The human gut microbiome: methodological issues in characterization and role in trimethylamine N-oxide formation

The human gut microbiome has been of great interest in recent years and is being increasingly incorporated into epidemiological research. While early work was primarily interested in characterizing the gut microbiota within healthy individuals, studies have gradually shifted focus towards identifying associations with disease risk and outcomes. The gut microbiome has already been linked to numerous diseases including colorectal cancer and liver disease, as well as allergic and immune diseases. Population-based approaches to studying the microbiome present both challenges and opportunities, including the need to better understand reliability of microbiome samples and the role of microbially-derived metabolites in disease risk. We used data from the Multiethnic Cohort study (MEC) to address two primary aims: 1) investigate the temporal variability and stability of the fecal microbiome, and 2) identify associations of plasma trimethylamine N-oxide (TMAO) and precursors with biomarkers of inflammation and cardiometabolic risk and the fecal microbiome. Temporal variation of the fecal microbiome was assessed within 50 MEC participants who each provided 5 stool samples over a 2-year period. We calculated the reliability of the overall fecal microbial community using permutational multivariate analysis of variance (PERMANOVA). Taxa and diversity were measured by intraclass correlation coefficients (ICCs). Analyses were additionally stratified based on antibiotics use during the study period. For our second aim, we calculated associations between TMAO and its precursors (choline, carnitine, and betaine) with inflammatory and cardiometabolic risk biomarkers using multivariable regression. Associations between TMAO and the fecal microbiome were also assessed by PERMANOVA and LASSO regression of standardized, centered log-ratio transformed genera. For our first aim, we found the fecal microbiome to be stable over the two-year study period, with inter-individual variation as the largest source of variation. The majority of microbiome measures were reliable (ICCs>0.40), although genera with very low abundances tended to be variable. Changes in abundances, rather than the complete loss or gain of taxa, were the main source of variation within individuals. Reliability was lower for participants who used antibiotics, although this was largely driven by samples with reported antibiotics use in the month prior to collection. In our second aim, we identified several associations between TMAO and its precursors with disease biomarkers. In particular, choline indicated risk to adverse health outcomes, while TMAO, choline, carnitine, and betaine were all associated with insulin resistance. Although TMAO explained a small percentage of the overall fecal microbiome variation (<1%), it was associated with several genus-level taxa, including Desulfovibrio and two within the family Lachnospiraceae, all of which contain bacteria capable of metabolizing choline into trimethylamine. In summary, we showed the fecal microbiome to be reliable for use in a population-based study, and when incorporating the fecal microbiome in an epidemiological study, found TMAO to be associated with choline-metabolizing bacteria as well as associations between TMAO and its precursors with inflammatory and cardiometabolic biomarkers.