Temporal relationship among adiposity, gut microbiota, and insulin resistance in a longitudinal human cohort

This study was based on the Guangzhou Nutrition and Health Study (GNHS), involving 4048 Chinese participants aged 40–75 years who lived in Guangzhou, China, for at least 5 years and were recruited between 2008 and 2013. Stool samples from 741 GNHS participants were collected between 2014 and 2017 (as baseline for the current study). Among them, stool samples were repeatedly collected in 505 participants between 2018 and 2019 (a follow-up collection). The median follow-up time was 3.15 years. We excluded participants with missing information on BMI at baseline or follow-up (n = 37), missing follow-up time (n = 2), or who had diabetes medications at baseline or follow-up (n = 40). Finally, 426 participants remained for subsequent analysis. The study design of this study is showed in Fig. 1A. This study was approved by the Ethics Committee of the School of Public Health at Sun Yat-sen University and Ethics Committee of Westlake University. All participants involved in this study provided written informed consent.

Anthropometric parameters, including height, weight, and waist circumference (WC), were measured by trained staff on site, and BMI was calculated as weight (kg)/height (m)². The information of sociodemographic, lifestyle covariates, and medication use was collected using a structured questionnaire during face-to-face interviews. Physical activity was assessed as total metabolic equivalent for task (MET) hours per day by a validated physical activity questionnaire which included 19 items [14]. Total energy intake was calculated based on the habitual dietary intakes over the past 12 months, which were collected using a validated food frequency questionnaire including 79 items [15]. Bristol stool score was immediately recorded in the stool sampler by the participants.

Fasting venous blood samples were drawn at both baseline and follow-up. Fasting insulin was measured by electrochemiluminescence immunoassay using a Roche Cobas 8000/e602 immunoanalyzer (Roche Diagnostics, Shanghai, China). Fasting glucose was measured through colorimetric methods using a Roche cobas 8000 c702 automated analyzer (Roche Diagnostics, Shanghai, China). Glycated hemoglobin A1c (HbA1c) was measured by high performance liquid chromatography using Bole D-10 Hemoglobin A1c Program on Bole D-10 Hemoglobin Testing System. Homeostasis model assessment of insulin resistance (HOMA-IR) was calculated based on fasting insulin and fasting glucose, which measures insulin resistance by using the mathematical model [16].

The same protocol for stool sample collection and processing was used at baseline and follow-up visits. Fecal samples from all participants were collected on the examination day at the site, and then stored at – 80 °C within 4 h. Fecal DNA extraction and metagenomic sequencing followed the same protocol at the two time points. Microbial DNA was isolated using the QIAamp DNA Stool Mini Kit (Qiagen, Hilden, Germany) based on the manufacturer’s instruction. DNA concentrations were determined by the Qubit quantification system (Thermo Scientific, Delaware, USA). The shotgun metagenome was sequenced through the Illumina HiSeq platform (Illumina Inc., CA, USA) using the 2 × 300-bp paired-end read protocol. We obtained an average of 44.6 million (min, 22.1 million; max, 100 million) paired-end raw reads for each sample. The detailed information on bioinformatics analysis of the metagenome data could be found in our previous paper [17]. PRINSEQ (version 0.20.447) was employed to filter the reads with low-quality scores, with the following filtering parameters: (1) trim the reads by quality score from the 5′ end and 3′ end with a quality threshold of 20; (2) remove read pairs when either read was < 60 bp, contained “N” bases or quality score mean bellow 30; and (3) deduplicate the reads. Reads that could be aligned to the human genome (H. sapiens, UCSC hg19) were removed (aligned with Bowtie2 v2.2.5 using –reorder –no-contain –dovetail) [18]. Taxonomic profiling of the metagenomic samples was performed using MetaPhlAn2 (version 2.6.02) which used a library of clade-specific markers to provide pan-microbial (bacterial, archaeal, viral and eukaryotic) quantification at the species level [19]. MetaPhlAn2 was run using default settings. Microbial species with a minimum detective relative abundance of 0.01% in at least 10% of the samples were included, which yielded 171 species. Functional profiling of the metagenomic samples was performed using HUMAnN2 (version 2.8.1) using default settings [20], which generated microbial pathways based on MetaCyc metabolic pathway database [21, 22]. We applied log-transformation for the relative abundances of both species and pathway features before subsequent analysis. Before log-transformation, we replaced the zeros with a very small value (0.00001 for gut microbiota data and 1e−09 for functional pathway data), which may not substantially change the distribution of the data.

We used a cross-lagged panel design to investigate the bidirectional relationship between BMI and gut microbial features (Fig. 1B). Cross-lagged path analysis is a form of path analysis that examines reciprocal, longitudinal relationships among a set of inter-correlated variables [23‐25]. This method tested the effect of baseline gut microbiota on subsequent BMI (ρ1 in Fig. 1B) and the effect of baseline BMI on subsequent gut microbiota (ρ2 in Fig. 1B) simultaneously, adjusted for autoregressive effects. Before performing the cross-lagged path analysis, we performed the linear regression analysis and got the residual of the baseline and follow-up values of BMI, adjusted for potential confounders, including age, sex, smoking status, alcohol status, education, income, physical activity, and total energy intake and then standardized the residual into Z-scores; gut microbial features were processed in the same manner with an additional adjustment for Bristol stool score. Pearson correlation coefficients of Z-transformed BMI and gut microbial features at baseline and follow-up were calculated, adjusted for the time interval (years) between two time points. The cross-lag path coefficients (ρ1 and ρ2) showed in Fig. 1B were estimated simultaneously based on the correlation matrix. All parameters in the cross-lagged path analysis were estimated through constructing structural equation model by R package lavaan (version 0.6–8) [26]. The validity of model fitting was evaluated by the standardized root mean square residual (SRMR) and comparative fit index (CFI) [27].

We performed a principal coordinates (PCo) analysis based on Bray-Curtis dissimilarity using metagenomic data at the species level, and the first two PCo (PCo1 and PCo2) were obtained to reflect the β-diversity of the gut microbiota. We calculated α-diversity indices (Observed species, Shannon index, Simpson index, Pielou’s evenness) based on the relative abundance of the species by using the R package vegan (version 2.5-7) [28]. We examined the temporal relationships between BMI and α-diversity, β-diversity, and individual microbes using the cross-lagged path analysis. The Benjamini-Hochberg (BH) method was used to control the false discovery rate (FDR). Given that high-dimensional tests were performed, associations with FDR < 0.25 were considered statistically significant for per-species test. The identified BMI-associated species were selected for subsequent analysis. The stratified analysis by sex for the association of BMI with the identified species was performed, and the heterogeneity in the effect sizes between females and males were tested using the Cochran-Q test [29]. We further did sensitivity analysis to investigate the temporal relationship between WC and gut microbiota. We assessed the prospective associations between dietary factors (vegetable intake, fruit intake, fish intake, red and processed meat intake, and dairy intake) and identified gut microbes using the multivariable linear regression models, adjusted for age, sex, BMI, smoking status, alcohol status, education, income, physical activity, total energy intake, Bristol stool score, time interval, and corresponding baseline microbe abundance. Each dietary factor was divided into higher and lower groups based on the median value.

To replicate the results from the above GNHS participants, we used the repeated-measured fecal metagenome data available 1 year apart from 43 healthy participants aged 18–40 years in the HMP cohort [9, 11]. We obtained the relative abundance of microbiota data through the R package curatedMetagenomicData (version 1.10.2) [30], and the corresponding phenotype data from the dbGaP (https://​dbgap.​ncbi.​nlm.​nih.​gov/​; study accession: phs000228.v4.p1) [31, 32]. The relative abundances of the species were log-transformed before formal analysis. Since BMI was only available at baseline, we assessed the prospective association of baseline BMI with follow-up microbes using multivariable linear regression models, adjusted for age, sex, race (white/not white), time interval, and corresponding baseline microbe abundance. We used the meta-analysis with a random effects model and inverse-variance weights to integrate the results from GNHS and HMP cohorts, and assessed the heterogeneity between them using I² and Cochran-Q test [29]. We considered the association between baseline BMI and follow-up microbes to be replicable if p_meta<0.05, p_{heterogeneity}>0.05, I²<50%, and the same direction of associations between the two cohorts. The meta-analysis and Cochran-Q test were conducted using the R package metafor (version 3.0-2) [33].

In the GNHS, to explore the impact of long-term weight change on gut microbiota, participants were divided into four weight change patterns: stable normal (n = 205), normal to adiposity (n = 23), adiposity to normal (n = 21), and stable adiposity (n = 156). Participants who were underweight at either time point were excluded (n = 21). Adiposity was defined as overweight or obesity in this study. According to the suggestion of Working Group On Obesity In China for Chinese populations [34], underweight, normal weight, overweight, and obesity were defined as BMI<18.5, 18.5≤BMI≤23.9, 24≤BMI≤27.9, and BMI≥28, respectively. The influence of weight change pattern on follow-up microbial features (α-diversity: Observed species, Shannon index, Simpson index, Pielou’s evenness; β-diversity: PCo1 and PCo2; and microbes) was assessed using multivariable linear regression models, adjusted for age, sex, smoking status, alcohol status, education, income, physical activity, total energy intake, Bristol stool score, time interval, and corresponding baseline microbial features. We only included BMI-associated microbes in the analysis of the association between weight change pattern and gut microbes. The stable normal group was served as the reference group when conducting the above analyses. The potential collinearity among covariates was assessed by variance inflation factor (VIF) using R function vif in R package car (version 3.0-12), and VIF > 10 indicates collinearity among variables. After fitting 16 multivariable linear regression models for microbial features (10 identified BMI-associated microbes plus Observed species, Shannon index, Simpson index, Pielou’s evenness, PCo1 and PCo2), we obtained 16 VIF for each covariate and the ranges of them were as follows: age (1.175–1.187), sex (1.486–1.525), smoking status (1.336–1.343), alcohol status (1.167–1.180), education (1.306–1.321), income (1.384–1.408), physical activity (1.068–1.074), total energy intake (1.130–1.137), Bristol stool score (1.086–1.124), time interval (1.064–1.077), corresponding baseline microbial features (1.029–1.102), which confirmed that there was no collinearity among covariates.

We explored the associations between the above identified species and HOMA-IR using linear mixed-effect models by R package lme4 (version 1.1-27.1) [35], adjusted for age, sex, smoking status, alcohol status, education, income, physical activity, and total energy intake. In the secondary analyses, we analyzed the association of identified species with fasting insulin, fasting glucose, and HbA1c using linear mixed-effect models, adjusted for the same covariates. We log-transformed insulin resistance-related phenotypes with skewed distribution (HOMA-IR, fasting insulin, and fasting glucose) before analysis. All phenotypes and microbes were then standardized into Z-scores. We also assessed whether dietary factors (vegetable intake, fruit intake, fish intake, red and processed meat intake, and dairy intake) contributed to the insulin resistance-related phenotypes by using multivariable linear regression models, adjusted for age, sex, BMI, smoking status, alcohol status, education, income, physical activity, total energy intake, time interval, and corresponding baseline levels of insulin resistance-related phenotypes. Each dietary factor was divided into higher and lower groups based on the median value.

To investigate whether there were potential mediation effects of gut microbes on the association between adiposity and insulin resistance, we performed a mediation analysis using the R package mediation (version 4.5.0) [36]. The baseline overweight/obesity status was served as the exposure, and follow-up gut microbes and insulin resistance-related phenotypes were served as mediators and outcomes, respectively (Fig. 1C). The corresponding baseline gut microbes and insulin resistance-related phenotypes were adjusted in the statistical models. The covariates included age, sex, smoking status, alcohol status, education, income, physical activity, total energy intake, Bristol stool score, and time interval. Before performing mediation analysis, we examined the prospective associations between weight group (underweight, normal weight, and adiposity) and insulin resistance-related phenotypes (HOMA-IR, fasting insulin, fasting glucose, and HbA1c) using multivariable linear regression models, adjusted for age, sex, smoking status, alcohol status, education, income, physical activity, total energy intake, time interval, and corresponding baseline levels of insulin resistance-related phenotypes.

One identified species Lachnospiraceae bacterium 3 1 57FAA CT1 belonged to Lachnospiraceae which is involved in the production of short-chain fatty acid (SCFA) through fermenting plant polysaccharides [37, 38]. In addition, Lachnospiraceae bacterium 3 1 57FAA CT1 has phylogenetic similarity to known butyrogenic gut bacteria [39]. The butyrate-producing pathway PWY-5022 (4-aminobutanoate degradation V) was extracted from the pathway data obtained from functional profiling of the metagenomic samples. We used a linear mixed-effect model to examine the association between Lachnospiraceae bacterium 3 1 57FAA CT1 and PWY-5022, adjusted for age, sex, smoking status, alcohol status, education, income, physical activity, total energy intake, and Bristol stool score. Additionally, the associations between Lachnospiraceae bacterium 3 1 57FAA CT1 and other pathways in addition to PWY-5022 were also assessed using linear mixed-effect models, adjusted for the same covariates described above. We only included microbial pathways with a minimum detective relative abundance of 0.01% in at least 10% of the samples for this analysis. Unless otherwise noted, FDR<0.05 was considered statistically significant in this study.

The mean (SD) age of these participants was 63.6 (5.4) years (31.9% men) when their first batch of stool samples were collected at baseline (Table 1). BMI was slightly increased from baseline to follow-up (p = 0.002), while WC was not statistically different between two time points (Additional file 1: Table S1). The prevalence of adiposity (overweight or obesity) for the baseline and follow-up is 41.5% and 42%, respectively. Insulin resistance-related phenotypes, including HOMA-IR, fasting insulin, HbA1c, and fasting glucose, were increased from baseline to follow-up (p<0.05; Additional file 1: Table S1).

The associations between the BMI and gut microbial α-diversity and β-diversity were not statistically significant in either direction (Additional file 1: Fig. S1A, Fig. S2A and Table S2). For the temporal relationship between BMI and individual gut microbes, we did not find any statistically significant associations for the path coefficients (ρ₁) from baseline gut microbes to follow-up BMI (Additional file 1: Fig. S1B and Table S3). For the path coefficients (ρ₂) from baseline BMI to follow-up gut microbes, ten species were identified (FDR<0.25; Fig. 2A and Additional file 1: Table S3). Among them, with the increase of BMI, the abundances of four species (Clostridium hathewayi, Parabacteroides unclassified, Lachnospiraceae bacterium 3 1 57FAA CT1, Lachnospiraceae bacterium 7 1 58FAA) were decreased, and the abundances of the other species (Megamonas hypermegale, Megamonas unclassified, Bacteroides caccae, Ruminococcus sp 5 1 39BFAA, Megasphaera unclassified, Adlercreutzia equolifaciens) were increased. Adlercreutzia equolifaciens belonged to Actinobacteria, Parabacteroides unclassified and Bacteroides caccae belonged to Bacteroidetes, and other species belonged to Firmicutes. When we used a more stringent cut-off value of 0.2 for FDR values, all identified ten microbes based on FDR<0.25 were also statistically significant (all FDR values less than 0.2 or very close to 0.2; Fig. 2A). All ten species had relatively high prevalence in our cohort (0.28–0.83; Additional file 1: Table S9). There were low to moderate correlations among these species (Additional file 1: Fig. S3). In addition, there were moderate correlations (from 0.12 to 0.46) of identified species between two time points (Additional file 1: Table S4). The stratified analysis by sex for the association of BMI with identified species suggested that there was no heterogeneity of the regression coefficients between females and males (FDR_{heterogeneity}>0.05; Additional file 1: Table S5) and there was a strong correlation between the regression coefficients for females and those for males (r = 0.815; Additional file 1: Fig. S4). The association of WC with gut microbes were similar compared with those of BMI with gut microbes, with correlation coefficients 0.88 and 0.99 for all gut microbial features and the identified microbes, respectively (Additional file 1: Fig. S5, Table S6 and S7). For the contributions of dietary factors (vegetable intake, fruit intake, fish intake, red and processed meat intake, and dairy intake) to the identified microbes, we only found that Ruminococcus sp 5 1 39BFAA was associated with fish intake (FDR<0.05; Additional file 1: Table S8), while other microbes were not correlated with dietary factors (FDR>0.05; Additional file 1: Table S8).

In the Human Microbiome Project (HMP) cohort, the prevalence of Megamonas hypermegale, Megamonas unclassified, and Megasphaera unclassified was very low (0.05, 0.06, and 0.09, respectively; Additional file 1: Table S9). Results of the prospective associations between baseline BMI and four species (Adlercreutzia equolifaciens, Parabacteroides unclassified, Lachnospiraceae bacterium 3 1 57FAA CT1, Lachnospiraceae bacterium 7 1 58FAA) were replicated in the HMP cohort, with a p-value for meta-analysis across the two studies < 0.05, p_{heterogeneity}>0.05, and I²<50% (Fig. 2B; Additional file 1: Table S9).

The prospective association of long-term weight change pattern with gut microbial β-diversity was statistically significant with reduced PCo2 comparing the normal to adiposity group with the stable normal group (Additional file 1: Fig. S1C). The long-term weight change pattern was not associated with any of the α-diversity indices (Additional file 1: Fig. S1C and Fig. S2B). Among microbes associated with the BMI, Megamonas unclassified was significantly higher in the normal to adiposity group compared with the stable normal group (FDR=0.01; Fig. 3A). Compared with the stable normal group, the abundances of Clostridium hathewayi and Lachnospiraceae bacterium 3 1 57FAA CT1 were significantly lower, and the abundances of Megamonas hypermegale, Megamonas unclassified, and Bacteroides caccae were higher in the stable adiposity group (FDR<0.05; Fig. 3B). The mean abundance of identified microbes was 0.13%, 0.03%, 0.24%, 2.04%, and 0.92% for Clostridium hathewayi, Lachnospiraceae bacterium 3 1 57FAA CT1, Megamonas hypermegale, Megamonas unclassified, and Bacteroides caccae, respectively.

Lachnospiraceae bacterium 3 1 57FAA CT1 [β = − 0.13, 95% CI (− 0.22, − 0.05), FDR = 0.014] and Clostridium hathewayi [β = − 0.1, 95% CI (− 0.18, − 0.02), FDR = 0.03] were negatively associated with HOMA-IR (Fig. 4 and Additional file 1: Table S10). In the secondary analyses, Lachnospiraceae bacterium 3 1 57FAA CT1 was negatively associated with fasting insulin [β = − 0.13, 95% CI (− 0.22, − 0.04), FDR = 0.037], and Megamonas unclassified had a positive association with HbA1c [β = 0.09, 95% CI (0.03, 0.15), FDR = 0.037] (Fig. 4 and Additional file 1: Table S10). For the contributions of dietary factors (vegetable intake, fruit intake, fish intake, red and processed meat intake, and dairy intake) to insulin resistance, we did not find any dietary factor that was prospectively associated with insulin resistance-related phenotypes (FDR > 0.05; Additional file 1: Table S11).

We first examined the associations between weight group (underweight, normal weight, and adiposity) and insulin resistance. We found that adiposity was positively associated with several insulin resistance-related phenotypes including HOMA-IR, fasting insulin, and HbA1c compared with the normal weight (FDR<0.05; Additional file 1: Table S12). The results of mediation analysis showed that the Lachnospiraceae bacterium 3 1 57FAA CT1 mediated the association of adiposity with HOMA-IR (17%, FDR = 0.024; Fig. 5A). Clostridium hathewayi did not have the mediation role in the association between adiposity and HOMA-IR (2.7%, FDR = 0.53; Fig. 5B). In the secondary analyses, the mediation effect of Lachnospiraceae bacterium 3 1 57FAA CT1 on the association between adiposity and fasting insulin was borderline significant (14.8%, FDR = 0.056; Fig. 5C). The mediation role of Megamonas unclassified on the association of adiposity with HbA1c was not statistically significant (6.1%, FDR = 0.354; Fig. 5D). As all average causal mediation effects (ACME) and average direct effect (ADE) were larger than zero, the mediation effects and direct effects in the overweight/obesity group were larger than those in the normal group for all above mediation analyses (Fig. 5). We further found that Lachnospiraceae bacterium 3 1 57FAA CT1 was associated with a higher level of the butyrate-producing pathway PWY-5022 [β = 0.12, p < 0.001; Additional file 1: Table S13). In addition, Lachnospiraceae bacterium 3 1 57FAA CT1 was also associated with several other microbial functional pathways, such as amino acid biosynthesis and degradation pathways (HISDEG-PWY, PWY-4981, PWY-5101, DAPLYSINESYN-PWY, PWY-5030, PWY-2942, PWY-5154, HSERMETANA-PWY, PWY-6630, PWY-5097, PWY-6629), anaerobic energy metabolism (PWY-7383), lipid metabolism (PWY-5667, PWY0-1319), and gluconeogenesis (PWY66-399) (FDR < 0.05; Additional file 1: Table S14).