Circulating Gut Microbiota Metabolite Trimethylamine N-Oxide (TMAO) and Changes in Bone Density in Response to Weight Loss Diets: The POUNDS Lost Trial

Type 2 diabetes is closely related to obesity and altered bone health (1,2). A growing number of studies have associated type 2 diabetes with increased fracture risk, even though individuals with type 2 diabetes have high bone mineral density (BMD). The fragility in apparently “strong” bones in patients with diabetes may result from microcrack accumulation or cortical porosity (2–4). Diet intervention is a mainstream approach to reduce body weight and diabetes risk. However, several clinical trials show that weight loss has been associated with reductions in BMD (5–7), although the long-term effects on bone health remain to be determined.

Emerging evidence has linked gut microbiota and its related metabolites, such as trimethylamine N-oxide (TMAO), to obesity (8–10). Our previous studies showed that weight loss diets were associated with decreased levels of choline and l-carnitine, which are precursors of TMAO. These changes were associated with improvements in adiposity, energy metabolism, and insulin resistance (11,12). Gut microbiota have also been linked with bone health (13–15). Alterations in gut microbiota by pharmaceutical agents and pre- and probiotics were suggested to improve bone health (14,16). Notably, TMAO was shown to increase levels of fasting insulin, which may increase BMD by affecting bone formation (2,17). However, it is not known whether the relationships between changes in the gut microbiome–related metabolites in participants on weight loss diets are associated with changes in BMD.

Therefore, in this study, we examined whether changes in TMAO, choline, and l-carnitine induced by weight loss were associated with changes in BMD among overweight or obese participants from the Preventing Overweight Using Novel Dietary Strategies trial (POUNDS Lost), a 2-year randomized dietary intervention trial (18). Given that dietary fat might affect the gut microbiota metabolism as well as bone health (19,20), we also examined the effect of potential interactions between dietary fat intake and changes in these metabolites on changes in BMD.

POUNDS Lost is a 2-year randomized intervention trial. The study design and sample collection have previously been described (18). In brief, 811 overweight and obese individuals were assigned to four energy-reduced diets that differed in macronutrient composition. The percentages of energy derived from fat, protein, and carbohydrates in the four diets were 20%, 15%, and 65% (low fat, average protein); 20%, 25%, and 55% (low fat, high protein); 40%, 15%, and 45% (high fat, average protein); and 40%, 25%, and 35% (high fat, high protein), respectively. Thus, two diets were low fat (20%) and two diets were high fat (40%) or two diets were average protein (15%) and two diets were high protein (25%). The two high-fat diets were also low-carbohydrate diets. The major exclusion criteria in this trial included the presence of diabetes, unstable cardiovascular disease, use of medications that affect body weight, and insufficient motivation (18). Of the 811 participants, 424 were recruited into the study of BMD. Of these, 264 who had measurements of plasma TMAO, choline, and l-carnitine levels were included in the final analysis (see flowchart in Supplementary Fig. 1). There were no significant differences in baseline values of BMD among individuals with or without data on these metabolites. The study had 80% power to detect a 0.7-kg difference in total body fat loss as an effect of the amount of protein or fat in the diet over the 2-year period, with assumption of a dropout rate of 40% (21).

The trial was conducted at two sites: one in Boston at the Harvard School of Public Health and Brigham and Women’s Hospital and another in Baton Rouge at the Pennington Biomedical Research Center of the Louisiana State University System. The study was approved by the human subjects committee at each institution and by a data and safety monitoring board appointed by the National Heart, Lung, and Blood Institute. All participants gave written informed consent.

Primary outcomes were changes in BMD during the 2-year intervention. Body weight was measured by calibrated hospital scales in the morning before breakfast. Height was measured at baseline. BMI was calculated as weight in kilograms divided by the square of height in meters. DXA (Hologic QDR-4500A bone densitometer; Hologic, Inc.) was used to measure the spine, total hip, and whole-body BMD, the method for which has previously been reported in detail (7).

At baseline, dietary intake was obtained from 5-day diet records and at 6 and 24 months from 24-h recalls collected by telephone on three nonconsecutive days in a random sample of 50% of the participants. To validate the self-reported adherence to the diet intervention, we examined the following biomarkers of nutrient intake: urinary nitrogen excretion for protein and respiratory quotient for fat. Baecke Physical Activity Questionnaire was used to determine an individual’s level of habitual physical activity. Analyses of serum glucose, insulin, and hemoglobin A_1c (HbA_1c) were performed at the Clinical Laboratory at Pennington. Fasting glucose and fasting insulin concentrations were measured using an immunoassay with chemiluminescent detection on the Immulite Analyzer (Diagnostic Products Corporation). HbA_1c was measured on a Synchron CX5 (Beckman Coulter, Brea, CA). HOMA of insulin resistance was calculated.

Plasma concentration of TMAO, choline, and l-carnitine was measured at Preventive Research Laboratory and Laboratory Diagnostic Core, Cleveland Clinic. The details of the measurements have previously been described (11). Briefly, fasting blood samples were obtained at baseline and 6 months and stored at −80°C. Stable isotope dilution high-performance liquid chromatography with electrospray ionization tandem mass spectrometry was used to measure circulating levels of TMAO, choline, and l-carnitine.

Data on TMAO, choline, and l-carnitine were log transformed to improve normality (11). We used the generalized linear model to analyze the association of TMAO, choline, and l-carnitine with BMD at the baseline measurement with adjustment for age, sex, race, and baseline BMI in model 1. We further adjusted for physical activity in model 2. Changes in circulating TMAO, choline, and l-carnitine concentrations from baseline to 6 months during the intervention were calculated to further examine the effect of each 1 log-transformed change in these metabolites on change in BMD at 6 months and 2 years. In model 1, data were adjusted for age, sex, race, diet group, BMI, concurrent weight change, value for the respective outcome traits at the baseline examination, and value for the respective metabolite (TMAO, choline, or l-carnitine) at baseline. In model 2, we additionally adjusted for physical activity. We also conducted longitudinal analyses to test the effect of changes in metabolites on the trajectory of changes in BMD over time by adding the interaction term of change in metabolites and time point using the linear mixed model (PROC MIXED). To determine whether the association of metabolites with changes in BMD can be modified by dietary intake, we added an interaction term of diet intake group (low/high-fat diet groups and low/high-protein diet groups) and change in metabolites in generalized linear models. All statistical comparisons were two sided, and a P value <0.05 was considered statistically significant. Statistical analyses were performed with SAS, version 9.4 (SAS Institute, Cary, NC).

The baseline characteristics of participants are shown in Table 1. At baseline, the mean ± SD age was 52.3 ± 8.9 years and the mean BMI was 32.4 ± 3.7 kg/cm². Of the 264 participants, 86% were white. Mean values of BMD in the spine, hip, and whole body were 1.06 ± 0.15, 1.02 ± 0.14, and 1.13 ± 0.11 g/cm², respectively. Mean values of fasting glucose and HbA_1c were 5.1 ± 0.7 mmol/L and 5.4 ± 0.4% (35.3 ± 4.2 mmol/mol), respectively. The prevalence of hyperglycemia (fasting glucose 5.6 mmol/L or HbA_1c 5.7%) at baseline was 27.3%. No significant differences were observed for age, sex, race, BMI, height, weight, baseline diabetes-related factors, baseline BMD, and dietary intake across the TMAO tertiles. The median plasma concentration of TMAO, choline, and l-carnitine at the baseline measurements was 2.7 mmol/L (interquartile range 1.8), 8.8 mmol/L (3.0), and 35.5 mmol/L (9.0), respectively. No association was found between levels of these metabolites and the spine, hip, and whole-body BMD at baseline (Table 2).

At 6 months, median changes in TMAO, choline, and l-carnitine were 0.0 mmol/L (interquartile range 2.6), −0.3 mmol/L (2.5), and 0.1 mmol/L (7.0) mmol/L, with P = 0.70, 0.006, and 0.30, respectively. We did not find significant differences in changes of TMAO, choline, or l-carnitine across different diet groups. We found that changes in TMAO concentration, rather than its precursors choline and l-carnitine, were associated with changes in whole-body BMD (P = 0.03) (Table 3). A greater reduction in TMAO levels at 6 months was associated with a greater loss in whole-body BMD at 6 months (Fig. 1).

We then analyzed whether the changes in plasma TMAO, choline, and l-carnitine levels from baseline to 6 months were associated with changes in BMD at 2 years. We found that the change in concentration of TMAO, but not choline and l-carnitine, was associated with the changes in the spine (P = 0.005) and whole-body BMD (P = 0.02) (Table 3). A greater reduction in TMAO levels at 6 months was associated with a greater loss in the spine and whole-body BMD at 2 years.

We found that participants with higher baseline HbA_1c tended to have less bone loss in the spine at 2 years (Supplementary Table 1). Thus, we made additional adjustments for baseline HbA_1c, fasting glucose, and parental history of diabetes in the sensitivity analysis, and the results were not essentially changed (Supplementary Table 2).

We also assessed the changes in BMD during the intervention period by different TMAO, choline, and l-carnitine groups using linear mixed models (Fig. 1). The association of change in TMAO level from baseline to 6 months and change in spine BMD was more pronounced at the 2-year visit than at the 6-month visit (P < 0.05). For changes in l-carnitine, similar results were found for change in spine and hip BMD (both P < 0.05). However, we did not find any significant change in association between the change in choline and change in BMD during the intervention period.

Finally, we investigated whether diet interventions modified the effect of change in gut microbiota metabolites on changes in BMD. We found significant interactions between dietary fat intake and changes in l-carnitine on changes in spine and hip BMD at 6 months. Participants with the least decrease in l-carnitine (tertile 3) tended to have less bone loss when assigned to a low-fat diet (Fig. 2). No similar interaction was found for dietary protein intake. We did not find an interaction between the changes in TMAO and choline and dietary intake on the change in BMD. We did not find a significant difference between men and women in the major outcomes reported in this study, which may be partly due to the relatively small sample size and power.

In the current study, we found that a greater reduction in circulating levels of TMAO from baseline to 6 months was associated with a greater loss in spine and whole-body BMD during a 2-year diet intervention study. In addition, we found that the associations of changes in spine and hip BMD with changes in l-carnitine, a precursor of TMAO, were significantly modified by dietary fat intake. Among participants with the tertile of the least decrease in l-carnitine, randomization to a low-fat diet was associated with less bone loss than in the high-fat group.

Weight loss diets have shown consistent benefits in improving cardiometabolic status among obese populations (22–24). However, several studies have shown that weight loss diets were associated with a decrease in BMD (7,25,26), though other studies did not find this relationship (27,28). Emerging evidence from recent studies has linked gut microbiota to both obesity and bone health (9,29). Metabolites related to gut microbiota, such as TMAO, choline, and l-carnitine, have been related to obesity and cardiometabolic disorders (10,11,30,31). In our previous study, we observed associations of changes in TMAO, choline, and l-carnitine with improvement of adiposity, insulin sensitivity, and glucose metabolism among overweight and obese adults during weight loss diet interventions (11,12). These gut microbiota metabolites may also affect bone mass (32).

TMAO is mainly derived from dietary l-carnitine and choline. l-carnitine and choline are abundant nutrients in animal foods, such as meat, fish, and dairy products. Gut microbiota converts the dietary l-carnitine and choline into trimethylamine (TMA), which is subsequently converted into TMAO by hepatic enzyme flavin-containing mono-oxygenates (33). Growing evidence has implicated gut microbiota in regulation of bone mass (13–15), though the underlying mechanisms remain poorly understood. We recently reported that greater decreases in choline and l-carnitine, the precursors of gut microbiota metabolite TMAO, were significantly associated with greater improvements in fasting insulin concentrations and insulin resistance (12). TMAO exacerbates impaired glucose tolerance and obstructs the hepatic insulin signaling pathway (17), and insulin affects bone formation by interacting with the IGF-1 receptor, which is present on osteoblasts in bone, resulting in higher BMD (2). We found that a greater reduction in TMAO during weight loss was associated with a greater loss in spine and whole-body BMD among obese individuals, who had high risk of type 2 diabetes. Of note, diabetes affects bone through multiple pathways, with contradictory effects. Patients with type 2 diabetes have increased fracture risk despite having higher BMD, and the fragility in apparently “strong” bones may relate to impaired bone repair (2,3). Given complex pathophysiological mechanisms underlying bone health in diabetes, further investigations are needed to explore the relations between TMAO and markers of bone fragility in patients with diabetes and the complex interactions with glycemic control. In addition, changes in TMAO levels may reflect alterations in gut microbiota and its related other metabolites. For example, gut microbiota metabolites short-chain fatty acids have been also linked to regulation of bone mass and may inhibit pathological bone loss (34). Further studies are warranted to examine the changes of other gut microbiota–related pathways during weight loss and assess their relations with bone mineral metabolism-related biomarkers in both population and experimental settings.

Intriguingly, we found that dietary fat intake significantly modified the association between changes in l-carnitine and changes in spine and hip BMD. Participants with the least decrease in l-carnitine showed less reduction in BMD in response to a low-fat diet. It is not surprising that, as the dietary precursor of TMAO, l-carnitine is more likely to interact with dietary fat. High-fat diet may increase the level of plasma carnitine by enhancing l-carnitine intestinal absorption and promoting biosynthesis (35,36). Participants with a high-fat diet had significantly higher concentrations of plasma total and free carnitine compared with those with a low-fat diet (35). An animal study showed that a high-fat diet altered the polyadenylation of γ-butyrobetaine hydroxylase (Bbox1) mRNAs, which promoted l-carnitine biosynthesis from γ-butyrobetaine (36). l-carnitine beneficially affects bone health (e.g., slows bone loss, improves bone microstructural properties, and stimulates bone formation) by decreasing bone turnover and increasing osteoblast activity (37,38). In addition, high-fat diet may also adversely affect bone health. For example, the finding of the National Health and Nutrition Examination Survey (NHANES) III showed that saturated fat intake was negatively associated with bone density (20). Taken together, we speculate that high-fat diet may interact with l-carnitine metabolism, thereby affecting l-carnitine function and eventually affecting the changes in BMD.

To the best of our knowledge, this is the first study to assess the association between the changes in circulating levels of metabolites of gut microbiota and bone health during a dietary weight loss intervention. The major strengths of our study included analyses of dynamic changes with repeated measurement of plasma TMAO, choline, and l-carnitine from baseline to 6 months and BMD over 2 years’ dietary intervention. Nevertheless, there are several limitations in our study. First, we did not collect data on gut microbiota in this study and could not evaluate the role of microbiota itself in regulating BMD. Second, the levels of metabolites from gut microbiota might be affected by endogenous or exogenous factors. Third, we did not replicate our observed interactions in other studies; thus, further studies are warranted to validate these findings. Fourth, given increased fat intake reflects decreased carbohydrate intake, it is difficult to distinguish which macronutrient plays the key role behind the observed interactions. Last, intakes of vitamin D and calcium were not validated because the design of POUNDS Lost is to test the effects of diets varying in macronutrient intakes and the validation study (biomarkers of nutrient intake) was specifically to assess macronutrients, not micronutrients.

In conclusion, we found that a decrease in TMAO was related to a reduction in BMD during diet-induced weight loss. Surprisingly, dietary fat modified the effects of changes in l-carnitine, a precursor of TMAO, on bone health. Our data suggest that, even though reduction of TMAO and its precursors during weight loss diet interventions may benefit cardiometabolic health, the weight loss diet–induced changes in gut microbiota metabolites may not be associated with the improvements in bone health among overweight and obese adults.

Acknowledgments. The authors thank all the participants in the study for their dedication and contribution to the research. The authors also thank the Preventive Research Laboratory and Laboratory Diagnostic Core, Cleveland Clinic, for the measurements.

Funding. The study is supported by National Institutes of Health (NIH) grants from the National Heart, Lung, and Blood Institute (Hl071981, Hl034594, Hl126024), the National Institute of Diabetes and Digestive and Kidney Diseases (DK115679, DK091718, DK100383, DK078616), and the Boston Obesity Nutrition Research Center (DK46200) and by United States–Israel Binational Science Foundation grant 2011036. L.Q. is a recipient of the American Heart Association Scientist Development award (0730094N). T.Z. is a recipient of a scholarship under the China Scholarship Council to pursue studies in the U.S. (201606240145).

The sponsors had no role in the design or conduct of the study.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. T.Z. contributed to the study concept and design, analysis and interpretation of data, drafting and revision of the manuscript, statistical analysis, and study supervision. Y.H., Y.C., X.L., D.S., J.A.D., X.P., and M.S.L. contributed to analysis and interpretation of data and drafting and revision of the manuscript. G.A.B. and F.M.S. contributed to acquisition of data, interpretation of data, and drafting and revision of the manuscript. L.Q. contributed to the study concept and design, acquisition of data, analysis and interpretation of data, drafting and revision of the manuscript, statistical analysis, obtainment of funding, and study supervision. L.Q. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented at the American Heart Association EPI|LIFESTYLE 2019 Scientific Sessions, Houston, TX, 5–8 March 2019.