AJCN. First published ahead of print February 10, 2016 as doi: 10.3945/ajcn.115.121269.

Plasma trimethylamine N-oxide concentration is associated with choline, phospholipids, and methyl metabolism1–3 Rima Obeid,4,5* Hussain M Awwad,4 Yannick Rabagny,4 Stefan Graeber,6 Wolfgang Herrmann,4 and Juergen Geisel4 4 Department of Clinical Chemistry and Laboratory Medicine, Saarland University Hospital, Homburg, Germany; 5Aarhus Institute of Advanced Studies, Aarhus University, Aarhus, Denmark; and 6Institute of Medical Biometry, Epidemiology and Medical Informatics, Saarland University, Homburg, Germany

ABSTRACT Background: Elevated plasma concentrations of the gut bacteria choline metabolite trimethylamine N-oxide (TMAO) are associated with atherosclerosis. However, the determinants of TMAO in humans require additional assessment. Objective: We examined cardiometabolic risk factors and pathways associated with TMAO concentrations in humans. Design: A total of 283 individuals (mean 6 SD age: 66.7 6 9.0 y) were included in this observational study. Plasma concentrations of trimethylamine, TMAO, choline, lipids, phospholipids, and methyl metabolites were measured. Results: Study participants were divided into 4 groups by median concentrations of TMAO and choline (4.36 and 9.7 mmol/L, respectively). Compared with the group with TMAO and choline concentrations that were less than the median (n = 82), the group with TMAO and choline concentrations that were at least the median (n = 83) was older and had lower high-density lipoprotein (HDL) cholesterol, phospholipids, and methylation potential, higher creatinine, betaine, S-adenosylhomocysteine (SAH), and S-adenosylmethionine (SAM), and higher percentages of men and subjects with diabetes. The difference in plasma TMAO concentrations between men and women (7.3 6 10.0 compared with 5.4 6 5.6 mmol/L, respectively) was NS after adjustment for age and creatinine (P = 0.455). The TMAO:trimethylamine ratio was higher in men (P , 0.001). Diabetes was associated with significantly higher plasma TMAO concentration (8.6 6 12.2 compared with 5.4 6 5.2 mmol/L) even after adjustments. Sex and diabetes showed an interactive effect on trimethylamine concentrations (P = 0.010) but not on TMAO concentrations (P = 0.950). Positive determinants of TMAO in a stepwise regression model that applied to the whole group were SAH, trimethylamine, choline, and female sex, whereas plasma phosphatidylcholine was a negative determinant. Conclusions: High TMAO and choline concentrations are associated with an advanced cardiometabolic risk profile. Diabetes is related to higher plasma TMAO concentrations but also to alterations in interrelated pathways such as lipids, phospholipids, and methylation. Elevated plasma TMAO concentrations likely reflect a specific metabolic pattern characterized by low HDL and phospholipids in addition to hypomethylation. This trial was registered at clinicaltrials.gov as NCT02586181 and NCT02588898. Am J Clin Nutr doi: 10.3945/ajcn.115.121269. Keywords: choline, metabolism, methyl, phospholipids, trimethylamine N-oxide

INTRODUCTION

Elevated plasma concentrations of the intestinal bacteria choline metabolite trimethylamine N-oxide (TMAO)6 are associated with diabetes (1), renal insufficiency (2), and vascular diseases (3–5). However, an adequate dietary intake of choline is necessary to maintain human health because the endogenous liver synthesis of phosphatidylcholine is not sufficient. These 2 contradictory aspects of choline challenge its role in health and render its function as a nutrient a potential double-edged sword. TMAO is produced from trimethylamine through the action of flavin-containing monooxygenase 3 (FMO3) in the liver. Trimethylamine is formed, in turn, from several trimethylaminecontaining compounds in foods by certain intestinal bacteria. Choline derivatives that are not absorbed in the upper part of the intestine are considered available for conversion to trimethylamine in the gut and later to TMAO in the host liver (Figure 1). A few studies have reported protective effects of the FMO3 system on liver and nerve functions (6, 7). A loss of function of FMO3 activity (trimethylaminuria) causes low plasma TMAO and high trimethylamine concentrations; however, to our knowledge, no other metabolic or physical symptoms have been closely investigated in affected patients. In contrast, a gain of function of FMO3 (i.e., a high TMAO concentration) in animal models has 1

Supported by a Marie-Curie Fellowship at Aarhus University, Denmark (to RO), and a scholarship from the Ministry of Higher Education, Syria (to HMA). The ultraperformance liquid chromatography system was funded by the German Research Foundation (Deutsche Forschungsgemeinschaft), which is a public research funding organization, and the University of Saarland (project 63813068). 2 Funding agents had no role in the design or performance of the study or in the interpretation of the results. 3 Supplemental Figures 1–3 and Supplemental Tables 1–3 are available from the “Online Supporting Material” link in the online posting of the article and from the same link in the online table of contents at http://ajcn. nutrition.org. *To whom correspondence should be addressed. E-mail: rima.obeid@ uks.eu. 6 Abbreviations used: FMO3, flavin-containing monooxygenase 3; GLM, general linear model; SAH, S-adenosylhomocysteine; SAM, S-adenosylmethionine; tHcy, total homocysteine; TMAO, trimethylamine N-oxide. Received August 13, 2015. Accepted for publication December 29, 2015. doi: 10.3945/ajcn.115.121269.

Am J Clin Nutr doi: 10.3945/ajcn.115.121269. Printed in USA. Ó 2016 American Society for Nutrition

Copyright (C) 2016 by the American Society for Nutrition

1 of 9

2 of 9

OBEID ET AL.

been associated with atherosclerosis through complex mechanisms that involve cholesterol metabolism. Plasma TMAO concentrations were shown to increase in apolipoprotein E2/2 mice that were fed a choline-rich diet, and this increase was associated with an enhanced cholesterol accumulation in macrophages (3). TMAO is a universal osmolyte with a tissue content in marine animals that is correlated with body lipid mass (i.e., percentage of digestive gland mass) and has been suggested to regulate lipid homeostasis (8). In humans, TMAO concentrations increased a few hours after a high-fat meal, and this early change was discussed as reflecting changes in the endogenous TMAO pool (9). Accumulating evidence has suggested that the FMO3-TMAO system is likely to be a key modifier of the balance between cholesterol intake, absorption, transport, and tissue storage (6, 10, 11). Thus, the plasma concentration of TMAO might serve as a potential marker for imbalances between lipids and other metabolic factors that, themselves, enhance atherosclerosis. Choline is a source of phospholipids that constitute integral parts of lipoprotein particles. In addition, through their amphipathic nature, phospholipids participate in solubilizing cholesterol and in reverse cholesterol transport. Alternatively, choline is oxidized to betaine, which is a precursor of S-adenosylmethionine (SAM) that, in turn, is used for endogenous phospholipids synthesis. The binary role of choline in methyl and phospholipid metabolisms poses a question regarding the potential interactions between TMAO, methyl, and phospholipid-producing pathways. Human studies are required to elucidate the determinants of plasma TMAO concentrations and the role of the FMO3-TMAO system in whole-body lipid metabolism. The aim of the current study was to investigate the cardiometabolic risk factors associated with plasma concentrations of TMAO and focuses on plasma TMAO and choline in relation to lipid, phospholipid, and methyl metabolites.

METHODS

Subjects and study design Participants in the current study (n = 283) were originally recruited into the following 2 studies: a diabetes case-control study (12) (clinicaltrials.gov; NCT02588898) and a vitamin-supplementation trial (baseline data; clinicaltrials.gov; NCT02586181) (Supplemental Figures 1–3) (13, 14). All participants were recruited at the Saarland University Hospital between 2009 and 2011. All laboratory analyses were performed at the Department of Clinical Chemistry and Laboratory Medicine, Saarland University Hospital. Fasting blood samples were collected in dry tubes or tubes that contained K+ EDTA for obtaining serum and EDTA-plasma samples, respectively. Inclusion in the current study was based on the availability of novel EDTA-plasma aliquots for measurements of phospholipids, choline metabolites, TMAO, and trimethylamine. All markers were measured #1 y after the sample collection except for TMAO and trimethylamine, which were measured #5 y after the sample collection. Exclusion criteria in the original studies were as follows: cancer, methotrexate use, recent vascular event or recent surgical procedure, renal dysfunction (glomerular filtration rate .50 mL $ min21$ 1.73 m22), hepatic dysfunction, gastric or ileum resection, or B-vitamin supplementation. Information on medical histories, medication used, and traditional risk factors was available. The

studies were conducted in accordance with the ethical principles stated in the Declaration of Helsinki. The study protocols were approved by the ethical committee of the Saarland region, and all participants provided written informed consent. Assays and quality assurance Concentrations of TMAO and trimethylamine were measured in the EDTA-plasma samples with the use of ultraperformance liquid chromatography–tandem mass spectrometry with isotope labeled internal standards (d9-TMAO and d9-trimethylamine). Samples were extracted with the use of methanol:acetonitrile (15:85) that contained 0.2% formic acid. After the addition of the internal standards, the samples were mixed, centrifuged at room temperature for 5 min at 10,000 3 g, and separated on an Acquity UPLC BEH HILIC column (100 3 2.1 mm) with an Acquity HILIC VanGuard precolumn (5 3 2.1 mm) (Waters Corp.). The column temperature was set to 308C. The flow rate was 0.4 mL/min. The gradient elution consisted of 15 mmol/L ammonium formate (solvent A; pH: 3.5) and acetonitrile (solvent B). TMAO and trimethylamine were identified in positive multiple-reaction monitoring mode with the use of characteristic precursor-product ion transitions as follows: m/z 75.90/58.10 and 59.98/44.02, respectively. d9-TMAO and d9-trimethylamine were monitored at m/z 84.90/66.05 and 129.77/48.09, respectively. The between-day CVs of the 2 markers in the quality-control samples were ,7% at 3 concentrations (between 1 and 90 mmol/L). The limits of detection were a concentration of 0.09 mmol/L for TMAO and a concentration of 0.12 mmol/L for trimethylamine. TMAO (15, 16) and trimethylamine (16) are stable on thawing and freezing. TMAO has been shown to be stable during storage at 2808C for 5 y (15) although information on the long term-stability of trimethylamine is currently not available. The concentrations of phospholipid species were measured in unthawed EDTA plasma with the use of an ultraperformance liquid chromatography–tandem mass spectrometry–based method as described previously (17). Phospholipid classes (phosphatidylcholine, phosphatidylethanolamine, sphingomyelin, and lysophosphatidylcholine) were considered as the sum of all identified species within one class. Phospholipids were extracted with the use of methanol and methyl tert-butyl ether. The organic extracts were vacuum dried and resuspended in methanol. The separation was performed on an Acquity UPLC BEH Shield RP18 column (100 3 2.1 mm) with an Acquity UPLC BEH Shield RP18 VanGuard precolumn (5 3 2.1 mm) (Waters Corp.) (17). The assay does not differentiate between phospholipid species of the same mass according to their exact contents of fatty acids (17). Plasma concentrations of free choline, betaine, dimethylglycine (18), SAM, S-adenosylhomocysteine (SAH) (19), and total homocysteine (tHcy) in addition to serum folate and creatinine were measured with the use of established methods. Plasma lipids and lipoproteins were measured with the use of routine reagents (Cobas System; Roche Diagnostics). Statistical analyses were conducted with the use of IBM SPSS Statistics 22 for Windows software (SPSS Inc.). A KolmogorovSmirnov test was used to check for a normality assumption. Age and cholesterol were normally distributed. All other continuous variables were log transformed to approach normal distribution before the application of general linear model (GLM) univariate

DETERMINANTS OF PLASMA TRIMETHYLAMINE N-OXIDE

3 of 9

FIGURE 1 Sources, metabolism, and elimination of TMAO and potential interactions with lipids, phospholipids, and methyl metabolisms. Gut bacteria, dietary intake, renal function, biliary and nonbiliary cholesterol and phospholipid transport, and methyl metabolism may affect plasma TMAO concentrations. FMO3, flavin-containing monooxygenase 3; PtdChol, phosphatidylcholine; PtdEth, phosphatidylethanolamine; SAH, S-adenosylhomocysteine; SAM, S-adenosylmethionine; TMA, trimethylamine; TMAO, trimethylamine N-oxide.

analyses or regression analyses. The results of continuous variables are shown as mean 6 SD. A GLM univariate analysis was used to test for differences in continuous variables between groups of TMAO and choline. When the test was significant, the post hoc Tamhane’s T2 test was used for multiple comparisons (Tables 1 and 2). GLM univariate analyses were also applied in Table 3 to study the differences between the means of fixed factors such as sex and diabetes and the interaction between sex and diabetes. Covariates such as age and log creatinine were entered in the univariate analysis as indicated in the tables. Stepwise multiple linear regression analyses were applied to identify significant predictors of TMAO in the whole group and according to the presence or absence of diabetes or statin use (Table 4). Differences between categorical variables were tested with the use of a chi-square test. Pearson’s test was applied on the log-transformed data (except for age) to study correlations between TMAO and trimethylamine with other markers. All tests were 2 sided, and P , 0.05 was considered statistically significant. RESULTS

The study included 283 participants (mean 6 SD age: 66.7 6 9.0 y). The majority of participants were taking glucose-lowering agents, statins, or antihypertensive or antiplatelet drugs at the time of the sample collection. Supplemental Table 1 shows the main characteristics of the whole group and according to the source

population. The studies differed in the proportions of women, patients with type 2 diabetes, and users of certain drugs. TMAO and choline concentrations in relation to cardiometabolic risk factors We investigated the associations between high TMAO and choline concentrations with metabolic and vascular risk factors. Four groups were studied according to 4 possible combinations of plasma TMAO and choline concentrations that were separated by the medians of the whole group (median choline concentration: 9.7 mmol/L; median TMAO concentration: 4.36 mmol/L) (Table 1). The group with lowTMAO and low-choline concentrations was considered the reference group (group 1). Marked differences in several blood biomarkers were observed between the groups in particular for group 1 compared with group 4 (high-TMAO and high-choline concentrations). Compared with group 1, group 4 had higher proportions of men, patients with diabetes, and statin users (Table 1). Participants in group 4 had the worst risk profile; group 4, compared with group 1, was significantly older (mean: 69.9 compared with 64.3 y, respectively) and had a lower mean plasma HDL-cholesterol concentration (1.33 compared with 1.73 mmol/L, respectively) and a higher mean creatinine concentration (94.6 compared with 75.1 mmol/L, respectively) (Table 1). Differences in HDL cholesterol between the groups persisted (P = 0.024) after adjustment for age, creatinine, sex, diabetes, and statin use.

4 of 9

OBEID ET AL.

TABLE 1 Biomarkers and conditions associated with atherosclerosis according to the combination of plasma TMAO and free choline1

Free choline, mmol/L TMAO, mmol/L n Women, n (%) Type 2 diabetes, n (%) Statin use, n (%) Age, y BMI, kg/m2 Total cholesterol, mmol/L Triglycerides, mmol/L HDL cholesterol, mmol/L LDL cholesterol, mmol/L Glucose, mmol/L Creatinine, mmol/L

Group 1

Group 2

Group 3

Group 4

P2

P3

P4

Low (,9.7) Low (,4.36) 82 45 (55) 15 (18) 15 (18) 64.3 6 8.76 26.3 6 3.6 5.52 6 0.98 1.42 6 0.90 1.73 6 0.48 3.16 6 0.96 5.88 6 1.22 75.1 6 18.6

High ($9.7) Low (,4.36) 57 26 (46) 17 (30) 13 (23) 66.8 6 8.1 28.1 6 4.7 5.59 6 1.04 1.44 6 0.72 1.63 6 0.44 3.32 6 0.85 5.99 6 1.44 77.8 6 14.1

Low (,9.7) High ($4.36) 59 26 (44) 26 (44) 18 (31) 65.5 6 9.0 27.5 6 4.2 5.13 6 0.88 1.41 6 0.93 1.45 6 0.437 3.03 6 0.78 6.71 6 1.787 79.6 6 18.6

High ($9.7) High ($4.36) 83 24 (29) 39 (47) 33 (40) 69.9 6 9.37,8 28.1 6 4.5 5.23 6 1.22 1.81 6 1.18 1.33 6 0.427,9 3.11 6 1.19 6.33 6 2.05 94.6 6 26.57–9

— — — 0.0045 ,0.0015 0.0025 0.001 0.043 0.136 0.131 ,0.001 0.718 0.024 ,0.001

— — — — — — — 0.344 0.180 0.692 0.016 0.718 0.024 —

— — — — — — — 0.226 0.442 0.517 0.024 0.497 0.664 —

1

Concentrations of plasma free choline and TMAO were divided by the corresponding medians in the entire group (n = 283; median choline concentration: 9.7 mmol/L; median TMAO concentration: 4.36 mmol/L) to get the 4 groups shown in the table. Group 1 was the reference group. Log-transformed continuous variables were used except for age and cholesterol. TMAO, trimethylamine N-oxide. 2 Unless otherwise noted, P values were determined with the use of a general linear model univariate analysis with the group included as a fixed factor; a post hoc Tamhane’s T2 test was applied when the results of the general linear model were significant. 3 P values were determined with the use of a general linear model univariate analyses including age and log-creatinine as covariates. 4 P values were determined with the use of a general linear model univariate analyses including age, log-creatinine, sex, diabetes and statin use as covariates. 5 P values were determined with the use of a chi-square test. 6 Mean 6 SD (all such values). 7–9 P , 0.05 (post hoc comparisons): 7compared with the reference group; 8compared with group 3; 9compared with group 2.

Plasma concentrations of trimethylamine were not significantly different between the 4 groups (Table 2). Furthermore, concentrations of plasma betaine, dimethylglycine, phosphatidylcholine, phosphatidylethanolamine, lysophosphatidylcholine, and SAH differed between the groups, and the differences persisted after multiple adjustments for age, creatinine, sex, diabetes, and statin use (Table 2). Differences in total sphingomyelin, tHcy, folate, and the SAM: SAH ratio were NS after multiple adjustments, and plasma SAM concentrations tended to be lower in group 1 (P-adjusted = 0.107). To further minimize the effects of disease and statin use on the associations between TMAO and other metabolites, we repeated the analyses shown in Tables 1 and 2 in a subgroup consisting of 149 participants who were free of diabetes and statin use (Supplemental Table 2). Between-group differences for HDL cholesterol, SAH, and the SAM:SAH ratio remained significant after adjustment for age, creatinine, and sex (Supplemental Table 2). Interactions of TMAO concentrations with sex and diabetes We further investigated the effects of sex and diabetes and their interactions on plasma lipids, TMAO, lipoproteins, phospholipids, and methyl-related metabolites (Table 3). The univariate analyses simultaneously included sex and diabetes as fixed factors and age and log creatinine as covariates. Sex showed a significant influence on HDL-cholesterol, trimethylamine, phosphatidylcholine, phosphatidylethanolamine, sphingomyelin, and folate concentrations (higher in women) as well as on the TMAO:trimethylamine ratio and betaine concentrations (lower in women) (Table 3). The presence of diabetes was related to significantly lower concentrations of HDL cholesterol, phosphatidylcholine, lysophosphatidylcholine, and sphingomyelin in addition to higher

concentrations of TMAO, SAM, and SAH. There were significant interactive effects of sex and diabetes on plasma trimethylamine and sphingomyelin but not on TMAO (Table 3). Determinants of TMAO from regression analyses We applied stepwise multiple regression analyses to identify the significant predictors of TMAO plasma concentrations in the whole group (n = 283) and in subgroups including individuals with diabetes (n = 98) or without diabetes (n = 185), those not taking statins (n = 203) or taking statins (n = 80), and individuals without statin use and diabetes (n = 149) (Table 4). SAH was a significant positive predictor of TMAO in all groups except for statin users. Trimethylamine was a significant positive determinant of TMAO in all groups except for subjects without diabetes. Plasma choline was a significant positive determinant in all groups except for patients with diabetes. Female sex was a positive predictor of TMAO in subjects without diabetes, whereas diabetes was a positive predictor of TMAO in nonstatin users. Negative determinants of TMAO were phosphatidylcholine in the whole group, cholesterol and lysophosphatidylcholine in statin users, and HDL cholesterol in the no-statin, no-diabetes subgroup. Simple correlations of TMAO and trimethylamine with other markers Plasma TMAO concentrations in the whole group showed positive correlations with age, creatinine, HDL cholesterol, betaine, choline, dimethylglycine, tHcy, SAH, and SAM (Supplemental Table 3). TMAO showed negative correlations with phosphatidylcholine, lysophosphatidylcholine, folate, and the SAM:SAH ratio. The correlations of trimethylamine with the previously

5 of 9

DETERMINANTS OF PLASMA TRIMETHYLAMINE N-OXIDE TABLE 2 Plasma choline-, phospholipids-, and methylation-related markers according to plasma TMAO and free choline1

Free choline, mmol/L TMAO, mmol/L Plasma choline-related markers Free choline, mmol/L TMAO, mmol/L Trimethylamine, mmol/L TMAO:trimethylamine ratio Betaine, mmol/L Dimethylglycine, mmol/L Plasma phospholipids Total phosphatidylcholine, mmol/L Total phosphatidylethanolamine, mmol/L Total lysophosphatidylcholine, mmol/L Total sphingomyelin, mmol/L Methyl-related markers tHcy, mmol/L Folate, ng/mL SAM, nmol/L SAH, nmol/L SAM:SAH ratio

Group 1

Group 2

Group 3

Group 4

P2

P3

P4

Low (,9.7) Low (,4.36)

High ($9.7) Low (,4.36)

Low (,9.7) High ($4.36)

High ($9.7) High ($4.36)

— —

— —

— —

7.9 2.8 0.43 36.5 32.6 2.91

6 6 6 6 6 6

1.2 1.0 0.30 117.8 9.1 0.83

11.9 2.8 0.34 22.0 37.8 3.97

6 6 6 6 6 6

1.75 1.0 0.26 33.5 14.6 1.75

8.1 7.5 0.46 55.8 34.1 3.32

6 6 6 6 6 6

1.16 4.25,6 0.44 101.05,6 9.7 1.37

13.0 12.1 0.66 58.4 40.6 4.27

6 6 6 6 6 6

2.75–7 13.25–7 1.33 80.85,6 11.65,7 1.785,7

,0.001 ,0.001 0.795 ,0.001 ,0.001 ,0.001

,0.001 ,0.001 0.318 ,0.001 0.014 0.014

,0.001 ,0.001 0.575 ,0.001 0.002 0.014

2005 25.9 166 411

6 6 6 6

384 12.8 39 83

2008 30.9 171 395

6 6 6 6

441 15.5 53 92

1772 23.1 145 378

6 6 6 6

3185,6 13.36 405,6 89

1901 31.3 146 368

6 6 6 6

475 18.77 435,6 945

0.002 0.016 ,0.001 0.007

0.007 0.106 0.002 0.409

0.036 0.037 0.027 0.771

14.0 12.6 113 13.8 9.3

6 6 6 6 6

4.7 7.0 21 5.7 4.0

13.9 13.9 126 16.5 8.3

6 6 6 6 6

5.0 7.5 305 7.15 2.1

13.7 6 6.3 13.8 7.9 125 6 265 16.3 6 6.55 8.4 6 2.6

16.2 11.2 146 23.6 7.0

6 6 6 6 6

8.37 7.6 395–7 12.65–7 2.15–7

0.035 0.086 ,0.001 ,0.001 ,0.001

0.138 0.354 0.027 0.002 0.147

0.138 0.257 0.107 0.008 0.169

All values are means 6 SDs. Concentrations of plasma free choline and TMAO were divided by medians (n = 283; median choline concentration: 9.7 mmol/L; median TMAO concentration: 4.36 mmol/L) to get the 4 groups. Group 1 was the reference group. Log-transformed continuous variables are used except for age. SAH, S-adenosylhomocysteine; SAM, S-adenosylmethionine; tHcy, total homocysteine; TMAO, trimethylamine N-oxide. 2 P values were determined with the use of a general linear model univariate analysis with the group included as a fixed factor; a post hoc Tamhane’s T2 test was applied when the results of the general linear model were significant. 3 P values were determined with the use of general linear model univariate analyses including age and log creatinine as covariates. 4 P values were determined with the use of general linear model univariate analyses including age, log-creatinine, sex, diabetes, and statin use as covariates. 5–7 P , 0.05 (post hoc comparisons): 5compared with the reference group; 6compared with group 2; 7compared with group 3. 1

mentioned markers were either NS or in the reverse direction compared with that with TMAO (Supplemental Table 3). TMAO concentrations and common drug use In patients with diabetes, the mean plasma concentration of TMAO showed no significant difference between metformin treated patients (n = 58) and untreated patients (n = 40) (mean 6 SD: 9.0 6 14.5 compared with 8.1 6 7.8 mmol/L, respectively; P = 0.890) even after multiple adjustments. Of other common medications, b-blockers (P , 0.001), calcium channel inhibitors (P = 0.003), statins (P = 0.012), and diuretics (P = 0.020) were associated with higher plasma TMAO concentrations. However, because of the variety of drug use and background diseases, the current study could not rule out associations between drugs and plasma TMAO concentrations. DISCUSSION

The potential role for choline and TMAO in atherosclerosis requires a careful assessment. Choline is essential for human health and carries no known adverse effects (20, 21). An elevated baseline plasma TMAO concentration has been associated with traditional atherosclerosis risk factors such as male sex, age, or a lower glomerular filtration rate and with major adverse vascular events after a 3-y follow-up (22). High plasma concentrations of both TMAO and free choline have been shown to predict future risk of vascular morbidities compared with when both markers are present at low concentrations (5). However, little is known from human studies regarding other metabolic conditions that

might be associated with high TMAO and choline concentrations. In the current article, we report the presence of complex associations between elevated plasma concentrations of TMAO and choline with several cardiometabolic traits such as age, low HDLcholesterol concentrations, and elevated creatinine concentrations (Table 1). In this study, low plasma TMAO and choline concentrations were linked to a favorable risk profile including higher plasma phospholipids and a higher methylation capacity than when both TMAO and choline concentrations were above the corresponding medians (Table 2). However, the interdependency of these metabolic pathways and the observational nature of the current study did not allow us to draw inferences regarding potential causal relations.

TMAO is part of a metabolic interplay between lipids, phospholipids, and methyl moieties The inverse association of TMAO and choline with plasma HDL-cholesterol (Table 1) and phospholipid concentrations and the direct association with methylation markers (betaine, dimethylglycine, and SAH) (Table 2) suggest that higher cholinecontaining phospholipids coincide with less choline being used as a methyl donor. This metabolic pattern suggests that TMAO is elevated under hypomethylation conditions, when less choline is used for phospholipid synthesis, or when phospholipid concentrations are low because of low intake or endogenous synthesis. Thus, our results suggest that there is an existence of a yet unknown role for HDL cholesterol, phospholipid synthesis, and methyl donors in determining plasma TMAO concentrations.

6 of 9

OBEID ET AL.

TABLE 3 Plasma TMAO-related biomarkers according to sex and diabetes1 Sex

Diabetes Men

No

P Sex

Diabetes

Sex 3 diabetes

9.5 4.8 1.00 1.07 0.41 1.00 2.13 22.8

— 0.072 0.069 0.083 ,0.001 0.993 0.371 —

— ,0.001 ,0.001 0.004 0.012 ,0.001 ,0.001 —

— 0.774 0.032 0.805 0.283 0.219 0.043 —

6 6 6 6 6 6

3.0 12.2 1.15 56.8 12.7 1.3

0.637 0.455 ,0.001 ,0.001 ,0.001 0.640

0.256 0.003 0.133 0.610 0.155 0.834

0.900 0.950 0.010 0.031 0.436 0.686

1909 32.6 140 366

6 6 6 6

447 17.5 37 80

,0.001 0.004 0.310 ,0.001

0.002 0.135 ,0.001 0.001

0.892 0.971 0.409 0.027

14.2 12.2 136 19.1 8.0

6 6 6 6 6

5.5 7.1 32 9.0 2.4

0.951 ,0.001 0.270 0.256 0.573

0.741 0.393 0.001 0.029 0.655

0.676 0.619 0.898 0.866 0.799

Variable

Women

Yes

n Age, y BMI, kg/m2 Total cholesterol, mmol/L Triglycerides, mmol/L HDL cholesterol, mmol/L LDL cholesterol, mmol/L Glucose, mmol/L Creatinine, mmol/L Plasma choline-related markers Free choline, mmol/L TMAO, mmol/L Trimethylamine, mmol/L TMAO:trimethylamine ratio Betaine, mmol/L Dimethylglycine, mmol/L Plasma phospholipids Total phosphatidylcholine, mmol/L Total phosphatidylethanolamine, mmol/L Total lysophosphatidylcholine, mmol/L Total sphingomyelin, mmol/L Methyl-related markers tHcy, mmol/L Serum folate, ng/mL SAM, nmol/L SAH, nmol/L SAM:SAH ratio

122 64.4 6 9.3 26.5 6 4.2 5.6 6 1.03 1.31 6 0.65 1.75 6 0.45 3.27 6 0.97 6.16 6 1.45 69.1 6 16.3

68.5 28.2 5.01 1.85 1.25 2.93 6.29 91.0

161 6 8.42 6 4.2 6 0.93 6 1.24 6 0.32 6 0.90 6 1.81 6 20.02

66.5 26.4 5.62 1.38 1.63 3.36 5.63 82.0

185 6 8.8 6 3.5 6 0.95 6 0.86 6 0.49 6 0.85 6 0.89 6 20.9

67.2 29.3 4.89 1.80 1.40 2.70 7.81 80.6

98 6 6 6 6 6 6 6 6

9.6 5.4 0.59 11.9 31.2 3.0

6 6 6 6 6 6

2.7 5.6 0.53 12.5 8.2 1.7

10.7 7.3 0.40 68.1 40.2 3.8

6 6 6 6 6 6

3.0 10.0 0.92 115.7 12.5 1.4

9.9 5.4 0.41 49.8 36.7 3.6

6 6 6 6 6 6

2.9 5.2 0.48 105.5 11.2 1.5

10.9 8.6 0.62 32.6 35.6 3.4

2062 28.3 155 442

6 6 6 6

363 13.9 43 84

1828 27.9 158 349

6 6 6 6

431 16.9 46 72

1937 25.7 165 400

6 6 6 6

405 14.1 46 94

13.3 21.7 119 15.0 8.9

6 6 6 6 6

7.0 16.7 35 9.8 2.3

15.5 18.3 135 19.8 7.7

6 6 6 6 6

5.7 12.6 29 8.8 3.4

14.8 13.0 124 17.0 8.4

6 6 6 6 6

6.8 7.6 32 9.7 3.3

1 All values are means 6 SDs. Log-transformed continuous variables were used except for age and cholesterol. P values were determined with the use of a general linear model univariate analysis that simultaneously included sex and diabetes as fixed factors and age and log creatinine as covariates. SAH, S-adenosylhomocysteine; SAM, S-adenosylmethionine; tHcy, total homocysteine; TMAO, trimethylamine N-oxide. 2 Age and creatinine were significantly different between men and women (general linear model univariate analysis).

There are only a few modifiable factors known to affect plasma TMAO. Dietary choline intake in the range recommended for promoting human health influences plasma TMAO (23). Nevertheless, only 11–15% of dietary choline (#714 mg egg phospholipids) is converted to TMAO (on the basis of a 24-h urine excretion of TMAO) (23). A 10-fold increase of choline intake causes a less proportional increase in plasma and urine TMAO and trimethylamine concentrations in animals (3–4-fold) (24). Gut bacteria is known to modify trimethylamine and TMAO (25), although taking probiotics has failed to lower TMAO and has shown mixed results on biomarkers of vascular risk (26–28). We have identified metabolically related pathways (i.e., phospholipids and methylation) that offer a potential modifiable effect on TMAO (Figure 1). TMAO interactions with sex and diabetes Sex had a significant influence on trimethylamine and phospholipids (higher in women) and the TMAO:trimethylamine ratio (lower in women) but not on TMAO after adjustment for age and creatinine (Table 3). Differences in the TMAO:trimethylamine ratio were likely due to differences in FMO3 activity or substrate load. Our results suggest that women might have less-efficient FMO3 enzyme activity than do men. Wang et al. (3) also did not observe sex-differences in plasma TMAO concentrations. In contrast, a higher expression of FMO3 has been reported in the

female mouse liver (1), which is not consistent with our findings or those from Wang et al. (3). Choline and methyl metabolisms are sex dependent (29–31). Estrogen enhances the endogenous synthesis of phosphatidylcholine from phosphatidylethanolamine (32), thereby causing higher phosphatidylcholine in women and, thus, enhanced HDL-mediated cholesterol clearance. Therefore, in addition to FMO3 activity, women might have different substrate inputs for TMAO synthesis. Our study confirmed earlier observations on the association between diabetes and elevated plasma TMAO concentrations (1) (Table 3). Diabetes also has an influence on lipoproteins (lower HDL-cholesterol concentrations), phospholipids (lower concentrations), and methylation markers (increased SAM and SAH concentrations) (Table 3). These diabetes-related conditions might independently affect TMAO. However, the cause of TMAO elevation in diabetes remains to be identified. Cardiometabolic predictors of TMAO plasma concentration To our knowledge, the regression models primarily showed 3 novel points (Table 4). First, phosphatidylcholine, lysophosphatidylcholine, or HDL cholesterol was identified as the sole negative predictor of plasma TMAO concentrations. Second, in addition to the direct substrates (i.e., choline and trimethylamine), SAH was a dominant

DETERMINANTS OF PLASMA TRIMETHYLAMINE N-OXIDE

7 of 9

Impact of the current study on future research

34) and potentially regulates cholesterol transport. Each day, #1300 mg cholesterol is exported from the bile to the gut (35) and an additional 350 mg passes to the gut via intestinal transport (36). This cholesterol must be emulsified to prevent crystallization. Phosphatidylcholine is secreted in the bile (37) where it serves as a cholesterol emulsifier (38) that promotes cholesterol transport (39). TMAO can be considered as an alternative means for solubilizing cholesterol in people with high cholesterol intake, low phospholipids, or low HDL cholesterol (male sex and diabetes) (Figure 1). This theory would also be relevant for tissue TMAO as a predictor of body fat mass (8) and for the role of TMAO in enhancing cholesterol uptake into macrophages (3) potentially by maintaining the solubility of cholesterol. The efficacy of converting carnitine, lecithin, phosphatidylcholine, or sphingomyelin to trimethylamine has not been studied to our knowledge. The transient increase in TMAO after a highfat meal (9) suggests that endogenous phospholipids or tissue stores of TMAO released to the bile after the meal might be a continuous source of TMAO particularly in men or in patients with diabetes. TMAO-like compounds exhibit antiparasite effects (40, 41) through the alteration of phosphatidylcholine metabolism in the parasite (40) and choline uptake into infected erythrocytes and, thus, the synthesis of other choline intermediates (42). Therefore, trimethylamine or TMAO might be used by gut bacteria to prevent the growth of other competing pathogens or to regulate the amount of phosphatidylcholine biosynthesis in the host. This possibility is consistent with the role of gut microflora overall in the health and immunity of the host (43). Finally, a few limitations of the current study deserve mention. The combination of 2 cohorts with different study designs might have introduced a selection bias. However, the studies were conducted during the same period, and the assays were performed with the use of identical methods. In addition, we did not collect data on dietary intakes of choline, carnitine, and cholesterol, which might have differed according to sex and diabetes status. Moreover, we could not rule out whether common medications that were used by patients with diabetes might have influenced the TMAO-concentration measures independent of the presence of diabetes. Finally, the trimethylamine stability in stored samples is currently not known, which may have introduced a preanalytic bias. In conclusion, our results show that high plasma TMAO and choline concentrations were associated with lower HDL cholesterol and plasma phospholipids and hypomethylation. The combination of high TMAO and choline concentrations reflect a more-advanced risk profile that is associated with higher age, reduced renal function, male sex, and diabetes. Diabetes was associated with higher TMAO concentrations that were accompanied by other metabolic conditions such as low HDL cholesterol and phospholipids and hypomethylation. Thus, an elevated plasma TMAO concentration appear to be predicted by a complex interplay between pathways with causal links to atherosclerosis. The potential modifying roles of methyl donors and phospholipids on plasma TMAO concentrations deserve additional investigation.

Animal models of FMO3 and TMAO metabolism have shown variability, and the extrapolation of results obtained therein to humans requires care (1, 3, 6). FMO3 has several substrates (33,

We thank Christiane Waldura for the technical help in measuring TMAO in plasma. We also thank Susanne Kirsch for the supportive role in conducting the original studies.

TABLE 4 Determinants of TMAO from stepwise multiple linear regression analyses1 Variable entered Whole group (n = 283, R2 = 0.36) SAH TMA Choline Phosphatidylcholine Sex, F No statins (n = 203, R2 = 0.31) SAH TMA Choline Diabetes, yes Statin users (n = 80, R2 = 0.52) TMA Choline Cholesterol Lysophosphatidylcholine No diabetes (n = 185, R2 = 0.28) SAH Choline Sex, F Type 2 diabetes (n = 98, R2 = 0.37) SAH Trimethylamine No statins and no diabetes (n = 149, R2 = 0.39) SAH HDL cholesterol Choline Trimethylamine Sex, F

b (95% CI)2

0.703 0.374 0.667 20.894 0.135

0.553 0.302 0.656 0.134

0.532 0.945 20.005 20.995

(0.397, 1.008) (0.216, 0.533) (0.274, 1.060) (21.160, 20.327) (0.031, 0.240)

(0.233, (0.102, (0.209, (0.013,

0.872) 0.501) 1.103) 0.254)

(0.258, 0.807) (0.203, 1.686) (20.008, 20.002) (21.681, 20.309)

P

,0.001 ,0.001 0.001 0.002 0.012

0.001 0.003 0.004 0.030

,0.001 0.014 0.001 0.006

0.831 (0.474, 1.189) 0.576 (0.144, 1.009) 0.164 (0.046, 0.282)

,0.001 0.010 0.007

0.654 (0.124, 1.185) 0.592 (0.336, 0.847)

0.017 ,0.001

0.735 20.821 0.576 0.259 0.208

(0.357, 1.114) (21.364, 20.278) (0.131, 1.021) (0.049, 0.469) (0.077, 0.339)

,0.001 0.002 0.012 0.016 0.002

1 The dependent variable was log TMAO. Adjusted R2 values are shown. All continuous variables were entered as logarithmic values except for age and cholesterol. Only variables entered in the final equation are shown. Variables in the regression model were as follows: female sex (one for women; zero for men) and age, creatinine, total cholesterol, triglycerides, HDL cholesterol, LDL cholesterol, free choline, trimethylamine, betaine, phosphatidylcholine, phosphatidylethanolamine, lysophosphatidylcholine, sphingomyelin, total homocysteine, S-adenosylmethionine, and SAH (all as continuous variables). The presence of diabetes and statin use was entered as a dichotomous variable (one if yes, and zero if no) in the whole group or as required in the subgroup analyses. SAH, S-adenosylhomocysteine; TMA, trimethylamine; TMAO, trimethylamine N-oxide. 2 Unstandardized b coefficients.

positive predictor of TMAO concentrations. Third, subgroup analyses have shown that diabetes and statin use alter the predictors of TMAO concentrations. If the regression analysis is used for hypothesis generation, our findings suggest that phosphatidylcholine or methyl donors (i.e., B vitamins) may potentially lower plasma TMAO concentrations.

8 of 9

OBEID ET AL.

The authors’ responsibilities were as follows—RO: participated in planning the studies, blood assays, and statistical analyses; wrote the first draft of the manuscript; and had access to the data and the main responsibility for the content of the manuscript and data analyses; HMA: developed the TMAO and choline assays, was responsible for the TMAO and choline blood measurements and sample management, and gave input on the final manuscript; YR: developed the phospholipid assay and was responsible for the measurements of phospholipids in plasma; SG: provided guidance and recommendations concerning the statistical analyses; WH: participated in planning part of the original studies and gave input on the final manuscript; and JG: supervised the studies and gave input on the Discussion. None of the authors reported a conflict of interest related to the study.

REFERENCES 1. Bennett BJ, de Aguiar Vallim TQ, Wang Z, Shih DM, Meng Y, Gregory J, Allayee H, Lee R, Graham M, Crooke R. et al. Trimethylamine-Noxide, a metabolite associated with atherosclerosis, exhibits complex genetic and dietary regulation. Cell Metab 2013;17:49–60. 2. Tang WH, Wang Z, Kennedy DJ, Wu Y, Buffa JA, Agatisa-Boyle B, Li XS, Levison BS, Hazen SL. Gut microbiota-dependent trimethylamine N-oxide (TMAO) pathway contributes to both development of renal insufficiency and mortality risk in chronic kidney disease. Circ Res 2015;116:448–55. 3. Wang Z, Klipfell E, Bennett BJ, Koeth R, Levison BS, Dugar B, Feldstein AE, Britt EB, Fu X, Chung YM, et al. Gut flora metabolism of phosphatidylcholine promotes cardiovascular disease. Nature 2011; 472:57–63. 4. Wang Z, Tang WH, Buffa JA, Fu X, Britt EB, Koeth RA, Levison BS, Fan Y, Wu Y, Hazen SL. Prognostic value of choline and betaine depends on intestinal microbiota-generated metabolite trimethylamine-Noxide. Eur Heart J 2014;35:904–10. 5. Tang WH, Wang Z, Shrestha K, Borowski AG, Wu Y, Troughton RW, Klein AL, Hazen SL. Intestinal microbiota-dependent phosphatidylcholine metabolites, diastolic dysfunction, and adverse clinical outcomes in chronic systolic heart failure. J Card Fail 2015;21:91–6. 6. Shih DM, Wang Z, Lee R, Meng Y, Che N, Charugundla S, Qi H, Wu J, Pan C, Brown JM, et al. Flavin containing monooxygenase 3 exerts broad effects on glucose and lipid metabolism and atherosclerosis. J Lipid Res 2015;56:22–37. 7. Lupachyk S, Watcho P, Stavniichuk R, Shevalye H, Obrosova IG. Endoplasmic reticulum stress plays a key role in the pathogenesis of diabetic peripheral neuropathy. Diabetes 2013;62:944–52. 8. Seibel BA, Walsh PJ. Trimethylamine oxide accumulation in marine animals: relationship to acylglycerol storage. J Exp Biol 2002;205: 297–306. 9. Boutagy NE, Neilson AP, Osterberg KL, Smithson AT, Englund TR, Davy BM, Hulver MW, Davy KP. Short-term high-fat diet increases postprandial trimethylamine-N-oxide in humans. Nutr Res 2015;35: 858–64. 10. Warrier M, Shih DM, Burrows AC, Ferguson D, Gromovsky AD, Brown AL, Marshall S, McDaniel A, Schugar RC, Wang Z, et al. The TMAO-generating enzyme flavin monooxygenase 3 is a central regulator of cholesterol balance. Cell Reports 2015;10:326–38. 11. Koeth RA, Wang Z, Levison BS, Buffa JA, Org E, Sheehy BT, Britt EB, Fu X, Wu Y, Li L, et al. Intestinal microbiota metabolism of L-carnitine, a nutrient in red meat, promotes atherosclerosis. Nat Med 2013;19:576–85. 12. Obeid R, Jung J, Falk J, Herrmann W, Geisel J, Friesenhahn-Ochs B, Lammert F, Fassbender K, Kostopoulos P. Serum vitamin B12 not reflecting vitamin B12 status in patients with type 2 diabetes. Biochimie 2013;95:1056–61. 13. Herrmann W, Kirsch SH, Kruse V, Eckert R, Graber S, Geisel J, Obeid R. One year B and D vitamins supplementation improves metabolic bone markers. Clin Chem Lab Med 2013;51:639–47. 14. Kirsch SH, Herrmann W, Kruse V, Eckert R, Graber S, Geisel J, Obeid R. One year B-vitamins increases serum and whole blood folate forms and lowers plasma homocysteine in older Germans. Clin Chem Lab Med 2015;53:445–452. 15. Wang Z, Levison BS, Hazen JE, Donahue L, Li XM, Hazen SL. Measurement of trimethylamine-N-oxide by stable isotope dilution liquid chromatography tandem mass spectrometry. Anal Biochem 2014;455:35–40.

16. Zhao X, Zeisel SH, Zhang S. Rapid LC-MRM-MS assay for simultaneous quantification of choline, betaine, trimethylamine, trimethylamine N-oxide, and creatinine in human plasma and urine. Electrophoresis 2015 Jun 17 (Epub ahead of print; DOI: 10.1002/ elps.201500055). 17. Rabagny Y, Herrmann W, Geisel J, Kirsch SH, Obeid R. Quantification of plasma phospholipids by ultra performance liquid chromatography tandem mass spectrometry. Anal Bioanal Chem 2011; 401:891–9. 18. Kirsch SH, Herrmann W, Rabagny Y, Obeid R. Quantification of acetylcholine, choline, betaine, and dimethylglycine in human plasma and urine using stable-isotope dilution ultra performance liquid chromatography-tandem mass spectrometry. J Chromatogr B Analyt Technol Biomed Life Sci 2010;878:3338–44. 19. Kirsch SH, Knapp JP, Geisel J, Herrmann W, Obeid R. Simultaneous quantification of S-adenosyl methionine and S-adenosyl homocysteine in human plasma by stable-isotope dilution ultra performance liquid chromatography tandem mass spectrometry. J Chromatogr B Analyt Technol Biomed Life Sci 2009;877:3865–70. 20. Zeisel SH, da Costa KA. Choline: an essential nutrient for public health. Nutr Rev 2009;67:615–23. 21. Leermakers ET, Moreira EM, Kiefte-de Jong JC, Darweesh SK, Visser T, Voortman T, Bautista PK, Chowdhury R, Gorman D, Bramer WM, et al. Effects of choline on health across the life course: a systematic review. Nutr Rev 2015;73:500–22. 22. Tang WH, Wang Z, Levison BS, Koeth RA, Britt EB, Fu X, Wu Y, Hazen SL. Intestinal microbial metabolism of phosphatidylcholine and cardiovascular risk. N Engl J Med 2013;368:1575–84. 23. Miller CA, Corbin KD, da Costa KA, Zhang S, Zhao X, Galanko JA, Blevins T, Bennett BJ, O’Connor A, Zeisel SH. Effect of egg ingestion on trimethylamine-N-oxide production in humans: a randomized, controlled, dose-response study. Am J Clin Nutr 2014;100: 778–86. 24. Zeisel SH, daCosta KA, Youssef M, Hensey S. Conversion of dietary choline to trimethylamine and dimethylamine in rats: dose-response relationship. J Nutr 1989;119:800–4. 25. Romano KA, Vivas EI, Amador-Noguez D, Rey FE. Intestinal microbiota composition modulates choline bioavailability from diet and accumulation of the proatherogenic metabolite trimethylamine-N-oxide. MBio 2015;6:e02481. 26. Ejtahed HS, Mohtadi-Nia J, Homayouni-Rad A, Niafar M, AsghariJafarabadi M, Mofid V, Akbarian-Moghari A. Effect of probiotic yogurt containing Lactobacillus acidophilus and Bifidobacterium lactis on lipid profile in individuals with type 2 diabetes mellitus. J Dairy Sci 2011;94:3288–94. 27. Jung SP, Lee KM, Kang JH, Yun SI, Park HO, Moon Y, Kim JY. Effect of Lactobacillus gasseri BNR17 on overweight and obese adults: a randomized, double-blind clinical trial. Korean J Fam Med 2013;34: 80–9. 28. Tripolt NJ, Leber B, Triebl A, Kofeler H, Stadlbauer V, Sourij H. Effect of Lactobacillus casei Shirota supplementation on trimethylamine-N-oxide levels in patients with metabolic syndrome: an open-label, randomized study. Atherosclerosis 2015;242:141–4. 29. Holm PI, Ueland PM, Vollset SE, Midttun O, Blom HJ, Keijzer MB, den Heijer M. Betaine and folate status as cooperative determinants of plasma homocysteine in humans. Arterioscler Thromb Vasc Biol 2005; 25:379–85. 30. Awwad HM, Kirsch SH, Geisel J, Obeid R. Measurement of concentrations of whole blood levels of choline, betaine, and dimethylglycine and their relations to plasma levels. J Chromatogr B Analyt Technol Biomed Life Sci 2014;957:41–5. 31. Poirier LA, Wise CK, Delongchamp RR, Sinha R. Blood determinations of S-adenosylmethionine, S-adenosylhomocysteine, and homocysteine: correlations with diet. Cancer Epidemiol Biomarkers Prev 2001;10:649–55. 32. Resseguie M, Song J, Niculescu MD, da Costa KA, Randall TA, Zeisel SH. Phosphatidylethanolamine N-methyltransferase (PEMT) gene expression is induced by estrogen in human and mouse primary hepatocytes. FASEB J 2007;21:2622–32. 33. Iwahana S, Iida H, Yashima E, Pescitelli G, Di BL, Petrovic AG, Berova N. Absolute stereochemistry of a 4 a-hydroxyriboflavin analogue of the key intermediate of the FAD-monooxygenase cycle. Chemistry 2014;20:4386–95.

DETERMINANTS OF PLASMA TRIMETHYLAMINE N-OXIDE 34. Ripp SL, Itagaki K, Philpot RM, Elfarra AA. Methionine S-oxidation in human and rabbit liver microsomes: evidence for a high-affinity methionine S-oxidase activity that is distinct from flavin-containing monooxygenase 3. Arch Biochem Biophys 1999;367:322–32. 35. Grundy SM, Metzger AL. A physiological method for estimation of hepatic secretion of biliary lipids in man. Gastroenterology 1972;62:1200–17. 36. Cheng SH, Stanley MM. Secretion of cholesterol by intestinal mucosa in patients with complete common bile duct obstruction. Proc Soc Exp Biol Med 1959;101:223–5. 37. Evans WH, Kremmmer T, Culvenor JG. Role of membranes in bile formation. Comparison of the composition of bile and a liver bile-canalicular plasma-membrane subfraction. Biochem J 1976;154:589–95. 38. van Erpecum KJ, Carey MC. Influence of bile salts on molecular interactions between sphingomyelin and cholesterol: relevance to bile formation and stability. Biochim Biophys Acta 1997;1345:269–82.

9 of 9

39. van der Velde AE, Vrins CL, van den Oever K, Kunne C, Oude Elferink RP, Kuipers F, Groen AK. Direct intestinal cholesterol secretion contributes significantly to total fecal neutral sterol excretion in mice. Gastroenterology 2007;133:967–75. 40. Ancelin ML, Vial HJ. Quaternary ammonium compounds efficiently inhibit Plasmodium falciparum growth in vitro by impairment of choline transport. Antimicrob Agents Chemother 1986;29:814–20. 41. Tischer M, Pradel G, Ohlsen K, Holzgrabe U. Quaternary ammonium salts and their antimicrobial potential: targets or nonspecific interactions? ChemMedChem 2012;7:22–31. 42. Ancelin ML, Vial HJ, Philippot JR. Inhibitors of choline transport into plasmodium-infected erythrocytes are effective antiplasmodial compounds in vitro. Biochem Pharmacol 1985;34:4068–71. 43. Hooper LV, Littman DR, Macpherson AJ. Interactions between the microbiota and the immune system. Science 2012;336:1268–73.