Population studies of TMAO and its precursors may help elucidate mechanisms

Over the past decade, a growing body of literature has implicated the metabolite trimethylamine N-oxide (TMAO) in cardiovascular disease (CVD) risk (1). These findings spurred research into TMAO, and investigations have expanded beyond CVD into metabolic, renal, and other outcomes. Much of this work has focused on patient-based or high-risk samples and there has been a relative lack of study in population-based samples. In this issue of the Journal, Fu et al. (2) report their findings on TMAO, microbiome, and cardiometabolic risk measures. Their work is a significant contribution to the field, as it addresses several important gaps in the existing literature.

TMAO is a metabolite of dietary choline and l-carnitine. Choline and l-carnitine that are not absorbed in the small intestine travel to the colon, where they can be converted to trimethylamine (TMA) by microbial enzymes. TMA is oxidized to TMAO in the liver by flavin-containing monooxygenase (FMO) enzymes, including FMO3 enzyme (3). Thus, TMAO is dependent upon dietary consumption of precursors, gut microbial enzymatic activity, and host genetics. A genome-wide association study (GWAS) of human and mouse data indicates a limited role for FMO3 variants on TMAO concentrations, with most variability likely reflecting consumption of dietary precursors and gut microbial function (4). Specific microbial enzymes involved in TMA-production pathways have been described, and microbes have been defined that typically carry the genes for these enzymes (5). Interestingly, there may be large intraindividual variability in the production of TMA/TMAO when challenged with a dietary precursor, such as eggs (6). These features of TMAO metabolism are illustrative of gut microbial and host co-metabolism, and highlight its potential relevance for precision nutrition approaches.

Several important questions remain. Much of the data supporting the TMAO–CVD association derives from studies in people with disease or at high risk of CVD, renal, or metabolic outcomes (1). A critical question is whether these associations will be observed in general-population studies. Associations in high-risk samples may reflect residual confounding by declining renal function, as TMAO is excreted in urine, or other subclinical measures. Of note, studies of general-risk populations have not consistently supported associations between TMAO and CVD risk factors (7, 8). In addition, the dependence of TMAO on consumption of dietary precursors leads to questions about associations between the precursors themselves and CVD-related risk measures. The evidence is mixed: a recent investigation of 3 large population–based cohorts found positive associations between dietary choline and CVD events (9), while a meta-analysis of earlier studies did not (10). Another quandary relates to fish consumption. Fish is a source for both choline and TMAO and is considered beneficial for cardiometabolic health. Recently, the International Study of Macro-/Micronutrients and Blood Pressure (INTERMAP) reported that TMAO was positively associated with blood pressure in low-fish-consuming Western individuals, but not in high-fish-consuming Japanese individuals (8), raising the possibility that geographic differences in dietary patterns or gut microbial composition may be important considerations.

Fu et al. used data from the geographically and ethnically diverse Multi-Ethnic Cohort (MEC) to examine plasma concentrations of TMAO, choline, and betaine with respect to several cardiometabolic risk markers. The team also tested associations between TMAO concentrations and gut microbial composition. This is the first population-based study of associations between TMAO and the gut microbial community, and one of a few investigations of TMAO and cardiometabolic risk factors in a population-based sample. The study is notable for its large size, multiethnic/racial composition, clinic-based measurement of cardiometabolic measures, detailed assessment of diet and other covariates, and rigorous analytic approach. TMAO was positively associated with insulin resistance (HOMA-IR) but was not associated with C-reactive protein (CRP) or with serum lipid concentrations or blood pressure, standard CVD risk factors. In contrast to the limited TMAO findings, choline and carnitine were positively, and betaine negatively, associated with a broader set of risk factors. The findings for choline are intriguing, considering the tight regulatory control on circulating concentrations (11), but, as the authors note, these findings are consistent with previous reports. Adjustment for the consumption of fish, red meat, and processed meat—all significantly positively associated with TMAO in their sample—did not meaningfully change the results.

The authors did not discuss mechanism, but the complexity of TMAO-related pathways is beginning to be illuminated. It has been shown that FMO3 expression is increased in obese/insulin-resistant mice and humans, which may suggest that insulin resistance precedes increases in TMAO concentrations (12). Fu et al.’s analysis was cross-sectional, and temporality cannot be established for the association between TMAO and insulin resistance. Also relevant is protein kinase R (PKR)-like endoplasmic reticulum kinase (PERK), the first identified TMAO receptor and a critical link for TMAO-induced hyperglycemia, which can be activated through signaling to the liver by TMA-producing microbes (13). These, and other, mechanistic findings highlight the enormous complexity of the interplay among diet, the gut microbiome, microbiota-generated metabolites, and cardiometabolic health and underscores our early stage in understanding these pathways.

Interestingly, betaine was associated with microbiota-related measures, including circulating lipoprotein-binding protein and gut microbial community composition. Betaine was significantly associated with both unweighted and weighted UniFrac, while TMAO was only associated with unweighted UniFrac. UniFrac is a multivariate approach to estimate similarity among microbial communities based on phylogenetically informed distances between people. Unweighted UniFrac focuses on distances based on the presence or absence of a microbe, rather than abundance (as weighted UniFrac does) and may reveal associations for microbes that are at low abundance but distinguish people based on presence or absence. These findings require replication and may reflect residual confounding (such as by diet), but they also highlight the possibility that the gut microbiome may contribute to our understanding of nutrient metabolism in unexpected ways.

Several features of the work represent important advances in the study of the gut microbiome in large-scale studies. The focus on a microbiota-generated metabolite, TMAO, with known enzymatic pathways allowed the investigators to interpret their findings for individual taxa within the context of what is known about the propensity for certain taxonomic groups to carry genes for TMA-producing enzymes (5). The authors also included study of archaea, through quantification of a specific methanogen, Methanobrevibacter smithii. Archaea have not received the level of attention as bacteria but may be important for nutrient metabolism, including TMA production, as supported by the authors’ findings.

The gut microbiome has emerged as a potential novel risk factor for cardiovascular, metabolic, and renal outcomes. One pathway by which the microbiome may influence human health is through the production of microbiota-generated metabolites. Much attention has focused on TMAO, but the number of such metabolites, particularly of dietary origin, is vast. We are only beginning to delineate these complex interconnections. Animal models have provided the strongest support for causal mechanisms, and observational cohort studies have established the human relevance of microbiome study. However, there has been a paucity of pathway and mechanistic work in population-based data. The article by Fu et al. provides an appealing model for such work. Additional advances can be made through deeper sequencing (e.g., whole-metagenomics) and increased integration of in vivo and in vitro models for functional confirmation of findings from human studies.

Study of the gut microbiome may significantly enhance our understanding of nutrient metabolism and specific pathways by which diet can influence health. In particular, the enormous individual variability in gut microbial structure and function, and the potential for differential metabolite production, highlights the relevance of the microbiome for precision nutrition approaches. Fu et al.’s results refine our understanding of the potential relevance for TMAO in metabolic dysfunction, which may underlie, at least partially, positive associations between TMAO and CVD. If so, TMAO monitoring in individuals with insulin resistance or prediabetes may be beneficial.

Notes

This work was supported by K01-HL127159 and P30-DK056350.