Cardiovascular diseases (CVDs) are the leading causes of death worldwide and account for one in three deaths in the United States alone [1
]. Classic CVD risk factors include high blood pressure, hyperlipidemia, smoking, diabetes, obesity, and lack of physical activity; however, advances in multiomic technologies have offered unbiased and unprecedented insights to the biological mechanisms of disease and have revealed new CVD risk factors such as the plasma metabolite trimethylamine n-oxide (TMAO). TMAO can be obtained in two ways including the direct consumption of TMAO containing foods and the meta-organismal metabolism of quaternary amines [3
]. In the latter method, unabsorbed dietary nutrients including choline, carnitine, phosphatidylcholine, and betaine are utilized by the gut microbiota to generate trimethylamine (TMA), which is absorbed and oxidized in the liver by flavin-containing monooxygenase 3 (FMO3) to TMAO [4
]. TMAO is robustly associated with CVD events and death in prospective studies conducted in metabolically unhealthy individuals; however, it is unclear whether TMAO is a relevant risk factor in healthy individuals [6
]. Furthermore, inter- and intra-variation in TMAO concentrations have been reported suggesting the regulation of TMAO is complex [8
]. For example, the canonical TMAO pathway, defined by diet, the gut microbiome, hepatic oxidation, and excretion contains modifiable (diet), non-modifiable (FMO3 genotype, kidney function), and semi-modifiable factors (the gut microbiome), introducing multiple levels of variability. It is unknown whether manipulating the diet, or the gut microbiome will alter TMAO concentrations and if the responses are dependent on factors such as sex, age, BMI, or health status.
Consuming eggs, meat, and the tissue of deep-sea fish are commonly associated with plasma TMAO due to their levels of phosphatidylcholine, carnitine, and TMAO, respectively [3
]. Although feeding studies have clearly demonstrated a post-prandial rise in TMAO after consumption of these foods or nutrients, the long-term effect on fasting concentrations is unclear. It is probable that both the post-prandial and long-term effects on TMAO concentrations are dependent, in part, on the metabolite and its form. Eggs are a nutrient-dense food with yolks abundant in the lipid phosphatidylcholine and have consistently demonstrated null effects on fasting TMAO concentrations. For instance, a randomized-controlled feeding trial providing eggs or egg whites found no difference in fasting TMAO concentrations after four weeks [12
]. In a randomized controlled feeding trial, which compared the effects of two eggs per day versus oatmeal, no significant differences in TMAO concentrations by breakfast food were observed [13
]. Red and white meat contain choline and carnitine mechanistically linking them to TMAO. A randomized two-arm cross-over design investigating the effect of red versus white meat observed effects of red (choline, 572 mg/d; carnitine, 258 mg/d) but not white meat (choline, 498 mg/d; carnitine, 56 mg/d) on TMAO after four weeks of consumption [10
The effect of individual foods on TMAO concentrations may vary due to a person’s overall diet pattern. Considering the fiber and carbohydrate content of the diet, a randomized two-arm cross-over feeding study providing low- or high-levels of resistant starch on either a low- or high-carbohydrate background diet resulted in increased TMAO concentrations with the high-resistant starch, low-carbohydrate diet confirming a nutrient by diet pattern interaction [14
]. Additionally, a Mediterranean diet pattern with 200 versus 500 g of unprocessed lean red meat per week identified lower TMAO concentrations on the 200 but not the 500 g diet, suggesting that reducing red meat is an effective strategy for lowering TMAO [15
]. Evidently, the nutrient form, the food matrix, and the diet pattern affect TMAO but how these dynamics play out in a free-living cohort has not been well documented.
The relationships among nutrients, foods, diet patterns, and TMAO may be due to the impact of each on the gut microbiome, which has been clearly implicated in the generation of TMAO. This has been demonstrated in rodents and humans such that antibiotics negate the effect of choline or carnitine feeding on postprandial TMAO concentrations and in humans such that, individuals consuming a vegan diet produced a reduced postprandial TMAO response after a carnitine meal challenge than matched omnivore controls [4
]. Furthermore, metagenomic screenings have identified TMA-generating gene clusters, which has led to the identification that, commonly, Firmicutes
but not Bacteroidetes
contain the TMA-generating gene cluster choline utilizing C and glycyl radical activating protein (CutC/D) [17
]. Gene clusters that utilize carnitine or its degradation product γ-butyrobetaine have also been characterized including a two-component Rieske-type oxygenase/reductase (CntA/B) that is commonly observed in Proteobacteria
]. Although gene clusters have been identified, a metagenome analysis failed to predict TMAO concentrations from TMA-generating genes in a cohort of 113 adults. This result suggests that other TMA-generating genes have yet to be discovered, that TMAO-reducing genes may be important, that the fecal microbiome provides a poor representation of TMAO dynamics, or that the gut microbiome’s role in regulating fasting TMAO concentrations is important but minor [21
]. To better understand which mechanistic steps drive fasting TMAO concentrations, studies need to consider diet, the gut microbiome, and host health in tandem. Additionally, it is unknown whether differences in the gut microbiomes of generally healthy adults relate to TMAO concentrations.
The objectives of our study are to assess the biological role of TMAO in a generally healthy cohort of adults living in the United States by characterizing the concentrations of TMAO and relating these concentrations to recent diet intake, the fecal microbiome, and cardiometabolic risk factors. Notably, our study utilizes data from a cross-sectional study equally stratified by sex, age, and BMI, allowing us to probe how these characteristics affect our outcomes.
TMAO has been associated with cardiovascular disease (CVD) in several studies, but studies of younger and generally healthy individuals have not always confirmed this association [4
]. For example, TMAO was not associated with atherosclerosis measures in healthy early-middle-aged adults enrolled in the Coronary Artery Risk Development in Young Adults (CARDIA) study [34
]. Furthermore, in a cohort of parents and children, TMAO was not related to metabolic syndrome nor markers of cardiovascular risk in either generation [35
]. Thus, determining the relationships between TMAO and risk factors of CVD remain an important area of research prior to adopting dietary modifications to reduce TMAO concentrations. We sought to address this gap and to assess the drivers of TMAO variation by studying diet, the fecal microbiome, and cardiometabolic markers—known cornerstones of the TMAO pathway, in a generally healthy cohort of adults.
Plasma TMAO is primarily derived from the microbial metabolism of choline, phosphatidylcholine, and carnitine which are abundant in red and white meat, and eggs and are commonly consumed in the typical American diet. Feeding studies demonstrate that consuming foods abundant in the TMAO precursors will lead to a transient post-prandial rise in TMAO concentrations, but studies assessing fasting TMAO concentrations report mixed results [3
]. To account for the intake of TMAO precursors as well as other dietary factors that may influence TMAO production, we assessed the relationship between recent dietary intake of animal-sourced foods, as well as vegetable, fruit, grain, and dairy food groups. We found no dietary associations between recent food intake and fasting metabolite concentrations in our healthy cohort. Although we did not observe a relationship between TMAO and meat intake, the cohort’s average intake of meat resembled national averages suggesting this was not due to reduced intake. For example, our cohort averaged 5.30 ± 4.18 oz. equivalent per day of animal sourced foods defined as the total of red and white meat, organ meat, seafood, and eggs, (males, 5.56 ± 4.96 oz. equivalent; females, 3.48 ± 2.85 oz. equivalent) compared to 2009–2010 What We Eat in America results showing males averaged 6.28 ± 0.13 oz. equivalent and females averaged 3.79 ± 0.099 oz. equivalent [38
]. Considering habitual diet, the average HEI score for our cohort was 62 (0–100 point scale, higher scores reflect compliance to the Dietary Guidelines for Americans) compared to national averages of 56 for adults 19–30 years and 59 for adults 31–59 years (Supplemental Table S3
]. While consumption of meat was comparable to national averages, overall diet quality was more aligned to the Dietary Guidelines. This may have been driven by intakes of food groups such as whole grains, fruits, and vegetables, which may reduce TMAO concentrations. As such, it is possible that other dietary factors including the food matrix, or the timing of the meal play critical roles in the ability to generate TMAO from dietary constituents. Dietary patterns may shape the gut microbiome and TMAO concentrations, but we were unable to discern interpretable diet factors (dietary factor analysis data not shown). Overall, our findings support that recent food intake is not associated with TMAO concentrations in generally healthy individuals.
The gut microbiome plays an indispensable role in the metabolism of choline and carnitine to TMAO [4
]. Therefore, we hypothesized that compositional differences would be identified between individuals with varying concentrations of TMAO. Measures of evenness and richness consistently revealed a direct relationship between α-diversity and TMAO. Alpha-diversity is commonly associated with healthy biological states making the direct relationship puzzling; however, accounts of TMAO concentrations increasing on healthy diets have been observed [40
]. Further, fiber introduces complex oligosaccharides to support a diverse gut microbiome, yet total fiber intake did not differ between the TMAO tertiles suggesting other factors may drive the observed α-diversity relationship.
When subjects were classified as high or low responders of TMAO concentrations (i.e., below or above the median of 3.32 μM), compositional differences were identified. Additionally, we identified several differentially abundant taxonomies between the lowest (<2.18 μM) and highest TMAO tertile (>3.9 μM). The genus Butyrivibrio
(Phylum Firmicutes, class Clostridia, order Clostridiales) is involved in butyrate production via the breakdown of plant polysaccharides and was more abundant in the low TMAO tertile. Although total fiber was not statistically different between the tertiles, diets of individuals in the first tertile may have been enriched in specific plant foods that promoted Butyrivibrio
(Phylum Firmicutes, class Clostridia, order Lachnospiraceae) is also a butyrate producer and was enriched in participants in the lowest tertile. In numerous strains of mice, Roseburia
was negatively associated with atherosclerotic plaque size. Follow up fecal transfer experiments providing a core community with or without the strain Roseburia intestinalis
demonstrated that mice given Roseburia
had reduced levels of TNF-α and vascular cell adhesion molecule 1 [42
]. Of the differentially abundant taxa, only the family Christensenellaceae (Phylum Firmicutes, class Clostridia, order Clostridiales) was more abundant in the highest tertile. Christensenellaceae has been reported to be highly heritable and widely associated with positive health states such as reduced adiposity [43
]. Interestingly, Christensenellaceae had a strong inverse relationship to BMI (β = −3.8, p
= 0.015); TMAO was not related to BMI (β = 0.829, Padj = 0.331). Relative to TMAO, we identified a stepwise increase in the Firmicute to Bacteroidetes ratio, a simplified marker generally associated with positive health states. We note that our resolution is limited to the genus level and that species within a genus may have different physiological effects.
Prospective cohort studies demonstrate that individuals with high baseline TMAO concentrations experience increased incidences of adverse cardiovascular events including heart attack and stroke [44
]. Mechanistically, TMAO has been shown to increase platelet reactivity and thrombosis and has been linked to many inflammatory pathways including the NF-κβ cascade and increased expression of inflammatory cytokines and chemokines [45
]. TNF-α is important in the activation of NF-κβ, which has been mechanistically linked with TMAO in mice, human aortic endothelial cells, and vascular smooth muscle cells. Additionally, TMAO may be linked to dysregulated glucose metabolism and altered cholesterol transport [46
]. We assessed the relationship between TMAO and cardiometabolic markers in a healthy population and identified that TMAO was strongly associated with tumor necrosis factor-alpha (TNF-α). This observation replicates the finding of Rohrmann et al., who observed few relationships between diet and fasting TMAO but did observe a relationship between TMAO and inflammation [37
]. Extending this finding, the significant relationship was unaffected when the abundances of the differentially abundant taxa were included as covariates suggesting that the inflammatory association was independent of the components of the gut microbiome related to TMAO. We tested whether we would be able to detect associations between TMAO and CVD risk factors within the clinically healthy range. We did not find associations between TMAO and the risk factors; however, CVD is an inflammatory disease that takes decades to develop, therefore, our TNF-α finding warrants further investigation.
Many reports including our own demonstrate that TMAO concentrations increase with age. Many CVD risk factors, and metabolic diseases increase with age making it difficult to ascertain the effect of age. Our study overcame this common limitation by including an equal stratification of young, middle, and older adults. Therefore, our observation that TMAO increases with age but is not related to cardiometabolic risk suggests that the effects of elevated TMAO may be difficult to detect in healthy populations or in early stages of cardiometabolic disease. Categorizing participants by quartiles or tertiles is commonplace in the epidemiological TMAO literature and adverse CVD outcomes are frequently reported in the highest group whose TMAO concentrations are above 5.67 [50
], 6.18 [16
], or 9.26 [51
] μM. The fasting values observed in our cohort were much lower such that the median concentration was 3.32 μM and the threshold for the highest TMAO-tertile was 3.91 μM with a mean of 6.18 μM. Potentially, elevated TMAO concentrations coupled with an underlying cardiometabolic disease (e.g., obesity, systemic inflammation, decreased kidney function, gut microbiome dysbiosis) may dictate the physiological effect of TMAO. Furthermore, our finding that TMAO increases with age in females, but not males, supports the epidemiological observation that cardiovascular risk increases in females after menopause [52
]. The FMO3 gene has an estrogen response element in its promoter and may influence this finding [49
]. Future work should investigate the effects of common FMO3 single nucleotide polymorphisms in the context of pre- and post-menopause, TMAO, and CVD risk.
Our study has several limitations. Due to the study’s cross-sectional nature, we were limited to describing associations amongst the diet, microbiome, CVD risk factors, and fasting plasma TMAO. Future studies should consider a prospective design, which allow the researcher to investigate TMAO’s role in the progression from health to disease. Our study used strict exclusion criteria to enroll generally healthy individuals. However, we were unable to confirm normal blood parameters prior to enrollment, which resulted in a small number of participants with undiagnosed hyperglycemia and/or hyperlipidemia. Additionally, our food analysis was confined by the food groups categorized by the ASA24. For example, when investigating the effect of non-processed meat (defined by the ASA24 as “total of beef, veal, pork, lamb, and game meat; excludes organ meat and cured meat”), we were unable to parse the intake of one food source from the other non-processed meats. Carnitine intake is a TMAO precursor, but the nutrient has not been well characterized in foods and is not quantified by the ASA24. Furthermore, by utilizing 16s rRNA microbial sequencing, we were unable to look at the strain phylogenetic level nor were we able to draw conclusions regarding the presence of TMA generating genes.