Changes in the type 2 diabetes gut mycobiome associate with metformin treatment across populations

Abstract

The human gut teems with a diverse ecosystem of microbes, yet non-bacterial portions of that community are overlooked in studies of metabolic diseases firmly linked to gut bacteria. Type 2 diabetes mellitus (T2D) is associated with compositional shifts in the gut bacterial microbiome and the mycobiome, the fungal portion of the microbiome. However, whether T2D and/or metformin treatment underpins fungal community changes is unresolved. To differentiate these effects, we curated a gut mycobiome cohort spanning 1,000 human samples across five countries and validated our findings in a murine experimental model. We use Bayesian multinomial logistic normal models to show that T2D and metformin both associate with shifts in the relative abundance of distinct gut fungi. T2D is associated with shifts in the Saccharomycetes and Sordariomycetes fungal classes, while the genera Fusarium and Tetrapisipora most consistently associate with metformin treatment. We confirmed the impact of metformin on individual gut fungi by administering metformin to healthy mice. Thus, metformin and T2D account for subtle, but significant and distinct variation in the gut mycobiome across human populations. This work highlights for the first time that metformin can confound associations of gut fungi with T2D and warrants the need to consider pharmaceutical interventions in investigations of linkages between metabolic diseases and gut microbial inhabitants.

Importance: This is the largest to-date multi-country cohort characterizing the human gut mycobiome, and the first to investigate potential perturbations in gut fungi from oral pharmaceutical treatment. We demonstrate the reproducible effects of metformin treatment on the human and murine gut mycobiome and highlight a need to consider metformin as a confounding factor in investigations between type 2 diabetes mellitus and the gut microbial ecosystem.

Keywords: Type 2 diabetes mellitus; gut microbiome; gut mycobiome; metagenomics.

Fig 1

Characterization of the gut mycobiome from metagenomics data across studies. (A) The geographic location of each study. The size of the bubbles corresponds with the total sample size. (B) An Upset plot demonstrating shared genera between studies. The gray dot matrix shows which study-treatment category combination shared fungal genera and the blue bar shows the total number shared. (C–E) Barplots of compositional abundance at the fungal class level from each study, geographic region, and treatment category without adjustment for between-study differences. The map in panel A was created using the Datawrapper web browser.

Fig 2

Metabolic disease is associated with a shift in genera belonging to the Saccharomycetes class. Subjects with T2D-NOMET were compared to NORM (A), and markers of metabolic disease were compared; (B) fasting blood glucose (FBG; n = 784), (C) fasting plasma insulin (FPI; n = 697), and (D) body mass index (BMI; n = 766). In all panels, black triangles and line ranges show posterior mean and posterior 95% credible intervals of the aggregate data set. Colored diamond points show posterior means of individual studies. Arrows show increased fungal abundance. The dashed vertical line denotes zero association.

Fig 3

Metformin’s associations with gut fungi in T2D in (A) the total data set, (B, D) randomized clinical trials (RCTsC) representing short-term metformin treatment, and (C) cross-sectional studies representing putative long-term metformin treatment. (A–C) Individual gut fungal genera associated with metformin treatment. The black triangle and lines show posterior mean and 95% credible intervals of the aggregate data set. Colored diamond points show posterior means of individual studies. The arrows show increased relative abundance at baseline or T2D-NOMET (left) or during metformin (MET) treatment (right). The dashed vertical line denotes zero association. (D) Proportional changes in the gut mycobiome before and after short-term metformin treatment are shown in barplots. In three RCTs with time-series sampling, proportional relative abundance is shown at baseline and after starting metformin treatment for each study. Paired samples from the same individual are shown as neighboring horizontal bars of proportional relative abundance with the duration of metformin treatment shown above the barplot for each study. Taxa are colored as Class:Family, and less relatively abundant taxa are merged in the “other” category.

Fig 4

Metformin is associated with similar differences in the human and mouse mycobiomes compared to normoglycemic individuals. Black triangles and line ranges show the posterior mean and posterior 95% credible intervals of the aggregate data set. Colored diamond points show posterior means of individual studies. Red squares show the posterior means of the mouse study. The arrows show increased abundance in NORM subjects or mice with placebo treatment (left) or T2D-MET subjects and metformin-treated mice (right). The dashed vertical line denotes zero association.