Association of gut microbiota with glycaemic traits and incident type 2 diabetes, and modulation by habitual diet: a population-based longitudinal cohort study in Chinese adults

Abstract

Aims/hypothesis: The gut microbiome is mainly shaped by diet, and varies across geographical regions. Little is known about the longitudinal association of gut microbiota with glycaemic control. We aimed to identify gut microbiota prospectively associated with glycaemic traits and type 2 diabetes in a geographically diverse population, and examined the cross-sectional association of dietary or lifestyle factors with the identified gut microbiota.

Methods: The China Health and Nutrition Survey is a population-based longitudinal cohort covering 15 provinces/megacities across China. Of the participants in that study, 2772 diabetes-free participants with a gut microbiota profile based on 16S rRNA analysis were included in the present study (age 50.8 ± 12.7 years, mean ± SD). Using a multivariable-adjusted linear mixed-effects model, we examined the prospective association of gut microbiota with glycaemic traits (fasting glucose, fasting insulin, HbA_1c and HOMA-IR). We constructed a healthy microbiome index (HMI), and used Poisson regression to examine the relationship between the HMI and incident type 2 diabetes. We evaluated the association of dietary or lifestyle factors with the glycaemic trait-related gut microbiota using a multivariable-adjusted linear regression model.

Results: After follow-up for 3 years, 123 incident type 2 diabetes cases were identified. We identified 25 gut microbial genera positively or inversely associated with glycaemic traits. The newly created HMI (per SD unit) was inversely associated with incident type 2 diabetes (risk ratio 0.69, 95% CI 0.58, 0.84). Furthermore, we found that several microbial genera that were favourable for the glycaemic trait were consistently associated with healthy dietary habits (higher consumption of vegetable, fruit, fish and nuts).

Conclusions/interpretation: Our results revealed multiple gut microbiota prospectively associated with glycaemic traits and type 2 diabetes in a geographically diverse population, and highlighted the potential of gut microbiota-based diagnosis or therapy for type 2 diabetes.

Data availability: The code for data analysis associated with the current study is available at https://github.com/wenutrition/Microbiota-T2D-CHNS.

Keywords: Glycaemic traits; Gut microbiota; Longitudinal cohort; Type 2 diabetes.

Fig. 1

Region-discriminating gut microbiota and dietary habits. (a) Comparison of dietary habits among participants from Northern and Southern China (n = 2772). For each dietary factor, data are presented as scaled mean values (i.e. mean values divided by the corresponding maximum mean value of two regions). (b) Dissimilarities in gut microbial composition between participants from Northern and Southern China represented by a Bray–Curtis dissimilarity matrix and principal coordinate analysis. The p value was determined by 1000 permutations. The level of confidence for the ellipses was 85%. The values on the axes represent the variance of the gut microbial composition at the genus level explained by principal components PCoA1 and PCoA2. (c) The microbial genera-based classifier achieved a high performance in regional prediction. The genus-level taxonomic abundance was used as the predictive features for the LightGBM model to predict the probability for each participant of belonging to the Southern region. (d) Receiver operator characteristic curves classifying participants’ staple food preference. We used the region-discriminating genera as input for the LightGBM model to predict the staple food preference. Staple food preference was calculated as the ratio of wheat intake to rice intake. A ratio ≥1 was considered as a wheat preference, otherwise a rice preference was inferred. Here, missing values were imputed using strategies (single mean imputation and multiple imputation). AUC indicates a tenfold cross-validated AUC. The range shown by the AUC is the 95% CI of the receiver operator characteristic curves

Fig. 2

Prospective association between the gut microbiota and glycaemic traits. Prospective association of baseline gut microbiota with (a) fasting glucose, (b) HbA_1c, (c) fasting insulin and (d) HOMA-IR. A total of 1829 participants were included in this analysis. A linear mixed-effects model was used to examine the prospective association of gut microbiota with the glycaemic traits fasting glucose, HbA_1c, fasting insulin and HOMA-IR, adjusting for the baseline glycaemic traits, demographic, anthropometric and lifestyle confounders. We independently examined the gut microbiota/glycaemic trait association in the Northern and Southern populations, and combined the effect estimates from the two regions using random-effects meta-analysis. Associations are expressed as the difference in glycaemic traits (in SD units) per SD difference for each genus. Superscript letters (a to g) indicate that the marked gut microbial genera were associated with at least two glycaemic traits. A p value <0.05 was considered as statistically significant. No individual gut microbial genera were found to be associated with glycaemic traits after adjusting for multiple testing

Fig. 3

Association of HMI with incident type 2 diabetes and modulation by dietary and lifestyle factors. (a) HMI and type 2 diabetes incidence (n = 1829). Poisson regression was used to examine the association of baseline HMI (per SD unit) with incident type 2 diabetes, adjusted for demographic, anthropometric, dietary and lifestyle factors. Subgroup analyses stratified by geographic region, age group, sex, BMI level and urbanisation level (city or rural) were performed to test the robustness of the model. (b) Association of dietary and lifestyle factors with gut microbiota (n = 2772). Linear regression was used to estimate the difference in glycaemic trait-related gut microbiota or HMI (in SD units) per SD change for continuous dietary or lifestyle factors (per-category change for categorical dietary or lifestyle factors), with adjustment for the confounders and mutually adjusted for the other tested dietary or lifestyle factors. Red arrows indicate gut microbiota that were positively associated with glycaemic traits; green arrows indicate gut microbiota that were inversely associated with glycaemic traits. The Benjamini–Hochberg method was used to control the FDR. An FDR value <0.05 was considered statistically significant