Gut Microbiota-Generated Trimethylamine N-Oxide from Dietary Choline and Glucose Homeostasis in Humans

Abstract

The causal relationship between gut microbiota-generated trimethylamine N-oxide (TMAO) and type 2 diabetes remains unclear. We tested the hypothesis that gut microbiota-generated increases in plasma TMAO concentrations will impair insulin sensitivity and glucose tolerance in healthy, sedentary humans. To address this, we performed two studies utilizing a randomized double-blind, placebo-controlled, crossover design. Eligible participants (age of 35 ± 16 years for the acute study and 46 ± 14 years for the short-term study) consumed a 1000mg/day dose of choline bitartrate and placebo (maltodextrin) the night before each testing session (for the acute study, n=20, 8 women) or for 4 weeks (for the short-term study, n=23, 13 women). An oral glucose tolerance test was performed the day after the acute study and before and after the short-term study. Insulin sensitivity was calculated using the Matsuda Index. Plasma TMAO was increased (both P<0.05) 400% and 110% following acute and 4 weeks of daily ingestion of 1000mg of choline bitartrate, respectively. There were no statistically significant differences in insulin sensitivity (Matsuda index) or insulin resistance (HOMA-IR) between the interventions in both the acute study and short-term study. In addition, there were no significant differences between conditions in fasting glucose, 2-hour glucose, area under the curve, and incremental area under the curve in the acute and short-term study. Taken together, the results of the present study suggest that the relationship between TMAO and glucose homeostasis may not be causal in nature.

New and Noteworthy

1. The gut microbiota-generated metabolite, trimethylamine N-oxide, has been associated with impaired glucose homeostasis and type 2 diabetes. However, the causal nature of this relationship remains uncertain.

2. We performed two randomized double-blind placebo-controlled studies involving acute and short-term dietary choline ingestion to determine whether increasing trimethylamine N-oxide impairs glucose homeostasis in humans. Insulin sensitivity and glucose tolerance were assessed using oral glucose tolerance testing.

3. We found that, despite elevations in trimethylamine N-oxide, there were no significant reductions in insulin sensitivity or glucose tolerance (or fasting glucose and insulin concentrations).

4. Taken together, our findings that the relationship between trimethylamine N-oxide and glucose homeostasis may not be causal but addition, including longer studies, are needed.

Introduction

Trimethylamine N-oxide (TMAO), a gut microbiota-derived metabolite of dietary nutrients such as choline and carnitine, has been implicated in the development of cardiovascular diseases,^1,2 although the causal nature of this relation has recently been questioned.^3,4 Nevertheless, given the close relationship between cardiovascular disease and type 2 diabetes (T2D) it is not surprising that a higher prevalence of T2D has been reported in patients with elevated TMAO concentrations.^5–12 In addition, feeding rodents TMAO exaggerated high-fat feeding glucose intolerance, impaired insulin signaling, and promoted adipose tissue inflammation.^13,14 In turn, suppressing TMAO improves glucose tolerance in mice.¹⁴ However, elevated TMAO concentrations have not been consistently associated with an increased risk of T2D in humans.^15–17 A recent meta-analysis¹⁵ indicated that TMAO concentrations were associated with T2D risk in cohort studies¹⁸ but not case-control studies.^10–12 In a pooled analysis, there was no significant association between TMAO concentrations and risk of T2D. Furthermore, a Mendelian randomization study suggested that genetically predicted higher TMAO concentrations are not causally linked to T2D.¹⁹ As such, whether TMAO directly impairs glucose homeostasis in humans remains unknown.

To address this gap, we conducted an acute and a short-term randomized controlled trial of choline ingestion to increase circulating TMAO and assess its impact on glucose tolerance. We hypothesized that gut microbiota-generated increases in TMAO following choline ingestion would reduce insulin sensitivity and impair glucose regulation in healthy adults.

Methods

We conducted two randomized, double-blind, placebo-controlled crossover trials of varying lengths (Clinicaltrials.gov identifier: NCT03327805) using the approach described by Zhu et al.²⁰ Details of endothelial function and arterial stiffness outcomes from these trials have been published previously.³ Eligible participants received either 1000 mg/day of choline bitartrate (providing approximately 450 mg of choline; NutriScience, Trumbull, CT, USA) or maltodextrin placebo. Participants were instructed to avoid consuming foods that contain high amounts of choline, such as red meat and fish the night before each testing session. In the acute choline trial, ingestion of choline occurred the evening prior to each testing session, whereas in the short-term trial, participants consumed the assigned treatment daily for four weeks. Washout intervals of one week (acute trial) and two weeks (short-term trial) separated the interventions.

Participants were eligible for inclusion if they were adults aged 18–79 years, weight-stable within ±5 pounds over the preceding six months, refrained from probiotic or antibiotic one month prior to the study, had no diagnosed cardiometabolic conditions, and reported a sedentary lifestyle defined as less than 60 minutes of exercise per week. Participants were required to be willing to take choline capsules and to be fully vaccinated against SARS-CoV-2.

Individuals were excluded if they followed a vegan or vegetarian diet; had diagnosed diabetes or were using diabetes medications; were pregnant; or had used prebiotics, probiotics, or antibiotics within the previous three months. Additional exclusion criteria included total cholesterol greater than 6.2 mmol/L, triglycerides greater than 4.5 mmol/L, blood pressure exceeding 159/99 mmHg or the use of antihypertensive medications, active smoking, a known choline allergy or intolerance, or a diagnosis of trimethylaminuria. or had a body mass index (BMI) greater than 35 kg/m². Both protocols received approval from the Virginia Tech Institutional Review Board (IRB # 17–562 and 18–535), and all volunteers gave written and verbal informed consent.

Testing sessions were scheduled following an overnight fast and after avoiding strenuous exercise for 24 hours. Body weight was recorded at each visit to confirm stability. Height and weight were measured in light clothing without shoes using a digital scale (Scale-Tronix 5002, Chicago, IL, USA) and a stadiometer; BMI was calculated as weight (kg) divided by height squared (m²). Blood pressure was assessed in triplicate over the brachial artery during supine rest using an automated oscillometric device (HEM 907XL, Omron Healthcare Inc., Lake Forest, IL, USA).

TMAO and Choline Analysis

Fasting blood samples were collected at each visit and processed as described previously.³ Briefly, plasma was isolated and stored at −80 °C for analysis of TMAO and choline, which was performed as described previously.²¹ Concentrations were determined in duplicate using UPLC-MS/MS (Waters Acquity system) with electrospray ionization in positive mode. Separation was performed on a Waters BEH HILIC column under isocratic conditions (20% A: 15 mM ammonium formate, pH 3.5; 80% B: acetonitrile) at 0.65 mL/min. Quantification was based on analyte/internal standard peak area ratios using external calibration curves prepared with TMAO-d₉ and choline-d₉.

Oral Glucose Tolerance Test

A venous catheter was inserted into an antecubital arm vein for blood sampling. Following the initial blood draw, participants ingested 75 g of glucose (75 SUN-DEX^™, Thermo Fisher Scientific Inc.). Subsequent blood samples were collected at five intervals, every 30 minutes over a 2-hour period. Plasma glucose was analyzed at each time point using the HemoCue^® 201 Glucose Analyzer (HemoCue^®, Brea, CA, USA). The HemoCue^® 201 Glucose Analyzer has acceptable analytical reliability (coefficient of variation of 2–3%), has good agreement with laboratory analytical methods, and linear measurement range between 0 and 400 mg/dL. Recovery accuracy is between 99 and 1006% .²² Quality control was performed according to manufacturer guidelines.²³ Insulin concentrations were measured by ultrasensitive insulin enzyme linked immune sorbent assay (ELISA) (#80-INSHU-E01.1, ALPCO Diagnostics, NH, USA). Indices of insulin sensitivity and resistance were calculated using the Matsuda index²⁴ and the Homeostatic Model Assessment for Insulin Resistance (HOMA-IR)²⁵, respectively. Stumvoll, Insulinogenic and disposition indices were calculated based on published guidelines.^26–28 Glucose area under the curve (AUC) and glucose incremental area under the curve (iAUC) were calculated using the trapezoid rule.

Statistical Analyses

Statistical analyses were conducted using IBM SPSS Statistics for Windows, version 29 (IBM Corp., Armonk, NY, USA; Released 2022). Normality of continuous variables within each group was assessed using the Shapiro–Wilk test, with significance set at p < 0.05. For the acute choline study, paired t-tests were used to compare mean values between conditions. In the short-term study, treatment effects were evaluated using a linear mixed-effects model (LMEM), with participant ID included as a random factor. Post hoc comparisons were adjusted using the Bonferroni method (p < 0.05). Pearson correlation analyses were performed to examine associations between changes in plasma TMAO and key outcome measures. Figures were created using GraphPad Prism (version 10.0.3 for Windows; GraphPad Software, La Jolla, CA, USA).

Results

Subject Characteristics

The participant flow (i.e., CONSORT Diagram) through the studies has been reported previously.³ Briefly, a total of 20 participants enrolled and completed the acute choline study, and 23 participants enrolled and completed the short-term study, with two individuals participating in both protocols. In the short-term study, one participant was excluded due to missing placebo TMAO data and one was removed because their plasma TMAO value was not physiologically tenable; additionally, three participants were non-compliant with pre-test dietary instructions during the placebo arm. The insulinogenic and disposition indices from 4 participants and the Stumvoll index from 1 participant were not included in the analysis, because of negative values yielded based on the 0-to-30-minute OGTT drop in glucose or insulin and a very high insulin in the 120 minutes for the Stumvoll index No enrolled participants withdrew from either study. Baseline characteristics of the participants can be found in table 1.

Table 1.

Baseline characteristics of all participants

	Acute study, n=20, 8 Females	Short-term study, n=23, 13 Females
Age, yr	35 ± 16	46 ± 14
Body Height, cm	171± 11	170± 6
Body Weight, kg	81.0 ± 31.7	76.8 ± 26.3
Body Mass Index, kg/m2	25.6 ± 3.5	25.2 ± 3.9
Systolic Blood Pressure, mmHg	116 ± 12	118 ± 12
Diastolic Blood Pressure, mmHg	64 ± 10	71 ± 7

Data presented as mean ± SD.

Acute study

Plasma TMAO was approximately 400% higher after acute ingestion of choline compared with placebo (placebo: 3.1 ± 1.6 μM, choline: 12.5 ± 10.7 μM, Figure 1A). Results presented as mean ±SD. However, the Matsuda index was not significantly different between conditions (placebo: 3.6 ± 1.1, choline: 3.8 ± 1.5, Figure 1B). Glucose concentrations increased similarly with the OGTT in the two conditions (Figure 1C). Fasting glucose concentration (placebo: 92 ± 8, choline: 93 ± 10mg/dL), 2-hour glucose concentration (placebo: 109 ± 29, choline: 104 ± 31 mg/dL), and both AUC (placebo: 15278 ± 2027 choline: 15385 ± 2850 units) and iAUC (placebo: 4232 ± 2002, choline: 4375 ± 3030 units) were similar in the two conditions (all P>0.05). In addition, we did not observe any significant difference in the HOMA-IR (placebo: 2.9 ± 1.0, choline: 3.1 ± 0.9 units, p=0.389), Insulinogenic (placebo: 1.3 ± 0.8, choline: 1.7 ± 2.1, p=0.372) and Disposition index (placebo: 4.6 ± 2.9, choline: 5.9 ± 7.0, p=0.458). Stumvoll index was significantly different (placebo: 0.09 ± 0.03, choline: 0.1 ± 0.03, p=0.043) between the two conditions.

Figure 1.

Individual changes in plasma trimethylamine N-oxide concentrations (A), Matsuda Index (B), and changes in glucose (C) and insulin (D) during the oral glucose tolerance Test following acute placebo or choline ingestion (1000 mg). Data presented are Estimated Marginal Means ± SE. ** p = 0.001 vs. Placebo. * p<0.01 vs. Time 0. TMAO; Trimethylamine N-oxide.

Short-term study

Fasting plasma TMAO concentrations increased approximately 110% following 4 weeks of ingestion of choline but no change was observed following placebo (Figure 2A). Matsuda index did not change with the intervention (P>0.05) (Figure 2B). Glucose concentrations increased significantly during the OGTT but there was no difference between the conditions (P>0.05) (Figure 2C). There was an order effect for fasting glucose concentrations and AUC (p=0.006), fasting glucose concentration increased significantly more in participants randomized to choline first. There were no significant changes in glucose tolerance, as measured by AUC, iAUC, HOMA-IR, Stumvoll, Disposition, or Insulinogenic Index, between the two intervention groups (Table 2).

Figure 2.

Individual changes in plasma trimethylamine N-oxide concentrations (A), Matsuda Index (B) and changes in glucose (C) and insulin (D) during the Oral Glucose Tolerance Test following short-term (4 weeks) placebo or choline ingestion (1000 mg). Data presented are Estimated Marginal Means ± SE. ** p=0.006 vs. Pre Choline; * p = 0.001 vs Time 0. TMAO; Trimethylamine N-oxide.

Table 2.

TMAO and glucose homeostasis before and following choline or placebo in the short-term study

	Placebo		Choline
	Pre	Post	Pre	Post
Plasma TMAO, uM	3.85 ± 0.88	5.00 ± 0.90	3.68 ± 0.88	8.59 ± 0.90
Matsuda Index	3.9 ± 0.27	3.7 ± 0.27	4.1 ± 0.28	4.0 ± 0.28
Fasting Glucose, mg/dL	90 ± 2	90 ± 2	90 ± 2	88 ± 2
2-hour Glu OGTT, mg/dL	97 ± 6	105 ± 6	108 ± 6	94 ± 6
Glucose AUC	14152 ± 581	14614 ± 574	14638 ± 590	14174 ± 595
Glucose iAUC	3300 ± 556	3862 ± 550	3826 ± 566	3603 ± 570
HOMA-IR	2.84 ± 0.39	3.09 ± 0.39	2.80 ± 0.40	3.60 ± 0.41
Stumvoll index	0.089 ± 0.005	0.087 ± 0.005	0.091 ± 0.005	0.097 ± 0.005
Insulinogenic index	2.00 ± 0.35	1.77 ± 0.34	1.54 ± 0.35	1.23 ± 0.36
Disposition index	8.00 ± 1.53	6.44 ± 1.48	5.74 ± 1.53	5.19 ± 1.54

Data presented as Estimated Marginal Means ± SE. TMAO, Trimethylamine-N-oxide; OGTT, Oral Glucose Tolerance Test; AUC, Area Under the Curve; iAUC, incremental Area Under the Curve; HOMA-IR, Homeostatic Model Assessment for Insulin Resistance.

There were no significant correlations between the increases in TMAO and any of the glucose homeostasis variables. In addition, neither age nor baseline TMAO concentration was correlated with the increases in TMAO.

Discussion

The major finding of the present study is that, contrary to our hypotheses, acute and short-term gut microbiota-generated increases in TMAO concentrations did not impair insulin sensitivity. In addition, fasting glucose concentration, glucose tolerance (AUC or iAUC) and insulin resistance (HOMA-IR) did not change following both the acute or short-term interventions.

TMAO concentrations have been reported to be higher in db/db diabetic mice compared with nondiabetic mice.²⁹ TMAO or choline feeding exacerbated hyperglycemia in mice fed a high fat diet for 3,^13, 9–12 weeks ^14,25 perhaps, in part, via promotion of endoplasmic reticulum stress. ⁶ However, TMAO appears to protect beta-cells exposed to glucolipotoxic conditions in vitro.³¹ In addition, Dumas et al. recently reported that trimethylamine, the microbial precursor to TMAO, can reduce metabolic inflammation and improve glycemia in high-fat-fed mice by inhibiting interleukin-1 receptor-associated kinase 4 activation of the Toll-like receptor pathway.³² The latter is a bit surprising given the observations linking TMAO with adverse cardiometabolic outcomes. Nevertheless, a neutral effect of TMAO on the risk of T2D, as observed in the present studies, could be explained, at least in part, by the directionally opposite effects of TMA and TMAO on glucose homeostasis.

The results of a recent systematic review and meta-analysis suggest that TMAO concentrations are elevated in T2D compared with normoglycemic humans.¹⁵ In addition, elevated TMAO concentrations were associated with a 49% increase in the odds of T2D. However, there was not a significant association between TMAO concentrations and T2D risk in the pooled analysis. Furthermore, a Mendelian randomization study suggested that genetically predicted higher TMAO concentrations are not causally linked to T2D.¹⁹ In the same study, chronic kidney disease and T2D were causally linked to genetically increased TMAO concentrations suggesting reverse causality.¹⁹ Taken together, the causal nature of the relationship between TMAO and T2D remains uncertain.

The strengths of our study include the randomized, double-blind, placebo-controlled, crossover trial design, the use of gut microbiota-generated increases in TMAO concentrations to test a causal link between TMAO concentrations and insulin sensitivity acutely and in the short term, and the phenotyping of glucose homeostasis in humans. However, there are some limitations that should also be acknowledged. First, our sample was limited to a relatively small number of participants and primarily a young, healthy Caucasian population, which suggests caution in interpreting our findings, as they may not generalize to other racial/ethnic groups or those with chronic diseases. Second, the increases in TMAO were highly heterogeneous. Whether more consistent and robust increases would have a different impact is not known. Third, the typical Western diet contains meat, fish, milk, and eggs which are the dietary precursors of TMAO. As a result, choline ingestion of 1,000mg might not be enough to sufficiently raise TMAO to concentrations necessary to influence glucose homeostasis. However, the average daily choline intake from foods and beverages in the US is 402mg for men and 278mg for women.³³ In a subset of our participants (n=11), choline intake was 308±112 mg/day in the acute study and intake and 298±63 mg/day in the short-term study (n=9). As such, we effectively doubled the average daily intake during each of the intervention periods. In addition, the average plasma TMAO concentration achieved in both studies (acute study: 12.5 ± 10.7 μM, short-term study: 8.59 EMM ± 0.90 μM) exceed that reported previously in patients with T2D (median and interquartile range of TMAO concentrations of people with diabetes were 4.4 [2.8–7.7] μmol/L). Nevertheless, further increases in plasma TMAO may be necessary. ⁸ Fourth, our studies were limited to acute and short-term gut microbiota-generated increases in TMAO following ingestion of choline bitartrate. Whether increases in TMAO over a longer period are necessary to impair glucose homeostasis is unclear. Finally, it is possible that increases in TMAO alone is insufficient to impair glucose homeostasis but might interact synergistically with other factors (e.g., obesity) to increase risk of T2D. Future studies are needed to better understand the nature of the relationship between TMAO and T2D.

In summary, acute and short-term gut microbiota–generated increases in TMAO did not alter insulin sensitivity or other measures of glucose homeostasis. These neutral metabolic effects contrast with mixed findings from preclinical studies, where TMAO has been reported to both worsen and, under some conditions, protect aspects of glucose homeostasis. Evidence in humans is similarly inconsistent, with observational and genetic studies offering conflicting insights into whether TMAO plays a causal role in T2D. Taken together, our findings add to a growing body of literature suggesting that the relationship between TMAO and glucose metabolism is complex and may not be directly causal. Further work, especially involving longer-term elevations in TMAO and more diverse populations, will be needed to clarify its metabolic significance.

Data Availability

The data are available from the corresponding authors upon reasonable request.