Does Our Gut Microbiome Predict Cardiovascular Risk? A Review of the Evidence from Metabolomics

Abstract

Millions of microbes are found in the human gut, and are collectively referred as the gut microbiota. Recent studies have estimated that the microbiota genome contains 100-fold more genes than the host genome. These microbiota contribute to digestion by processing energy substrates unutilized by the host, with about half of the total genome of the gut microbiota being related to central carbon and amino acid metabolism as well as the biosynthesis of secondary metabolites. Therefore, the gut microbiome and its interaction with the host influences many aspects of health and disease, including influencing the composition of biofluids such as urine and blood plasma. Metabolomics is uniquely suited to capture these functional host-microbe interactions. This review aims at providing an overview of recent metabolomics evidence of gut microbiota-host metabolic interactions with a specific focus on cardiovascular disease and related aspects of the metabolic syndrome. Furthermore, the emphasis is given on the complexities of translating these metabolite signatures as potential clinical biomarkers, as the composition and activity of gut microbiome change with many factors, particularly with diet with special reference to trimethylamine-oxide.

Introduction - Metabolomics, Metabonomics and the gut microbiome

The word metabolomics^1-3, and the related term metabonomics^4-5, are used to describe the use of analytical chemistry techniques to measure the metabolite compliment of a cell, tissue, or organism, often in order to derive a better understanding of a disease process or a genetic modification. The field was in part developed by the use of NMR spectroscopy to follow changes in metabolites found in urine in order to monitor potential drug toxicity as part of drug safety assessment. Using such an approach Phipps and colleagues ⁶made the intriguing observation that changes in diet in rats could alter the urinary excretion of phenolic compounds, in particular hippuric acid (HA) and meta-(hydroxyphenyl)-propionic acid (mHPPA). Through a series of different experiments involving different rat strains and diets, the authors were able to determine this was a diet dependent effect, although the diets used did not have a major difference in the content of aromoatic amino acids. The authors suggested that in fact the major difference in the excretion of HA or mHPPA arose from diet-induced alterations in gut microflora. Since this first report of the gut microbiome influencing the urinary metabolomics profile, a number of studies have observed how the metabolome is influenced by different gut microbes, and there is increasing evidence that some of these interactions between the gut microflora and the host may result or contribute to disease pathology⁷.

Developing the observations of Phipps and colleagues ⁶further, Nicholls and co-workers ⁸housed germ free mice for 21 days in metabolism cages and observed their urinary profiles as the mice established a gut microflora. These changes were fairly dramatic with increases in the concentrations of hippurate, trimethylamine N-oxide (TMAO), phenylacetylglycine and 3-hydroxypropionic acid. It is now well established for many types of metabolic profiling studies that two metabolomes must be considered for mammals, particularly when examining urinary profile changes ^9-11. Since these reports of gut microflora affecting the metabolism of amines and aromatic metabolites, a wide range of studies have suggested a role for the gut microbioticain a diverse range of diseases and modifications ¹²including obesity, type 2 diabetes and aspects of the metabolic syndrome ^13-15, drug metabolism⁹, autism spectral disorder ^16-18,colitis ¹⁹and Chron's disease²⁰. Just how large the influence of the gut microbiome has on host metabolism was examined by Wikoff and colleagues ²¹, who showed that hundreds of metabolites were perturbed in concentration between germ free and control conventionally raised mice. Indeed certain classes of metabolites were entirely absent from the metabolome of blood plasma from germ free mice. Many of the identified modifications associated with gut microflora resembled phase 2 drug metabolites.

However, because of the complex interactions between the huge quantities and diverse range of microbes found in the gut and the human body while we know that gut microbiotica do interact with a diverse range of disease, we still don't understand the underlying mechanisms. This brief review examines the potential role of the gut microbiome in cardiovascular disease. We first assess how the gut microbiome interacts with diet in healthy individuals, and then its potential role in altering risk of cardiovascular disease as well as the related co-morbidity type 2 diabetes and aspects of the metabolic syndrome. These studies have centred on a set of metabolites associated with choline metabolism and in particular the formation of TMAO, a metabolite that has been linked to a variety of interventions and diseases including CVD.

TMAO, the gut microbiome and metabolomics

One of the most important pathways linked to gut microbial metabolism involves the turnover of choline to produce amines such as trimethylamine (TMA) and TMAO. Given this central importance we felt it important to consider how these metabolites are produced in the mammalian body and how physiological changes impact on flux through this pathway. TMAO is a ubiquitous metabolite found in a variety of biological fluids and tissues and features as a major source of variation in many metabolomics studies for a range of physiological and pathophysiological states^22-24. One of the main roles of TMAO in the body is as a renal osmolyte^25-26 and elevation of TMAO concentrations in urine and plasma has previously been associated with renal medullary damage and chronic renal failure^26-27. The increased concentration of blood TMAO could reflect renal medullary damage as a result of effects on the renin-angiotensin system and hypertension associated with CVD²⁸.

TMAO is the result of the N-oxidation of its TMA precursor by liver flavin-containing monooxygenase (FMO) enzymes in the liver, in particular associated with the FMO3 isoform²⁹. TMA itself can be formed by endogenous N-dealkylation of choline by liver cytochrome P450 enzymes, but recent interest has concentrated on the bacterial degradation of choline by the gut microbiome^30-33. Dietary sources of choline include red meat, eggs, fish and brassica vegetables and choline can be found in its free form or in an esterified form (phosphocholine, glycerophosphocholine, phosphatidylcholine and sphingomyelin)³⁴. However, these forms of dietary choline have different bioavailabilities³⁵, with the lipid-soluble forms bypassing the liver whereas the water-soluble forms enter the enterohepatic circulation³⁴. Choline can also undergo metabolism to betaine via a two-step process involving choline dehydrogenase (EC 1.1.99.1) (Figure 1). Betaine acts as a methyl donor in the conversion of homocysteine to methionine in the presence of betaine homocysteine methyl transferase (EC 2.1.1.5) and it has been reported that higher concentrations of plasma homocysteine correlate with increased CVD risk³⁶.

Figure 1.

Choline metabolism: The figure shows key metabolic pathways involved in the degradation of choline and production of urinary amines, in part via the gut microbiome.

Key: BHCMT: betaine-homocysteine methyl transferase (EC 2.1.1.5), CA: creatinase (EC 3.5.3.3), CHAT: choline O-acetyltransferase (EC 2.3.1.6), CHCD: Choline dehydrogenase (EC 1.1.99.1), CK: creatine kinase (EC 2.7.3.2), CNA: creatininase (EC 3.5.2.10), DMA: dimethylamine, DMDH: dimethylglycine dehydrogenase (EC 1.5.99.2), FMO3: flavin-containing monoxygenase, MTHFD: methylenetetrahydrofolate dehydrogenase(EC 1.5.1.5), MTHFR:Methylene-tetrahydrofolate reductase (EC 1.5.1.20), PD: phospholipase D (EC 3.1.4.4), PEMT: Phosphatidylethanolaminemethyltransferase (EC 2.1.1.17, 2.1.1.71), SDH: sarcosine dehydrogenase (EC 1.5.99.1), TMA: trimethylamine, TMAO: Trimethylamine-N-oxide.

*Bacterial degradation of choline by the gut microbiome.

Gender related differences in urinary TMAO concentrations have been identified inmurine models with female rats ²² and mice ²³ showing higher concentrations of TMAO. Urinary TMAO has also been shown to fluctuate at different stages of the rat estrus cycle, which has in turn been linked to changing estrogen concentrations²⁴. In humans, the FMO3 enzymes responsible for the N-oxidation of TMA have been shown to be induced by estrogen³⁷ but suppressed by testosterone³⁸, suggesting that gender based variation in the rate of choline catabolism to form methylamine derivatives is the result of hormonal imprinting of liver enzymes such as FMO3. Further evidence that estrogens have an important role in the control of TMAO metabolism is the metabolic condition Trimethylaminuria or Fish Odour Syndrome which is caused by an inherited defect in the FMO3 enzyme³⁹. This condition can be exacerbated around puberty and in females the symptoms worsen before and during menstruation and with the use of oral contraceptives as a result of the hormonal inhibition of the FMO3 enzyme⁴⁰. It is postulated that premenopausal female subjects also have a different choline requirement to males in order to support fetal development ³⁴ particularly at the point where choline turnover links to folate metabolism and with the conversion of homocysteine to methionine (Figure 1). Estrogens are also believed to be a mediator of increased de novo biosynthesis of phosphatidylcholine, catalysed by phosphatidylethanolamine N-methyltransferase (PEMT, EC 2.1.1.17) which has been shown to be greater in female mice ⁴¹.

The gut microbiome, cardiovascular disease and the metabolic syndrome

One of the most compelling studies to link gut microbiome changes to increased obesity and the risk of developing type 2 diabetes was conducted by Turnbaugh and colleagues ⁴² using obese and lean mice. They demonstrated that the microbiome associated with obese mice was capable of extracting more calorific value from food compared with lean animals, and this ability was transferable by the transplantation of faeces into germ free mice. Furthermore, when the innate immune system is compromised as in Toll-like receptor 5 (TLR-5) knock-out mice, it has been demonstrated the gut microbiome can contribute to the development of type 2 diabetes and obesity ⁴³. In humans while the gut microbiome is shared to a degree between family members, it can also be influenced by obesity, which is associated with a decrease in diversity of the microbial community in weight discordant and concordant twins ⁴⁴. Oral bacteria have also been associated with increased CVD, and bacteria associated with oral cavities have been found in atherosclerotic plaques ⁴⁵.

From these studies there is clearly an important contribution that the gut microbiome plays in a variety of metabolic disorders including obesity, insulin resistance and CVD. These relationships are also supported in the metabolomics literature. Dumas and co-workers ⁴⁶ used ¹H NMR spectroscopy to examine the plasma and urine metabolic profiles of mice (129S6) prone to developing non-alcoholic fatty liver disease (NAFLD), determining that liver pathology was associated with changes in choline metabolism, in particular with increased excretion of amines such as dimethylamine, trimethylamine and TMAO derived in part by gut microbial metabolism. The authors suggested that the degradation of choline by the gut microbiome, reduces the availability of choline and mimics a choline deficient diet, which in turn is known to induce NAFLD. Dumas and colleagues ⁴⁷ further went on to define the interaction between host and microbial metabolism by conducting chromosomal mapping in control and diabetic rats, and then cross correlating this information with metabolic profiles measured in plasma using ¹H NMR spectroscopy. Using this approach they identified a correlation between benzoate and uridine diphosphate glucuronosyltransferase, identifying a metabotypic quantitative trait loci (mQTL) that relied on a strong gut microbe / host interaction.

Velagapudi and colleagues ⁴⁸ have examined the blood serum from germ free and conventionally raised mice, and correlated changes in blood metabolites with alterations in the lipidome of the liver and adipose tissue. As with similar studies they revealed profound changes in blood serum metabolism with conventionally raised animals having increased pyruvic acid, citric acid, fumaric acid and malic acid, and reduced concentrations of cholesterol and fatty acids. More remarkably, triglyceride and phosphocholine metabolism was also altered in the liver and adipose tissue, with conventionally raised mice having improved clearance rate of triglycerides. Given that bile acids have a key role in terms of lipid uptake from the gut, Sayin and colleagues ⁴⁹ have identified a novel key regulatory role for gut microflora in terms of bile acid synthesis. They demonstrated a signalling pathway between bile acid synthesis and activation of the nuclear receptor farnesoid X receptor (FXR) which relied on gut microbiotica regulated by the production of beta and alpha-muricholic acid in the gut. Activation of FXR in turn reduced the production of bile acids in the liver. Li and colleagues ⁵⁰ further defined the role of bile acid metabolism by gut microflora by showing that tempol, an anti-oxidant which reduces obesity in mice, alters the gut microbiome to increase the production of tauro-beta-muricholic acid, which is a FXR antagonist. When they compared FXR-null mice with wildtype controls on a high fat diet, the mutant mice showed less obesity suggesting that intestinal FXR mediates the effects of tempol.

Wang and colleagues applied metabolomics to identify potential metabolites of CVD by looking at 50 individuals who went into hospital for elective cardiac evaluation and then subsequently experienced a myocardial infarction, stroke or death over the next three years, along with 50 age-matched controls³⁰. The analysis was performed using a non-targeted Liquid Chromatography – Mass Spectrometry (LC-MS) approach, and to validate the approach a further group of 25 subjects and 25 controls were analysed in a validation group. This analysis identified 18 metabolites in both studies that robustly discriminated the disease group from the controls in both the biomarker discovery and validation tests. Of these three metabolites of phosphatidylcholine metabolism, choline, TMAO and betaine were particularly strong at predicting subsequent CVD disease. To test causality, these metabolites were then supplemented into the diet of ApoE null mice which increased the inflammatory state of the CVD system by upregulating multiple macrophage receptors linked to atherosclerosis. Given, as we have seen above, that the production of TMAO is regulated in part by gut microflora, the authors then examined what would happen in germ free animals, where TMAO production was prevented, and so too was the stimulation of the inflammatory state in ApoE mice following choline feeding, as well as producing a reduction in aortic plaque size in the mice.

While the study as a whole makes an impressive argument for the role of gut microflora in CVD, mediated through the break-down of choline to produce TMAO, it did raise a number of questions. As we have seen there is an important influence of diet on the urinary and blood plasma concentrations of TMAO, betaine and choline. Firstly red meat consumption will increase TMAO excretion, and increased red meat consumption is associated with CVD, so could TMAO actually be a biomarker for red meat consumption in the human cohorts examined? Furthermore, metabolomics investigations have found that plasma and urine concentrations of TMAO increase dramatically with increasing consumption of fish, but the evidence is that increased fish consumption is protective from CVD^51-55.

Furthermore, metabolites produced by the oxidation of choline have been shown to be poor markers of plaque formation in the mouse. We have previously shown that the oxidation of choline is increased by feeding C57BL mice a diet containing a high fat and cholate containing diet, but decreased in the LDL receptor knock out mouse, a transgenic mouse model with a known predisposition to CVD ⁵⁶. Furthermore, betaine supplementation is reported to attenuate atherosclerotic lesion in ApoE^-/- mice ⁵⁷. In animal models, it also reduces plasma homocysteine concentration⁵⁸, a known atherosclerotic risk factor that is associated with enhanced vascular inflammation and oxidative stress⁵⁹. These studies raise the question as to whether the correlation observed by Wang and colleagues³⁰ is causative, as claimed in the study, or a surrogate marker of the effect. Furthermore, the dietary variation we describe above will affect the utility of TMAO as a potential biomarker since the sources cannot be delineated from a measure of the plasma. Hence, the background affect for use as a biomarker is limited to situations where exclusion criteria are applied before sampling and such criteria will be strict. However, the relationship between CVD and TMAO concentrations has been borne out in a larger study of over 4000 individuals, where TMAO was again found to be correlated with subsequent CVD events ³¹, even after correcting for major risk factors including age, sex, smoking, systolic blood pressure, LDL-cholesterol, HDL-cholesterol and type 2 diabetes status. Thus, if TMAO is a surrogate marker for another pathological process it must be separate to these classic risk factors.

To further characterise the role of TMAO in CVD, Bennett and colleagues ⁶⁰ examined the regulation of TMAO concentrations in mice, identifying that flavinmonoxygenases (FMO) 1 and 3 are responsible for the conversion of trimethylamine (TMA) to TMAO, with FMO3 having the highest activity. Intriguingly the expression of FMO3 is induced by dietary bile acids involving the farnesoid X receptor, linking lipid uptake to TMAO concentrations. In mice the induction of FMO1 and 3 was in part negatively regulated by the production of testosterone, which the authors use to explain the greater susceptibility of female mice to atherosclerosis, in marked contrast to human disease.

High urine concentrations of TMAO directly correlate to the consumption of a high meat content diet ⁶¹ and this can be attributed to high concentrations of not only choline but also carnitine (found in high concentrations in red meat, fish and dairy products as well as some energy drinks and diet supplements) which has also been shown to be associated with bacterial degradation to the TMA precursor⁶². It has also been observed that vegetarian and vegan subjects who consume a single meal of meat had lower blood levels of TMAO than those eating a regular omnivorous diet which was attributed to the lower levels of the intestinal bacteria responsible for the catabolism of carnitine to TMA in vegetarian and vegan subjects ⁶². However, these studies were rather under powered and further clinical studies are required to fully understand the relationship between the dietary sources of TMA (whether from dietary choline and/or carnitine) the role of the liver FMO3 enzymes responsible for the N-oxidation of TMA to TMAO and the potential ‘imprinting’ of the gut microbiome from different baseline diets (i.e. omnivorous, pescetarian, vegetarian and vegan). Furthermore, Ussher and colleagues have reviewed both the positive and negative effects of carnitine supplementation, suggesting that potential negative effects associated with raised production of TMAO contributing to CVD maybe balanced by increased glucose metabolism in muscle and heart⁶³.

In addition to interactions where the gut microbiome influences the health of the host, there are also examples where surgical intervention on the host in turn affects the microbiome. Ashrafian and colleagues ⁶⁴ examined Roux-en-Y gastric bypass (RYGB) surgery in the Wistar rat using a combination of ¹H NMR spectroscopy and LC-MS based metabolomics to study metabolic changes in blood plasma and cardiac tissue. In humans the RYGB procedure has a profound effect on patient health, often dramatically improving insulin resistance and reducing cardiovascular risk. This approach identified perturbations in a variety of metabolic pathways including bile acids, phosphocholines, amino acids, nucleosides and amines in blood plasma and glycogen and amino acids in the heart. The authors suggest that these changes were in part mediated by the alterations in gut microflora following the gastric bypass, demonstrating the complexities between the host/gut microbiome interactions.

Conclusions

There is increasing evidence the gut microbiome has a profound effect on the systemic health of the host for a range of diseases. Obesity, type 2 diabetes and the metabolic syndrome are some of the best characterised pathologies that have been identified as examples of these host/gut microbiome interactions, and metabolomics is increasingly being used to understand how the gut microbiome may aggravate disease processes. This has identified a number of metabolites that are readily observed in human blood plasma and urine that could be used to follow these interactions. There is also evidence that some of these metabolites could be used to identify patients at risk of disease, although given the dynamic nature of the gut microbiome one must take care with these correlations, particularly in terms of whether these changes are causative or a consequence of the pathology. Furthermore, diet has an even more profound effect on the gut microbiome and we must be vigilant that suggested biomarkers of obesity and cardiovascular risk are not proxies for dietary changes.