Urinary Equol and Equol‐Predicting Microbial Species Are Favorably Associated With Cardiometabolic Risk Markers in Chinese Adults

Abstract

Background

The association between soy isoflavones intake and cardiometabolic health remains inconclusive. We investigated the associations of urinary biomarkers of isoflavones including daidzein, glycitein, genistein, equol (a gut microbial metabolite of daidzein), and equol‐predicting microbial species with cardiometabolic risk markers.

Methods and Results

In a 1‐year study of 305 Chinese community‐dwelling adults aged ≥18 years, urinary isoflavones, fecal microbiota, blood pressure, blood glucose and lipids, and anthropometric data were measured twice, 1 year apart. Brachial‐ankle pulse wave velocity was also measured after 1 year. A linear mixed‐effects model was used to analyze repeated measurements. Logistic regression was used to calculate the adjusted odds ratio (aOR) and 95% CI for the associations for arterial stiffness. Each 1 μg/g creatinine increase in urinary equol concentrations was associated with 1.47%, 0.96%, and 3.32% decrease in triglycerides, plasma atherogenic index, and metabolic syndrome score, respectively (all P<0.05), and 0.61% increase in high‐density lipoprotein cholesterol (P=0.025). Urinary equol was also associated with lower risk of arterial stiffness (aOR, 0.28 [95% CI, 0.09–0.90]; P _trend=0.036). We identified 21 bacterial genera whose relative abundance was positively associated with urinary equol (false discovery rate‐corrected P<0.05) and constructed a microbial species score to reflect the overall equol‐predicting capacity. This score (per 1‐point increase) was inversely associated with triglycerides (percentage difference=−1.48%), plasma atherogenic index (percentage difference=−0.85%), and the risk of arterial stiffness (aOR, 0.27 [95% CI, 0.08–0.88]; all P<0.05).

Conclusions

Our findings suggest that urinary equol and equol‐predicting microbial species may improve cardiometabolic risk parameters in Chinese adults.

Nonstandard Abbreviations and Acronyms

ALDS

Anhui Liver Diseases Study

ALVS

Anhui Lifestyle Validation Study

FDR

false discovery rate

LMM

linear mixed‐effects model

Research Perspective.

What Is New ?

• Epidemiological studies on the association between intake of soy isoflavones and cardiometabolic outcomes have yielded conflicting results. In addition to the challenge in the measurement of dietary isoflavones in epidemiological studies, interindividual variability in equol‐producing gut microbial metabolism of isoflavone daidzein could be another reason for the inconsistency.

• This study aimed to dissect gut microbiota associated with equol production at the population level and to investigate whether urinary biomarkers of soy isoflavones and equol‐predicting microbial species are associated with cardiometabolic risk markers among Chinese community‐dwelling adults.

• We found that urinary levels of equol but not of other isoflavones (ie, daidzein, glycitein, and genistein) were favorably associated with several cardiometabolic risk markers. We identified multiple bacterial genera that were positively associated with urinary equol concentrations and constructed a score to reflect the overall equol‐predicting potential of host gut microbiota, and this score showed a protective association with cardiometabolic health.

What Question Should Be Addressed Next?

• Our findings need to be confirmed in large‐scale cohort studies with longer follow‐up periods and deeper sequencing (eg, metagenomics) to profile gut microbiota. If validated, clinical studies are warranted to develop potential strategies to maximize equol production by modulating the gut microbiome to improve cardiometabolic health.

There is a substantial increase in the global prevalence of cardiovascular diseases, along with a notable rise in their disability‐adjusted life‐years and years of life lost. ¹ , ² It is well established that diet plays a major role in cardiometabolic health. Soy foods are commonly consumed in Asia, particularly in China, and may exhibit atheroprotective properties due to the presence of soy isoflavones. ³ Soy isoflavones preferentially bind to estrogen receptor β, expressed in the vasculature, showing selective estrogenic or antiestrogenic effects. ⁴ Moreover, human interventional epidemiological studies indicate that supplementation of soy isoflavones may improve lipid homeostasis and vascular function. ⁵ , ⁶ , ⁷ , ⁸

However, human observational epidemiological studies have reported inconsistent results on the association between soy isoflavones intake and cardiometabolic risk. Several studies suggested an inverse association between intake of soy or soy isoflavones and cardiometabolic outcomes, ⁹ , ¹⁰ whereas other studies showed a null association. ¹¹ , ¹² , ¹³ In addition to difficulty in accurately assessing dietary isoflavones in observational epidemiologic studies, interindividual variation in equol‐producing gut microbial metabolism of isoflavone daidzein might be another reason for such a discrepancy. Most previous studies ¹⁰ , ¹² , ¹⁴ used a food frequency questionnaire to collect habitual diet, which could have been subjected to measurement error. Also, food composition tables on isoflavones were incomplete, which may undermine the accuracy of the estimation of dietary isoflavone intake. Although few epidemiological studies ¹⁵ , ¹⁶ on cardiovascular outcomes have used biomarkers of isoflavones such as genistein, glycitein, and daidzein in urine or blood samples, these studies are all based on a single measurement without considering the impact of day‐to‐day variations in urinary isoflavones.

Of note, equol is a gut microbial catabolite of daidzein with higher bioavailability than other isoflavones. However, about half of the population is unable to produce equol because of the lack of equol‐producting microbiota in their gastrointestinal tract. ¹⁷ , ¹⁸ Thus, identifying the equol‐predicting microbial species to develop strategies to maximize equol production by modulating the gut microbiome may provide future therapeutic approaches to improve cardiometabolic health. Nonetheless, limited studies have dissected the equol‐predicting microbial taxa at the population level in China, and these studies were all cross‐sectional. ¹⁹ , ²⁰ , ²¹ Although previous culture‐based studies have identified several equol‐predicting microbial species, ²² , ²³ such as Adlercreutzi, Coprococcus, Alistipes, and Faecalibacterium, these studies may have not accounted for all microbes that contribute to the production of equol, given the known difficulties in culturing many of the microbes comprising the human gastrointestinal microbiome. ²⁴

Therefore, we repeatedly measured the urinary biomarkers of isoflavones, fecal microbiota, and cardiometabolic risk markers at baseline and after 1 year in 305 Chinese community‐dwelling adults. We aimed to identify gut microbiota associated with equol production and to investigate the associations of urinary isoflavones biomarkers (daidzein, glycitein, genistein, and equol) and equol‐predicting microbial species (EqMS) with cardiometabolic risk markers.

METHODS

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Study Population

The participants in the study were selected from the ALDS (Anhui Liver Diseases Study). The ALDS is an ongoing prospective cohort study started in 2020, enrolling approximately 3200 community‐dwelling adults aged ≥18 years in Lu’an, China. A multistage cluster sampling design was used in the ALDS to ensure the representativeness of the study population. Participants were followed with face‐to‐face interviews every 3 to 5 years to obtain updated information on their diet, lifestyle, and liver health‐related information. During 2021 and 2022, the ALVS (Anhui Lifestyle Validation Study) was performed to validate the questionnaires used in the ALDS and other cohorts in Anhui. The ALVS is a 1‐year study that recruited a subset of 754 participants from the ALDS and 138 participants from other cities in Anhui (Figure S1).

In this study, we used data from 754 participants from the ALDS/ALVS, in which 482 participants have completed the baseline survey and a 1‐year follow‐up interview. Among the 482 participants, 177 were excluded because of abnormal energy intake (<600 or >3500 kcal/d for women and <800 or >4200 kcal/d for men, n=79), missing data on urinary isoflavones (n=97), and history of hormone replacement therapy (n=1). After exclusion, 305 participants were included in the final analysis. All participants provided written informed consent, and the study protocol was approved by the ethics committee of Anhui Medical University (protocol number: 20210730).

Urinary Biomarkers Measurement

Participants were requested to provide first morning void urine samples at baseline and after 1 year of follow‐up. Urinary excretion of isoflavones including daidzein, glycitein, genistein, and equol were measured using liquid chromatography–tandem mass spectrometry. The quantification process used the chromatographic peak area internal standard method. Values under the detection limits were recorded as half of the detection limit. ²⁵ Urinary concentrations of isoflavones were corrected by urinary creatinine (micrograms per gram creatinine).

Covariate Measurement

A structured questionnaire was used to collect demographic information such as age, sex, income, marital status, educational level, and lifestyle habits including smoking, alcohol consumption, tea drinking, physical activities, history of chronic medical conditions, and medication use (eg, hypertension, dyslipidemia, diabetes). Physical activity was quantified as the metabolic equivalent tasks in hours per week. Hypertension and diabetes were identified if participants had been told by a health care professional that they had such diseases or took prescribed medications due to the diseases. Diabetes (a fasting plasma glucose ≥126 mg/dL) and hypertension (systolic blood pressure ≥140 or diastolic blood pressure ≥90 mm Hg) were also identified through laboratory test or physical examinations at baseline. Diet was repeatedly assessed at baseline and after 1 year using a 141‐item food frequency questionnaire. The Dietary Approaches to Stop Hypertension scores were calculated to assess overall diet quality. ²⁶

Fecal Sample Collection Microbiome Profiling

Participants provided fecal samples twice (1 year apart) 1 to 3 days after each urine sample collection. Fecal samples were collected at home using a commode specimen collection system and a stool collection container (Fisher Scientific) by the participants and were delivered to the nearest Community Health Center within 4 hours. Upon arrival, each sample was immediately stored in −80 °C freezers until nucleic acid extraction. To reduce the impact of probiotic or antibiotic exposure, we only included participants without probiotic or antibiotic use at least 1.5 months before fecal sample collection in the study. ²⁷

The fecal microbiome was profiled using 16S ribosomal RNA gene sequencing. The microbial community DNA was extracted using MagPure Stool DNA KF kit B (Magen, China). We quantified DNA with a Qubit Fluorometer by using Qubit dsDNA BR Assay kit (Invitrogen), and the quality was checked by running aliquot on 1% agarose gel. Variable regions V4 of bacterial 16S ribosomal RNA gene was amplified with degenerate polymerase chain reaction (PCR) primers. Both forward and reverse primers were tagged with Illumina adapter, pad, and linker sequences. PCR enrichment was performed in a 50 μL reaction containing a 30 ng template, fusion PCR primer, and PCR master mix. PCR cycling conditions were as follows: 95 °C for 3 minutes, 30 cycles of 95 °C for 45 seconds, 56 °C for 45 seconds, 72 °C for 45 seconds, and final extension at 72 °C for 10 minutes. The PCR products were purified using Agencourt AMPure XP beads and eluted in Elution buffer. Libraries were qualified by the Agilent Technologies 2100 bioanalyzer. The validated libraries were used for sequencing on an Illumina HiSeq 2500 platform (BGI, Shenzhen, China) and generating 2×250 bp paired‐end reads. Species with an average relative abundance >0.001% were selected for the analysis. A total of 87 genera were identified.

Cardiometabolic Risk Markers and Arterial Stiffness Measurement

Fasting blood samples were repeatedly collected at the same time points when urine samples were collected. Serum glucose, triglycerides, total cholesterol, low‐density lipoprotein cholesterol, and high‐density lipoprotein cholesterol (HDL‐C) were quantified by electrochemiluminescence. Systolic blood pressure and diastolic blood pressure were measured 3 times with the participants in a seated position after resting for at least 5 minutes, and the mean values of the previous parameters were used in the present study. Waist circumference was measured twice, 1 year apart, by a professional investigator. Arterial stiffness was assessed by measuring brachial‐ankle pulse wave velocity after 1 year of follow‐up. The method for measuring brachial‐ankle pulse wave velocity has been detailed elsewhere, ²⁸ and brachial‐ankle pulse wave velocity ≥1630 cm/s was defined as arterial stiffness. ²⁹ Considering multiple testing, we also calculated 2 indices by aggregating cardiometabolic risk parameters into composite scores to reflect the overall cardiometabolic risk. Plasma atherogenic index is defined as the log‐transformed value of the ratio of triglycerides to HDL‐C. The metabolic syndrome score was calculated based on the following parameters: systolic blood pressure, diastolic blood pressure, HDL‐C, triglycerides, blood glucose, and waist circumference, which has been detailed elsewhere. ³⁰

Statistical Analysis

Characteristics of participants according to urinary equol level at baseline and after 1‐year were compared using 1‐way ANOVA or χ² tests. A linear mixed‐effects model (LMM) was used to estimate the associations between levels of urinary isoflavones and cardiometabolic risk markers, adjusting for age, sex, education, family income, body mass index, menopausal status, Dietary Approaches to Stop Hypertension score, total energy intake, physical activity, medication use due to hypertension, dyslipidemia, or diabetes, alcohol drinking, smoking, and batch effect (first, second). Urinary isoflavones and serum markers (glucose, triglycerides, total cholesterol, low‐density lipoprotein cholesterol, and HDL‐C) were log‐transformed to normalize the distribution in the LMM. We reported percentage differences and 95% CIs in cardiometabolic risk markers for each 1 μg/g creatinine increase in urinary isoflavones or 1‐point increase in EqMS score. Multivariate logistic regression was used to assess the association between the mean urinary equol concentration at 2 measurements and the risk of arterial stiffness. The adjusted odds ratios (aORs) and 95% CIs for the risk of arterial stiffness were calculated by comparing the extreme tertiles of urinary equol levels.

The diversity within the gut microbiota was assessed using metrics including the abundance‐based coverage estimator, Chao1, Shannon, and Simpson indices to measure α‐diversity. Additionally, we used principal coordinate analysis and analysis of similarities based on the weighted UniFrac metric to evaluate β‐diversity. We used LMM to identify bacterial genera that can predict equol levels using repeated measures of fecal microbiome and urinary equol at baseline and after 1 year. The relative abundances of taxonomic features were transformed by the arcsin square root before association analyses. The models were adjusted for age, sex, body mass index, total energy intake, physical activity, family income, smoking, alcohol consumption, education, Dietary Approaches to Stop Hypertension score, and batch effect. To correct for multiple testing, the Benjamini‐Hochberg false discovery rate (FDR) method was applied. Bacterial genera associated with equol at an FDR <0.05 were considered as statistically significant.

To reflect the overall equol‐predicting capacity of host gut microbiota, we constructed an EqMS score. This score summarizes the relative abundance of microbial genera that were positively associated with equol levels in the LMM analyses at an P _FDR<0.05. Microbial genera detected in ≥50% of the samples were categorized as high (median levels or high) or low (less than the median level) according to the median relative abundance, whereas microbial genera detected in <50% of the samples were dichotomously categorized according to the presence or absence of the genus. We assigned 1 point for higher abundance or presence of genera or 0 otherwise. The score of all genera were then summed to calculate a total score. We investigated the association between EqMS score and cardiometabolic risk markers. We also created a weighted EqMS score in the sensitivity analysis, which was weighted by regression coefficients (β) derived from the aforementioned LMM that identified equol‐predicting bacterial genera, and summed to calculate a weighted EqMS score.

Although we have an a priori hypothesis that high urinary isoflavones (biomarkers of soy isoflavones), particularly equol and equol‐predicting microbial species, may improve cardiometabolic health, a sensitivity analysis was conducted to correct P values (P _FDR) considering multiple comparisons. We also conducted a sensitivity analysis to investigate the weighted EqMS score in relation to cardiometabolic risk. Statistical analyses were performed using R software version 4.2.0.

RESULTS

Participant Characteristics

The analysis encompassed 305 participants including 138 men and 167 women with a median age of 50 years (interquartile range, 37–59 years). The median urinary equol concentrations in the population was 87.2 μg/g creatinine (interquartile range, 6.7–580.2 μg/g creatinine). Individuals with higher urinary equol concentrations at baseline were younger and had a lower prevalence of diabetes and hypertension (Table 1).

Table 1.

Characteristics of 305 Participants by Tertiles of Urinary Equol Concentration at Baseline and After 1‐Year of Follow‐Up in Huoshan, China

Characteristics	Urinary equol at baseline			Urinary equol after 1 year of follow‐up
	Tertile 1 (n=101)	Tertile 2 (n=102)	Tertile 3 (n=102)	Tertile 1 (n=101)	Tertile 2 (n=102)	Tertile 3 (n=102)
Equol concentration, μg/g creatinine	0.1 (0.1–1.1)	33.8 (15.2–101.7)	1213.0 (490.4–2587.3)	0.6 (0.1–2.5)	21.4 (9.1–58.2)	725.7 (377.8–1582.9)
Age, y* , †	50 (42–58)	53 (39–63)	44 (34–57)	50 (37–58)	52 (40–63)	53 (37–60)
Female, %	57.4	51.0	55.9	50.5	57.8	54.9
BMI, kg/m2	25.1±3.8	24.3±3.3	24.2±3.7	24.7±3.4	23.9±3.5	23.9±3.5
Physical activities, METs, h/wk	153.2 (95.3–190.0)	144.6 (101.2–204.3)	139.5 (95.1–204.5)	138.0 (105.9–184.5)	147.2 (92.6–192.6)	152.1 (87.8–233.0)
Total energy intake, kcal/d	2257 (1800–2855)	1965 (1555–2795)	2139 (1706–2717)	1897 (1557–2477)	1959 (1582–2567)	1835 (1513–2412)
Smoking, %
Never smokers	73.27	64.71	78.22	67.33	74.51	63.73
Former smokers	6.93	9.80	3.96	10.89	5.88	10.78
Current smokers	19.80	25.49	17.82	21.78	19.61	25.49
Alcohol drinking, %
Never drinkers	79.59	82.83	82.47	83.00	81.19	83.17
Former drinkers	4.08	1.01	2.06	2.00	5.94	0.99
Current drinkers	16.33	16.16	15.46	15.00	12.87	15.84
Household per capita income, %
<5000 yuan	15.46	16.00	12.37	8.08	9.09	11.22
5000–10 000 yuan	20.62	25	19.59	14.14	19.19	11.22
10 000–20 000 yuan	28.87	23.00	25.77	27.27	24.24	26.53
>20 000 yuan	35.05	36.00	42.27	50.51	47.47	51.02
Education level, %
Unschooled	8.91	19.61	13.73	8.91	14.71	14.71
Primary school	33.66	25.49	22.55	25.74	26.47	25.49
Junior high school or above	57.43	54.90	63.73	65.35	58.82	59.80
Hypertension, %	50.50	52.90	42.20	49.00	52.94	43.14
Diabetes*, %	12.90	15.70	4.90	12.00	17.65	13.73
Dyslipidemia*, %	52.50	45.10	35.30	50.00	43.14	34.31
DASH score	24 (22–27)	24 (22–26)	24 (22–27)	24 (21–26)	24 (21–26)	25 (22–27)

Values are standardized to the age distribution of the study population. Continuous variables are expressed as the mean±SD or median (interquartile range) according to the distribution of the variables, whereas categorical variables are presented as percent. P values were calculated from the 1‐way ANOVA or Kruskal–Wallis test for continuous variables and χ² test for categorical variables. Values of polytomous variables may not sum to 100% due to rounding. BMI indicates body mass index; DASH, Dietary Approaches to Stop Hypertension; and METs, metabolic equivalent tasks.

^*Urinary equol at baseline: P value<0.05.

^† Value is not age adjusted.

Urinary Isoflavones and Cardiometabolic Risk

Each 1 μg/g creatinine decrease in urinary equol concentrations was associated with 1.47% (95% CI, −2.52 to −0.41; P=0.007) decrease in triglycerides and 0.61% increase in HDL‐C (95% CI, 0.07–1.14; P=0.025). In addition, urinary equol was inversely associated with metabolic syndrome score and plasma atherogenic index, with the percentage differences of −0.96% (95% CI, −1.50 to −0.42; P=0.001), and −3.32% (95% CI, −5.46 to −1.13; P=0.003), respectively. We did not find any statistically significant association between other urinary isoflavones and cardiometabolic risk markers (Table 2, Figure S2). The participants with the highest tertile of urinary equol levels had lower risk of arterial stiffness compared with those in the lowest tertile (aOR, 0.28 [95% CI, 0.09–0.90]; P _trend=0.036).

Table 2.

Percentage Difference and 95% CI for the Repeated‐Measures Associations Between Urinary Isoflavone Metabolites and Cardiometabolic Risk Markers in Huoshan, China (N=610)

Cardiometabolic risk markers	Percentage difference (95% CI)
	Equol	P value	Daidzein	P value	Genistein	P value	Glycitein	P value
SBP	−0.17 (−0.42 to 0.08)	0.177	−0.02 (−0.60 to 0.55)	0.931	−0.08 (−0.58 to 0.42)	0.739	−0.13 (−0.53 to 0.27)	0.523
DBP	−0.11 (−0.37 to 0.15)	0.401	0.05 (−0.52 to 0.62)	0.875	0.07 (−0.43 to 0.57)	0.780	−0.03 (−0.43 to 0.37)	0.871
TC	0.04 (−0.36 to 0.44)	0.840	−0.37 (−1.29 to 0.56)	0.435	−0.03 (−0.85 to 0.79)	0.937	−0.15 (−0.80 to 0.50)	0.642
Triglycerides	−1.47 (−2.52 to −0.41)	0.007	−1.54 (−3.88 to 0.85)	0.204	−1.28 (−3.34 to 0.83)	0.232	−0.64 (−2.30 to 1.06)	0.458
HDL‐C	0.61 (0.07 to 1.14)	0.025	0.83 (−0.46 to 2.15)	0.208	0.61 (−0.53 to 1.75)	0.294	0.55 (−0.35 to 1.47)	0.232
LDL‐C	−0.31 (−0.85 to 0.22)	0.250	−0.10 (−1.39 to 1.21)	0.884	0.05 (−1.08 to 1.20)	0.928	0.46 (−0.96 to 0.86)	0.913
Glucose	−0.03 (−0.41 to 0.35)	0.890	0.44 (−0.49 to 1.37)	0.354	0.36 (−0.45 to 1.17)	0.385	0.58 (−0.07 to 1.22)	0.079
AIP	−0.96 (−1.50 to −0.42)	0.001	−1.11 (−2.32 to 0.13)	0.079	−0.98 (−2.06 to 0.10)	0.076	−0.59 (−1.45 to 0.28)	0.184
MSs	−3.32 (−5.46 to −1.13)	0.003	−3.79 (−8.70 to 1.40)	0.149	−5.34 (−9.66 to −0.82)	0.021	−2.89 (−6.38 to 0.74)	0.117

Linear mixed‐effects model with repeated measures was adjusted for age (continuous), sex (women, men), education (unschooled, primary school, junior high school or above), annual household per capita income (<5000, 5000–10 000, 10 000–20 000, ≥20 000 yuan), body mass index (<28.0 and ≥28.0 kg/m²), drinking status (never, past, current drinking), smoking status (never, past, current smoking), total energy intake (kcal/d, continuous), physical activity (metabolic equivalent tasks, h/wk, continuous), menopausal status (no, yes, women only), Dietary Approaches to Stop Hypertension diet index (continuous), batch effect (first, second). Additionally adjusted for diabetes medication use (no, yes), cholesterol‐lowering medication use (no, yes), hypertension medication use (no, yes) when analyzing corresponding risk markers. Values were log‐transformed to approximate a normal distribution of the residuals, and percentage difference was calculated as (e^β−1)×100%. AIP indicates plasma atherogenic index; DBP, diastolic blood pressure; HDL‐C, high‐density lipoprotein cholesterol; LDL‐C, low‐density lipoprotein cholesterol; MSs, metabolic syndrome score; SBP, systolic blood pressure; and TC, total cholesterol.

Equol, Fecal Microbiome, and Cardiometabolic Risk

Individuals with higher urinary equol concentrations exhibited higher gut microbial α‐diversity compared with those with lower urinary equol concentrations (P<0.01 for all α‐diversity metrics; Figure S3). The gut microbial β‐diversity between participants with high and low urinary equol was significantly differed (R=0.116, P=0.001; Figure S4). The α‐diversity in the metabolic syndrome group was lower than that in the nonmetabolic syndrome group (P<0.01 for all α‐diversity metrics; Figure S5), whereas the β‐diversity differed between the 2 groups (R=0.129, P=0.002; Figure S6).

Of the 87 identified bacterial genera, the relative abundance of 21 genera was positively associated with urinary equol concentrations at P _FDR<0.05 (Table 3). These genera belong to the following 4 phyla families: Firmicutes (Intestinimonas, Mogibacterium, Mitsuokella, Coprococcus, Sporobacter, Oscillibacter, Eubacterium, Ruminococcus, Faecalibacterium), Actinobacteria (Adlercreutzia), Proteobacteria (Oxalobacter, Desulfovibrio, Bilophila), and Bacteroidetes (Odoribacter, Butyricimonas, Coprobacter, Barnesiella, Alistipes, Paraprevotella, Alloprevotella, Prevotella). All of these genera were included in the development of the EqMS score. We did not find any genera whose relative abundance was statistically associated with metabolic syndrome after correcting for the P values.

Table 3.

Percentage Differences in Urinary Equol Concentrations for Each 1% Increase in the Relative Abundance of the Bacterial Genus Among 485 Participants in Huoshan, China

Phylum; class; order; family; genus	Bacterial genera at baseline		Bacterial genera after 1 year of follow‐up		Percentage difference (95% CI)	P value	P FDR value
	Prevalence (%)	Average abundance (%)	Prevalence (%)	Average abundance (%)
Firmicutes; Clostridia; Clostridiales; Ruminococcaceae; Intestinimonas	41.94	0.02	…	…	140.0 (55.9–269.7)	<0.001	0.001
Actinobacteria; Actinobacteria; Coriobacteriales; Coriobacteriaceae; Adlercreutzia	55.24	0.02	50.63	0.02	114.0 (56.2–193.4)	<0.001	<0.001
Proteobacteria; Betaproteobacteria; Burkholderiales; Oxalobacteraceae; Oxalobacter	…	…	30.80	0.02	106.4 (33.5–219.2)	0.001	0.007
Bacteroidetes; Bacteroidia; Bacteroidales; Porphyromonadaceae; Odoribacter	70.97	0.09	64.14	0.04	70.1 (39.5–107.6)	<0.001	<0.001
Firmicutes; Clostridia; Clostridiales; Clostridiales; Mogibacterium	…	…	29.54	0.02	66.4 (12.4–146.3)	0.011	0.042
Bacteroidetes; Bacteroidia; Bacteroidales; Porphyromonadaceae; Butyricimonas	57.66	0.08	49.37	0.03	58.7 (31.3–91.8)	<0.001	<0.001
Firmicutes; Negativicutes; Selenomonadales; Veillonellaceae; Mitsuokella	4.84	0.04	3.80	0.02	38.1 (9.9–73.5)	0.005	0.028
Bacteroidetes; Bacteroidia; Bacteroidales; Porphyromonadaceae; Coprobacter	16.53	0.02	12.24	0.03	33.6 (6.8–67.3)	0.011	0.042
Firmicutes; Clostridia; Clostridiales; Lachnospiraceae; Coprococcus	75.81	0.19	83.97	0.26	31.4 (19.4–44.6)	<0.001	<0.001
Firmicutes; Clostridia; Clostridiales; Ruminococcaceae; Sporobacter	61.29	0.16	63.29	0.27	28.6 (18.4–39.7)	<0.001	<0.001
Bacteroidetes; Bacteroidia; Bacteroidales; Porphyromonadaceae; Barnesiella	45.56	0.21	42.19	0.11	28.4 (16.3–41.7)	<0.001	<0.001
Proteobacteria; Deltaproteobacteria; Desulfovibrionales; Desulfovibrionaceae; Desulfovibrio	39.11	0.14	44.73	0.32	19.8 (10.6–29.8)	<0.001	<0.001
Proteobacteria; Deltaproteobacteria; Desulfovibrionales; Desulfovibrionaceae; Bilophila	78.63	0.18	80.17	0.19	16.7 (4.1–30.8)	0.008	0.034
Bacteroidetes; Bacteroidia; Bacteroidales; Rikenellaceae; Alistipes	85.48	0.98	84.81	0.54	12.6 (7.0–18.6)	<0.001	<0.001
Bacteroidetes; Bacteroidia; Bacteroidales; Prevotellaceae; Paraprevotella	37.5	0.30	35.44	0.25	11.3 (3.4–19.8)	0.005	0.023
Bacteroidetes; Bacteroidia; Bacteroidales; Prevotellaceae; Alloprevotella	9.68	0.27	10.13	0.30	9.3 (1.8–17.3)	0.014	0.050
Firmicutes; Clostridia; Clostridiales; Eubacteriaceae; Eubacterium	62.50	0.25	67.09	0.90	8.8 (3.4–14.5)	0.001	0.007
Firmicutes; Clostridia; Clostridiales; Ruminococcaceae; Oscillibacter	90.73	0.36	91.14	0.71	8.5 (2.3–15.1)	0.007	0.029
Firmicutes; Clostridia; Clostridiales; Ruminococcaceae; Ruminococcus	84.68	1.23	84.39	1.28	7.5 (3.0–12.2)	0.001	0.006
Firmicutes; Clostridia; Clostridiales; Ruminococcaceae; Faecalibacterium	100.00	7.71	98.73	8.98	4.7 (2.5–6.9)	<0.001	<0.001
Bacteroidetes; Bacteroidia; Bacteroidales; Prevotellaceae; Prevotella	89.52	13.99	81.01	10.64	1.7 (0.7–2.8)	0.002	0.009

P values were estimated from linear mixed‐effects model after adjustment for age (continuous), sex (women, men), education (unschooled, primary school, junior high school or above), annual household per capita income (<5000, 5000–10 000, 10 000–20 000, >20 000 yuan), body mass index (<28.0 and ≥28.0 kg/m²), drinking status (never, past, current drinking), smoking status (never, past, current smoking), total energy intake (kcal/d, continuous), physical activity (metabolic equivalent tasks, h/wk, continuous), Dietary Approaches to Stop Hypertension diet index (continuous), batch effect (first, second). All taxa with FDR‐adjusted P<0.05 are included in the table. FDR indicates false discovery rate.

We found an inverse association between the EqMS score and triglyceride levels, with the percentage differences of −1.48% (95% CI, −2.89 to −0.05; P=0.042). Per 1‐point increase in EqMS score was associated with a 0.85% (95% CI, −1.59 to −0.11; P=0.025) decrease in plasma atherogenic index. In addition, EqMS score showed a protective association with systolic blood pressure (P=0.052) and metabolic syndrome score (P=0.098) with borderline statistical significance (Table 4). The aOR of arterial stiffness by comparing highest with the lowest tertile of the EqMS score was 0.27 (95% CI, 0.08–0.88; P _trend=0.031; Figure S7).

Table 4.

Percentage Difference and 95% CI for the Repeated‐Measures Associations Between Equol‐Predicting Microbial Score and Cardiometabolic Risk Markers (N=485)

Outcome variables	Percentage difference (95% CI)	P value
SBP	−0.32 (−0.64 to 0.00)	0.052
DBP	−0.27 (−0.61 to 0.06)	0.113
TC	0.06 (−0.46 to 0.59)	0.826
Triglycerides	−1.48 (−2.89 to −0.05)	0.042
HDL‐C	−0.09 (−0.77 to 0.58)	0.783
LDL‐C	0.23 (−0.46 to 0.93)	0.508
Glucose	−0.07 (−0.57 to 0.43)	0.779
AIP	−0.85 (−1.59 to −0.11)	0.025
MSs	−2.37 (−5.11 to 0.44)	0.098

Linear mixed‐effects model was adjusted for age (continuous), sex (women, men), education (unschooled, primary school, junior high school or above), household per capita income (<5000, 5000–10 000, 10 000–20 000, >20 000 yuan), body mass index (<28.0 and ≥28.0 kg/m²), drinking status (never, past, current drinking), smoking status (never, past, current smoking), total energy intake (kcal/d, continuous), physical activity (metabolic equivalent tasks, h/wk, continuous), Dietary Approaches to Stop Hypertension diet index (continuous), batch effect (first, second). Additionally adjusted for diabetes medication use (no, yes), cholesterol‐lowering medication use (no, yes), hypertension medication use (no, yes) when analyzing corresponding risk markers. Values were log‐transformed to approximate a normal distribution of the residuals, and percentage difference was calculated as (e^β−1)×100%. AIP indicates plasma atherogenic index; DBP, diastolic blood pressure; HDL‐C, high‐density lipoprotein cholesterol; LDL‐C, low‐density lipoprotein cholesterol; MSs, metabolic syndrome score; SBP, systolic blood pressure; and TC, total cholesterol.

In sensitivity analysis, upon FDR correction, the protective associations of urinary equol with triglycerides and HDL‐C remained, whereas other observed associations became nonsignificant (data not shown). When using weighted EqMS score, the protective association between EqMS and cardiometabolic risk remained (Table S1, Figure S8).

DISCUSSION

In this community‐based study, we repeatedly measured urinary isoflavones, gut microbiota, and cardiometabolic risk markers at baseline and after 1 year. We found that equol, a gut microbiota‐derived metabolite of daidzein, was favorably associated with several cardiometabolic risk markers in Chinese community‐dwelling adults. We did not find any statistically significant associations for other urinary isoflavones. We identified 21 bacterial genera whose relative abundance was associated with higher urinary concentrations of equol and constructed a microbial species score to reflect the overall equol‐predicting potential. This score showed an inverse association with triglycerides and the risk of arterial stiffness. The protective association remained when we used a weighted microbial species score. Our findings may aid in developing new strategies to improve cardiometabolic health through enhancing equol‐producing capacity by modulating the host gut microbiome.

Epidemiological evidence on the association between intake of soy isoflavones and cardiometabolic health remains inconclusive. A meta‐analysis of randomized controlled trials reported a significant improvement in triglycerides and low‐density lipoprotein cholesterol after supplementation of soy isoflavones, ⁸ whereas a cohort study found a null association between habitual intake of isoflavones and serum lipids. ¹² Differences in dietary estimation and interindividual variability in isoflavone absorption and metabolism may partly explain such inconsistency. ³ In this study, higher urinary equol, but not other isoflavones, was associated with lower triglyceride and higher HDL‐C levels. Similarly, Zheng et al noted a lower prevalence of dyslipidemia in equol producers. ¹⁹ Animal experiments revealed that atherosclerotic mice treated with equol exhibited diminished atherosclerotic lesions, along with notable reductions in triglycerides, total cholesterol, and low‐density lipoprotein cholesterol levels and significant increases in HDL‐C levels, ³¹ which are in line with our study. Although no associations of cardiometabolic risk markers with other urinary isoflavones were found, this is consistent with previous findings. ¹⁶ This finding highlights the increased biological activity of equol when compared with other urinary isoflavones.

This metabolite boasts a longer half‐life and has greater biological activity than its counterparts. Numerous studies have reported its potential in regulating hormone levels and mitigating oxidative stress, thereby exerting diverse benefits on the body. ²² , ²³ , ³² The favorable association between equol and cardiometabolic risk markers could be biologically plausible though the following pathways. First, as a polyphenol, the hydroxyl group in equol can serve as a hydrogen donor for free radicals, thereby reducing the number of free radicals. ³² Second, equol can exert estrogenic or antiestrogenic effects by binding to the estrogen receptor (estrogen receptors ER‐α and ER‐β) ³³ and indirectly regulate glucose and lipid metabolism through lipogenesis, lipolysis, and adipogenesis. Ricketts et al found that equol also works by activating peroxisome proliferator‐activaed receptor‐α, which leads to decreased triglyceride concentrations by increasing fatty acid oxidation in the liver or skeletal muscle. ³⁴ Moreover, results from in vitro studies indicated that equol contributes to vascular relaxation through the production of vasodilators like nitric oxide. ³⁵ We also observed an inverse association between the equol level and arterial stiffness, which was in line with the above mechanistic findings.

Interestingly, the proportion of individuals with the ability to produce equol in our study population (ie, equol producers ³⁶ ) was 70% higher than that in Western populations (20%–30%). ¹⁸ The differences among subjects in equol‐producting capacity could be largely determined by specific gut microbiota involved in producing equol. ³⁷ In our study, we identified microbial species, including Adlercreutzia, which is a well‐documented gut microbe involved in equol production. ³⁸ Specific genera, including Coprococcus, Alistipes, and Faecalibacterium, showed a strong positive connection with equol production, which have been reported in other studies. ²¹ , ³⁹ Consistent with our results, several intervention studies observed that the relative abundance of bacterial genera, including Ruminococcus, Desulfovibrio, Alistipes, and Odoribacter, increased after consuming soy foods or soy isoflavones. ⁴⁰ , ⁴¹ Moreover, equol producers have a higher abundance of the equol‐producting bacteria Adlercreutzia and Coprococcus. ¹⁹ , ²¹ In addition, we found several genera associated with urinary excretion of equol, including Intestinimonas, Mogibacterium, and Butyricimona, which have not been reported in previous studies. In summary, our study replicated previous studies at the genus level on equol‐predicting microbial taxa and dissected new bacterial genera that may contribute to equol production in a Chinese population.

In addition to gut microbiota, other factors including age, sex, dietary habits, and genetic polymorphisms may also affect the equol‐producing status among individuals. ³⁷ For example, Asian populations generally begin to consume soy foods in early life, have higher intake levels of isoflavones and fiber, and have a higher ratio of carbohydrates to total energy in the diet and thus exhibit a higher prevalence of equol producers than that in Western populations, ¹⁸ whereas short‐term supplementation of soy foods fails to alter equol‐production status. ⁴² , ⁴³ This suggests that equol production status may also depend on long‐term habits of consuming soy foods. Additionally, studies have found that the equol producer phenotype is associated with individual intake of different isoflavone components and their compositions. ⁴⁴ We therefore adjusted for these factors when we identified the potential gut microbiota involved in producing equol and investigated their association with cardiometabolic risk, and found a protective association between equol or equol‐predicting microbial species and cardiometabolic risk parameters. This observation further suggests that the gut microbiota may improve cardiometabolic health by promoting the production of equol.

Our study has several strengths including the use of internal exposure (ie, urinary isoflavones) and the repeated measurement of urinary isoflavones, gut microbiota, and cardiometabolic risk markers. Our study also has several limitations. First, despite the 1‐year data with repeated measures, we were unable to assess the long‐term effects of soy isoflavone intake and gut microbiota changes on cardiometabolic health given the 1‐year follow‐up period. Second, we cannot entirely rule out the possibility of residual confounding because of unmeasured or inaccurately measured covariates. Third, our study was conducted in a Chinese population, which might limit its applicability to other populations with different genetic backgrounds and dietary habits. Fourth, compared with 24‐hour urine samples, the use of first morning spot urine might not accurately reflect the urinary excretion of isoflavone. However, there is a strong correlation between measures from 24‐hour urine and first morning void urine samples, with correlation coefficients of 0.70 for genistein, 0.81 for genistein, 0.84 for daidzein, and 0.93 for equol. ⁴⁵ Last, our study could be enhanced with the use of deeper sequencing (eg, metagenomics) to evaluate fecal microbial gene expression. Therefore, large‐scale cohort studies with deeper sequencing and longer follow‐up periods are warranted.

In summary, we replicated previous reports at the genus level on equol‐predicting microbial species and found several novel bacterial genera that could be involved in equol production in a Chinese population, which needs further investigation. We showed that high urinary excretion of equol and equol‐predicting microbial species are favorably associated with cardiometabolic risk markers. These findings, if validated, will contribute to developing personalized nutritional prevention approaches for cardiometabolic diseases by targeting gut flora.

Sources of Funding

This work was supported by the National Natural Science Foundation of China (82373673 and 82103796), research funds of the Center for Big Data and Population Health of the Institute of Health Metrics (JKS2022018), grants from Anhui Medical University (2021xkjT007), and Postgraduate Innovation Research and Practice Program of Anhui Medical University (YJS20230048 and YJS20230150).

Disclosures

None.

Supporting information

Table S1

Figures S1–S8