IFN-signaling genes + IFN-γ | IFN-signaling genes + IFN-γ | IFN-signaling genes + IFN-γ | IFN-signaling genes + IFN-γ | IFN-signaling genes + IFN-γ | IFN-signaling genes + IFN-γ | IFN-signaling genes + IFN-γ | IFN-signaling genes + IFN-γ | IFN-signaling genes + IFN-γ | IFN-signaling genes + IFN-γ | IFN-signaling genes + IFN-γ | IFN-signaling genes + IFN-γ | IFN-signaling genes + IFN-γ |
| CD274 | 0.21 | 0.25 | 0.07 | 0.08 | 0.20 | 0.12 | 0.28 | −0.06 | 0.25 | 0.30 | 0.09 | 0.27 | 0.23 | −0.40 | 0.10 | −0.08 | −0.23 | −0.18 | −0.18 | 0.02 | 0.07 | −0.10 | 0.08 | 0.49b | 0.39a | 0.32 | −0.20 | 0.15 |
| GBP1 | 0.13 | 0.10 | 0.09 | −0.05 | 0.03 | 0.02 | 0.24 | −0.13 | −0.07 | −0.02 | 0.02 | 0.01 | 0.03 | −0.16 | 0.21 | −0.08 | −0.07 | −0.14 | 0.15 | 0.11 | 0.37 | 0.01 | 0.26 | 0.24 | 0.03 | 0.33 | −0.23 | 0.25 |
| GBP2 | −0.14 | −0.16 | 0.34 | 0.15 | 0.30 | −0.10 | 0.03 | −0.21 | −0.45 | 0.22 | 0.14 | 0.14 | −0.05 | −0.38 | 0.13 | −0.06 | 0.03 | −0.03 | 0.26 | −0.10 | 0.01 | 0.04 | 0.11 | 0.25 | 0.13 | 0.33 | −0.27 | 0.31 |
| GBP5 | 0.20 | 0.20 | 0.11 | −0.15 | 0.06 | 0.09 | 0.14 | −0.14 | −0.13 | 0.19 | 0.19 | 0.17 | −0.16 | −0.18 | 0.25 | −0.08 | −0.18 | −0.16 | −0.03 | 0.06 | 0.19 | 0.11 | 0.29 | 0.30 | 0.24 | 0.48a | −0.27 | 0.19 |
| FCGR1A /B/CP | 0.03 | −0.09 | −0.11 | −0.04 | 0.08 | −0.13 | 0.17 | −0.04 | 0.06 | 0.24 | 0.20 | 0.24 | 0.07 | −0.13 | −0.22 | −0.02 | 0.18 | 0.12 | 0.30 | 0.00 | 0.33 | −0.25 | −0.20 | 0.49b | 0.48a | 0.35 | −0.07 | 0.15 |
| IFI16 | −0.06 | 0.13 | −0.01 | −0.05 | 0.15 | −0.24 | 0.28 | −0.45 | −0.29 | 0.31 | 0.06 | 0.13 | −0.17 | −0.28 | 0.08 | 0.05 | 0.05 | −0.29 | 0.24 | 0.11 | 0.29 | −0.31 | −0.17 | 0.23 | 0.13 | 0.16 | −0.19 | 0.16 |
| IFIT3 | 0.14 | 0.04 | 0.03 | 0.00 | 0.05 | −0.30 | 0.15 | 0.26 | 0.11 | 0.28 | 0.44 | 0.42 | 0.04 | −0.18 | 0.29 | −0.03 | 0.16 | −0.12 | 0.31 | −0.06 | 0.16 | −0.35 | −0.15 | 0.27 | 0.00 | 0.19 | −0.21 | 0.09 |
| IFITM1 | 0.01 | −0.01 | −0.25 | −0.24 | −0.09 | −0.22 | −0.05 | −0.30 | −0.12 | −0.07 | 0.10 | −0.12 | 0.18 | 0.01 | −0.19 | −0.03 | −0.01 | −0.22 | 0.08 | −0.09 | 0.26 | −0.20 | −0.07 | 0.16 | 0.12 | −0.02 | −0.32 | 0.37 |
| IFITM3 | 0.06 | −0.01 | −0.39 | −0.25 | −0.24 | −0.34 | 0.11 | 0.11 | 0.05 | 0.53a | 0.53a | 0.58a | 0.26 | −0.15 | 0.13 | −0.23 | 0.35 | 0.23 | 0.54a | −0.28 | 0.16 | −0.21 | −0.12 | −0.09 | −0.20 | −0.31 | −0.13 | 0.21 |
| TAP1 | −0.06 | −0.01 | 0.23 | 0.17 | 0.25 | −0.26 | 0.11 | −0.18 | −0.47a | 0.15 | 0.09 | −0.04 | −0.45 | −0.19 | −0.04 | −0.53a | 0.21 | 0.27 | 0.18 | −0.11 | −0.11 | 0.01 | 0.04 | 0.10 | −0.01 | 0.18 | −0.25 | 0.16 |
| IFN-γ (protein) | 0.52** | 0.02 | −0.24 | −0.20 | −0.27 | 0.00 | 0.18 | 0.01 | 0.16 | 0.20 | 0.29 | 0.26 | 0.21 | 0.03 | 0.08 | −0.13 | −0.10 | −0.08 | −0.15 | −0.07 | 0.20 | −0.08 | 0.07 | 0.32 | 0.29 | 0.38 | −0.38 | 0.15 |
## Discussion
In this study, we compared mRNA levels of a large panel of immune-related genes between Dutch South Asian and Dutch Europid patients with T2D. We found that mRNA levels of 8 of 10 ($80\%$) measured B-cell markers and 10 of 25 ($40\%$) measured IFN-signaling genes in addition to IFN-γ protein levels were higher in Dutch South Asians compared to Dutch Europids. IPA showed that the IFN signaling pathway was the most activated canonical pathway and IFN-γ the top activating transcriptional regulator explaining these differences. Accordingly, IFN-γ protein levels were higher in Dutch South Asians compared to Dutch Europids. These ethnic differences were more pronounced in women than in men. We hypothesize that an enhanced IFN-signaling pathway may contribute to the more severe disease progression and accelerated risk for T2D-associated complications that is generally found in the South Asian compared to the Europid population.
The finding that the IFN-signaling pathway is more activated in Dutch South Asian compared to Dutch Europid patients with T2D is in seeming contrast with one of our previous findings. In overweight prediabetic men, we found lower mRNA levels of several type I IFN-signaling genes in subcutaneous white adipose tissue and skeletal muscle biopsies of South Asian compared with Europid men, without differences in blood mRNA levels of IFN-signaling genes [23]. Although both type I IFNs and IFN-γ (ie, type II IFN) regulate the antiviral response, they are in fact structurally unrelated, bind to a different receptor, and have distinct physiological functions [24, 25]. In turn, both studies point toward dysregulated IFN signaling in metabolically compromised Dutch South Asians, perhaps starting with impaired type I IFN signaling in metabolic tissues in individuals with prediabetes, followed by an overall more activated inflammatory response including accelerated IFN-signaling pathways once the disease progresses.
Indeed, the involvement of IFN signaling in the development and progression of T2D has been supported by previous preclinical and clinical studies. In prediabetic obese mice, virally induced IFN-γ (using murine cytomegalovirus) drives the progression from prediabetes to T2D by causing insulin resistance in skeletal muscles through downregulation of the insulin receptor [26]. Also in primary human adipocytes, IFN-γ induces insulin resistance by downregulating the insulin receptor, as well as insulin receptor substrate-1 and GLUT4 [27]. These preclinical findings are corroborated by an observational study showing that patients with T2D, compared to controls, have higher IFN-γ levels that positively correlate with HbA1c levels [28]. Moreover, in a cohort study consisting of 157 overweight Dutch Europid individuals, IFN-stimulated genes were found to be upregulated in the whole-blood transcriptome of insulin-resistant compared with insulin-sensitive individuals [29]. These data thus suggest that IFN-signaling pathways may play a role in the pathophysiology of T2D. Unfortunately, in the present study we did not study the severity of peripheral insulin resistance (eg, via hyperinsulinemic-euglycemic clamp). This would have enabled us to study possible associations of IFN signaling genes with insulin resistance. Notably, in patients with T2D, IFN-γ is also shown to be positively related to diabetes-associated complications such as nephropathy [30] and diabetic foot ulcers [31]. Thus, a higher activation of the IFN-signaling pathway could contribute to the increased risk of South Asians to develop T2D, but also a more severe T2D progression with a higher risk of T2D-associated complications. Nonetheless, we cannot exclude that a higher activation of the IFN-signaling pathway is reflective of a longer disease progression in South Asian patients with T2D, since the Dutch South Asians included in our study had a longer T2D duration compared to the Dutch Europids.
Next to higher expression of IFN-signaling genes, we found higher mRNA levels of B-cell markers, and lower mRNA levels of interleukin-10 (IL-10), in Dutch South Asian compared to Europid patients with T2D. B cells can contribute to insulin resistance via antigen presentation to T cells, changes in cytokine secretion, and pathogenic antibody production, thereby activating T cells (that secrete, eg, IFN-γ) and polarizing macrophages toward a proinflammatory phenotype (32–34). B-cell–deficient mice on a high-fat diet, compared to wild-type controls on the same diet, display lower blood glucose levels and less insulin resistance [33], accompanied by decreased systemic and adipose tissue inflammation, and decreased adipose tissue IFN-γ expression [34]. In healthy individuals, B cells are an important source of anti-inflammatory IL-10 secretion, and circulating IL-10 is positively associated with insulin sensitivity [35]. On the other hand, an impaired IL-10 response on a proinflammatory stimulus is related to the presence of metabolic syndrome and T2D in old individuals [36], and B cells from patients with T2D show diminished IL-10 secretion on stimulation of its toll-like receptors [37]. In summary, B cells can become involved in inflammation and insulin resistance, among others, by decreasing IL-10 secretion, and activating T cells that produce IFN-γ [32]. The results of the present study suggest that this inflammatory pathway is enhanced in Dutch South Asian compared to Europid patients with T2D, and future studies should investigate whether the phenotype of B cells in South Asians indeed associates with enhanced insulin resistance.
We observed that the ethnic differences in immune mRNA levels were more pronounced in female patients with T2D than in male patients with T2D. Generally, premenopausal females are at lower risk of developing cardiometabolic diseases compared to men of the same age. However, after menopause or once T2D has developed, this sex-dependent benefit diminishes and the cardiometabolic risk in women accelerates [38, 39]. Consequently, postmenopausal women with T2D are at greater risk of developing cardiovascular comorbidities compared to age-matched men with T2D [39, 40]. Interestingly, the age at menopause differs across populations, with lower menopausal ages reported among South Asians (ie, age 44-49 years) compared with Europids (ie, age 50-54 years) [41, 42]. In this study, we cannot exclude that more South Asian women have reached the postmenopausal state than Europid women, which, together with their earlier diabetes onset, could result in a more disturbed immune system. Another factor to consider when speculating about the origin of the enhanced circulating mRNA levels of IFN-signaling genes in Dutch South Asian women is adipose tissue. Total body fat percentage was slightly lower in Dutch South Asian women than in Dutch Europid women, and did overall not correlate with mRNA levels of IFN-signaling genes, indicating that total fat mass might not explain the found differences. However, previously it has been reported that healthy young South Asian individuals, compared to Europids, have larger subcutaneous adipocytes [43, 44] and increased adipose tissue macrophage infiltration [45], regardless of total body fat or visceral fat content. Hypertrophic adipocytes that are overloaded with triglycerides are at risk for hypoxia [46]. Induction of hypoxia-inducible factor-1 (HIF-1) and endoplasmic reticulum stress cause adipocyte cell death as well as the infiltration of immune cells from the innate and adaptive immune system, such as B cells, T cells, and macrophages resulting in a local and systemic proinflammatory environment [47, 48]. Hypothetically, dysfunctional adipose tissue, regardless of adipose tissue mass, may be driving insulin resistance and the progression toward T2D.
One of the strengths of our study is the large array of immune-related genes that were measured both in men and women, allowing us to detect transcriptional regulators and canonical pathways driving the observed differences in mRNA levels. Unfortunately, mRNA levels of immune genes were measured only in whole-blood samples, without isolating cells using flow cytometry analyses, lacking the ability to relate mRNA expression data with immune cell numbers and composition in blood. Furthermore, this could have been followed by ex vivo stimulation experiments on blood cells derived from both ethnicities to study the functionality of immune cells. Moreover, we did not measure inflammatory molecules in insulin target tissues, such as white adipose tissue and skeletal muscle, and as a result we cannot relate differences in circulating immune genes with those in metabolically active tissues. Similarly, we lack information on the menopausal state of women included in the present study. Concerning the study design, it remains a challenge to optimally match Dutch South Asian and Europid individuals. In this study, Dutch South Asians were younger, had a lower BMI, but had a longer diabetes duration, meaning a longer diabetic treatment period, which may have influenced the results. All these mentioned baseline differences can thus be confounders in the present study and results should be interpreted with this in mind. In addition, although medication use did not differ between ethnicities, we cannot exclude that use of prior medication (metformin, sulfonylurea, insulin, antihypertensives, and statins) may have influenced the results. Last, it is important to consider that an ethnicity is defined by cultural traditions. Therefore, we cannot exclude the influence of differences in behavior and lifestyle on the results in this study.
In conclusion, we show that circulating mRNA levels of IFN-signaling genes and B-cell markers are higher in Dutch South Asian than in Europid patients with T2D. We propose that increased inflammation, involving both B cells and IFN-signaling pathways, in Dutch South Asian patients with T2D is possibly contributing to the rapid progression of T2D and its complications in this population. Only future intervention studies can show whether targeting the IFN pathway for the treatment of T2D using anti-inflammatory therapies will be beneficial in the Dutch South Asian population. In this respect, the relatively novel antidiabetics glucagon-like peptide-1 receptor agonists and sodium-glucose cotransporter-2 inhibitors, which have been shown to exert cardiorenal protective effects, are particularly interesting.
## Financial Support
This work was supported by Novo Nordisk A/S (Bagsvaerd, Denmark), the Dutch Diabetes Foundation (No. 2015.81.1808 to M.R.B.), a Maria Zambrano fellowship by the Ministerio de Universidades y la Unión Europea–NextGenerationEU (No. RR_C_2021_04 to B.M.T.), and the Netherlands Cardiovascular Research Initiative: an initiative with support of the Dutch Heart Foundation (No. CVON2017 GENIUS-2 to P.C.N.R.).
## Disclosures
The authors have nothing to disclose.
## Data Availability
Some or all data sets generated during and/or analyzed during the present study are not publicly available but are available from the corresponding author on reasonable request.
## Clinical Trial Information
ClinicalTrials.gov numbers NCT01761318 and NCT02660047 (registered January 4, 2013).
## References
1. Sattar N, Gill JMR. **Type 2 diabetes in migrant south Asians: mechanisms, mitigation, and management**. *Lancet Diab Endocrinol* (2015) **3** 1004-1016
2. Shah A, Kanaya AM. **Diabetes and associated complications in the South Asian population**. *Curr Cardiol Rep* (2014) **16** 476. PMID: 24643902
3. McKeigue PM, Shah B, Marmot MG. **Relation of central obesity and insulin resistance with high diabetes prevalence and cardiovascular risk in South Asians**. *Lancet* (1991) **337** 382-386. PMID: 1671422
4. Bakker LEH, Sleddering MA, Schoones JW, Meinders AE, Jazet IM. **Pathogenesis of type 2 diabetes in South Asians**. *Eur J Endocrinol* (2013) **169** R99-R114. PMID: 23939919
5. Boon MR, Bakker LE, van der Linden RA. **High prevalence of cardiovascular disease in South Asians: central role for brown adipose tissue?**. *Crit Rev Clin Lab Sci* (2015) **52** 150-157. PMID: 25955567
6. Donath MY, Shoelson SE. **Type 2 diabetes as an inflammatory disease**. *Nat Rev Immunol* (2011) **11** 98-107. PMID: 21233852
7. Wu H, Ballantyne CM. **Metabolic inflammation and insulin resistance in obesity**. *Circ Res* (2020) **126** 1549-1564. PMID: 32437299
8. Wu H, Ballantyne CM. **Skeletal muscle inflammation and insulin resistance in obesity**. *J Clin Invest* (2017) **127** 43-54. PMID: 28045398
9. Larsen CM, Faulenbach M, Vaag A. **Interleukin-1-receptor antagonist in type 2 diabetes mellitus**. *N Engl J Med* (2007) **356** 1517-1526. PMID: 17429083
10. Fleischman A, Shoelson SE, Bernier R, Goldfine AB. **Salsalate improves glycemia and inflammatory parameters in obese young adults**. *Diabetes Care* (2008) **31** 289-294. PMID: 17959861
11. Goldfine AB, Fonseca V, Jablonski KA, Pyle L, Staten MA, Shoelson SE. **The effects of salsalate on glycemic control in patients with type 2 diabetes: a randomized trial**. *Ann Intern Med* (2010) **152** 346-357. PMID: 20231565
12. Goldfine AB, Fonseca V, Jablonski KA. **Salicylate (salsalate) in patients with type 2 diabetes: a randomized trial**. *Ann Intern Med* (2013) **159** 1-12. PMID: 23817699
13. Chambers JC, Eda S, Bassett P. **C-Reactive protein, insulin resistance, central obesity, and coronary heart disease risk in Indian Asians from the United Kingdom compared with European Whites**. *Circulation* (2001) **104** 145-150. PMID: 11447077
14. Forouhi NG, Sattar N, McKeigue PM. **Relation of C-reactive protein to body fat distribution and features of the metabolic syndrome in Europeans and South Asians**. *Int J Obesity* (2001) **25** 1327-1331
15. Petersen KF, Dufour S, Feng J. **Increased prevalence of insulin resistance and nonalcoholic fatty liver disease in Asian-Indian men**. *Proc Natl Acad Sci U S A* (2006) **103** 18273-18277. PMID: 17114290
16. Peters MJL, Ghouri N, McKeigue P, Forouhi NG, Sattar N. **Circulating IL-6 concentrations and associated anthropometric and metabolic parameters in South Asian men and women in comparison to European whites**. *Cytokine* (2013) **61** 29-32. PMID: 23026295
17. Bizino MB, Jazet IM, Westenberg JJM. **Effect of liraglutide on cardiac function in patients with type 2 diabetes mellitus: randomized placebo-controlled trial**. *Cardiovasc Diabetol* (2019) **18** 55. PMID: 31039778
18. van Eyk HJ, Paiman EHM, Bizino MB. **A double-blind, placebo-controlled, randomised trial to assess the effect of liraglutide on ectopic fat accumulation in South Asian type 2 diabetes patients**. *Cardiovasc Diabetol* (2019) **18** 87. PMID: 31288820
19. Bizino MB, Jazet IM, de Heer P. **Placebo-controlled randomised trial with liraglutide on magnetic resonance endpoints in individuals with type 2 diabetes: a pre-specified secondary study on ectopic fat accumulation**. *Diabetologia* (2020) **63** 65-74. PMID: 31690988
20. Assarsson E, Lundberg M, Holmquist G. **Homogenous 96-Plex PEA immunoassay exhibiting high sensitivity, specificity, and excellent scalability**. *PloS One* (2014) **9** e95192. PMID: 24755770
21. Straat ME, Martinez-Tellez B, van Eyk HJ. *J Clin Endocrinol Metab.* (2023) **108** 931-940
22. Bizino MB, Jazet IM, van Eyk HJ. **Efficacy of liraglutide on glycemic endpoints in people of Western European and South Asian descent with T2DM using multiple daily insulin injections: results of the MAGNA VICTORIA studies**. *Acta Diabetol* (2021) **58** 485-493. PMID: 33399989
23. van Dam AD, Hanssen MJW, van Eenige R. **South Asian men have lower expression of IFN signalling genes in white adipose tissue and skeletal muscle compared with white men**. *Diabetologia* (2017) **60** 2525-2528. PMID: 28887664
24. Lee AJ, Ashkar AA. **The dual nature of type I and type II interferons**. *Front Immunol* (2018) **9** 2061. PMID: 30254639
25. Schroder K, Hertzog PJ, Ravasi T, Hume DA. **Interferon-γ: an overview of signals, mechanisms and functions**. *J Leukocyte Biol* (2004) **75** 163-189. PMID: 14525967
26. Šestan M, Marinović S, Kavazović I. **Virus-induced interferon-γ causes insulin resistance in skeletal muscle and derails glycemic control in obesity**. *Immunity* (2018) **49** 164-177.e6. PMID: 29958802
27. McGillicuddy FC, Chiquoine EH, Hinkle CC. **Interferon γ attenuates insulin signaling, lipid storage, and differentiation in human adipocytes via activation of the JAK/STAT pathway**. *J Biol Chem* (2009) **284** 31936-31944. PMID: 19776010
28. Abdel-Moneim A, Semmler M, Abdel-Reheim ES, Zanaty MI, Addaleel W. **Association of glycemic status and interferon-γ production with leukocytes and platelet indices alterations in type2 diabetes**. *Diabetes Metab Syndr* (2019) **13** 1963-1969. PMID: 31235122
29. Kalafati M, Kutmon M, Evelo CT. **An interferon-related signature characterizes the whole blood transcriptome profile of insulin-resistant individuals—the CODAM study**. *Genes Nutr* (2021) **16** 22. PMID: 34886800
30. Nosratabadi R, Arababadi MK, Hassanshahi G. **Evaluation of IFN-gamma serum level in nephropatic type 2 diabetic patients**. *Pak J Biol Sci* (2009) **12** 746-749. PMID: 19634484
31. Hammad R, Elmadbouly AA, Ahmad IH. **T-natural killers and interferon gamma/interleukin 4 in augmentation of infection in foot ulcer in type 2 diabetes**. *Diabetes Metab Syndr Obes* (2021) **14** 1897-1908. PMID: 33958881
32. Winer DA, Winer S, Chng MHY, Shen L, Engleman EG. **B lymphocytes in obesity-related adipose tissue inflammation and insulin resistance**. *Cell Mol Life Sci* (2014) **71** 1033-1043. PMID: 24127133
33. Winer DA, Winer S, Shen L. **B cells promote insulin resistance through modulation of T cells and production of pathogenic IgG antibodies**. *Nat Med* (2011) **17** 610-617. PMID: 21499269
34. DeFuria J, Belkina AC, Jagannathan-Bogdan M. **B cells promote inflammation in obesity and type 2 diabetes through regulation of T-cell function and an inflammatory cytokine profile**. *Proc Natl Acad Sci U S A* (2013) **110** 5133-5138. PMID: 23479618
35. Straczkowski M, Kowalska I, Nikolajuk A, Krukowska A, Gorska M. **Plasma interleukin-10 concentration is positively related to insulin sensitivity in young healthy individuals**. *Diabetes Care* (2005) **28** 2036-2037. PMID: 16043753
36. van Exel E, Gussekloo J, de Craen AJ, Frölich M, Bootsma-Van Der Wiel A, Westendorp RG;. **Low production capacity of interleukin-10 associates with the metabolic syndrome and type 2 diabetes: the Leiden 85-Plus study**. *Diabetes* (2002) **51** 1088-1092. PMID: 11916930
37. Jagannathan M, McDonnell M, Liang Y. **Toll-like receptors regulate B cell cytokine production in patients with diabetes**. *Diabetologia* (2010) **53** 1461-1471. PMID: 20383694
38. Carr MC. **The emergence of the metabolic syndrome with menopause**. *J Clin Endocrinol Metab* (2003) **88** 2404-2411. PMID: 12788835
39. Mascarenhas-Melo F, Marado D, Palavra F. **Diabetes abrogates sex differences and aggravates cardiometabolic risk in postmenopausal women**. *Cardiovasc Diabetol* (2013) **12** 61. PMID: 23570342
40. Wannamethee SG, Papacosta O, Lawlor DA. **Do women exhibit greater differences in established and novel risk factors between diabetes and non-diabetes than men? The British Regional Heart Study and British Women's Heart Health Study**. *Diabetologia* (2012) **55** 80-87. PMID: 21861177
41. Murphy L, Sievert L, Begum K. **Life course effects on age at menopause among Bangladeshi sedentees and migrants to the UK**. *Am J Hum Biol* (2013) **25** 83-93. PMID: 23175465
42. Dratva J, Gómez Real F, Schindler C. **Is age at menopause increasing across Europe? Results on age at menopause and determinants from two population-based studies**. *Menopause* (2009) **16** 385-394. PMID: 19034049
43. Chandalia M, Lin P, Seenivasan T. **Insulin resistance and body fat distribution in South Asian men compared to Caucasian men**. *PLoS One* (2007) **2** e812. PMID: 17726542
44. Balakrishnan P, Grundy SM, Islam A, Dunn F, Vega GL. **Influence of upper and lower body adipose tissue on insulin sensitivity in South Asian men**. *J Invest Med* (2012) **60** 999-1004
45. Munoz A, Abate N, Chandalia M. **Adipose tissue collagen and inflammation in nonobese Asian Indian men**. *J Clin Endocrinol Metab* (2013) **98** E1360-E1363. PMID: 23780376
46. Hosogai N, Fukuhara A, Oshima K. **Adipose tissue hypoxia in obesity and its impact on adipocytokine dysregulation**. *Diabetes* (2007) **56** 901-911. PMID: 17395738
47. de Luca C, Olefsky JM. **Inflammation and insulin resistance**. *FEBS Lett* (2008) **582** 97-105. PMID: 18053812
48. Shoelson SE, Lee J, Goldfine AB. **Inflammation and insulin resistance**. *J Clin Invest* (2006) **116** 1793-1801. PMID: 16823477
|
---
title: 'Associations between self-reported sleep patterns and health, cognition and
amyloid measures: results from the Wisconsin Registry for Alzheimer’s Prevention'
authors:
- Lianlian Du
- Rebecca Langhough
- Bruce P Hermann
- Erin Jonaitis
- Tobey J Betthauser
- Karly Alex Cody
- Kimberly Mueller
- Megan Zuelsdorff
- Nathaniel Chin
- Gilda E Ennis
- Barbara B Bendlin
- Carey E Gleason
- Bradley T Christian
- David T Plante
- Rick Chappell
- Sterling C Johnson
journal: Brain Communications
year: 2023
pmcid: PMC9999364
doi: 10.1093/braincomms/fcad039
license: CC BY 4.0
---
# Associations between self-reported sleep patterns and health, cognition and amyloid measures: results from the Wisconsin Registry for Alzheimer’s Prevention
## Abstract
Previous studies suggest associations between self-reported sleep problems and poorer health, cognition, Alzheimer’s disease pathology and dementia-related outcomes. It is important to develop a deeper understanding of the relationship between these complications and sleep disturbance, a modifiable risk factor, in late midlife, a time when Alzheimer’s disease pathology may be accruing. The objectives of this study included application of unsupervised machine learning procedures to identify distinct subgroups of persons with problematic sleep and the association of these subgroups with concurrent measures of mental and physical health, cognition and PET-identified amyloid. Dementia-free participants from the Wisconsin Registry for Alzheimer’s Prevention ($$n = 619$$) completed sleep questionnaires including the Insomnia Severity Index, Epworth Sleepiness Scale and Medical Outcomes Study Sleep Scale. K-means clustering analysis identified discrete sleep problem groups who were then compared across concurrent health outcomes (e.g. depression, self-rated health and insulin resistance), cognitive composite indices including episodic memory and executive function and, in a subset, Pittsburgh Compound B PET imaging to assess amyloid burden. Significant omnibus tests ($P \leq 0.05$) were followed with pairwise comparisons. Mean (SD) sample baseline sleep assessment age was 62.6 (6.7). Cluster analysis identified three groups: healthy sleepers [$$n = 262$$ ($42.3\%$)], intermediate sleepers [$$n = 229$$ ($37.0\%$)] and poor sleepers [$$n = 128$$ ($20.7\%$)]. All omnibus tests comparing demographics and health measures across sleep groups were significant except for age, sex and apolipoprotein E e4 carriers; the poor sleepers group was worse than one or both of the other groups on all other measures, including measures of depression, self-reported health and memory complaints. The poor sleepers group had higher average body mass index, waist–hip ratio and homeostatic model assessment of insulin resistance. After adjusting for covariates, the poor sleepers group also performed worse on all concurrent cognitive composites except working memory. There were no differences between sleep groups on PET-based measures of amyloid. Sensitivity analyses indicated that while different clustering approaches resulted in different group assignments for some (predominantly the intermediate group), between-group patterns in outcomes were consistent. In conclusion, distinct sleep characteristics groups were identified with a sizable minority ($20.7\%$) exhibiting poor sleep characteristics, and this group also exhibited the poorest concurrent mental and physical health and cognition, indicating substantial multi-morbidity; sleep group was not associated with amyloid PET estimates. Precision-based management of sleep and related factors may provide an opportunity for early intervention that could serve to delay or prevent clinical impairment.
Du et al. identified sleep profile clusters based on self-reported sleep symptoms in late middle-aged adults. Sleep profiles were associated with differences in concurrent health outcomes, subjective reports of functioning and cognitive performance, but not PET amyloid estimates. Addressing sleep problems may benefit health and cognitive outcomes.
## Graphical Abstract
Graphical Abstract
## Introduction
Alzheimer’s disease, the most common cause of dementia, accounts for an estimated 60–$80\%$ of prevalent cases.1 Approximately $40\%$ of worldwide dementia cases are thought attributable to potentially modifiable risk factors,2 and emerging evidence suggests meaningful associations between various sleep disturbances (SDS) and dementia. The 2020 report of the Lancet Commission identified sleep as a putative risk factor for dementia2 with recent meta-analyses indicating that 60–$70\%$ of people with cognitive impairment or dementia have SDS3 including evidence that SDS are associated with a higher risk of all-cause dementia (RR 1.2; $95\%$ CI 1.1–1.3)4 and clinically diagnosed Alzheimer’s disease (1.6, 1.3–1.9)5 compared to those with no SDS. Midlife insomnia and late-life terminal insomnia or long sleep duration have also been associated with a higher late-life dementia risk.6 Some investigations report that short and long sleep durations are associated with worse outcomes for older adults including greater amyloid-β burden and cognitive decline,7 but other studies8-13 do not find these associations.
Concerning in this literature are the variable definitions of SDS. Evidence regarding the growing burden of SDS may be limited predominantly to measures of sleep duration,14-16 or be limited to one or two measures of sleep,6,7,9,17-20 or may be defined broadly, or rely on different self-reported sleep characteristics such as short or long sleep duration, poor sleep quality, circadian rhythm abnormality, insomnia, or obstructive sleep apnoea. Another potential contributor to divergent findings is that ‘SDS’ often serves as an umbrella term encompassing different aspects of sleep dysfunction (e.g. insufficient quantity and poor quality) or related impairment in daytime functioning (e.g. daytime sleepiness). Therefore, the need exists to systematically examine the association between multiple indicators of SDS with meaningful outcome measures including cognition and Alzheimer’s disease risk. In addition, SDS are associated with a multitude of health indicators including poor self-rated health,21 depression,22 subjective memory problems,23 increased body mass index (BMI)24 and insulin resistance (IR).25 The presence, aggregation and association of these conditions, also viewed as modifiable risk factors, with identified sleep problems and cognitive and disease outcomes remain to be determined.
Methodologically, many studies have used the traditional variable-centred approach, investigating relationships between two or more variables in a given sample. Clarification of the nature of the relationship between multiple variables of interest is important, but uncertainty regarding how identified relationships may apply to all study participants potentially places limits on their clinical applicability. In contrast, person-centred analysis, an alternative approach, focusses on the identification of subgroups of individuals based on the dependent variable(s) of interest as well as the aggregation of comorbid variables inherent in each subgroup.26 Harnessing the heterogeneity inherent in study populations and phenotyping them into more homogeneous ‘groups’ offer to improve ecological validity and clinical utility and to advance understanding of the co-occurrence and interplay among multiple risk factors in these discrete groups.27 To the best of our knowledge based on literature reviews, this is the first report to understand the association between the individual part of sleep and Alzheimer’s disease risk in a group where people are unimpaired at baseline. Person-centred analyses have been used in sleep studies including with targeted clinical populations (e.g. apnoea28), individuals with comorbid psychiatric (major depression29) or other health conditions (COVID-1930) and among diverse sociodemographic groups (children,31 Caribbean Blacks32 and Australian patients33). Only four studies19,34-36 examined clusters of sleep problems in the general population; however, none of them examined the association between sleep, cognition and Alzheimer’s disease risk reflected in amyloid-β burden. In addition, three studies applied latent class analysis which only works on categorical variables19,34,35 and included only one sleep measurement.19,35 Most SDS were assessed using continuous variables; therefore, applying K-means and latent profile analysis (LPA) (two person-centred methods37,38) to sleep problems offers a better opportunity to identify the distribution of persons with heterogeneous patterns of sleep and their linked mental, physical and cognitive comorbidities.
In this study, we examine cross-sectional associations between sleep, health, cognition and positron emission tomography (PET) amyloid indicators from the Wisconsin Registry for Alzheimer’s Prevention (WRAP), a well-characterized midlife cohort at risk for Alzheimer’s disease. Our first aim is to identify discrete subgroups or clusters of participants with variable patterns of sleep efficiency using person-centred methods. The second aim is to determine to what degree the identified sleep clusters are associated with concurrent measures of cognition, mental and physical health assessed by self-report and objective measures, as well as PET-assessed amyloid. We hypothesize that discrete subgroups of increasingly severe sleep abnormality would be identified with linked risks of health and cognitive abnormalities and amyloid positivity.
## Participants and study design
Participants were drawn from the WRAP, a longitudinal study designed to identify midlife factors associated with the development of Alzheimer’s disease.39,40 Enrolment of participants began in 2001, with the first follow-up visit occurring 2 to 4 years after the baseline visit and all additional visits occurring at 2-year intervals thereafter. WRAP participants were free of dementia at enrolment (mean age 54 years). All study procedures were approved by the University of Wisconsin School of Medicine and Public Health Institutional Review Board and are in concordance with the Declaration of Helsinki.
At each study visit, participants completed comprehensive neuropsychological assessment and multiple questionnaires related to a broad array of factors, including lifestyle, modifiable risk factors, medical history and memory functioning. Sleep measures were added in two stages to the WRAP assessment protocol. To be eligible for the primary analyses, participants needed to have completed the full set of sleep measures at least once and be free of dementia at time of sleep assessment ($$n = 619$$). To be eligible for secondary analyses, participants needed to have completed at least one of the sleep questionnaires described below and had completed a Pittsburgh Compound B (PiB) PET scan.
## Sleep assessment
The assessment protocol was expanded in 2012 to incorporate two self-report sleep measures [the Medical Outcomes Study Sleep Scale (MOS)41 and the Epworth Sleepiness Scale (ESS)42]; in 2014, the Insomnia Severity Index (ISI)43 was added to a specific visit, and in 2016, these assessments were included at all study visits.
## The Sleep Scale from the Medical Outcomes Study
This scale comprises 12 questions about the past 4 weeks, from which eight scores were computed.44 The first question asks how long it takes to fall asleep, with possible responses in 15-min increments ranging from 1 = ‘0–15 minutes’ to 5 =‘More than 60 minutes’.44 The second question asks the average number of hours slept each night, which is entered freely.44 Responses to the remaining 10 questions are on a 6-point scale ranging from 1 = ‘all of the time’ to 6 = ‘none of the time’.44 Responses are summed to give scores for six sleep domains: SDS, somnolence (SOM), sleep adequacy (ADQ), snoring, awaking short of breath or with a headache, and two indices of sleep problems summarizing six (Index I) (SPI1) or nine (Index II) (SPI2) items.45 Multi-item scores show good internal consistency, with Cronbach’s alpha 0.71 to 0.81.46Supplementary Table 1 indicates which items contribute to each score, with some items contributing to more than one score. We define people to have the optimal sleep if 7 h ≤self-reported sleep duration ≤8.47
## The Epworth Sleepiness Scale
The ESS42 assesses sleep propensity and daytime sleepiness.44 Participants rate how likely they are to doze off or fall asleep in eight common situations that vary in their soporific qualities, such as watching TV, talking to someone or lying down.44 Responses are on a 4-point scale ranging from 0 = ‘would never doze’ to 3 = ‘high chance of dozing’. Responses are summed to produce a total score ranging from 0 to 24, with higher scores indicating greater daytime sleepiness.44 The ESS has been shown to have good internal consistency (Cronbach α = 0.73–0.88) and test–retest reliability (correlation of measures across a 5-month interval = 0.82).42 A threshold of ≥11 as defined here (http://epworthsleepinessscale.com/about-the-ess/) is applied for determining ESS abnormal.
## The Insomnia Severity Index
Insomnia severity is assessed with the ISI,43 a validated clinical measure that asks about symptoms of insomnia in the last 2 weeks. The first three items assess early, middle and late insomnia symptoms. The last four items measure sleep satisfaction/dissatisfaction, SDS noticeability, sleep worry and sleep interference with daily life, respectively. For these items, Likert scores of 0 represent ‘very satisfied’ or ‘not at all worried/noticeable/interfering’, whereas scores of 4 represent ‘very dissatisfied’ or ‘very much worried/noticeable/interfering’. Its internal consistency, concurrent validity and sensitivity to clinical improvements in insomnia patients are well established.43 Responses are summed to produce a total score ranging from 0 to 28, with higher scores indicating increasing insomnia severity. A threshold of ≥10 as defined here48 is applied for determining ISI abnormal.
All sleep scores are average scores across the multiple items and inverted (higher score means better sleep).
## Sleep disorder diagnosis information
The presence of diagnosed sleep disorders, such as insomnia, restless leg syndrome and obstructive sleep apnoea, is determined by the questions ‘Have you ever been told by a doctor or other health professional that you have any of the following?’ and ‘Do you use a Continuous positive airway pressure (CPAP) machine or other appliance when you sleep to treat your sleep apnoea if you have apnoea?’.
## Apolipoprotein E, cognitive composites and cognitive status
Apolipoprotein E (APOE) genotype is expressed as a binary categorical variable, with participants classified as carriers (one or more ɛ4 alleles present) or non-carriers (no ɛ4 allele present).
As sleep may affect various aspects of cognition differently,49 we include five cognitive composite indices, reflecting the average of domain-specific standardized test scores (Z-scores) administered as part of the WRAP battery.40 The cognitive composites include working memory,50 immediate memory, delayed memory, executive function (EF)51 and a Preclinical Alzheimer Cognitive Composite (PACC).52,53 The tests contributing to each composite are shown in Supplementary Table 2.
Cognitive status is determined for each visit using a consensus review process that incorporated internal as well as published norms. A multi-disciplinary panel reviews cases to determine whether mild cognitive impairment or dementia was present.54
## Self-reported health measurements
Self-rated health55 is measured using a 5-point scale (1 = poor, 2 = fair, 3 = good, 4 = very good and 5 = excellent) in response to the question ‘How would you rate your current health?’. Depressive symptoms [20-item Center for Epidemiologic Studies Depression Scale (CES-D)]56 are completed by each participant. Self-rated memory is measured using a 7-point scale in response to the question ‘Overall, how would you rate your memory in terms of the kinds of problems that you have?’. The scores are summarized as 1–3 = major problems, 4 = neutral and 5–7 = no problems.
## Objective health measurements
Two measures of obesity [BMI and waist–hip ratio (WHR)] are calculated. To evaluate BMI (/m2), height (m) and weight (kg) are measured. The WHR is calculated using the circumferences of the two target areas (waist and hip). The homeostatic model assessment of insulin resistance (HOMA-IR) is used to measure IR and is calculated as follows: fasting insulin (μU/mL) × fasting glucose (mmol/L)/22.5. A high HOMA-IR denotes low insulin sensitivity. HOMA-IR values are log-transformed into normally distributed values prior to analysis.
## Medication use
Medication data is obtained at each visit through a combination of self-report, medical records and research staff review of medications brought to the study visit. The total number of prescriptions is counted for each participant at each visit, up to 15 per type.57
## Pittsburgh Compound B PET imaging
A subset of WRAP participants complete [11C]PiB58 amyloid PET imaging at the University of Wisconsin—Madison Waisman Brain Imaging Laboratory. Detailed imaging methods have been previously described.59,60 Amyloid burden is assessed as global cortical PiB distribution volume ratio (DVR)20 for continuous analyses, and one DVR threshold of ≥1.2 as defined previously61 is applied for determining PiB positivity (A+). Estimated amyloid chronicity (i.e. estimated years A+) is calculated at time of sleep assessment using previously published methods.62,63
## Statistical analysis
All analyses were performed in R version 4.0.0. For all analyses, significant omnibus tests ($P \leq 0.05$) were followed with unadjusted pairwise comparisons.
## Aim 1 analyses
In preparation for our cluster analyses, all sleep variables were mean-centred and scaled such that higher scores indicated better sleep. We then used K-means cluster analysis to characterize subgroups of sleep variables in WRAP participants ($$n = 619$$), conducted using ‘factoextra’ package in R.64 The cluster assignment was based on the minimum distance (sum of the deviation of each variable) of a participant from the centroid of the cluster. The optimal number of clusters was identified using the elbow method by looking at the total within-cluster sum of square (WSS). To characterize the sleep group for each clustering-based subgroup of participants, the effect size (ε2) of the sleep problems used in cluster analysis was noted in the right column of Supplementary Table 3. The relative contributions of the different problems in the grouping of participants were large, medium and small when ε2≥ 0.26, ε2≥ 0.08 and ε2≥ 0.01, respectively.65 Given the high correlation among sleep variables, we conducted preliminary cluster analyses, sequentially excluding subsets of the scales and examining fit statistics and consistency across solutions. Based on the best WSS and Calinski–Harabasz Index values, the following subset of scales was selected in primary analyses: SPI1, SDS, ADQ, SOM, self-reported sleep duration, ESS and ISI.
To characterize how sleep groups differed across sleep characteristics, we used chi-square for categorical variables and Kruskal–Wallis tests for Likert-scale variables [median (Q1–Q3) reported]. Post hoc pairwise group differences at unadjusted $P \leq 0.05$ were reported.
Three sensitivity analyses were conducted to investigate the consistency of sleep group assignments and to examine whether between sleep group patterns in our outcomes were stable across different sample selection criteria. Alternative 1: we used LPA to characterize sleep subgroups (‘Mclust’ package in R). Briefly, LPA was a data-driven approach using continuous variables and indicators to identify subgroups of individuals. In this statistical approach, subgroup membership was determined by examining the pattern of interrelationships among indicator variables (maximizing homogeneity within each subgroup and heterogeneity between subgroups).66 Alternative 2 (cognitively unimpaired subset only): we reduced the original set to include only those who were cognitively unimpaired ($$n = 21$$ with mild cognitive impairment were removed; leaving $$n = 598$$), and K-means cluster analysis was used in this subset. Alternative 3 (expanded set with imputed ISI): as previously noted, the primary cluster analysis was based on the first visit with MOS, ESS and ISI. Since the MOS and ESS questionnaires were added to the battery several years before the ISI, we opted to enlarge ‘baseline sleep’ in sensitivity analyses to include those who had not yet completed an ISI but had completed MOS and ESS at least once. The imputation method used the sleep data on a person both before and after the ‘missing value’. The next observation carried backward assigned the person’s next known sleep score after the ‘missing’ one to the ‘missing value’. If the person did not have the next value, the last observation carried forward, assigned the person’s last previous known sleep score to the ‘missing value’, was used.67 The resulting enlarged set included $$n = 1237$$ available.
## Aim 2 analyses
To analyse differences between sleep groups in demographic characteristics and concurrent health measures ($$n = 619$$), we used chi-square or Fisher’s exact test for categorical variables, analysis of variance (ANOVA) for continuous variables [mean (SD) reported] and Kruskal–Wallis tests for Likert-scale variables [median (Q1–Q3) reported]. Post hoc pairwise group differences at unadjusted $P \leq 0.05$ were reported. We excluded people who took insulin medication ($$n = 9$$) when comparing the HOMA-IR difference across the sleep groups (see Wallace et al.68). Linear regression was used to assess the relationship between sleep group and concurrent cognitive composite scores after adjusting for covariates [age, sex, education, WRAT3 reading score and the number of prior exposures to the cognitive tests (the practice effect)].
Similarly, to analyse differences between sleep groups and concurrent amyloid burden, we examined data from the subset that had completed at least one PiB PET study [n(%) = 108 ($17.4\%$)]. Kruskal–Wallis tests were used to assess the difference between sleep groups in estimated concurrent global PiB DVR and amyloid chronicity, and Fisher’s exact test was used to analyse the concurrent amyloid PET status difference between sleep groups. In sensitivity analyses, we tested whether there was significant difference of amyloid burden at the most recent PET scan across the alternative sleep group assignments. In the imputed data set, 285 ($23.0\%$) had at least one PiB PET scan, and we tested the difference in estimated concurrent and most recent global PiB DVR and amyloid chronicity among sleep groups.
We compared corrected Akaike information criteria (AICc) model fit statistics across otherwise identical models and considered |ΔAICc| values <2 to represent comparable models. Linear regression was performed for the association between sleep groups and concurrent cognitive composite scores after we removed stroke ($$n = 10$$), epilepsy/seizures ($$n = 13$$), multiple sclerosis ($$n = 5$$) and Parkinson’s disease ($$n = 2$$). Since APOE genotype associates with cognition, additional linear regression was performed including APOE e4 carriers in the model, and we compared model fits with the fits of the model in Aim 2 with the participants who have APOE data ($$n = 538$$). ΔAICc values were reported.
## Aim 3 analyses
Last, we replicated analyses from an earlier WRAP PET publication.20 That study showed significant yet small associations between less adequate sleep, more sleep problems and greater SOM on the MOS and greater amyloid PET burden in Alzheimer’s disease–sensitive brain regions among 98 cognitively unimpaired adults (aged 62.4 ± 5.7 years) at their fourth WRAP visit. Participants were identified for the present analysis if they had completed WRAP Visit 4 (including sleep assessment), had completed a PiB PET scan and were non-demented; 315 individuals met these inclusion criteria. To match with the data set in Sprecher et al., we then excluded 95 people, leaving a sample size $$n = 220$.$ We performed the same linear regression previously performed in Sprecher et al. ,20 including age, sex, APOE e4 genotype, family history of Alzheimer’s disease and BMI as covariates.
## Data availability
All data and materials used within this study will be made available, upon reasonable request, to research groups wishing to reproduce/confirm our results.
## Identifying sleep groups
A total of 619 participants were eligible for Aim 1 analyses. Mean (SD) sleep baseline age was 62.6 (6.7). Correlations among sleep variables are shown in Supplementary Fig. 1 and range from 0.87 (SPI1 and ADQ) to 0.15 (ESS and SDS).
## Sleep groups identified by K-means cluster analysis (primary)
K-means cluster analysis identified three groups: healthy sleepers [HS, $$n = 262$$ ($42.3\%$)], intermediate sleepers [IS, $$n = 229$$ ($37.0\%$)] and poor sleepers [PS, $$n = 128$$ ($20.7\%$)]. The cluster solution and differences in contributing variables were shown in Fig. 1. Differences across sleep groups were shown for all sleep variables in Supplementary Table 3. In post hoc pairwise comparisons, all sleep group pairs differed at $P \leq 0.001.$ The relative contributions of all cluster analysis variables except ESS in the grouping of participants were large, and the three largest contributors were SPI1, ADQ and ISI. PS group also had a lower score on additional sleep variables, including more self-reported restless leg syndrome and apnoea.
**Figure 1:** *K-means clustering of participants. Left panel: Clusters are distributed along the principal components. Observations are represented by points in the plot. Right panel: Mean Z-scores of each variable within each group. The Z-scores are calculated for the current sample, yielding a sample sum of 0 and a standard deviation of 1; thus, the groups tend to approximately mirror each other around the y = 0 axis when the group sizes are similar. SPI1, Sleep Problem Index I; SDS, sleep disturbance scale; ADQ, sleep adequacy; SOM, somnolence; SRSD, self-reported sleep duration; ESS, Epworth Sleepiness Scale; ISI, Insomnia Severity Index. The optimal number of clusters given by the elbow method was 3. The total within-cluster sum of square (WSS) was 2282.454.*
As shown in Table 1, there were no significant differences between sleep groups on age, sex and APOE e4 carrier status. The PS group was slightly more racially diverse and had lower mean education years and lower WRAT3 reading.
**Table 1**
| Unnamed: 0 | Healthy sleepers (N = 262) | Intermediate sleepers (N = 229) | Poor sleepers (N = 128) | P-valuea | Difference pairs |
| --- | --- | --- | --- | --- | --- |
| Age (years) [mean (SD)] | 63.24 (6.77) | 62.25 (6.68) | 62.07 (6.66) | 0.151 | |
| Male (%) | 74 (28.2) | 80 (34.9) | 33 (25.8) | 0.129 | |
| Race (non-Hispanic White) (%) | 25 (9.54) | 191 (16.6) | 93 (27.3) | <0.001 | All pairs |
| College (%) | 179 (68.3) | 141 (61.6) | 56 (43.8) | <0.001 | PS versus HS, IS |
| APOE e4 carriers positive (%) | 78 (33.2) | 85 (41.5) | 44 (44.9) | 0.072 | |
| WRAT3 reading [median (IQR)] | 109.00 (102.00, 115.00) | 109.00 (100.50, 115.00) | 105.00 (98.00, 112.00) | <0.001 | PS versus HS, IS |
| Concurrent health | | | | | |
| CES-D score [median (IQR)] | 2.00 (1.00, 5.00) | 6.00 (3.00, 11.00) | 12.00 (7.00, 18.00) | <0.001 | All pairs |
| SRH | | | | <0.001 | All pairs |
| Poor | 0 (0.0) | 4 (1.8) | 2 (1.6) | | |
| Fair | 8 (3.1) | 15 (6.6) | 22 (17.3) | | |
| Good | 80 (30.7) | 105 (46.5) | 62 (48.8) | | |
| Very good | 131 (50.2) | 88 (38.9) | 35 (27.6) | | |
| Excellent | 42 (16.1) | 14 (6.2) | 6 (4.7) | | |
| Num_prescription [median (IQR)] | 2.00 (0.00, 4.00) | 3.00 (1.00, 5.00) | 4.00 (1.00, 7.00) | <0.001 | All pairs |
| Self-rated memory | | | | <0.001 | All pairs |
| Major problems | 11 (4.2) | 33 (14.4) | 31 (24.2) | | |
| Neutral | 35 (13.4) | 44 (19.2) | 30 (23.4) | | |
| No problems | 216 (82.4) | 152 (66.4) | 67 (52.3) | | |
| Mild cognitive impairment (%) | 5 (2.1) | 8 (3.7) | 8 (3.7) | 0.089 | |
## Sleep groups identified in sensitivity analyses
As noted in the ‘Methods’ section, we used three additional approaches to identify alternative sleep group assignments. Each identified three groups with varying distributions. Alternative 1 (LPA): HS [$$n = 214$$ ($34.6\%$)], IS [$$n = 198$$ ($32.0\%$)] and PS [$$n = 207$$ ($33.4\%$)] (Supplementary Fig. 2). Alternative 2 (restricting sample to the unimpaired subset): HS [$$n = 257$$ ($43.0\%$)], IS [$$n = 220$$ ($36.8\%$)] and PS [$$n = 121$$ ($20.2\%$)]. Alternative 3 (larger imputed set): HS [$$n = 589$$ ($47.6\%$)], IS [$$n = 429$$ ($34.7\%$)] and PS [$$n = 219$$ ($17.7\%$)]. The participant demographic, concurrent health and sleep variables in these 619 participants (original data), unimpaired subset and imputed set were similar (shown in Supplementary Table 4). The Z-scores of these sleep variables across the three groups identified by K-means and LPA in these data sets were similar (Supplementary Figs. 3 and 4). The disagreements between K-means and LPA cluster results and 598 participants in three data sets were shown in Supplementary Fig. 5.
## Examining associations between sleep groups and concurrent health
The PS group had the worse concurrent health, including self-reported depression, health and memory complaints, than the HS and/or IS groups (Table 1). PS had higher BMI, WHR and HOMA-IR than HS and IS (Fig. 2).
**Figure 2:** *Comparison of metabolic biomarkers at sleep baseline among sleep profiles. (A) BMI, (B) WHR and (C) HOMA-IR. HOMA-IR is calculated by glucose (mg/dL) × insulin (mIU/mL)/405 (excluding people on insulin therapy). In the figure, HOMA-IR is log-transformed. BMI, body mass index; WHR, waist–hip ratio; HOMA-IR, homeostatic model assessment of insulin resistance. Post hoc significant pairwise group differences at unadjusted P < 0.05 are shown in the figure.*
Sensitivity analyses comparing the concurrent health variables across sleep clusters described in the ‘Methods’ section showed that patterns for the LPA groups were the same in BMI compared with the primary K-means groups (Supplementary Fig. 6); patterns of association using the second and third clustering alternatives were consistent to those in the primary sleep group analyses (Supplementary Figs. 7 and 8); the only observed discrepancies were that differences between HS and PS for WHR (unimpaired subset) and between IS and PS for BMI (imputed ISI) were not significant.
## Examining associations between sleep groups and objective cognitive performance
All omnibus tests of sleep groups’ differences on cognitive composite scores were significant. After adjusting for covariates, PS performed worse on all cognitive outcomes except working memory (Fig. 3). The IS group performed worse on average than the HS group on PACC3 (Cohen’s $d = 0.32$), immediate learning (Cohen’s $d = 0.27$) and delayed recall (Cohen’s $d = 0.31$). The same pairwise differences were observed for PS versus HS with these effect sizes: PACC3, 0.51; immediate learning, 0.36; and delayed recall, 0.38. Last, the PS group performed worse on average than HS (effect size = 0.37) and IS (effect size = 0.25) on EF.
**Figure 3:** *Comparison of cognitive composite scores among sleep profile. PACC3 = Preclinical Alzheimer’s Cognitive Composite (comprised of averaged Z-scores for three tests: Auditory Verbal Learning Test (AVLT) Total Learning, Logical Memory Delayed Recall and Digit Symbol Substitution); EF composite = averaged Z-scores for Stroop Color–Word, Trail Making Test Part B and Wechsler Adult Intelligence Scale—Revised (WAIS-R) Digit Symbol; immediate learning composite = averaged Z-scores for AVLT Total Learning, Wechsler Memory Scale-R Logical Memory-I and Brief Visuospatial Memory Test—Revised (BVMT-R); delayed recall composite = averaged Z-scores for AVLT Delayed Learning, Wechsler Memory Scale-R Logical Memory-II and BVMT-R Delayed. All composite scores are covariate-adjusted (age, sex, WRAT3 reading, college and practice effect). Analysis of covariance was used to compare the difference of composite scores in healthy sleepers (HS), intermediate sleepers (IS) and poor sleepers (PS). P-value *0.05, **0.01 and ***0.001.*
The number of participants with stroke, epilepsy/seizures, multiple sclerosis and Parkinson’s disease in HS/IS/PS group was $\frac{3}{5}$/2, $\frac{3}{5}$/5, $\frac{2}{1}$/2 and $\frac{1}{1}$/1, separately. After excluding these conditions, linear regression showed similar results in Aim 2 except the PS group versus IS group on EF was not significant. The effect size between IS and HS group increased from 0.32 to 0.36 on PACC3, from 0.27 to 0.30 on immediate learning and from 0.31 to 0.33 on delayed recall, and the effect size between PS and HS decreased by 0.06 on PACC3, 0.02 on immediate learning and 0.04 on delayed recall. In sensitivity analyses, linear regression was performed in alternative unimpaired subset and imputed data set. In the subset of primary data set ($$n = 538$$) who have APOE data, LM model of cognitive composite scores showed better fit except EF after adding the APOE e4 carriers to the model, including covariates and age terms (AICc decreased 1.2 on PACC3 model, AICc decreased 3.4 on immediate learning model, AICc decreased 3.7 on delayed recall model and AICc increased 0.15 on EF model). All the regression results were shown in Supplementary Tables 5–9.
## Examining associations between sleep groups and PiB PET amyloid
In the data set used for the primary analyses ($$n = 619$$), 109 ($17.4\%$) had at least one PiB PET scan. There were no significant differences between sleep groups on estimate DVR at the time of sleep assessment (Fig. 4). The difference in amyloid status between sleep groups was not statistically significant. After replicating the analyses using the sleep clusters from LPA (Supplementary Fig. 9), in the unimpaired subset (Supplementary Fig. 10) or in the larger PiB PET scan sample ($$n = 287$$) (Supplementary Fig. 11), the results were consistent.
**Figure 4:** *Comparison of PET amyloid measures at sleep baseline among sleep profiles. (A) Estimates build on the PiB DVR and (B) amyloid chronicity described in Koscik et al.62 The number (%) of PiB(+) in healthy sleepers (HS), intermediate sleepers (IS) and poor sleepers (PS) is 9 (17.3%), 7 (17.5%) and 1 (6.2) (Kruskal–Wallis χ2 = 1.26, P = 0.53). PiB(+) defined as any estimate PiB DVR within a person ≥1.2. The mean (95% CI) estimate age PiB positive in HS, IS and PS is 77.61 (73.09, 81.97), 78.28 (75.51, 81.44) and 80.93 (75.61, 85.38). PiB, Pittsburgh Compound B; DVR, distribution volume ratio. Post hoc significant pairwise group differences at unadjusted P < 0.05 are shown in the figure.*
## Replication study comparing current data set with previous results
Participant demographic and cognitive characteristics in the prior20 study and in this replication study with bigger sample size were compared in Table 2. The mean age in this replication study was 63.7 years (SD = 5.78, range = 50.2–74) at the time of the PiB PET scan, 1.3 years on average older than previous study. Mean interval between PET scan and questionnaire completion was 1.28 (SD 1.10) years, 0.59 years longer than previous study, and the association did not change when interval was added as a covariate. Sex, years of education, BMI, CES-D and cognitive scores were similar.
**Table 2**
| Unnamed: 0 | Sprecher et al.a (N = 98) | Replication sample (N = 220b) |
| --- | --- | --- |
| Age at PiB PET scan, years | 62.4 (5.7; 50–73) | 63.7 (5.8; 50.2–74.0) |
| Age at sleep assessment, years | 63.0 (5.6; 51–73) | 63.6 (5.6; 51.4–73.7) |
| Interval between PiB PET scan and sleep assessment, years | 0.69 (0.98; 0–3.7) | 1.28 (1.10; 0–3.6) |
| Female, % | 67.3 | 66.0 |
| APOE e4 positive, % | 34.7 | 38.6 |
| FH positive, % | 75.5 | 74.1 |
| BMI, kg/m2, mean (SD) | 28.7 (5.7) | 28.6 (5.7) |
| Education, years | 16.6 (2.8; 12–25) | 16.2 (2.1; 12–20) |
| CES-D | 5.78 (5.48; 0–27) | 5.76 (5.56; 0–27) |
| MMSE | 29.31 (1.22; 23–30) | 29.38 (0.96; 24–30) |
| AVLT total | 50.21 (8.66; 30–67) | 50.51 (8.68; 28–67) |
| AVLT delayed recall | 10.36 (2.96; 0–15) | 10.63 (2.83; 1–15) |
| Trails A timec | 10.11 (2.18; 5–17) | 10.44 (2.45; 5–17) |
| Trails B timec | 10.26 (2.51; 6–17) | 10.56 (2.48; 3–17) |
| Digit symbol scaled scorec | 13.35 (2.1; 9–19) | 13.30 (2.1; 8–19) |
| PiB DVR | | 1.14 (0.18; 0.9–2.07) |
After adjusting for covariates (age, sex, APOE e4 status, family history of Alzheimer’s disease and BMI), poorer sleep was not significantly associated ($P \leq 0.05$) with greater PiB DVR on any of the sleep measures examined. The results were summarized in Supplementary Table 10 together with the Sprecher results. Scale score versus PiB DVR were plotted in Fig. 5 for the three variables with smallest P-values in the Sprecher et al. paper. While the previous publication reported significant associations between PiB DVR and ADQ [beta (SE) = −0.002 (0.001), $$P \leq 0.014$$] and SOM [beta (SE) = 0.003 (0.001), $$P \leq 0.033$$], these associations were weaker and NS in the current PiB PET data set [ADQ beta (SE) = −0.001 (0.0006), $$P \leq 0.06$$; SOM beta (SE) = 0.0006 (0.0007), $$P \leq 0.38$$)].
**Figure 5:** *Association between sleep adequacy, somnolence, Problem Index I scores and mean PiB DVR. (A) Results from Sprecher’s paper (top panel). (B) Replication results (bottom panel). Raw data is plotted, and the regression line with ribbon is adjusted for age, sex, the epsilon 4 allele of the apolipoprotein E gene, family history of Alzheimer’s disease and body mass index (the line without ribbon is the regression line without covariates adjustment). Per prior research,20 the sleep scores (x-axis) were scaled to a range of 0–100 with higher values indicating more of the concept being measured (higher score is better for sleep adequacy, worse for somnolence and Problem Index I scores). The mean PiB DVR value (y-axis) is distribution volume ratio, which reflects the equilibrium distribution of PiB. The test statistic values alongside P-values from separate regression models with covariate adjustment are shown in the figure. Results were considered statistically significant when P < 0.05. DVR, distribution volume ratio; PiB, Pittsburgh Compound B.*
## Discussion
In this study, we applied person-centred data-driven analysis techniques to cognitively unimpaired WRAP participants’ self-reported sleep characteristics. This novel approach allowed us to identify more homogeneous subgroups that differed on self-reported sleep quality, insomnia and daytime sleepiness. We then demonstrated how these HS, IS and PS subgroups also differed in concurrent health, cognition and amyloid PET. Four major findings resulted, and each finding was discussed in greater detail below. Overall, the evidence suggested that conceptualizing sleep as a single dimension or the total score of multiple items could not fully capture the changes and variability in self-reported sleep patterns and supported the multi-dimensional sleep health perspective.69 Participants in the HS group reported better concurrent mental and physical health and demonstrated better cognition than those in the other groups. The PS group was more likely to report increased depressive symptoms, more physical health problems and worse cognition. However, there were no significant associations between the sleep variables and amyloid PET measures.
## Finding 1: sleep groups differed on self-reported sleep quality and select demographics
Three groups of participants were identified which were heterogeneous with respect to self-reported sleep characteristics. The group with the highest proportion of participants was HS ($42.3\%$), which exhibited sufficient standings on all measured sleep characteristics. MOS scales SPI1, ADQ and ISI were the three sleep dimensions that contributed most to the cluster analysis. The PS group with more sleep problems, lower ADQ and more serious insomnia was slightly more racially diverse and had overall lower education levels and lower estimated premorbid ability level. A review of the association between race/ethnicity and sleep patterns summarized that the racialized group had objectively measured and self-reported worse sleep duration and quality than Whites.70 Other studies also reported race and education associations with sleep quality.34,71-73 For example, Stamatakis and colleagues71 found that short sleep duration was more common among those with lower education levels and among racialized racial and ethnic groups. Racialized groups were more likely to experience social disadvantage due to historical and contemporary forms of race-based institutional and interpersonal discriminatory policies and practices.70 The social disparities in sleep patterns provided further evidence that these disparities might be associated with disparities in other areas, such as cardiovascular and metabolic health.74,75 But more data were needed to assess the social disparities in sleep. There were no significant differences between sleep groups in age, sex and APOE e4 carriers. One study32 reported that younger men were more likely than their older counterparts to be assigned to the ‘poor sleep quality’ class. The mean age in that study was 42.7, which was much younger than our study. A review of recent studies showed women tend to experience the most significant sleep problems during the peri-menopausal period.76 However, given the age of our sample, nearly all women would have been post-menopausal at the time of the sleep data. Our study was consistent with Drogos et al.77 who reported no relationship between the presence of APOE e4 allele and subjective sleep complaints in a healthy population screened for dementia.
## Finding 2: sleep groups differed on concurrent health outcomes and subjective reports of functioning
The PS group had higher measured BMI, WHR and HOMA-IR than the HS and IS groups. These associations were also reported in other studies using variable-centred approaches.78-80 For example, shorter sleep duration was associated with higher BMI in a sample of 1042 individuals from Brazil, including both genders (20–80 years)80; partial sleep deprivation during only a single night induced IR in multiple metabolic pathways in nine healthy subjects.79 We also observed associations between sleep groups and concurrent self-reported health, depression and memory, such that those with the poorest self-reported sleep characteristics were worse on all these measures. Again, these results were consistent with other studies that used variable-centred approaches (e.g. Tsapanou et al., Paunio et al. and Carpi et al.23,81,82). For example, the onset of poor sleep predicted incident depression in a sample of 12 063 individuals using logistic regression models.81 But our study showed the co-aggregation of these risk variables using person-centred approaches. Some associations were found using different person-centred approaches in older adults.19,35 For example, Yu and colleagues distinguished four sleep profiles based on self-report measures of sleep problems: inadequate sleep, disturbed sleep, trouble falling asleep and multiple problems, using latent class analysis in a community sample of elderly (mean age 67 years). The ‘multiple problems’ group had significantly higher levels of depression and anxiety relative to the control group. However, our study was in a longitudinal preclinical study enriched with Alzheimer’s disease parental history from midlife, which was important to show these associations in this sample.
## Finding 3: sleep groups differed in objective cognitive performance
After adjusting for covariates (age, sex, college and WRAT3 reading score), PS performed worse on all cognitive outcomes except working memory. Poor sleep was considered a potential risk factor for cognitive decline,83,84 extreme sleep durations in later life were associated with worse average cognition in female nurses aged 70 and older free of stroke and depression at the initial cognitive assessment and poor sleep quality was associated with mild cognitive impairment in a group of 1793 participants ($51\%$ men; 63.8 ± 7.5 years) of the population-based Heinz Nixdorf Recall study. Observational studies3,5,85-87 with self-report sleep measures supported links between sleep and cognitive decline using variable-centred approaches. But the results were mixed. For example, one study found self-reported poor sleep quality was related to cognitive changes, whereas daytime sleepiness was not related.18 In contrast to other studies,14,15,88 McSorley et al.9 did not find evidence that self-reported sleep duration was a significant contributor to cognitive function. Cross et al.89 found adults with probable insomnia disorder exhibited declarative memory deficits compared with insomnia symptoms only or no insomnia symptoms; however, adults with insomnia symptoms exhibited better performance on a task of mental flexibility than both other groups. However, these studies assumed that all individuals at certain levels of risk factors were at equal risk of adverse outcome and therefore the association between a risk factor and outcome was the same across the entire population.
## Finding 4: sleep groups were not associated with PET measures of amyloid
No significant associations between sleep and amyloid PET measures were found using our person-centred approach or when we repeated a previous variable-centred analysis. Specifically, a previous study from this cohort found that in cognitively healthy adults ($$n = 98$$), less adequate sleep, more sleep problems and greater SOM were associated with higher amyloid burden in the angular gyrus, frontal medial orbital cortex, cingulate gyrus and precuneus when these individual sleep variables were included in separate models that adjusted for age, sex, APOE e4, family history of Alzheimer’s disease and BMI.20 In Aim 3, using the larger sample size now available, none of the individual sleep variables were significantly associated with amyloid burden. As shown in Fig. 5, we had a wider range of values on the SOM and the SPI1 than the values in the sample for the 2015 paper; associations remained non-significant after removing these higher points in sensitivity analyses. Similar to our current results, a study of 143 community-dwelling participants aged ≥70 years12 found no significant relationship between amyloid-PET burden and nighttime sleep duration, daytime sleep duration, 24-h sleep duration, naps, restless leg syndrome, daytime sleepiness, insomnia symptoms or sleep apnoea risk before and after adjustment for APOE e4 and depressive symptoms using logistic regression models. Conversely, a cross-sectional study of 184 cognitively normal participants older than 60 years found that longer sleep latency was associated with PET-measured higher amyloid burden, independent of the APOEe4 status.90 Another cross-sectional study conducted by Spira et al.91 studied 70 community-dwelling subjects (mean age 76) and found a greater amyloid-β burden associated with both self-reported shorter sleep duration and poorer sleep quality using regression models. All these studies included older people than ours and used variable-centred methods with different sleep variables. Another study92 showed that greater amyloid-β burden was linked to significantly greater self-reported sleep problems and/or a significantly greater mismatch between participants’ subjective evaluation of sleep, relative to their actual objective sleep. As a result, individuals expressed lower subjective (perceived) sleep quality than their objective quality of sleep showed. Given the inconsistent associations between sleep characteristics and amyloid, more studies with both subjective and objective sleep measurements are needed to understand whether amyloid development is associated with specific sleep problems (such as Obstructive sleep apnea) or profiles of problems.
The current findings show the heterogeneity in self-reported sleep characteristics and support the importance of establishing good sleep among late middle-aged non-demented adults. Sleep could be a risk factor for mental and physical health, and objective cognition, but not amyloid burden, which has implications in clinical trial design and early intervention or prevention efforts. Subjective sleep measures could be a mediator or moderator in the health- and cognition-related research. In addition, clustering techniques should be considered when looking at the association between a risk factor and outcomes.
## Strengths
The strength of this study was the use of a novel analytic approach to leverage the underlying heterogeneity in self-reported sleep characteristics, identifying distinct groups of sleep subtypes among late middle-aged adults in a large prospective study. Moreover, the present study examined the association between SDS and five different cognitive domains. In addition, we used several strategies to investigate the robustness of our results. For example, we applied two person-centred methods to define the different sleep profiles to different cross-sectional subsets and observed that results were consistent with our primary approach.
## Limitations
First, this cross-sectional study does not determine whether poor sleep profiles precede cognitive decline (causation cannot be inferred). Considering the bi-directional relationship between sleep and cognitive decline,93 more severity of behavioural problems and cognitive impairment might actively worsen sleep, and poorer sleep might in turn worsen cognitive and physical functions. Second, similar to most of the previous studies,13,14,91,94-96 this study was based on self-reported information rather than objective measurement, which can be affected by reporting bias. For example, self-reported sleep time tended to overestimate sleep time.97 In some cases, self-reported sleep measures were only modestly correlated or even uncorrelated with objective sleep measures.98 Since objective and subjective measures of sleep have been shown to correlate differently with Alzheimer’s disease biomarkers,92 both are important to study in understanding Alzheimer’s disease–related decline. In the future, we hope to add objective sleep measurements to our study. However, we included the broadest array of health, mood, cognition, and other variables. Finally, the imputation method in sensitivity analysis was potentially flawed if the missingness is not random. Other imputation methods like multiple imputation could be used in the future. While our approach was useful in consolidating sleep measures into groups with common patterns, it is possible that some sleep characteristics have synergistic associations with cognition that are not captured by our approach. A key next step is to determine if these sleep trajectories align with more objective indicators of individuals’ sleep patterns and understand how an intervention impacts the meaning of the results. Future studies with larger sample size on sleep characteristics and cognition, also the pathologies of Alzheimer’s disease, not only amyloid plaque, but neurofibrillary tangles, are needed.
## Conclusion
This study indicates that clustering techniques can be used to identify sleep characteristic subtypes which are associated with concurrent mental, physical and cognitive health, but not beta amyloid. Although not all PS will develop mild cognitive impairment or progress to clinical dementia, this group appears to be at increased risk of cognitive decline. Future research will follow this group over time and will also examine how other risk factors differ between sleep groups. The ability to identify persons’ risk factors has implications for clinical trial design and early intervention or prevention efforts.
## Supplementary material
Supplementary material is available at Brain Communications online.
## Funding
This research was supported by the National Institutes of Health (RF-1 AG027161, R01AG021155, R01AG037639, R01AG062285 and R01AG054059) and supported by the Alzheimer’s Association (AARF-19-643973). This study was also supported in part by a core grant to the Waisman Center from the National Institute of Child Health and Human Development (P50 HD105353).
## Competing interests
D.V.P. has served as a consultant for Teva Pharmaceuticals Australia, a consultant for Harmony Biosciences and consultant/medical advisory board member for Jazz Pharmaceuticals.
## References
1. **2021 Alzheimer’s disease facts and figures**. *Alzheimers Dement* (2021) **17** 327-406. PMID: 33756057
2. Livingston G, Huntley J, Sommerlad A. **Dementia prevention, intervention, and care: 2020 report of the Lancet Commission**. *Lancet.* (2020) **396** 413-446. PMID: 32738937
3. Wennberg AMV, Wu MN, Rosenberg PB, Spira AP. **Sleep disturbance, cognitive decline, and dementia: A review**. *Semin Neurol* (2017) **37** 395-406. PMID: 28837986
4. Shi L, Chen SJ, Ma MY. **Sleep disturbances increase the risk of dementia: A systematic review and meta-analysis**. *Sleep Med Rev* (2018) **40** 4-16. PMID: 28890168
5. Bubu OM, Brannick M, Mortimer J. **Sleep, cognitive impairment, and Alzheimer’s disease: A systematic review and meta-analysis**. *Sleep* (2017) **40** zsw032
6. Sindi S, Kåreholt I, Johansson L. **Sleep disturbances and dementia risk: A multicenter study**. *Alzheimers Dement* (2018) **14** 1235-1242. PMID: 30030112
7. Winer JR, Deters KD, Kennedy G. **Association of short and long sleep duration with amyloid-β burden and cognition in aging**. *JAMA Neurol* (2021) **78** 1187-1196. PMID: 34459862
8. Tworoger SS, Lee S, Schernhammer ES, Grodstein F. **The association of self-reported sleep duration, difficulty sleeping, and snoring with cognitive function in older women**. *Alzheimer Dis Assoc Disord* (2006) **20** 41-48. PMID: 16493235
9. McSorley VE, Bin YS, Lauderdale DS. **Associations of sleep characteristics with cognitive function and decline among older adults**. *Am J Epidemiol* (2019) **188** 1066-1075. PMID: 30759177
10. Lutsey PL, Bengtson LGS, Punjabi NM. **Obstructive sleep apnea and 15-year cognitive decline: The Atherosclerosis Risk in Communities (ARIC) study**. *Sleep* (2016) **39** 309-316. PMID: 26446113
11. Lutsey PL, Misialek JR, Mosley TH. **Sleep characteristics and risk of dementia and Alzheimer’s disease: The atherosclerosis risk in communities study**. *Alzheimers Dement* (2018) **14** 157-166. PMID: 28738188
12. Gabelle A, Gutierrez LA, Jaussent I. **Absence of relationship between self-reported sleep measures and amyloid load in elderly subjects**. *Front Neurol* (2019) **10** 989. PMID: 31572295
13. Fjell AM, Sørensen Ø, Amlien IK. **Poor self-reported sleep is related to regional cortical thinning in aging but not memory decline-results from the Lifebrain consortium**. *Cereb Cortex* (2021) **31** 1953-1969. PMID: 33236064
14. Ohara T, Honda T, Hata J. **Association between daily sleep duration and risk of dementia and mortality in a Japanese community**. *J Am Geriatr Soc* (2018) **66** 1911-1918. PMID: 29873398
15. Ma Y, Liang L, Zheng F, Shi L, Zhong B, Xie W. **Association between sleep duration and cognitive decline**. *JAMA Netw Open* (2020) **3**
16. Sabia S, Fayosse A, Dumurgier J. **Association of sleep duration in middle and old age with incidence of dementia**. *Nat Commun* (2021) **12** 2289. PMID: 33879784
17. Suh SW, Han JW, Lee JR. **Sleep and cognitive decline: a prospective nondemented elderly cohort study**. *Ann Neurol* (2018) **83** 472-482. PMID: 29394505
18. Waller KL, Mortensen EL, Avlund K. **Subjective sleep quality and daytime sleepiness in late midlife and their association with age-related changes in cognition**. *Sleep Med* (2016) **17** 165-173. PMID: 26188599
19. Yu J, Mahendran R, Abdullah FNM, Kua EH, Feng L. **Self-reported sleep problems among the elderly: A latent class analysis**. *Psychiatry Res* (2017) **258** 415-420. PMID: 28867406
20. Sprecher KE, Bendlin BB, Racine AM. **Amyloid burden is associated with self-reported sleep in non-demented late middle-aged adults**. *Neurobiol Aging* (2015) **36** 2568-2576. PMID: 26059712
21. Frange C, de Queiroz SS, da Silva Prado JM, Tufik S, de Mello MT. **The impact of sleep duration on self-rated health**. *Sleep Sci* (2014) **7** 107-113. PMID: 26483912
22. Franzen PL, Buysse DJ. **Sleep disturbances and depression: Risk relationships for subsequent depression and therapeutic implications**. *Dialogues Clin Neurosci* (2008) **10** 473-481. PMID: 19170404
23. Tsapanou A, Vlachos GS, Cosentino S. **Sleep and subjective cognitive decline in cognitively healthy elderly. Results from two cohorts**. *J Sleep Res* (2019) **28**
24. Grandner MA, Schopfer EA, Sands-Lincoln M, Jackson N, Malhotra A. **The relationship between sleep duration and body mass index depends on age**. *Obes Silver Spring Md* (2015) **23** 2491-2498
25. de Souza JFT, Dáttilo M, de Mello MT, Tufik S, Antunes HKM. **High-intensity interval training attenuates insulin resistance induced by sleep deprivation in healthy males**. *Front Physiol* (2017) **8** 992. PMID: 29270126
26. Kusurkar RA, Mak-van der Vossen M, Kors J. **‘One size does not fit all’: The value of person-centred analysis in health professions education research**. *Perspect Med Educ* (2021) **10** 245-251. PMID: 33284407
27. Lanza ST, Rhoades BL, Greenberg MT, Cox M. **Modeling multiple risks during infancy to predict quality of the caregiving environment: Contributions of a person-centered approach**. *Infant Behav Dev* (2011) **34** 390-406. PMID: 21477866
28. Ye L, Pien GW, Ratcliffe SJ. **The different clinical faces of obstructive sleep apnoea: A cluster analysis**. *Eur Respir J* (2014) **44** 1600-1607. PMID: 25186268
29. Selvi Y, Boysan M, Kandeger A. **Heterogeneity of sleep quality in relation to circadian preferences and depressive symptomatology among major depressive patients**. *J Affect Disord* (2018) **235** 242-249. PMID: 29660638
30. Robillard R, Dion K, Pennestri M. **Profiles of sleep changes during the COVID-19 pandemic: Demographic, behavioural and psychological factors**. *J Sleep Res* (2021) **30**
31. Matricciani L, Paquet C, Fraysse F, Wake M, Olds T. **Sleep profiles of Australian children aged 11–12 years and their parents: Sociodemographic characteristics and lifestyle correlates**. *Sleep Med* (2020) **73** 53-62. PMID: 32771928
32. Lincoln KD, Ailshire J, Nguyen A, Taylor RJ, Govia I, Ifatunji MA. **Profiles of sleep and depression risk among Caribbean Blacks**. *Ethn Health* (2021) **26** 981-999. PMID: 31137946
33. Matricciani L, Paquet C, Fraysse F. **Sleep and cardiometabolic risk: A cluster analysis of actigraphy-derived sleep profiles in adults and children**. *Sleep* (2021) **44**
34. Fatima Y, Bucks RS, Mamun AA. **Sleep trajectories and mediators of poor sleep: Findings from the longitudinal analysis of 41,094 participants of the UK Biobank cohort**. *Sleep Med* (2020) **76** 120-127. PMID: 33157426
35. Gadie A, Shafto M, Leng Y, Kievit RA. **How are age-related differences in sleep quality associated with health outcomes? An epidemiological investigation in a UK cohort of 2406 adults**. *BMJ Open* (2017) **7**
36. Smith CE, Lee S. **Identifying diverse forms of (un)healthy sleep: Sleep profiles differentiate adults’ psychological and physical well-being**. *Soc Sci Med* (2022) **292**
37. Alashwal H, El Halaby M, Crouse JJ, Abdalla A, Moustafa AA. **The application of unsupervised clustering methods to Alzheimer’s disease**. *Front Comput Neurosci* (2019) **13** 31. PMID: 31178711
38. Blanken AE, Jang JY, Ho JK. **Distilling heterogeneity of mild cognitive impairment in the National Alzheimer Coordinating Center Database using latent profile analysis**. *JAMA Netw Open* (2020) **3**
39. Sager MA, Hermann B, La Rue A. **Middle-aged children of persons with Alzheimer’s disease: APOE genotypes and cognitive function in the Wisconsin Registry for Alzheimer’s Prevention**. *J Geriatr Psychiatry Neurol* (2005) **18** 245-249. PMID: 16306248
40. Johnson SC, Koscik RL, Jonaitis EM. **The Wisconsin Registry for Alzheimer’s Prevention: A review of findings and current directions**. *Alzheimers Dement (Amst)* (2017) **10** 130-142. PMID: 29322089
41. Stewart AL, Ware JE. *Measuring functioning and well being: The medical outcomes study approach* (1992)
42. Johns MW. **A new method for measuring daytime sleepiness: The Epworth sleepiness scale**. *Sleep* (1991) **14** 540-545. PMID: 1798888
43. Bastien CH, Vallières A, Morin CM. **Validation of the Insomnia Severity Index as an outcome measure for insomnia research**. *Sleep Med* (2001) **2** 297-307. PMID: 11438246
44. Sprecher KE. (2017)
45. Spritzer KL, Hays RD. **MOS Sleep Scale: A manual for use and scoring, version 1.0**. (2003)
46. Viala-Danten M, Martin S, Guillemin I, Hays RD. **Evaluation of the reliability and validity of the Medical Outcomes Study sleep scale in patients with painful diabetic peripheral neuropathy during an international clinical trial**. *Health Qual Life Outcomes* (2008) **6** 113. PMID: 19091084
47. Wolfe F, Michaud K, Li T. **Sleep disturbance in patients with rheumatoid arthritis: Evaluation by medical outcomes study and visual analog sleep scales**. *J Rheumatol* (2006) **33** 1942-1951. PMID: 16960928
48. Morin CM, Belleville G, Bélanger L, Ivers H. **The Insomnia Severity Index: Psychometric indicators to detect insomnia cases and evaluate treatment response**. *Sleep* (2011) **34** 601-608. PMID: 21532953
49. Nebes RD, Buysse DJ, Halligan EM, Houck PR, Monk TH. **Self-reported sleep quality predicts poor cognitive performance in healthy older adults**. *J Gerontol B Psychol Sci Soc Sci* (2009) **64B** 180-187
50. Koscik RL, La Rue A, Jonaitis EM. **Emergence of mild cognitive impairment in late middle-aged adults in the Wisconsin registry for Alzheimer’s prevention**. *Dement Geriatr Cogn Disord* (2014) **38** 16-30. PMID: 24556849
51. Clark LR, Koscik RL, Nicholas CR. **Mild cognitive impairment in late middle age in the Wisconsin registry for Alzheimer’s prevention study: Prevalence and characteristics using robust and standard neuropsychological normative data**. *Arch Clin Neuropsychol* (2016) **31** 675-688. PMID: 27193363
52. Donohue MC, Sperling RA, Salmon DP. **The preclinical Alzheimer cognitive composite: Measuring amyloid-related decline**. *JAMA Neurol* (2014) **71** 961-970. PMID: 24886908
53. Jonaitis EM, Koscik RL, Clark LR. **Measuring longitudinal cognition: Individual tests versus composites**. *Alzheimers Dement Diagn Assess Dis Monit* (2019) **11** 74-84
54. Langhough Koscik R, Hermann BP, Allison S. **Validity evidence for the research category, “cognitively unimpaired—declining,” as a risk marker for mild cognitive impairment and Alzheimer’s disease**. *Front Aging Neurosci* (2021) **13**
55. Walker JD, Maxwell CJ, Hogan DB, Ebly EM. **Does self-rated health predict survival in older persons with cognitive impairment?**. *J Am Geriatr Soc* (2004) **52** 1895-1900. PMID: 15507068
56. Radloff LS. **The CES-D Scale: A self-report depression scale for research in the general population**. *Appl Psychol Meas* (1977) **1** 385-401
57. Du L, Koscik RL, Chin NA. **Prescription medications and co-morbidities in late middle-age are associated with greater cognitive declines: Results from WRAP**. *Front Aging* (2021) **2**
58. Klunk WE, Lopresti BJ, Ikonomovic MD. **Binding of the positron emission tomography tracer Pittsburgh compound-B reflects the amount of amyloid-β in Alzheimer’s disease brain but not in transgenic mouse brain**. *J Neurosci* (2005) **25** 10598-10606. PMID: 16291932
59. Johnson SC, Christian BT, Okonkwo OC. **Amyloid burden and neural function in people at risk for Alzheimer’s disease**. *Neurobiol Aging* (2014) **35** 576-584. PMID: 24269021
60. Betthauser TJ, Cody KA, Zammit MD. **In vivo characterization and quantification of neurofibrillary tau PET radioligand 18F-MK-6240 in humans from Alzheimer disease dementia to young controls**. *J Nucl Med* (2019) **60** 93-99. PMID: 29777006
61. Racine AM, Koscik RL, Berman SE. **Biomarker clusters are differentially associated with longitudinal cognitive decline in late midlife**. *Brain* (2016) **139** 2261-2274. PMID: 27324877
62. Koscik RL, Betthauser TJ, Jonaitis EM. **Amyloid duration is associated with preclinical cognitive decline and tau PET**. *Alzheimers Dement (Amst)* (2020) **12**
63. Betthauser TJ, Koscik RL, Jonaitis EM. **Amyloid and tau imaging biomarkers explain cognitive decline from late middle-age**. *Brain* (2020) **143** 320-335. PMID: 31886494
64. Kassambara A, Mundt F.. (2020)
65. Mangiafico SS. (2016)
66. Oberski DL. **Review of latent variable modeling with R**. *J Educ Behav Stat* (2016) **41** 226-233
67. Engels JM, Diehr P. **Imputation of missing longitudinal data: A comparison of methods**. *J Clin Epidemiol* (2003) **56** 968-976. PMID: 14568628
68. Wallace TM, Levy JC, Matthews DR. **Use and abuse of HOMA modeling**. *Diabetes Care* (2004) **27** 1487-1495. PMID: 15161807
69. Buysse DJ. **Sleep health: Can we define it? Does it matter?**. *Sleep* (2014) **37** 9-17. PMID: 24470692
70. Johnson DA, Jackson CL, Williams NJ, Alcántara C. **Are sleep patterns influenced by race/ethnicity—a marker of relative advantage or disadvantage? Evidence to date**. *Nat Sci Sleep* (2019) **11** 79-95. PMID: 31440109
71. Stamatakis KA, Kaplan GA, Roberts RE. **Short sleep duration across income, education and race/ethnic groups: Population prevalence and growing disparities over 34 years of follow-up**. *Ann Epidemiol* (2007) **17** 948-955. PMID: 17855122
72. Song Y, Ancoli-Israel S, Lewis CE, Redline S, Harrison SL, Stone KL. **The association of race/ethnicity with objectively measured sleep characteristics in older men**. *Behav Sleep Med* (2011) **10** 54-69. PMID: 22250779
73. Whinnery J, Jackson N, Rattanaumpawan P, Grandner MA. **Short and long sleep duration associated with race/ethnicity, sociodemographics, and socioeconomic position**. *Sleep* (2014) **37** 601-611. PMID: 24587584
74. Kingsbury JH, Buxton OM, Emmons KM, Redline S. **Sleep and its relationship to racial and ethnic disparities in cardiovascular disease**. *Curr Cardiovasc Risk Rep* (2013) **7** 387-394
75. Grandner MA, Williams NJ, Knutson KL, Roberts D, Jean-Louis G. **Sleep disparity, race/ethnicity, and socioeconomic position**. *Sleep Med* (2016) **18** 7-18. PMID: 26431755
76. Harrington YA, Parisi JM, Duan D, Rojo-Wissar DM, Holingue C, Spira AP. **Sex hormones, sleep, and memory: Interrelationships across the adult female lifespan**. *Front Aging Neurosci* (2022) **14**
77. Drogos LL, Gill SJ, Tyndall AV. **Evidence of association between sleep quality and APOE ɛ4 in healthy older adults: A pilot study**. *Neurology* (2016) **87** 1836-1842. PMID: 27777343
78. Crispim CA, Zalcman I, Dáttilo M. **The influence of sleep and sleep loss upon food intake and metabolism**. *Nutr Res Rev* (2007) **20** 195-212. PMID: 19079870
79. Donga E, van Dijk M, van Dijk JG. **A single night of partial sleep deprivation induces insulin resistance in multiple metabolic pathways in healthy subjects**. *J Clin Endocrinol Metab* (2010) **95** 2963-2968. PMID: 20371664
80. Moraes W, Poyares D, Zalcman I. **Association between body mass index and sleep duration assessed by objective methods in a representative sample of the adult population**. *Sleep Med* (2013) **14** 312-318. PMID: 23391395
81. Paunio T, Korhonen T, Hublin C. **Poor sleep predicts symptoms of depression and disability retirement due to depression**. *J Affect Disord* (2015) **172** 381-389. PMID: 25451441
82. Carpi M, Cianfarani C, Vestri A. **Sleep quality and its associations with physical and mental health-related quality of life among university students: A cross-sectional study**. *Int J Environ Res Public Health* (2022) **19** 2874. PMID: 35270566
83. Devore EE, Grodstein F, Duffy JF, Stampfer MJ, Czeisler CA, Schernhammer ES. **Sleep duration in midlife and later life in relation to cognition**. *J Am Geriatr Soc* (2014) **62** 1073-1081. PMID: 24786726
84. Dlugaj M, Weinreich G, Weimar C. **Sleep-disordered breathing, sleep quality, and mild cognitive impairment in the general population**. *J Alzheimers Dis* (2014) **41** 479-497. PMID: 24643134
85. Devore EE, Grodstein F, Schernhammer ES. **Sleep duration in relation to cognitive function among older adults: A systematic review of observational studies**. *Neuroepidemiology* (2016) **46** 57-78. PMID: 26735553
86. Spira AP, Chen-Edinboro LP, Wu MN, Yaffe K. **Impact of sleep on the risk of cognitive decline and dementia**. *Curr Opin Psychiatry* (2014) **27** 478-483. PMID: 25188896
87. Yaffe K, Falvey CM, Hoang T. **Connections between sleep and cognition in older adults**. *Lancet Neurol* (2014) **13** 1017-1028. PMID: 25231524
88. Chen JC, Espeland MA, Brunner RL. **Sleep duration, cognitive decline, and dementia risk in older women**. *Alzheimers Dement* (2016) **12** 21-33. PMID: 26086180
89. Cross NE, Carrier J, Postuma RB. **Association between insomnia disorder and cognitive function in middle-aged and older adults: A cross-sectional analysis of the Canadian longitudinal study on aging**. *Sleep* (2019) **42**
90. Brown BM, Rainey-Smith SR, Villemagne VL. **The relationship between sleep quality and brain amyloid burden**. *Sleep* (2016) **39** 1063-1068. PMID: 27091528
91. Spira AP, Gamaldo AA, An Y. **Self-reported sleep and β-amyloid deposition in community-dwelling older adults**. *JAMA Neurol* (2013) **70** 1537-1543. PMID: 24145859
92. Winer JR, Morehouse A, Fenton L. **Tau and β-amyloid burden predict actigraphy-measured and self-reported impairment and misperception of human sleep**. *J Neurosci* (2021) **41** 7687-7696. PMID: 34290080
93. Guarnieri B, Sorbi S. **Sleep and cognitive decline: A strong bidirectional relationship. It is time for specific recommendations on routine assessment and the management of sleep disorders in patients with mild cognitive impairment and dementia**. *Eur Neurol* (2015) **74** 43-48. PMID: 26159605
94. Lo JC, Groeger JA, Cheng GH, Dijk DJ, Chee MWL. **Self-reported sleep duration and cognitive performance in older adults: A systematic review and meta-analysis**. *Sleep Med* (2016) **17** 87-98. PMID: 26847980
95. Bokenberger K, Ström P, Dahl Aslan AK. **Association between sleep characteristics and incident dementia accounting for baseline cognitive status: A prospective population-based study**. *J Gerontol A Biol Sci Med Sci* (2017) **72** 134-139. PMID: 27402049
96. Brachem C, Winkler A, Tebrügge S. **Associations between self-reported sleep characteristics and incident mild cognitive impairment: The Heinz Nixdorf Recall Cohort Study**. *Sci Rep* (2020) **10** 6542. PMID: 32300149
97. Jackson CL, Patel SR, Jackson WB II, Lutsey PL, Redline S. **Agreement between self-reported and objectively measured sleep duration among white, black, Hispanic, and Chinese adults in the United States: Multi-Ethnic Study of Atherosclerosis**. *Sleep* (2018) **41**
98. Unruh ML, Redline S, An MW. **Subjective and objective sleep quality and aging in the sleep heart health study**. *J Am Geriatr Soc* (2008) **56** 1218-1227. PMID: 18482295
|
---
title: 'Gut Bacteria Involved
in Ellagic Acid Metabolism To
Yield Human Urolithin Metabotypes Revealed'
authors:
- 'Carlos
E. Iglesias-Aguirre'
- Rocío García-Villalba
- David Beltrán
- María Dolores Frutos-Lisón
- Juan C. Espín
- Francisco A. Tomás-Barberán
- María V. Selma
journal: Journal of Agricultural and Food Chemistry
year: 2023
pmcid: PMC9999415
doi: 10.1021/acs.jafc.2c08889
license: CC BY 4.0
---
# Gut Bacteria Involved
in Ellagic Acid Metabolism To
Yield Human Urolithin Metabotypes Revealed
## Abstract
We aimed to elucidate the gut bacteria that characterize the human urolithin metabotypes A and B (UM-A and UM-B). We report here a new bacterium isolated from the feces of a healthy woman, capable of producing the final metabolites urolithins A and B and different intermediates. Besides, we describe two gut bacterial co-cultures that reproduced the urolithin formation pathways upon in vitro fermentation of both UM-A and UM-B. This is the first time that the capacity of pure strains to metabolize ellagic acid cooperatively to yield urolithin profiles associated with UM-A and UM-B has been demonstrated. The urolithin-producing bacteria described herein could have potential as novel probiotics and in the industrial manufacture of bioactive urolithins to develop new ingredients, beverages, nutraceuticals, pharmaceuticals, and (or) functional foods. This is especially relevant in UM-0 individuals since they cannot produce bioactive urolithins.
## Introduction
Urolithins (Uros) have gained recognition as one of the main drivers for the health effects related to the intake of ellagitannins (ETs) and ellagic acid (EA)-rich foods such as nuts, pomegranates, many tropical fruits, and berries. The human gut microbiota converts these polyphenols into Uros. To date, 13 Uros and their conjugated metabolites (glucuronides and sulfates) have been described in different human fluids and tissues (blood, urine, feces, breastmilk, prostate, colon, and breast tissues).1−3 Uro production capacity and, consequently, at least partly, the health effects associated with ET consumption vary among individuals because not everyone has the gut bacteria needed to produce all the Uros.4,5 Three Uro metabotypes (UMs, i.e., UM-A, UM-B, and UM-0) associated with three different Uro production profiles have been described in western and eastern populations.6−8 The Uro production pathways for each UM have been elucidated using human samples and fecal fermentation studies in batch or using a dynamic gastrointestinal simulation model (TWIN-SHIME). Differences in the Uro profiles have been observed between UMs and along the large intestine, showing predominant Uro production in the distal colon region.9−12 One of the main differences between the metabolic profiles associated with UMs is the final Uros produced. UM-A individuals only yield Uro-A as the final metabolite of the EA metabolic pathway, whereas UM-B subjects produce Uro-A and, distinctively, IsoUro-A and Uro-B. Finally, UM-0 individuals cannot produce Uros (only the precursor Uro-M5 has been detected so far). Remarkably, the percentage of UM-0 in Spanish and Chinese healthy populations rounds to $10\%$.7,8 UM-0 prevalence could be even higher ($60\%$) in the US population, according to a study with 100 participants, where $33\%$ did not produce or $27\%$ were low producers of Uro-A.13 The type of UM depends on the gut microbiota composition of each person.5,14 There has been a substantial advance in the research on the specific bacteria involved in Uro production and the compositional and functional characterization of the gut microbiota associated with UMs (UM-A, UM-B, and UM-0).14−16 Recent studies demonstrated that more than $30\%$ of the discriminating genera between UM-A and UM-B belonged to the Eggerthellaceae family.14 Certainly, genera from this family, such as Gordonibacter and Ellagibacter, harbor intestinal species that can transform EA into some intermediary Uros.17−20 However, the human gut bacteria producing Uro-A and Uro-B (i.e., the main metabolite markers of UM-A and UM-B, respectively), and many intermediate Uros within each UM, are still unknown. Therefore, the present study is aimed to elucidate the gut bacteria that characterize the human UMs.
## Chemicals
As described elsewhere, Uros were chemically synthesized (Villapharma, Murcia, Spain)10 or purchased from Dalton Pharma Services (Toronto, Canada). Purity was higher than $95\%$ in all tested compounds.
## Isolation of Uro-Producing Bacteria
A healthy female donor (aged 30), who was previously demonstrated to produce Uros in vivo, provided the stool samples. The study conformed to ethical guidelines outlined in the Declaration of Helsinki and its amendments. The protocol (included in the project AGL2015-64124-R) was approved by the Spanish National Research Council’s Bioethics Committee (Spain). The donor gave written informed consent following the Declaration of Helsinki. As explained elsewhere, Uros were identified in feces and urine after walnut consumption.21 The feces were prepared to isolate Uro-producing bacteria following a protocol previously described with some modifications.19,20 Briefly, after $\frac{1}{10}$ (w/v) fecal dilution in nutrient broth (Oxoid, Basingstoke, Hampshire, UK) supplemented with $0.05\%$ l-cysteine hydrochloride (PanReac Química, Barcelona, Spain), the filtrated sample was homogenized and further diluted in Wilkins–Chalgren anaerobe medium (WAM, Oxoid). The metabolic activity was evaluated by adding to the broth Uro-C (Dalton Pharma Services) dissolved in propylene glycol (PanReac Química SLU, Barcelona, Spain) to reach a final concentration of 15 μM. After anaerobic incubation, a portion of the culture, having metabolic activity, was seeded on WAM agar. Colonies were collected and inoculated into 5 mL of WAM containing 15 μM Uro-C, and after incubation, their capacity to convert Uro-C was assayed. Uro-C-transforming colonies were subcultured until single strains were isolated. The isolation procedure and plate incubation were achieved in an anaerobic chamber (Concept 400, Baker Ruskin Technologies Ltd., Bridgend, South Wales, UK) at 37 °C. Samples (5 mL) were prepared for HPLC–DAD–MS analyses of Uros. We isolated pure bacterial cultures (*Enterocloster bolteae* strain CEBAS S4A9), which showed the capacity to transform Uro-C. This strain was phylogenetically identified, and its metabolic characteristics were analyzed as described below.
## Identification of the Isolated Uro-Producing Bacteria
The almost-complete 16S rRNA gene sequence of the isolated bacterial strain (E. bolteae CEBAS S4A9) and the phylogenetic analysis were achieved as previously described.20 A phylogenetic tree, including the isolated strain E. bolteae CEBAS S4A9, the most closely related species and known Uro-producing genera (Gordonibacter and Ellagibacter), was constructed using the neighbor-joining treeing method.19
## Conversion Testing of EA and Intermediary Uros
The isolated strain E. bolteae CEBAS S4A9 and representative strains of the closest relatives (E. bolteae DSM 29485, DSM 15670T, *Enterocloster asparagiformis* DSM 15981T, *Enterocloster citroniae* DSM 19261T, and *Enterocloster clostridioformis* DSM 933T) obtained from the DSMZ culture collection were used to investigate their capacity to produce final Uros in the presence of EA and other Uro intermediaries. Briefly, isolated and DSMZ strains were separately incubated on a WAM agar plate for 6 days. A single colony was cultivated in a 5 mL WAM tube. Diluted inoculum (2 mL) was transferred to WAM (20 mL), obtaining an initial load of 104 CFU mL–1. EA, Uro-M6, Uro-D, Uro-C, Uro-A, IsoUro-A, and Uro-B were dissolved in propylene glycol and added to the 20 mL cultures to obtain a final concentration of 15 μM each. After incubation in an anoxic environment at 37 °C, aliquots (5 mL) were taken periodically for high-performance liquid chromatography (HPLC) analyses as described below.
## In Vitro Conversion of EA with Gut Bacteria To Reproduce UMs
Gordonibacter urolithinfaciens DSM 27213T, Ellagibacter isourolithinifaciens DSM 104140T obtained from the DSMZ culture collection, and the isolated strain E. bolteae CEBAS S4A9 were cultivated anaerobically in 5 mL WAM tubes. First, 2 mL of a diluted aliquot of G. urolithinfaciens DSM 27213T and E. bolteae CEBAS S4A9 strains was transferred to WAM (100 mL). Similarly, 2 mL of diluted aliquots of E. isourolithinifaciens DSM 104140T and E. bolteae CEBAS S4A9 strains was transferred to WAM (100 mL). Finally, EA dissolved in propylene glycol was added to the 100 mL cultures to obtain a final concentration of 25 μM. During incubation in an anoxic environment at 37 °C, aliquots (5 mL) were taken for HPLC analyses as described below. Incubations were made in triplicate, and the experiment was repeated twice.
## Sample Clean-Up and HPLC–DAD–MS Analyses
As previously described, aliquots (5 mL) collected during the incubation of single and combined bacterial strains were extracted and analyzed by HPLC–DAD–ESI-Q (MS).19 Briefly, fermented medium (5 mL) was extracted with ethyl acetate (5 mL) (Labscan, Dublin, Ireland), acidified with $1.5\%$ formic acid (PanReac), vortexed for 2 min, and centrifuged at 3500g for 10 min. The organic phase was separated and evaporated, and the dry samples were then re-dissolved in methanol (250 μL) (Romil, Barcelona, Spain). An HPLC system (1200 Series, Agilent Technologies, Madrid, Spain) equipped with a photodiode-array detector (DAD) and a single quadrupole mass spectrometer detector in series (6120 Quadrupole, Agilent Technologies, Madrid, Spain) was used. Calibration curves were obtained for EA, Uro-M6, Uro-D, Uro-C, Uro-A, Uro-B, and IsoUro-A with good linearity (R2 > 0.998).
## Identification of Uro-Producing Bacteria
One bacterial strain isolated from a human fecal sample, named strain CEBAS S4A9, obtained from a 1:104 dilution plated on WAM agar, showed the capacity to convert Uro-C into Uro-A under anaerobic conditions. A nearly complete 16S rRNA gene sequence (1389 bp) was obtained for isolating CEBAS S4A9. The sequence was aligned with the closest accepted members of this family. The phylogenetic tree, representing minimum evolutionary distances (Jukes–Cantor), showed that strain CEBAS S4A9 grouped with the other members of the *Enterocloster genus* (Figure 1). The closest relatives of strain CEBAS S4A9 are E. bolteae DSM 15670T ($99.8\%$ 16S rRNA gene sequence similarity), E. asparagiformis DSM 15981T ($98.0\%$), E. citroniae DSM 19261T ($97.0\%$), and E. clostridioformis DSM 933T ($97.7\%$). A higher distance was observed with other known Uro-producing bacteria in the phylogenetic tree, i.e., G. pamelaeae DSM 19378T ($80.9\%$), Gordonibacter urolithinfaciens DSM 27213T ($78.2\%$), and Ellagibacter isourolithinifaciens DSM 104140T ($80.0\%$).
**Figure 1:** *Phylogenetic tree showing the relationship
between the strain E. bolteae CEBAS
S4A9 and other Uro-producing bacteria
(green color). The tree was constructed using the neighbor-joining
method based on 16S rRNA gene sequences. The distance matrix was calculated
by the Jukes–Cantor method. GenBank accession numbers are presented
in parentheses. Bar, 0.05 substitutions per nucleotide position. Numbers
at nodes (≥70%) indicate support for internal branches within
the tree obtained by bootstrap analysis (percentages of 500 re-samplings).*
## Analysis of Uros Produced by Enterocloster Species
The HPLC–MS analyses showed that, in contrast to G. urolithinfaciens DSM 27213T and E. isourolithinifaciens DSM 104140T, the isolate E. bolteae CEBAS S4A9 and the closest relatives (E. bolteae DSM 29485, DSM 15670T, E. asparagiformis DSM 15981T, E. citroniae DSM 19261T, and E. clostridioformis DSM 933T) did not metabolize EA (Table 1). However, all Enterocloster species tested, except E. clostridioformis DSM 933T, metabolized Uro-M6 to other Uros, such as Uro-A, via Uro-M7. Table 1 shows the specific Uros produced by each microbial species after incubation with the different precursors. G. urolithinfaciens DSM 27213T and E. isourolithinifaciens DSM 104140T also transformed Uro-M6, but G. urolithinfaciens rendered Uro-C, whereas E. isourolithinifaciens produced IsoUro-A via Uro-C. Uro-D was also transformed by most of the Enterocloster species tested, rendering a novel metabolite that we named urolithin G (Uro-G; 3,4,8-trihydroxy-urolithin), whose structure was recently established.22 Uro-G showed an Rt at 12.58 min that did not coincide, under the same assay conditions, with the already known trihydroxy-urolithins, Uro-C (Rt 12.44 min), Uro-CR (Rt 13.17 min), and Uro-M7 (Rt 13.59 min) (Figure 2), suggesting a new metabolite (7 in Figure 2A,B). In contrast, G. urolithinfaciens DSM 27213T transformed Uro-D until Uro-C, whereas E. isourolithinifaciens DSM 104140T transformed Uro-D until IsoUro-A via Uro-C (Table 1). Most of the Enterocloster species tested completely transformed Uro-C into Uro-A except E. clostridioformis, which gave negative reactions for Uro production. E. isourolithinifaciens DSM 104140T also transformed Uro-C but only until IsoUro-A. Unlike E. isourolithinifaciens DSM 104140T and G. urolithinfaciens DSM 27213T, Enterocloster species further converted IsoUro-A into Uro-B. Interestingly, the type strain of E. bolteae did not transform IsoUro-A into Uro-B, unlike the other strains of E. bolteae tested (CEBAS S4A9 and DSM 29485) (Table 1). None of these bacterial strains dehydroxylated Uro-A or Uro-B (data not shown). As reported previously, all metabolites were identified by direct comparison (UV spectra and MS) with standards and confirmed by their spectral properties and molecular masses.10 **Figure 2:** *HPLC–DAD chromatogram of in vitro metabolism of EA by (A) G. urolithinfaciens DSM 27213T and E. bolteae CEBAS S4A9 strains, which mimic the Uro
metabotype A (UM-A) and by (B) E. isourolithinifaciens DSM 104140T and E. bolteae CEBAS S4A9 strain co-culture, which mimics the Uro metabotype B
(UM-B). 1: Uro-M5; 2: Uro-D; 3: Uro-E; 4: EA; 5: Uro-M6; 6: Uro-C;
7: Uro-G; 8: Uro-M7; 9: IsoUro-A; 10: Uro-A; 11: Uro-B.* TABLE_PLACEHOLDER:Table 1
## In Vitro Catabolism of EA by Human Gut Bacteria Co-Culture Reproducing
UMs
The in vitro co-culture of G. urolithinfaciens DSM 27213T and E. bolteae CEBAS S4A9 strains (co-culture 1) and that of E. isourolithinifaciens DSM 104140T and E. bolteae CEBAS S4A9 strains (co-culture 2) were followed to study their Uro production patterns from EA (Figure 2). The HPLC–DAD chromatogram at 15 h of incubation showed the production of Uro-M5, Uro-D, Uro-E, Uro-M6, Uro-C, Uro-G, Uro-M7, and Uro-A from EA by the bacterial co-culture 1 (potential UM-A reproducer) (Figure 2A). In the case of the bacterial co-culture 2 (potential UM-B reproducer), the HPLC–DAD chromatogram showed the production of Uro-M5, Uro-C, Uro-G, Uro-M7, IsoUro-A, Uro-A, and Uro-B from EA (Figure 2B). In both chromatograms, Uro-M5 was barely detected. Uro-E and the novel Uro-G were quantified using the Uro-M7 standard, whereas Uro-M5 was quantified using Uro-M6 as there were no standards for these Uros.
When EA was incubated with co-culture 1 (potential UM-A reproducer), Uros started to be detected after 15 h of incubation (Figure 3A). Uro-D was only detected at this time. Uro-M6 and Uro-E (tetrahydroxy-urolithins) also appeared at 15 h with a concentration of 3.34 and 0.21 μM, respectively (Figure 3A). Regarding trihydroxy-urolithins, Uro-C also peaked at 15 h of incubation and reached a plateau. Uro-M7 and Uro-G started to be detected at 15 h of incubation and then progressively decreased. Concerning dihydroxy-urolithins, only Uro-A was detected, reaching a maximum concentration of 18.71 μM (Figure 3A). Most EA disappeared on the third day of incubation, with remaining nonmetabolized EA concentrations in the medium being lower than 0.06 μM after 5 days (Figure 3A). When EA was incubated with co-culture 2 (potential UM-B reproducer), EA started to be converted to Uro-M7 and Uro-C via Uro-M5 and Uro-M6. The maximal Uro-M7, Uro-C, and Uro-G concentrations were achieved on the third day. Then, a plateau was maintained but only in the case of Uro-C. Uro-A started to be detected at 15 h of incubation and reached a concentration of 18.86 μM (Figure 3B). Similarly, IsoUro-A and Uro-B started to be detected at 15 h of incubation, and then a plateau was reached. Most EA was metabolized, and no EA was detected after 7 days (Figure 3B).
**Figure 3:** *Time course production of Uros from EA. (A) Metabolism
of EA by G. urolithinfaciens DSM 27213T and E. bolteae CEBAS S4A9
strain co-culture. (B) Metabolism
of EA by E. isourolithinifaciens DSM
104140T and E. bolteae CEBAS
S4A9 strain co-culture.*
## Discussion
The specific gut microbial ecology of UMs can indirectly affect the health benefits attributed to ETs and EA consumption.14 In the present study, we have revisited the metabolic capacity of known Uro-producing genera (Gordonibacter and Ellagibacter) using different intermediary Uros as substrates. The genus Gordonibacter, predominant in UM-A individuals,14 metabolizes EA into Uro-M5, Uro-M6, and Uro-C.17,18Ellagibacter, another genus from the Eggerthellaceae family, predominant in UM-B individuals, can also convert EA into some Uros (Uro-M5, Uro-M6, Uro-C, and IsoUro-A).19,20 In the present study, we observed that the Gordonibacter and *Ellagibacter* genera also converted Uro-D and Uro-M6 into Uro-C because of their 4- and 10-dehydroxylase activities, respectively (Table 1 and Figure 4). However, the Gordonibacter and Ellagibacter strains could not produce Uro-A from Uro-C, neither Uro-B from Uro-A nor any other Uro conversion involving the dehydroxylation activity at the 9-position, including the conversion of Uro-M6 into Uro-M7, or that of Uro-D into the novel Uro-G, which is described here for the first time (Figure 4). Similarly, Gordonibacter and Ellagibacter did not produce the intermediaries Uro-E or Uro-M7 from EA because of their lack of dehydroxylation activity at the 9-position. Consequently, other unknown bacteria from the gut were necessary to complete the EA metabolism associated with human UMs (Table 1 and Figure 4).
**Figure 4:** *Proposed metabolic pathway
of EA by strains from Gordonibacter (1), Ellagibacter (2), and Enterocloster (3)
genera including the isolate E. bolteae CEBAS S4A9. (1) Gordonibacter urolithinfaciens and G. pamelaeae; (2) Ellagibacter isourolithinifaciens; (3) Enterocloster bolteae, E. asparagiformis, and E. citroniae.*
We report here a new bacterium isolated from the feces of a healthy woman, capable of producing the final metabolites Uro-A and Uro-B from Uro-C and IsoUro-A, respectively. The comparison of the 16S rRNA gene sequence of the strain showed that the isolate belongs to the *Enterocloster bolteae* species ($99.8\%$ similarity with the type strain E. bolteae DSM 15670) from the family Lachnospiraceae. Before creating the family Lachnospiraceae, this large group was recognized as the *Clostridium cluster* XIVa or *Clostridium coccoides* group. The clade with C. bolteae, Clostridium asparagiformis, Clostridium citroniae, Clostridium clostridioforme, and *Clostridium aldenensis* has recently been reclassified as Enterocloster gen. nov., and the species as *Enterocloster bolteae* comb. nov., *Enterocloster asparagiformis* comb. nov., *Enterocloster citroniae* comb. nov., *Enterocloster clostridioformis* comb. nov., and *Enterocloster aldensis* comb. nov., respectively.23 Some Lachnospiraceae species, such as Butyrivibrio and Blautia, are known for being benign members of gut microbiomes and their plant-degrading capabilities, including the metabolism of phenolic compounds.5 We show here that the isolate E. bolteae CEBAS S4A9 and its closest relatives, such as E. bolteae DSM 29485, DSM 15670T, E. asparagiformis DSM 15981T, and E. citroniae DSM 19261T, produced the final Uros Uro-A and Uro-B. (Table 1). However, none could metabolize EA, not even into intermediate Uros such as Uro-M5, unlike Gordonibacter and Ellagibacter. Therefore, although phylogenetically far, genera from these two families (Lachnospiraceae and Eggerthellaceae) have complementary activities in the EA catabolism to produce Uros. Gordonibacter transformed EA into Uros by lactone-ring cleavage, decarboxylation, and further catechol dehydroxylations at 4- and 10-positions. Ellagibacter shared with Gordonibacter the lactone-ring cleavage and decarboxylation but dehydroxylated at the 4-, 8-, and 10-positions (Table 1 and Figure 4). Ellagibacter did not produce Uro-B from Uro-G or Uro-A despite having 8-dehydroxylase capacity. In contrast, it can produce Uro-A from Uro-G. This suggests that it can only dehydroxylate on catechol rings. On the contrary, the *Enterocloster* genera catalyzed the dehydroxylation of hydroxyl groups at 9- and 10-positions, regardless of whether they were in a catechol ring (Table 1 and Figure 4). Uro-G was only obtained after Uro-D incubation with the Enterocloster species that harbor 9-dehydroxylase activity. This supports that Uro-G corresponds to 3,4,8-trihydroxy-urolithin.22 We tested and patented two bacterial combinations to reproduce the Uro profiles that characterize the human UM-A and UM-B, i.e., group 1 combined G. urolithinfaciens DSM 27213T and E. bolteae CEBAS S4A9 strains, whereas group 2 combined E. isourolithinifaciens DSM 104140T and E. bolteae CEBAS S4A9 strains.22 The metabolic capabilities of the two co-cultures were followed in vitro to study the time course production of the potential intermediate catabolites in the route from EA to Uro-A or Uro-B (Figure 3). Besides, the similarities with the Uro profiles of UM-A and UM-B individuals were also analyzed. Uro metabolic profiles of UM-A individuals described in vivo6−8 and in fecal fermentation studies9−12 showed a lack of 8-dehydroxylase activity and were consistent with those found in vitro during the incubation of EA with co-culture 1 (Figures 2A and 3A). In the case of co-culture 2 (Figures 2B and 3B), the Uro profile obtained resembled the metabolic profile of UM-B individuals described in vivo6−8 and in human fecal fermentation studies.9−12 In the present study, bacterial combinations 1 and 2 included the E. bolteae CEBAS S4A9 strain. However, similar results would have been obtained with other Enterocloster species such as E. bolteae (strain CEBAS S4A9, DSM 15670T), E. asparagiformis DSM 15981T, and E. citroniae DSM 19261T because of their implication in Uro metabolism, unlike E. clostridioformis DSM 933T (Table 1).
We report here for the first time the capacity of pure strains to metabolize EA cooperatively to render Uro profiles associated with UM-A and UM-B. The Uro-producing bacteria described herein could have potential as novel probiotics and in the industrial manufacture of bioactive Uros to develop new ingredients, beverages, nutraceuticals, pharmaceuticals, and (or) functional foods. This is especially relevant in those individuals with UM-0 since they cannot produce bioactive Uros. Uro-A administration has recently been assayed for safety requirements and Generally Recognized as Safe (GRAS) by the Food and Drug Administration (FDA: 20-12-2018. GRAS Notice No. GRN 000791).24 The impact and safety of oral supplementation with Uro-A were recently investigated in a randomized clinical trial in middle-aged adults. Results showed that oral administration of Uro-A improved muscle strength and exercise performance measures accompanied by an impact on mitochondrial biomarkers.13 However, in the case of Uro-producing bacteria, further research is necessary to probe well-established health effects on the host as well as safety requirements before being considered among the next-generation probiotics.
## References
1. García-Villalba R., Giménez-Bastida J. A., Cortés-Martín A., Ávila-Gálvez M. Á., Tomás-Barberán F. A., Selma M. V., Espín J. C., González-Sarrías A.. **Urolithins: A Comprehensive Update on Their Metabolism, Bioactivity, and Associated Gut Microbiota**. *Mol. Nutr. Food Res.* (2022) **66**. DOI: 10.1002/mnfr.202101019
2. Cortés-Martín A., García-Villalba R., García-Mantrana I., Rodríguez-Varela A., Romo-Vaquero M., Collado M. C., Tomás-Barberán F. A., Espín J. C., Selma M. V.. **Urolithins in Human Breast Milk after Walnut Intake and Kinetics of**. *J. Agric. Food Chem.* (2020) **68** 12606-12616. DOI: 10.1021/acs.jafc.0c04821
3. Tomás-Barberán F. A., González-Sarrías A., García-Villalba R., Núñez-Sánchez M. A., Selma M. V., García-Conesa M. T., Espín J. C.. **Urolithins, the Rescue of “Old” Metabolites to Understand a “New” Concept: Metabotypes as a Nexus among Phenolic Metabolism, Microbiota Dysbiosis, and Host Health Status**. *Mol. Nutr. Food Res.* (2017) **61**. DOI: 10.1002/mnfr.201500901
4. Iglesias-Aguirre C. E., Cortés-Martín A., Ávila-Gálvez M. Á., Giménez Bastida J. A., Selma M. V., González-Sarrías A., Espín J. C.. **Main Drivers of (Poly)Phenols Effects on Human Health: Metabolites Production and(or) Gut Microbiota-Associated Metabotypes?**. *Food Funct.* (2021) **12** 10324-10355. DOI: 10.1039/D1FO02033A
5. Cortés-Martín A., Selma M. V., Tomás-Barberán F. A., González-Sarrías A., Espín J. C.. **Where to Look into the Puzzle of Polyphenols and Health? The Postbiotics and Gut Microbiota Associated with Human Metabotypes**. *Mol. Nutr. Food Res.* (2020) **64**. DOI: 10.1002/mnfr.201900952
6. Tomás-Barberán F. A., García-Villalba R., González-Sarrías A., Selma M. V., Espín J. C.. **Ellagic Acid Metabolism by Human Gut Microbiota: Consistent Observation of Three Urolithin Phenotypes in Intervention Trials, Independent of Food Source, Age, and Health Status**. *J. Agric. Food Chem.* (2014) **62** 6535-6538. DOI: 10.1021/jf5024615
7. Xian W., Yang S., Deng Y., Yang Y., Chen C., Li W., Yang R.. **Distribution of Urolithins Metabotypes in Healthy Chinese Youth: Difference in Gut Microbiota and Predicted Metabolic Pathways**. *J. Agric. Food Chem.* (2021) **69** 13055-13065. DOI: 10.1021/acs.jafc.1c04849
8. Cortés-Martín A., García-Villalba R., González-Sarrías A., Romo-Vaquero M., Loria-Kohen V., Ramírez-De-Molina A., Tomás-Barberán F. A., Selma M. V., Espín J. C.. **The Gut Microbiota Urolithin Metabotypes Revisited: The Human Metabolism of Ellagic Acid Is Mainly Determined by Aging**. *Food Funct.* (2018) **9** 4100-4106. DOI: 10.1039/C8FO00956B
9. García-Villalba R., Vissenaekens H., Pitart J., Romo-Vaquero M., Espín J. C., Grootaert C., Selma M. V., Raes K., Smagghe G., Possemiers S., Van Camp J., Tomás-Barberán F. A.. **Gastrointestinal Simulation Model TWIN-SHIME Shows Differences between Human Urolithin-Metabotypes in Gut Microbiota Composition, Pomegranate Polyphenol Metabolism, and Transport along the Intestinal Tract**. *J.
Agric. Food Chem.* (2017) **65** 5480-5493. DOI: 10.1021/acs.jafc.7b02049
10. García-Villalba R., Espín J. C., Tomás-Barberán F. A.. **Chromatographic and Spectroscopic Characterization of Urolithins for Their Determination in Biological Samples after the Intake of Foods Containing Ellagitannins and Ellagic Acid**. *J. Chromatogr. A* (2016) **1428** 162-175. DOI: 10.1016/j.chroma.2015.08.044
11. García-Villalba R., Selma M. V., Espín J. C., Tomás-Barberán F. A.. **Identification of Novel Urolithin Metabolites in Human Feces and Urine after the Intake of a Pomegranate Extract**. *J. Agric. Food
Chem.* (2019) **67** 11099-11107. DOI: 10.1021/acs.jafc.9b04435
12. Romo-Vaquero M., García-Villalba R., González-Sarrías A., Beltrán D., Tomás-Barberán F. A., Espín J. C., Selma M. V.. **Interindividual Variability in the Human Metabolism of Ellagic Acid: Contribution of**. *J. Funct. Foods* (2015) **17** 785-791. DOI: 10.1016/j.jff.2015.06.040
13. Singh A., D’Amico D., Andreux P. A., Dunngalvin G., Kern T., Blanco-Bose W., Auwerx J., Aebischer P., Rinsch C.. **Direct Supplementation with Urolithin A Overcomes Limitations of Dietary Exposure and Gut Microbiome Variability in Healthy Adults to Achieve Consistent Levels across the Population**. *Eur. J. Clin. Nutr.* (2022) **76** 297-308. DOI: 10.1038/s41430-021-00950-1
14. Romo-Vaquero M., Cortés-Martín A., Loria-Kohen V., Ramírez-de-Molina A., García-Mantrana I., Collado M. C., Espín J. C., Selma M. V.. **Deciphering the Human Gut Microbiome of Urolithin Metabotypes: Association with Enterotypes and Potential Cardiometabolic Health Implications**. *Mol. Nutr. Food Res.* (2019) **63**. DOI: 10.1002/mnfr.201800958
15. Cortés-Martín A., Romo-Vaquero M., García-Mantrana I., Rodríguez-Varela A., Collado M. C., Espín J. C., Selma M. V.. **Urolithin Metabotypes Can Anticipate the Different Restoration of the Gut Microbiota and Anthropometric Profiles during the First Year Postpartum**. *Nutrients* (2019) **11** 2079. DOI: 10.3390/nu11092079
16. García-Mantrana I., Calatayud M., Romo-Vaquero M., Espín J. C., Selma M. V., Collado M. C.. **Urolithin Metabotypes Can Determine the Modulation of Gut Microbiota in Healthy Individuals by Tracking Walnuts Consumption over Three Days**. *Nutrients* (2019) **11** 2483. DOI: 10.3390/nu11102483
17. Selma M. V., Tomás-Barberán F. A., Beltrán D., García-Villalba R., Espín J. C.. *Int.
J. Syst. Evol. Microbiol.* (2014) **64** 2346-2352. DOI: 10.1099/ijs.0.055095-0
18. Selma M. V., Beltrán D., García-Villalba R., Espín J. C., Tomás-Barberán F. A.. **Description of Urolithin Production Capacity from Ellagic Acid of Two Human Intestinal**. *Food Funct.* (2014) **5** 1779-1784. DOI: 10.1039/C4FO00092G
19. Selma M. V., Beltrán D., Luna M. C., Romo-Vaquero M., García-Villalba R., Mira A., Espín J. C., Tomás-Barberán F. A.. **Isolation of Human Intestinal Bacteria Capable of Producing the Bioactive Metabolite Isourolithin A from Ellagic Acid**. *Front. Microbiol.* (2017) **8** 1521. DOI: 10.3389/fmicb.2017.01521
20. Beltrán D., Romo-Vaquero M., Espín J. C., Tomás-Barberán F. A., Selma M. V.. *Int. J. Syst. Evol. Microbiol.* (2018) **68** 1707-1712. DOI: 10.1099/ijsem.0.002735
21. Selma M. V., Romo-Vaquero M., García-Villalba R., González-Sarrías A., Tomás-Barberán F. A., Espín J. C.. **The Human Gut Microbial Ecology Associated with Overweight and Obesity Determines Ellagic Acid Metabolism**. *Food Funct.* (2016) **7** 1769-1774. DOI: 10.1039/C5FO01100K
22. Selma M. V., Tomás-Barberán F. A., Beltrán D., García-Villalba R., Espín J. C.. **Microbial Method for Reproducing the Human Urolithin Metabotypes**. (2022)
23. Haas K. N., Blanchard J. L.. **Reclassification of the**. *Int. J. Syst. Evol.
Microbiol.* (2020) **70** 23-34. DOI: 10.1099/ijsem.0.003698
24. Singh A., Andreux P., Blanco-Bose W., Ryu D., Aebischer P., Auwerx J., Rinsch C.. **Orally Administered Urolithin A Is Safe And Modulates Muscle And Mitochondrial Biomarkers In Elderly**. *Innov. Aging* (2017) **1** 1223-1224. DOI: 10.1093/geroni/igx004.4446
|
---
title: 'Sex-based epidemiological and immunovirological characteristics of people
living with HIV in current follow-up at a tertiary hospital: a comparative retrospective
study, Catalonia, Spain, 1982 to 2020'
authors:
- Sara Toyos
- Leire Berrocal
- Ana González-Cordón
- Alexy Inciarte
- Lorena de la Mora
- María Martínez-Rebollar
- Montserrat Laguno
- Emma Fernández
- Juan Ambrosioni
- Iván Chivite
- Elisa de Lazzari
- José Luis Blanco
- Esteban Martínez
- José M Miró
- Josep Mallolas
- Berta Torres
journal: Eurosurveillance
year: 2023
pmcid: PMC9999459
doi: 10.2807/1560-7917.ES.2023.28.10.2200317
license: CC BY 4.0
---
# Sex-based epidemiological and immunovirological characteristics of people living with HIV in current follow-up at a tertiary hospital: a comparative retrospective study, Catalonia, Spain, 1982 to 2020
## Abstract
### Background
Epidemiological and immunovirological features of people living with HIV (PLWH) can vary by sex.
### Aim
To investigate, particularly according to sex, characteristics of PLWH who consulted a tertiary hospital in Barcelona, Spain, in 1982–2020.
### Methods
PLWH, still in active follow-up in 2020 were retrospectively analysed by sex, age at diagnosis, age at data extraction (December 2020), birth place, CD4+ cell counts, and virological failure.
### Results
In total, 5,377 PLWH (comprising 828 women; $15\%$) were included. HIV diagnoses in women appeared to decrease from the 1990s, representing $7.4\%$ ($\frac{61}{828}$) of new diagnoses in 2015–2020. From 1997, proportions of new HIV diagnoses from patients born in Latin America seemed to increase; moreover, for women born outside of Spain, the median age at diagnosis appeared to become younger than for those born in Spain, with significant differences observed in 2005–2009 and 2010–2014 (31 vs 39 years ($$p \leq 0.001$$), and 32 vs 42 years ($p \leq 0.001$) respectively), but not in 2015–2020 (35 vs 42 years; $$p \leq 0.254$$). Among women, proportions of late diagnoses (CD4+ cells/mm3 < 350) were higher than men (significantly in 2015–2020: $62\%$ ($\frac{32}{52}$) vs $46\%$ ($\frac{300}{656}$); $$p \leq 0.030$$). Initially, virological failure rates were higher in women than men, but they were similar in 2015–2020 ($12\%$ ($\frac{6}{52}$) vs $8\%$ ($\frac{55}{659}$); $$p \leq 0.431$$). Women ≥ 50 years old represented $68\%$ ($\frac{564}{828}$) of women actively followed up in 2020.
### Conclusions
Women still have higher rates of late HIV diagnoses than men. Among currently-followed-up women, ≥ 50 year-olds, who need age-adapted care represent a high percentage. Stratifying PLWH by sex matters for HIV prevention and control interventions.
## Introduction
Worldwide, women and girls in 2021 represented ca $54\%$ of people living with HIV (PLWH) and $49\%$ of all new HIV diagnoses were in women. Across different parts of the globe, numbers of women living with HIV (WLWH) and their age at diagnosis can vary. In sub-Saharan Africa, 15 to 24-year-old women are estimated twice as likely to be living with HIV as men of the same age. In all other areas of the world, men living with HIV (MLWH) outnumber WLWH [1].
Studying WLWH and MLWH separately is important for several reasons. In terms of pathogenesis, issues related to both gender and sex may cause differences between women and men. Gender-specific behaviours and or/socio-cultural constructs could influence how HIV is acquired in each group or how it affects respective individuals. Sex-related biological aspects, such as genetic and/or hormonal factors may also lead to heterogeneous host-responses to the virus in men and women, with, for example, variable levels of immune activation at acquisition of the virus/onset of disease and differences in tolerances to antiretroviral medications [2]. In terms of preventive strategies, WLWH may receive less attention and may be excluded in countries where the epidemic disproportionally affects men who have sex with men [3].
According to the Spanish Bulletin of Epidemiological Surveillance of HIV and AIDS published in 2022, the rate of new HIV diagnoses in *Spain is* similar to that of countries in the western part of the World Health Organization (WHO) European Region. In 2021, women accounted for $11\%$ of all new HIV diagnoses in Spain, with a rate of 1.3 new HIV diagnoses per 100,000 population [4], which is slightly lower than the 1.6 new HIV diagnoses per 100,000 population reported among women across the countries of the European Economic Area in 2021 [5].
Concerning the region of Catalonia (7.5 million inhabitants), the Hospital Clínic in Barcelona, a tertiary hospital, has been one of the reference centres for HIV care since the beginning of the HIV epidemic. In 2020, 329 cases of HIV were diagnosed in Catalonia [6], and in the same year 110 treatment-naïve HIV patients had their first visit at the Hospital Clínic (according to local hospital data).
The objective of this study was to investigate epidemiological and immunovirological characteristics of PLWH at diagnosis, at Hospital Clínic, and the response of these patients to antiretroviral treatment (ART). The aim was also to understand how these features differ among PLWH, in particular according to sex, and how they evolved with time since the beginning of the HIV epidemic.
## Study type
This was a descriptive, retrospective and comparative observational study of PLWH who were followed-up in the HIV unit of Hospital Clínic from the beginning of the HIV epidemic until December 2020, when data extraction was performed.
## Source of data
Epidemiological data, such as date and place of birth, and date of HIV diagnosis, have been routinely registered into a clinical-history database of Hospital Clínic approved by the local ethics review committee since 1982. Laboratory data, such as CD4+ cell count, HIV viral load, and type of ART, have been routinely registered in the same database since 1990.
## Description of the historical and active cohorts and their use
The ‘historical cohort’ of Hospital Clínic includes all PLWH who visited Hospital Clínic since the HIV epidemic started. The ‘active cohort’ of Hospital Clínic consists of PLWH who are currently in follow-up, i.e. HIV patients who had at least one laboratory test 12 months before the data extraction was performed (December 2020).
The historical cohort was used to describe all patients with a new HIV diagnosis who ever visited our hospital. These were described according to each year of the study and stratified according to sex.
The active cohort served as a base for more extensive epidemiological, immunovirological and clinical analyses. Data in the active cohort were stratified by sex and retrospectively analysed according to successive periods within the study. The time intervals of these periods (all starting on 1 January and ending on 31 December of the given years) were determined by considering the introduction of combined ART, commercialisation of newer drugs, and changes in policies on when to start ART. Two periods pre-introduction of combined ART and four periods post-introduction of combined ART were established: period 1, from January 1982 to December 1989; period 2, from January 1990 to December 1996; period 3, from January 1997 to December 2004; period 4, from January 2005 to December 2009; period 5, from January 2010 to December 2014; and period 6, from January 2015 to December 2020.
## Epidemiological variables and laboratory parameters of active cohort patients
For the active cohort, the epidemiological variables studied were sex (referring to, throughout the manuscript, as ‘sex assigned at birth’; male or female), year of HIV diagnosis, age at HIV diagnosis, age at data extraction (December 2020), mode of HIV acquisition (sexual, transfusion of blood products, vertical transmission, or people who inject drugs) and place of birth (Spain, rest of Europe, Africa, Latin America, United States, Asia, or unknown).
Laboratory, antiretroviral and clinical data were only analysed for a subset of the active cohort starting 1990. Indeed, in patients diagnosed with HIV before 1990, missing data on the baseline CD4+ cell count, HIV viral load, and antiretroviral treatment (ART) were frequent occurrences; hence, the period from 1 January 1982 to December 1989 was excluded from analysis. Moreover, for patients transferred to Hospital Clínic from other centres, the baseline CD4+ cell count and HIV viral load were not always reported in the electronic records; therefore, for data accuracy, laboratory, antiretroviral and clinical data were analysed only for treatment-naïve HIV patients visiting in our hospital, i.e. patients who had never started ART in other centres before being transferred or before their first visit to our hospital.
The laboratory parameters taken into account were the CD4+ cell count (cells/mm3) at diagnosis, nadir CD4+ cell count (cells/mm3), and HIV viral load (copies/mL) at diagnosis. Late diagnosis was defined as CD4+ cell count < 350 cells/mm3 at diagnosis. In relation to treatment, we analysed the type of ART regimen at the initiation of treatment, number of changes in ART during the follow-up, and virological suppression and virological failure after the initiation of ART. Virological failure was defined as two consecutive viral loads of > 50 copies/mL after achieving viral suppression.
## Statistical analyses
Qualitative variables were expressed as the frequency and percentage. Quantitative variables were expressed as the median and interquartile range (IQR), as some of them were not normally distributed. Dunn’s test was used to perform multiple pairwise comparisons. For comparisons of two groups in different time periods, logistic and linear regression models with the interaction of both group and period variables were performed for each variable of interest as the dependent variable. Application criteria for both regressions were checked, including the normal distribution of the residuals for linear regression. All tests were two-tailed, and statistical significance was set at $p \leq 0.05.$ The statistical analyses were performed using Stata 17 software (StataCorp LLC, College Station, TX, United States (US)).
## Demographic evolution of the historical cohort
From the first cases registered in the early 1980s to those registered until December 2020, 11,617 PLWH were followed-up in the Hospital Clínic. Women accounted for $18\%$ ($$n = 2$$,084) of all PLWH during this period. The number of female patients with a new HIV diagnosis who ever visited Hospital Clínic appeared to increase in the late 1980s, then to progressively decrease after 1992 (Figure 1).
**Figure 1:** *Number of new HIV diagnoses per year and by sex at Hospital Clínic (A) over the whole study time (n = 11,617) and (B) in the last 10 years of the study (n = 2,926), Barcelona, Spain, 1982–2020*
## Epidemiological characteristics of the active cohort
Only PLWH who underwent active follow-up until December 2020, when the analysis was performed ($$n = 5$$,377), were included in the more detailed study of demographic and epidemiological characteristics. Women represented $15\%$ ($\frac{828}{5}$,377) of the active cohort with a median follow-up of 18 years (IQR: 10.5–23.0 years). As observed in the historical cohort, a progressive decrease in the number of HIV diagnoses in women was observed since the 1990s. New diagnoses in women represented $33.1\%$ ($\frac{274}{828}$) of the total diagnoses in 1990–1996 vs $7.4\%$ ($\frac{61}{828}$) in 2015–2020.
The analysis of epidemiological data was performed in all women ($$n = 828$$) and men ($$n = 4$$,549) in the active cohort. Most of the women who were currently in follow-up at Hospital Clínic ($73.1\%$, $\frac{605}{828}$) were diagnosed with HIV between the ages of 20 and 39 years and an age at diagnosis of ≥ 50 years was observed in $7.4\%$ ($\frac{61}{828}$) of the women, with no overall differences with the men ($75.5\%$, 3,$\frac{434}{4}$,549 and $6.2\%$, $\frac{282}{4}$,549, respectively). Women were younger at diagnosis than men, when diagnosed before 1997, but this trend reversed with time, and women appeared significantly older than men at diagnosis (36 vs 32 years old; $$p \leq 0.005$$) in the last study period from 2015 to 2020 (Table 1).
**Table 1**
| Comparison of women and men living with HIV | Comparison of women and men living with HIV.1 | Comparison of women and men living with HIV.2 | Comparison of women and men living with HIV.3 | Comparison of women and men living with HIV.4 | Comparison of women and men living with HIV.5 | Comparison of women and men living with HIV.6 | Comparison of women and men living with HIV.7 | Comparison of women and men living with HIV.8 | Comparison of women and men living with HIV.9 | Comparison of women and men living with HIV.10 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Women (n = 828)a and men’s (n = 4,549)a age at diagnosis, 1982–2020 | Women (n = 828)a and men’s (n = 4,549)a age at diagnosis, 1982–2020 | Women (n = 828)a and men’s (n = 4,549)a age at diagnosis, 1982–2020 | Women (n = 828)a and men’s (n = 4,549)a age at diagnosis, 1982–2020 | Women (n = 828)a and men’s (n = 4,549)a age at diagnosis, 1982–2020 | Women (n = 828)a and men’s (n = 4,549)a age at diagnosis, 1982–2020 | Women (n = 828)a and men’s (n = 4,549)a age at diagnosis, 1982–2020 | Women (n = 828)a and men’s (n = 4,549)a age at diagnosis, 1982–2020 | Women (n = 828)a and men’s (n = 4,549)a age at diagnosis, 1982–2020 | Women (n = 828)a and men’s (n = 4,549)a age at diagnosis, 1982–2020 | Women (n = 828)a and men’s (n = 4,549)a age at diagnosis, 1982–2020 |
| Periods | Age in years at diagnosis | Age in years at diagnosis | Age in years at diagnosis | Age in years at diagnosis | Age in years at diagnosis | Age in years at diagnosis | Age in years at diagnosis | Regression coefficient | 95% CI | p value |
| Periods | Womenn = 828 | Womenn = 828 | Womenn = 828 | Menn = 4,549 | Menn = 4,549 | Menn = 4,549 | Menn = 4,549 | Regression coefficient | 95% CI | p value |
| Periods | N | Median (IQR) | Median (IQR) | N | N | Median (IQR) | Median (IQR) | Regression coefficient | 95% CI | p value |
| 1982–1989 | 96 | 23 (20 to 26) | 23 (20 to 26) | 215 | 215 | 25 (22 to 29) | 25 (22 to 29) | −2.45 | −4.68 to −0.22 | 0.031 |
| 1990–1996 | 274 | 27 (24 to 32) | 27 (24 to 32) | 536 | 536 | 30 (26 to 36) | 30 (26 to 36) | −3.63 | −4.98 to −2.28 | < 0.001 |
| 1997–2004 | 222 | 33 (28 to 41) | 33 (28 to 41) | 842 | 842 | 33 (28 to 39) | 33 (28 to 39) | 0.33 | −1.04 to 1.70 | 0.636 |
| 2005–2009 | 100 | 35 (29 to 43) | 35 (29 to 43) | 893 | 893 | 33 (28 to 40) | 33 (28 to 40) | 1.82 | −0.10 to 3.73 | 0.063 |
| 2010–2014 | 75 | 34 (27 to 43) | 34 (27 to 43) | 1112 | 1112 | 33 (27 to 40) | 33 (27 to 40) | 1.78 | −0.39 to 3.94 | 0.109 |
| 2015–2020 | 61 | 36 (30 to 44) | 36 (30 to 44) | 951 | 951 | 32 (27 to 39) | 32 (27 to 39) | 3.44 | 1.04 to 5.84 | 0.005 |
| TOTAL | 828 | 30 (25 to 37) | 30 (25 to 37) | 4549 | 4549 | 32 (27 to 39) | 32 (27 to 39) | −1.95 | −2.66 to −1.23 | <0.001 |
| Laboratory parameters for women (n = 549)b and men (n = 2,787)b, 1990–2020 | Laboratory parameters for women (n = 549)b and men (n = 2,787)b, 1990–2020 | Laboratory parameters for women (n = 549)b and men (n = 2,787)b, 1990–2020 | Laboratory parameters for women (n = 549)b and men (n = 2,787)b, 1990–2020 | Laboratory parameters for women (n = 549)b and men (n = 2,787)b, 1990–2020 | Laboratory parameters for women (n = 549)b and men (n = 2,787)b, 1990–2020 | Laboratory parameters for women (n = 549)b and men (n = 2,787)b, 1990–2020 | Laboratory parameters for women (n = 549)b and men (n = 2,787)b, 1990–2020 | Laboratory parameters for women (n = 549)b and men (n = 2,787)b, 1990–2020 | Laboratory parameters for women (n = 549)b and men (n = 2,787)b, 1990–2020 | Laboratory parameters for women (n = 549)b and men (n = 2,787)b, 1990–2020 |
| Periods | CD4+ cells/mm3 ‒ basal | CD4+ cells/mm3 ‒ basal | CD4+ cells/mm3 ‒ basal | CD4+ cells/mm3 ‒ basal | CD4+ cells/mm3 ‒ basal | CD4+ cells/mm3 ‒ basal | CD4+ cells/mm3 ‒ basal | Comparison | Comparison | Comparison |
| Periods | Womenn = 540c | Womenn = 540c | Womenn = 540c | Menn = 2,737c | Menn = 2,737c | Menn = 2,737c | Menn = 2,737c | Regression coefficient | 95% CI | p value |
| Periods | N | Median (IQR) | Median (IQR) | N | N | Median (IQR) | Median (IQR) | Regression coefficient | 95% CI | p value |
| 1990‒1996 | 176 | 292 (200 to 407) | 292 (200 to 407) | 136 | 136 | 347 (171 to 540) | 347 (171 to 540) | −49.05 | −105.58 to 7.49 | 0.089 |
| 1997‒2004 | 176 | 291 (151 to 492) | 291 (151 to 492) | 532 | 532 | 302 (134 to 495) | 302 (134 to 495) | 4.08 | −38.98 to 47.14 | 0.853 |
| 2005‒2009 | 80 | 233 (110 to 495) | 233 (110 to 495) | 635 | 635 | 383 (235 to 550) | 383 (235 to 550) | −67.87 | −126.61 to −9.12 | 0.024 |
| 2010‒2014 | 56 | 381 (207 to 482) | 381 (207 to 482) | 778 | 778 | 376 (242 to 509) | 376 (242 to 509) | −6.38 | −74.89 to 62.14 | 0.855 |
| 2015‒2020 | 52 | 313 (153 to 445) | 313 (153 to 445) | 656 | 656 | 379 (219 to 550) | 379 (219 to 550) | −72.11 | −143.45 to −0.77 | 0.048 |
| TOTAL | 540 | 289 (176 to 457) | 289 (176 to 457) | 2737 | 2737 | 365 (213 to 528) | 365 (213 to 528) | −42.92 | −66.32 to −19.52 | <0.001 |
| Periods | CD4+ cells/mm3 ‒ nadir | CD4+ cells/mm3 ‒ nadir | CD4+ cells/mm3 ‒ nadir | CD4+ cells/mm3 ‒ nadir | CD4+ cells/mm3 ‒ nadir | CD4+ cells/mm3 ‒ nadir | CD4+ cells/mm3 ‒ nadir | Regression coefficient | 95% CI | p value |
| Periods | Womenn = 548c | Womenn = 548c | Womenn = 548c | Menn = 2,786c | Menn = 2,786c | Menn = 2,786c | Menn = 2,786c | Regression coefficient | 95% CI | p value |
| Periods | N | Median (IQR) | Median (IQR) | N | N | Median (IQR) | Median (IQR) | Regression coefficient | 95% CI | p value |
| 1990‒1996 | 184 | 227 (131 to 291) | 227 (131 to 291) | 152 | 152 | 225 (130 to 332) | 225 (130 to 332) | −13.30 | −50.47 to 23.87 | 0.483 |
| 1997‒2004 | 177 | 207 (92 to 293) | 207 (92 to 293) | 554 | 554 | 208 (99 to 311) | 208 (99 to 311) | −11.67 | −40.95 to 17.61 | 0.435 |
| 2005‒2009 | 80 | 201 (98 to 295) | 201 (98 to 295) | 641 | 641 | 288 (198 to 380) | 288 (198 to 380) | −56.28 | −96.49 to −16.07 | 0.006 |
| 2010‒2014 | 56 | 304 (189 to 379) | 304 (189 to 379) | 781 | 781 | 326 (208 to 439) | 326 (208 to 439) | −32.35 | −79.27 to 14.56 | 0.176 |
| 2015‒2020 | 51 | 269 (157 to 433) | 269 (157 to 433) | 658 | 658 | 355 (209 to 497) | 355 (209 to 497) | −63.44 | −112.73 to −14.14 | 0.012 |
| TOTAL | 548 | 225 (114 to 310) | 225 (114 to 310) | 2786 | 2786 | 289 (169 to 409) | 289 (169 to 409) | −68.04 | −84.60 to −51.47 | <0.001 |
| Periods | Log HIV viral load ‒ basal | Log HIV viral load ‒ basal | Log HIV viral load ‒ basal | Log HIV viral load ‒ basal | Log HIV viral load ‒ basal | Log HIV viral load ‒ basal | Log HIV viral load ‒ basal | Regression coefficient | 95% CI | p value |
| Periods | Womenn = 424c | Womenn = 424c | Womenn = 424c | Menn = 2,755c | Menn = 2,755c | Menn = 2,755c | Menn = 2,755c | Regression coefficient | 95% CI | p value |
| Periods | N | Median (IQR) | Median (IQR) | N | N | Median (IQR) | Median (IQR) | Regression coefficient | 95% CI | p value |
| 1990‒1996 | 78 | 4.29 (3.19 to 4.82) | 4.29 (3.19 to 4.82) | 139 | 139 | 4.81 (4.18 to 5.26) | 4.81 (4.18 to 5.26) | −0.65 | −0.91 to −0.39 | < 0.001 |
| 1997‒2004 | 163 | 4.83 (4.21 to 5.42) | 4.83 (4.21 to 5.42) | 544 | 544 | 5.10 (4.55 to 5.51) | 5.10 (4.55 to 5.51) | −0.30 | −0.46 to −0.14 | < 0.001 |
| 2005‒2009 | 77 | 4.36 (3.82 to 5.03) | 4.36 (3.82 to 5.03) | 637 | 637 | 4.69 (4.15 to 5.17) | 4.69 (4.15 to 5.17) | −0.22 | −0.44 to 0.01 | 0.055 |
| 2010‒2014 | 55 | 4.19 (3.47 to 4.94) | 4.19 (3.47 to 4.94) | 779 | 779 | 4.56 (4.02 to 5.1) | 4.56 (4.02 to 5.1) | −0.44 | −0.69 to −0.18 | 0.001 |
| 2015‒2020 | 51 | 4.31 (3.53 to 5.03) | 4.31 (3.53 to 5.03) | 656 | 656 | 4.66 (3.94 to 5.31) | 4.66 (3.94 to 5.31) | −0.32 | −0.59 to −0.06 | 0.017 |
| TOTAL | 424 | 4.47 (3.77 to 5.15) | 4.47 (3.77 to 5.15) | 2755 | 2755 | 4.73 (4.13 to 5.28) | 4.73 (4.13 to 5.28) | −0.27 | −0.36 to −0.17 | < 0.001 |
| Proportion of individuals with late diagnosis and virological failure among women (n = 549)b and men (n = 2,787)b, 1990–2020 | Proportion of individuals with late diagnosis and virological failure among women (n = 549)b and men (n = 2,787)b, 1990–2020 | Proportion of individuals with late diagnosis and virological failure among women (n = 549)b and men (n = 2,787)b, 1990–2020 | Proportion of individuals with late diagnosis and virological failure among women (n = 549)b and men (n = 2,787)b, 1990–2020 | Proportion of individuals with late diagnosis and virological failure among women (n = 549)b and men (n = 2,787)b, 1990–2020 | Proportion of individuals with late diagnosis and virological failure among women (n = 549)b and men (n = 2,787)b, 1990–2020 | Proportion of individuals with late diagnosis and virological failure among women (n = 549)b and men (n = 2,787)b, 1990–2020 | Proportion of individuals with late diagnosis and virological failure among women (n = 549)b and men (n = 2,787)b, 1990–2020 | Proportion of individuals with late diagnosis and virological failure among women (n = 549)b and men (n = 2,787)b, 1990–2020 | Proportion of individuals with late diagnosis and virological failure among women (n = 549)b and men (n = 2,787)b, 1990–2020 | Proportion of individuals with late diagnosis and virological failure among women (n = 549)b and men (n = 2,787)b, 1990–2020 |
| Periods | Late diagnosis | Late diagnosis | Late diagnosis | Late diagnosis | Late diagnosis | Late diagnosis | Late diagnosis | Odds ratio | 95% CI | p value |
| Periods | Women(total = 540)c | Women(total = 540)c | Women(total = 540)c | Men(total = 2,737)c | Men(total = 2,737)c | Men(total = 2,737)c | Men(total = 2,737)c | Odds ratio | 95% CI | p value |
| Periods | n | N | % | n | N | N | % | Odds ratio | 95% CI | p value |
| 1990‒1996 | 109 | 176 | 62 | 68 | 136 | 136 | 50 | 1.63 | 1.03 to 2.56 | 0.035 |
| 1997‒2004 | 102 | 176 | 58 | 303 | 532 | 532 | 57 | 1.04 | 0.74 to 1.47 | 0.816 |
| 2005‒2009 | 52 | 80 | 65 | 280 | 635 | 635 | 44 | 2.36 | 1.45 to 3.83 | 0.001 |
| 2010‒2014 | 25 | 56 | 45 | 350 | 778 | 778 | 45 | 0.99 | 0.57 to 1.70 | 0.960 |
| 2015‒2020 | 32 | 52 | 62 | 300 | 656 | 656 | 46 | 1.90 | 1.06 to 3.39 | 0.030 |
| TOTAL | 320 | 540 | 60 | 1301 | 2737 | 2737 | 48 | 1.61 | 1.33 to 1.94 | < 0.001 |
| Periods | Virological failure | Virological failure | Virological failure | Virological failure | Virological failure | Virological failure | Virological failure | Odds ratio | 95% CI | p value |
| Periods | Women(total = 549)c | Women(total = 549)c | Women(total = 549)c | Men(total = 2,787)c | Men(total = 2,787)c | Men(total = 2,787)c | Men(total = 2,787)c | Odds ratio | 95% CI | p value |
| Periods | n | N | % | n | N | N | % | Odds ratio | 95% CI | p value |
| 1990‒1996 | 137 | 184 | 74 | 104 | 152 | 152 | 68 | 1.35 | 0.84 to 2.17 | 0.222 |
| 1997‒2004 | 95 | 177 | 54 | 265 | 554 | 554 | 48 | 1.26 | 0.90 to 1.77 | 0.177 |
| 2005‒2009 | 30 | 80 | 38 | 163 | 641 | 641 | 25 | 1.76 | 1.08 to 2.86 | 0.023 |
| 2010‒2014 | 13 | 56 | 23 | 102 | 781 | 781 | 13 | 2.01 | 1.05 to 3.87 | 0.036 |
| 2015‒2020 | 6 | 52 | 12 | 55 | 659 | 659 | 8 | 1.43 | 0.59 to 3.50 | 0.431 |
| TOTAL | 281 | 549 | 51 | 689 | 2787 | 2787 | 25 | 3.19 | 2.65 to 3.85 | < 0.001 |
Among women, younger age at diagnosis was observed for those diagnosed before 1997 compared to women diagnosed later. In the first two periods (spanning 1982 to 1996), the median ages at diagnosis were 23 and 27 years, while in the four subsequent periods these were 33, 35, 34, and 36 years ($p \leq 0.001$ for all periods). When we grouped age at diagnosis by place of birth, women born in a foreign country with an HIV diagnosis after 2005 were younger than women born in Spain, and this difference was significant in both the 2005–2009 (31 vs 39 years old; $$p \leq 0.001$$) and 2010–2014 (32 vs 42 years old; $p \leq 0.001$) periods (Table 2 and Figure 2). Also, for all new diagnoses among women under 40 years in 2015–2020, almost half of them ($49\%$, $\frac{18}{37}$) were born in Latin America, in contrast with $27\%$ ($\frac{10}{37}$) of women born in Spain for the same period.
At the time of data extraction, women of 50 years or older accounted for $68\%$ ($\frac{564}{828}$) of all WLWH in active follow-up, compared with $42\%$ (1,$\frac{928}{4}$,549) of men in the same age group (Figure 3).
**Figure 3:** *Age groups of women and mena in active follow-up stratified by the periods of HIV diagnosis, Barcelona, Spain, 1982–2020 (n = 5,377)*
The main mode of HIV acquisition for women diagnosed in the first part of the study (1982–1989), was intravenous drug injection ($59\%$, $\frac{57}{96}$, with three women among the 96 lacking information on mode of HIV acquisition). After that period, sexual contact was the primary mode of acquisition ($70\%$, $\frac{192}{274}$ in 1990‒1996; $82\%$, $\frac{181}{222}$ in 1997‒2004; $88\%$, $\frac{88}{100}$ in 2005‒2009; $81\%$, $\frac{61}{75}$ in 2010‒2014 and $89\%$, $\frac{54}{61}$ in 2015‒2020). A similar phenomenon was observed in men (Supplementary Figure 1).
Regarding the place of birth, a progressive increase in the percentage of foreign-born men and women, composed mainly of people originating from Latin America, has been observed since 1997. Overall, in the active cohort, men born in Latin America accounted for $34\%$ (1,$\frac{273}{3}$,698) of the MLWH with available information on place of birth, and women born in Latin America accounted for $14\%$ ($\frac{117}{820}$) of WLWH with such information. Men born in Latin America represented only $9\%$ ($\frac{33}{388}$) of the active cohort before 1997. However, in the period 2015–2020, they represented $52\%$ ($\frac{466}{894}$) of those with a new diagnosis, in contrast to $36\%$ ($\frac{318}{894}$) for men born in Spain. The same phenomenon was observed for women in the active cohort where, in terms of HIV diagnoses, proportions of women originating from Latin America outnumbered those of women born in Spain in the last two periods of the study, with respectively $40\%$ ($\frac{30}{75}$) vs $37\%$ ($\frac{28}{75}$) in the 2010–2014 period, and $43\%$ ($\frac{26}{60}$) vs $37\%$ ($\frac{22}{60}$) in the 2015–2020 period (Figure 4).
**Figure 4:** *Respective proportions of men and women living with HIV according to their place of birth and period of diagnosis, Barcelona, Spain, 1982–2020 (n = 4,518)a*
## HIV viral load, CD4+ cell count at diagnosis, and nadir CD4+ cell count
As explained in the Methods section, only treatment-naïve WLWH ($$n = 549$$) and MLWH ($$n = 2$$,787) who were diagnosed with HIV after 1990 were included in the immunovirological data analyses.
The viral load at diagnosis was lower in the women (median: 4.47 log10 copies/mL; IQR: 3.77–5.15) than in the men (median: 4.73 log10 copies/mL; IQR: 4.13–5.28), and this difference was statistically significant in all periods (Table 1), except for in the period of 2005–2009, in which, despite clinical relevance, the p value was close to being statistically significant though not ($$p \leq 0.055$$) (Figure 5-C).
**Figure 5:** *Characteristics of treatment-naïve HIV patients in terms of (A) nadir CD4+ cell count (cells/mm3) (n = 3,334)a, (B) CD4+ cell count at diagnosis (cells/mm3) (n = 3,277)a, (C) basal log viral load (n = 3,179)a, (D) late diagnosis (n = 3,277)a, Barcelona, Spain, 1990‒2020*
The median basal CD4+ cell count at diagnosis for women was 289 cells/mm3 (IQR: 176–457), and there was no difference between the periods. The median basal CD4+ cell count at diagnosis in men was 365 cells/mm3 (IQR: 213–528). These values were significantly lower in the women than in the men in the periods 2005–2009 ($$p \leq 0.024$$) and 2015–2020 ($$p \leq 0.048$$) (Table 1 and Figure 5-B).
The nadir CD4+ cell count was lower in the women (median: 225 cells/mm3; IQR: 114–310) than in the men (median: 289 cells/mm3; IQR: 169–409), and the difference was statistically significant for the same periods as those for the low basal CD4+ cell count (Table 1 and Figure 5-A).
Overall, a higher percentage of women than men presented with a late diagnosis (CD4+ cells/mm3 < 350 at the first visit). In the last period (2015–2020), $62\%$ of the women had a late diagnosis, whereas $46\%$ of the men had a late diagnosis ($$p \leq 0.030$$) (Table 1), with no differences observed between the Spanish and foreign-born women (Table 2). Women who presented with a late diagnosis in the periods 1997–2004 and 2015–2020 were significantly older than those who did not, with medians of 35 years old (IQR: 30‒44) vs 31 years old (IQR: 26‒37) ($$p \leq 0.006$$) and 37 years old (IQR: 32–45) vs 33 years old (IQR: 25.5‒42) ($$p \leq 0.039$$), respectively. Except for the period 2010–2014, the percentage of late diagnosis in women remained stable at ca $60\%$. In contrast, there was a progressive decrease in the percentage of a late diagnosis in men after 2004, after which it remained stable at ca $45\%$ of all new diagnoses (Table 1 and Figure 5-D).
## Antiretroviral therapy (ART) and virological failure
The analysis of the first ART was performed in only the actively-followed-treatment-naïve HIV patients who were diagnosed with HIV after 1990. The first regimen of ART changed over time according to drug approval and treatment guidelines. Of the women diagnosed in the period 2015–2020, $75\%$ ($\frac{38}{51}$) received the first regimen based on integrase strand transfer inhibitors (INSTI), and a similar percentage of men did ($73\%$; $\frac{479}{655}$), which is compared with $21\%$ ($\frac{12}{56}$) and $22\%$ ($\frac{174}{777}$) in women and men, respectively, in the previous period (2010–2014). Changes in ART were more frequent in women than in men who were diagnosed in the first period until 2009. After that year, there were no differences in the numbers of ARTs administered between the women and men (Supplementary Figure 2).
Once ART was started, $99\%$ ($\frac{547}{549}$) of treatment-naïve WLWH achieved an undetectable viral load at any point. Virological failure (defined as two consecutive viral loads of > 50 copies/mL after starting ART and achieving viral suppression) after 2005 was significantly higher in the women than in the men until December 2014 ($$p \leq 0.023$$ and $$p \leq 0.036$$ in the periods 2005–2009 and 2010–2014, respectively) (Table 1), with no differences between women born in Spain and foreign-born women (Table 2). However, the proportion of women with virological failure sharply decreased in the period 2015–2020, when it was not significantly different from the proportion of men with virological failure ($12\%$ vs $8\%$; $$p \leq 0.431$$) (Table 1 and Figure 6).
**Figure 6:** *Line graph representing the proportion of individuals with virological failure among treatment-naïve HIV patients after introduction of ART and viral suppression achievement, Barcelona, Spain, 1990–2020 (n = 3,336)*
## Discussion
Women represented $18\%$ of all PLWH who visited our cohort at Hospital Clínic (Barcelona), which is a similar result to what has been recently published in a Spanish national cohort analysis, where women represented $16.6\%$ of the total cohort [7].
The number of women who visited our centre and who had a diagnostic examination between 2015 and 2020 has remained low and was much lower than that of men.
When we compared women and men who are in active follow-up at our centre, we observed a similar distribution in both the current age and age at diagnosis. A younger age at diagnosis was observed in the first period of the study, probably because intravenous drug injection was the main mode of HIV acquisition in that period. When we compared age by the place of birth in women, a younger age at diagnosis was observed in the foreign-born women than in the Spanish women after 2005. This finding is relevant because of the childbearing potential of younger women and the possibility that many of them may be nulliparous at the time of diagnosis. It would then be necessary to provide adequate information and adapt initial ART in these cases [8]. Possible reasons for the younger age at diagnosis in foreign-born women (most of them being Latin American) than in Spanish women could be a higher perception of risk in this community [9] or an earlier screening in the reproductive health controls due to pregnancies at younger ages among the immigrant population in Spain, as stated in the last published report of the Spanish National Institute of Statistics [10]. If these were the causes, we would expect to see a higher rate of a late diagnosis in Spanish women; however, we did not observe any differences between the foreign-born and women born in Spain in the rate of a late diagnosis. Another possible explanation for the younger age at diagnosis of foreign-born women could be exposure to risks for HIV at an earlier age than women born in Spain, however, to the best of our knowledge, we are not aware of reports to support this hypothesis.
The introduction and constant improvement in ART has translated into a prolonged life expectancy in PLWH. HIV is now considered a chronic condition, and PLWH present distinct comorbidities according to age, in addition to possible adverse long-term effects of antiretroviral medication and the HIV infection itself. The fact that women aged > 50 years accounted for almost $70\%$ of all women in active follow-up in our cohort raises concerns about the need to screen for pathologies that are prevalent in women at this age, such as osteoporosis, dyslipidaemia, and cardiovascular diseases, to include them in preventive strategies (i.e. breast cancer screening), in addition to addressing potential long-term adverse effects of previous or current ART.
A progressive increase in the number of HIV diagnoses in people from Latin America has been observed, which coincides with the migratory flow into Spain during the past 20 years [11]. The Spanish HIV surveillance agency reported that $33.9\%$ of new HIV diagnoses in Spain in 2020 were in foreign-born individuals [4].
These findings are consistent with those presented in recent years by various Spanish working groups [7,12,13], and they have important clinical implications because they have been related to differences in aspects such as risk behaviours and diagnostic delays (although not demonstrated in our cohort), coinfection with other pathogens at diagnosis, presentation of disease, adherence to treatment, or treatment failure [12,14].
Sexual transmission was the main mode of HIV acquisition in both the men and women in the active cohort. Intravenous drug injection as a mode of acquisition in our cohort was mainly limited to those patients still in follow-up who were diagnosed in early years, in accordance with what has been observed in the rest of the European Union (EU), and in contrast to that observed in the Eastern part of the WHO European Region, where rates of people who inject drugs in PLWH remain high.
The percentage of individuals in our cohort with a late diagnosis was high and was significantly higher in women than in men. In addition, according to the year of diagnosis, the percentage of those with a late diagnosis has remained stable over time in the women’s cohort, being as high as $62\%$ in the 2015–2020 period, in comparison with the progressive decrease observed in men. Our findings differ from those reported in the EU, where a decrease in the percentage of those with a late diagnosis in the past 10 years has occurred for all groups and modes of transmission [15].
Some studies performed in Spain have described the ‘migrant status’ as a risk factor for a late diagnosis of HIV [16]. However, we did not observe any differences between the Spanish- and foreign-born women. These findings are in line with those recently presented by the European Centre for Disease Prevention and Control (ECDC) and the WHO Regional Office for Europe, in which the possible risk factors and predictors for a late diagnosis among European women were analysed. No association was demonstrated between a late diagnosis and migrant status, but there was an association with age (older age at diagnosis was associated with a higher risk of a late diagnosis) and the mode of acquisition only in the Eastern part of the WHO European Region (where women who acquired HIV through drug injection were less likely to have a late diagnosis than women who acquired HIV through heterosexual transmission) [17].
Having a late diagnosis is associated with high morbidity and short-term mortality [18,19]. According to the 2020 data from the ECDC, $51\%$ of those with a new infection were diagnosed with HIV while having a CD4+ cell count of less than 350 cells/mm3. A late diagnosis was more frequent in women, older adults, people who acquired HIV through heterosexual sex, people who inject drugs, and migrants from South and South-east Asia, sub-Saharan Africa, and central and eastern Europe. The last epidemiological surveillance bulletin for HIV and AIDS in Spain reported that the percentage of late diagnoses was $49.8\%$ and that the percentage was higher in women than in men ($54.4\%$ vs $49.2\%$) [4]. These findings are similar to those observed in our cohort.
It is striking that 40 years after the onset of the HIV pandemic, the percentage of those with a late diagnosis did not decrease considerably. Although evidence suggests that the current tools of European health services (including sexual health programmes, quality of care in primary care centres, notification systems for sexual partners, and tests for patients with indicator diseases) are effective interventions that favour an early diagnosis, their coverage is still limited [15,20]. This means that further efforts towards HIV prevention across Europe are warranted to achieve universal health coverage for all and meet the Sustainable Development Goal 3 target of ending AIDS by 2030 [21].
From a virological point of view, women presented significantly lower viral loads than men at the time of entering the cohort throughout all periods. Lower viral loads at diagnosis in women have been widely described in the literature since the early years of research on HIV [22-24]. This finding has been recognised as of special importance in terms of curative strategies [25].
Women achieved viral suppression in the majority of cases, as did men. However, women experience more episodes of virological failure. Some studies performed in the United States have reported higher rates of virological failure in women in relation to ethnic and sociodemographic factors [26]. Studies in the first decade of the present century reported poorer treatment adherence in women [27], possibly related to a higher prevalence of adverse effects in women than in men [28]. Differences in ART tolerance have been reported in a few studies [29]. Adverse effects are a well-documented reason for changing ART, and actual evidence demonstrates the need for ART changes to improve the quality of life of PLWH [30]. In our study, changes in ART before 2009 were more frequent in women than in men, but men and women exhibited similar rates after that year. A Swedish study recently described that the frequency of reported side effects significantly decreased from 2011 to 2017, coinciding with a shift in ART prescriptions from efavirenz to dolutegravir [31]. Currently, as many as $62\%$ of the women in our cohort are undergoing INSTI treatment, and episodes of virological failure have been far less frequent in our cohort since 2015, which coincides with the introduction of second-generation INSTIs. This could be explained by the better tolerance to ART exhibited by this group.
However, weight gain after the initiation of ART is a growing concern in PLWH. In a pooled analysis of eight randomised clinical trials from 2003 to 2019 by Sax et al. [ 2020], regimens containing INSTI were associated with greater weight gain than non-nucleoside reverse transcriptase inhibitor (NNRTI)-based or protease inhibitor (PI)-based regimens [32]. Demographic factors, such as female sex and black race/ethnicity have also been associated with greater weight gain [33,34]. This last issue requires research efforts to elucidate the causes of weight gain with these regimens that specifically affect women and to be able to offer other options that are equivalent in efficacy and tolerance.
The main strengths of this study are the large number of PLWH included, long-term follow-up, and analysis of the results by periods, which helps to consider the different epidemiological contexts throughout the HIV pandemic and its evolution in our area. However, we are aware of the limitations of this study. Some data were missing owing to its retrospective nature, and we did not include all PLWH who had visited the hospital for data accuracy reasons and only analysed patients who were in current follow-up. Finally, the fact that this was a single-centre study makes it difficult to extrapolate results to other contexts/areas. There is a high percentage of Latin American immigrants in Barcelona (cf.d with the rest of Spain); however, immigrants from other areas are lacking or underrepresented.
## Conclusions
In conclusion, the diagnosis of HIV in women in our area has decreased significantly in recent years. However, a steady and progressive increase in Latin American women with HIV has been observed, and in 2015–2020, they represent almost half of new HIV diagnoses at a younger age (i.e. < 40 years old) in women. The percentage of those with a late diagnosis remained high in our cohort in general, but the women had a higher percentage of a late diagnosis than the men. Finally, a high percentage of women aged ≥ 50 years are currently in follow-up. WLWH are still underrepresented in the scientific literature in demographic, epidemiological, pharmacological, and long-term follow-up. Determining these characteristics in particular settings is of great importance for establishing preventive and clinical interventions that are frequently not prioritised in favour of other populations with the highest risk.
## References
1. 1UNAIDS. The Joint United Nations Programme on HIV and AIDS (UNAIDS). Global HIV&AIDS statistics. Fact sheet, 2022. [Accessed 28 Oct 2022]. Available from: https://www.unaids.org/en/resources/fact-sheet
2. Scully EP. **Sex Differences in HIV Infection.**. *Curr HIV/AIDS Rep* (2018.0) **15** 136-46. DOI: 10.1007/s11904-018-0383-2
3. Bozzette SA, Berry SH, Duan N, Frankel MR, Leibowitz AA, Lefkowitz D. **The care of HIV-infected adults in the United States.**. *N Engl J Med* (1998.0) **339** 1897-904. DOI: 10.1056/NEJM199812243392606
4. 4Unidad de vigilancia de VIH, ITS y hepatitis. Vigilancia Epidemiológica del VIH y sida en España 2021: Sistema de Información sobre Nuevos Diagnósticos de VIH y Registro Nacional de Casos de Sida. [Epidemiological Surveillance of HIV and AIDS in Spain 2021: Information System on New HIV Diagnoses and National Registry of AIDS Cases]. Centro Nacional de Epidemiología. Instituto de Salud Carlos III/ División de control de VIH, ITS, Hepatitis virales y tuberculosis. Madrid: Ministerio de Sanidad; Nov 2022. Spanish.
5. 5WHO Regional Office for Europe (WHO/Europe), European Centre for Disease Prevention and Control (ECDC). HIV/AIDS surveillance in Europe 2022 – 2021 data. Copenhagen: WHO Europe, Stockholm: ECDC; 2022.
6. 6Centre d’Estudis Epidemiològics sobre les Infeccions de Transmissió Sexual i Sida de Catalunya (CEEISCAT). Vigilància epidemiològica del VIH i la SIDA a Catalunya. Informe anual 2020. [Epidemiological surveillance of HIV and AIDS in Catalonia. Annual report 2020]. Badalona; 2021. Catalan.
7. Muñoz Hornero C, Muriel A, Montero M, Iribarren JA, Masía M, Muñoz L. **Differences in epidemiology and mortality between men and women with HIV infection in the CoRIS cohort from 2004 to 2014.**. *Enferm Infecc Microbiol Clin (Engl Ed)* (2021.0) **39** 372-82. DOI: 10.1016/j.eimce.2021.07.003
8. Kowalska JD, Aebi-Popp K, Loutfy M, Post FA, Perez-Elias MJ, Johnson M. **Promoting high standards of care for women living with HIV: position statement from the Women Against Viruses in Europe Working Group.**. *HIV Med* (2018.0) **19** 167-73. DOI: 10.1111/hiv.12565
9. Hoyos J, Fernández-Balbuena S, de la Fuente L, Sordo L, Ruiz M, Barrio G. **Never tested for HIV in Latin-American migrants and Spaniards: prevalence and perceived barriers.**. *J Int AIDS Soc* (2013.0) **16** 18560. DOI: 10.7448/IAS.16.1.18560
10. 10Instituto Nacional de Estadística (INE). INEbase. Encuesta de Fecundidad. Año 2018. Resultados definitivos. [INEbase. Fertility Survey. Year 2018. Final results]. 2018. Spanish.
11. 11Instituto Nacional de Estadística (INE). INEbase: Principales series de población desde 1998. Población extranjera por nacionalidad, comunidades, sexo y año n.d. [INEbase: Main population series since 1998. Foreign population by nationality, communities, sex and year]. Spanish. [Accessed 28 Oct 2022]. Available from: https://www.ine.es/jaxi/Datos.htm?path=/t20/e245/p08/l0/&file=02005.px#!tabs-grafico.
12. Reyes-Urueña J, Campbell C, Hernando C, Vives N, Folch C, Ferrer L. **Differences between migrants and Spanish-born population through the HIV care cascade, Catalonia: an analysis using multiple data sources.**. *Epidemiol Infect* (2017.0) **145** 1670-81. DOI: 10.1017/S0950268817000437
13. Ayerdi Aguirrebengoa O, Vera Garcia M, Puerta López T, Clavo Escribano P, Ballesteros Martín J, Lejarrag Cañas C. **Changes in the profile of newly HIV-diagnosed men who have sex with men, Madrid, 2014 to 2019.**. *Euro Surveill* (2021.0) **26**. DOI: 10.2807/1560-7917.ES.2021.26.47.2001501
14. 14European Centre for Disease Prevention and Control (ECDC). HIV and migrants: Monitoring implementation of the Dublin Declaration on partnership to fight HIV/AIDS in Europe and Central Asia (2018 progress report). Stockholm: ECDC; 2019.
15. 15European Centre for Disease Prevention and Control (ECDC). Public health guidance on HIV, hepatitis B and C testing in the EU/EEA: An integrated approach. Stockholm: 2018.
16. Carnicer-Pont D, Vives N, Casabona I Barbarà J. **Epidemiología de la infección por virus de la inmunodeficiencia humana. Retraso en el diagnóstico. [Epidemiology of human inmunodeficiency virus infection: delay in diagnosis]**. *Enferm Infecc Microbiol Clin* (2011.0) **29** 144-51. DOI: 10.1016/j.eimc.2010.11.010
17. Mårdh O, Quinten C, Kuchukhidze G, Seguy N, Dara M, Amato-Gauci AJ. **HIV among women in the WHO European Region - epidemiological trends and predictors of late diagnosis, 2009-2018.**. *Euro Surveill* (2019.0) **24**. DOI: 10.2807/1560-7917.ES.2019.24.48.1900696
18. Croxford S, Kitching A, Desai S, Kall M, Edelstein M, Skingsley A. **Mortality and causes of death in people diagnosed with HIV in the era of highly active antiretroviral therapy compared with the general population: an analysis of a national observational cohort.**. *Lancet Public Health* (2017.0) **2** e35-46. DOI: 10.1016/S2468-2667(16)30020-2
19. Ang LW, Toh MPHS, Wong CS, Boudville IC, Archuleta S, Lee VJM. **Short-term mortality from HIV-infected persons diagnosed from 2012 to 2016: Impact of late diagnosis of HIV infection.**. *Medicine (Baltimore)* (2021.0) **100**. DOI: 10.1097/MD.0000000000026507
20. 20World Health Organization (WHO). HIV/AIDS. Consolidated guidelines on HIV testing services 2019. Geneva: WHO. [Accessed 12 Nov 2019]. Available from: https://www.who.int/hiv/pub/ guidelines/hiv-testing-services/en/
21. 21United Nations (UN). Resolution adopted by the General Assembly on 25 September 2015. 70/1. Transforming our world: the 2030 Agenda for Sustainable Development. A/RES/70/1 2015.
22. Farzadegan H, Hoover DR, Astemborski J, Lyles CM, Margolick JB, Markham RB. **Sex differences in HIV-1 viral load and progression to AIDS.**. *Lancet* (1998.0) **352** 1510-4. DOI: 10.1016/S0140-6736(98)02372-1
23. Moore AL, Kirk O, Johnson AM, Katlama C, Blaxhult A, Dietrich M. **Virologic, immunologic, and clinical response to highly active antiretroviral therapy: the gender issue revisited.**. *J Acquir Immune Defic Syndr* (2003.0) **32** 452-61. DOI: 10.1097/00126334-200304010-00017
24. Gandhi M, Bacchetti P, Miotti P, Quinn TC, Veronese F, Greenblatt RM. **Does patient sex affect human immunodeficiency virus levels?**. *Clin Infect Dis* (2002.0) **35** 313-22. DOI: 10.1086/341249
25. 25World Health Organization (WHO). Guidelines for managing advanced HIV disease and rapid initiation of antiretroviral therapy, July 2017. Geneva; 2017.
26. Bhagwat P, Kapadia SN, Ribaudo HJ, Gulick RM, Currier JS. **Racial Disparities in Virologic Failure and Tolerability During Firstline HIV Antiretroviral Therapy.**. *Open Forum Infect Dis* (2019.0) **6**. DOI: 10.1093/ofid/ofz022
27. Puskas CM, Forrest JI, Parashar S, Salters KA, Cescon AM, Kaida A. **Women and vulnerability to HAART non-adherence: a literature review of treatment adherence by gender from 2000 to 2011.**. *Curr HIV/AIDS Rep* (2011.0) **8** 277-87. DOI: 10.1007/s11904-011-0098-0
28. Kempf M-C, Pisu M, Dumcheva A, Westfall AO, Kilby JM, Saag MS. **Gender differences in discontinuation of antiretroviral treatment regimens.**. *J Acquir Immune Defic Syndr* (2009.0) **52** 336-41. DOI: 10.1097/QAI.0b013e3181b628be
29. Soon GG, Min M, Struble KA, Chan-Tack KM, Hammerstrom T, Qi K. **Meta-analysis of gender differences in efficacy outcomes for HIV-positive subjects in randomized controlled clinical trials of antiretroviral therapy (2000-2008).**. *AIDS Patient Care STDS* (2012.0) **26** 444-53. DOI: 10.1089/apc.2011.0278
30. Maiese EM, Johnson PT, Bancroft T, Goolsby Hunter A, Wu AW. **Quality of life of HIV-infected patients who switch antiretroviral medication due to side effects or other reasons.**. *Curr Med Res Opin* (2016.0) **32** 2039-46. DOI: 10.1080/03007995.2016.1227776
31. Mellgren Å, Eriksson LE, Reinius M, Marrone G, Svedhem V. **Longitudinal trends and determinants of patient-reported side effects on ART-a Swedish national registry study.**. *PLoS One* (2020.0) **15**. DOI: 10.1371/journal.pone.0242710
32. Sax PE, Erlandson KM, Lake JE, Mccomsey GA, Orkin C, Esser S. **Weight Gain Following Initiation of Antiretroviral Therapy: Risk Factors in Randomized Comparative Clinical Trials.**. *Clin Infect Dis* (2020.0) **71** 1379-89. DOI: 10.1093/cid/ciz999
33. Bakal DR, Coelho LE, Luz PM, Clark JL, De Boni RB, Cardoso SW. **Obesity following ART initiation is common and influenced by both traditional and HIV-/ART-specific risk factors.**. *J Antimicrob Chemother* (2018.0) **73** 2177-85. DOI: 10.1093/jac/dky145
34. Yuh B, Tate J, Butt AA, Crothers K, Freiberg M, Leaf D. **Weight change after antiretroviral therapy and mortality.**. *Clin Infect Dis* (2015.0) **60** 1852-9. DOI: 10.1093/cid/civ192
|
---
title: Weight-dependent and weight-independent effects of dulaglutide on blood pressure
in patients with type 2 diabetes
authors:
- Keith C. Ferdinand
- Julia Dunn
- Claudia Nicolay
- Flora Sam
- Emily K. Blue
- Hui Wang
journal: Cardiovascular Diabetology
year: 2023
pmcid: PMC9999488
doi: 10.1186/s12933-023-01775-x
license: CC BY 4.0
---
# Weight-dependent and weight-independent effects of dulaglutide on blood pressure in patients with type 2 diabetes
## Abstract
### Background
Patients with type 2 diabetes (T2D) treated with glucagon-like peptide-1 receptor agonists may experience reductions in weight and blood pressure. The primary objective of the current study was to determine the weight-dependent and weight-independent effects of ~ 6 months treatment with dulaglutide 1.5 mg treatment in participants with T2D.
### Methods
Mediation analysis was conducted for five randomized, placebo-controlled trials of dulaglutide 1.5 mg to estimate the weight-dependent (i.e., mediated by weight) and weight-independent effects from dulaglutide vs. placebo on change from baseline for systolic blood pressure (SBP), diastolic blood pressure (DBP), and pulse pressure. A random-effects meta-analysis combined these results. To investigate a dose response between dulaglutide 4.5 mg and placebo, mediation analysis was first conducted in AWARD-11 to estimate the weight-dependent and weight-independent effects of dulaglutide 4.5 mg vs. 1.5 mg, followed by an indirect comparison with the mediation result for dulaglutide 1.5 mg vs. placebo.
### Results
Baseline characteristics were largely similar across the trials. In the mediation meta-analysis of placebo-controlled trials, the total treatment effect of dulaglutide 1.5 mg after placebo-adjustment on SBP was − 2.6 mmHg ($95\%$ CI − 3.8, − 1.5; $p \leq 0.001$) and was attributed to both a weight-dependent effect (− 0.9 mmHg; $95\%$ CI: − 1.4, − 0.5; $p \leq 0.001$) and a weight-independent effect (− 1.5 mmHg; $95\%$ CI: − 2.6, − 0.3; $$p \leq 0.01$$), accounting for $36\%$ and $64\%$ of the total effect, respectively. For pulse pressure, the total treatment effect of dulaglutide (− 2.5 mmHg; $95\%$ CI: − 3.5, − 1.5; $p \leq 0.001$) was $14\%$ weight-dependent and $86\%$ weight-independent. For DBP there was limited impact of dulaglutide treatment, with only a small weight-mediated effect. Dulaglutide 4.5 mg demonstrated an effect on reduction in SBP and pulse pressure beyond that of dulaglutide 1.5 mg which was primarily weight mediated.
### Conclusions
Dulaglutide 1.5 mg reduced SBP and pulse pressure in people with T2D across the placebo-controlled trials in the AWARD program. While up to one third of the effect of dulaglutide 1.5 mg on SBP and pulse pressure was due to weight reduction, the majority was independent of weight. A greater understanding of the pleotropic effects of GLP-1 RA that contribute to reduction in blood pressure could support developing future approaches for treating hypertension.
Trial registrations (clinicaltrials.gov) NCT01064687, NCT00734474, NCT01769378, NCT02597049, NCT01149421, NCT03495102
### Supplementary Information
The online version contains supplementary material available at 10.1186/s12933-023-01775-x.
## Introduction
Elevated blood pressure (BP) is highly prevalent and is reported in about three-fourths of patients with type 2 diabetes (T2D) [1]. The role of elevated BP in both macrovascular and microvascular complications is well-established in patients with T2D [2, 3] and is recognized as a common and robust predisposing risk factor for cardiovascular disease [4, 5]. Importantly, relatively small reductions in the mean systolic blood pressure (SBP) (2–5 mmHg) are sufficient to reduce cardiovascular events and death in a population (Fig. 1) [6]. In a meta-analysis of recent blood pressure trials, the Blood Pressure Lowering Treatment Trialists’ Collaboration reported that a 5-mmHg reduction in SBP lowers the risk of major cardiovascular events by $10\%$, and the benefit was independent of the baseline presence of cardiovascular disease [7]. In a large meta-regression analysis of patients with diabetes, the risk of stroke decreased by $13\%$ for each 5-mmHg reduction in SBP and by $11.5\%$ for each 2-mmHg reduction in diastolic blood pressure (DBP) [8]. Similarly, results from the UK Prospective Diabetes Study (UKPDS) showed that among patients with mean SBP ranging from 159–160 mmHg, a 10-mmHg reduction in SBP decreased diabetes-related mortality by $15\%$ and all-cause mortality by $12\%$ [9].Fig. 1Distribution of systolic blood pressure before and after intervention. BP blood pressure, CHD coronary heart disease. Adapted from Whelton PK, et al. [ 6] The current American College of Cardiology/American Heart Association multi-society guideline recommends individualized BP treatment targets for patients with T2D based on cardiovascular risk, with a goal of < $\frac{130}{80}$ mmHg; individualized targets should account for patient tolerance of the BP level [10]. However, up to half of patients do not achieve these goals [11, 12]. Since small reductions in BP occur with glucagon-like peptide-1 receptor agonist (GLP-1 RA) treatment, it has been proposed that these reductions in BP and the specific mechanisms contributing to BP reduction with GLP-1 RAs may be relevant to decreases in major adverse cardiovascular events (MACE) observed with GLP-1 RA treatment [13, 14]. The specific mechanisms contributing to BP reduction with GLP-1 RAs are unknown, but improvement in arterial stiffness is a probable mechanism [15].
Treatment of patients with T2D with GLP-1 RAs also results in modest weight reduction [16] which may contribute to the long-term reduction in BP, as well as prevent the onset of hypertension [17, 18]. However, previous large metanalyses have reported conflicting results [18, 19]. In a randomized, placebo-controlled clinical trial to characterize the effect of dulaglutide 1.5 mg vs. placebo on BP and heart rate in participants with T2D using 24-h ambulatory BP monitoring, participants who received dulaglutide demonstrated a 2.5 mmHg decrease in 24-h SBP vs. an increase of 0.2 mmHg with placebo at week 26, with limited change in DBP. There was no significant association found between weight reduction and BP reduction occurring with dulaglutide treatment in this single trial [20].
The current study aimed to better characterize the relationship between dulaglutide treatment and changes in BP and pulse pressure in participants with T2D. The primary objective was to determine the weight-dependent and weight-independent effects of treatment through a mediation meta-analysis of the all the dulaglutide placebo-controlled trials. The secondary objective was to investigate the effect of higher dose dulaglutide treatment on the weight-dependent and weight-independent effects of blood pressure change.
## Studies
The analyses included six pivotal randomized, double-blind trials of dulaglutide 1.5 mg in participants with T2D that measured sitting SBP and DBP from vital sign data around the timeline of 6 months (week 24 to week 26). Five placebo-controlled studies were used to estimate the effects between dulaglutide 1.5 mg and placebo. AWARD-1 (NCT01064687), AWARD-5 (NCT00734474), AWARD-8 (NCT01769378), and AWARD-10 (NCT02597049) were phase 3, placebo-controlled trials which investigated the safety and glycemic efficacy of dulaglutide with various background glycemic therapies (Table 1). Ferdinand et al. ( NCT01149421) was a phase 2, randomized, double-blind, placebo-controlled trial which evaluated BP and heart rate effects of dulaglutide vs. placebo in participants with T2D with and without hypertension and BP < $\frac{140}{90}$ mmHg. In addition, AWARD-11 (NCT03495102) was a phase 3, non-placebo-controlled trial to evaluate safety and glycemic efficacy of dulaglutide 3.0 mg and 4.5 mg to dulaglutide 1.5 mg. Table 1Study design for placebo-controlled trials included in the meta-analysisParametersAWARD-1AWARD-5AWARD-8AWARD-10AWARD-11Ferdinand et alPhasePhase IIIPhase II/IIIPhase IIIPhase IIIPhase IIIPhase IIRandomizationRandomizedRandomizedRandomizedRandomizedRandomizedRandomizedBlindingBlindingDouble-blindDouble-blindDouble-blindDouble-blindDouble-blindPrimary EndpointA1cA1cA1cA1cA1c24-h SBPStudy Treatment Period52 weeks24 months24 weeks24 weeks52 weeks26 weeksLast scheduled visit with PBO26 weeks6 months24 weeks24 weeks52 weeks (no PBO)26 weeksBackground therapy (Add-ons)Met + TZDMet monoSU monoSGLT2i with or without metforminMet monoStable OAMKey inclusion/ exclusion criteria Age ≥ 18 years18–75 years ≥ 18 years ≥ 18 years ≥ 18 years ≥ 18 years T2D durationNA ≥ 6 monthsNANAfor ≥ 6 monthsNA A1c7.0–11.07.0–9.57.5–9.57.0–9.57.5–117–9.5 BMI23–4525–40 ≤ 45 ≤ 45 ≥ 25NA MedicationStable OAMDiet & exercise / metformin and/or other OAMStable SUSGLT2i with or without metformin for ≥ 3 monthsStable metformin for ≥ 3 monthsOAMBMI body mass index, NA not applicable for the study’s design, Met metformin, mono monotherapy, OAM oral antihyperglycemic medication, PBO placebo, SBP systolic blood pressure, SGLT2i sodium-glucose cotransporter-2 inhibitors, SU sulfonylurea, T2D type 2 diabetes, TZD thiazolidinediones
## Statistical analysis
The primary objective included mediation analyses of each of the five placebo-controlled trials (Ferdinand et al., AWARD-1, AWARD-5, AWARD-8, and AWARD-10) followed by a meta-analysis pooling the individual mediation results in a random-effect model. In the mediation analyses, the total effect of dulaglutide 1.5 mg vs. Placebo was decomposed into a weight-dependent (i.e., mediated by weight) and weight-independent effect on changes from baseline for SBP, DBP, and pulse pressure. The total effect, weight-dependent effect, and weight-independent effect were estimated via a series of multiple regression models adjusted for covariates including baseline weight, baseline blood pressure, hypertension diagnosis at baseline, and study-specific covariates. To provide more perspective, we also report the estimated “% weight-independent” calculated as the percentage of weight-independent effect with respect to the total effect. A post-hoc sensitivity meta-analysis was completed without AWARD-8 as its background medication differs from other studies. The mediation analyses assumed no other unknown or unmeasured confounding factors besides the adjusted covariates. The computational details of the mediation analysis were provided in Additional file 1: Supplemental Methods.
For the secondary objective, a mediation analysis for dose response was first conducted as described above on AWARD-11 for dulaglutide 4.5 mg vs. dulaglutide 1.5 mg. AWARD-11 also evaluated dulaglutide 3.0 mg vs. dulaglutide 1.5 mg, but these comparisons were not examined in the current report as there was less difference in SBP between dulaglutide 3.0 mg and dulaglutide 1.5 mg at week 26. An adjusted indirect comparison (Bucher method) [21] of dulaglutide 4.5 mg vs. placebo was then conducted using the mediation analysis results from AWARD-11 (4.5 mg vs. 1.5 mg) and AWARD-5 (1.5 mg vs. placebo), as these trials had similar background therapy (Additional file 1: Supplemental Methods). This analysis provided an estimation of the weight-dependent, weight-independent, and total effects of higher dose dulaglutide compared to placebo. A sensitivity analysis of the indirect comparison of dulaglutide 4.5 mg vs. placebo was also conducted that included a subset of participants from Ferdinand et al. which had similar background therapy as AWARD-11 and AWARD-5.
Analyses were based on the intention-to-treat populations from each study, excluding patients who discontinued study drug before 6 months. All analyses were exploratory: descriptive and mediation analysis (PROC CAUSALMED) was performed using SAS v9.4, and the meta-analysis (packages meta and metafor) and indirect comparison were performed using R v3.4.4.
## Baseline characteristics and concomitant medications
Baseline characteristics were largely similar across studies (Additional file 1: Table S1). Mean age ranged from 54 to 58 years, percentage of male patients ranged from 44 to $59\%$, and percentage of patients identified as white ranged from 51 to $89\%$. Duration of T2D ranged from 6.8 to 9.2 years and baseline SBP ranged from 127 to 132 mmHg. At baseline, 59–$72\%$ of participants had a hypertension diagnosis. Concomitant antihyperglycemic and antihypertensive medications at baseline are presented (Additional file 1: Table S2). The majority (88–$100\%$) of participants included in the individual trials received metformin as background therapy, except for participants in AWARD-8, who predominantly received a sulfonylurea due to differences in trial design (Additional file 1: Table S2). Blood pressure, hemoglobin A1c, and body weight results for the dulaglutide 1.5 mg group were comparable for studies used in the indirect comparison analysis for the effect of higher dose dulaglutide (Additional file 1: Table S3).
## SBP change from baseline (dulaglutide 1.5 mg vs. placebo)
In the mediation meta-analysis of placebo-controlled trials, the estimated overall total effect of dulaglutide 1.5 mg was − 2.6 mmHg (Table 2; Fig. 2), significantly reducing SBP compared to placebo ($95\%$ CI: − 3.8, − 1.5; $p \leq 0.001$); $36\%$ of dulaglutide’s total effect on BP change was weight-dependent, with an estimated treatment group difference of − 0.9 mmHg ($95\%$ CI: − 1.4, − 0.5; $p \leq 0.001$). Consequently, the weight-independent effect of dulaglutide 1.5 mg comprised $64\%$ of the total effect, with an estimated treatment effect of − 1.5 mmHg ($95\%$ CI: − 2.6, − 0.3; $$p \leq 0.013$$). No significant heterogeneity was detected in the meta-analysis. Results from the post-hoc sensitivity meta-analysis excluding AWARD-8 were consistent with the primary meta-analysis results (Additional file 1; Table S4).Table 2Summary of findings for SBP and PP on dulaglutide treatment effect in participants with T2DWeight-dependent effect (mmHg)Weight-independent effect (mmHg)Total effect (mmHg)% Weight-independent (%)Mediation meta-analysis of placebo-controlled trials (AWARD-1, 5, 8, 10, 11, and Ferdinand et al.) Dula 1.5 mg vs. PBOSBP− 0.9 (− 1.4, − 0.5)− 1.5 (− 2.6, − 0.3)− 2.6 (− 3.8, − 1.5)64PP− 0.4 (− 0.6, − 0.1)− 2.0 (− 3.0, − 1.0)− 2.5 (− 3.5, − 1.5)86Mediation analysis for dose response (AWARD-11) Dula 4.5 mg vs. Dula 1.5 mgSBP− 0.7 (− 1.1, − 0.4)− 0.3 (− 1.6, 1.0)− 1.0 (− 2.2, 0.3)29PP− 0.4 (− 0.6, − 0.2)− 0.9 (− 1.8, 0.2)− 1.2 (− 2.2, − 0.1)70Indirect comparison of Dula 4.5 mg vs. PBO (AWARD-5 and 11) Dula 4.5 mg vs. PBOSBP− 2.0 (− 2.9, − 1.1)− 1.5 (− 4.2, 1.2)− 3.5 (− 6.2, − 0.8)43PP− 1.1 (− 1.7, − 0.5)− 1.8 (− 4.0, 0.4)− 2.9 (− 5.1, − 0.7)62Dula dulaglutide, PBO placebo, PP pulse pressure, SBP systolic blood pressure% Weight-Independent was calculated as (1 − Weight-dependent Effect / Total Effect) × $100\%$ and was reported only when total effect p-value < 0.05 or when the weight-dependent and weight-independent effects had the same signFig. 2Mediation of dulaglutide 1.5 mg effects on SBP: meta-analysis of placebo-controlled trials for weight dependent vs. weight independent effects. Percent attributed as weight-independent was calculated as (1 − Weight-dependent Effect / Total Effect) × $100\%$ and was reported only when total effect p-value < 0.05 or when the weight-dependent and weight-independent effects had the same sign. CI confidence interval, NA not applicable, REM random-effect model, SBP systolic blood pressure
## Pulse pressure change from baseline (dulaglutide 1.5 mg vs. placebo)
In the mediation meta-analysis of placebo-controlled trials, the estimated overall total effect of dulaglutide 1.5 mg was − 2.5 mmHg, demonstrating significantly reduced pulse pressure compared to placebo ($95\%$ CI: − 3.5, − 1.5; $p \leq 0.001$) (Table 2; Fig. 3); $14\%$ of dulaglutide’s total effect on pulse pressure change was weight-dependent, with an estimated overall treatment group difference of − 0.4 mmHg ($95\%$ CI: − 0.6, − 0.1; $$p \leq 0.005$$). The weight-independent effect of dulaglutide 1.5 mg comprised $86\%$ of the total effect, with an overall estimated treatment effect of − 2.0 mmHg ($95\%$ CI: − 3.0, − 1.0; $p \leq 0.001$). The post-hoc sensitivity meta-analysis for pulse pressure showed consistent results (Additional file 1: Table S5).Fig. 3Mediation of dulaglutide 1.5 mg effects on pulse pressure: meta-analysis of placebo-controlled trials for weight dependent vs. weight independent effects. Percent attributed as weight-independent was calculated as (1 − Weight-dependent Effect / Total Effect) × $100\%$ and was reported only when total effect p-value < 0.05 or when the weight-dependent and weight-independent effects had the same sign. CI confidence interval, REM random-effect model
## DBP change from baseline (dulaglutide 1.5 mg vs. placebo)
In the mediation meta-analysis of placebo-controlled trials, limited effect was seen for DBP. This is likely due to that DBP changes from baseline were generally small in the individual studies (effect estimates ranged from − 0.9 to 1.1 mmHg; all $p \leq 0.05$ (Additional file 1; Table S6). The overall total effect was minimal and not significant (− 0.2 mmHg; $95\%$ CI: − 1.0, 0.5; $$p \leq 0.56$$); there was a small but significant weight-dependent effect that decreased DBP (− 0.6 mmHg; $95\%$ CI: − 0.8, − 0.3; $p \leq 0.001$), while the weight-independent effect for increased DBP was not significant (+ 0.5 mmHg; $95\%$ CI: − 0.3, 1.2; $$p \leq 0.24$$). The post-hoc sensitivity meta-analysis for DBP showed consistent results (Additional file 1; Table S6).
## The dose effect of dulaglutide 4.5 mg vs. dulaglutide 1.5 mg and placebo
The mediation analysis for dose response for estimation of difference between dulaglutide 4.5 mg vs. 1.5 mg was conducted on AWARD-11 at week 26. The dose response for SBP estimated total effect for 4.5 mg vs. 1.5 mg was − 1.0 mmHg ($95\%$ CI: − 2.2, 0.3; $$p \leq 0.15$$), weight-dependent effect was − 0.7 mmHg ($95\%$ CI: − 1.1, − 0.4; $p \leq 0.001$), and the weight-independent effect was − 0.3 mmHg ($95\%$ CI: − 1.6, 1.0; $$p \leq 0.67$$) (Table 2). For dulaglutide 4.5 mg, $71\%$ of the additional effect beyond dulaglutide 1.5 mg on SBP reduction was weight dependent. When comparing the dose response of dulaglutide 4.5 mg to 1.5 mg at week 26 for pulse pressure, the total effect was − 1.2 mmHg ($95\%$ CI: − 2.2, − 0.1; $$p \leq 0.02$$), the weight-dependent effect was − 0.4 mmHg ($95\%$ CI: − 0.6, − 0.2; $$p \leq 0.001$$), and weight-independent effect was − 0.9 mmHg ($95\%$ CI: − 1.8, 0.2; $$p \leq 0.11$$). For dulaglutide 4.5 mg vs. 1.5 mg, $30\%$ of the effect on pulse pressure reduction was weight-dependent, and $70\%$ of the effect was weight-independent. For dose response between dulaglutide 4.5 mg and 1.5 mg, DBP did not demonstrate a significant total effect (0.3; $95\%$ CI: − 0.6, 1.2; $$p \leq 0.56$$) nor weight-independent effect (0.6; $95\%$ CI: − 0.2, 1.5, $$p \leq 0.18$$) while there was a small weight-dependent effect (− 0.3; $95\%$ CI: − 0.6, − 0.2) $p \leq 0.001$) (Additional file 1; Table S7).
The indirect comparison analysis of dulaglutide 4.5 mg vs. placebo using AWARD-11 and AWARD-5 estimated the total effect for SBP change to be − 3.5 mmHg ($95\%$ CI: − 6.2, − 0.8; $$p \leq 0.01$$; Table 2), the weight-dependent effect to be − 2.0 mmHg ($95\%$ CI: − 2.9, − 1.1; $p \leq 0.001$), and the weight-independent effect to be − 1.5 mmHg ($95\%$ CI: − 4.2, 1.2; $$p \leq 0.28$$). Of the total effect of dulaglutide 4.5 mg vs. placebo, $57\%$ was weight dependent, and $43\%$ was weight independent. Results of the sensitivity analysis for SBP from the indirect comparison for dulaglutide 4.5 mg vs. placebo, which included the additional data from Ferdinand et al., were consistent with the primary analysis (Additional file 1; Table S8).
Additionally, the indirect comparison of the effect of dulaglutide 4.5 mg vs. placebo estimated the total effect for pulse pressure to be − 2.9 mmHg ($95\%$ CI: − 5.1, − 0.7; $$p \leq 0.01$$; Tables 2 and S8). The weight-dependent effect was − 1.1 mmHg ($95\%$ CI: − 1.7, − 0.5; $p \leq 0.001$), and the weight-independent effect − 1.8 mmHg ($95\%$ CI: − 4.0, 0.4; $$p \leq 0.11$$). A total of $38\%$ of the effect was weight-dependent, and $62\%$ of the effect was weight independent. In the sensitivity analysis for the indirect comparison of dulaglutide 4.5 mg vs. placebo there was modest variation from the primary analysis as the total effect was an additional 0.5 mmHg reduction in pulse pressure which was due to a larger weight-independent effect. While the absolute weight-dependent effect was increased moderately compared to the primary analysis, it explained only $29\%$ of the reduction in pulse pressure, and $71\%$ of the reduction was weight-independent (Additional file 1; Table S9).
The indirect comparison for the effect of dulaglutide 4.5 mg vs. placebo for DBP was in-line with the results for dulaglutide 1.5 mg vs. placebo; the estimated total effect was − 0.6 mmHg ($95\%$ CI: − 2.2, 1.0; $$p \leq 0.46$$), the weight-dependent effect was -0.9 mmHg ($95\%$ CI: − 1.3, − 0.5; $p \leq 0.001$), while the weight-independent effect was 0.3 (− 1.3, 1.9), $$p \leq 0.71$$) (Additional file 1; Table S7). The sensitivity analysis for the indirect comparison demonstrated similar results for DBP (Additional file 1; Table S7).
## Discussion
Our primary findings demonstrate that dulaglutide treatment has both weight-dependent and weight-independent effects on reduction in SBP and pulse pressure in participants with T2D. Both SBP and pulse pressure decreased consistently with dulaglutide treatment, and the majority of the effect on blood pressure with dulaglutide 1.5 mg treatment was weight-independent, as weight reduction mediated only $36\%$ and $14\%$ of the effect on SBP and pulse pressure, respectively. Additional reductions of SBP and pulse pressure were observed with dulaglutide 4.5 mg. The greater effect was mostly weight-dependent and was likely driven by the greater weight reduction known to occur with higher dose dulaglutide [22].
## Possible mechanisms of dulaglutide’s effects on CV system
Dulaglutide is a highly efficacious treatment for hyperglycemia in T2D that was developed based on incretin physiology, as GLP-1 is released after nutrient ingestion and stimulates glucose-dependent secretion of insulin. Pleotropic effects of GLP-1 RAs include suppression of glucagon, delayed gastric emptying, and improved satiety [23]. Like other GLP-1 RAs, chronic treatment with dulaglutide is associated with reduced body weight [24], which has been proposed as one of the mechanisms underlying the concomitant decrease in BP. While weight reduction is a favorable characteristic of a T2D treatment, our analysis showed that most of the BP reduction that occurred with dulaglutide 1.5 mg was weight-independent, highlighting additional effects of dulaglutide.
In the REWIND CV outcomes trial evaluating dulaglutide 1.5 mg vs. placebo, 3-point major adverse cardiac events (MACE-3) were reduced by $12\%$ in the dulaglutide arm over a median of 5.4 years, even though $68\%$ of participants at baseline only had CV risk factors and had not experienced a CV event prior to starting the trial [25]. Over the course of the REWIND trial, the estimated treatment difference for SBP was a 1.7-mmHg reduction for dulaglutide compared to placebo [25]. There was also a 1.9-beat-per-minute increase in heart rate with dulaglutide treatment that persisted through the REWIND trial, which was similar to the results of other GLP-1 RA cardiovascular outcomes trials [26]. While the mechanism of the slight increase in heart rate with GLP-1 RA treatment is uncertain, clinical evidence supports that GLP-1 receptor activation does not affect cardiac sympathetic activity [27], and the MACE reduction seen in REWIND and other CVOTs is reassuring. A mediation analysis of the REWIND trial data did not demonstrate a significant relationship between MACE reduction and reduced SBP [28]. A similar analysis of data from the LEADER trial evaluating liraglutide vs. placebo also did not find SBP reduction to be a mediator of MACE reduction [29], however, pulse pressure was not included in either of these mediation analyses.
Increases in pulse pressure are caused by arterial stiffening [30]. Diabetes and obesity are established risk factors for elevated pulse pressure and accelerate the progression of arterial stiffening that occurs with age [30]. Dulaglutide and other GLP-1 RAs improve arterial stiffness in people with T2D [31, 32]. Numerous mechanisms contribute to the arterial stiffness that occurs in patients with obesity, insulin resistance and type 2 diabetes, including endothelial cell and vascular smooth muscle cell dysfunction [33]. Liu et al. reviewed the GLP-1 receptor activation-induced cell signaling that directly improves endothelial cell and vascular smooth muscle dysfunction [34]. Importantly, GLP-1 receptor activation increases endothelial-dependent relaxation and reduces endothelial-induced contractions of vascular smooth muscle to reduce blood pressure [34]. Various other mechanisms have been identified in preclinical models including that GLP-1 receptor agonism directly induces secretion of atrial natriuretic peptide [19]. However, findings are not consistent in clinical studies [35]. More recent research demonstrates that dulaglutide treatment in people with early T2D occurs with increased number and function of circulating endothelial progenitor cells (EPCs), which are important for maintaining endothelial structure by repairing vascular injury. This increase in EPCs predicted a decrease in arterial stiffness supporting the potential clinical relevance on this finding [15]. Other mechanisms proposed for GLP-1 RAs on cardiovascular function and BP reduction include suppression of oxidative stress, anti-inflammatory activity, renal anti-fibrotic effects, and central nervous system control [36–38]. The area postrema, a circumventricular organ located in the dorsal medulla of the brain which densely expresses GLP-1 receptors, may have a role mediating the effects of GLP-1 RAs [39]; preclinical models of hypertension support an antihypertensive effect of GLP-1 RAs by activating these neurons and suppressing sympathetic nerve activity [40].
## Integration of our findings with previous studies
In the current study, data from similar time points (24–36 weeks) were used to limit potential confounding by the length of GLP-1 RA treatment on the mediation analysis results. Another meta-analysis found that weight reduction partially mediated SBP reduction with GLP-1 RA treatment [16]; however, the effect may have been confounded by the broad range of timing used in the study (8 weeks to > 5 years). A third meta-analysis with an endpoint time range of 12–56 weeks did not find an effect of weight reduction on SBP [19]. After 4 weeks, treatment with once weekly dulaglutide contributes to robust decreases in SBP [20, 22] and pulse pressure [20], but partial attrition occurs by week 26. Due to the partial attrition of the BP effect from 4 to 26 weeks and that at 4 weeks the BP reductions are before significant weight reduction, at least a portion of this early decrease in BP is likely due to a different mechanism than the reductions seen at the later time points investigated in the current study.
## Possible explanations for different results in AWARD-8
AWARD-8 was the only individual study included in the primary analysis that did not show a significant effect of treatment for either SBP or pulse pressure; participants in AWARD-8 also had the lowest mean weight reduction of the trials included in this analysis. Differences in the trial design that could account for the different response include the concomitant medications taken by the participants; while participants in AWARD-8 were taking a sulfonylurea, with very few participants receiving concomitant metformin, participants in the other trials were predominantly on metformin with or without another oral anti-diabetes medication. In a retrospective cohort comparing monotherapy after one year of treatment with sulfonylurea, SBP was 1.3 mmHg higher vs. metformin, and this difference was thought to be due to the varying treatment effects on weight [41]. The literature regarding the effect of metformin on BP suggests there may be a small BP-lowering effect that occurs specifically in individuals with underlying hypertension [42]. Even with potentially different effects on BP between sulfonylurea and metformin, participants were required to be on stable background therapy before enrolling in the included trials. AWARD-8 also had the smallest sample size for a single study which limits any generalizability to the findings from this individual study.
## Active weight loss vs. weight reduction maintenance
While there is a robust decrease in BP during active weight loss, the benefit is attenuated during the weight reduction maintenance phase. This attenuation is hypothesized to be due to a lack of sustained neurohormonal responses to the active weight loss during weight reduction maintenance [17]. In the Look AHEAD trial, participants with T2D who undertook an intensive lifestyle-based weight-reduction intervention achieved an $8.6\%$ weight reduction by the end of one year compared to $0.7\%$ in the standard diabetes education arm, concurrent with a 7-mmHg decrease in SBP [43]. After approximately 10 years of follow-up, most of the weight reduction ($6.0\%$) was maintained in the intensive arm but only 2 mmHg of the SBP reduction persisted [44]. Pulse pressure also decreases early on after lifestyle-induced weight reduction phase in people with obesity [45]. While long-term follow up from bariatric surgery also demonstrates a reversal of the early reduction in SBP and DBP, there is long-term slowing of the age-associated increase in pulse pressure [46]. In the REWIND trial, after a median of 5.4 years of follow up the dulaglutide treatment occurred with associated reductions in SBP (− 1.7 mmHg) and weight (− 1.5 kg) [25], but it is not known if weight reduction contributes to the long-term SBP effect observed in the REWIND trial.
## Weight-independent and weight-dependent effects of dulaglutide on blood pressure
The Ferdinand et al. 2014 trial included in this meta-analysis reported a differential dose effect of dulaglutide 0.75 mg vs. 1.5 mg on SBP and pulse pressure [19]. In the current study, we also found a dose effect for SBP and pulse pressure reduction when comparing dulaglutide 4.5 mg and 1.5 mg at 26 weeks in AWARD-11. The higher dose reduced SBP by an additional − 1.0 mmHg compared to lower dose of which $71\%$ was dependent on weight reduction and reduced pulse pressure by an additional 1.2 mmHg of which $30\%$ was dependent on weight reduction. Interestingly, the weight-independent effect of dulaglutide 1.5 mg and dulaglutide 4.5 mg were similar (− 1.5 mmHg for SBP and − 2.0 mmHg and − 1.8 mmHg for pulse pressure, respectively), suggesting that the weight-independent effect of dulaglutide may be maximized at the 1.5 mg dose. Thus, that may be the limit of the weight-independent effects of dulaglutide.
## Study limitations
There are specific considerations which limit the interpretation of the current study. Many of the clinical trials included were multinational impacting the racial and ethnic representation; therefore, it is uncertain if the current findings apply to specific populations with high rates of T2D and high blood pressure such as African American or Black adults. Participants’ blood pressure control at baseline was potentially better than many clinical practice populations and may also influence generalizability of the findings. Additionally, because the trials included in this analysis were primarily designed to assess hyperglycemia, strict definitions for hypertension diagnosis at baseline were not used, partially limiting interpretation. Our findings are specific to approximately 6 months of dulaglutide treatment; additional research is necessary to understand the interaction of dulaglutide and weight reduction on SBP and pulse pressure in longer-term studies. Results from the indirect comparison should be interpreted with caution as despite the similarities in baseline demographics and background treatment there might be unmeasured confounding factors that could have influenced the results.
## Conclusions
In conclusion, as elevated SBP and pulse pressure are risk factors for cardiovascular and microvascular complications in patients with T2D, treatment options like dulaglutide and other GLP-1 RAs that reduce these are favorable. The current findings indicate that a portion of the SBP and pulse pressure reduction observed with dulaglutide treatment is not weight mediated, and further research is needed to understand the mechanisms of the additional benefit. Understanding the mechanisms by which dulaglutide improves SBP and pulse pressure, whether dependent or independent of weight reduction, could provide insight into developing future treatment regimens for elevated blood pressure.
## Supplementary Information
Additional file 1. Supplemental Methods and Tables S1-S9.
## References
1. Muntner P, Whelton PK, Woodward M, Carey RM. **A comparison of the 2017 American College of Cardiology/American Heart Association blood pressure guideline and the 2017 American Diabetes Association diabetes and hypertension position statement for U.S. adults with diabetes**. *Diabetes Care* (2018) **41** 2322-2329. DOI: 10.2337/dc18-1307
2. An J, Nichols GA, Qian L, Munis MA, Harrison TN, Li Z. **Prevalence and incidence of microvascular and macrovascular complications over 15 years among patients with incident type 2 diabetes**. *BMJ Open Diabetes Res Care* (2021) **9** e001847. DOI: 10.1136/bmjdrc-2020-001847
3. Tseng LN, Tseng YH, Jiang YD, Chang CH, Chung CH, Lin BJ. **Prevalence of hypertension and dyslipidemia and their associations with micro- and macrovascular diseases in patients with diabetes in Taiwan: an analysis of nationwide data for 2000–2009**. *J Formos Med Assoc* (2012) **111** 625-636. DOI: 10.1016/j.jfma.2012.09.010
4. Flack JM, Adekola B. **Blood pressure and the new ACC/AHA hypertension guidelines**. *Trends Cardiovasc Med* (2020) **30** 160-164. DOI: 10.1016/j.tcm.2019.05.003
5. Kannel WB. **Framingham study insights into hypertensive risk of cardiovascular disease**. *Hypertens Res* (1995) **18** 181-196. DOI: 10.1291/hypres.18.181
6. Whelton PK, He J, Appel LJ, Cutler JA, Havas S, Kotchen TA. **Primary prevention of hypertension: clinical and public health advisory from The National High Blood Pressure Education Program**. *JAMA* (2002) **288** 1882-1888. DOI: 10.1001/jama.288.15.1882
7. **Pharmacological blood pressure lowering for primary and secondary prevention of cardiovascular disease across different levels of blood pressure: an individual participant-level data meta-analysis**. *Lancet* (2021) **397** 1625-1636. DOI: 10.1016/S0140-6736(21)00590-0
8. Reboldi G, Gentile G, Angeli F, Ambrosio G, Mancia G, Verdecchia P. **Effects of intensive blood pressure reduction on myocardial infarction and stroke in diabetes: a meta-analysis in 73,913 patients**. *J Hypertens* (2011) **29** 1253-1269. DOI: 10.1097/HJH.0b013e3283469976
9. Adler AI, Stratton IM, Neil HA, Yudkin JS, Matthews DR, Cull CA. **Association of systolic blood pressure with macrovascular and microvascular complications of type 2 diabetes (UKPDS 36): prospective observational study**. *BMJ* (2000) **321** 412-419. DOI: 10.1136/bmj.321.7258.412
10. Whelton PK, Carey RM, Aronow WS, Casey DE, Collins KJ, Dennison Himmelfarb C. **2017 ACC/AHA/AAPA/ABC/ACPM/AGS/APhA/ASH/ASPC/NMA/PCNA guideline for the prevention, detection, evaluation, and management of high blood pressure in adults: a report of the American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines**. *Hypertension* (2018) **71** e13-e115. PMID: 29133356
11. Ali MK, Bullard KM, Saaddine JB, Cowie CC, Imperatore G, Gregg EW. **Achievement of goals in U.S. diabetes care, 1999–2010**. *N Engl J Med* (2013) **368** 1613-1624. DOI: 10.1056/NEJMsa1213829
12. Grenier J, Goodman SG, Leiter LA, Langer A, Teoh H, Bhatt DL. **Blood pressure management in adults with type 2 diabetes: insights from the diabetes mellitus status in Canada (DM-SCAN) survey**. *Can J Diabetes* (2018) **42** 130-137. DOI: 10.1016/j.jcjd.2017.01.005
13. Berra C, Manfrini R, Regazzoli D, Radaelli MG, Disoteo O, Sommese C. **Blood pressure control in type 2 diabetes mellitus with arterial hypertension. The important ancillary role of SGLT2-inhibitors and GLP1-receptor agonists**. *Pharmacol Res* (2020) **160** 105052. DOI: 10.1016/j.phrs.2020.105052
14. Palmer SC, Tendal B, Mustafa RA, Vandvik PO, Li S, Hao Q. **Sodium-glucose cotransporter protein-2 (SGLT-2) inhibitors and glucagon-like peptide-1 (GLP-1) receptor agonists for type 2 diabetes: systematic review and network meta-analysis of randomised controlled trials**. *BMJ* (2021) **372** m4573. DOI: 10.1136/bmj.m4573
15. Xie D, Li Y, Xu M, Zhao X, Chen M. **Effects of dulaglutide on endothelial progenitor cells and arterial elasticity in patients with type 2 diabetes mellitus**. *Cardiovasc Diabetol* (2022) **21** 200. DOI: 10.1186/s12933-022-01634-1
16. Ryan D, Acosta A. **GLP-1 receptor agonists: nonglycemic clinical effects in weight loss and beyond**. *Obesity (Silver Spring)* (2015) **23** 1119-1129. DOI: 10.1002/oby.21107
17. Hall ME, Cohen JB, Ard JD, Egan BM, Hall JE, Lavie CJ. **Weight-loss strategies for prevention and treatment of hypertension: a scientific statement from the American Heart Association**. *Hypertension* (2021) **78** e38-e50. DOI: 10.1161/HYP.0000000000000202
18. Hu M, Cai X, Yang W, Zhang S, Nie L, Ji L. **Effect of hemoglobin a1c reduction or weight reduction on blood pressure in glucagon-like peptide-1 receptor agonist and sodium-glucose cotransporter-2 inhibitor treatment in type 2 diabetes mellitus: a meta-analysis**. *J Am Heart Assoc* (2020) **9** e015323. DOI: 10.1161/JAHA.119.015323
19. Katout M, Zhu H, Rutsky J, Shah P, Brook RD, Zhong J. **Effect of GLP-1 mimetics on blood pressure and relationship to weight loss and glycemia lowering: results of a systematic meta-analysis and meta-regression**. *Am J Hypertens* (2013) **27** 130-139. DOI: 10.1093/ajh/hpt196
20. Ferdinand KC, White WB, Calhoun DA, Lonn EM, Sager PT, Brunelle R. **Effects of the once-weekly glucagon-like peptide-1 receptor agonist dulaglutide on ambulatory blood pressure and heart rate in patients with type 2 diabetes mellitus**. *Hypertension* (2014) **64** 731-737. DOI: 10.1161/HYPERTENSIONAHA.114.03062
21. Bucher HC, Guyatt GH, Griffith LE, Walter SD. **The results of direct and indirect treatment comparisons in meta-analysis of randomized controlled trials**. *J Clin Epidemiol* (1997) **50** 683-691. DOI: 10.1016/S0895-4356(97)00049-8
22. Frias JP, Bonora E, Nevarez Ruiz L, Li YG, Yu Z, Milicevic Z. **Efficacy and safety of dulaglutide 3.0 mg and 4.5 mg versus dulaglutide 1.5 mg in metformin-treated patients with type 2 diabetes in a randomized controlled trial (AWARD-11)**. *Diabetes Care* (2021) **44** 765-773. DOI: 10.2337/dc20-1473
23. Jendle J, Grunberger G, Blevins T, Giorgino F, Hietpas RT, Botros FT. **Efficacy and safety of dulaglutide in the treatment of type 2 diabetes: a comprehensive review of the dulaglutide clinical data focusing on the AWARD phase 3 clinical trial program**. *Diabetes Metab Res Rev* (2016) **32** 776-790. DOI: 10.1002/dmrr.2810
24. Bonora E, Frias JP, Tinahones FJ, Van J, Malik RE, Yu Z. **Effect of dulaglutide 3.0 and 4.5 mg on weight in patients with type 2 diabetes: exploratory analyses of AWARD-11**. *Diabetes Obes Metab* (2021) **23** 2242-2250. DOI: 10.1111/dom.14465
25. 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
26. Das SR, Everett BM, Birtcher KK, Brown JM, Januzzi JL, Kalyani RR. **2020 Expert consensus decision pathway on novel therapies for cardiovascular risk reduction in patients with type 2 diabetes: a report of the American College of Cardiology Solution Set Oversight Committee**. *J Am Coll Cardiol* (2020) **76** 1117-1145. DOI: 10.1016/j.jacc.2020.05.037
27. Bharucha AE, Charkoudian N, Andrews CN, Camilleri M, Sletten D, Zinsmeister AR. **Effects of glucagon-like peptide-1, yohimbine, and nitrergic modulation on sympathetic and parasympathetic activity in humans**. *Am J Physiol Regul Integr Comp Physiol* (2008) **295** R874-R880. DOI: 10.1152/ajpregu.00153.2008
28. Konig M, Riddle MC, Colhoun HM, Branch KR, Atisso CM, Lakshmanan MC. **Exploring potential mediators of the cardiovascular benefit of dulaglutide in type 2 diabetes patients in REWIND**. *Cardiovasc Diabetol* (2021) **20** 194. DOI: 10.1186/s12933-021-01386-4
29. Buse JB, Bain SC, Mann JFE, Nauck MA, Nissen SE, Pocock S. **Cardiovascular risk reduction with liraglutide: an exploratory mediation analysis of the LEADER trial**. *Diabetes Care* (2020) **43** 1546-1552. DOI: 10.2337/dc19-2251
30. Schillaci G, Pucci G. **Aging and pulse pressure widening: the inseparable duo?**. *J Hypertens* (2015) **33** 2389-2391. DOI: 10.1097/HJH.0000000000000788
31. Lambadiari V, Pavlidis G, Kousathana F, Varoudi M, Vlastos D, Maratou E. **Effects of 6-month treatment with the glucagon like peptide-1 analogue liraglutide on arterial stiffness, left ventricular myocardial deformation and oxidative stress in subjects with newly diagnosed type 2 diabetes**. *Cardiovasc Diabetol* (2018) **17** 8. DOI: 10.1186/s12933-017-0646-z
32. Tuttolomondo A, Cirrincione A, Casuccio A, Del Cuore A, Daidone M, Di Chiara T. **Efficacy of dulaglutide on vascular health indexes in subjects with type 2 diabetes: a randomized trial**. *Cardiovasc Diabetol* (2021) **20** 1. DOI: 10.1186/s12933-020-01183-5
33. Jia G, Aroor AR, DeMarco VG, Martinez-Lemus LA, Meininger GA, Sowers JR. **Vascular stiffness in insulin resistance and obesity**. *Front Physiol* (2015) **6** 231. DOI: 10.3389/fphys.2015.00231
34. Liu L, Liu J, Huang Y. **Protective effects of glucagon-like peptide 1 on endothelial function in hypertension**. *J Cardiovasc Pharmacol* (2015) **65** 399-405. DOI: 10.1097/FJC.0000000000000176
35. Lovshin JA, Barnie A, DeAlmeida A, Logan A, Zinman B, Drucker DJ. **Liraglutide promotes natriuresis but does not increase circulating levels of atrial natriuretic peptide in hypertensive subjects with type 2 diabetes**. *Diabetes Care* (2015) **38** 132-139. DOI: 10.2337/dc14-1958
36. Yaribeygi H, Farrokhi FR, Abdalla MA, Sathyapalan T, Banach M, Jamialahmadi T. **The effects of glucagon-like peptide-1 receptor agonists and dipeptydilpeptidase-4 inhibitors on blood pressure and cardiovascular complications in diabetes**. *J Diabetes Res* (2021) **2021** 6518221. DOI: 10.1155/2021/6518221
37. Helmstadter J, Frenis K, Filippou K, Grill A, Dib M, Kalinovic S. **Endothelial GLP-1 (glucagon-like peptide-1) receptor mediates cardiovascular protection by liraglutide in mice with experimental arterial hypertension**. *Arterioscler Thromb Vasc Biol* (2020) **40** 145-158. DOI: 10.1161/atv.0000615456.97862.30
38. Mohamad HE, Abdelhady MA, Abdel Aal SM, Elrashidy RA. **Dulaglutide mitigates high dietary fructose-induced renal fibrosis in rats through suppressing epithelial-mesenchymal transition mediated by GSK-3β/TGF-β1/Smad3 signaling pathways**. *Life Sci* (2022) **309** 120999. DOI: 10.1016/j.lfs.2022.120999
39. Yamamoto H, Kishi T, Lee CE, Choi BJ, Fang H, Hollenberg AN. **Glucagon-like peptide-1-responsive catecholamine neurons in the area postrema link peripheral glucagon-like peptide-1 with central autonomic control sites**. *J Neurosci* (2003) **23** 2939-2946. DOI: 10.1523/JNEUROSCI.23-07-02939.2003
40. Katsurada K, Nakata M, Saito T, Zhang B, Maejima Y, Nandi SS. **Central glucagon-like peptide-1 receptor signaling via brainstem catecholamine neurons counteracts hypertension in spontaneously hypertensive rats**. *Sci Rep* (2019) **9** 12986. DOI: 10.1038/s41598-019-49364-x
41. Roumie CL, Liu X, Choma NN, Greevy RA, Hung AM, Grijalva CG. **Initiation of sulfonylureas versus metformin is associated with higher blood pressure at one year**. *Pharmacoepidemiol Drug Saf* (2012) **21** 515-523. DOI: 10.1002/pds.3249
42. Anabtawi A, Miles JM. **Metformin: nonglycemic effects and potential novel indications**. *Endocr Pract* (2016) **22** 999-1007. DOI: 10.4158/EP151145.RA
43. Look ARG, Pi-Sunyer X, Blackburn G, Brancati FL, Bray GA, Bright R. **Reduction in weight and cardiovascular disease risk factors in individuals with type 2 diabetes: one-year results of the look AHEAD trial**. *Diabetes Care* (2007) **30** 1374-1383. DOI: 10.2337/dc07-0048
44. Wing RR, Bolin P, Brancati FL, Bray GA, Clark JM, Coday M. **Cardiovascular effects of intensive lifestyle intervention in type 2 diabetes**. *N Engl J Med* (2013) **369** 145-154. DOI: 10.1056/NEJMoa1212914
45. Kwagyan J, Tabe CE, Xu S, Maqbool AR, Gordeuk VR, Randall OS. **The impact of body mass index on pulse pressure in obesity**. *J Hypertens* (2005) **23** 619-624. DOI: 10.1097/01.hjh.0000160220.71350.5f
46. Sjöström CD, Peltonen M, Sjöström L. **Blood pressure and pulse pressure during long-term weight loss in the obese: the Swedish Obese Subjects (SOS) Intervention Study**. *Obes Res* (2001) **9** 188-195. DOI: 10.1038/oby.2001.20
|
---
title: 'Association between very advanced maternal age women with gestational diabetes
mellitus and the risks of adverse infant outcomes: a cohort study from the NVSS
2014–2019'
authors:
- Lin Lu
- Lidan He
- Jifen Hu
- Jianhua Li
journal: BMC Pregnancy and Childbirth
year: 2023
pmcid: PMC9999489
doi: 10.1186/s12884-023-05449-0
license: CC BY 4.0
---
# Association between very advanced maternal age women with gestational diabetes mellitus and the risks of adverse infant outcomes: a cohort study from the NVSS 2014–2019
## Abstract
### Background
To evaluate the association between gestational diabetes mellitus (GDM) and infant outcomes in women of very advanced maternal age (vAMA) (≥45 years).
### Methods
This cohort study utilized data from the National Vital Statistics System (NVSS) database (2014–2019) in the United States. Preterm birth was the primary outcome, which was subdivided into extremely preterm, very preterm, and moderate or late preterm. The secondary outcomes were neonatal intensive care unit (NICU) admission, low birthweight and small for gestational age. Univariate and multivariate logistic regression analyses were used to explore the association between GDM and infant outcomes among vAMA women. Subgroup analyses were performed based on race and use of infertility treatment. Odds ratios (ORs) and $95\%$ confidence intervals (CIs) were estimated.
### Results
A total of 52,544 vAMA pregnant women were included. All analysis made comparisons between women with vAMA and GDM and women with vAMA and no GDM. Women with GDM had a significantly higher risk of preterm birth than those without GDM (OR = 1.26, $95\%$CI = 1.18–1.36, $P \leq 0.001$). Compared with women without GDM, those with GDM had a significantly increased risk of moderate or late preterm birth (OR = 1.27, $95\%$CI = 1.18–1.37, $P \leq 0.001$); no significant association of GDM with extremely preterm birth and very preterm birth was observed. Women with GDM had a significantly greater risk of NICU admission than those without (OR = 1.33, $95\%$CI = 1.23–1.43, $P \leq 0.001$). GDM was associated with a significantly lower risk of low birthweight (OR = 0.91, $95\%$CI = 0.84–0.98, $$P \leq 0.010$$), and no significant association was found between GDM and small for gestational age (OR = 0.95, $95\%$CI = 0.87–1.03, $$P \leq 0.200$$) in vAMA women.
### Conclusion
vAMA women with GDM had an increased risk of preterm birth, especially moderate or late preterm birth. NICU admission and low birthweight were also associated with GDM among vAMA women.
### Supplementary Information
The online version contains supplementary material available at 10.1186/s12884-023-05449-0.
## Background
Due to socio-economic development and advances in assisted reproductive technology (ART), the trend of more frequent births among older women, particularly those of very advanced maternal age (≥45 years) (vAMA), is likely to continue [1]. A considerable amount of studies have reported that rising maternal age is one of the key drivers for the increased prevalence of gestational diabetes mellitus (GDM) [2–4]. GDM is traditionally defined as carbohydrate intolerance leading to hyperglycemia of varying severity with onset or first detection during pregnancy [5]. AMA has been identified as a risk factor for GDM [6]. Women of vAMA have a higher incidence of GDM than those under 45 years [7]. Besides, the risks of adverse perinatal outcomes for women aged ≥40 years increase with age [8, 9], and a prior review shows that vAMA women have elevated risks of adverse perinatal outcomes [10]. Extensive research has shown that GDM is associated with increased risks of adverse perinatal outcomes [11–13], such as preterm birth, pre-eclampsia/eclampsia, growth abnormalities, and respiratory distress. Thus, vAMA women who plan to become pregnant may need to pay attention to the risks of GDM and adverse outcomes. Healthcare givers can counsel women of vAMA, especially those with GDM.
Preterm birth (< 37 weeks of gestation) is a common adverse infant outcome, resulting in approximately 1 million infant deaths each year [14]. Even if premature infants survive, they are accompanied by long-lasting diseases that contribute to a global health burden [15, 16]. It has previously been observed that diabetes is a significant risk factor for spontaneous and indicated preterm delivery [11, 17–19]. Diboun et al. [ 20] indicated that GDM may be used as a novel predictor of preterm delivery. In the study of Billionnet et al. [ 11], the risk of preterm birth was illustrated to be higher in the GDM group than in the no diabetes group. Regarding other infant outcomes, Venkatesh et al. [ 21] reported that from 2014 through 2020, the frequency of neonatal intensive care unit (NICU) admission increased, while no significant change was shown in small for gestational age for American women with GDM aged 15–44 years. Although many studies delved into the relationship between GDM and preterm birth, the association of GDM with preterm birth for women of vAMA awaits exploration, which may help identify the population with a high risk of preterm birth and devise prevention and intervention strategies to improve outcomes in vAMA women. The relationships between GDM and NICU admission, low birthweight and small for gestational age are also under-researched.
This study aimed to evaluate the associations of GDM with preterm birth, NICU admission, low birthweight and small for gestational age in vAMA women using the National Vital Statistics System (NVSS) database (2014–2019) in the United States. Given that these associations may vary by race and use of infertility treatment, we further performed subgroup analyses.
## Study design and population
This was a cohort study. All data of pregnant women aged 45 or older who were tested for GDM and did not have pre-gestational diabetes were extracted from the NVSS 2014–2019. The NVSS database provides data on births and deaths as well as maternal characteristics in 50 states, New York City, District of Columbia, and 5 territories (Puerto Rico, Virgin Islands, Guam, American Samoa, and Northern Mariana Islands) of the United States [22]. Participants were excluded according to the following criteria: [1] women with infections presenting or treated during this pregnancy; [2] women with missing information on gestational weeks, neonatal weight, and NICU admission records.
## Variables
Preterm birth was the primary outcome of this study, which was defined as births before 37 completed weeks of gestation. The World Health Organization (WHO) further subdivided preterm birth based on gestational age: extremely preterm (< 28 weeks), very preterm (28 to < 32 weeks), and moderate or late preterm (32 to < 37 weeks) [23]. Secondary outcomes were NICU admission, low birthweight and small for gestational age. Low birthweight was defined as a birthweight < 2500 g, and small for gestational age was defined as a birthweight less than the 10th percentile. The following variables were collected: maternal age at delivery (years), race [Asian, Black (Black or African American), White, other (American Indian or Alaska Native, Native Hawaiian or Other Pacific Islander, and more than one race)], education [less than 12 grade, high school/general educational development (GED), some college or associate degree (AA), bachelor or higher], pre-pregnancy weight (lb), pre-pregnancy body mass index (BMI) (BMI < 18.5 kg/m2, underweight; BMI = 18.5–24.9 kg/m2, normal; BMI = 25.0–29.9 kg/m2, overweight; BMI = 30.0–34.9 kg/m2, obesity), delivery weight (lb), weight gain (lb), smoking before pregnancy (yes or no), smoking status 1st/2nd/3rd trimester (mother-reported smoking in the three trimesters of pregnancy, yes or no), hypertension eclampsia (yes or no), gestational hypertension (yes or no), pre-pregnancy hypertension (yes or no), number of prenatal visits, the Special Supplemental Nutrition Program for Women, Infants, and Children (WIC, receipt of WIC food for the mother during this pregnancy, yes or no), plurality, prior birth now living, prior birth now dead, prior other terminations, total birth order, gestational age (weeks), newborn sex (female or male), birth weight (g), infertility treatment used (yes or no), pregnancy method (natural pregnancy, pregnancy via ART), method of delivery [spontaneous, non-spontaneous (forceps, vacuum, cesarean)], preterm birth [extremely preterm, very preterm, moderate or late preterm; spontaneous, indicated (forceps, vacuum, cesarean)], NICU admission, low birthweight (yes or no), and small for gestational age (yes or no). WIC is a program intended to help low income pregnant women, infants, and children through age 5 receive proper nutrition by providing vouchers for food, nutrition counseling, health care screenings and referrals; it is administered by the U.S. Department of Agriculture (https://ftp.cdc.gov/pub/Health_Statistics/NCHS/Dataset_Documentation/DVS/natality/UserGuide2019-508.pdf). Infertility treatment referred to using fertility enhancing drugs, artificial insemination, intrauterine insemination, or using ART. ART included in vitro fertilization (IVF), gamete intrafallopian transfer (GIFT), and zygote intrafallopian transfer (ZIFT). Information on variables is available at https://www.cdc.gov/nchs/nvss/index.htm.
## Statistical analysis
Continuous data were tested for normality using the Kolmogorov-Smirnov test, and the continuous data of normal distribution were described as mean ± standard deviation (Mean ± SD), and the t-test was used for comparisons between groups. Non-normally distributed continuous variables were shown by median and quartile [M (Q1, Q3)], and the Wilcoxon rank sum test was used for comparisons between groups. Categorical data of groups were compared with the Pearson’s χ2 test, and expressed as cases and the constituent ratio [n (%)]. Statistical power was calculated (power = 1). Missing data were imputed using multiple imputation (Supplementary Table 1). Data before imputation were also used for multivariate analyses to conduct sensitivity analyses.
In order to study the association between GDM and preterm birth among vAMA women, we established three models, and odds ratios (ORs) with $95\%$ confidence intervals (CIs) were estimated. Model 1 was a univariate model. Model 2 was a multivariate model adjusting for maternal age at delivery, race, education, and newborn sex. Then all variables were included in a multivariable model for stepwise regression, and the following variables were adjusted for in Model 3: maternal age at delivery, race, education, newborn sex, pre-pregnancy weight, pre-pregnancy BMI, delivery weight, weight gain, smoking before pregnancy, hypertension eclampsia, gestational hypertension, pre-pregnancy hypertension, number of prenatal visits, plurality, total birth order, prior birth now living, prior other terminations, birth weight, pregnancy method, and method of delivery. Subgroup analyses were then performed based on race and use of infertility treatment to demonstrate if and how the association between GDM and preterm birth in vAMA women varied by race and use of infertility treatment. Further, preterm birth was subdivided into extremely preterm, very preterm, and moderate or late preterm birth. Logistic regression was used to investigate the association between GDM and different stages of preterm birth. Model 1 was a univariate model. Model 2 was a multivariate model correcting for maternal age at delivery, race, education, and newborn sex. Model 3 was a multivariate model correcting for maternal age at delivery, race, education, newborn sex, delivery weight, smoking status 2nd trimester, hypertension eclampsia, gestational hypertension, pre-pregnancy hypertension, number of prenatal visits, WIC, plurality, prior other terminations, total birth order, birth weight, pregnancy method, and method of delivery. As for the associations of GDM with NICU admission, low birthweight and small for gestational age in vAMA women, analytical methods were the same as those for the association between GDM and preterm birth, and subgroup analyses by race and use of infertility treatment were also conducted for these outcomes.
All statistical analyses were two-sided, and $P \leq 0.05$ was considered to be statistically significant. All analyses were completed by SAS 9.4 software (SAS Institute Inc., Cary, NC, USA).
## Participant characteristics
There were a total of 53,484 pregnant women of vAMA in the NVSS database (2014–2019). After excluding women with infections presenting or treated during this pregnancy ($$n = 830$$), and women with missing information on gestational weeks ($$n = 41$$), newborn birth weight ($$n = 39$$), and NICU admission records ($$n = 30$$), 52,544 pregnant women were included in this study. The follow-up time was 37.62 ± 2.97 weeks. Figure 1 presents the flow chart of participant selection. Among them, 7563 women had GDM, while 44,981 pregnant women did not have GDM. The average age of them was 46.39 ± 1.63 years. The proportion of Asians, Blacks, Whites, and other races was $13.48\%$ [7082], $15.42\%$ [8103], $68.61\%$ [36053], and $2.49\%$ [1306], respectively. The median weight gain during pregnancy was 27.00 (19.00, 36.00) lb. Of the included women, $30.00\%$ [15761] used infertility treatment. Women who had a premature birth accounted for $24.00\%$ [12609] of the total, with $1.31\%$ [689] having an extremely preterm birth, $2.79\%$ [1466] having a very preterm birth, and $19.90\%$ [10454] having a moderate or late preterm birth; $18.10\%$ [9513] of newborns were admitted to the NICU, $18.58\%$ [9761] had a low birthweight, and $9.97\%$ [5241] were small for gestational age. Among preterm birth, the proportions of spontaneous preterm birth and indicated preterm birth were 23.08 and $76.92\%$, respectively. More details for participant characteristics are shown in Table 1.Fig. 1Flow chart of participant selection. NVSS, the National Vital Statistics System; NICU, neonatal intensive care unitTable 1Characteristics of the included populationVariablesTotal ($$n = 52$$,544)Pregnant women without GDM ($$n = 44$$,981)Pregnant women with GDM ($$n = 7563$$)StatisticsPMaternal age at delivery, years, Mean ± SD46.39 ± 1.6346.40 ± 1.6446.37 ± 1.62t = 1.230.220Race, n (%)χ2 = 152.230< 0.001 Asian7082 (13.48)5736 (12.75)1346 (17.80) Black8103 (15.42)7050 (15.67)1053 (13.92) White36,053 (68.61)31,103 (69.15)4950 (65.45) Other1306 (2.49)1092 (2.43)214 (2.83)Education, n (%)χ2 = 375.705< 0.001 Less than 12 grade5975 (11.37)4730 (10.52)1245 (16.46) High school/GED6920 (13.17)5738 (12.76)1182 (15.63) Some college or AA10,203 (19.42)8631 (19.19)1572 (20.79) Bachelor or higher29,446 (56.04)25,882 (57.54)3564 (47.12)Pre-pregnancy weight, lb., Mean ± SD157.39 ± 35.63156.01 ± 34.77165.63 ± 39.36t = − 19.99< 0.001Pre-pregnancy BMI, n (%)χ2 = 914.794< 0.001 Underweight926 (1.76)844 (1.88)82 (1.08) Normal22,601 (43.01)20,307 (45.15)2294 (30.33) Overweight16,180 (30.79)13,781 (30.64)2399 (31.72) Obesity12,837 (24.43)10,049 (22.34)2788 (36.86)Delivery weight, lb., Mean ± SD185.27 ± 36.18184.45 ± 35.49190.15 ± 39.66t = −11.72< 0.001Weight gain, lb., M (Q1, Q3)27.00 (19.00, 36.00)28.00 (20.00, 36.00)23.00 (15.00, 33.00)Z = -23.502< 0.001Smoking before pregnancy, n (%)χ2 = 2.2180.136 No51,401 (97.82)44,020 (97.86)7381 (97.59) Yes1143 (2.18)961 (2.14)182 (2.41)*Smoking status* 1st trimester, n (%)χ2 = 0.1780.673 No51,671 (98.34)44,238 (98.35)7433 (98.28) Yes873 (1.66)743 (1.65)130 (1.72)*Smoking status* 2nd trimester, n (%)χ2 = 0.0010.978 No51,771 (98.53)44,319 (98.53)7452 (98.53) Yes773 (1.47)662 (1.47)111 (1.47)*Smoking status* 3rd trimester, n (%)χ2 = 0.0040.947 No51,805 (98.59)44,349 (98.59)7456 (98.59) Yes739 (1.41)632 (1.41)107 (1.41)Hypertension eclampsia, n (%)χ2 = 5.4450.020 No52,234 (99.41)44,730 (99.44)7504 (99.22) Yes310 (0.59)251 (0.56)59 (0.78)Gestational hypertension, n (%)χ2 = 346.242< 0.001 No46,470 (88.44)40,260 (89.50)6210 (82.11) Yes6074 (11.56)4721 (10.50)1353 (17.89)Pre-pregnancy hypertension, n (%)χ2 = 320.794< 0.001 No49,818 (94.81)42,967 (95.52)6851 (90.59) Yes2726 (5.19)2014 (4.48)712 (9.41)Number of prenatal visits, M (Q1, Q3)12.00 (10.00, 14.00)12.00 (9.00, 14.00)12.00 (10.00, 15.00)$Z = 11.083$< 0.001WIC, n (%)χ2 = 279.140< 0.001 No41,933 (79.81)36,437 (81.01)5496 (72.67) Yes10,611 (20.19)8544 (18.99)2067 (27.33)Plurality, M (Q1, Q3)1.00 (1.00, 1.00)1.00 (1.00, 1.00)1.00 (1.00, 1.00)$Z = 0.4570.648$Prior birth now living, M (Q1,Q3)1.00 (0.00, 3.00)1.00 (0.00, 3.00)1.00 (0.00, 3.00)$Z = 5.848$< 0.001Prior birth now dead, Mean ± SD0.03 ± 0.280.03 ± 0.270.04 ± 0.30t = −2.300.021Prior other terminations, M (Q1, Q3)0.00 (0.00, 1.00)0.00 (0.00, 1.00)0.00 (0.00, 2.00)$Z = 6.863$< 0.001Total birth order, M (Q1, Q3)3.00 (2.00, 5.00)3.00 (2.00, 5.00)4.00 (2.00, 5.00)$Z = 8.346$< 0.001Gestational age, weeks, Mean ± SD37.62 ± 2.9737.66 ± 3.0037.35 ± 2.79t = 8.75< 0.001Newborn sex, n (%)χ2 = 2.4260.119 Female25,891 (49.27)22,227 (49.41)3664 (48.45) Male26,653 (50.73)22,754 (50.59)3899 (51.55)Birth weight, g, Mean ± SD3061.08 ± 709.213061.90 ± 710.493056.17 ± 701.57t = 0.650.516Infertility treatment used, n (%)χ2 = 8.4870.004 No36,783 (70.00)31,596 (70.24)5187 (68.58) Yes15,761 (30.00)13,385 (29.76)2376 (31.42)Pregnancy method, n (%)χ2 = 9.3800.002 Natural pregnancy38,681 (73.62)33,222 (73.86)5459 (72.18) *Pregnancy via* ART13,863 (26.38)11,759 (26.14)2104 (27.82)Method of delivery, n (%)χ2 = 89.588< 0.001 Spontaneous19,594 (37.29)17,142 (38.11)2452 (32.42) Non-spontaneous32,950 (62.71)27,839 (61.89)5111 (67.58)Preterm birth, n (%)χ2 = 59.518< 0.001 No39,935 (76.00)34,452 (76.59)5483 (72.50) Yes12,609 (24.00)10,529 (23.41)2080 (27.50) Extremely preterm689 (1.31)622 (1.38)67 (0.89) Very preterm1466 (2.79)1254 (2.79)212 (2.80) Moderate or late preterm10,454 (19.90)8653 (19.24)1801 (23.81)NICU admission, n (%)χ2 = 89.877< 0.001 No43,031 (81.90)37,131 (82.55)5900 (78.01) Yes9513 (18.10)7850 (17.45)1663 (21.99)Low birthweight, n (%)χ2 = 4.7270.030 No42,783 (81.42)36,693 (81.57)6090 (80.52) Yes9761 (18.58)8288 (18.43)1473 (19.48)Small for gestational age, n (%)χ2 = 1.7210.190 No47,303 (90.03)40,526 (90.10)6777 (89.61) Yes5241 (9.97)4455 (9.90)786 (10.39)GDM Gestational diabetes mellitus, SD Standard deviation, GED General educational development, AA Associate degree, BMI Body mass index, WIC the Special Supplemental Nutrition Program for Women, Infants, and Children, NICU Neonatal intensive care unit
## Comparisons of characteristics between women with and without GDM
The results illustrated that there were significant differences between pregnant women with and without GDM in race ($P \leq 0.001$), education ($P \leq 0.001$), pre-pregnancy weight ($P \leq 0.001$), pre-pregnancy BMI ($P \leq 0.001$), delivery weight ($P \leq 0.001$), weight gain ($P \leq 0.001$), number of prenatal visits ($P \leq 0.001$), prior other terminations ($P \leq 0.001$), total birth order ($P \leq 0.001$), WIC ($P \leq 0.001$), gestational age ($P \leq 0.001$), preterm birth ($P \leq 0.001$), prior birth now living ($P \leq 0.001$), prior birth now dead ($$P \leq 0.021$$), hypertension eclampsia ($$P \leq 0.020$$), gestational hypertension ($P \leq 0.001$), pre-pregnancy hypertension ($P \leq 0.001$), NICU admission ($P \leq 0.001$), method of delivery ($P \leq 0.001$), low birthweight ($$P \leq 0.030$$), infertility treatment used ($$P \leq 0.004$$), pregnancy method ($$P \leq 0.002$$) (Table 1).
## Association between GDM and preterm birth in vAMA women
Women with GDM had a significantly higher risk of preterm birth than those without GDM, according to multivariate analysis (OR = 1.26, $95\%$CI = 1.18–1.36, $P \leq 0.001$). Based on sensitivity analysis, the results were consistent before and after imputation. For different races, it was demonstrated that GDM was associated with a significantly increased risk of preterm birth in Asian (OR = 1.28, $95\%$CI = 1.08–1.54, $$P \leq 0.006$$) and White women (OR = 1.32, $95\%$CI = 1.21–1.45, $P \leq 0.001$). GDM was correlated to a significantly greater risk of preterm birth among women without (OR = 1.33, $95\%$CI = 1.21–1.45, $P \leq 0.001$) and with (OR = 1.16, $95\%$CI = 1.02–1.31, $$P \leq 0.020$$) infertility treatment (Table 2, Fig. 2a-c).Table 2Association between GDM and preterm birth in vAMA womenVariablesModel 1aModel 2bModel 3cOR ($95\%$CI)POR ($95\%$CI)POR ($95\%$CI)PAfter imputation GDM NoRefRefRef Yes1.24 (1.18, 1.31)< 0.0011.26 (1.19, 1.33)< 0.0011.26 (1.18, 1.36)< 0.001Before imputation GDM NoRefRefRef Yes1.29 (1.22, 1.37)< 0.0011.31 (1.23, 1.39)< 0.0011.28 (1.19, 1.39)< 0.001Race Asian1.30 (1.13, 1.48)0.0011.32 (1.16, 1.51)< 0.0011.28 (1.08, 1.54)0.006 Black0.96 (0.83, 1.11)0.5490.96 (0.83, 1.11)0.5941.06 (0.88, 1.27)0.545 White1.32 (1.24, 1.42)< 0.0011.34 (1.25, 1.44)< 0.0011.32 (1.21, 1.45)< 0.001 Other0.90 (0.64, 1.27)0.5510.91 (0.64, 1.29)0.5810.96 (0.62, 1.50)0.865Infertility treatment used Non-infertility treatment used1.20 (1.12, 1.29)< 0.0011.22 (1.13, 1.31)< 0.0011.33 (1.21, 1.45)< 0.001 Infertility treatment used1.29 (1.17, 1.41)< 0.0011.27 (1.15, 1.39)< 0.0011.16 (1.02, 1.31)0.020GDM Gestational diabetes mellitus, vAMA Very advanced maternal age, OR Odds ratio, CI Confidence interval, Ref Reference, BMI Body mass indexFor analysis after and before imputation and subgroup analysis by infertility treatment used:a Model 1 was an univariate model;b Model 2 adjusted for maternal age at delivery, race, education, and newborn sex;c Model 3 adjusted for maternal age at delivery, race, education, newborn sex, pre-pregnancy weight, pre-pregnancy BMI, delivery weight, weight gain, smoking before pregnancy, hypertension eclampsia, gestational hypertension, pre-pregnancy hypertension, number of prenatal visits, plurality, total birth order, prior birth now living, prior other terminations, birth weight, pregnancy method, and method of deliveryFor subgroup analysis by race:a Model 1 was an univariate model;b Model 2 adjusted for maternal age at delivery, education, and newborn sex;c Model 3 adjusted for maternal age at delivery, education, newborn sex, pre-pregnancy weight, pre-pregnancy BMI, delivery weight, weight gain, smoking before pregnancy, hypertension eclampsia, gestational hypertension, pre-pregnancy hypertension, number of prenatal visits, plurality, total birth order, prior birth now living, prior other terminations, birth weight, pregnancy method, and method of deliveryFig. 2Forest plot for the association between GDM and preterm birth in vAMA women. a the association after and before imputation; b the association in women of different races; c the association in women with and without infertility treatment. GDM, gestational diabetes mellitus; vAMA, very advanced maternal age; OR, odds ratio; CI, confidence interval; BMI, body mass index Further, preterm birth was subdivided into extremely preterm birth, very preterm birth, and moderate or late preterm birth. Multivariate analysis showed that compared with women without GDM, those with GDM had a significantly higher risk of moderate or late preterm birth (OR = 1.27, $95\%$CI = 1.18–1.37, $P \leq 0.001$); no significant association was observed between GDM and extremely preterm birth and between GDM and very preterm birth (Table 3, Fig. 3).Table 3Association between GDM and different stages of preterm birth in vAMA womenPreterm birthModel 1aModel 2bModel 3cOR ($95\%$CI)POR ($95\%$CI)POR ($95\%$CI)PExtremely preterm0.68 (0.53–0.87)0.0030.68 (0.53–0.88)0.0031.45 (0.86–2.42)0.161Very preterm1.06 (0.92–1.23)0.4251.06 (0.92–1.24)0.4151.21 (0.96–1.52)0.107Moderate or late preterm1.31 (1.23–1.39)< 0.0011.33 (1.26–1.41)< 0.0011.27 (1.18–1.37)< 0.001GDM Gestational diabetes mellitus, vAMA Very advanced maternal age, OR Odds ratio, CI Confidence interval, WIC the Special Supplemental Nutrition Program for Women, Infants, and Childrena Model 1 was an univariate model;b Model 2 adjusted for maternal age at delivery, race, education, and newborn sex;c Model 3 adjusted for maternal age at delivery, race, education, newborn sex, delivery weight, smoking status 2nd trimester, hypertension eclampsia, gestational hypertension, pre-pregnancy hypertension, number of prenatal visits, WIC, plurality, prior other terminations, total birth order, birth weight, pregnancy method, and method of deliveryFig. 3Forest plot for the association between GDM and different stages of preterm birth in vAMA women. GDM, gestational diabetes mellitus; vAMA, very advanced maternal age; OR, odds ratio; CI, confidence interval; WIC, the Special Supplemental Nutrition Program for Women, Infants, and Children
## Association between GDM and NICU admission in vAMA women
Women with GDM had a significantly greater risk of NICU admission than those without, as illustrated by multivariate analysis (OR = 1.33, $95\%$CI = 1.23–1.43, $P \leq 0.001$). Based on sensitivity analysis, the results were consistent before and after imputation. According to subgroup analysis, GDM was associated with a significantly increased risk of NICU admission in Asian (OR = 1.22, $95\%$CI = 1.01–1.48, $$P \leq 0.045$$), Black (OR = 1.35, $95\%$CI = 1.11–1.63, $$P \leq 0.003$$), White (OR = 1.34, $95\%$CI = 1.22–1.47, $P \leq 0.001$), and other (OR = 1.65, $95\%$CI = 1.02–2.66, $$P \leq 0.040$$) races; a significant elevated risk of NICU admission was found in women with GDM, regardless of whether they received infertility treatment (OR = 1.31, $95\%$CI = 1.16–1.49, $P \leq 0.001$) or not (OR = 1.33, $95\%$CI = 1.21–1.46, $P \leq 0.001$) (Table 4).Table 4Association between GDM and NICU admission in vAMA womenVariablesModel 1aModel 2bModel 3cOR ($95\%$CI)POR ($95\%$CI)POR ($95\%$CI)PAfter imputation GDM NoRefRefRef Yes1.33 (1.26, 1.42)< 0.0011.37 (1.29, 1.46)< 0.0011.33 (1.23, 1.43)< 0.001Before imputation GDM NoRefRefRef Yes1.37 (1.28, 1.46)< 0.0011.40 (1.31, 1.50)< 0.0011.32 (1.22, 1.44)< 0.001Race Asian1.35 (1.17,1.57)< 0.0011.37 (1.18, 1.59)< 0.0011.22 (1.01, 1.48)0.045 Black1.15 (0.99, 1.34)0.0701.16 (1.00, 1.35)0.05811.35 (1.11, 1.63)0.003 White1.40 (1.30, 1.50)< 0.0011.434 (1.33, 1.55)< 0.0011.34 (1.22, 1.47)< 0.001 Other1.23 (0.85, 1.77)0.2821.28 (0.88, 1.87)0.1951.65 (1.02, 2.66)0.040Infertility treatment used Non-infertility treatment used1.28 (1.18, 1.38)< 0.00011.31 (1.21, 1.42)< 0.0011.33 (1.21, 1.46)< 0.001 Infertility treatment used1.40 (1.27, 1.54)< 0.00011.39 (1.26, 1.53)< 0.0011.31 (1.16, 1.49)< 0.001GDM Gestational diabetes mellitus, NICU Neonatal intensive care unit, vAMA Very advanced maternal age, OR Odds ratio, CI Confidence interval, Ref Reference, BMI Body mass indexFor analysis after and before imputation and subgroup analysis by infertility treatment used:a Model 1 was an univariate model;b Model 2 adjusted for maternal age at delivery, race, education, and newborn sex;c Model 3 adjusted for maternal age at delivery, race, education, newborn sex, pre-pregnancy weight, pre-pregnancy BMI, delivery weight, weight gain, gestational hypertension, pre-pregnancy hypertension, number of prenatal visits, plurality, total birth order, birth weight, pregnancy method, method of delivery, and preterm birthFor subgroup analysis by race:a Model 1 was an univariate model;b Model 2 adjusted for maternal age at delivery, education, and newborn sex;c Model 3 adjusted for maternal age at delivery, education, newborn sex, pre-pregnancy weight, pre-pregnancy BMI, delivery weight, weight gain, gestational hypertension, pre-pregnancy hypertension, number of prenatal visits, plurality, total birth order, birth weight, pregnancy method, method of delivery, and preterm birth
## Association between GDM and low birthweight in vAMA women
Multivariate analysis demonstrated that in contrast to women without GDM, those with GDM had a significantly decreased risk of low birthweight (OR = 0.91, $95\%$CI = 0.84–0.98, $$P \leq 0.010$$). Based on sensitivity analysis, the results were consistent before and after imputation. GDM was related to a significantly lower risk of low birthweight among Blacks (OR = 0.80, $95\%$CI = 0.66–0.96, $$P \leq 0.019$$). In women without infertility treatment, GDM was associated with a significantly reduced risk of low birthweight (OR = 0.87, $95\%$CI = 0.79–0.96, $$P \leq 0.006$$) (Table 5).Table 5Association between GDM and low birthweight in vAMA womenVariablesModel 1aModel 2bModel 3cOR ($95\%$CI)POR ($95\%$CI)POR ($95\%$CI)PAfter imputation GDM NoRefRefRef Yes1.07 (1.01, 1.14)0.0301.10 (1.03, 1.17)0.0040.91 (0.84, 0.98)0.010Before imputation GDM NoRefRefRef Yes1.10 (1.03, 1.18)0.0051.12 (1.05, 1.20)0.0010.91 (0.84, 0.98)0.020Race Asian1.24 (1.08, 1.43)0.0031.28 (1.11, 1.48)0.0010.95 (0.80, 1.13)0.571 Black0.85 (0.73, 0.99)0.0420.87 (0.74, 1.02)0.0830.80 (0.66, 0.96)0.019 White1.09 (1.01, 1.18)0.0241.12 (1.03, 1.21)0.0050.91 (0.83, 1.00)0.051 Other0.75 (0.50, 1.13)0.1620.81 (0.54, 1.24)0.3330.68 (0.40, 1.17)0.163Infertility treatment used Non-infertility treatment used0.94 (0.86, 1.02)0.1120.97 (0.89, 1.05)0.430.87 (0.79, 0.96)0.006 Infertility treatment used1.27 (1.15, 1.40)< 0.0011.24 (1.13, 1.37)< 0.0010.96 (0.85, 1.09)0.551GDM Gestational diabetes mellitus, vAMA Very advanced maternal age, OR Odds ratio, CI Confidence interval, Ref Reference, BMI Body mass indexFor analysis after and before imputation and subgroup analysis by infertility treatment used:a Model 1 was an univariate model;b Model 2 adjusted for maternal age at delivery, race, education, and newborn sex;c Model 3 adjusted for maternal age at delivery, race, education, newborn sex, pre-pregnancy weight, pre-pregnancy BMI, delivery weight, weight gain, smoking before pregnancy, smoking status 1st trimester, hypertension eclampsia, gestational hypertension, pre-pregnancy hypertension, number of prenatal visits, plurality, prior birth now living, prior other terminations, pregnancy method, and method of deliveryFor subgroup analysis by race:a Model 1 was an univariate model;b Model 2 adjusted for maternal age at delivery, education, and newborn sex;c Model 3 adjusted for maternal age at delivery, education, newborn sex, pre-pregnancy weight, pre-pregnancy BMI, delivery weight, weight gain, smoking before pregnancy, smoking status 1st trimester, hypertension eclampsia, gestational hypertension, pre-pregnancy hypertension, number of prenatal visits, plurality, prior birth now living, prior other terminations, pregnancy method, and method of delivery
## Association between GDM and small for gestational age in vAMA women
No significant association was found between GDM and a risk of small for gestational age, as exhibited by multivariate analysis (OR = 0.95, $95\%$CI = 0.87–1.03, $$P \leq 0.200$$). Based on sensitivity analysis, the results were consistent before and after imputation. As regards different races, White women with GDM had a significantly decreased risk of small for gestational age (OR = 0.89, $95\%$CI = 0.80–0.99, $$P \leq 0.043$$). There was no significant association between GDM and small for gestational age in women with non-infertility treatment and infertility treatment (Table 6).Table 6Association between GDM and small for gestational age in vAMA womenVariablesModel 1aModel 2bModel 3cOR ($95\%$CI)POR ($95\%$CI)POR ($95\%$CI)PAfter imputation GDM NoRefRefRef Yes1.06 (0.97, 1.14)0.1901.07 (0.98, 1.16)0.1260.95 (0.87, 1.03)0.200Before imputation GDM NoRefRefRef Yes1.04 (0.95, 1.13)0.4021.05 (0.96, 1.14)0.3240.92 (0.84, 1.01)0.068Race Asian1.27 (1.07, 1.52)0.0071.29 (1.08, 1.55)0.0041.08 (0.89 1.31)0.438 Black0.92 (0.75, 1.12)0.3930.93 (0.76, 1.14)0.4950.89 (0.72, 1.09)0.260 White1.01 (0.91, 1.12)0.8251.02 (0.92, 1.14)0.6870.89 (0.80, 0.99)0.043 Other1.26 (0.79, 2.00)0.3341.35 (0.84, 2.15)0.2121.29 (0.78, 2.16)0.324Infertility treatment used Non-infertility treatment used0.96 (0.87, 1.07)0.4530.98 (0.88, 1.08)0.6550.91 (0.81, 1.01)0.076 Infertility treatment used1.21 (1.07, 1.38)0.0031.19 (1.05, 1.36)0.0071.03 (0.90, 1.18)0.705GDM Gestational diabetes mellitus, vAMA Very advanced maternal age, OR Odds ratio, CI Confidence interval, Ref ReferenceFor analysis after and before imputation and subgroup analysis by infertility treatment used:a Model 1 was an univariate model;b Model 2 adjusted for maternal age at delivery, race, education, and newborn sex;c Model 3 adjusted for maternal age at delivery, race, education, newborn sex, pre-pregnancy weight, weight gain, smoking status 1st trimester, hypertension eclampsia, gestational hypertension, pre-pregnancy hypertension, plurality, prior birth now living, pregnancy method, and method of deliveryFor subgroup analysis by race:a Model 1 was an univariate model;b Model 2 adjusted for maternal age at delivery, education, and newborn sex;c Model 3 adjusted for maternal age at delivery, education, newborn sex, pre-pregnancy weight, weight gain, smoking status 1st trimester, hypertension eclampsia, gestational hypertension, pre-pregnancy hypertension, plurality, prior birth now living, pregnancy method, and method of delivery
## Discussion
The present study assessed the association between GDM and adverse infant outcomes (preterm birth, NICU admission, low birthweight and small for gestational age) in vAMA women applying data from the NVSS database. GDM was identified to be positively associated with the risk of preterm birth, especially moderate or late preterm birth; the risks of NICU admission and low birthweight were correlated with GDM in women of very advanced age. According to a prior meta-analysis, women ≥35 years old were more likely to have GDM and worse perinatal outcomes including preterm delivery, low birthweight infants and higher rates of NICU admission [24], which was also supported by Frick et al. [ 25] and Carolan et al. [ 26] Fuchs et al. found that women aged 40 years and over had a greater risk of preterm birth [27]. The risk of small-for-gestational-age infants was approximately doubled in vAMA women compared with women aged 35–39 years [28]. As for the relationship between GDM and infant outcomes, Billionnet et al. [ 11] reported the elevated risks of preterm birth and macrosomia in women with an average age of 30 years having GDM versus those having no diabetes. A cohort study of 46,230 deliveries found that GDM was correlated with a significantly higher risk of spontaneous preterm birth [29]. GDM was associated with mild increases in birth size, as shown by other authors [30]. Few previous studies have focused on the association between GDM and infant outcomes among vAMA women. This study filled this gap, and identified vAMA women with higher risks of infant outcomes. Measures targeting GDM may be adopted to manage these risks for women of vAMA.
As for the possible causes of the association between GDM and preterm delivery in vAMA women, GDM has been associated with polyhydramnios [31, 32], and polyhydramnios can lead to increased uterine tension, and thus induce uterine contractions and cause premature birth. As shown by Buen et al., polyhydramnios acts as a risk factors for preterm delivery [33]. The relatively poor intrauterine environment of vAMA women, which is not conducive to the growth and development of the fetus, may contribute to the positive association of GDM and preterm birth. Additionally, consistent declines in β cell function and insulin secretion are symbols of aging in humans [34–37], and aging effects interact with diabetes to accelerate the progression of many common diabetes complications [38], which may make the association of GDM with premature birth more significant among vAMA women. Considering the finding that over $75\%$ of preterm births were indicated preterm deliveries, clinical practice patterns may play an important role. Indicated delivery is usually chosen for preterm birth in women of very advanced age in clinical practice. Since the physical strength, productivity, cervical elasticity and dilatation ability of vAMA women are inferior to those of young people, indicated preterm delivery (forceps, vacuum, cesarean) may reduce the risk of adverse pregnancy outcomes and complications among these women. Of note, we further showed that the above relationship existed between GDM and moderate or late preterm birth. In clinical practice, more attention should be paid to vAMA women with GDM, and appropriate measures can be taken to reduce risks. The correlations of GDM with NICU admission and low birthweight in vAMA women were also revealed in the current study. More investigations are warranted to consolidate our findings.
Interestingly, we found that the association between GDM and preterm birth varied by race in vAMA women. GDM was associated with a significantly higher risk of preterm delivery among Asians and Whites, while no association was identified among Blacks and other races. Some studies pointed out that in the United States, non-Hispanic Black race (compared with non-Hispanic White race) was a risk factor for preterm birth [39, 40], which did not seem to cohere with our results. However, increased age may have an important influence on the relationship of GDM and preterm birth in different races, which these studies did not take into consideration. Thus, it is worth paying more attention to the effect of vAMA on this relationship. The relationship between GDM and low birthweight was shown to vary by race and use of infertility treatment, and White women with GDM were at a significantly reduced risk of having small-for-gestational-age infants. Further studies are required for validation, and corresponding management of infant outcomes can be undertaken for populations with different risks.
Our study has several strengths. A large, nationally representative sample size ($$n = 52$$,544) with adequate power (power = 1) was utilized to assess the associations between GDM and adverse infant outcomes (preterm birth, NICU admission, low birthweight and small for gestational age) in pregnant women of very advanced age, making the results reliable. Different stages of preterm birth were also analyzed, and the aforementioned associations were further evaluated according to race and use of infertility treatment, which provides additional insights into these associations for different populations.
A few limitations of the present study need to be noted. First, this study was retrospective in nature, and some data were missing during data collection. To address this, missing data were imputed using multiple imputation, and sensitivity analyses confirmed the reliability of the results. Additionally, the level of evidence for this study is low. Prospective studies are needed for verification. Second, there are no data on some covariates, such as caesarean (elective and emergency), physical activity, psychological and social stress and depression during pregnancy, blood glucose, medication use (especially psychotropic medication), and socioeconomic factors (apart from education and WIC), which may influence our results. Third, this study focused on the U.S. population and has limited generalizability. Future studies are required to investigate the relationship between GDM and infant outcomes in vAMA women with consideration of the above covariates, so as to confirm our findings. This relationship can also be evaluated in populations from other countries. Notably, $76.92\%$ of preterm births were caused by indicated delivery. Hypertension and the pregnancy method may be related to indicated preterm delivery in vAMA women. Future research can investigate the indications for indicated preterm delivery, and assess whether hypertension and the pregnancy method are associated with indicated preterm delivery.
Based on our findings, vAMA women with GDM had higher risks of preterm birth and NICU admission than those without. Greater attention should be paid to vAMA women with GDM and early interventions should be taken to lower the risks. Improving GDM may be a viable approach. Since populations grouped by age and use of infertility treatment had different risks of adverse infant outcomes, individualized measures should be developed. vAMA women with pregnancy planning should be informed of increased risks of preterm birth and NICU admission when they had GDM and corresponding advice can be provided by clinicians or healthcare givers.
## Conclusion
GDM was associated with an increased risk of preterm birth, especially moderate or late preterm birth; the risks of NICU admission and low birthweight were correlated with GDM among vAMA women. More investigations are warranted to verify this conclusion.
## Supplementary Information
Additional file 1: Supplementary Table 1. Proportion of missing values.
## References
1. Hodes-Wertz B, Druckenmiller S, Smith M, Noyes N. **What do reproductive-age women who undergo oocyte cryopreservation think about the process as a means to preserve fertility?**. *Fertil Steril* (2013) **100** 1343-1349. DOI: 10.1016/j.fertnstert.2013.07.201
2. Lavery JA, Friedman AM, Keyes KM, Wright JD, Ananth CV. **Gestational diabetes in the United States: temporal changes in prevalence rates between 1979 and 2010**. *BJOG* (2017) **124** 804-813. DOI: 10.1111/1471-0528.14236
3. Buiten RA, Ploumen EH, Zocca P, Doggen CJM, Jessurun GAJ, Schotborgh CE. **Thin composite-wire-strut Zotarolimus-eluting stents versus ultrathin-strut Sirolimus-eluting stents in BIONYX at 2 years**. *JACC Cardiovasc Interv* (2020) **13** 1100-1109. DOI: 10.1016/j.jcin.2020.01.230
4. Yong HY, Mohd Shariff Z, Mohd Yusof BN, Rejali Z, Tee YYS, Bindels J. **Independent and combined effects of age, body mass index and gestational weight gain on the risk of gestational diabetes mellitus**. *Sci Rep* (2020) **10** 8486. DOI: 10.1038/s41598-020-65251-2
5. Alberti KG, Zimmet PZ. **Definition, diagnosis and classification of diabetes mellitus and its complications. Part 1: diagnosis and classification of diabetes mellitus provisional report of a WHO consultation**. *Diabet Med* (1998) **15** 539-553. DOI: 10.1002/(SICI)1096-9136(199807)15:7<539::AID-DIA668>3.0.CO;2-S
6. Plows JF, Stanley JL, Baker PN, Reynolds CM, Vickers MH. **The pathophysiology of gestational diabetes mellitus**. *Int J Mol Sci* (2018) **19** 3342. DOI: 10.3390/ijms19113342
7. Zhang M, Wang Y, Qi X. **Effect of very advanced maternal age on pregnant women and fetuses**. *J Coll Physicians Surg Pak* (2021) **30** 542-5. PMID: 34027866
8. Kortekaas JC, Kazemier BM, Keulen JKJ, Bruinsma A, Mol BW, Vandenbussche F. **Risk of adverse pregnancy outcomes of late- and postterm pregnancies in advanced maternal age: a national cohort study**. *Acta Obstet Gynecol Scand* (2020) **99** 1022-1030. DOI: 10.1111/aogs.13828
9. Rademaker D, Hukkelhoven CWPM, van Pampus MG. **Adverse maternal and perinatal pregnancy outcomes related to very advanced maternal age in primigravida and multigravida in the Netherlands: a population-based cohort**. *Acta Obstet Gynecol Scand* (2021) **100** 941-948. DOI: 10.1111/aogs.14064
10. Leader J, Bajwa A, Lanes A, Hua X, Rennicks White R, Rybak N. **The effect of very advanced maternal age on maternal and neonatal outcomes: a systematic review**. *J Obstet Gynaecol Can* (2018) **40** 1208-1218. DOI: 10.1016/j.jogc.2017.10.027
11. Billionnet C, Mitanchez D, Weill A, Nizard J, Alla F, Hartemann A. **Gestational diabetes and adverse perinatal outcomes from 716,152 births in France in 2012**. *Diabetologia.* (2017) **60** 636-644. DOI: 10.1007/s00125-017-4206-6
12. Silva-Zolezzi I, Samuel TM, Spieldenner J. **Maternal nutrition: opportunities in the prevention of gestational diabetes**. *Nutr Rev* (2017) **75** 32-50. DOI: 10.1093/nutrit/nuw033
13. Mustad VA, Huynh DTT, López-Pedrosa JM, Campoy C, Rueda R. **The role of dietary carbohydrates in gestational diabetes**. *Nutrients.* (2020) **12** 385. DOI: 10.3390/nu12020385
14. Liu L, Oza S, Hogan D, Chu Y, Perin J, Zhu J. **Global, regional, and national causes of under-5 mortality in 2000-15: an updated systematic analysis with implications for the sustainable development goals**. *Lancet (London, England)* (2016) **388** 3027-3035. DOI: 10.1016/S0140-6736(16)31593-8
15. Walani SR. **Global burden of preterm birth**. *Int J Gynaecol Obstet* (2020) **150** 31-33. DOI: 10.1002/ijgo.13195
16. **Global, regional, and national age-sex specific mortality for 264 causes of death, 1980–2016: a systematic analysis for the global burden of disease study 2016**. *Lancet* (2017) **390** 1151-1210. DOI: 10.1016/S0140-6736(17)32152-9
17. Farrar D, Simmonds M, Bryant M, Sheldon TA, Tuffnell D, Golder S. **Hyperglycaemia and risk of adverse perinatal outcomes: systematic review and meta-analysis**. *BMJ (Clinical research ed)* (2016) **354** i4694. PMID: 27624087
18. Delnord M, Zeitlin J. **Epidemiology of late preterm and early term births - an international perspective**. *Semin Fetal Neonatal Med* (2019) **24** 3-10. DOI: 10.1016/j.siny.2018.09.001
19. Berger H, Melamed N, Davis BM, Hasan H, Mawjee K, Barrett J. **Impact of diabetes, obesity and hypertension on preterm birth: population-based study**. *PLoS One* (2020) **15** e0228743. DOI: 10.1371/journal.pone.0228743
20. Diboun I, Ramanjaneya M, Majeed Y, Ahmed L, Bashir M, Butler AE. **Metabolic profiling of pre-gestational and gestational diabetes mellitus identifies novel predictors of pre-term delivery**. *J Transl Med* (2020) **18** 366. DOI: 10.1186/s12967-020-02531-5
21. Venkatesh KK, Lynch CD, Powe CE, Costantine MM, Thung SF, Gabbe SG. **Risk of adverse pregnancy outcomes among pregnant individuals with gestational diabetes by race and ethnicity in the United States, 2014-2020**. *JAMA.* (2022) **327** 1356-1367. DOI: 10.1001/jama.2022.3189
22. Deputy NP, Kim SY, Conrey EJ, Bullard KM. **Prevalence and changes in preexisting diabetes and gestational diabetes among women who had a live birth - United States, 2012-2016**. *MMWR Morb Mortal Wkly Rep* (2018) **67** 1201-1207. DOI: 10.15585/mmwr.mm6743a2
23. Blencowe H, Cousens S, Oestergaard MZ, Chou D, Moller AB, Narwal R. **National, regional, and worldwide estimates of preterm birth rates in the year 2010 with time trends since 1990 for selected countries: a systematic analysis and implications**. *Lancet.* (2012) **379** 2162-2172. DOI: 10.1016/S0140-6736(12)60820-4
24. Pinheiro RL, Areia AL, Mota Pinto A, Donato H. **Advanced maternal age: adverse outcomes of pregnancy, A Meta-Analysis**. *Acta Méd Port* (2019) **32** 219-226. DOI: 10.20344/amp.11057
25. Frick AP. **Advanced maternal age and adverse pregnancy outcomes. Best Pract Res**. *Clin obstetrics and gynaecology* (2021) **70** 92-100
26. Carolan M. **Maternal age ≥45 years and maternal and perinatal outcomes: a review of the evidence**. *Midwifery.* (2013) **29** 479-489. DOI: 10.1016/j.midw.2012.04.001
27. Fuchs F, Monet B, Ducruet T, Chaillet N, Audibert F. **Effect of maternal age on the risk of preterm birth: a large cohort study**. *PLoS One* (2018) **13** e0191002. DOI: 10.1371/journal.pone.0191002
28. Arya S, Mulla ZD, Plavsic SK. **Outcomes of women delivering at very advanced maternal age**. *J Women's Health (Larchmt)* (2018) **27** 1378-1384. DOI: 10.1089/jwh.2018.7027
29. Hedderson MM, Ferrara A, Sacks DA. **Gestational diabetes mellitus and lesser degrees of pregnancy hyperglycemia: association with increased risk of spontaneous preterm birth**. *Obstet Gynecol* (2003) **102** 850-856. PMID: 14551018
30. Kong L, Nilsson IAK, Gissler M, Lavebratt C. **Associations of maternal diabetes and body mass index with offspring birth weight and prematurity**. *JAMA Pediatr* (2019) **173** 371-378. DOI: 10.1001/jamapediatrics.2018.5541
31. **190: gestational diabetes mellitus**. *Obstet Gynecol* (2018) **131** e49-e64. DOI: 10.1097/AOG.0000000000002501
32. Moore LE. **Amount of polyhydramnios attributable to diabetes may be less than previously reported**. *World J Diabetes* (2017) **8** 7-10. DOI: 10.4239/wjd.v8.i1.7
33. Buen M, Amaral E, Souza RT, Passini R, Lajos GJ, Tedesco RP. **Maternal work and spontaneous preterm birth: a multicenter observational study in Brazil**. *Sci Rep* (2020) **10** 9684. DOI: 10.1038/s41598-020-66231-2
34. Kirkman MS, Briscoe VJ, Clark N, Florez H, Haas LB, Halter JB. **Diabetes in older adults**. *Diabetes Care* (2012) **35** 2650-2664. DOI: 10.2337/dc12-1801
35. Krishnamurthy J, Ramsey MR, Ligon KL, Torrice C, Koh A, Bonner-Weir S. **p16INK4a induces an age-dependent decline in islet regenerative potential**. *Nature.* (2006) **443** 453-457. DOI: 10.1038/nature05092
36. Dai C, Hang Y, Shostak A, Poffenberger G, Hart N, Prasad N. **Age-dependent human β cell proliferation induced by glucagon-like peptide 1 and calcineurin signaling**. *J Clin Invest* (2017) **127** 3835-3844. DOI: 10.1172/JCI91761
37. Utzschneider KM, Carr DB, Hull RL, Kodama K, Shofer JB, Retzlaff BM. **Impact of intra-abdominal fat and age on insulin sensitivity and beta-cell function**. *Diabetes.* (2004) **53** 2867-2872. DOI: 10.2337/diabetes.53.11.2867
38. LeRoith D, Biessels GJ, Braithwaite SS, Casanueva FF, Draznin B, Halter JB. **Treatment of diabetes in older adults: an Endocrine Society* clinical practice guideline**. *J Clin Endocrinol Metab* (2019) **104** 1520-1574. DOI: 10.1210/jc.2019-00198
39. Hong X, Bartell TR, Wang X. **Gaining a deeper understanding of social determinants of preterm birth by integrating multi-omics data**. *Pediatr Res* (2021) **89** 336-343. DOI: 10.1038/s41390-020-01266-9
40. McKinnon B, Yang S, Kramer MS, Bushnik T, Sheppard AJ, Kaufman JS. **Comparison of black-white disparities in preterm birth between Canada and the United States**. *CMAJ* (2016) **188** E19-e26. DOI: 10.1503/cmaj.150464
|
---
title: Ellagic acid ameliorates aging-induced renal oxidative damage through upregulating
SIRT1 and NRF2
authors:
- Niloufar Naghibi
- Asie Sadeghi
- Sajjadeh Movahedinia
- Mahdis Rahimi Naiini
- Mohammad Amin Rajizadeh
- Faegheh Bahri
- Mahdieh Nazari-Robati
journal: BMC Complementary Medicine and Therapies
year: 2023
pmcid: PMC9999491
doi: 10.1186/s12906-023-03907-y
license: CC BY 4.0
---
# Ellagic acid ameliorates aging-induced renal oxidative damage through upregulating SIRT1 and NRF2
## Abstract
### Background
Aging is associated with impaired renal function and structural alterations. Oxidative stress plays a vital role in renal senescence and damage. Sirtuin 1 (SIRT1) is thought to protect cells from oxidative stress through nuclear factor erythroid 2-related factor 2 (NRF2). Ellagic acid (EA), a natural antioxidant, has been demonstrated to have renoprotective roles in vitro and in vivo. This study investigated if SIRT1 and NRF2 mediate the protective effects of EA in aged kidneys.
### Methods
Male Wistar rats were divided into three groups including young (4 months), old, and old + EA (25 months). Young and old groups received EA solvent, while the old + EA group was treated with EA (30 mg/kg) by gavage for 30 days. Then, the level of renal oxidative stress, SIRT1 and NRF2 expression, kidney function parameters, and histopathological indices were measured.
### Results
Treatment with EA significantly increased the level of antioxidant enzymes and reduced malondialdehyde concentration ($P \leq 0.01$). Moreover, EA administration remarkably upregulated mRNA and protein levels of SIRT1 and NRF2 as well as deacetylated NRF2 protein ($P \leq 0.05$). Additionally, EA treated rats improved kidney function and histopathological scores ($P \leq 0.05$).
### Conclusions
These findings suggest that ellagic acid exerts protective effects on aged kidneys by activating SIRT1 and NRF2 signaling.
### Supplementary Information
The online version contains supplementary material available at 10.1186/s12906-023-03907-y.
## Background
Aging is a physiological process associated with irreversible functional impairment and predisposition to chronic diseases. As with other organ systems, the functional capabilities of the kidney decline progressively with aging, which primarily manifests as reductions in the glomerular filtration rate (GFR) [1]. On the structural level, the kidney experiences several changes, including nephrosclerosis, glomerular basement membrane thickening, and accumulation of extracellular matrix with aging. These alterations impair the ability of the kidney to recover from injury, thus resulting in the increased susceptibility of the aged kidneys to acute kidney injury (AKI) and the development of progressive chronic kidney disease (CKD) [2].
The aging process is closely connected to enhanced mitochondrial dysfunction and reduced antioxidant capacity leading to redox signaling disruption and a wide range of phenotypic changes, including altered gene expression, arrested cell proliferation and growth, and cellular senescence [3, 4]. In addition, the reduction in mitochondrial function during aging results in a decrease in cellular NAD+ levels, which would be expected to compromise the activities of NAD+-dependent enzymes, including protein deacetylases of the Sirtuin (SIRT) family [5]. In mammals, seven sirtuins constitute an evolutionarily conserved group of enzymes involved in various interrelated cellular processes, such as metabolism, mitochondrial biogenesis, autophagy, and apoptosis [6]. SIRT1 is the most well-studied member of the sirtuin family in the kidney, which is widely expressed in tubular cells and podocytes [7]. By targeting various transcriptional factors for deacetylation, SIRT1 provides renoprotections by reducing interstitial fibrosis, inhibiting tubular and glomerular cell apoptosis, suppressing inflammation, and improving mitochondrial function, and regulating blood pressure [8]. Therefore, SIRT1 dysfunction associated with aging may contribute to the initiation and progression of age-related kidney diseases [6].
Nuclear factor erythroid 2-related factor 2 (NRF2) is a crucial regulator of oxidative stress response that induces transcription of target genes encoding proteins associated with redox regulation through binding to antioxidant response element (ARE) in ARE-driven gene promoters [9]. In normal conditions, NRF2 is thought to be sequestered by Kelch-like ECH-associated protein 1 (Keap1) in the cytosol. Still, upon a redox imbalance, NRF2 dissociates from Keap1 and translocates to the nucleus to initiate the transcription of genes involved in antioxidant defense. However, NRF2 has been described to be dysfunctional during aging, resulting in impaired oxidative stress response signaling [10]. Multiple posttranslational modifications including acetylation-deacetylation regulate the function of NRF2. Consistent with the view that NRF2 is positively regulated by deacetylation, SIRT1 has been described to deacetylate NRF2, which increases the stability and transcriptional activity of NRF2, thus improves the resistance of cells to oxidative damage [11, 12]. In addition, recent findings have shown that SIRT1 can protect cells from oxidative stress by upregulating the level of NRF2 protein [13]. Therefore, pharmacological targeting of SIRT1 in the kidney may counteract the pathological changes involved in kidney aging.
Ellagic acid (EA, 2,3,7,8-tetrahydroxybenzopyrano (5,4,3-cde) benzopyrano-5,10-dione), a naturally occurring polyphenolic compound, is commonly found in nuts, vegetables, and fruits, such as pomegranate, raspberries, strawberries, and grapes. It is well established that EA has a wide range of biological activities, including antioxidant, anti-inflammatory, anticancer, and antidiabetic properties [14]. Recent findings have suggested that EA ameliorates oxidative stress through the upregulation of NRF2 and, thus, the induction of antioxidant enzymes [15]. Furthermore, EA has been shown to protect kidneys against carbon tetrachloride-induced oxidative damage via activation of NRF2-driven antioxidant signal pathway [16]. In addition, EA has recently been reported to prevent iron oxide-induced nephrotoxicity by inducing the expression of SIRT1 in renal tissues [17]. Despite these beneficial effects of EA, the role of EA in the aging kidney has not been well defined. Therefore, in this study, we aimed to investigate the possible effect of EA on renal SIRT1 and NRF2 and kidney function in aged rats.
## Chemicals and reagents
All antibodies used in this research were purchased from Santa Cruz Biotechnology company (CA, USA). Chemicals were provided by Sigma-Aldrich company (MO, USA). The origin of other reagents and kits has been described in experimental methods.
## Animals and experimental design
In this experimental study, a total of 21 male Wistar rats including 14 aged (25-month-old) and 7 young (4-month-old) rats were used. Animals were housed in standard cages (2–3 rats per cage) under controlled environmental conditions (22 ± 1 °C; $60\%$ humidity, and 12 h light/dark cycle) with free access to standard food and tap water. All experiments were approved by the Ethics Committee of Kerman University of Medical Sciences (IR.KMU.AH.REC.1400.041), and performed according to the guide for the care and use of laboratory animals by National Institutes of Health (NIH). Old animals were randomly divided into two groups of seven rats. EA (Sigma–Aldrich, USA) was dissolved in saline containing $0.1\%$ DMSO. Then young and aged control rats received EA solvent by gavage, while the rats in EA group were treated with EA (30 mg/kg). All treatments were given once daily and continued up to 30 days. The dose of EA was selected based on previous studies [18, 19].
## Sample collection
At the end of experimental period, rats were placed in individual metabolic cages and 24 h urine was collected to measure albumin and creatinine concentration. Then, animals were deeply anesthetized with ketamine (50 mg/kg) and xylazine (5 mg/kg) and blood samples were collected via cardiac puncture for urea and creatinine measurement. Kidneys were immediately taken and washed with cold isotonic saline. The left kidney was fixed in $10\%$ formalin for histopathological examinations and the right kidney stored at -80℃ for biochemical analysis.
## Assessment of renal oxidative stress level
The concentration of malondialdehyde (MDA) as a marker of oxidative stress and the levels of total antioxidant capacity (TAC), catalase (CAT) and superoxide dismutase (SOD) activity were measured using commercial kits and according to the manufacturer’s protocol (Kiazist, Iran). Briefly, kidney tissues were homogenized in lysis buffer containing protease inhibitors (Sigma–Aldrich, USA). After centrifugation by a 3-18KS Sigma centrifuge (Sigma, Germany), supernatants were collected for next analysis. MDA level was quantified by measuring thiobarbituric acid reactive substances produced in the reaction of MDA with thiobarbituric acid. TAC level was measured based on the capacity to convert Cu2+ to Cu+ ion. The activity of catalase was determined according to the reaction of the enzyme with methanol in the presence of hydrogen peroxide and measurement of generated formaldehyde. SOD activity was assayed by measuring the dismutation of superoxide radicals generated by the xanthine/xanthine oxidase system. Protein concentration of lysates was measured using Bradford method. Then the levels of oxidative stress markers were normalized to protein content [20, 21].
## Detection of acetylated NRF2
The kidney tissue was lysed in ice-cold lysis buffer (50 mM Tris, 150 mM NaCl, $1\%$ NP-40) (1:5 w/v) in the presence of protease inhibitor. The lysate was centrifuged at 12,000 rpm for 15 min at 4℃. Then the supernatant was immunoprecipitated with antibody specific to NRF2 (sc-33649, 1:50) at 4℃ overnight. The immunocomplexes were then collected and subjected to western blotting using anti-acetyl lysine antibody (sc-9441, 1:700).
## Western blot analysis
The kidney tissue was homogenized in ice-cold RIPA buffer (1:10 w/v) containing protease inhibitor. The homogenate was centrifuged at 12,000 rpm for 15 min at 4℃. Then the resulting supernatant was collected. Protein concentration was determined by bicinchoninic acid assay. Equal amounts of extracted proteins were separated by $10\%$ sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) and transferred onto the PVDF membranes. The membranes were blocked with $5\%$ non-fat milk in TBST buffer at 4℃ overnight. Then the membranes were incubated with primary antibodies specific to SIRT1 (sc-74465, 1:500), NRF2 (sc-365949, 1:500), and β-actin (sc-47778, 1:1000) for 1 h at room temperature. After washing in TBST, membranes were incubated with horseradish peroxidase (HRP)-conjugated secondary antibodies (sc-2357, 1:10,000) for 1 h at room temperature. Then membranes were washed in TBST and detected with ECL kit (Bio-Rad, USA) for protein bands. The detected bands were quantified using ImageJ analyzing software. β-actin was used as an internal reference.
## RNA extraction and Real-time PCR
Total RNA was extracted from kidney samples using Trizol reagent (GeneAll Biotechnology, Korea) according to the manufacturer’s instructions. The concentration and purity of RNA were determined by a NanoDrop ND-1000 spectrophotometer (Thermo Scientific, USA). Then complementary DNA (cDNA) was synthesized according to the protocol of cDNA synthesis kit (Parstous, Iran). Real-time PCR amplification was performed on Mic PCR system (BMS, Australia) using SYBR Green master mix (Ampliqon, Denmark), cDNA and primers. Primer sequences were as follows: SIRT1, 5′-GGTAGTTCCTCGGTGTCCT -3′ (forward) and 5′-ACCCAATAACAATGAGGAGGTC-3′ (reverse); NRF2, 5′-CACATCCAGACAGACACCAGT-3′ (forward) and 5′-CTACAAATGGGAATGTCTCTGC-3′ (reverse); GAPDH, 5′-AGGTTGTCTCCTGTGACTTC -3′ (forward) and 5′-CTGTTGCTGTAGCCATATTC-3′ (reverse). PCR conditions were 95℃ for 10 min followed by 40 cycles of 95 °C for 40 s, 62 °C for 25 s, and 72 °C for 15 s. *Relative* gene expression was calculated using ΔΔCT method. GAPDH gene was used as internal control [20].
## Measurement of renal function
Blood urea and creatinine levels were analyzed by commercial kits (Pars Azmun, Iran). Urinary concentration of creatinine was also measured. Urine albumin was determined by immunoturbidometric kit (BioSystems, Spain). To estimate GFR, creatinine clearance was calculated using the standard formula: [urine creatinine (mg/dl) × urine volume (ml/24 h)] /[serum creatinine (mg/dl) × 1440 (min/24 h)].
## Histopathological Examination
Fixed kidney tissues in $10\%$ formalin were embedded in paraffin and cut into 4-μm sections. Then renal sections were stained with hematoxylin and eosin (H&E). Moreover, to detect collagen fibers or fibrosis, Masson’s trichrome staining was performed. For each section, five fields were observed under a CX41 Olympus light microscope and digitally photographed (Olympus, Japan). Necrosis, inflammation, tubular atrophy, tubulointerstitial fibrosis were then analyzed and scored as described previously [20].
## Statistical analysis
All data were presented as mean ± SD. SPSS software 20.0 (IBM, USA) were employed to perform data analysis. Statistical significance was determined using Mann–Whitney with the level of significance set at $$P \leq 0.05.$$
## Alleviation of renal oxidative stress after treatment with EA
The obtained results in Fig. 1 showed a marked increase in the level of MDA and a significant decline in the levels of TAC, SOD and CAT in renal tissues of aged rats compared with those of young animals ($P \leq 0.001$). However, administration of EA could attenuate oxidative stress in senescent kidneys, evidenced by a significant reduction in the level of MDA and a remarkable elevation in the levels of TAC, SOD and CAT in comparison with untreated aged renal tissues ($P \leq 0.01$).Fig. 1Effect of ellagic acid (EA) on renal level of A malondialdehyde (MDA) B total antioxidant capacity (TAC) C superoxide dismutase (SOD) and D catalase (CAT) in different groups ($$n = 7$$). All data are mean ± SD. Statistical significance expressed as ** $p \leq 0.01$, *** $p \leq 0.001$
## Upregulation of SIRT1 and NRF2 at mRNA level after treatment with EA
Results indicated a significant reduction in the level of SITR1 and NRF2 gene expression in the kidneys of old compared with young rats ($P \leq 0.001$). However, the level of SIRT1 and NRF2 mRNA was upregulated by 1.7-fold and twofold, respectively in aged animals following treatment with EA compared with old control rats ($P \leq 0.05$ and $P \leq 0.01$, respectively) (Fig. 2).Fig. 2Effect of ellagic acid (EA) on renal level of A SIRT1 and B NRF2 mRNA transcripts in different groups ($$n = 7$$). All data are mean ± SD. Statistical significance expressed as * $p \leq 0.05$, ** $p \leq 0.01$, *** $p \leq 0.001$
## Alterations of SIRT1, NRF2 and acetylated NRF2 at protein level after treatment with EA
Western blot analysis of SIRT1 and NRF2 revealed a significant decrease in the level of these proteins in renal tissues of aged rats compared with young animals ($P \leq 0.001$). Conversely, old rats showed a significant elevated level of acetylated NRF2 (Ac-NRF2) in comparison with young group ($P \leq 0.001$). However, SIRT1 and NRF2 proteins were significantly increased in the kidneys of senescent rats treated with EA compared with aged control group ($P \leq 0.01$). In contrast, there was a significant reduction in the level of Ac-NRF2 in the renal tissues of aged rats received EA in comparison to those of old control animals ($P \leq 0.01$) (Fig. 3).Fig. 3Effect of ellagic acid (EA) on renal level of A SIRT1, B NRF2 and C Acetyl-NRF2 proteins in different groups ($$n = 7$$). D Representative immunoblot images of SIRT1, NRF2, Acetyl-NRF2 and β-Actin. All data are mean ± SD. Statistical significance expressed as ** $p \leq 0.01$, *** $p \leq 0.001$
## Amelioration of renal function after treatment with EA
Old rats showed significant elevated serum levels of urea ($P \leq 0.001$) and creatinine ($P \leq 0.01$) as well as increased urine level of albumin compared with young animals ($P \leq 0.001$). Additionally, creatinine clearance rate was decreased remarkably with aging ($P \leq 0.001$). Comparing with senescent control group, EA administration reduced serum urea ($P \leq 0.05$) and creatinine ($P \leq 0.01$) and improved creatinine clearance rate significantly ($P \leq 0.05$). However, urinary albumin excretion did not change after treatment with EA ($P \leq 0.05$) (Table 1).Table 1Kidney function parameters in different groupsParameterYoungOldOld + EAUrea (mg/dl)54.1 ± 5.268.5 ± 5.8***59.2 ± 4.7#Creatinine (mg/dl)0.64 ± 0.0660.82 ± 0.075**0.67 ± 0.054##Creatinine clearance rate (ml/min/kg)2.5 ± 0.301.5 ± 0.25***2.0 ± 0.27#Albuminuria (μg/day)13.9 ± 2.520.2 ± 2.9***17.8 ± 2.1Body weight (g)329 ± 13400 ± 15***414 ± 17Kidney weight (g)1.0 ± 0.131.4 ± 0.17***1.3 ± 0.14Data are presented as mean ± SD for $$n = 7$$ rats in each groupEA Ellagic acid** $p \leq 0.01$ vs young control*** $p \leq 0.001$ vs young control# $p \leq 0.05$ vs old control## $p \leq 0.01$ vs old control
## Alterations of histopathological scores after treatment with EA
Histopathologically, renal sections of aged rats displayed elevated levels of inflammation, tubular atrophy and fibrosis compared with those of young animals ($P \leq 0.01$) (Fig. 4). However, treatment with EA caused significant improvement in these histopathological features in comparison with senescent control tissues ($P \leq 0.05$) (Table 2).Fig. 4Histopatholoy of kidney tissue in different studied rat groups. Representative images of Hematoxylin and Eosin (original magnification × 200) stain of [1] cortical and [2] medullary areas and [3] Trichrome stain (original magnification × 200) are shown (scale bars = 100 μm). A Young control rats renal tissue with absent to minimal inflammation, no atrophy or fibrosis; B Old control rats kidney tissue exhibiting marked inflammation (long arrow), evidence of hyalin cast formation (arrow head), and mild tubulointerstitial atrophy and fibrosis (short arrow). C Treated old rats renal tissue with mild inflammation (long arrow), minimal atrophy and fibrosis (short arrow)Table 2Grading of histopathological features in different groupsHistopathological featureYoungOldOld + EAInflammation0.0 ± 0.03.0 ± 0.4**2.2 ± 0.3#Necrosis (%)0.0 ± 0.00.0 ± 0.00.0 ± 0.0Tubular atrophy (%)0.0 ± 0.014.0 ± 2.4***4.0 ± 1.0##*Tubulointerstitial fibrosis* (%)0.0 ± 0.010.4 ± 2.8**2.6 ± 0.8#Pathological changes were scored as described in Materials and Methods. Data are presented as mean ± SD for $$n = 7$$ rats in each groupEA Ellagic acid** $p \leq 0.01$ vs young control*** $p \leq 0.001$ vs young control# $p \leq 0.05$ vs old control## $p \leq 0.01$ vs old control
## Discussion
Renal aging is associated with disruptions in cellular homeostasis, leading to a reduction in major protective factors such as SIRT1 and a decrease in responsiveness to physiological damage, including oxidative stress [22]. Therefore, targeting the SIRT1 signaling pathway may be a potential strategy to prevent or slow kidney aging. In the current study, we showed that treatment with ellagic acid (EA) could upregulate the expression of SIRT1 and ameliorate oxidative stress through induction of NRF2 expression and deacetylation. Moreover, EA administration could attenuate aging-induced renal dysfunction and lesions.
Increased oxidative stress and mitochondrial dysfunction are believed to be significant factors contributing to aging [5]. Mitochondria are the main producer of ROS in cells. However, a significant reduction occurs in electron flow through the mitochondrial respiratory chain with aging, which favors increased ROS generation. Therefore, senescent cells have a high content of ROS and accumulative oxidative damage to biomolecules, particularly DNA and Proteins [23]. The removal of oxidative DNA damage depletes intracellular NAD+ pools and impairs the activity of NAD+-dependent enzymes, such as sirtuins [5].
SIRT1, the most extensively studied sirtuin, has been increasingly recognized to play various roles in gene silencing, stress resistance, apoptosis, inflammation, and aging [24]. Emerging evidence shows that SIRT1 expression decreases with aging and may be involved in age-associated diseases [25]. Consistent with our data, Kwon et al. reported that SIRT1 expression was reduced in the kidney of 24-month-old mice compared with 6-month-old mice [26]. In a similar study, SIRT1 expression was indicated to be lower in the renal tissues of 24-month-old mice compared to 2 and 12-month-old mice [27]. However, in the present study, EA administration was found to augment the level of SIRT1 at both gene and protein expression in aged kidneys. EA, a natural polyphenolic compound, has received considerable attention because of its various biological properties, such as radical scavenging, anti-inflammatory, antiviral, cancer, and diabetes-prevention activities [14]. A recent study reported that EA administration significantly enhanced the expression of SIRT1 in rat kidneys and ameliorated cisplatin-induced nephrotoxicity [28]. Additionally, EA has been shown to protect renal tissues against iron oxide-induced damage by promoting the expression of SIRT1 [17]. These findings suggested that SIRT1 plays a crucial role in the protective effects of EA on renal tissue damage.
NRF2 is a redox-sensitive transcription factor that regulates the basal expression of antioxidant genes and confers cytoprotection against oxidative stress. In normal conditions, Keap1, a cysteine-rich protein, is associated with NRF2. Exposure to reactive oxidants leads to the oxidation of critical cysteine residues, thus resulting in NRF2 release and translocation to the nucleus, where it binds to DNA-responsive elements and activates the transcription of several antioxidant genes and major ROS scavenging enzymes. It is well established that aging is associated with a gradual decline in NRF2 level and pathway responsiveness, promoting oxidative injury in senescence tissues [9, 29]. In this sense, the results of the current study indicated a redox imbalance in aged kidneys as evidenced by the downregulation of NRF2 and diminished levels of SOD, CAT, and TAC and elevated levels of MDA. Similar features of oxidative damage associated with reduced NRF2 were previously observed in the kidney of 24-month-old rats [30]. The renal protective effect of NRF2 is supported by the fact that NRF2 gene ablation increased renal oxidative stress and inflammation in the experimental model of diabetes [31].
Moreover, NRF2 knockout mice exhibited more severe kidney injury during ischemic and nephrotoxic insults than wild-type mice [32]. In contrast, pharmacological interventions using NRF2 activators attenuated markers of kidney damage from oxidative stress in various experimental models [33]. In the current study, we found that EA administration reversed the downregulation of NRF2 and subsequent oxidative stress induced by aging in the kidney tissues, suggesting EA potential effect in activating the NRF2 signaling pathway in aged kidneys. In this context, a recent study reported that EA could markedly prevent kidney damage against carbon tetrachloride-induced oxidative stress through the upregulation of NRF2 [16]. Moreover, EA was confirmed to protect against diabetic nephropathy by modulating the transcription and activity of NRF2 [34].
Activation of the NRF2 signaling pathway is the primary mechanism to combat oxidative stress. Although NRF2 activity is mainly controlled by Keap1, other forms of regulation of NRF2 function include acetylation-deacetylation of NRF2 [35]. In the present study, we found that EA administration increased SIRT1-mediated deacetylation of NRF2 in senescent renal tissues. In this regard, a recent investigation reported that SIRT1 exerts its antioxidant activity by promoting nuclear translocation, DNA binding, transcriptional activity, and target genes expression of NRF2 in a deacetylase-dependent manner [13]. Additionally, acetylation of NRF2 was shown to reduce NRF2 stability and impaired antioxidant defenses. Therefore, SIRT1-mediated deacetylation of NRF2 was proposed to increase NRF2 stability and enhance antioxidant gene expression [36, 37]. In line with our results, the administration of resveratrol to aged mice was demonstrated to upregulate the expression of SIRT1 and NRF2, leading to reduced renal oxidative damage and dysfunction [38]. Moreover, activation of SIRT1 in the kidneys of diabetic mice could elevate NRF2 antioxidant signaling and provide remarkable protection against diabetic nephropathy-induced renal oxidative stress [11]. In contrast, the depletion of SIRT1 was associated with a reduction in the transcriptional activity of NRF2, indicating that SIRT1 promoted the activation of NRF2 signaling pathway [39].
Aging is characterized by progressive structural and functional deterioration of kidneys. With aging, many subjects exhibit reductions in glomerular filtration rate (GFR) and renal blood flow (RBF), which occur in concert with a decline in renal mass, tubulointerstitial fibrosis, and increased glomerulosclerosis [2]. Our results revealed impaired renal function in aged rats as indicated by reduced levels of GFR and increased levels of serum urea and creatinine, and albuminuria levels. Histologically, in the present study, senescent renal tissues displayed elevated levels of inflammation, tubular atrophy, and fibrosis, which agree with the findings of a previous study [27]. Increased oxidative stress is considered to be the main pathogenic factor underlying these features and contributing to the elevated oxidative stress; reduced levels of SIRT1 and NRF2 in aged kidneys have been suggested [2, 9]. In this regard, a recent investigation demonstrated that mice with catalytically inactive SIRT1 had a lower glomerular numbers and GFR [40].
Additionally, the diminishment of NRF2 antioxidant capacity aggravates renal tubular apoptosis and atrophy as well as interstitial fibrosis under oxidative stress conditions [41]. However, pharmacological intervention using SIRT1 and NRF2 activator resveratrol improved renal function, proteinuria, glomerulosclerosis, tubular fibrosis, and inflammation in aged kidneys [38, 42]. Here, we showed that EA administration ameliorated age-associated renal dysfunction and histopathological alterations, which is thought to be mediated through SIRT1. Similar findings of improved renal function and histopathological features mediated by SIRT1 were also reported in several rat models of nephrotoxicity [17, 28].
The findings of this study demonstrated that EA has nephroprotective effects through SIRT1/NRF2 pathway in the aging kidneys. However, the reverse experiment was not performed through SIRT1 inhibition to provide more evidence which is one limitation of this study. Therefore, additional experiments using pharmacological inhibitors are required to confirm our conclusion further. Another limitation of our study is that we did not determine NRF2 nuclear translocation and activity, which can be explored in future studies.
## Conclusion
Taken together, our results provided evidence that EA administration upregulated SIRT1 and NRF2 in aged kidneys. SIRT1 activation further promoted NRF2 deacetylation and subsequent activation, leading to increased antioxidant levels in senescent renal tissues. Moreover, treatment with EA improved renal function and ameliorated histopathological features in aged kidneys. These findings provided an experimental basis for the application of EA to delay the process of aging in kidneys. However, additional clinical studies for EA safety are required.
## Supplementary Information
Additional file 1.
## References
1. Fang Y, Gong AY, Haller ST, Dworkin LD, Liu Z, Gong R. **The ageing kidney: molecular mechanisms and clinical implications**. *Ageing Res Rev* (2020.0) **63** 101151. DOI: 10.1016/j.arr.2020.101151
2. O'Sullivan ED, Hughes J, Ferenbach DA. **Renal aging: causes and consequences**. *J Am Soc Nephrol* (2017.0) **28** 407-420. DOI: 10.1681/ASN.2015121308
3. Schmitt R, Melk A. **Molecular mechanisms of renal aging**. *Kidney Int* (2017.0) **92** 569-579. DOI: 10.1016/j.kint.2017.02.036
4. Luo J, Mills K, le Cessie S, Noordam R, van Heemst D. **Ageing, age-related diseases and oxidative stress: what to do next?**. *Ageing Res Rev* (2020.0) **57** 100982. DOI: 10.1016/j.arr.2019.100982
5. Covarrubias AJ, Perrone R, Grozio A, Verdin E. **NAD(+) metabolism and its roles in cellular processes during ageing**. *Nat Rev Mol Cell Biol* (2021.0) **22** 119-141. DOI: 10.1038/s41580-020-00313-x
6. Kitada M, Kume S, Takeda-Watanabe A, Kanasaki K, Koya D. **Sirtuins and renal diseases: relationship with aging and diabetic nephropathy**. *Clin Sci (Lond)* (2013.0) **124** 153-164. DOI: 10.1042/CS20120190
7. Morigi M, Perico L, Benigni A. **Sirtuins in renal health and disease**. *J Am Soc Nephrol* (2018.0) **29** 1799-1809. DOI: 10.1681/ASN.2017111218
8. Kong L, Wu H, Zhou W, Luo M, Tan Y, Miao L, Cai L. **Sirtuin 1: a target for kidney diseases**. *Mol Med (Cambridge, Mass)* (2015.0) **21** 87-97. DOI: 10.2119/molmed.2014.00211
9. Schmidlin CJ, Dodson MB, Zhang DD. **Filtering through the role of NRF2 in kidney disease**. *Arch Pharm Res* (2020.0) **43** 361-369. DOI: 10.1007/s12272-019-01177-2
10. 10.Ucar BI, Ucar G, Saha S, Buttari B, Profumo E, Saso L. Pharmacological protection against ischemia-reperfusion injury by regulating the Nrf2-Keap1-ARE signaling pathway. Antioxidants (Basel, Switzerland). 2021;10 (6).
11. Ma F, Wu J, Jiang Z, Huang W, Jia Y, Sun W, Wu H. **P53/NRF2 mediates SIRT1's protective effect on diabetic nephropathy**. *BBA Mol Cell Res* (2019.0) **1866** 1272-1281
12. Kratz EM, Sołkiewicz K, Kubis-Kubiak A, Piwowar A. **Sirtuins as important factors in pathological states and the role of their molecular activity modulators**. *Int J Mol Sci* (2021.0) **22** 630. DOI: 10.3390/ijms22020630
13. Huang K, Huang J, Xie X, Wang S, Chen C, Shen X, Liu P, Huang H. **Sirt1 resists advanced glycation end products-induced expressions of fibronectin and TGF-β1 by activating the Nrf2/ARE pathway in glomerular mesangial cells**. *Free Radic Biol Med* (2013.0) **65** 528-540. DOI: 10.1016/j.freeradbiomed.2013.07.029
14. Ríos JL, Giner RM, Marín M, Recio MC. **A pharmacological update of ellagic acid**. *Planta Med* (2018.0) **84** 1068-1093. DOI: 10.1055/a-0633-9492
15. Hseu YC, Chou CW, Senthil Kumar KJ, Fu KT, Wang HM, Hsu LS, Kuo YH, Wu CR, Chen SC, Yang HL. **Ellagic acid protects human keratinocyte (HaCaT) cells against UVA-induced oxidative stress and apoptosis through the upregulation of the HO-1 and Nrf-2 antioxidant genes**. *Food Chem Toxicol* (2012.0) **50** 1245-1255. DOI: 10.1016/j.fct.2012.02.020
16. Aslan A, Gok O, Beyaz S, Ağca CA, Erman O, Zerek A. **Ellagic acid prevents kidney injury and oxidative damage via regulation of Nrf-2/NF-κB signaling in carbon tetrachloride induced rats**. *Mol Biol Rep* (2020.0) **47** 7959-7970. DOI: 10.1007/s11033-020-05873-x
17. Mohammed ET, Hashem KS, Abdelazem AZ, Foda F. **Prospective protective effect of ellagic acid as a SIRT1 activator in iron oxide nanoparticle-induced renal damage in rats**. *Biol Trace Elem Res* (2020.0) **198** 177-188. DOI: 10.1007/s12011-020-02034-w
18. Mehrzadi S, Fatemi I, Malayeri AR, Khodadadi A, Mohammadi F, Mansouri E, Rashno M, Goudarzi M. **Ellagic acid mitigates sodium arsenite-induced renal and hepatic toxicity in male Wistar rats**. *Pharmacol Rep* (2018.0) **70** 712-719. DOI: 10.1016/j.pharep.2018.02.007
19. Mehrzadi S, Mombeini MA, Fatemi I, Kalantari H, Kalantar M, Goudarzi M. **The effect of ellagic acid on renal injury associated with acrylamide in experimental rats**. *Physiol Pharmacol* (2022.0) **26** 313-321
20. Bahri F, Khaksari M, Movahedinia S, Shafiei B, Rajizadeh MA, Nazari-Robati M. **Improving SIRT1 by trehalose supplementation reduces oxidative stress, inflammation, and histopathological scores in the kidney of aged rats**. *J Food Biochem* (2021.0) **45** e13931. DOI: 10.1111/jfbc.13931
21. Daei S, Ziamajidi N, Abbasalipourkabir R, Aminzadeh Z, Vahabirad M. **Silver nanoparticles exert apoptotic activity in bladder cancer 5637 cells through alteration of Bax/Bcl-2 genes expression**. *Chonnam Med J* (2022.0) **58** 102-109. DOI: 10.4068/cmj.2022.58.3.102
22. Kitada M, Koya D. **Renal protective effects of resveratrol**. *Oxi Med Cell Longev* (2013.0) **2013** 568093
23. 23.Cui H, Kong Y, Zhang H. Oxidative stress, mitochondrial dysfunction, and aging. J Signal Transduct. 2012:646354.
24. Hwang JW, Yao H, Caito S, Sundar IK, Rahman I. **Redox regulation of SIRT1 in inflammation and cellular senescence**. *Free Radic Biol Med* (2013.0) **61** 95-110. DOI: 10.1016/j.freeradbiomed.2013.03.015
25. Hong YA, Kim JE, Jo M, Ko G-J. **The role of sirtuins in kidney diseases**. *Int J Mol Sci* (2020.0) **21** 6686. DOI: 10.3390/ijms21186686
26. Kwon Y, Kim J, Lee C-Y, Kim H. **Expression of SIRT1 and SIRT3 varies according to age in mice**. *Anat Cell Biol* (2015.0) **48** 54-61. DOI: 10.5115/acb.2015.48.1.54
27. Lim JH, Kim EN, Kim MY, Chung S, Shin SJ, Kim HW, Yang CW, Kim Y-S, Chang YS, Park CW, Choi BS. **Age-associated molecular changes in the kidney in aged mice**. *Oxi Med Cell Longev* (2012.0) **2012** 171383
28. Neamatallah T, Eid BG, Bagher AM, Nasrullah MZ, Ali SS, Mohammedsaleh ZM, El-Shitany NA. **Activation of sirtuin 1 by ellagic acid ameliorates cisplatin-induced apoptosis**. *J Pharm Res Int* (2021.0) **33** 21-32. DOI: 10.9734/jpri/2021/v33i1531283
29. Schmidlin CJ, Dodson MB, Madhavan L, Zhang DD. **Redox regulation by NRF2 in aging and disease**. *Free Radic Biol Med* (2019.0) **134** 702-707. DOI: 10.1016/j.freeradbiomed.2019.01.016
30. Zhang G, Kang Y, Zhou C, Cui R, Jia M, Hu S, Ji X, Yuan J, Cui H, Shi G. **Amelioratory effects of testosterone propionate on age-related renal fibrosis via suppression of TGF-β1/Smad signaling and activation of Nrf2-ARE signaling**. *Sci Rep* (2018.0) **8** 10726. DOI: 10.1038/s41598-018-29023-3
31. Ruiz S, Pergola PE, Zager RA, Vaziri ND. **Targeting the transcription factor Nrf2 to ameliorate oxidative stress and inflammation in chronic kidney disease**. *Kidney Int* (2013.0) **83** 1029-1041. DOI: 10.1038/ki.2012.439
32. Liu M, Grigoryev DN, Crow MT, Haas M, Yamamoto M, Reddy SP, Rabb H. **Transcription factor Nrf2 is protective during ischemic and nephrotoxic acute kidney injury in mice**. *Kidney Int* (2009.0) **76** 277-285. DOI: 10.1038/ki.2009.157
33. Choi BH, Kang KS, Kwak MK. **Effect of redox modulating NRF2 activators on chronic kidney disease**. *Molecules* (2014.0) **19** 12727-12759. DOI: 10.3390/molecules190812727
34. Altamimi JZ, AlFaris NA, Alshammari GM, Alagal RI, Aljabryn DH, Aldera H, Alrfaei BM, Alkhateeb MA, Yahya MA. **Ellagic acid protects against diabetic nephropathy in rats by regulating the transcription and activity of Nrf2**. *J Funct Foods* (2021.0) **79** 104397. DOI: 10.1016/j.jff.2021.104397
35. 35.Guerrero-Hue M, Rayego-Mateos S, Vázquez-Carballo C, Palomino-Antolín A, García-Caballero C, Opazo-Rios L, Morgado-Pascual JL, Herencia C, Mas S, Ortiz A, Rubio-Navarro A, Egea J, Villalba JM, Egido J, Moreno JA. Protective role of Nrf2 in renal disease. Antioxidants (Basel). 2020;10 (1).
36. Mercado N, Thimmulappa R, Thomas CM, Fenwick PS, Chana KK, Donnelly LE, Biswal S, Ito K, Barnes PJ. **Decreased histone deacetylase 2 impairs Nrf2 activation by oxidative stress**. *Biochem Biophys Res Commun* (2011.0) **406** 292-298. DOI: 10.1016/j.bbrc.2011.02.035
37. Ding YW, Zhao GJ, Li XL, Hong GL, Li MF, Qiu QM, Wu B, Lu ZQ. **SIRT1 exerts protective effects against paraquat-induced injury in mouse type II alveolar epithelial cells by deacetylating NRF2 in vitro**. *Int J Mol Med* (2016.0) **37** 1049-1058. DOI: 10.3892/ijmm.2016.2503
38. Kim EN, Lim JH, Kim MY, Ban TH, Jang I-A, Yoon HE, Park CW, Chang YS, Choi BS. **Resveratrol, an Nrf2 activator, ameliorates aging-related progressive renal injury**. *Aging (Albany NY)* (2018.0) **10** 83-99. DOI: 10.18632/aging.101361
39. Huang K, Chen C, Hao J, Huang J, Wang S, Liu P, Huang H. **Polydatin promotes Nrf2-ARE anti-oxidative pathway through activating Sirt1 to resist AGEs-induced upregulation of fibronetin and transforming growth factor-β1 in rat glomerular messangial cells**. *Mol Cell Endocrinol* (2015.0) **399** 178-189. DOI: 10.1016/j.mce.2014.08.014
40. Bellin AR, Zhang Y, Thai K, Rosenblum ND, Cullen-McEwen LA, Bertram JF, Gilbert RE. **Impaired SIRT1 activity leads to diminution in glomerular endowment without accelerating age-associated GFR decline**. *Physiol Rep* (2019.0) **7** e14044. DOI: 10.14814/phy2.14044
41. Wu C-F, Liu C-C, Tsai Y-C, Chen C-C, Wu M-T, Hsieh T-J. **Diminishment of Nrf2 antioxidative defense aggravates nephrotoxicity of melamine and oxalate coexposure**. *Antioxidants (Basel, Switzerland)* (2021.0) **10** 1464. PMID: 34573096
42. Jang I-A, Kim EN, Lim JH, Kim MY, Ban TH, Yoon HE, Park CW, Chang YS. **Choi BS Effects of resveratrol on the renin-angiotensin system in the aging kidney**. *Nutrients* (2018.0) **10** 1741. DOI: 10.3390/nu10111741
|
---
title: Adjusting phosphate feeding regimen according to daily rhythm increases eggshell
quality via enhancing medullary bone remodeling in laying hens
authors:
- Jiakun Yan
- Jiajie Wang
- Jie Chen
- Hao Shi
- Xujie Liao
- Chong Pan
- Yanli Liu
- Xin Yang
- Zhouzheng Ren
- Xiaojun Yang
journal: Journal of Animal Science and Biotechnology
year: 2023
pmcid: PMC9999492
doi: 10.1186/s40104-023-00829-0
license: CC BY 4.0
---
# Adjusting phosphate feeding regimen according to daily rhythm increases eggshell quality via enhancing medullary bone remodeling in laying hens
## Abstract
### Background
Body phosphorus metabolism exhibits a circadian rhythm over the 24-h daily cycle. The egg laying behavior makes laying hens a very special model for investigating phosphorus circadian rhythms. There is lack of information about the impact of adjusting phosphate feeding regimen according to daily rhythm on the phosphorus homeostasis and bone remodeling of laying hens.
### Methods and results
Two experiments were conducted. In Exp. 1, Hy-Line Brown laying hens ($$n = 45$$) were sampled according the oviposition cycle (at 0, 6, 12, and 18 h post-oviposition, and at the next oviposition, respectively; $$n = 9$$ at each time point). Diurnal rhythms of body calcium/phosphorus ingestions and excretions, serum calcium/phosphorus levels, oviduct uterus calcium transporter expressions, and medullary bone (MB) remodeling were illustrated. In Exp. 2, two diets with different phosphorus levels ($0.32\%$ and $0.14\%$ non-phytate phosphorus (NPP), respectively) were alternately presented to the laying hens. Briefly, four phosphorus feeding regimens in total (each included 6 replicates of 5 hens): [1] fed $0.32\%$ NPP at both 09:00 and 17:00; [2] fed $0.32\%$ NPP at 09:00 and $0.14\%$ NPP at 17:00; [3] fed $0.14\%$ NPP at 09:00 and $0.32\%$ NPP at 17:00; [4] fed $0.14\%$ NPP at both 09:00 and 17:00. As a result, the regimen fed $0.14\%$ NPP at 09:00 and $0.32\%$ NPP at 17:00, which was designed to strengthen intrinsic phosphate circadian rhythms according to the findings in Exp. 1, enhanced ($P \leq 0.05$) MB remodeling (indicated by histological images, serum markers and bone mineralization gene expressions), elevated ($P \leq 0.05$) oviduct uterus calcium transportation (indicated by transient receptor potential vanilloid 6 protein expression), and subsequently increased ($P \leq 0.05$) eggshell thickness, eggshell strength, egg specific gravity and eggshell index in laying hens.
### Conclusions
These results underscore the importance of manipulating the sequence of daily phosphorus ingestion, instead of simply controlling dietary phosphate concentrations, in modifying the bone remodeling process. Body phosphorus rhythms will need to be maintained during the daily eggshell calcification cycle.
### Supplementary Information
The online version contains supplementary material available at 10.1186/s40104-023-00829-0.
## Introduction
Phosphate metabolism is involved in a variety of biologic processes [1] and its influx-storage-efflux balance is essential for life [2]. The two extremes of phosphate homeostasis (deficiency and excess) have been inextricably linked to tissue mineralization disorders including but not limit to osteopenia [3], osteoporosis [4], nephrocalcinosis [5] and vascular calcification [6]. Generally, dietary phosphate regimen must be precisely managed to maintain the balance of body phosphate homeostasis when problems occur during the intestinal absorption (phosphate influx) and/or bone mineralization (phosphate storage) and/or renal resorption (phosphate influx) processes [7]. In current practice, dietary phosphate regimen interventions are mostly conducted in two approaches: [1] direct controlling the concentrations of phosphate in ingested food [8]; [2] using phosphate binders (e.g., sevelamer carbonate) to remove available phosphate [9, 10] or using phosphate utilization promoters (e.g., calcitriol) to increase available phosphate [10, 11]. However, the effectiveness of these approaches has been questioned since the body exhibits significant diurnal variations in serum phosphate concentrations [12, 13], and a simple dietary intervention (which ignores the diurnal variations) may not enough for supporting a normal phosphate homeostasis [12, 13].
The phosphate metabolism exhibits a circadian rhythm over the 24-h daily cycle [14, 15]. Circadian rhythm disturbance of intestinal, renal and skeletal functions could induce abnormal phosphate rhythm and subsequently cause tissue mineralization disorders [16, 17]. For example, chronic kidney disease-mineral bone disorder, which exhibits a decreased serum phosphate circadian rhythm (amplitude decreases and phase shift) [18, 19], is often accompanied by circadian rhythm disturbances in the bone [20] and kidney [16]. In this case, simply increasing/decreasing dietary phosphate concentrations, could only overall increase/decrease serum phosphate concentrations, but could not normalize the circadian rhythm of serum phosphate [18]. These results indicating that the daily rhythms of body phosphate metabolism will need to be carefully maintained when developing dietary phosphate intervention strategies in humans and animals.
The egg laying behavior makes laying hens a very special model for investigating phosphate circadian rhythms [21]. Especially, the diurnal eggshell formation process in laying hens was support by a strong circadian rhythm in the remodeling of medullary bone (MB), which is unique to avian and dinosaurs and could be quickly absorbed and reformed during the 24-h egg-laying cycle [22]. It is well documented that circulating level of phosphate is highly dynamic during the MB cycle [21], and alterations in dietary phosphate levels could influence the remodeling of MB [23]. In aged laying hens, the remodeling efficiency of MB decreases and the frequency of cracked eggs increases (unqualified eggshell formation) [24]. In the field, attempts have been made to evaluate the most appropriate dietary phosphate levels for maintaining MB remodeling and increasing eggshell quality in aged laying hens [25]. However, in conventional feeding systems, laying hens are provided with a single diet with a constant phosphate level throughout the entire day without considering body phosphate circadian rhythms.
So, we hypothesized that a daily dynamic phosphate feeding regimen that enhancing body phosphate circadian rhythms would optimize the remodeling process of MB and thereby improve eggshell quality in laying hens. To test this hypothesis, we illustrated the circadian rhythms of phosphate metabolism and diets with different phosphate levels were alternately presented to the laying hens to meet the specific phosphate requirement during different stages of the daily egg-laying cycle. Our objectives were to reveal the multi-organ interactions on phosphate metabolism rhythms and to develop a simple daily phosphate regimen for improving eggshell quality in aged laying hens.
## Experimental animals, dosage regimen and sample collection
The animals used (Hy-Line Brown laying hens) were all purchased from Julong Poultry Farm (Wugong, Shaanxi, China), and were individually housed in cages with raised wire floors (depth × width × height = 45 cm × 35 cm × 45 cm) at the Animal Nutrition & Healthy Feeding Research Laboratory (Northwest A&F University, Yangling, China). A photoperiod of 16-h-light:8-h-dark was applied (lights-on, 05:30; lights-off, 21:30). The hens were fed twice daily (09:00 and 17:00).
## Exp. 1
Hy-Line Brown laying hens were fed with a regular diet (corn-soybean meal-based; containing $0.32\%$ non-phytate phosphorus (NPP); Table 1) start from 35 weeks of age. On the last day of age 40 weeks, a total of 60 hens that laid eggs between 07:30−08:30 were randomly selected to evaluate the daily phosphorus rhythms. Of them, 45 hens were euthanized for sample collection, and the other 15 hens were used to study the feed intake and calcium/phosphorus excretion rhythms. For sample collection, the 45 hens were sampled according the oviposition cycle: at oviposition, at 6, 12, 18 h post-oviposition, and at the next oviposition, respectively, with 9 hens sampled at each of the time point. The following samples were collected: blood (for serum), uterine (stored at −80 ℃, for Western-blotting analysis), femur (in $4\%$ paraformaldehyde, for histological analysis) and kidney (stored at −80 ℃, for Western-blotting analysis). For the other 15 hens, the feed intake was recoded and the excreta was collected at the following intervals: from oviposition to 6 h post-oviposition, from 7 to 12 h post-oviposition, from 13 to 18 h post-oviposition, from 19 h post-oviposition to the next oviposition. Table 1Composition and nutrient concentrations of basal diet (%, unless noted, as-is basis)ItemLow phosphorusRegular phosphorusIngredients Corn56.6956.69 Soybean meal25.7725.77 Distillers dried grains with solubles4.004.00 Calcium carbonate9.739.04 Dicalcium phosphate-1.15 Soybean oil1.511.51 Sodium chloride0.260.26 DL-Methionine0.180.18 Choline chloride0.150.15 Montmorillonite0.710.25 Premix111 In total100.00100.00Nutrient levels Metabolizable energy, kcal/kg (calculated)2,6002,600 Crude protein (calculated)16.516.5 *Total phosphorus* (calculated/analyzed)$\frac{0.34}{0.340.53}$/0.49 Non-phytate phosphorus (calculated)0.140.32 Calcium (calculated/analyzed)$\frac{3.50}{3.473.50}$/3.521Provided per kilogram of diet: manganese 60 mg, copper 8 mg, zinc 80 mg, iodine 0.35 mg, selenium 0.3 mg, vitamin A 8000 IU, vitamin E 30 mg, vitamin K3 1.5 mg, thiamine 4 mg, riboflavin 13 mg, pantothenic acid 15 mg, nicotinamide 20 mg, pyridoxine 6 mg, biotin 0.15 mg, folic acid 1.5 mg, and cobalamin 0.02 mg
## Exp. 2
At the age of 70 weeks, a total of 120 hens were randomly selected to evaluate dietary interventions of body phosphorus rhythms. The hens were fed with 4 phosphorus regimens: [1] RR, provided with regular phosphorus diet at both 09:00 and 17:00 (conventional feeding without considering daily rhythms of body phosphorus metabolism); [2] RL, provided with regular phosphorus diet at 09:00 and low phosphorus diet at 17:00 (dynamic feeding converse to the body phosphorus rhythms found in Exp. 1); [3] LR, provided with low phosphorus diet at 09:00 and regular phosphorus diet at 17:00 (dynamic feeding consistent with the body phosphorus rhythms found in Exp. 1); [4] LL, provided with low phosphorus diet at both 09:00 and 17:00 (direct restriction without considering daily rhythms of body phosphorus metabolism). Each feeding regimens included 6 replicates, and each replicate contained 5 hens. The regular and the low phosphorus diet contained $0.32\%$ and $0.14\%$ NPP, respectively. The feeding trial lasted for 12 weeks (according to the literature, changes in eggshell and bone mineralization status could be observed in 8 to 12 weeks after dietary phosphorus interventions in laying hens) [26, 27]. On the last 3 d of the feeding trial, all the eggs were collected for egg quality analysis. On the last day of the feeding trial, two egg-laying hens were randomly selected from each replicate (sampled at 6 and 18 h post-oviposition, respectively). The following samples were collected: blood (for serum), uterine (stored at −80 ℃, for Western-blotting analysis), femur (left side, stored in $4\%$ paraformaldehyde for histological analysis; right side, stored at −80 ℃ for the determination of mineralization status and gene expressions).
## Serum biochemical assay
Blood samples (3 mL, from wing veins) were clotted at 37 ℃ for 60 min in water bath and centrifuged (594 g, 15 min) for serum samples (stored at −80 ℃). Serum levels of calcium (catalogue no. C004-2), phosphorus (catalogue no. C006-1) and tartrate-resistant acid phosphatase (TRAP; catalogue no. A058-1) were analyzed using commercial kits purchased from Nanjing Jiancheng Bioengineering Institute (Nanjing, Jiangsu, China). Serum levels of C-terminal telopeptide of type I collagen (CTX-I) was analyzed using a commercial kit (catalogue no. ml060903) purchased from Shanghai Enzyme-linked Biotechnology Co., Ltd (Shanghai, China). The spectrophotometric reactions were detected using a Synergy HT plate reader (BioTek, Winooski, VT, USA; for calcium and CTX-I analysis) or a UV-1800 spectrophotometer (Shimadzu, Kyoto, Japan; for phosphorus and TARP analysis).
## Determination of calcium and phosphorus contents
The excreta samples were oven dried, air equilibrated, and ground before analysis. The femur samples were cut at $25\%$ and $75\%$ from the proximal femur of the length of the bone to separate mid-diaphysis ($50\%$), then this part was oven dried, defatted, cut longitudinally, and MB was removed by scraping with aid of scalpel. The pretreated samples were ashed using a muffle furnace (550 ℃; 6 h) and dissolved in hydrochloric acid. Calcium and phosphorus contents of the samples were determined using the ethylenediaminetetraacetic acid method and the ammonium-vanadium-molybdate method, respectively [28]. The results were calculated based on oven-dried-basis for excreta and ash-basis for femur samples.
## Histological analysis
The paraformaldehyde fixed femur samples were decalcified (6 weeks; in $10\%$ EDTA phosphate buffer solution), dehydrated (in ethyl alcohol), hyalinized (in xylene), and the proximal diaphysis was embedded in paraffin wax. Sample sections (5 μm thick; mounted on glass slides) were stained with toluidine blue and scanned using an optical microscope (BX46; Olympus, Japan). Three photomicrographs (40× magnification) were taken from each sample. The area percentage of MB is expressed as the percentage of MB in the region of the marrow cavity where the MB is located (ImageJ 1.8.0. software).
## Western-blotting analysis
The protein isolation and western blotting procedures were conducted as previously described [29]. Primary antibodies (rabbit) to type 2a sodium-phosphate co-transporter (NPt2a, catalogue no. A9460), type III sodium-dependent phosphate transporter 1 (PiT1, catalogue no. A4117), type III sodium-dependent phosphate transporter 2 (PiT2, catalogue no. A6739), transient receptor potential vanilloid 6 (TRPV6, catalogue no. A16128), and calbindin D‐28k (CaBP-D28k, catalogue no. A0802) were purchased from ABclonal Technology (Wuhan, Hubei, China). Primary antibody (mouse) to β-actin (ACTB, catalogue no. CW0096) was purchased from CWBIO Co., Ltd. (Beijing, China). The secondary goat anti-rabbit IgG (catalogue no. DY60202) was purchased from Diyi Biotechnology Co., Ltd. (Shanghai, China) and the secondary goat anti-mouse IgG (catalogue no. bs-0296G-HRP) was purchased from Bioss Biotechnology Co., Ltd. (Beijing, China). The protein bands were visualized with a DNR imaging system (Micro Chemi, Israel) and the blot density was normalized to ACTB.
## Quantitative real-time PCR
The quantitative real‐time RCR analysis was performed as previously described [30]. The sequences (Table S1) of primers used in quantitative real-time PCR analysis were designed with the Primer3 program. All reactions were run in triplicate. Relative mRNA expressions were calculated using the chicken ACTB (β-actin) gene as an internal reference (2−ΔΔCt method).
## Statistical analysis
Data analysis was performed using SPSS version 23.0 (IBM Corp., Chicago, IL, USA). The individual laying hen was considered as the statistical unit. Two-tailed Students’ t-tests was conducted for the comparisons between two groups. One-way ANOVA, followed by Duncan’s multiple-range post hoc test, was conducted to determine the differences among multiple groups. The results are presented as means and standard error of the mean (SEM). Statistical significance was considered at $P \leq 0.05.$
## Diurnal rhythms of serum calcium and phosphorus levels and uterine protein expressions in laying hens
Serum calcium level of the laying hens was increased after oviposition and peaked at 6 h post-oviposition (Fig. 1A, $P \leq 0.05$). Then, the serum calcium level was gradually decreased to its lowest level at 18 h post-oviposition ($P \leq 0.05$). Thereafter, increased at the next oviposition ($P \leq 0.05$). Serum phosphorus level (Fig. 1B) and uterine CaBP-D28k expression (Fig. 1C) of the laying hens were gradually increased after oviposition ($P \leq 0.05$), peaked at 18 h post-oviposition, and then deceased at the next oviposition ($P \leq 0.05$). Uterine expression of TRPV6 was higher at 12 h and 18 h post-oviposition (the period of eggshell fast deposition) when compared to the other time points during the egg laying cycle (Fig. 1C, $P \leq 0.05$).Fig. 1Diurnal rhythms of serum calcium and phosphorus levels and uterine protein expressions in 40-week-old Hy-Line Brown laying hens during the egg laying cycle. A) *Serum calcium* ($$n = 9$$ per group); B) serum phosphorus ($$n = 9$$ per group); C) protein expression of TRPV6 and CaBP-D28k ($$n = 4$$ per group). White and black bars represent the light and dark. Data are presented as the mean ± SEM. $P \leq 0.05$ by one-way ANOVA followed by Duncan's multiple range tests and the letters (a–d) indicate significant differences among all treatment groups ($P \leq 0.05$). ACTB, β-actin; CaBP-D28k, Calbindin D‐28k; TRPV6, Transient receptor potential vanilloid 6
## Diurnal rhythms of MB remodeling in laying hens
Serum concentrations of bone resorption markers TRAP (Fig. 2A) and CTX-I (Fig. 2B) were significantly changed ($P \leq 0.05$) during the daily egg laying cycle, and peaked at 12 h and 18 h post-oviposition (the period of eggshell fast deposition), respectively. The area percentage of MB was higher ($P \leq 0.05$) at 6 h and 12 h post-oviposition when compared to the other time points during the egg laying cycle (Fig. 2C, D).Fig. 2Diurnal rhythms of MB remodeling in 40-week-old Hy-Line Brown laying hens. A) Serum TARP ($$n = 9$$ per group); B) serum CTX-I ($$n = 9$$ per group); C) area percentage of MB analyzed by toluidine blue staining ($$n = 3$$ per group); D) representative images of toluidine blue staining of femur transverse sections, bar represents 100 µm. White and black bars represent the light and dark. Data are presented as the mean ± SEM. $P \leq 0.05$ by one-way ANOVA followed by Duncan's multiple range tests and the letters (a–c) indicate significant differences among all treatment groups ($P \leq 0.05$). CTX-I, C-terminal telopeptide of type I collagen; MB, medullary bone; TRAP, tartrate-resistant acid phosphatase
## Diurnal rhythms of the intake and excretion of calcium and phosphorus in laying hens
Calcium and phosphorus intake of the laying hens were majorly happened during the lighting period (Fig. 3A, B). The excreta collected during 0 to 6 h post-oviposition had increased ($P \leq 0.05$) calcium concentration when compared to those collected from the other period during the egg laying cycle (Fig. 3C). The excreta collected during 7 to 12 h post-oviposition had increased ($P \leq 0.05$) calcium concentration when compared to those collected during 13 to 18 h post-oviposition. Accordingly, body calcium excretion was the highest during 0 to 6 h post-oviposition and was the lowest during 13 to 18 h post-oviposition (Fig. 3D, $P \leq 0.05$). Body phosphorus excretion was majorly happened during the lighting period (Fig. 3E). The excreta collected during 13 to 18 h post-oviposition had the highest phosphorus concentration and the excreta collected during 0 to 6 h post-oviposition had the lowest phosphorus concentration (Fig. 3F, $P \leq 0.05$). Kidney expression of NPt2a of the laying hens was gradually increased after oviposition (Fig. 3G, $P \leq 0.05$), peaked at 12 h post-oviposition, and then deceased all the way down until the next oviposition ($P \leq 0.05$).Fig. 3Diurnal rhythms of the intake and excretion of calcium and phosphorus in 40-week-old Hy-Line Brown laying hens. A) Calcium intake ($$n = 15$$ per group); B) phosphorus intake ($$n = 15$$ per group); C) total excretion of calcium ($$n = 15$$ per group); D) dry excrete concretion of calcium ($$n = 15$$ per group); E) total excretion of phosphorus ($$n = 15$$ per group); F) dry excrete concretion of phosphorus ($$n = 15$$ per group); G) representative western blots and statistical analysis of protein abundances of ACTB, PiT1, PiT2 and NPt2a in the kidney ($$n = 4$$ per group), all samples were normalized to their respective ACTB levels of each sample. White and black bars represent the light and dark. Data are presented as the mean ± SEM. $P \leq 0.05$ by one-way ANOVA followed by Duncan's multiple range tests and the letters (a–d) indicate significant differences among all treatment groups ($P \leq 0.05$). ACTB, β-actin; NPt2a, Type 2a sodium-phosphate co-transporters; PiT1, Type III sodium-dependent phosphate transporter 1; PiT2, Type III sodium-dependent phosphate transporter 2
## Daily dynamic phosphorus feeding regimen increased uterine calcium transportation and eggshell quality in laying hens
Laying hens on the RL and LR phosphorus feeding regimen had increased eggshell thickness ($P \leq 0.05$) when compared to those on the regimens of RR and LL (Fig. 4A). Laying hens on the LR phosphorus feeding regimen had: [1] increased ($P \leq 0.05$) eggshell strength, egg specific gravity and eggshell index when compared to those on the regimens of RR and LL (Fig. 4B−D); [2] increased ($P \leq 0.05$) uterine TRPV6 expression when compared to those on the regimens of RR and RL (Fig. 4E). Laying hens on the RL phosphorus feeding regimen had: [1] increased ($P \leq 0.05$) eggshell strength, egg specific gravity and eggshell index when compared to those on the LL regimen (Fig. 4B−D); [2] decreased ($P \leq 0.05$) uterine TRPV6 expression when compared to those on the regimens of LL (Fig. 4E).Fig. 4Daily dynamic phosphorus feeding regimen increased uterine calcium transportation and eggshell quality in 70-week-old Hy-Line Brown laying hens for 12 weeks. A) Eggshell thickness ($$n = 62$$−66 per group); B) eggshell strength ($$n = 62$$−66 per group); C) egg specific gravity ($$n = 61$$−66 per group); D) shell index ($$n = 62$$−63 per group); D) western blot analysis and statistical analysis of protein abundances of ACTB, TRPV6 and CaBP-D28k in the uterus collected from 18 h post-oviposition ($$n = 3$$ per group), all samples were normalized to their respective ACTB levels of each sample. Data are presented as the mean ± SEM. $P \leq 0.05$ by one-way ANOVA followed by Duncan's multiple range tests and the letters (a−c) indicate significant differences among all treatment groups ($P \leq 0.05$). RR, provided with regular phosphorus diet at both 09:00 and 17:00 for 12 weeks; RL, provided with regular phosphorus diet at 09:00 and low phosphorus diet at 17:00 for 12 weeks; LR, provided with low phosphorus diet at 09:00 and regular phosphorus diet at 17:00 for 12 weeks; LL, provided with low phosphorus diet at both 09:00 and 17:00 for 12 weeks. ACTB, β-actin; CaBP-D28k, Calbindin D‐28k; TRPV6, Transient receptor potential vanilloid 6
## Daily dynamic phosphorus feeding regimen improved MB remodeling in laying hens
At 6 h post-oviposition: [1] laying hens on the LR phosphorus feeding regimen had increased serum calcium (Fig. 5A, $P \leq 0.05$) when compared to those on the RR, RL and LL regimens; [2] laying hens on the RR phosphorus feeding regimen had increased serum phosphorus (Fig. 5B, $P \leq 0.05$) when compared to those on the LL regimen. At 18 h post-oviposition: [1] laying hens on the RR and RL phosphorus feeding regimens had increased MB calcium content (Fig. 5C, $P \leq 0.05$) when compared to those on LL regimen; [2] laying hens on the RL phosphorus feeding regimen had increased MB percentage (Fig. 5F, G, $P \leq 0.05$) when compared to those on the LR and LL regimens; [3] laying hens on the LR phosphorus feeding regimen had increased serum phosphorus (Fig. 5B, $P \leq 0.05$) when compared to those on the RR, RL and LL regimens. Within each feeding regimen: [1] laying hens on the RL phosphorus feeding regimen had increased serum phosphorus (Fig. 5B, $P \leq 0.05$) at 18 h post-oviposition when compared to 6 h post-oviposition; [2] laying hens on the LR phosphorus feeding regimen had decreased serum calcium (Fig. 5A, $P \leq 0.05$), MB calcium content (Fig. 5C, $P \leq 0.05$), MB calcium/phosphorus ratio (Fig. 5E, $P \leq 0.05$), and MB percentage (Fig. 5F, G, $P \leq 0.05$), and increased serum phosphorus (Fig. 5B, $P \leq 0.05$) at 18 h post-oviposition when compared to 6 h post-oviposition; [3] laying hens on the LL phosphorus feeding regimen had increased serum phosphorus (Fig. 5B, $P \leq 0.05$), and decreased MB calcium content (Fig. 5C, $P \leq 0.05$) and MB calcium/phosphorus ratio (Fig. 5E, $P \leq 0.05$) at 18 h post-oviposition when compared to 6 h post-oviposition. No difference was observed among treatments on MB phosphorus content (Fig. 5D, $P \leq 0.05$).Fig. 5Daily dynamic phosphorus feeding regimen improved MB remodeling in 70-week-old Hy-Line Brown laying hens for 12 weeks. A) serum calcium ($$n = 6$$ per group); B) serum phosphorus ($$n = 6$$ per group); C) calcium content of MB ($$n = 6$$ per group); D) phosphorus content of MB ($$n = 6$$ per group); E) calcium/phosphorus ratio of MB ($$n = 6$$ per group); F) area percentage of MB analyzed by toluidine blue staining ($$n = 3$$ per group); G) representative images of toluidine blue staining of femur transverse sections, bar represents 100 µm. Data are presented as the mean ± SEM. $P \leq 0.05$ by one-way ANOVA followed by Duncan's multiple range tests, the letters (A, B) and the letters (a, b) indicate significant differences among all treatment groups chosen from 6 h post-oviposition and 18 h post-oviposition, respectively ($P \leq 0.05$). The significance of difference between 6 h and 18 h post-oviposition for each group was analyzed using two-tailed students’ t-tests, *$P \leq 0.05$, **$P \leq 0.01$, ***$P \leq 0.001.$ RR, provided with regular phosphorus diet at both 09:00 and 17:00 for 12 weeks; RL, provided with regular phosphorus diet at 09:00 and low phosphorus diet at 17:00 for 12 weeks; LR, provided with low phosphorus diet at 09:00 and regular phosphorus diet at 17:00 for 12 weeks; LL, provided with low phosphorus diet at both 09:00 and 17:00 for 12 weeks; MB, medullary bone At 6 h post-oviposition, laying hens on the LR phosphorus feeding regimen had: [1] increased MB alkaline phosphatase (ALPL) mRNA expression (Fig. 6A, $P \leq 0.05$) when compared to those on the RR and RL regimens; [2] increased MB runt related transcription factor 2 (RUNX2) mRNA expression (Fig. 6B, $P \leq 0.05$) when compared to those on the RR and LL regimens; [3] increased MB gamma-carboxyglutamate protein (BGLAP) mRNA expression (Fig. 6C, $P \leq 0.05$) when compared to those on the RL and LL regimens. At 18 h post-oviposition, laying hens on the LL phosphorus feeding regimen had increased MB ALPL mRNA expression (Fig. 6A, $P \leq 0.05$) when compared to those on RR, RL and LL regimens. Within the LR phosphorus feeding regimen, laying hens had decreased MB ALPL, RUNX2, BGLAP and collagen type I alpha 2 chain (COL1A2) mRNA expressions at 18 h post-oviposition when compared to 6 h post-oviposition (Fig. 6A−D).Fig. 6Daily dynamic phosphorus feeding regimen improved the rhythm of MB mineralization markers in 70-week-old Hy-Line Brown laying hens. A) ALPL mRNA ($$n = 6$$ per group); B) RUNX2 mRNA ($$n = 6$$ per group); C) BGLAP mRNA ($$n = 6$$ per group); D) SLC1A2 mRNA ($$n = 6$$ per group). Data are presented as the mean ± SEM. $P \leq 0.05$ by one-way ANOVA followed by Duncan's multiple range tests, the letters (A, B) and the letters (a, b) indicate significant differences among all treatment groups chosen from 6 h post-oviposition and 18 h post-oviposition, respectively ($P \leq 0.05$). The significance of difference between 6 h and 18 h post-oviposition for each group was analyzed using two-tailed students’ t-tests, *$P \leq 0.05$, **$P \leq 0.01.$ RR, provided with regular phosphorus diet at both 09:00 and 17:00 for 12 weeks; RL, provided with regular phosphorus diet at 09:00 and low phosphorus diet at 17:00 for 12 weeks; LR, provided with low phosphorus diet at 09:00 and regular phosphorus diet at 17:00 for 12 weeks; LL, provided with low phosphorus diet at both 09:00 and 17:00 for 12 weeks. ALPL, alkaline phosphatase; BGLAP, gamma-carboxyglutamate protein; COL1A2, collagen type I alpha 2 chain; MB, medullary bone; RUNX2, runt related transcription factor 2
## Discussion
The phosphate metabolism of laying hens exhibits a characteristic circadian rhythm over the course of the 24-h egg-laying cycle [21]. A nadir of serum phosphate concentration was observed at the time of oviposition (which happened during 07:30−08:30 in the current study). After oviposition, serum phosphate concentration was gradually increased and raised to the peak at 01:30−02:30 in the next morning (18-h after oviposition). The presence of serum phosphate oscillations was initially thought to be a simple result of the eggshell formation cycle [31]. However, recent advances in phosphorus nutrition mechanisms indicting that serum phosphate oscillations may play vital roles in regulating the fast MB remodeling process which directly determined the quality of eggshell formation [32]. What need to be studied is whether the MB remodeling and eggshell formation efficiency could be improved by strengthening the phosphate circadian rhythms, especially when the serum phosphate lost its rhythms in cases of metabolic diseases, aging and nutritional disorders [17, 33, 34]. Indeed, abnormal phosphate rhythm has repeatedly been described as a typical symptom in bone and kidney related diseases in humans and animals [19, 32]. Not surprisingly, the management of phosphate rhythm is difficult, and, despite multifaceted approaches have been tested regarding nutritional intervention [35, 36] and metabolic regulation [37], it remains unsuccessful or at least inefficient. The daily egg-laying physiology makes avian systems excellent models for understanding the underlining mechanisms and developing effective strategies for managing body phosphate rhythms.
Daily oscillation of serum phosphate concentrations is the sum of multi-organ interactions among intestine, kidneys, and bone [38]. What is particular in laying hens is the existence of MB (which provides a rapidly accessible reservoir for calcium and phosphate) [21] and oviduct uterus (which secrets all necessary ionic and organic precursors for the eggshell formation process) [39]. The eggshell formation process can be divided into three stages: initial phase (from 5 to 10 h after oviposition), growth phase (from 10 to 22 h after oviposition) and terminal phase (from 22 to 24 h after oviposition) [40]. In the current study, increased protein productions of calcium tunnels/transporters (i.e., TRPV6 and CaBP-D28k) were observed in oviduct uterus during the eggshell formation period, indicating an increasing of calcium secretion. As a result, serum calcium decreased and the absorption of MB was stimulated. When the eggshell formation was mostly completed in the early morning, serum calcium started to increase as a result of decreased calcium secretion and increased feed intake. Accordingly, the MB was reformed. The above-mentioned changes clearly demonstrated the interactions between MB and body calcium-phosphate nutritional conditions. Of particular note, in this study, serum phosphate oscillation in laying hens was apparently not simply derived by MB remodeling, because kidney protein productions of NPt2a (a major phosphate transporter in the kidney) was significantly increased before serum phosphate concentration reached the peak. Seemly, during the fast-deposition period of eggshell formation, the high-phosphate status in serum was achieved by both MB releasing and kidney resorption. These observations brought up a concept that the increased serum phosphate concentration is physiologically required by the laying hens for eggshell formation. In this sense, improving phosphate circadian rhythms may benefit MB remodeling and eggshell formation.
Manipulating the sequence of nutrient ingestion has been shown as effective in preventing circadian misalignment and metabolic disorders [41]. In the current study, layer on the LR phosphate feeding regimen (fed $0.14\%$ NPP at 09:00 and $0.32\%$ NPP at 17:00) had increased levels of egg specific gravity, shell index, eggshell thickness and eggshell strength, when compared to layers on the RR phosphate feeding regimen (fed $0.32\%$ NPP at both 09:00 and 17:00). Possibly, the LR phosphate feeding regimen induced an increase in uterus TRPV6 protein production, and thereby increased uterus calcium secretion, which directly determines the quality of the fast-deposition eggshell [42]. Impaired eggshell quality was also observed in layers on the LL phosphate feeding regimen, which represent a direct deprivation of dietary inorganic phosphate without considering the body phosphate rhythms. These results support the hypothesis that eggshell quality could be improved by a sequential phosphate feeding regimen that designed to enhancing body phosphate circadian rhythms in aged laying hens. We previously showed that when laying hens were fed with the LR regimen, egg production performance was well supported with significant decreases in phosphorus excretion [43]. Thus, dynamic phosphorus feeding regimen has the potential to be used in commercial laying hen farms. To give concrete conclusions, many more different feeding regimes, with more dietary phosphorus levels and feeding time points, will need to be tested in future studies. In the current study, phosphorus feeding regimens were simply adjusted by the changes in dietary phosphorus levels. Indeed, part of the phosphorus may be given via drinking water [44]. So, it is worth exploring to adjust phosphorus feeding regimens by controlling water phosphorus concentrations. In humans, the sequential nutrition of phosphate has rarely been studied [12, 13]. Even in patients with serious phosphate metabolic diseases, phosphate is simply intervened by controlling overall dietary concentrations [45, 46]. Future studies will need to further illustrate the mechanisms and potential applications of sequential phosphate regimens in humans and animals.
MB is a special bone tissue that forms in the bone marrow cavities of egg-laying birds which provides calcium for eggshell formation [40]. In the present study, layers on the LR phosphate feeding regimen had increased MB remodeling ability (shown as increased absorption rate during eggshell formation period), when compared to those on all the other phosphate regimens. These results well explained the increased shell quality in layers on the LR phosphate feeding regimen. In retrospect, the importance of phosphate nutrition in regulating bone remodeling has been well documented in humans and animals [47, 48]. The fact that daily dynamic phosphate feeding regimen regulated MB remodeling further suggesting the possibility of using sequential phosphate nutrition technologies in control bone mineralization diseases [49]. In humans and animals, bone remodeling markers exhibit typical diurnal variations, and circadian disruption (by sleep alterations or metabolism disorders) is associated with the bone remodeling disorder osteoporosis [50, 51]. Especially, considering the circadian disruption and progressive bone loss in aged individuals, strategies to strengthen intrinsic calcium-phosphate circadian rhythms are highly warranted [49, 52]. Such strategies should preferably involve nutritional interventions (e.g., daily dynamic dietary regimens) instead of pharmaceutical drugs that are costly and may cause adverse effects [41, 53].
## Conclusion
In conclusion, we demonstrate that a simple daily dynamic phosphate feeding regimen (i.e., fed $0.14\%$ NPP at 09:00 and $0.32\%$ NPP at 17:00), which was designed to strengthen intrinsic phosphate circadian rhythms, enhanced MB remodeling, elevated oviduct uterus calcium secretion, and subsequently increased eggshell quality in laying hens. These results underscore the importance of manipulating the sequence of phosphate ingestion, instead of simply controlling dietary phosphate, in modifying the bone remodeling process.
## Supplementary Information
Additional file 1: Table S1. Sequences of primers used for the quantitative real-time PCR analysis1.
## References
1. Bowler MW, Cliff MJ, Waltho JP, Blackburn GM. **Why did nature select phosphate for its dominant roles in biology?**. *New J Chem.* (2010.0) **34** 784-94. DOI: 10.1039/b9nj00718k
2. Levi M, Gratton E, Forster IC, Hernando N, Wagner CA, Biber J. **Mechanisms of phosphate transport**. *Nat Rev Nephrol.* (2019.0) **15** 482-500. DOI: 10.1038/s41581-019-0159-y
3. 3.Khosla S, Melton LJ 3rd. Clinical practice. Osteopenia. N Engl J Med. 2007;356(22):2293–300. 10.1056/NEJMcp070341.
4. Rachner TD, Khosla S, Hofbauer LC. **Osteoporosis: now and the future**. *Lancet.* (2011.0) **377** 1276-87. DOI: 10.1016/S0140-6736(10)62349-5
5. Shavit L, Jaeger P, Unwin RJ. **What is nephrocalcinosis?**. *Kidney Int.* (2015.0) **88** 35-43. DOI: 10.1038/ki.2015.76
6. Sage AP, Tintut Y, Demer LL. **Regulatory mechanisms in vascular calcification**. *Nat Rev Cardiol.* (2010.0) **7** 528-36. DOI: 10.1038/nrcardio.2010.115
7. Kalantar-Zadeh K, Fouque D. **Nutritional management of chronic kidney disease**. *N Engl J Med.* (2017.0) **377** 1765-76. DOI: 10.1056/NEJMra1700312
8. Sullivan C, Sayre SS, Leon JB, Machekano R, Love TE, Porter D. **Effect of food additives on hyperphosphatemia among patients with end-stage renal disease a randomized controlled trial**. *JAMA-J Am Med Assoc.* (2009.0) **301** 629-35. DOI: 10.1001/jama.2009.96
9. Liabeuf S, Ryckelynck JP, El Esper N, Urena P, Combe C, Dussol B. **Randomized clinical trial of sevelamer carbonate on serum klotho and fibroblast growth factor 23 in CKD**. *Clin J Am Soc Nephrol.* (2017.0) **12** 1930-40. DOI: 10.2215/CJN.03030317
10. Selim S, Abdel-Megeid NS, Khalifa HK, Fakiha KG, Majrashi KA, Hussein E. **Efficacy of various feed additives on performance, nutrient digestibility, bone quality, blood constituents, and phosphorus absorption and utilization of broiler chickens fed low phosphorus diet**. *Animals.* (2022.0) **12** 1742. DOI: 10.3390/ani12141742
11. Imel EA, DiMeglio LA, Hui SL, Carpenter TO, Econs MJ. **Treatment of X-linked hypophosphatemia with calcitriol and phosphate increases circulating fibroblast growth factor 23 concentrations**. *J Clin Endocrinol Metab.* (2010.0) **95** 1846-50. DOI: 10.1210/jc.2009-1671
12. Isakova T, Block G. **The phosphate bucket list**. *Kidney Int.* (2018.0) **93** 1033-5. DOI: 10.1016/j.kint.2018.01.010
13. Barreto FC, Barreto DV, Massy ZA, Drueke TB. **Strategies for phosphate control in patients with CKD**. *Kidney Int Rep.* (2019.0) **4** 1043-56. DOI: 10.1016/j.ekir.2019.06.002
14. Markowitz M, Rotkin L, Rosen JF. **Circadian rhythms of blood minerals in humans**. *Science.* (1981.0) **213** 672-4. DOI: 10.1126/science.7256269
15. Becker GJ, Walker RG, Hewitson TD, Pedagogos E. **Phosphate levels–time for a rethink?**. *Nephrol Dial Transplant.* (2009.0) **24** 2321-4. DOI: 10.1093/ndt/gfp220
16. Motohashi H, Tahara Y, Whittaker DS, Wang HB, Yamaji T, Wakui H. **The circadian clock is disrupted in mice with adenine-induced tubulointerstitial nephropathy**. *Kidney Int.* (2020.0) **97** 728-40. DOI: 10.1016/j.kint.2019.09.032
17. Egstrand S, Olgaard K, Lewin E. **Circadian rhythms of mineral metabolism in chronic kidney disease-mineral bone disorder**. *Curr Opin Nephrol Hypertens.* (2020.0) **29** 367-77. DOI: 10.1097/MNH.0000000000000611
18. Nordholm A, Egstrand S, Gravesen E, Mace ML, Morevati M, Olgaard K. **Circadian rhythm of activin A and related parameters of mineral metabolism in normal and uremic rats**. *Pflugers Arch.* (2019.0) **471** 1079-94. DOI: 10.1007/s00424-019-02291-2
19. Svajger BA, Riddoch JLH, Pruss CM, Laverty KJ, Ward E, Holden RM. **Development of experimental chronic kidney disease and vascular calcification alters diurnal variation of phosphate and its hormonal regulators**. *Physiol Rep.* (2020.0) **8** e14626. DOI: 10.14814/phy2.14626
20. Haarhaus M, Evenepoel P. **Differentiating the causes of adynamic bone in advanced chronic kidney disease informs osteoporosis treatment**. *Kidney Int.* (2021.0) **100** 546-58. DOI: 10.1016/j.kint.2021.04.043
21. Kerschnitzki M, Zander T, Zaslansky P, Fratzl P, Shahar R, Wagermaier W. **Rapid alterations of avian medullary bone material during the daily egg-laying cycle**. *Bone.* (2014.0) **69** 109-17. DOI: 10.1016/j.bone.2014.08.019
22. 22.Wang M, O’Connor JK, Bailleul AM, Li Z. Evolution and distribution of medullary bone: evidence from a new Early Cretaceous enantiornithine bird. Natl Sci Rev. 2020;7(6):1068–78.
23. Bar A, Hurwitz S. **Egg shell quality, medullary bone ash, intestinal calcium and phosphorus absorption, and calcium-binding protein in phosphorus-deficient hens**. *Poult Sci.* (1984.0) **63** 1975-9. DOI: 10.3382/ps.0631975
24. Habig C, Weigend A, Baulain U, Petow S, Weigend S. **Influence of age and phylogenetic background on blood parameters associated with bone metabolism in laying hens**. *Front Physiol.* (2021.0) **12** 678054. DOI: 10.3389/fphys.2021.678054
25. Bar A, Razaphkovsky V, Vax E. **Re-evaluation of calcium and phosphorus requirements in aged laying hens**. *Br Poult Sci.* (2002.0) **43** 261-9. DOI: 10.1080/00071660120121481
26. Sohail SS, Roland DA. **Influence of dietary phosphorus on performance of Hy-line W36 hens**. *Poult Sci.* (2002.0) **81** 75-83. DOI: 10.1093/ps/81.1.75
27. Pongmanee K, Kuhn I, Korver DR. **Effects of phytase supplementation on eggshell and bone quality, and phosphorus and calcium digestibility in laying hens from 25 to 37 wk of age**. *Poult Sci.* (2020.0) **99** 2595-607. DOI: 10.1016/j.psj.2019.12.051
28. 28.Nie W, Yang Y, Yuan J, Wang Z, Guo Y. Effect of dietary nonphytate phosphorus on laying performance and small intestinal epithelial phosphate transporter expression in Dwarf pink-shell laying hens. J Anim Sci Biotechnol. 2013;4(1):34.
29. Yan J, Pan C, Liu Y, Liao X, Chen J, Zhu Y. **Dietary vitamin D3 deprivation suppresses fibroblast growth factor 23 signals by reducing serum phosphorus levels in laying hens**. *Anim Nutr.* (2022.0) **9** 23-30. DOI: 10.1016/j.aninu.2021.07.010
30. 30.Liu Y, Shen J, Yang X, Sun Q, Yang X. Folic acid reduced triglycerides deposition in primary chicken hepatocytes. J Agric Food Chem. 2018;66(50):13162–72.
31. 31.Nys Y, Le Roy N. Calcium homeostasis and eggshell biomineralization in female chicken. In: Feldman D, editor. Vitamin D. Fourth Edition. Academic; 2018. p. 361–82. 10.1016/B978-0-12-809965-0.00022-7.
32. 32.Nys Y, N’Guyen TM, Williams J, Etches RJ. Blood levels of ionized calcium, inorganic phosphorus, 1,25-dihydroxycholecalciferol and gonadal hormones in hens laying hard-shelled or shell-less eggs. J Endocrinol. 1986;111(1):151–7.
33. Trivedi H, Szabo A, Zhao S, Cantor T, Raff H. **Circadian variation of mineral and bone parameters in end-stage renal disease**. *J Nephrol.* (2015.0) **28** 351-9. DOI: 10.1007/s40620-014-0124-6
34. Masood T, Kushwaha RS, Singh R, Sailwal S, Pandey H, Varma A. **Circadian rhythm of serum 25 (OH) vitamin D, calcium and phosphorus levels in the treatment and management of type-2 diabetic patients**. *Drug Discov Ther.* (2015.0) **9** 70-4. DOI: 10.5582/ddt.2015.01002
35. Portale AA, Halloran BP, Morris RC. **Dietary intake of phosphorus modulates the circadian rhythm in serum concentration of phosphorus. Implications for the renal production of 1,25-dihydroxyvitamin D**. *J Clin Invest.* (1987.0) **80** 1147-54. DOI: 10.1172/JCI113172
36. Frost TJ, Roland DA, Marple DN. **The effects of various dietary phosphorus levels on the circadian patterns of plasma 1,25-dihydroxycholecalciferol, total calcium, ionized calcium, and phosphorus in laying hens**. *Poult Sci.* (1991.0) **70** 1564-70. DOI: 10.3382/ps.0701564
37. Isakova T, Xie H, Barchi-Chung A, Smith K, Sowden N, Epstein M. **Daily variability in mineral metabolites in CKD and effects of dietary calcium and calcitriol**. *Clin J Am Soc Nephrol.* (2012.0) **7** 820-8. DOI: 10.2215/CJN.11721111
38. Hernando N, Gagnon K, Lederer E. **Phosphate transport in epithelial and nonepithelial tissue**. *Physiol Rev.* (2021.0) **101** 1-35. DOI: 10.1152/physrev.00008.2019
39. 39.Gautron J, Stapane L, Le Roy N, Nys Y, Rodriguez-Navarro AB, Hincke MT. Avian eggshell biomineralization: an update on its structure, mineralogy and protein tool kit. BMC Mol Cell Biol. 2021;22:11.
40. Stapane L, Le Roy N, Ezagal J, Rodriguez-Navarro AB, Labas V, Combes-Soia L. **Avian eggshell formation reveals a new paradigm for vertebrate mineralization via vesicular amorphous calcium carbonate**. *J Biol Chem.* (2020.0) **295** 15853-69. DOI: 10.1074/jbc.RA120.014542
41. Ruan W, Yuan X, Eltzschig HK. **Circadian rhythm as a therapeutic target**. *Nat Rev Drug Discov.* (2021.0) **20** 287-307. DOI: 10.1038/s41573-020-00109-w
42. Yang JH, Zhao ZH, Hou JF, Zhou ZL, Deng YF, Dai JJ. **Expression of TRPV6 and CaBP-D28k in the egg shell gland (uterus) during the oviposition cycle of the laying hen**. *Br Poult Sci.* (2013.0) **54** 398-406. PMID: 23796121
43. Liao X, Yan J, Chen J, Huang Z, Xiao T, Li C. **A simple daily dynamic feeding regimen for reducing phosphorus consumption and excretion in laying hens**. *Anim Nutr.* (2022.0) **11** 132-41. DOI: 10.1016/j.aninu.2022.07.003
44. Sharma RK, Ravikanth K, Maini S, Rekhe DS, Rastogi SK. **Influence of calcium and phosphorus supplements with synergistic herbs on egg shell quality in late layers**. *Vet World.* (2009.0) **2** 231-3
45. Isakova T, Barchi-Chung A, Enfield G, Smith K, Vargas G, Houston J. **Effects of dietary phosphate restriction and phosphate binders on FGF23 levels in CKD**. *Clin J Am Soc Nephrol.* (2013.0) **8** 1009-18. DOI: 10.2215/CJN.09250912
46. Isaka Y, Hamano T, Fujii H, Tsujimoto Y, Koiwa F, Sakaguchi Y. **Optimal phosphate control related to coronary artery calcification in dialysis patients**. *J Am Soc Nephrol.* (2021.0) **32** 723-35. DOI: 10.1681/ASN.2020050598
47. 47.Trautvetter U, Jahreis G, Kiehntopf M, Glei M. Consequences of a high phosphorus intake on mineral metabolism and bone remodeling in dependence of calcium intake in healthy subjects – a randomized placebo-controlled human intervention study. Nutr J. 2016;15(1):7.
48. Chande S, Bergwitz C. **Role of phosphate sensing in bone and mineral metabolism**. *Nat Rev Endocrinol.* (2018.0) **14** 637-55. DOI: 10.1038/s41574-018-0076-3
49. Winter EM, Kooijman S, Appelman-Dijkstra NM, Meijer OC, Rensen PC, Schilperoort M. **Chronobiology and chronotherapy of osteoporosis**. *JBMR Plus.* (2021.0) **5** e10504. DOI: 10.1002/jbm4.10504
50. Swanson C, Shea SA, Wolfe P, Markwardt S, Cain SW, Munch M. **24-hour profile of serum sclerostin and its association with bone biomarkers in men**. *Osteoporosis Int.* (2017.0) **28** 3205-13. DOI: 10.1007/s00198-017-4162-5
51. Swanson CM, Shea SA, Kohrt WM, Wright KP, Cain SW, Munch M. **Sleep restriction with circadian disruption negatively alter bone turnover markers in women**. *J Clin Endocrinol Metab.* (2020.0) **105** 2456-63. DOI: 10.1210/clinem/dgaa232
52. Imerb N, Thonusin C, Chattipakorn N, Chattipakorn SC. **Aging, obese-insulin resistance, and bone remodeling**. *Mech Ageing Dev.* (2020.0) **191** 111335. DOI: 10.1016/j.mad.2020.111335
53. Lewis P, Oster H, Korf HW, Foster RG, Erren TC. **Food as a circadian time cue - evidence from human studies**. *Nat Rev Endocrinol.* (2020.0) **16** 213-23. DOI: 10.1038/s41574-020-0318-z
|
---
title: MiR-760 targets HBEGF to control cartilage extracellular matrix degradation
in osteoarthritis
authors:
- Yingchun Zhu
- Chi Zhang
- Bo Jiang
- Qirong Dong
journal: Journal of Orthopaedic Surgery and Research
year: 2023
pmcid: PMC9999495
doi: 10.1186/s13018-023-03664-1
license: CC BY 4.0
---
# MiR-760 targets HBEGF to control cartilage extracellular matrix degradation in osteoarthritis
## Abstract
The present study was developed to explore whether microRNA (miR)-760 targets heparin-binding EGF-like growth factor (HBEGF) to control cartilage extracellular matrix degradation in osteoarthritis. Both miR-760 and HBEGF expression levels were analysed in human degenerative cartilage tissues and in interleukin (IL)-1β/tumour necrosis factor (TNF)-α-treated chondrocytes in vitro. A series of knockdown and overexpression assays were then used to gauge the functional importance of miR-760 and HBEGF in OA, with qPCR and western immunoblotting analyses. Bioinformatics assays were used to identify putative miR-760 target genes, with these predictions then being validated through RNA pulldown and luciferase reporter assays. A murine anterior cruciate ligament transection model of OA was then established to prove the in vivo relevance of these findings. These experiments revealed that human degenerative cartilage tissues exhibited significant increases in miR-760 expression with a concomitant drop in HBEGF levels. IL-1β/TNF-α-treated chondrocytes also exhibited significant increases in miR-760 expression with a concomitant drop in HBEGF expression. When chondrocytes were transfected with either miR-760 inhibitor or HBEGF overexpression constructs, this was sufficient to interfere with degradation of the extracellular matrix (ECM). Moreover, miR-760 was confirmed to control chondrocyte matrix homeostasis by targeting HBEGF, and the overexpression of HBEGF partially reversed the effects of miR-760 mimic treatment on the degradation of the cartilage ECM. When OA model mice were administered an intra-articular knee injection of an adenoviral vector encoding a miR-760 mimic construct, cartilage ECM degradation was aggravated. Conversely, the overexpression of HBEGF in OA model mice partially reversed the effects of miR-760 overexpression, restoring appropriate ECM homeostasis. In summary, these data indicated that the miR-760/HBEGF axis plays a central role in orchestrating the pathogenesis of OA, making it a candidate target for therapeutic efforts in OA.
### Supplementary Information
The online version contains supplementary material available at 10.1186/s13018-023-03664-1.
## Introduction
Osteoarthritis (OA) is the second most common cause of musculoskeletal disability and a prominent driver of joint pain in affected patients [1]. At present, therapeutic strategies used to treat OA primarily focus on relieving symptoms through a combination of surgery, injections, and/or physical therapy. However, these approaches cannot restore normal joint function or prevent further disease progression in treated patients [2]. Moreover, the mechanisms driving the onset and pathological progression of OA are limited, hampering efforts to develop novel therapeutic interventions.
Prior research efforts have largely explored the biomechanical and morphological alterations exhibited by OA patients, as these changes, together with cellular metabolic dysregulation-induced nutritional dysfunction, are thought to contribute to eventual joint degeneration [3, 4]. However, more recent work has suggested that altered cellular functionality and cell death within joints may be a more critical driver of OA pathogenesis [5]. Under physiological conditions, chondrocytes maintain appropriate joint homeostasis by secreting extracellular matrix (ECM) components and a range of other regulatory mediators. As such, chondrocyte function is a key determinant of joint function [6, 7]. Inflammation is a hallmark of the pathological progression of OA and a range of other degenerative diseases, with interleukin (IL)-1β, tumour necrosis factor (TNF)-α and other inflammatory cytokines functioning to promote altered ECM metabolism and the senescence or apoptosis of target chondrocytes [8, 9]. As such, strategies aimed at modulating chondrocyte ECM metabolic activity may be an effective means of treating OA.
MicroRNAs (miRNAs) are 20–24 nucleotide transcripts that can control diverse processes including proliferation, differentiation, and immune response induction through the post-transcriptional regulation of specific target mRNAs [10–13]. Several miRNAs have been found to play important regulatory roles in OA [14, 15]. For example, miR-335-5p has been shown to activate autophagic activity, thereby mitigating chondrocyte inflammation and slowing OA disease progression [16], while miR-1271 can influence the ECM metabolizing activity of chondrocytes to alter OA incidence and severity [17]. As such, miRNAs are important regulators of joint homeostasis that warrant further study in an effort to highlight novel therapeutic targets and to gain additional insight into the regulatory pathways that shape the progression of OA.
MiR-760 was recently reported to be involved in the development of inflammation. MiR-760-3p could regulate Map3k8 to activate the NF-κB pathway in cerebral ischemia [18], while the ceRNA network circNTRK2/miR-760/LAT was dysregulated in obese patients, which was also associated with inflammation [19]. In addition, miR-760 was found to target Myo18b in the context of rheumatoid arthritis to control skeletal muscle cell proliferation [20]. Osteoarthritis is the most common arthritic disease, and whether miR-760 plays a similar regulatory role in OA has yet to be established. Accordingly, the present study was developed to assess the expression and function of miR-760 in human degenerative cartilage tissues and IL-1β/TNF-α-treated chondrocytes to gain insight into its regulatory role in the context of abnormal chondrocyte-regulated matrix homeostasis. Through these experiments, the miR-760/heparin-binding EGF-like growth factor (HBEGF) axis was identified as a novel regulator of the progression of OA, underscoring its potential utility as a target for therapeutic intervention in patients affected by this debilitating disease.
## Ethics statement
The Ethical Committee of the Ningbo City First Hospital approved all human studies, which were conducted in accordance with the Declaration of Helsinki (Approval Number: 2021-R142). Informed consent was obtained before experimentation with human subjects. All animal studies were approved by the Animal Care and Use Committee of Ningbo University and were performed as per the NIH Guide for the Care and Use of Laboratory Animals (Approval Number: 11430).
## Human tissue collection and cell culture
Human tissue and cell culture samples were harvested from 20 end-stage symptomatic OA patients undergoing total knee joint replacement surgery. A schematic representing the target locations for the collection of degenerative cartilage samples from internal worn areas and nondegenerative samples from external nonabraded areas is shown in Fig. 1A. Twenty paired clinical OA and control tissues were used for miR-760 and HEBGF expression analysis. Nondegenerative samples from three different donors were used for cell extraction and culture. Human chondrocytes were prepared by cutting isolated samples of articular cartilage into pieces and digesting them for 6 h in DMEM containing collagenase II (2 mg/mL) in a 37 °C incubator. Supernatants were then passed through a 0.075 mm filter, and the filtrate was then centrifuged. Then, the bottom sediment was washed two times using PBS and cultured in DMEM containing $10\%$ fetal bovine serum (FBS; Thermo Fisher Scientific, MA, USA) and $1\%$ penicillin‒streptomycin in a 37 °C humidified $5\%$ CO2 incubator. Human chondrocytes were collected from 3 patients and used for in vitro experiments. Fig. 1Degenerative OA-associated cartilage exhibited higher levels of miR-760 expression than nondegenerative cartilage. A Schematic overview of the classification of degenerative and nondegenerative cartilage tissue samples. B Representative safranin-O/fast green staining of cartilage from control or OA patients. Scale bar: 100 µm. C MiR-760 expression levels in OA and non-OA cartilage samples. $$n = 20$$ (twenty different donors). Student's t test (t test) was used for comparisons between two groups. * $p \leq 0.05.$ D, E The expression of miR-760 in chondrocytes stimulated with TNF-α (10 ng/mL) and IL-1β (10 ng/mL) for 24 h, 48 h, and 72 h. $$n = 3$$ (1 technical replicate on 3 different donors). Data were analysed by one-way analysis of variance (ANOVA) followed by Tukey’s post hoc test for comparison between the control and treatment groups. *, $p \leq 0.05$ compared with 0 h. MiR-760 regulates chondrocyte-mediated degradation of the extracellular matrix. F, I After 24 h of transfection, the efficiency of the overexpression or knockdown of miR-760 in chondrocytes transfected with the indicated constructs. $$n = 3$$ with 3 technical replicates. Student's t test (t test) was used for comparisons between two groups. G, J After 24 h, cartilage metabolism-related gene expression was analysed in chondrocytes via qPCR following the overexpression or inhibition of miR-760. $$n = 3$$ with 3 technical replicates. Student's t test (t test) was used for comparisons between two groups. * $p \leq 0.05.$ H, K After 48 h, cartilage metabolism-related protein expression was examined in chondrocytes following the overexpression or inhibition of miR-760 via Western immunoblotting. $$n = 3$$; three different donors. miRNAs, microRNAs; qPCR, real-time quantitative PCR. OA, osteoarthritis; TNF, tumour necrosis factor; IL, interleukin
## Cell treatment
Primary human articular chondrocytes were treated with TNF-α (10 ng/mL) and IL-1β (10 ng/mL) (Sigma-Aldrich) when 70–$80\%$ confluent, and total RNA was then harvested from these cells after 0, 24, 48, or 72 h to assess inflammation-induced changes in miR-760 and HBEGF levels.
## Transfection
Human chondrocytes were added to 6-well plates and incubated to $70\%$ confluence, after which they were transfected with miR-760 mimic/inhibitor, overexpression HBEGF (OE HBEGF), knockdown HBEGF (shHBEGF), or corresponding negative control constructs (Ruibo, Guangzhou, China) using Lipofectamine 3000 (Invitrogen, CA, USA). At 24 h and 48 h post-transfection, cells were harvested for downstream use. Transfection efficiency was assessed through qPCR and Western immunoblotting. The above constructs sequences were as follows:mimic miR-760 sequence: 5'-CGGCUCUGGGUCUGUGGGGA-3', 5'-UCCCCACAGACCCAGAGCCG-3'; negative control of mimic miR-760 sequence: 5'-UUCUCCGAACGUGUCACGUTT-3', 5'-ACGUGACACGUUCGGAGAATT-3'; inhibitor miR-760 sequence: 5'-UCCCCACAGACCCAGAGCCG-3'; negative control of inhibitor miR-760 sequence: 5'-CAGUACUUUUGUGUAGUACAA-3'. In addition, the pLVX vector was used for HBEGF overexpression and knockdown. Vector sequence diagrams are described in Additional file 1.
## RNA extraction and quantitative real-time PCR analysis
In this study, mRNA was isolated from cultured cells and knee articular cartilage tissues. The cultured cells were rinsed with PBS and lysed in RNA-Solv® Reagent (Omega Bio-tek, Norcross, GA, USA). The knee cartilage samples were placed in paired RNase-Free 1.5 EP tubes with four ground beads (5 mm in diameter) and frozen with liquid nitrogen. Subsequently, the tissues were pulverized and homogenized using Tissuelyser-24 (Jingxin, Shanghai, China). The TissueLyser was operated twice for 30 s at 45 Hz. The above tissue powder (50–100 mg) was lysed in Omega RNA-Solv® Reagent and RNA was isolated using the E.Z.N.A.® Total RNA Kit I (Omega Bio-tek) according to manufacturer’s protocol. MiRNA levels were extracted using a miRNA Isolation Kit (Ambion). RNA was stored at − 80 °C. Reverse transcription was performed using 1.0 µg total RNA and then used to prepare cDNA using miRNA and HiFiScript cDNA kits (CWBIO, Beijing, China), which were used to investigate the expression of miRNA and mRNA, respectively. All qPCRs were performed in a 20 µL volume using appropriate primers (1 µL; Sangon Biotech, Shanghai, China), cDNA (1 µL), and a ROX-containing UltraSYBR Mixture (CWBIO) with an ABI 7500 Sequencing Detection instrument (Applied Biosystems, CA, USA). The thermocycler settings were as follows: 40 cycles of 95 °C for 5 s and 60 °C for 24 s. U6 was used as an internal control for microRNA, whereas β-actin served as the control for messenger RNA. The cycle threshold (Ct) values were collected and normalized to the level of U6 or β-actin, with three samples per group. The relative mRNA level of each target gene was calculated by using the 2−ΔΔCt method. Primer sequences are shown in Table 1.Table 1Primer sequences for qPCRGenePrimersMiR-760Forward: UUCUCCGAACGUGUCACGUTTReverse: ACGUGACACGUUCGGAGAATTMMP3Forward: AGTCTTCCAATCCTACTGTTGCTReverse: TCCCCGTCACCTCCAATCCMMP13Forward: ACTGAGAGGCTCCGAGAAATGReverse: GAACCCCGCATCTTGGCTTADAMTS4Forward: GAGGAGGAGATCGTGTTTCCAReverse: CCAGCTCTAGTAGCAGCGTCCOL2A1Forward: TGGACGATCAGGCGAAACCReverse: GCTGCGGATGCTCTCAATCTAggrecanForward: ACTCTGGGTTTTCGTGACTCTReverse: ACACTCAGCGAGTTGTCATGGHBEGFForward: ATCGTGGGGCTTCTCATGTTTReverse: TTAGTCATGCCCAACTTCACTTTCBLForward: TGGTGCGGTTGTGTCAGAACReverse: GGTAGGTATCTGGTAGCAGGTCCAMK2GForward: ACCCGTTTCACCGACGACTAReverse: CTCCTGCGTGGAGGTTTTCTTMAP2K1Forward: CAATGGCGGTGTGGTGTTCReverse: GATTGCGGGTTTGATCTCCAGADCY1Forward: AGGCACGACAATGTGAGCATCReverse: TTCATCGAACTTGCCGAAGAGRPS6KA3Forward: CGCTGAGAATGGACAGCAAATReverse: TCCAAATGATCCCTGCCCTAATU6Forward: CTCGCTTCGGCAGCACAReverse: AACGCTTCACGAATTTGCGTβ-actinForward: AGATGTGGATCAGCAAGCAGReverse: GCGCAAGTTAGGTTTTGTCA
## Western immunoblotting
Expression of proteins of interest in chondrocytes and articular cartilage tissues were explored through western blot analysis. Briefly, cultured chondrocytes and previously ground tissues powder (100 mg) were lysed for 1 h using ice cold 1 × RIPA lysis buffer (Beyotime) supplemented with phenylmethanesulfonyl fluoride (PMSF, 1 mM, Beyotime). The lysates were centrifugated for 10 min at 4 °C and 12,000 rpm. The supernatant was transferred to a new tube for protein quantification. A standard curve was prepared with gradient concentration of BSA according to the instructions of a BCA quantification kit (Beyotime, China). For western blotting assay, the protein samples were mixed with 5 × loading buffer and heated for 10 min at 98 °C. Thereafter, 30 ug protein samples were separated by $10\%$–$12.5\%$ SDS-PAGE gels and transferred to polyvinylidene fluoride membranes (Bio-Rad, USA). The membrane was rinsed with TBS-T, following with $5\%$ nonfat milk for 1.5 h. Then, the membranes were incubated with primary antibodies specific to HBEGF (1:200, Abcam, Cat# ab92620), MMP3 (1:1000, Abcam, Cat# ab137659), MMP13 (1:1000, Abcam, Cat# ab84594), ADAMTS4 (1:1000, Abcam, Cat# ab84792), COL2A1 (1:1000, Abcam, Cat# ab34712), Aggrecan (1:100, ABclonal, Cat# A8536), or β-actin (1:2000) (Cell Signaling Technology, Cat# 4970) at 4 °C overnight. The blots were then rinsed and probed with secondary HRP-conjugated anti-rabbit or anti-mouse antibodies (Beyotime Institute of Biotechnology, Nantong, China), and proteins were detected using an ECL kit (Santa Cruz Biotechnology, TX, USA).
## Predictive bioinformatics analyses
Two databases were used to predict miR-760 target genes and downstream signalling pathways, miRDIP (http://ophid.utoronto.ca/miRDIP/index.jsp) and DAVID (DAVID Functional Annotation Bioinformatics Microarray Analysis (ncifcrf.gov)) [21]. Briefly, the potential targets of miRNAs were predicted with the miRDIP database, and the genes with a high minimum score (top $5\%$) were selected as the targets of the common miRNAs. Subsequently, the potential targets of differentially expressed miRNAs were used for KEGG enrichment analysis, and the genes were annotated and analysed by the DAVID database.
## Luciferase reporter assays
HEK293T cells were cotransfected with miR-760 mimic or control constructs together with 3ʹ-UTR-Luc reporter plasmids containing wild-type (WT) or mutated miR-760 binding sites from the HBEGF 3'-UTR using Lipofectamine 3000 (Invitrogen). For mutant HBEGF reporter constructs, we used the previously predicted binding site sequence of miR-760 and HBEGF (http://ophid.utoronto.ca/mirDIP/index.jsp#r). Then, we altered the gene at the binding site and transferred the sequence into an overexpression vector to form a mutant of the gene (detailed sequences are shown in Supplemental Materials 2). A Luciferase Reporter Gene system (Sigma‒Aldrich, MO, USA) was then used based on the provided directions to detect luciferase activity.
## Adenovirus preparation
To determine whether miR-760 or HBEGF plays a role in OA progression in vivo, adeno-associated virus (AAV, virus titer 1.3 × 1012vg/mL, catalogue number: GC20200717HZCY-AAV01) was obtained from Hanheng (Hanheng Biotechnology Co., Ltd., Shanghai, China) to overexpress miR-760 or HBEGF in the OA mouse model. The vector sequence diagram is described in Additional file 1.
## Experimental animal model
Referring to the common experimental animal randomization method and the 3R principle, 36 C57BL/6 male mice (6 weeks old) were randomized into the following three groups: a sham surgery group (incision of the right knee, lack of ACL transection surgery) and two groups in which the right knee articular cartilage underwent anterior cruciate ligament transection (ACLT) to establish an animal model of OA ($$n = 12$$). The two OA experimental groups were administered intra-articular injections of either AAV-miR-760 mimic or AAV-miR-760 mimic + AAV-OE HBEGF immediately after ACLT surgery (0 weeks). The injection procedure was repeated after 4 weeks, and the mice were sacrificed at 4 weeks after the second injection. The right knee joints of each group were harvested for the extraction of RNA ($$n = 3$$) and proteins ($$n = 3$$), with OA progression being assessed based on the expression of HBEGF, MMP3, MMP13, ADAMTS4, COL2A1, and Aggrecan. Briefly, mice were sacrificed with the nape facing up. Then, the front legs were immobilized and the skin and soft tissue were removed on the hind leg to make an incision at the knee joint. After exposing the tibial plateau, the surface resembling a regular translucent sphere (articular cartilage) was severed and processed for RNA and protein studies. In addition, the right knee joints of the remaining mice ($$n = 6$$) were dissected and processed for safranin-O/fast green staining and immunohistochemistry staining.
## Histological analysis and OARSI score
After surgery, mice that received different treatments were kept separately, and each treatment group had two cages (3 mice in a 400 square inch cage). At 8 weeks post-surgery, cartilage specimens were fixed in $4\%$ paraformaldehyde for paraffin embedding. Each paraffin-embedded cartilage sample was sectioned at 5 μm, and every tenth section was stained with $0.1\%$ safranin O solution and $0.001\%$ Fast Green solution (Sigma‒Aldrich, St. Louis, MO, USA). For simple histologic scoring of OA in the mouse, we used an approved 0–6 subjective scoring system [22]. Histologic scores were evaluated in a blinded manner according to a grading scale (0 for normal cartilage, 0.5–4 for moderately degenerated cartilage, and 5–6 for severely degenerated cartilage).
## IHC staining
The paraffin-embedded tissue sections were deparaffinized and rehydrated following standard procedures. Sections were incubated with $3\%$ H2O2 to block endogenous peroxidase activity and antigen retrieval was performed in citrated buffer at 110 ℃, for 5 min in a pressure cooker. After the citrated buffer reached room temperature, the sections were removed and incubated overnight with the primary antibodies COL2A1 (1:200, bioss, bs-10589R) and SOX9 (1:1000, Abcam, Cat# ab185966) at 4 ℃, followed by incubation with an HRP conjugated secondary antibody (Beyotime Institute of Biotechnology, Inc., Nantong, China) for 2 h at room temperature. Peroxidase binding for both COL2A1 and SOX9 was visualized using diaminobenzidine. Then, the nuclei were counterstained with hematoxylin, while the slides were dehydrated, mounted, and analyzed with a light microscope. For the quantitative analysis, all positively stained cells, including those in the femoral condyle and tibial plateau area, on the articular surface per specimen were counted, and the percentage of positive cells was calculated using Image-Pro Plus 6.0.
## Statistical analysis
SPSS 22.0 (SPSS, IL, USA) was used to analyse data, which are given as the means ± SEMs. Data were compared between groups using independent samples t tests or one-way ANOVAs with Tukey’s post hoc test as appropriate. $P \leq 0.05$ was the significance threshold.
## MiR-760 expression is relatively high in human OA tissues
Initially, degenerative and nondegenerative joint tissue samples were harvested from 20 OA patients for analysis (Fig. 1A). Safranin O and fast green staining (Fig. 1B) showed degenerative cartilage in OA-affected parts of the cartilage. Subsequent qPCR assays performed using these samples revealed significant increases in miR-760 expression in degenerative cartilage relative to nondegenerative cartilage (Fig. 1C), suggesting that this miRNA may play a role in the progression of OA.
## IL-1β and TNF-α treatment of chondrocytes promotes the time-dependent upregulation of miR-760
Inflammation is an important driver of the pathogenesis of OA. To model the relationship between such inflammation and miR-760 expression dynamics, we harvested primary human chondrocytes from nondegenerative cartilage samples and stimulated them with IL-1β/TNF-α. Subsequent qPCR analyses revealed that treatment with these inflammatory cytokines drove the time-dependent upregulation of miR-760 in these chondrocytes (Fig. 1D and E). These data were in line with the above data derived from human tissue samples.
## MiR-760 regulates chondrocyte-mediated degradation of the extracellular matrix
To gain insight into the functional role of miR-760 in the context of OA pathogenesis, we next transfected chondrocytes transfected with miR-760 mimic or inhibitor constructs, and the overexpression and knockdown efficiency was confirmed by qPCR (Fig. 1F and I). Next, the expression of anabolic enzymes associated with ECM synthesis (COL2A1, Aggrecan) and catabolic enzymes associated with ECM degradation (ADAMTS4, MMP-3, MMP-13) was assessed in these cells at the mRNA (Fig. 1G and J) and protein levels (Fig. 1H and K). MiR-760 inhibitor transfection significantly enhanced COL2A1 and aggrecan expression while suppressing MMP-3, MMP-13, and ADAMTS4 expression in these chondrocytes, and miR-760 mimic transfection yielded the opposite phenotype. As such, miR-760 may serve as a negative regulator of OA development.
## HBEGF serves as a miR-760 target gene in chondrocytes
To gain further insight regarding the mechanisms whereby miR-760 shapes the pathogenesis of OA, we used the miRDIP (http://ophid.utoronto.ca/miRDIP/index.jsp) and DAVID (DAVID Functional Annotation Bioinformatics Microarray Analysis (ncifcrf.gov)) databases to predict possible miR-760 target genes. First, we found 188 downstream genes that are highly bound to miR-760 and 11 signalling pathways associated with these 188 genes. Detailed information is displayed in Additional file 2. Second, we selected the three signalling pathways with the highest correlation (low P value), namely, long-term potentiation, the ErbB signalling pathway and the GnRH signalling pathway, leading to the identification of 6 possible targets. To investigate the correlation between miR-760 and these six genes, we transfected mimic miR-760 or inhibitor miR-760 into chondrocytes. After 24 h, chondrocyte RNA was extracted to detect the mRNA expression of the CBL, CAMK2G, HBEGF, MAP2K1, RPS6KA3, and ADCY1 genes. The results showed that after overexpression of miR-760 in chondrocytes, the expression of CAMK2G and HBEGF was significantly reduced, whereas knockdown of miR-760 resulted in the opposite phenotype (Fig. 2A and B). As shown in Fig. 2C, the CAM2KG and HBEGF protein expression was significantly reduced after overexpression of miR-760 in chondrocytes, while miR-760 inhibitor transfection only significantly enhanced HBEGF protein expression. A luciferase assay was then used to confirm the ability of miR-760 to bind the predicted HBEGF (or CAM2KG) 3'-UTR sequence by transfecting cells with miR-760 mimic or control constructs together with WT or mutant HBEGF (or CAM2KG) reporter constructs. Significantly reduced luciferase activity was observed following WT reporter and miR-760 mimic cotransfection, while no corresponding reduction was observed for the mutated reporter (Fig. 2D). As shown in Fig. 2E, no significant changes in luciferase activity were observed in both WT or mutant CAM2KG reporter and miR-760 mimic cotransfection. Hence, we inferred that miR-760 is most likely to directly target HBEGF and mediate the function of HBEGF in OA. As such, HBEGF was selected as a validated miR-760 target gene for further study. Fig. 2HBEGF serves as a direct target of miR-760 and mediates the function of miR-760 in chondrocytes. A, B After 24 h of transfection, the expression of 6 different putative miR-760 target genes in chondrocytes was assessed via qPCR following miR-760 overexpression or inhibition. $$n = 3$$ (1 technical replicate on 3 different donors). Student's t test (t test) was used for comparisons between two groups. * $p \leq 0.05.$ C After 48 h, CAM2KG and HBEGF protein levels were analysed via western immunoblotting following miR-760 inhibition or overexpression. $$n = 3$$ (1 technical replicate on 3 different donors). D, E A dual-luciferase reporter assay was used to analyse interactions between miR-760 and HBEGF (or CAM2KG) mRNA. HEK-293 T cells were cotransfected with miR-760 mimic or NC and a luciferase reporter construct containing WT or MUT HBEGF (or WT or MUT CAM2KG). $$n = 3$$ with 2 technical replicates. Data were analysed by one-way analysis of variance (ANOVA) followed by Tukey’s post hoc test for comparison between the control and treatment groups. * $p \leq 0.05.$ F Analyses of HBEGF expression in OA-associated degenerative and nondegenerative cartilage tissue samples. $$n = 20$.$ Student's t test (t test) was used for comparisons between two groups. * $p \leq 0.05.$ G, H HBEGF expression levels in TNF-α (10 ng/mL)- and IL-1β (10 ng/mL)-treated chondrocytes. $$n = 3$$ (1 technical replicate on 3 different donors). Data were analysed by one-way analysis of variance (ANOVA) followed by Tukey’s post hoc test for comparison between the control and treatment groups. *, $p \leq 0.05$ compared with 0 h. miRNAs, microRNAs; qPCR, real-time quantitative PCR; WT, wild type; MUT, mutant; OA, osteoarthritis; TNF, tumour necrosis factor; IL, interleukin; NC, negative control
## HBEGF is downregulated in OA patient cartilage tissue samples and IL-1β/TNF-α-treated chondrocytes
To assess the potential link between HBEGF and OA progression, we assessed its expression using human samples and primary chondrocytes as described above. In qPCR analyses, significant reductions in HBEGF expression were observed in degenerative cartilage samples from OA patients relative to nondegenerative samples (Fig. 2F). Similarly, time-dependent HBEGF downregulation was observed in primary human chondrocytes treated with IL-1β/TNF-α (Fig. 2G and H).
## HBEGF controls the ability of chondrocytes to regulate ECM homeostasis
To assess the functional role of HBEGF within chondrocytes, we next transfected these cells with HBEGF-specific overexpression (OE) or shRNA constructs. Successful HBEGF overexpression (Fig. 3A) and knockdown (Fig. 3B) in primary human chondrocytes were then confirmed by qPCR at 24 h post-transfection. Western immunoblotting was further used to confirm the overexpression or knockdown of HBEGF in appropriately transfected cells at 48 h post-transfection (Fig. 3C). MMP-3, MMP-13, and ADAMTS4 were downregulated at both the mRNA and protein levels in chondrocytes transfected with the OE HBEGF construct (Fig. 3D and E), with concomitant increases in COL2A1 and aggrecan expression. When HBEGF was silenced in chondrocytes, these cells exhibited decreases in COL2A1 and aggrecan expression together with increased MMP-3, MMP-13, and ADAMTS4 levels (Fig. 3F and G). Fig. 3HBEGF controls the homeostasis of chondrocyte extracellular matrix. A–C After 24 and 48 h of transfection, HBEGF overexpression and knockdown efficiency at the mRNA level (A-B, $$n = 3$$ with 3 technical replicates, Student's t test (t test) was used for comparison between two groups. * $p \leq 0.05$) and protein levels (C, $$n = 3$$; three different donors) were assessed in transfected chondrocytes. D, E Chondrogenesis-related gene expression at the mRNA level (D, $$n = 3$$ with 3 technical replicates; Student's t test was used for comparisons between two groups. * $p \leq 0.05$) and protein levels (E, $$n = 3$$; three different donors) were assessed in chondrocytes following HBEGF overexpression. F, G Chondrogenesis-related gene expression at the mRNA level (F, $$n = 3$$ with 3 technical replicates; Student's t test was used for comparisons between two groups. * $p \leq 0.05$) and protein levels (G, $$n = 3$$; three different donors) were assessed in chondrocytes following HBEGF knockdown. OE, overexpression; Sh, short hairpin RNA; NC, negative control; ECM, extracellular matrix
## HBEFG reverses the effects of miR-760 overexpression on OA progression in vitro
Next, primary human chondrocytes were transfected with both miR-760 mimic and HBEGF overexpression constructs. At 24 h post-transfection, MMP-3, MMP-13, ADAMTS4, aggrecan, and COL2A1 mRNA levels were assessed via qPCR, revealing that HBEGF overexpression partially reversed the effects of miR-760 overexpression on these ECM metabolism-related genes (Fig. 4A). Similarly, at 48 h post-transfection, western immunoblotting confirmed the ability of HBEGF overexpression to increase aggrecan and COL2A1 protein levels while suppressing MMP-3, MMP-13, and ADAMTS4 expression in miR-760 mimic-transfected cells (Fig. 4B). Conversely, when chondrocytes were cotransfected with miR-760 inhibitor and shHBEGF constructs, similar mRNA and protein level data were observed to those shown in Fig. 4C and D, with HBEGF knockdown thus reversing the beneficial effects of miR-760 inhibition, further confirming the identity of HBEGF as a miR-760 target gene. Fig. 4HBEGF reverses the effects of miR-760 on OA progression. A, B After 24 and 48 h of transfection, OA-associated cartilage metabolism gene expression at the mRNA (A, $$n = 3$$ with 3 technical replicates; data were analysed by one-way analysis of variance (ANOVA) followed by Tukey’s post hoc test for comparison between the control and treatment groups. * $p \leq 0.05$) and protein (B, $$n = 3$$; three different donors) levels were assessed in chondrocytes following miR-760 mimic transfection plus HBEGF overexpression. C, D OA-associated cartilage metabolism gene expression at the mRNA (C, $$n = 3$$ with 3 technical replicates; data were analysed by one-way analysis of variance (ANOVA) followed by Tukey’s post hoc test for comparison between the control and treatment groups. * $p \leq 0.05$) and protein (D, $$n = 3$$; three different donors) levels were assessed in chondrocytes following miR-760 inhibitor transfection plus HBEGF knockdown. The miR-760/HBEFG axis regulates OA progression in vivo. Sham: incision of the right knee of mice that did not undergo ACLT surgery. ACLT: right knee articular cartilage that underwent anterior cruciate ligament transection. E, F Changes in cartilage metabolism-related gene expression at the mRNA level (E, $$n = 3$$ with 3 technical replicates; data were analysed by one-way analysis of variance (ANOVA) followed by Tukey’s post hoc test for comparison between the control and treatment groups. * $p \leq 0.05$) and protein level (F, $$n = 3$$; three different animals) in the control group (Sham + PBS) and two OA experimental groups (ACLT + AVV-miR-760 mimic or ACLT + AAV-miR-760 mimic + AAV-OE HBEGF). G Safranin-O/fast green staining and COL2A1, SOX9 IHC staining in the control mice (Sham + PBS), and ACLT-induced OA articular cartilage (medial tibia) of mice injected with miR-760 mimic or miR-760 mimic + HBEGF overexpression. $$n = 6$$; six different animals. Scale bars = 50 μm. H OARSI scoring was performed according to staining results in the control group (Sham + PBS) and two OA experimental groups (ACLT + AVV-miR-760 mimic or ACLT + AAV-miR-760 mimic + AAV-OE HBEGF); $$n = 6$$; six different animals. Data were analysed by one-way analysis of variance (ANOVA) followed by Tukey’s post hoc test for comparison between the control and treatment groups. ** $p \leq 0.01.$ I, J Positive chondrocyte percentages of COL2A1 and SOX9 in the control group (Sham + PBS) and two OA experimental groups (ACLT + AVV-miR-760 mimic or ACLT + AAV-miR-760 mimic + AAV-OE HBEGF). Six mice were evaluated for each group, and eight sections at different sites were measured for each mouse. Data were analysed by one-way analysis of variance (ANOVA) followed by Tukey’s post hoc test for comparison between the control and treatment groups. * $p \leq 0.05$, ***$p \leq 0.001.$ OA, osteoarthritis; IHC, immunohistochemistry; ACLT, anterior cruciate ligament transection; miRNAs, microRNAs; ECM, extracellular matrix. OE, overexpression; Sh, short hairpin RNA; NC, negative control
## The miR-760/HB-EFG axis regulates OA progression in vivo
To extend the above in vitro findings, we assessed the ability of the miR-760/HB-EFG axis to regulate in vivo OA progression by establishing a murine ACLT-induced model of OA and providing these animals with intra-articular injections of adenoviral vectors encoding miR-760 mimic constructs with or without OE HBEGF vectors. At 8 weeks post-OA modelling, RNA and tissue samples from the knee joints of these animals were isolated for analysis, confirming the successful overexpression of HBEGF in mice in the appropriate treatment groups (Fig. 4E). qPCR and WB (Fig. 4E and F) were performed to detect the expression of catabolic enzymes (MMP-3, MMP-13 and ADAMTS4) and ECM composition (Aggrecan and COL2A1) in mouse cartilage tissues. The results showed that the injection of mimic miR-760 + OE HBEGF alleviated the degenerative changes in the cartilage matrix, such as decreased catabolic enzymes and enhanced ECM composition, in the mouse model of OA.
In addition, safranin O and fast green and immunohistochemistry staining (Fig. 4G) were performed. Quantitative analysis with Osteoarthritis Research Society International (OARSI) scoring showed that mimic miR-760 treatment significantly increased OARSI scores, whereas mimic miR-760 + OE HBEGF treatment lowered OARSI scores (Fig. 4H). Details about the type of OARSI scoring (mean/max/sum) are included in Additional file 2. Immunohistochemistry staining results showed that the rate of COL2A1- and SOX9-positive chondrocytes was decreased in the mimic miR-760 group, whereas the mimic miR-760 + OE HBEGF group exhibited increased ECM composition expression (Fig. 4I and J). Taken together, HBEGF overexpression partially reversed miR-760 mimic-induced exacerbation of OA severity. As such, miR-760 was able to regulate HBEGF expression and thereby drive OA progression by altering chondrocyte homeostasis and associated ECM metabolism in these animals (Fig. 5). Fig. 5A schematic overview of the putative mechanisms whereby miR-760 regulates HBEGF expression and OA progression by influencing chondrocyte-associated ECM homeostasis
## Discussion
OA is a highly complex and age-related, disabling joint disease characterized by progressive loss of hyaline articular cartilage, concomitant sclerotic changes in the subchondral bone and advancement in osteophytes [3]. A range of stressful stimuli including overuse, mechanical stress, excessive loading, and joint injury can alter articular cartilage and synovial membrane tissues, driving degeneration, osteophyte development, sclerosis, and synovial inflammation that ultimately induce the progression of OA [23, 24]. Altered cartilage metabolism is a hallmark of OA onset [25], with imbalanced ECM metabolism being the primary cause of articular cartilage loss in OA patients. The maintenance of appropriate ECM homeostasis is dependent on the tightly regulated expression of catabolic and anabolic enzymes, with MMP-3, MMP-13, and MMP-9 being among the most important catabolic regulators in this setting, whereas COL2A1, proteoglycans, and aggrecan are key anabolic proteins. The present study was developed to explore the effect of miR-760 on chondrocyte-mediated maintenance of ECM homeostasis and associated OA progression.
Several miRNAs have been shown to shape the pathogenesis of OA [26, 27]. For example, miR-132 regulates PTEN/PI3K/AKT signalling activity to shape OA-related chondrocyte activity [28], while miR-126 functions by controlling MAPK signalling in a rabbit model of OA to regulate the regeneration of cartilaginous tissue. Moreover, miR-107 can target caspase-1 to influence knee articular cartilage degradation [29, 30], whereas miR-103a-3p prevents OA by targeting FGF18 [31], and the miR-296-3p/PTEN axis shapes the development of OA [32]. MiR-599 can also target Casz1 and thereby alleviate inflammation while inhibiting chondrocyte apoptosis [33]. MiR-760 targets c-Myc to suppress fat metabolism [34], in addition to controlling cellular proliferative and migratory activity by targeting the BATF3/AP-1/cyclin D1 pathway [35]. MiR-760 has previously been reported to suppress G protein-coupled receptor kinase-interacting protein 1 expression, thereby inhibiting cell growth [36]. Here, a role for miR-760 in OA development was additionally identified, and a combination of bioinformatics analyses and preliminary screening led to the identification of HBEGF as an important miR-760 target gene that was downregulated in both OA patient degenerative cartilage tissues and in chondrocytes stimulated with IL-1β/TNF-α.
HBEGF signals by binding to the epidermal growth factor receptor (EGFR), which plays important roles in tumorigenesis [37, 38], metabolic diseases [39, 40], diabetes [41], pain-related diseases [42], and Alzheimer’s disease [43]. HBEGF can modulate the pathogenesis of OA and is reportedly linked to inflammation-associated muscle injury and related regenerative activity [44]. Mice harbouring a cartilage-specific loss of EGFR expression exhibit more rapid knee OA development, whereas cartilage degeneration was suppressed in mice in which the EGFR pathway was hyperactivated, with such hyperactivation similarly reversing other OA-related pathological changes observed following surgical medial meniscus destabilization [45]. Here, HBEGF silencing was further confirmed to induce imbalanced ECM metabolic activity in chondrocytes, with OA corresponding to a reduction in HBEGF expression, whereas the overexpression of HBEGF was sufficient to suppress excessive extracellular catabolic enzyme expression in vivo and in vitro.
To our knowledge, this study is the first to examine the regulatory role of the miR-760/HBEGF signalling pathway in the context of OA. However, this study is subject to some limitations. To avoid the effects of spontaneous osteoarthritis on the joints of mice, we used a 6-week-old mouse model with a relatively normal osteochondral morphology that was relatively conducive to operation. However, the 6-week-old mouse is very young for use in an OA model, which may affect its normal bone development and repair. In addition, mimic NC (vehicle therapy) injection was missing, given that sham surgery mice injected with PBS may not rule out the effect of miRNA itself on mouse joints. Finally, as an EGFR ligand, HBEGF can bind to the EGFR receptor and thus regulate the expression of downstream signalling pathways. Consequently, the signalling pathways downstream of the miR-760/HBEGF signalling axis also remain to be identified.
## Conclusions
In summary, these experiments revealed that miR-760 is upregulated in degenerated articular tissue samples from OA patients, with the treatment of chondrocytes using inflammatory cytokines similarly driving time-dependent miR-760 upregulation. Functional analyses demonstrated the ability of miR-760 to regulate chondrocyte-mediated ECM degradation, thereby controlling the onset and progression of OA. The overexpression of miR-760 in chondrocytes was associated with significant increases in MMP3, MMP13, and ADAMTS4 expression with concomitant aggrecan and COL2A1 downregulation, whereas miR-760 inhibition had the opposite effect. Mechanistic analyses revealed that miR-760 was capable of binding the HBEGF mRNA, thereby regulating HBEGF mRNA and protein levels within chondrocytes, with HBEGF in turn controlling OA-related metabolic genes, reversing the deleterious changes in anabolic and catabolic enzyme expression induced by miR-760. In line with these in vitro results, overexpressing miR-760 in the articular cavity of OA model mice was sufficient to drive OA progression, while simultaneous miR-760 and HBEGF overexpression was associated with the loss of miR-760-driven OA progression. Together these data thus highlight the miR-760/HBEGF axis as an important mediator of OA development that may be amenable to therapeutic intervention. However, mice in which miR-760 and/or HBEGF have been knocked out will be required to guide preclinical efforts to further confirm the role of this miR-760/HBEGF axis in the regulation of OA.
## Supplementary Information
Additional file 1. Vector sequence diagram description. Additional file 2. Supplementary table and figure.
## References
1. Sharma L. **Osteoarthritis of the Knee**. *N Engl J Med* (2021) **384** 51-59. DOI: 10.1056/NEJMcp1903768
2. Katz JN, Arant KR, Loeser RF. **Diagnosis and treatment of hip and knee osteoarthritis: a review**. *JAMA* (2021) **325** 568-578. DOI: 10.1001/jama.2020.22171
3. Chen D, Shen J, Zhao W, Wang T, Han L, Hamilton JL, Im HJ. **Osteoarthritis: toward a comprehensive understanding of pathological mechanism**. *Bone Res* (2017) **5** 16044. DOI: 10.1038/boneres.2016.44
4. Rogers EL, Reynard LN, Loughlin J. **The role of inflammation-related genes in osteoarthritis**. *Osteoarthritis Cartilage* (2015) **23** 1933-1938. DOI: 10.1016/j.joca.2015.01.003
5. Li YS, Xiao WF, Luo W. **Cellular aging towards osteoarthritis**. *Mech Ageing Dev* (2017) **162** 80-84. DOI: 10.1016/j.mad.2016.12.012
6. Long D, Blake S, Song XY, Lark M, Loeser RF. **Human articular chondrocytes produce IL-7 and respond to IL-7 with increased production of matrix metalloproteinase-13**. *Arthritis Res Ther* (2008) **10** R23. DOI: 10.1186/ar2376
7. Billinghurst RC, Dahlberg L, Ionescu M, Reiner A, Bourne R, Rorabeck C, Mitchell P, Hambor J, Diekmann O, Tschesche H, Chen J, Van Wart H, Poole AR. **Enhanced cleavage of type II collagen by collagenases in osteoarthritic articular cartilage**. *J Clin Invest* (1997) **99** 1534-1545. DOI: 10.1172/JCI119316
8. Zhou Z, Du D, Chen A, Zhu L. **Circular RNA expression profile of articular chondrocytes in an IL-1β-induced mouse model of osteoarthritis**. *Gene* (2018) **644** 20-26. DOI: 10.1016/j.gene.2017.12.020
9. Yang G, Li S, Li B, Cheng L, Jiang P, Tian Z, Sun S. **Protective effects of garlic-derived S-allylmercaptocysteine on IL-1**. *Biomed Res Int* (2017) **2017** 8686207. DOI: 10.1155/2017/8686207
10. Giordano L, Porta GD, Peretti GM, Maffulli N. **Therapeutic potential of microRNA in tendon injuries**. *Br Med Bull* (2020) **133** 79-94. DOI: 10.1093/bmb/ldaa002
11. Sales G, Calura E. **Micro-RNA quantification, target gene identification, and pathway analysis**. *Methods Mol Biol* (2021) **2284** 207-229. DOI: 10.1007/978-1-0716-1307-8_12
12. Gargano G, Oliviero A, Oliva F, Maffulli N. **Small interfering RNAs in tendon homeostasis**. *Br Med Bull* (2021) **138** 58-67. DOI: 10.1093/bmb/ldaa040
13. Gargano G, Oliva F, Oliviero A, Maffulli N. **Small interfering RNAs in the management of human rheumatoid arthritis**. *Br Med Bull* (2022) **142** 34-43. DOI: 10.1093/bmb/ldac012
14. Coutinho de Almeida R, Ramos YFM, Mahfouz A, den Hollander W, Lakenberg N, Houtman E, van Hoolwerff M, Suchiman HED, Rodríguez Ruiz A, Slagboom PE, Mei H, Kiełbasa SM, Nelissen RGHH, Reinders M, Meulenbelt I. **RNA sequencing data integration reveals an miRNA interactome of osteoarthritis cartilage**. *Ann Rheum Dis* (2019) **78** 270-277. DOI: 10.1136/annrheumdis-2018-213882
15. Oliviero A, Della Porta G, Peretti GM, Maffulli N. **MicroRNA in osteoarthritis: physiopathology, diagnosis and therapeutic challenge**. *Br Med Bull* (2019) **130** 137-147. DOI: 10.1093/bmb/ldz015
16. Zhong G, Long H, Ma S, Shunhan Y, Li J, Yao J. **miRNA-335-5p relieves chondrocyte inflammation by activating autophagy in osteoarthritis**. *Life Sci* (2019) **226** 164-172. DOI: 10.1016/j.lfs.2019.03.071
17. Shen S, Wu Y, Chen J, Xie Z, Huang K, Wang G, Yang Y, Ni W, Chen Z, Shi P, Ma Y, Fan S. **CircSERPINE2 protects against osteoarthritis by targeting miR-1271 and ETS-related gene**. *Ann Rheum Dis* (2019) **78** 826-836. DOI: 10.1136/annrheumdis-2018-214786
18. Zhang H, Deng J, Huang K, He Y, Cai Z, He Y. **circNup188/miR-760–3p/Map3k8 axis regulates inflammation in cerebral ischemia**. *Mol Cell Probes* (2022) **64** 101830. DOI: 10.1016/j.mcp.2022.101830
19. Li G, Zhang H, Cao K, Zhai Z, Wang Z, Yang L, Han J. **Transcriptome of visceral adipose tissue identifies an inflammation-related ceRNA network that regulates obesity**. *Mol Cell Biochem* (2022) **477** 1095-1106. DOI: 10.1007/s11010-022-04362-y
20. Tang X, Wang J, Zhou S, Zhou J, Jia G, Wang H, Xin C, Fu G, Zhang J. **miR-760 regulates skeletal muscle proliferation in rheumatoid arthritis by targeting Myo18b**. *Mol Med Rep* (2019) **20** 4843-4854. DOI: 10.3892/mmr.2019.10775
21. Ma Y, Wu Y, Chen J, Huang K, Ji B, Chen Z, Wang Q, Ma J, Shen S, Zhang J. **miR-10a-5p promotes chondrocyte apoptosis in osteoarthritis by targeting HOXA1**. *Mol Ther Nucleic Acids* (2019) **14** 398-409. DOI: 10.1016/j.omtn.2018.12.012
22. Glasson SS, Chambers MG, Van Den Berg WB, Little CB. **The OARSI histopathology initiative - recommendations for histological assessments of osteoarthritis in the mouse**. *Osteoarthritis Cartilage* (2010) **18** S17-S23. DOI: 10.1016/j.joca.2010.05.025
23. Timmermans RGM, Bloks NGC, Tuerlings M, van Hoolwerff M, Nelissen RGHH, van der Wal RJP, van der Kraan PM, Blom AB, van den Bosch MHJ, Ramos YFM, Meulenbelt I. **A human**. *Osteoarthr Cartil Open* (2021) **4** 100231. DOI: 10.1016/j.ocarto.2021.100231
24. Loeser RF, Goldring SR, Scanzello CR, Goldring MB. **Osteoarthritis: a disease of the joint as an organ**. *Arthritis Rheum* (2012) **64** 1697-1707. DOI: 10.1002/art.34453
25. Zheng G, Zhan Y, Li X, Pan Z, Zheng F, Zhang Z, Zhou Y, Wu Y, Wang X, Gao W, Xu H, Tian N, Zhang X. **TFEB, a potential therapeutic target for osteoarthritis via autophagy regulation**. *Cell Death Dis* (2018) **9** 858. DOI: 10.1038/s41419-018-0909-y
26. Tuerlings M, van Hoolwerff M, Houtman E, Suchiman EHED, Lakenberg N, Mei H, van der Linden EHMJ, Nelissen RRGHH, Ramos YYFM, Coutinho de Almeida R, Meulenbelt I. **RNA sequencing reveals interacting key determinants of osteoarthritis acting in subchondral bone and articular cartilage: identification of IL11 and CHADL as attractive treatment targets**. *Arthritis Rheumatol* (2021) **73** 789-799. DOI: 10.1136/annrheumdis-2018-213882
27. van Hoolwerff M, Metselaar PI, Tuerlings M, Suchiman HED, Lakenberg N, Ramos YFM, Cats D, Nelissen RGHH, Broekhuis D, Mei H, de Almeida RC, Meulenbelt I. **Elucidating epigenetic regulation by identifying functional cis-acting long noncoding RNAs and their targets in osteoarthritic articular cartilage**. *Arthritis Rheumatol* (2020) **72** 1845-1854. DOI: 10.1002/art.41396
28. Zhang W, Hu C, Zhang C, Luo C, Zhong B, Yu X. **MiRNA-132 regulates the development of osteoarthritis in correlation with the modulation of PTEN/PI3K/AKT signaling**. *BMC Geriatr* (2021) **21** 175. DOI: 10.1186/s12877-021-02046-8
29. Li S, Stöckl S, Lukas C, Herrmann M, Brochhausen C, König MA, Johnstone B, Grässel S. **Curcumin-primed human BMSC-derived extracellular vesicles reverse IL-1β-induced catabolic responses of OA chondrocytes by upregulating miR-126-3p**. *Stem Cell Res Ther* (2021) **12** 252. DOI: 10.1186/s13287-021-02317-6
30. Qian J, Fu P, Li S, Li X, Chen Y, Lin Z. **miR-107 affects cartilage matrix degradation in the pathogenesis of knee osteoarthritis by regulating caspase-1**. *J Orthop Surg Res* (2021) **16** 40. DOI: 10.1186/s13018-020-02121-7
31. Wu Y, Hong Z, Xu W, Chen J, Wang Q, Chen J, Ni W, Mei Z, Xie Z, Ma Y, Wang J, Lu J, Chen C, Fan S, Shen S. **Circular RNA circPDE4D protects against osteoarthritis by binding to miR-103a-3p and regulating FGF18**. *Mol Ther* (2021) **29** 308-323. DOI: 10.1016/j.ymthe.2020.09.002
32. Zhou Z, Ma J, Lu J, Chen A, Zhu L. **Circular RNA CircCDH13 contributes to the pathogenesis of osteoarthritis via CircCDH13/miR-296-3p/PTEN axis**. *J Cell Physiol* (2021) **236** 3521-3535. DOI: 10.1002/jcp.30091
33. Zhou JL, Deng S, Fang HS, Du XJ, Peng H, Hu QJ. **Circular RNA circANKRD36 regulates Casz1 by targeting miR-599 to prevent osteoarthritis chondrocyte apoptosis and inflammation**. *J Cell Mol Med* (2021) **25** 120-131. DOI: 10.1111/jcmm.15884
34. Yang X, Zhang C, Tie H, Luo J, Wang Y, Wu Q. **miR-760 exerts an antioncogenic effect in esophageal squamous cell carcinoma by negatively driving fat metabolism via targeting c-Myc**. *J Cell Biochem* (2020) **121** 2950-2961. DOI: 10.1002/jcb.29540
35. Ao L, Liu Y, Wang D, Huang L, Li F, Liu J, Zhang C, Shen Z, Gao Q, Yuan W, Zhang Y. **MiR-760 suppresses human colorectal cancer growth by targeting BATF3/AP-1/cyclinD1 signaling**. *J Exp Clin Cancer Res* (2018) **37** 83. DOI: 10.1186/s13046-018-0757-8
36. Ge L, Wang Y, Duan QH, Liu SS, Liu GJ. **MicroRNA-760 acts as a tumor suppressor in gastric cancer development**. *World J Gastroenterol* (2019) **25** 6619-6633. DOI: 10.3748/wjg.v25.i45.6619
37. Xiao W, Zheng S, Xie X, Li X, Zhang L, Yang A, Wang J, Tang H, Xie X. **SOX2 promotes brain metastasis of breast cancer by upregulating the expression of FSCN1 and HBEGF**. *Mol Ther Oncolyt* (2020) **17** 118-129. DOI: 10.1016/j.omto.2020.03.001
38. Wang L, Gao Y, Tong D, Wang X, Guo C, Guo B, Yang Y, Zhao L, Zhang J, Yang J, Qin Y, Liu L, Huang C. **MeCP2 drives hepatocellular carcinoma progression via enforcing HOXD3 promoter methylation and expression through the HB-EGF/EGFR pathway**. *Mol Oncol* (2021) **15** 3147-3163. DOI: 10.1002/1878-0261.13019
39. Kim S, Subramanian V, Abdel-Latif A, Lee S. **Role of heparin-binding epidermal growth factor-like growth factor in oxidative stress-associated metabolic diseases**. *Metab Syndr Relat Disord* (2020) **18** 186-196. DOI: 10.1089/met.2019.0120
40. Thornton KJ, Kamanga-Sollo E, White ME, Dayton WR. **Active G protein-coupled receptors (GPCR), matrix metalloproteinases 2/9 (MMP2/9), heparin-binding epidermal growth factor (hbEGF), epidermal growth factor receptor (EGFR), erbB2, and insulin-like growth factor 1 receptor (IGF-1R) are necessary for trenbolone acetate-induced alterations in protein turnover rate of fused bovine satellite cell cultures**. *J Anim Sci* (2016) **94** 2332-2343. DOI: 10.2527/jas.2015-0178
41. Xiang J, Zhang S, Xu R, Chu H, Biswas S, Yu S, Miao D, Li W, Li S, Brown AJ, Yang H, Xu Y, Li B, Liu H. **Elevated HB-EGF expression in neural stem cells causes middle age obesity by suppressing Hypocretin/Orexin expression**. *FASEB J* (2021) **35** e21345. DOI: 10.1096/fj.202001945R
42. Wangzhou A, Paige C, Neerukonda SV, Naik DK, Kume M, David ET, Dussor G, Ray PR, Price TJ. **A ligand-receptor interactome platform for discovery of pain mechanisms and therapeutic targets**. *Sci Signal* (2021) **14** eabe1648. DOI: 10.1126/scisignal.abe1648
43. da Rocha JF, Bastos L, Domingues SC, Bento AR, Konietzko U, da Cruz E, Silva OAB, Vieira SI. **APP Binds to the EGFR ligands HB-EGF and EGF, acting synergistically with EGF to promote ERK signaling and neuritogenesis**. *Mol Neurobiol* (2021) **58** 668-688. DOI: 10.1007/s12035-020-02139-2
44. Hanata N, Nagafuchi Y, Sugimori Y, Kobayashi S, Tsuchida Y, Iwasaki Y, Shoda H, Fujio K. **Serum amphiregulin and heparin-binding epidermal growth factor as biomarkers in patients with idiopathic inflammatory myopathy**. *J Clin Med* (2021) **10** 3730. DOI: 10.3390/jcm10163730
45. Wei Y, Luo L, Gui T, Yu F, Yan L, Yao L, Zhong L, Yu W, Han B, Patel JM, Liu JF, Beier F, Levin LS, Nelson C, Shao Z, Han L, Mauck RL, Tsourkas A, Ahn J, Cheng Z, Qin L. **Targeting cartilage EGFR pathway for osteoarthritis treatment**. *Sci Transl Med* (2021) **13** 3946. DOI: 10.1126/scitranslmed.abb3946
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.