ABSTRACT
Background
The role of trimethylamine-N-oxide (TMAO) in type 2 diabetes (T2D) is currently partially understood and controversial.
Objective
The aim of this study was to investigate associations between TMAO and related metabolites with T2D risk in subjects at high risk of cardiovascular disease.
Design
This is a case-cohort design study within the Prevención con Dieta Mediterránea (PREDIMED) study, with 251 incident T2D cases and a random sample of 694 participants (641 noncases and 53 overlapping cases) without T2D at baseline (median follow-up: 3.8 y). We used liquid chromatography–tandem mass spectrometry to measure plasma TMAO, l-carnitine, betaine, lyso-phosphatidylcholine (LPC) and lyso-phosphatidylethanolamine (LPE) species, phosphocholine, α-glycerophosphocholine, and choline at baseline and after 1 y. We examined associations with the use of weighted Cox proportional hazard models, accounting for the weighted case-cohort design by the Barlow method.
Results
After adjustment for recognized T2D risk factors and multiple testing, individuals in the highest quartile of baseline TMAO and α-glycerophosphocholine had a lower risk of T2D [HR (95% CI): 0.52 (0.29, 0.89) and 0.46 (0.24, 0.89), respectively]. The HR (95% CI) comparing the extreme quartiles of betaine was 0.41 (0.23, 0.74). Similar trends were observed for C16:0 LPC, C18:1 LPC, C18:0 LPC, C20:4 LPC, C22:6 LPC, C18:1 LPC plasmalogen, and C16:0 LPE. After correcting for multiple comparisons, participants in the highest quartile of 1-y changes in oleic acid LPC plasmalogen concentrations had a lower T2D risk than the reference quartile.
Conclusion
Whether the associations between plasma TMAO and certain metabolite concentrations with T2D risk reflect its pathophysiology or represent an epiphenomenon needs to be elucidated. This trial is registered at http://www.controlled-trials.com as ISRCTN35739639.
Keywords: trimethylamine-N-oxide, metabolites, type 2 diabetes, case-cohort, Mediterranean diet, PREDIMED
INTRODUCTION
Plasma concentrations of trimethylamine-N-oxide (TMAO) are determined by diet, the gut microbiome, and liver flavin-containing monooxygenase 3 (FMO3) activities (1). They are directly associated with the consumption of animal-derived foods (2, 3) containing choline, phosphatidylcholine, and l-carnitine, which are processed by gut bacteria resulting in the release of various metabolites including trimethylamine (TMA), into the blood. TMA is then transported to the liver where it is converted to TMAO, which is involved in various physiologic and pathophysiologic processes such as the deposition and removal of cholesterol from the artery endothelium (4, 5).
TMAO has been proposed as a molecule mediating the development of type 2 diabetes (T2D) (6). Higher plasma TMAO concentrations and alterations in interrelated pathways, such as phospholipid modification and methylation, have been associated with T2D (7). In this context, plasma choline concentrations have been found to be positively related, whereas betaine concentrations are inversely related to glucose concentrations (8) and lowered in insulin-resistant subjects (9). Betaine can be derived from choline or from l-carnitine metabolism (10), with l-carnitine associated with better insulin sensitivity in diabetics and with insulin-mediated glucose uptake in normoglycemic subjects (11).
Substantial inverse associations between dietary patterns consisting of healthy foods and the risk of T2D have been reported (12, 13). Recently, in a secondary outcome analysis of the Prevention of Disease with Mediterranean Diet [Prevención con Dieta Mediterránea (PREDIMED)] trial, a Mediterranean diet (MedDiet) reduced the risk of T2D by 30% compared with the control group (14). Although the benefits of this dietary pattern for T2D prevention is clearly observed in the PREDIMED trial, the biological mechanisms underlying these benefits are not completely understood. Consequently, an approach through metabolite-profiling technology (metabolomics) (15) can strengthen existing pathophysiologic evidence, providing further support for dietary prevention of T2D.
To our knowledge, no prospective study has assessed the association between TMAO with T2D risk. Thus, the role of TMAO in T2D development is not completely understood. Taking this into account, the primary aim of the present prospective study, nested in the framework of the PREDIMED trial, was to examine possible associations between baseline and 1-y changes in the concentrations of TMAO and several metabolites involved in relevant pathways [l-carnitine, betaine, lyso-phosphatidylcholine (LPC) and lyso-phosphatidylethanolamine (LPE) species, phosphocholine, α-glycerophosphocholine, and choline], with the risk of incident T2D. In addition, we aimed to examine whether a MedDiet modified these associations.
METHODS
Study design and participants
This study used a case-cohort design nested within the PREDIMED trial (ISRCTN35739639), a multicenter, single-blind, controlled trial, conducted in Spanish primary health care centers. The design of the PREDIMED trial has been described in detail elsewhere (16, 17). In brief, 7447 participants at high cardiovascular disease (CVD) risk were allocated to a MedDiet supplemented with extra-virgin olive oil, a MedDiet supplemented with mixed nuts, or a control diet consisting of advice to reduce fat intake. For the present study, we considered the 3541 participants who were free of T2D at study inception. The present case-cohort study comprises a random selection of 694 nondiabetic participants (20%) from the eligible subjects of the PREDIMED cohort without T2D at study inception and with available blood samples, together with all incident cases of T2D that occurred during the follow-up with available plasma samples (251 out of the 273 cases). Of the 892 participants included in our analyses, 641 were in the subcohort (including 53 overlapping cases between the subcohort and the total cases) and 198 comprised the rest of the T2D cases, which gave a total of 251 cases (Supplemental Figure 1). Of these, 686 out of the 892 participants had available samples after 1 y of follow-up and were included in the 1-y change analyses (Supplemental Figure 1). The institutional review boards of the recruitment centers approved the study protocol, and participants provided written informed consent.
Study samples and metabolomics profiling
Fasting (for ≥8 h) plasma EDTA samples were collected from subjects and stored at − 80°C. In June 2015, pairs of samples for each participant (baseline and at the end of the 1-y follow-up) were randomly ordered and shipped on dry ice to the Broad Institute, Inc., Boston, Massachusetts, for metabolomics assays. Liquid chromatography–tandem mass spectrometry techniques were used to perform semiquantitative profiling of several metabolites in blood plasma (TMAO, l-carnitine, betaine, LPC and LPE species, phosphocholine, α-glycerophosphocholine, and choline) and identify relations between them. A system composed of a Shimadzu Nexera X2 U-HPLC (Shimadzu Corp.) coupled to a Q Exactive hybrid quadrupole orbitrap mass spectrometer (Thermo Fisher Scientific) was used (18–24). Metabolite identities were confirmed by using authentic reference standards. Raw data were processed with the use of TraceFinder software (Thermo Fisher Scientific) and Progenesis QI (Nonlinear Dynamics).
Ascertainment of T2D cases
Information was collected through contact with participants and primary health care physicians, annual follow-up visits, yearly ad hoc reviews of medical charts, and annual consultation of the National Death Index. Because T2D was a prespecified secondary outcome of the PREDIMED trial, it was identified at baseline by clinical diagnosis or use of antidiabetic medication. The diagnosis of new-onset T2D during follow-up has been described elsewhere (14) and briefly followed the American Diabetes Association criteria (25), namely 2 confirmations of fasting plasma glucose ≥7.0 mmol/L or 2-h plasma glucose ≥11.1 mmol/L after a 75-g oral-glucose load.
Assessment of covariates and other variables
At baseline and yearly during the follow-up, the participants completed a 47-item questionnaire related to lifestyle variables, smoking status, medical history, and medication use. A validated Spanish version of the Minnesota Leisure-Time Physical Activity Questionnaire was administered in order to evaluate physical activity (26). To assess the degree of adherence to the MedDiet, a 14-item validated questionnaire was filled in for each participant (27). BMI was calculated as weight divided by height squared (kg/m2). Participants’ triacylglycerol and total, HDL, and LDL cholesterol concentrations were measured by using fasting plasma at baseline. Blood glucose and insulin concentrations were centrally assessed at baseline and at the end of the 1 y of follow-up. Insulin resistance was estimated by the HOMA-IR method with the use of the following equation (28): HOMA-IR = [fasting insulin (μIU/mL) × fasting glucose (mmol/L)]/22.5.
Statistical analysis
Baseline characteristics of cases and noncases are described as means ± SDs for quantitative variables and percentages or numbers for categorical variables. We applied a natural logarithmic transformation to approximate a normal distribution of metabolite levels. Person-time of follow-up was calculated as the interval between the baseline visit and date of T2D event, death, or date of the last contact, whichever came first. We used Cox proportional hazard models, with Barlow weights (to account for the overrepresentation of cases), to estimate HRs and their 95% CIs for the risk of T2D. A crude model and 2 multivariable-adjusted Cox regression models were fitted as follows: 1) multivariable model 1 adjusted for age (years), sex (male or female), BMI (kg/m2), intervention group, and baseline fasting glucose (milligrams per deciliter) (adding a quadratic term to account for the departure from linearity) and 2) multivariable model 2 additionally adjusted for smoking (never, current, or former), leisure-time physical activity (metabolic equivalent tasks in minutesper day), baseline dyslipidemia (yes or no), and hypertension (yes or no). We stratified the models according to recruitment center. Baseline metabolites were analyzed as both continuous variables (1-SD increment in their transformed levels) and by using quartiles (using cutoffs defined among noncases). To appraise the linear trend across quartiles, the median metabolite concentration within each quartile was included in the Cox regression models as a continuous variable. To account for multiple testing, we adjusted P values of the multivariable-adjusted associations between quartiles or 1-SD increment in metabolite level and T2D risk with the use of the Benjamini-Hochberg false discovery rate procedure (29). A false discovery rate P value < 0.05 was considered to be significant.
We also examined the associations of 1-y changes in metabolites with T2D risk. We used the same models as in the baseline value analyses but further adjusted for baseline metabolite levels. With respect to metabolites, we first calculated the ratio between 1-y and baseline values and then normalized this ratio with the natural logarithmic transformation. To test the robustness of our results in relation to the association between TMAO and T2D risk, we conducted 2 sensitivity analyses: 1) testing the association between baseline values and T2D risk after excluding early cases (<1 y) and 2) testing the associations between the mean values at baseline and 1-y follow-up and subsequent T2D risk (T2D cases that occurred from baseline through the 1-y follow-up were excluded). To examine whether the association between baseline and 1-y changes in metabolites and incident T2D varied by intervention group, we stratified the analysis described above by intervention group (both merged MedDiet interventions compared with the control). We also added a multiplicative term (1 df) between intervention assignment (merged MedDiet compared with the control group) and metabolites (continuous) into the multivariable Cox models stratified on intervention assignment to test for interactions by means of likelihood ratio tests. In addition, we compared differences in 1-y changes in metabolites in the MedDiet group with changes in the control group (adjusted for the aforementioned covariates) with the use of ANCOVA. Finally, we applied multiple linear regression analyses to examine relations between TMAO concentrations and precursors (choline, betaine, and l-carnitine) at baseline and at the 1-y follow-up in the whole group, adjusting for age, sex, and HOMA-IR. Statistical analyses were performed with the use of Stata 13.1 (StataCorp.). A 2-sided P value < 0.05 was considered significant.
RESULTS
Baseline characteristics
The median follow-up of the study population was 3.8 y. The baseline characteristics of the 892 subjects (251 cases and 641 noncases) included in the present case-cohort study are shown in Table 1. The mean age of participants at baseline was 66.5 y, and the mean ± SD BMI was 30.1 ± 3.5. Compared with noncases, those participants who developed T2D were more likely to be men, current smokers, and to have a higher prevalence of hypertension in addition to a higher BMI and fasting glucose and triacylglycerol concentrations (Table 1).
TABLE 1.
Baseline characteristics of the study population1
| Total | Cases | Noncases | P | |
|---|---|---|---|---|
| n | 892 | 251 | 641 | |
| Age, y | 66.5 ± 5.7 | 66.4 ± 5.7 | 66.5 ± 5.7 | 0.781 |
| Sex, % female | 61.2 | 55.0 | 63.6 | 0.017 |
| BMI, kg/m2 | 30.1 ± 3.5 | 30.8 ± 3.3 | 29.8 ± 3.6 | <0.001 |
| Physical activity, METs/d | 240.7 ± 234.6 | 249.2 ± 233.5 | 237.4 ± 235.1 | 0.500 |
| Intervention group, % | ||||
| MedDiet + EVOO | 30.6 | 29.9 | 30.9 | 0.425 |
| MedDiet + nuts | 36.3 | 37.2 | 33.8 | |
| Control group | 33.1 | 36.3 | 31.8 | |
| Hypertension, % | 91.7 | 96.0 | 90.0 | 0.003 |
| Dyslipidemia, % | 84.3 | 79.7 | 86.1 | 0.018 |
| Smoking, % | ||||
| Never | 59.0 | 52.6 | 61.5 | 0.006 |
| Former | 22.4 | 22.3 | 22.5 | |
| Current | 18.6 | 25.1 | 16.0 | |
| Score for adherence to MedDiet2 | 8.5 ± 2.0 | 8.4 ± 2.0 | 8.6 ± 1.9 | 0.186 |
| Fasting blood glucose, mg/dL | 103.3 ± 17.6 | 118.6 ± 18.0 | 97.8 ± 13.8 | <0.001 |
| Total cholesterol, mg/dL | 222.2 ± 39.3 | 221.7 ± 42.3 | 222.4 ± 38.1 | 0.846 |
| HDL cholesterol, mg/dL | 54.7 ± 13.0 | 52.6 ± 12.7 | 55.7 ± 13.1 | 0.014 |
| LDL cholesterol, mg/dL | 139.7 ± 32.9 | 137.0 ± 31.6 | 141.0 ± 33.3 | 0.201 |
| Triacylglycerol, mg/dL | 140.1 ± 85.0 | 169.0 ± 121.0 | 128.6 ± 62.0 | <0.001 |
1Values are means ± SDs unless otherwise indicated. Chi-square test was used for comparison of categorical variables, and Student's t test was used for comparison of continuous variables. EVOO, extra-virgin olive oil; MedDiet, Mediterranean diet; MET, metabolic equivalent.
2This score is based on the 14-item dietary screener.
Baseline metabolites and risk of T2D
Overall group
The associations between plasma metabolites with the risk of T2D in the overall group are presented in Table 2. In the group adjusted for age, sex, BMI, intervention group, glucose, smoking, leisure-time physical activity, dyslipidemia, and hypertension (multivariable model 2), the estimated HR for incident T2D reached significance only in the highest, compared with the lowest, quartile of TMAO concentrations (HR: 0.52; 95% CI: 0.29, 0.89). In both sensitivity analyses, TMAO concentrations in the highest quartile were significantly associated with lower T2D risk (1: HR: 0.52; 95% CI: 0.29, 0.93; 2: HR: 0.44; 95% CI: 0.20, 0.96) compared with the lowest quartile. In the group adjusted for age, sex, BMI, intervention group, and glucose (multivariable model 1), the highest quartile of phosphocholine was associated with a lower risk of T2D (HR: 0.51; 95% CI: 0.27, 0.99), but these associations were no longer significant after further adjustment. With regard to betaine, the estimated HR for incident T2D in the highest compared with the lowest quartile was 0.41 (95% CI: 0.23, 0.74; P = 0.003). Concerning lyso-choline species, significant inverse associations with T2D incidence were observed for palmitic acid LPC, oleic acid LPC, stearic acid LPC, arachidonic acid LPC, docosahexaenoic acid LPC, and oleic acid LPC plasmalogen, either were modeled continuously (per 1 SD) or as quartiles (Table 2). Notably, per 1-SD increase in oleic acid LPC plasmalogen concentrations, a 54% lower risk (HR: 0.46; 95% CI: 0.35, 0.59) of T2D was observed (P < 0.001), and individuals in the highest quartile had an HR of 0.15 (95% CI: 0.06, 0.35; P < 0.001).
TABLE 2.
Associations of baseline individual metabolite levels with the risk of type 2 diabetes in the PREDIMED trial, 2003–2010, in the overall group1
| P | |||||||||
|---|---|---|---|---|---|---|---|---|---|
| Quartile of plasma metabolite level | FDR-adjusted P | HR per 1-SD | FDR- | ||||||
| Metabolite | 1 | 2 | 3 | 4 | P-trend | value (Q4 vs. Q1) | increment | Unadjusted | adjusted |
| Trimethylamine | |||||||||
| N-oxide | 0.032 | 0.078 | |||||||
| Cases, n | 73 | 65 | 68 | 44 | |||||
| Crude model | Ref | 0.98 (0.64, 1.53) | 1.01 (0.66, 1.56) | 0.60 (0.37, 0.98) | 0.039 | 0.88 (0.75, 1.04) | 0.162 | ||
| MV1 | Ref | 0.96 (0.60, 1.54) | 0.95 (0.58, 1.56) | 0.51 (0.29, 0.88) | 0.012 | 0.86 (0.72, 1.04) | 0.123 | ||
| MV2 | Ref | 0.98 (0.60, 1.59) | 0.91 (0.55, 1.51) | 0.52 (0.29, 0.89) | 0.012 | 0.83 (0.69, 1.01) | 0.059 | ||
| Phosphocholine | 0.124 | 0.329 | |||||||
| Cases, n | 75 | 66 | 57 | 49 | |||||
| Crude model | Ref | 0.99 (0.64, 1.52) | 0.77 (0.49, 1.21) | 0.56 (0.32, 0.98) | 0.029 | 0.90 (0.77, 1.06) | 0.224 | ||
| MV1 | Ref | 1.11 (0.68, 1.80) | 0.82 (0.48, 1.39) | 0.51 (0.27, 0.99) | 0.036 | 0.88 (0.74, 1.05) | 0.157 | ||
| MV2 | Ref | 1.19 (0.71, 1.97) | 0.89 (0.51, 1.56) | 0.56 (0.29, 1.09) | 0.079 | 0.90 (0.75, 1.08) | 0.268 | ||
| Choline | 0.493 | 0.557 | |||||||
| Cases, n | 52 | 67 | 63 | 65 | |||||
| Crude model | Ref | 1.20 (0.75, 1.90) | 1.04 (0.63, 1.72) | 1.34 (0.83, 2.18) | 0.313 | 1.03 (0.86, 1.23) | 0.729 | ||
| MV1 | Ref | 1.33 (0.80, 2.24) | 1.07 (0.61, 1.88) | 1.20 (0.70, 2.07) | 0.698 | 0.94 (0.77, 1.14) | 0.538 | ||
| MV2 | Ref | 1.42 (0.84, 2.42) | 1.13 (0.63, 2.02) | 1.23 (0.71, 2.13) | 0.663 | 0.94 (0.77, 1.14) | 0.523 | ||
| C14:0 LPC | 0.650 | 0.879 | |||||||
| Cases, n | 72 | 65 | 55 | 56 | |||||
| Crude model | Ref | 1.05 (0.68, 1.64) | 0.91 (0.57, 1.45) | 0.91 (0.57, 1.44) | 0.590 | 1.01 (0.85, 1.19) | 0.910 | ||
| MV1 | Ref | 1.02 (0.61, 1.74) | 0.89 (0.53, 1.49) | 0.93 (0.55, 1.55) | 0.669 | 1.03 (0.86, 1.24) | 0.741 | ||
| MV2 | Ref | 1.05 (0.62, 1.78) | 0.88 (0.52, 1.49) | 0.88 (0.52, 1.50) | 0.548 | 1.01 (0.84, 1.23) | 0.879 | ||
| C16:1 LPC | 0.318 | 0.526 | |||||||
| Cases, n | 78 | 61 | 58 | 50 | |||||
| Crude model | Ref | 0.89 (0.57, 1.38) | 0.84 (0.54, 1.31) | 0.72 (0.45, 1.16) | 0.181 | 0.91 (0.77, 1.08) | 0.317 | ||
| MV1 | Ref | 0.77 (0.47, 1.26) | 0.93 (0.57, 1.52) | 0.75 (0.45, 1.26) | 0.387 | 0.91 (0.76, 1.11) | 0.367 | ||
| MV2 | Ref | 0.77 (0.47, 1.28) | 1.03 (0.63, 1.68) | 0.75 (0.44, 1.27) | 0.451 | 0.93 (0.77, 1.12) | 0.461 | ||
| C16:0 LPC | 0.012 | 0.003 | |||||||
| Cases, n | 102 | 62 | 51 | 36 | |||||
| Crude model | Ref | 0.60 (0.39, 0.92) | 0.58 (0.37, 0.91) | 0.41 (0.24, 0.69) | 0.001 | 0.69 (0.56, 0.83) | <0.001 | ||
| MV1 | Ref | 0.53 (0.33, 0.85) | 0.58 (0.36, 0.93) | 0.41 (0.23, 0.74) | 0.002 | 0.69 (0.56, 0.84) | <0.001 | ||
| MV2 | Ref | 0.50 (0.31, 0.82) | 0.56 (0.35, 0.92) | 0.42 (0.24, 0.75) | 0.002 | 0.69 (0.56, 0.85) | <0.001 | ||
| C18:1 LPC | 0.008 | 0.003 | |||||||
| Cases, n | 111 | 59 | 48 | 30 | |||||
| Crude model | Ref | 0.52 (0.34, 0.79) | 0.44 (0.28, 0.68) | 0.31 (0.18, 0.51) | <0.001 | 0.66 (0.56, 0.77) | <0.001 | ||
| MV1 | Ref | 0.51 (0.32, 0.82) | 0.50 (0.32, 0.80) | 0.34 (0.19, 0.60) | <0.001 | 0.67 (0.56, 0.80) | <0.001 | ||
| MV2 | Ref | 0.52 (0.32, 0.85) | 0.54 (0.33, 0.86) | 0.36 (0.20, 0.65) | <0.001 | 0.69 (0.58, 0.82) | <0.001 | ||
| C18:0 LPC | 0.013 | 0.003 | |||||||
| Cases, n | 108 | 64 | 48 | 30 | |||||
| Crude model | Ref | 0.55 (0.36, 0.84) | 0.42 (0.27, 0.66) | 0.37 (0.21, 0.64) | <0.001 | 0.59 (0.48, 0.74) | <0.001 | ||
| MV1 | Ref | 0.59 (0.37, 0.96) | 0.48 (0.29, 0.79) | 0.34 (0.18, 0.68) | <0.001 | 0.63 (0.50, 0.79) | <0.001 | ||
| MV2 | Ref | 0.58 (0.36, 0.95) | 0.47 (0.28, 0.79) | 0.37 (0.19, 0.74) | 0.001 | 0.65 (0.51, 0.82) | <0.001 | ||
| C20:4 LPC | 0.013 | 0.003 | |||||||
| Cases, n | 103 | 56 | 47 | 43 | |||||
| Crude model | Ref | 0.52 (0.34, 0.80) | 0.48 (0.31, 0.76) | 0.45 (0.28, 0.73) | <0.001 | 0.67 (0.56, 0.80) | <0.001 | ||
| MV1 | Ref | 0.49 (0.30, 0.78) | 0.54 (0.33, 0.87) | 0.43 (0.25, 0.75) | 0.001 | 0.66 (0.54, 0.80) | <0.001 | ||
| MV2 | Ref | 0.48 (0.29, 0.79) | 0.57 (0.35, 0.93) | 0.44 (0.25, 0.78) | 0.003 | 0.67 (0.54, 0.81) | <0.001 | ||
| C22:6 LPC | 0.014 | 0.028 | |||||||
| Cases, n | 93 | 72 | 52 | 33 | |||||
| Crude model | Ref | 1.03 (0.68, 1.55) | 0.71 (0.47, 1.09) | 0.39 (0.23, 0.66) | <0.001 | 0.76 (0.65, 0.89) | 0.001 | ||
| MV1 | Ref | 1.12 (0.70, 1.79) | 0.98 (0.63, 1.54) | 0.40 (0.22, 0.74) | 0.009 | 0.78 (0.66, 0.94) | 0.007 | ||
| MV2 | Ref | 1.13 (0.71, 1.82) | 1.07 (0.67, 1.70) | 0.41 (0.22, 0.77) | 0.022 | 0.80 (0.67, 0.96) | 0.016 | ||
| C18:1 LPC | |||||||||
| plasmalogen | 0.008 | 0.003 | |||||||
| Cases, n | 120 | 61 | 48 | 14 | |||||
| Crude model | Ref | 0.56 (0.37, 0.86) | 0.38 (0.24, 0.59) | 0.18 (0.09, 0.37) | <0.001 | 0.42 (0.33, 0.55) | <0.001 | ||
| MV1 | Ref | 0.68 (0.43, 1.09) | 0.44 (0.27, 0.72) | 0.14 (0.06, 0.32) | <0.001 | 0.45 (0.34, 0.58) | <0.001 | ||
| MV2 | Ref | 0.67 (0.42, 1.08) | 0.45 (0.27, 0.74) | 0.15 (0.06, 0.35) | <0.001 | 0.46 (0.35, 0.59) | <0.001 | ||
| (Continued) | |||||||||
| C16:0 LPE | 0.026 | 0.004 | |||||||
| Cases, n | 94 | 73 | 50 | 32 | |||||
| Crude model | Ref | 0.97 (0.64, 1.45) | 0.59 (0.38, 0.92) | 0.45 (0.26, 0.78) | 0.001 | 0.73 (0.61, 0.87) | 0.001 | ||
| MV1 | Ref | 0.86 (0.54, 1.36) | 0.56 (0.35, 0.89) | 0.49 (0.27, 0.88) | 0.003 | 0.74 (0.61, 0.89) | 0.002 | ||
| MV2 | Ref | 0.89 (0.55, 1.44) | 0.55 (0.34, 0.88) | 0.46 (0.25, 0.84) | 0.002 | 0.73 (0.60, 0.89) | 0.002 | ||
| C18:1 LPE | 0.075 | 0.078 | |||||||
| Cases, n | 73 | 66 | 61 | 49 | |||||
| Crude model | Ref | 0.89 (0.57, 1.38) | 0.84 (0.54, 1.30) | 0.60 (0.37, 0.95) | 0.034 | 0.81 (0.69, 0.95) | 0.011 | ||
| MV1 | Ref | 0.77 (0.47, 1.26) | 0.79 (0.50, 1.26) | 0.57 (0.34, 0.96) | 0.045 | 0.82 (0.70, 0.99) | 0.040 | ||
| MV2 | Ref | 0.79 (0.48, 1.32) | 0.86 (0.53, 1.39) | 0.59 (0.35, 1.00) | 0.073 | 0.84 (0.69, 1.00) | 0.058 | ||
| l-Carnitine | 0.280 | 0.040 | |||||||
| Cases, n | 52 | 79 | 69 | 49 | |||||
| Crude model | Ref | 1.81 (1.12, 2.91) | 1.44 (0.89, 2.34) | 0.91 (0.54, 1.55) | 0.379 | 0.90 (0.77, 1.05) | 0.198 | ||
| MV1 | Ref | 1.76 (1.07, 2.90) | 1.09 (0.64, 1.88) | 0.75 (0.41, 1.36) | 0.108 | 0.84 (0.70, 1.01) | 0.071 | ||
| MV2 | Ref | 1.77 (1.06, 2.94) | 1.07 (0.61, 1.88) | 0.69 (0.37, 1.26) | 0.066 | 0.81 (0.67, 0.97) | 0.025 | ||
| α-Glycerophos- | 0.035 | 0.004 | |||||||
| phocholine | |||||||||
| Cases, n | 81 | 56 | 71 | 40 | |||||
| Crude model | Ref | 0.70 (0.44, 1.12) | 0.81 (0.51, 1.28) | 0.57 (0.32, 1.02) | 0.091 | 0.70 (0.54, 0.92) | 0.011 | ||
| MV1 | Ref | 0.58 (0.35, 0.98) | 0.75 (0.46, 1.20) | 0.47 (0.25, 0.90) | 0.044 | 0.61 (0.45, 0.81) | 0.001 | ||
| MV2 | Ref | 0.54 (0.31, 0.93) | 0.75 (0.46, 1.23) | 0.46 (0.24, 0.89) | 0.051 | 0.61 (0.45, 0.83) | 0.002 | ||
| Betaine | 0.012 | 0.008 | |||||||
| Cases, n | 76 | 69 | 59 | 43 | |||||
| Crude model | Ref | 0.82 (0.53, 1.26) | 0.82 (0.52, 1.29) | 0.48 (0.29, 0.78) | 0.005 | 0.83 (0.69, 0.98) | 0.035 | ||
| MV1 | Ref | 0.90 (0.57, 1.43) | 0.74 (0.45, 1.21) | 0.41 (0.23, 0.73) | 0.002 | 0.77 (0.64, 0.94) | 0.009 | ||
| MV2 | Ref | 0.87 (0.54, 1.39) | 0.73 (0.43, 1.24) | 0.41 (0.23, 0.74) | 0.003 | 0.75 (0.61, 0.91) | 0.004 | ||
1Values are HRs (95% CIs) unless otherwise indicated. A natural logarithmic transformation was applied to the raw values of individual metabolites. Cox regression analysis was used. MV1 adjusted for age (years), sex (male or female), BMI (kg/m2), intervention group (MedDiet + EVOO or MedDiet + nuts), and baseline fasting glucose (milligrams per deciliter, centered on the sample mean and adding the quadratic term); MV2 additionally adjusted for smoking (never, current, or former), leisure-time physical activity (metabolic equivalent tasks in minutes per day), dyslipidemia, and hypertension. FDR-controlled adjustments were conducted by applying the method of Benjamini and Hochberg. EVOO, extra-virgin olive oil; FDR, false discovery rate; LPC, lyso-phosphatidylcholine; LPE, lyso-phosphatidylethanolamine; MedDiet, Mediterranean diet; MV, multivariable model; PREDIMED, Prevención con Dieta Mediterránea; Q, quartile; Ref, reference.
With regard to lyso-ethanolamine species, concentrations of palmitic acid LPE (per 1-SD increase and highest quartile) were significantly associated with a reduced risk of T2D. Per 1-SD increase in l-carnitine, α-glycerophosphocholine, and betaine concentrations acid, a 19% (HR: 0.81; 95% CI: 0.67, 0.97; P = 0.025), 39% (HR: 0.61; 95% CI: 0.45, 0.83; P = 0.002), and 25% (HR: 0.75; 95% CI: 0.61, 0.91; P = 0.004) lower risk of T2D was found, respectively. These associations remained significant after accounting for multiple comparisons.
MedDiet and control group
In stratified analysesxz by intervention group, the estimated HR for incident T2D in the highest compared with the lowest quartile of α-glycerophosphocholine concentrations was 0.27 (95% CI: 0.12, 0.62; P = 0.002) in the MedDiet group (Supplemental Table 1) and 1.08 (95%: CI 0.14, 1.16; P > 0.05) in the control group (Supplemental Table 2). In the MedDiet group, the HR associated with a 1-SD increment in α-glycerophosphocholine concentration was 0.55 (95% CI: 0.37, 0.82; P = 0.003, P-interaction = 0.041) (Figure 1A, Supplemental Table 3), whereas in the control group no significant associations were observed (Figure 1B).
FIGURE 1.
HRs (95% CIs) for type 2 diabetes according to baseline metabolites analyzed as continuous (per 1-SD increment) by both Mediterranean diet intervention groups (merged) (A) and the control group (B). All HRs were adjusted for age (years), sex (male or female), BMI (kg/m2), baseline fasting glucose (milligrams per deciliter; mean + quadratic term of the centered mean), smoking (never, current, or former), leisure-time physical activity (metabolic equivalent tasks in minutes per day), dyslipidemia, and hypertension. A natural logarithmic transformation was applied to the raw value. FDR-controlled adjustments were conducted by applying the method of Benjamini and Hochberg. Alpha-glycelop, α-glycerophosphocholine; FDR, false discovery rate; LPC, lysophosphatidylcholine; LPE, lysophosphatidylethanolamine; plas, plasmalogen; TMAO, trimethylamine N-oxide.
One-year changes in levels of metabolites and the risk of T2D
Associations between 1-y changes in metabolite levels across quartiles and the risk of T2D are shown in Table 3. In the highest quartile of increase in the concentrations of TMAO, a significant inverse association with T2D risk was found (HR: 0.49; 95% CI: 0.25, 0.98). Several lyso-choline species (palmitic acid LPC, oleic acid LPC, and stearic acid LPC), including oleic acid LPC plasmalogen, palmitic acid LPE, and α-glycerophosphocholine, were significantly associated with a decreased risk of T2D. After adjustment for multiple testing, only oleic acid LPC plasmalogen remained significant. We repeated the analyses with the use of a 1-SD increment in 1-y change in metabolite levels and, after adjustment for multiple testing, we found that per 1-SD increase in palmitic acid LPC, oleic acid LPC, stearic acid LPC, and oleic acid LPC plasmalogen concentrations, the risk of T2D decreased by 35% (HR: 0.65; 95% CI: 0.49, 0.86), 37% (HR: 0.63; 95% CI: 0.47, 0.85), 32% (HR: 0.68; 95% CI: 0.50, 0.92), and 42% (HR: 0.58; 95% CI: 0.42, 0.81), respectively. These significant inverse associated with lower risk in the control group. However, P-interaction values between palmitic acid LPC, oleic acid LPC, associations observed in the overall group persisted in the MedDiet group, whereas only C18:1 LPC was found to be significantly stearic acid LPC, and the intervention group (MedDiet compared with the control group) on T2D were nonsignificant (Supplemental Table 4). There was a tendency toward higher T2D risk for those individuals in the control group with higher 1-y changes in betaine concentrations, but the results were not significant (HR: 1.25; 95% CI: 0.81, 1.92; P-interaction = 0.012). There were no significant differences in mean 1-y changes in metabolite levels between the MedDiet and control groups (data not shown).
TABLE 3.
Associations of 1-y changes in individual metabolite levels with the risk of type 2 diabetes in the PREDIMED trial, 2003–2010, in the overall group1
| Quartile of plasma metabolite level | ||||||
|---|---|---|---|---|---|---|
| Metabolite | 1 | 2 | 3 | 4 | P-trend | FDR-adjusted P (Q4 vs. Q1) |
| Trimethylamine | ||||||
| N-oxide | 0.081 | 0.100 | ||||
| Cases, n | 44 | 44 | 45 | 48 | ||
| MV | Ref | 0.58 (0.31, 1.10) | 0.64 (0.32, 1.25) | 0.49 (0.25, 0.98) | ||
| Phosphocholine | 0.484 | 0.598 | ||||
| Cases, n | 44 | 44 | 45 | 48 | ||
| MV | Ref | 0.75 (0.42, 1.36) | 1.00 (0.59, 1.71) | 0.74 (0.40, 1.37) | ||
| Choline | 0.895 | 0.924 | ||||
| Cases, n | 49 | 44 | 50 | 37 | ||
| MV | Ref | 1.11 (0.61, 2.02) | 0.83 (0.46, 1.51) | 1.04 (0.54, 2.00) | ||
| C14:0 LPC | 0.587 | 0.769 | ||||
| Cases, n | 52 | 47 | 39 | 43 | ||
| MV | Ref | 0.77 (0.42, 1.40) | 0.79 (0.43, 1.47) | 0.79 (0.38, 1.64) | ||
| C16:1 LPC | 0.657 | 0.811 | ||||
| Cases, n | 57 | 35 | 43 | 45 | ||
| MV | Ref | 0.47 (0.24, 0.91) | 0.60 (0.33, 1.10) | 0.86 (0.44, 1.68) | ||
| C16:0 LPC | 0.014 | 0.100 | ||||
| Cases, n | 55 | 44 | 42 | 40 | ||
| MV | Ref | 0.57 (0.31, 1.04) | 0.38 (0.21, 1.69) | 0.47 (0.24, 0.93) | ||
| C16:0 LPC | 0.016 | 0.100 | ||||
| Cases, n | 44 | 53 | 36 | 44 | ||
| MV | Ref | 0.76 (0.42, 1.38) | 0.55 (0.29, 1.01) | 0.44 (0.21, 0.90) | ||
| C18:0 LPC | 0.018 | 0.100 | ||||
| Cases, n | 49 | 45 | 51 | 34 | ||
| MV | Ref | 0.63 (0.36, 1.12) | 0.47 (0.25, 0.90) | 0.45 (0.22, 0.92) | ||
| C20:4 LPC | 0.047 | 0.130 | ||||
| Cases, n | 48 | 41 | 42 | 49 | ||
| MV | Ref | 0.74 (0.41, 1.35) | 0.53 (0.27, 1.03) | 0.54 (0.28, 1.04) | ||
| C22:6 LPC | 0.334 | 0.598 | ||||
| Cases, n | 38 | 41 | 50 | 50 | ||
| MV | Ref | 0.99 (0.55, 1.79) | 0.55 (0.27, 1.10) | 0.72 (0.36, 1.47) | ||
| C18:1 LPC | ||||||
| plasmalogen | <0.001 | 0.032 | ||||
| Cases, n | 52 | 57 | 40 | 32 | ||
| MV | Ref | 0.76 (0.43, 1.33) | 0.30 (0.15, 0.58) | 0.33 (0.17, 0.66) | ||
| C16:0 LPE | 0.052 | 0.100 | ||||
| Cases, n | 44 | 51 | 45 | 34 | ||
| MV | Ref | 0.87 (0.48, 1.58) | 0.91 (0.51, 1.62) | 0.49 (0.25, 0.98) | ||
| C18:1 LPE | 0.408 | 0.811 | ||||
| Cases, n | 42 | 50 | 32 | 56 | ||
| MV | Ref | 1.19 (0.66, 2.12) | 0.69 (0.38, 1.27) | 0.85 (0.43, 1.67) | ||
| l-Carnitine | 0.725 | 0.811 | ||||
| Cases, n | 35 | 42 | 61 | 42 | ||
| MV | Ref | 1.16 (0.58, 2.32) | 2.05 (1.03, 4.11) | 0.86 (0.38, 1.91) | ||
| α-Glycerophos- | ||||||
| phocholine | 0.084 | 0.100 | ||||
| Cases, n | 49 | 47 | 51 | 33 | ||
| MV | Ref | 0.40 (0.21, 0.78) | 0.43 (0.22, 0.83) | 0.45 (0.21, 0.98) | ||
| Betaine | 0.732 | 0.924 | ||||
| Cases, n | 41 | 47 | 45 | 48 | ||
| MV | Ref | 1.41 (0.77, 2.59) | 0.90 (0.47, 1.73) | 1.03 (0.54, 1.95) | ||
1Values are HRs (95% CIs). We first calculated the ratio between 1-y and baseline levels of individual metabolites and then normalized this ratio with the natural logarithmic transformation. Cox regression analysis was used. The MV was adjusted for baseline levels of metabolites, age (years), sex (male or female), intervention group (MedDiet + EVOO or MedDiet + nuts), BMI (kg/m2), baseline fasting glucose (milligrams per deciliter, centered on the sample mean and adding the quadratic term), smoking (never, current, or former), leisure-time physical activity (metabolic equivalent tasks in minutes per day), dyslipidemia, and hypertension. FDR-controlled adjustments were conducted by applying the method of Benjamini and Hochberg. EVOO, extra-virgin olive oil; FDR, false discovery rate; LPC, lyso-phosphatidylcholine; LPE, lyso-phosphatidylethanolamine; MedDiet, Mediterranean diet; MV, multivariable model; PREDIMED, Prevención con Dieta Mediterránea; Q, quartile; Ref, reference.
Predictors of TMAO concentrations from regression analyses
A multiple linear regression was used to assess potential relations between TMAO and choline, betaine, and l-carnitine, while adjusting for age, sex, and HOMA-IR (Supplemental Tables 5 and 6). Baseline and 1-y changes in l-carnitine concentrations were positively associated with TMAO concentrations at baseline and at 1 y, respectively.
DISCUSSION
With the use of a case-cohort design within the PREDIMED trial and aimed at identifying plasma metabolites potentially related to T2D in 892 individuals at high CVD risk, we observed that higher baseline concentrations of TMAO, l-carnitine, betaine, α-glycerophosphocholine, and several LPC and LPE species were associated with lower risk of T2D development, independently of recognized T2D risk factors (i.e., age, sex, BMI, blood glucose, smoking). At baseline, the participants in the highest quartile of TMAO plasma concentrations had a 48% lower risk of developing T2D compared with the lowest quartile. Notably, the association between TMAO and T2D persisted and remained significant after sensitivity analyses.
l-Carnitine affects insulin-mediated glucose uptake and oxidation in diabetics and healthy controls, improving insulin sensitivity and blood glucose concentrations in patients with T2D (11) and mitochondrial utilization of fatty acids (30). Betaine is associated with a lower risk of diabetes, presumably because it serves as a methyl donor in the methionine cycle and its administration decreases homocysteine concentrations (31–33). Recently, plasma betaine concentrations were reported to be reduced in insulin-resistant humans and were associated with insulin sensitivity (9). In addition, lower plasma concentrations of betaine were reported in patients with metabolic syndrome compared with a healthy population (34). The possible protective role of l-carnitine and betaine in T2D development was also clearly supported by our findings.
The gut microbiome is involved in gastrointestinal and immune function as well as the digestion of nutrients, and may thus affect the risk of obesity and T2D (35). l-Carnitine, choline, and betaine can be metabolized by the microbiome to TMA, which is absorbed by the gut and further oxidized to TMAO in the liver (Figure 2). Among these substrates, l-carnitine was identified in our study as the sole positive predictor of plasma TMAO concentrations (10, 36). l-Carnitine may be transformed to TMAO through 2 independent pathways catalyzed by l-carnitine dehydrogenase or carnitine oxygenase (10). TMAO is, on the other hand, considered to be a risk factor for CVD (36). However, circulating TMAO is influenced by the gut microbiome, kidney function, as well as a FMO3 genotype, factors that may confound the relation between TMAO and chronic disease (37). Whether TMAO is a causal factor for disease development and progression or simply a biomarker of metabolic adaptation is unclear and needs to be explored further.
FIGURE 2.
Open homeostasis pattern. aThe role of the Mediterranean diet; bthe role of the microbiome; cmetabolites absorbed, produced and transformed, or methylated; dthe role of T2D pathogenicity target tissues. MEDDIET, Mediterranean diet; T2D, type 2 diabetes; TMAO, trimethylamine-N-oxide.
With regard to the inverse associations between TMAO plasma concentrations and the risk of developing T2D, our findings are compatible with a recent prospective study (38) involving a consecutive sample of 37 obese subjects undergoing bariatric surgery, 17 of whom had diabetes. One year after surgery, TMAO plasma concentrations increased by 2-fold compared with preoperative concentrations, which was significantly correlated with the corresponding 1-y decrease in glycated hemoglobin after the bariatric surgery (r = –0.39, P = 0.025). In our study, Spearman's correlation analysis between TMAO and 1-y increase in HOMA-IR, a value used to quantify insulin resistance, showed a significant negative correlation (r = –0.13, P = 0.026).
On the other hand, a study conducted in 283 individuals that examined cardiometabolic risk factors and pathways associated with TMAO concentration reported that the presence of diabetes was associated with higher concentrations of plasma TMAO (7). Similarly, Shan et al. (39) measured plasma concentrations of TMAO in a case-control study in 2694 participants and noted that higher plasma TMAO was associated with increased odds of T2D. In addition, in a larger study in 4007 patients undergoing coronary angiography, increased plasma concentrations of TMAO were related to increased concentrations of fasting glucose and diabetes (40). These studies (7, 39, 40) were not designed to investigate associations between TMAO and T2D prospectively, whereas our data showed that higher TMAO concentrations can be prospectively associated with decreased T2D risk. However, differences in study design and the populations studied can only partly explain the overall discrepancies. Because TMAO was reported to be highly variable in plasma and urine (log-normal reference change values ranging from 403% to 80% in plasma) of overweight people with T2D over a 2-y period (41), further research is needed to explain these variations. Overall, the exact mechanisms associating plasma TMAO concentrations and T2D remain unclear (10). On the other hand, the possibility that changes in TMAO just represent an epiphenomenon cannot be excluded.
TMAO and choline concentration changes were linked to risk profiles with altered concentrations of phospholipids and methylation markers, with total or specific LPC concentrations tending to be lower among diabetics (7, 42, 43). LPC may play an important role in glucose homeostasis by stimulating adipocyte glucose uptake, potentiating glucose-stimulated insulin secretion and lowering blood glucose concentrations (43, 44). In our study, Spearman's correlations between baseline metabolite levels and 1-y changes in HOMA-IR values indicated a significant negative correlation between oleic acid LPC plasmalogen and HOMA-IR (r = –0.14, P = 0.010). Choline plasmalogens with 18:1 in the sn-2 position are strongly correlated with a wide range of risk factors for metabolic syndrome (45). Nonalcoholic fatty liver disease shares common characteristics with metabolic syndrome, such as insulin resistance (46), and has been associated with decreased plasma plasmalogens (47). In our study, higher oleic acid LPC plasmalogen plasma concentrations may reflect a decrease in insulin resistance during follow-up and a potential improvement in liver function, resulting in increased secretion of plasmalogens into the circulation. Because the role of LPCs in diabetes pathophysiology is not completed understood, we cannot exclude the possibility that they do not offer any protection against the disease.
Another choline compound, α-glycerophosphocholine, which in our study was found to be associated with a lower risk of T2D, can produce phosphorylcholine and acetylcholine (Ach), a neurotransmitter responsible for initiating skeletal muscle contractions (48). Activation of muscarinic Ach receptors in skeletal muscle cells stimulates glucose uptake via a mechanism independent of the insulin-stimulated pathway (49). Because all muscle movements are related to contraction, and contraction is related to available Ach stores, maximizing Ach may optimize muscular performance and affect insulin sensitivity, glucose transporter type 4 availability for glucose uptake, etc.
We did not observe any significant difference in the association of most 1-y change metabolites with T2D risk between the MedDiet intervention and control groups. Furthermore, the intervention diets did not appear to significantly change the study metabolite levels during the intervention. We found a stronger inverse association only for α-glycerophosphocholine with T2D risk in the MedDiet intervention compared with the control group with a significant test of this interaction. This finding needs further investigation.
The results of the present study should be interpreted in the context of its limitations and strengths. First, participants were elderly Mediterranean individuals at high CVD risk and this may limit the generalizability of the findings to other age groups or populations. Second, even though we adjusted for several potential confounders, residual confounding may exist. With regard to strengths, the prospective evaluation of the association between metabolite levels and incident T2D, in the frame of a case-cohort design, minimizes biases that can affect case-control studies.
In conclusion, our study documented, for the first time to our knowledge, a strong inverse prospective association between plasma TMAO concentrations and relevant metabolite profile with incident T2D risk in an elderly population at high CVD risk. Because TMAO is a metabolite that has shown controversial associations with several chronic diseases, these results may be reported in the literature to avoid bias but must be interpreted cautiously and need to be replicated in the future in other populations. The potential mechanisms linking the aforementioned metabolites and diabetes risk must also be further investigated.
Supplementary Material
Acknowledgements
The authors’ responsibilities were as follows—FBH and JS-S: designed the research; CP, MB, MR-C, MG-F, ET, DC, RE, ER, M Fitó, FA, M Fiol, JL, LS-M, EG-G, and JS-S: conducted the research; DC, RE, M Fitó, FA, M Fiol, JL, LS-M, EG-G, and JS-S: coordinated the subject recruitment at the outpatient clinics; CP and MB: analyzed the data; CP, MB, FBH, and JS-S: performed the statistical analysis and interpreted the data; CC: analyzed metabolomics data; CP: drafted the manuscript; FBH and JS-S: supervised the study; CP, MB, and JS-S: had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis; and all authors including EY, LL, GAF, and CR: revised the manuscript for important intellectual content and read and approved the final manuscript. None of the authors have any conflicts of interest related to this article.
Notes
Supported by research grant R01-DK-102896 from the NIH. The Prevención con Dieta Mediterránea (PREDIMED) trial was supported by the official funding agency for biomedical research of the Spanish government, the Instituto de Salud Carlos III, through grants provided to research networks specifically developed for the trial [grant RTIC G03/140 (to RE)] and through the Centro de Investigación Biomédica en Red de Fisiopatología de la Obesidad y Nutrición, and by grants from Centro Nacional de Investigaciones Cardiovasculares (grant CNIC 06/2007), the Fondo de Investigación Sanitaria Fondo Europeo de Desarrollo Regional (grants PI04-2239, PI 05/2584, CP06/00100, PI07/0240, PI07/1138, PI07/0954, PI 07/0473, PI10/01407, PI10/02658, PI11/01647, P11/02505, and PI13/00462), the Ministerio de Ciencia e Innovación (grants AGL-2009–13906-C02 and AGL2010–22319-C03), the Fundación Mapfre 2010, Consejería de Salud de la Junta de Andalucía (grant PI0105/2007), the Public Health Division of the Department of Health of the Autonomous Government of Catalonia, Generalitat Valenciana (grants ACOMP06109, GVA-COMP2010–181, GVACOMP2011–151, CS2010-AP-111, and CS2011-AP-042), and the Regional Government of Navarra (grant P27/2011). CP was supported by a postdoctoral fellowship granted by the Autonomous Government of Catalonia (PERIS 2016-2020; Incorporació de Científics I Tecnòlegs, SLT002/0016/00428). MG-F was supported by a postdoctoral fellowship granted by the Lilly Foundation European Association of Diabetes (EASD) through the Institut d'Investigacions Sanitàries Pere i Virgili (IISPV), Tarragona, Spain. The Fundación Patrimonio Comunal Olivarero and Hojiblanca (Malaga, Spain), the California Walnut Commission (Sacramento, California), Borges (Reus, Spain), and Morella Nuts (Reus, Spain) donated the olive oil, walnuts, almonds, and hazelnuts, respectively, used in the study; however, none of these funding sources played a role in the design, collection, analysis, interpretation or publication of the data.
Supplemental Figure 1 and Supplemental Tables 1–6 are available from the “Supplementary data” link in the online posting of the article and from the same link in the online table of contents at https://academic.oup.com/ajcn/.
Abbreviations used: Ach, acetylcholine; CVD, cardiovascular disease; FMO3, flavin-containing monooxygenase 3; LPC, lyso-phosphatidylcholine; LPE, lyso-phosphatidylethanolamine; MedDiet, Mediterranean diet; PREDIMED, Prevención con Dieta Mediterránea; T2D, type 2 diabetes; TMA, trimethylamine; TMAO, trimethylamine-N-oxide.
REFERENCES
- 1. Velasquez MT, Ramezani A, Manal A, Raj DS. Trimethylamine N-oxide: the good, the bad and the unknown. Toxins (Basel) 2016;8:326. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Rohrmann S, Linseisen J, Allenspach M, von Eckardstein A, Müller D. Plasma levels of trimethylamine-N-oxide are directly associated with dairy food consumption and low-grade inflammation in a German adult population. J Nutr 2016;146:283–9. [DOI] [PubMed] [Google Scholar]
- 3. Cho CE, Taesuwan S, Malysheva OV, Bender E, Tulchinsky NF, Yan J, Sutter JL, Caudill MA. Trimethylamine-N-oxide (TMAO) response to animal source foods varies among healthy young men and is influenced by their gut microbiota composition: a randomized controlled trial. Mol Nutr Food Res 2017;61. [DOI] [PubMed] [Google Scholar]
- 4. 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. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. 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-N-oxide, a metabolite associated with atherosclerosis, exhibits complex genetic and dietary regulation. Cell Metab 2013;17:49–60. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Kim Y, Keogh J, Clifton P. A review of potential metabolic etiologies of the observed association between red meat consumption and development of type 2 diabetes mellitus. Metabolism 2015;64:768–79. [DOI] [PubMed] [Google Scholar]
- 7. Obeid R, Awwad HM, Rabagny Y, Graeber S, Herrmann W, Geisel J. Plasma trimethylamine N-oxide concentration is associated with choline, phospholipids, and methyl metabolism. Am J Clin Nutr 2016;103:703–11. [DOI] [PubMed] [Google Scholar]
- 8. Konstantinova SV, Tell GS, Vollset SE, Nygård O, Bleie Ø, Ueland PM. Divergent associations of plasma choline and betaine with components of metabolic syndrome in middle age and elderly men and women. J Nutr 2008;138:914–20. [DOI] [PubMed] [Google Scholar]
- 9. Ejaz A, Martinez-Guino L, Goldfine AB, Ribas-Aulinas F, De Nigris V, Ribó S, Gonzalez-Franquesa A, Garcia-Roves PM, Li E, Dreyfuss JM et al.. Dietary betaine supplementation increases Fgf21 levels to improve glucose homeostasis and reduce hepatic lipid accumulation in mice. Diabetes 2016;65:902–12. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Fennema D, Phillips IR, Shephard EA. Trimethylamine and trimethylamine N-oxide, a flavin-containing monooxygenase 3 (FMO3)-mediated host-microbiome metabolic axis implicated in health and disease. Drug Metab Dispos 2016;44:1839–50. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Mingrone G, Greco AV, Capristo E, Benedetti G, Giancaterini A, De Gaetano A, Gasbarrini G. L-carnitine improves glucose disposal in type 2 diabetic patients. J Am Coll Nutr 1999;18:77–82. [DOI] [PubMed] [Google Scholar]
- 12. Alhazmi A, Stojanovski E, McEvoy M, Garg ML. The association between dietary patterns and type 2 diabetes: a systematic review and meta-analysis of cohort studies. J Hum Nutr Diet 2014;27:251–60. [DOI] [PubMed] [Google Scholar]
- 13. McEvoy CT, Cardwell CR, Woodside JV, Young IS, Hunter SJ, McKinley MC. A posteriori dietary patterns are related to risk of type 2diabetes: findings from a systematic review and meta-analysis. J Acad Nutr Diet 2014;114:1759–75. [DOI] [PubMed] [Google Scholar]
- 14. Salas-Salvadó J, Bulló M, Estruch R, Ros E, Covas MI, Ibarrola-Jurado N, Corella D, Arós F, Gómez-Gracia E, Ruiz-Gutiérrez V et al.. Prevention of diabetes with Mediterranean diets: a subgroup analysis of a randomized trial. Ann Intern Med 2014;160:1–10. [DOI] [PubMed] [Google Scholar]
- 15. Cornelis MC, Hu FB. Systems epidemiology: a new direction in nutrition and metabolic disease research. Curr Nutr Rep 2013;2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Estruch R, Ros E, Salas-Salvadó J, Covas MI, Corella D, Arós F, Gómez-Gracia E, Ruiz-Gutiérrez V, Fiol M, Lapetra J et al.. Primary prevention of cardiovascular disease with a Mediterranean diet. N Engl J Med 2013;368:1279–90. [DOI] [PubMed] [Google Scholar]
- 17. Martínez-González MÁ, Corella D, Salas-Salvadó J, Ros E, Covas MI, Fiol M, Wärnberg J, Arós F, Ruíz-Gutiérrez V, Lamuela-Raventós RM et al.. Cohort profile: design and methods of the PREDIMED study. Int J Epidemiol 2012;41:377–85. [DOI] [PubMed] [Google Scholar]
- 18. Wang TJ, Larson MG, Vasan RS, Cheng S, Rhee EP, McCabe E, Lewis GD, Fox CS, Jacques PF, Fernandez C et al.. Metabolite profiles and the risk of developing diabetes. Nat Med 2011;17:448–53. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Rhee EP, Cheng S, Larson MG, Walford GA, Lewis GD, McCabe E, Yang E, Farrell L, Fox CS, O'Donnell CJ et al.. Lipid profiling identifies a triacylglycerol signature of insulin resistance and improves diabetes prediction in humans. J Clin Invest 2011;121:1402–11. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Lewis GD, Farrell L, Wood MJ, Martinovic M, Arany Z, Rowe GC, Souza A, Cheng S, McCabe EL, Yang E et al.. Metabolic signatures of exercise in human plasma. Sci Transl Med 2010;2:33ra37. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Gohil VM, Sheth SA, Nilsson R, Wojtovich AP, Lee JH, Perocchi F, Chen W, Clish CB, Ayata C, Brookes PS et al.. Nutrient-sensitized screening for drugs that shift energy metabolism from mitochondrial respiration to glycolysis. Nat Biotechnol 2010;28:249–55. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Rhee EP, Souza A, Farrell L, Pollak MR, Lewis GD, Steele DJ, Thadhani R, Clish CB, Greka A, Gerszten RE et al.. Metabolite profiling identifies markers of uremia. J Am Soc Nephrol 2010;21:1041–51. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Shaham O, Slate NG, Goldberger O, Xu Q, Ramanathan A, Souza AL, Clish CB, Sims KB, Mootha VK. A plasma signature of human mitochondrial disease revealed through metabolic profiling of spent media from cultured muscle cells. Proc Natl Acad Sci USA 2010;107:1571–5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Zhong L, D'Urso A, Toiber D, Sebastian C, Henry RE, Vadysirisack DD, Guimaraes A, Marinelli B, Wikstrom JD, Nir T et al.. The histone deacetylase Sirt6 regulates glucose homeostasis via Hif1alpha. Cell 2010;140:280–93. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. American Diabetes Association. Diagnosis and classification of diabetes mellitus. Diabetes Care 2008;31:55–60. [Google Scholar]
- 26. Elosua R, Marrugat J, Molina L, Pons S, Pujol E; MARATHOM Investigators. Validation of the Minnesota Leisure Time Physical Activity Questionnaire in Spanish men. Am J Epidemiol 1994;139:1197–209. [DOI] [PubMed] [Google Scholar]
- 27. Schröder H, Fitó M, Estruch R, Martínez-González MA, Corella D, Salas-Salvadó J, Lamuela-Raventós R, Ros E, Salaverría I, Fiol M et al.. A short screener is valid for assessing Mediterranean diet adherence among older Spanish men and women. J Nutr 2011;141;1140–5. [DOI] [PubMed] [Google Scholar]
- 28. Matthews DR, Hosker JP, Rudenski AS, Naylor BA, Treacher DF, Turner RC. Homeostasis model assessment: insulin resistance and beta-cell function from fasting plasma glucose and insulin levels in man. Diabetologia 1985;28:412–9. [DOI] [PubMed] [Google Scholar]
- 29. Benjamini Y, Hochberg Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J R Stat Soc Ser B Methodol 1995;57:289–300. [Google Scholar]
- 30. Stephens FB, Constantin-Teodosiu D, Greenhaff PL. New insights concerning the role of carnitine in the regulation of fuel metabolism in skeletal muscle. J Physiol 2007;581:431–44. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. Schwab U, Törrönen A, Meririnne E, Saarinen M, Alfthan G, Aro A, Uusitupa M. Orally administered betaine has an acute and dose-dependent effect on serum betaine and plasma homocysteine levels in healthy humans. J Nutr 2006;136:34–8. [DOI] [PubMed] [Google Scholar]
- 32. Huang T, Ren J, Huang J, Li D. Association of homocysteine with type 2 diabetes: a meta-analysis implementing Mendelian randomization approach. BMC Genomics 2013;14:867. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33. Cho NH, Lim S, Jang HC, Park HK, Metzger BE. Elevated homocysteine as a risk factor for the development of diabetes in women with a previous history of gestational diabetes mellitus: a 4-year prospective study. Diabetes Care 2005;28:2750–5. [DOI] [PubMed] [Google Scholar]
- 34. Lever M, Atkinson W, Slow S, Chambers ST, George PM. Plasma and urine betaine and dimethyl-glycine variation in healthy young male subjects. Clin Biochem 2009;42:706–12. [DOI] [PubMed] [Google Scholar]
- 35. 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-N-oxide. Eur Hear J 2014;35:904–10. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36. Dambrova M, Latkovskis G, Kuka J, Strele I, Konrade I, Grinberga S, Hartmane D, Pugovics O, Erglis A, Liepinsh E. Diabetes is associated with higher trimethylamine N-oxide plasma levels. Exp Clin Endocrinol Diabetes 2016;124(4):251–6. [DOI] [PubMed] [Google Scholar]
- 37. Cho CE, Caudill MA. Trimethylamine-N-oxide: friend, foe, or simply caught in the cross-fire? Trends Endocrinol Metab 2017;28:121–30. [DOI] [PubMed] [Google Scholar]
- 38. Trøseid M, Hov JR, Nestvold TK, Thoresen H, Berge RK, Svardal A, Lappegård KT. Major increase in microbiota-dependent proatherogenic metabolite TMAO one year after bariatric surgery. Metab Syndr Relat Disord 2016;14:197–201. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39. Shan Z, Sun T, Huang H, Chen S, Chen L, Luo C, Yang W, Yang X, Yao P, Cheng J et al.. Association between microbiota-dependent metabolite trimethylamine-N-oxide and type 2 diabetes. Am J Clin Nutr 2017;106:888–94. [DOI] [PubMed] [Google Scholar]
- 40. 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. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41. McEntyre CJ, Lever M, Chambers ST, George PM, Slow S, Elmslie JL, Florkowski CM, Lunt H, Krebs JD. Variation of betaine, N,N-dimethylglycine, choline, glycerophosphorylcholine, taurine and trimethylamine-N-oxide in the plasma and urine of overweight people with type 2 diabetes over a two-year period. Ann Clin Biochem 2015;52:352–60. [DOI] [PubMed] [Google Scholar]
- 42. Wallace M, Morris C, O'Grada CM, Ryan M, Dillon ET, Coleman E, Gibney ER, Gibney MJ, Roche HM, Brennan L. Relationship between the lipidome, inflammatory markers and insulin resistance. Mol Biosyst 2014;10:1586–95. [DOI] [PubMed] [Google Scholar]
- 43. Floegel A, Stefan N, Yu Z, Mühlenbruch K, Drogan D, Joost HG, Fritsche A, Häring HU, Hrabě de Angelis M, Peters A et al.. Identification of serum metabolites associated with risk of type 2 diabetes using a targeted metabolomic approach. Diabetes 2013;62:639–48. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44. Yea K, Kim J, Yoon JH, Kwon T, Kim JH, Lee BD, Lee HJ, Lee SJ, Kim JI, Lee TG et al.. Lysophosphatidylcholine activates adipocyte glucose uptake and lowers blood glucose levels in murine models of diabetes. J Biol Chem 2009;284:33833–40. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45. Nishimukai M, Maeba R, Yamazaki Y, Nezu T, Sakurai T, Takahashi Y, Hui SP, Chiba H, Okazaki T, Hara H. Serum choline plasmalogens, particularly those with oleic acid in sn-2, are associated with proatherogenic state. J Lipid Res 2014;55:956–65. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 46. Pacifico L, Nobili V, Anania C, Verdecchia P, Chiesa C. Pediatric nonalcoholic fatty liver disease, metabolic syndrome and cardiovascular risk. World J Gastroenterol 2011;17:3082–91. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 47. Puri P, Wiest MM, Cheung O. The plasma lipidomic signature of nonalcoholic steatohepatitis. Hepatology 2009;50:1827–38. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 48. Amenta F, Tayebati SS. Pathways of acetylcholine synthesis, transport and release as targets for treatment of adult-onset cognitive dysfunction. Curr Med Chem 2008;15:488–98. [DOI] [PubMed] [Google Scholar]
- 49. Merlin J, Evans BA, Csikasz RI, Bengtsson T, Summers RJ, Hutchinson DS. The M3-muscarinic acetylcholine receptor stimulates glucose uptake in L6 skeletal muscle cells by a CaMKK-AMPK-dependent mechanism. Cell Signal 2010;22:1104–13. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.