Dietary Betaine Supplementation Increases Fgf21 Levels to Improve Glucose Homeostasis and Reduce Hepatic Lipid Accumulation in Mice

Insulin resistance occurs years before the onset of type 2 diabetes and also predicts disease development (1). Fortunately, both intensive lifestyle programs (exercise, dietary modification) and pharmacotherapy are effective for type 2 diabetes prevention in insulin resistant humans (2). Thus, a key imperative is to identify individuals at highest risk to maximize success and cost-effectiveness of preventive strategies.

Advances in mass spectrometry–based metabolomics technology have permitted identification of novel metabolic signatures associated with risk for type 2 diabetes and other components of the metabolic syndrome (3–5). To identify biomarkers associated with insulin resistance and type 2 diabetes risk, we analyzed the fasting plasma metabolome in metabolically phenotyped human volunteers. We now report that insulin resistant humans have reduced plasma levels of N,N,N-trimethylglycine, or glycine betaine (hereafter referred to as betaine). Moreover, Walford et al. (6) report that plasma betaine levels not only are associated with reduced incidence of type 2 diabetes in a prospective cohort study but also predict successful response to prevention strategies.

Betaine is a modified amino acid found in many foods, with particularly high content in whole grains (7). Both betaine intake and plasma levels are inversely correlated with several metabolic syndrome markers (8,9). Owing to its dipolar zwitterion structure, betaine functions as a key intracellular osmolyte. In mammals, betaine metabolism is dominant in the liver, where it serves as a methyl donor in the methionine cycle; the enzyme betaine-homocysteine methyltransferase catalyzes the reaction between betaine and homocysteine, yielding dimethylglycine and methionine (10). Betaine administration decreases homocysteine levels (11), a cardiovascular risk factor (12), and is used as treatment for homocysteinemia. Betaine decreases hepatic lipid content and improves glucose tolerance in rodents (13–16). However, the molecular mechanisms underlying these effects remain unknown.

Given the multiple lines of evidence supporting betaine regulation of metabolism, we tested the impact of betaine in mice with diet-induced obesity. We demonstrate that long-term betaine supplementation increases circulating levels of the systemic metabolic regulator fibroblast growth factor (Fgf)21, improves glucose homeostasis, reduces hepatic lipid, and increases whole-body energy expenditure. Betaine fails to improve metabolic health in mice lacking Fgf21, demonstrating that Fgf21 is required for the beneficial effects of betaine.

Protocols were approved by the Joslin Diabetes Center Committee on Human Subjects, and informed consent was obtained from all participants. Plasma was obtained after an overnight fast from 40 healthy individuals with normal glucose tolerance. Insulin sensitivity was quantified by insulin-modified intravenous glucose tolerance with minimal-model analysis (17). Mass spectrometry (Metabolon, Inc.) was used to analyze 193 distinct metabolites. Plasma glucose was measured using glucose oxidase (2300 Stat Plus, YSI, Yellow Springs, OH) and insulin by radioimmunoassay (Diagnostic Systems Laboratories, Webster, TX).

All studies were approved by the Institutional Animal Care and Use Committee of Joslin Diabetes Center and the University of Barcelona. Male C57BL/6 mice (Harlan) were fed chow or a 45% or 60% kcal-from-fat diet (D12450B, D12451, and D12492, respectively; Research Diets) for 16 weeks. Betaine (Sigma-Aldrich) was administered in water (1% wt/vol) 1 week prior to starting high-fat feeding and was maintained throughout. Fgf21^−/− mice (18) had been backcrossed with C57BL/6 mice for >10 generations and were treated using the same protocol and 45% fat diet.

Glucose (1.5 g/kg i.p.) and insulin (0.75 units/kg i.p.) tolerance tests were performed after overnight and 4-h fasts, respectively. Mice were fasted overnight and anesthetized prior to sacrifice. Hyperinsulinemic-euglycemic clamps were performed at the Mouse Metabolic Phenotyping Core, University of Massachusetts. After overnight fasting, a primed (150 mU/kg) continuous (15 pmol/kg/min) infusion of human insulin was initiated, together with 20% glucose at variable rates to maintain basal glucose. Insulin-stimulated glucose uptake was normalized to plasma insulin at the end of the 2-h clamp. Plasma insulin and Fgf21 were determined by ELISA (Crystal Chem and Millipore, respectively). Plasma betaine was quantified by liquid chromatography–mass spectrometry (LC-MS) after fractionation using an Acquity UPLC BEH HILIC column (Waters) and detection with a QqQ/MS 6490 mass spectrometer (Agilent). Betaine content of diets was quantified in methanol extracts using LC-MS (4000 QTRAP; AB Sciex). d9-betaine (C/D/N Isotopes) or ¹³C₂-betaine was used as internal standards for absolute quantification for plasma and chow, respectively. For 60% mouse high-fat diet studies, plasma and liver metabolites were analyzed by LC-MS (19). Liver betaine, S-adenosylmethionine, S-adenosylhomocysteine, methionine, and choline were quantified using stable-isotope dilution liquid chromatography–electrospray ionization tandem mass spectrometry (Baylor Research Institute, Dallas, TX) (20). Hepatic lipid was quantified in chloroform-methanol extracts (50 mg) using a Triglyceride Assay kit (Sigma-Aldrich).

Total RNA was isolated with TRI Reagent (Sigma-Aldrich). cDNA was synthesized using a high-capacity cDNA kit (Applied Biosystems) and analyzed by real-time PCR using SYBR Green (Promega). Oligo sequences can be provided upon request.

Mice were placed in individual chambers (Oxymax OPTO-M3 system; Columbus Instruments) with ad libitum access to food and water. After 24-h adaptation, O₂ consumption, CO₂ production, and activity were measured.

Mitochondrial function was assessed in freshly isolated tissue using high-resolution respirometry (Oxygraph-2k, Oroboros Instruments, Innsbruck, Austria) (21). Inguinal white adipose tissue (iWAT) was mechanically permeabilized (Shredder SG3, PBI Pressure BioSciences, Inc.) in respirometry medium (0.5 mmol/L EGTA, 3 mmol/L MgCl₂, 60 mmol/L K-lactobionate, 20 mmol/L taurine, 10 mmol/L KH₂PO₄, 20 mmol/L HEPES, 110 mmol/L sucrose, and 0.1% (wt/vol) BSA, pH 7.1). Oxygen flux was measured by adding malate (final 2 mmol/L), glutamate (10 mmol/L), and ADP (5 mmol/L) for complex I (CI) or succinate (10 mmol/L) for combined CI+II activity. Subsequently, oligomycin A (2 μg/mL) was added to inhibit ATP synthase (leak-dependent respiration). Finally, electron transport–independent flux was assessed after sequential addition of rotenone (0.1 μmol/L) and anti–mycin A (2.4 μmol/L) and subtracted from values of each previous step. Values are expressed relative to protein.

Tissues were fixed, paraffin embedded, and stained with hematoxylin-eosin or Oil Red O (liver). For the pancreas, three nonconsecutive sections were labeled with anti-insulin (Dako) and anti-glucagon IgG (Sigma-Aldrich) and detected with Cy3- and Cy2-conjugated secondary antibodies (Jackson ImmunoResearch). Area of islets and adipocytes was measured in 30 images from three sections per animal using ImageJ by a blinded investigator.

Primary hepatocytes were isolated from C57BL/6 mice after liver perfusion with collagenase and cultured overnight in gelatin-coated plates containing DMEM (Sigma-Aldrich), 10% FBS, and 1 μmol/L insulin and subsequently incubated with betaine.

All data are presented as mean ± SEM. Between-group differences were analyzed using GraphPad Prism, using two-tailed t test or one-way ANOVA and a post hoc Tukey honestly significant difference test. Where stated, data were analyzed by two-way ANOVA with a post hoc Holm-Sidak test. Metabolomic data were log transformed and analyzed using R software (www.r-project.org); association with log insulin sensitivity index (S_I) was tested using Pearson correlation. P values were corrected for multiple comparisons using false discovery rate (www.jstor.org/stable/2346101). P < 0.05 was considered significant for all analyses.

The plasma metabolome was analyzed in the fasting state from individuals across a spectrum of insulin sensitivity in order to evaluate the impact of insulin resistance. All participants underwent intravenous glucose tolerance tests to quantify insulin sensitivity (S_I) (22) (range 0.49–14.28). Participants were categorized as insulin sensitive (IS) or resistant (IR) based on S_I values above or below the median value of 4.78 for a larger population of normoglycemic individuals studied at the Joslin Diabetes Center (17). IR individuals had higher fasting plasma glucose and insulin levels (Table 1).

We performed unbiased analysis of plasma metabolomics using mass spectroscopy and assessed correlation with insulin sensitivity (logS_I) to identify early signatures of insulin resistance (Supplementary Data Table 2). The metabolite with the highest correlation coefficient was betaine (r = 0.55, P < 0.0005) (Fig. 1A). Betaine concentrations were also significantly lower in IR compared with IS individuals (14% reduction, P < 0.01) (Fig. 1B). Given the importance of betaine for maintenance of the methionine cycle and systemic metabolism, we hypothesized that reductions in plasma betaine could contribute to metabolic defects associated with insulin resistance. We thus determined the impact of betaine supplementation on in vivo metabolism in mice with diet-induced obesity.

Male C57BL/6 mice were fed chow (10% kcal from fat) (LF) or a matched moderate high-fat diet (45% kcal from fat [HF]) for 16 weeks; high-fat–fed mice were unsupplemented or treated with 1% betaine in water (HF and HF-B, respectively). As expected, high-fat feeding led to greater weight gain, fasting hyperglycemia, and hyperinsulinemia (Table 2). Betaine levels were 50% lower in mice made insulin resistant with high-fat feeding (Table 2), consistent with lower levels in insulin-resistant humans. Lower betaine levels in high-fat–fed mice cannot be attributed to dietary betaine content, as measured betaine content (per gram) is 19% higher in the high-fat diet and calculated daily betaine intake is similar for both groups (Supplementary Fig. 1). Supplementation increased betaine levels by 2.6- and 6.5-fold compared with LF and HF groups (Table 2), respectively, and induced a small decrease in body weight versus HF mice (Fig. 2A). Consistent with previous data (15,16), betaine-treated mice had lower fasting insulin (Table 2) and improved glucose tolerance (Fig. 2B); plasma insulin was significantly lower in HF-B mice after intraperitoneal glucose (Fig. 2C). Insulin tolerance was also improved (Fig. 2D). Pancreatic β-cell area was lower in HF-B versus HF (Fig. 2E), indicating that betaine normalized the β-cell expansion with high-fat feeding.

We observed similar results in an independent cohort fed a higher fat diet (60% kcal from fat); betaine supplementation resulted in reduced body weight and leptin, 50% reduction in both fasted and refed insulin, and improved glucose tolerance (Supplementary Fig. 2A–F). Moreover, insulin sensitivity was improved, as assessed by insulin tolerance testing, and there was a trend toward a higher glucose infusion rate during hyperinsulinemic-euglycemic clamp (2.1-fold; 5.40 ± 1.36 vs. 2.53 ± 0.76 mg/kg/min/ng/ml, P = 0.08) (Supplementary Fig. 2G and H).

Indirect calorimetry was performed to assess betaine effects on energy homeostasis. Betaine administration increased oxygen consumption by 10% (P < 0.05) in both light and dark cycles (Fig. 2F). Physical activity did not differ (Supplementary Fig. 3A), while respiratory exchange ratio tended to increase in HF-B (P = 0.056) (Supplementary Fig. 3B). Together with the human data, these observations prompted us to investigate molecular mechanisms mediating betaine action, focusing on HF and HF-B groups.

Since the liver plays a major role in betaine metabolism, we analyzed betaine effects in this tissue. Supplementation had no major impact in phosphorylated Akt in fasted/refed conditions (Supplementary Fig. 4); however, consistent with previous rodent studies (13–16), betaine decreased liver weight and triglyceride (TAG) content in mice fed either 45% (Fig. 3A and B) or 60% fat diets (Supplementary Fig. 5A and B). Fasting plasma TAG levels were also reduced (Fig. 3C). Given these robust reductions in hepatic lipid, we used LC-MS to quantify key one-carbon pathway metabolites in liver from 60% fat-fed mice. Hepatic betaine, S-adenosylmethionine, and S-adenosylhomocysteine levels were increased in betaine-supplemented mice (Supplementary Fig. 5C); methionine, homocysteine, choline, and cystathionine were not significantly altered.

We next examined potential transcriptional contributions to improved lipid content. Quantitative RT-PCR revealed no major changes in genes regulating lipogenesis or lipid oxidation or in one-carbon metabolic pathway enzymes (Supplementary Fig. 5D). However, expression of proinflammatory genes Ccl2, Tnfa, Saa2, and Tlr4 (full gene names in Supplementary Table 1) was consistently downregulated in betaine-treated mice (Fig. 3D). Since betaine has antioxidant properties (23,24), we assessed tissue oxidative stress by measuring the ratio of oxidized to reduced glutathione. Betaine supplementation decreased the oxidized-to-reduced glutathione ratio by 20% (P < 0.05, Supplementary Fig. 5E), suggesting reduced oxidative stress.

Given the robust effect of betaine supplementation on both liver and whole-body metabolism, we hypothesized that a hepatokine was mediating betaine effects. We measured hepatokines known to impact glucose homeostasis and energy expenditure, including Angptl3, Angptl4, Angptl6, Enho, Hgf, Igf1, Sepp1, and Fgf21. Among these, only Fgf21 mRNA was increased (1.6-fold, P < 0.05) after betaine treatment (Fig. 3E). Circulating Fgf21 levels were also increased in HF-B mice (Fig. 3F). For determination of whether increased liver Fgf21 expression was due to a direct transcriptional effect, primary hepatocytes were incubated with betaine (5–25 mmol/L for 5–24 h). Expression of Fgf21 was not increased in any of these conditions (Fig. 3G and H), despite increases in Fgf21 in response to tunicamycin, a known activator of Fgf21 expression (positive control for these experiments) (Fig. 3H). Thus, it is unlikely that effects of betaine on Fgf21 are acute and/or direct. Since Fgf21 is also expressed in extrahepatic tissues, including WAT and brown adipose tissue (BAT), we tested whether betaine modulated Fgf21 expression in these tissues. We found no differences in expression of Fgf21 or its receptor Fgfr1 in either tissue (Supplementary Fig. 6). By contrast, expression of the coreceptor β-klotho was elevated in iWAT (Supplementary Fig. 6), suggesting a potential increase in Fgf21 signaling with betaine supplementation.

Circulating Fgf21 exerts most of its whole-body effects via adipose tissue (25–27) and induces a browning program in WAT (28), prompting us to investigate betaine effects on subcutaneous iWAT. Histological analysis showed a shift in adipocyte distribution toward smaller size, with 25% decrease (P < 0.05) in mean cell size in HF-B mice (Fig. 4A). mRNA expression of Ucp1 and Ppargc1a was increased by 3.9- and 1.5-fold, respectively (P < 0.05 for both), with betaine (Fig. 4B), but expression of the additional browning markers Cidea or Prdm16 did not differ. Betaine did not alter Ucp1 or Ppargc1a in BAT or epididymal WAT (eWAT) (Fig. 4C), indicating that this effect was specific for iWAT. Decreased adipocyte cell size and increased oxidative gene expression together suggested that betaine improved oxidative metabolism in iWAT. Consistent with this hypothesis and increased in vivo oxygen consumption, HF-B mice showed enhanced mitochondrial oxidative capacity in iWAT. Specifically, we observed increases in oxygen consumption by 1.5-fold (P < 0.05) after the addition of substrates for both C1 (malate and glutamate) and CII (succinate) and in leak-dependent respiration (Fig. 4D).

Betaine-induced improvements in glucose homeostasis, hepatic lipid accumulation, energy expenditure, and iWAT size and oxidative profile are all consistent with a potential role of Fgf21 in mediating betaine effects. To test this hypothesis, we investigated the impact of betaine supplementation in mice null for the Fgf21 gene (Fgf21^−/−). Male Fgf21^−/− mice were fed 45% kcal from fat either without or with 1% betaine in water for 16 weeks (KO and KO-B, respectively). Supplementation increased plasma levels by fivefold in KO mice (Fig. 5A). Unexpectedly, betaine administration to Fgf21^−/− mice led to higher body weight (Fig. 5B) with no difference in food intake (Fig. 5C); these mice also showed increased fasting glucose and numerically higher insulin levels (Fig. 5D and E). Neither glucose nor insulin tolerance was improved in KO-B vs. KO (Fig. 5F and G). Furthermore, betaine failed to decrease liver weight, TAG content (Fig. 5H and I), or inflammatory markers in KO mice (Supplementary Fig. 7). We next analyzed the effects of betaine in iWAT from Fgf21^−/− mice. As shown in Fig. 5J, betaine treatment did not increase expression of Ucp1 or Ppargc1a in iWAT from KO-B versus KO mice. Finally, mitochondrial respiration did not differ between groups (Fig. 5K). Together, these data indicate that Fgf21 is necessary for the beneficial effects of betaine to improve glucose homeostasis, reduce liver TAG accumulation, and improve oxidative metabolism in adipose tissue.

Betaine is the top-ranking plasma metabolite correlated with insulin sensitivity (S_I) in humans with normal glucose tolerance, as revealed by unbiased metabolomics analysis. Moreover, increasing betaine levels via long-term dietary supplementation improve systemic metabolism in mice, with prevention of impaired glucose homeostasis and liver fat accumulation, improved insulin sensitivity, increased energy expenditure, and increased oxidative capacity of iWAT, even in mice made obese and insulin resistant with continued high-fat feeding. Thus, plasma betaine concentrations can serve as a biomarker of insulin resistance and a target for prevention and treatment.

More broadly, our data provide support for the emerging concept that one-carbon metabolic pathways may contribute to the pathogenesis of insulin resistance, type 2 diabetes, and related metabolic disease. Prior cross-sectional analysis of data from Framingham Heart Study participants demonstrated inverse correlations between plasma betaine and several phenotypes associated with metabolic disease, including BMI, insulin, blood pressure, and lipids (29). Deficiency of the betaine metabolite dimethylglycine has also been recently associated with higher plasma glucose (30). In new longitudinal prospective studies of Diabetes Prevention Program (DPP) participants, Walford et al. (6) report that higher baseline plasma betaine levels are also associated with a significant reduction in incident diabetes for up to 10 years. Moreover, plasma betaine levels are increased with lifestyle intervention, and increases in betaine predict the success of these interventions to reduce diabetes incidence.

While we do not fully understand the mechanisms responsible for these interesting relationships, reduced betaine levels in insulin-resistant humans could reflect reduced intake of betaine-rich foods, as suggested by prior dietary studies (8,31). However, our findings of reduced betaine levels in mice fed a defined high-fat diet, despite similar betaine intake, indicate additional mechanisms. Alterations in intestinal absorption, metabolism of choline-related metabolites by intestinal microbiota (32,33), or osmotic dilution or increased losses could also contribute to reduced plasma betaine in insulin resistance. Additional complexity arises from genetic variation within this pathway, as evidenced by the recently described relationship between variation at the dimethylglycine dehydrogenase (DMGDH) locus, insulin resistance, and incident diabetes (30). Additional studies will be required to clarify the relative importance of each of these mechanisms as contributors to reduced betaine levels in human insulin resistance.

Our results support the efficacy of betaine dietary supplementation to improve systemic metabolism. Metabolites from the one-carbon metabolism pathway, including choline, methionine, folic acid, and betaine, singly or in combination, have been used to prevent and treat fatty liver in rodents (15,16,34,35). Our results demonstrate that betaine also yields beneficial effects in nonhepatic tissues, with increases in systemic energy expenditure and enhanced oxidative metabolism in subcutaneous WAT. Thus, both hepatic and extrahepatic mechanisms may contribute to betaine-mediated improvements in insulin sensitivity and glucose tolerance.

One potential mediator of betaine-induced improvements in systemic metabolism is the hormone Fgf21. Fgf21 regulates whole-body glucose and lipid metabolism and is under investigation as a novel metabolic disease therapeutic (36,37). Plasma Fgf21 is elevated during fasting and in response to low protein intake via Peroxisome proliferator-activated receptor alpha (PPARA) and general control nonderepressible 2-dependent regulation (38,39); Fgf21 is also increased in obesity and nonalcoholic fatty liver disease and correlates with hepatic fat, suggesting the possibility of resistance to hormone action (38,40–42). Further increases in Fgf21 in rodents via transgenic overexpression or injection of recombinant Fgf21 improve glucose homeostasis, decrease hepatic lipid, increase whole-body energy expenditure, and reduce body weight (43–45). This is of particular relevance, as our results demonstrate that long-term betaine supplementation yields a more modest increase in both hepatic and circulating Fgf21 by 1.6-fold, while reducing hepatic TAG content. We do not fully understand the mechanisms responsible for betaine-mediated increases in Fgf21 expression or secretion. Our experiments in primary hepatocytes indicate that betaine-induced increases in Fgf21 expression are likely not to be acute or direct but, rather, indirectly mediated by improvements in hepatic insulin action, transcriptional effects on PPARA or related nuclear receptor complexes (38,39,46,47), reduced lipid accumulation, or reduced oxidative stress or via extrahepatic metabolic effects.

Fgf21-induced browning in subcutaneous WAT (28) could contribute to enhanced energy expenditure. Consistent with this effect, betaine treatment increased expression of genes regulating oxidative metabolism, including Ucp1 and Ppargc1a, in subcutaneous WAT; other browning markers were not modified. In parallel, high-resolution respirometry revealed increased iWAT mitochondrial oxidative capacity in betaine-treated mice. While we do not fully understand the mechanisms mediating this effect, increased mitochondrial oxygen consumption and leak-dependent respiration could be mediated in part by the modestly increased Ucp1 and Ppargc1a expression in betaine-treated adipose. Effects of Fgf21 have recently been shown to be Ucp1 independent (48,49), suggesting important contributions of Ppargc1a and other regulators of oxidative metabolism (48,50). Expression of Ucp1 and Ppargc1a did not change with betaine in either BAT or eWAT, consistent with greater susceptibility of iWAT to browning stimuli (51) and Fgf21-mediated induction of thermogenic programs (28). Although we cannot rule out a direct effect of betaine in eWAT (16), the absence of Ucp1 or Ppargc1a induction in iWAT from betaine-treated Fgf21^−/− mice indicates that Fgf21 is necessary for long-term betaine effects.

Improvements in whole-body glucose homeostasis, liver fat, energy expenditure, and iWAT are all consistent with Fgf21 as a mediator of betaine action. Indeed, betaine supplementation failed to improve metabolic health in Fgf21^−/− mice, demonstrating that Fgf21 is necessary for betaine metabolic effects. Liver-derived Fgf21 is considered the main source of circulating Fgf21 and responsible for its beneficial systemic effects (40); betaine-induced increases in hepatic Fgf21 mRNA paralleled increased plasma levels. However, we recognize that these mice bear a whole-body knockout, and thus we cannot fully exclude a potential role for extrahepatic Fgf21. Unexpectedly, betaine administration to Fgf21^−/− mice led to worsening of some metabolic parameters. A plausible explanation is that betaine demethylation ultimately yields glycine, which can be oxidized and provide additional calories. Thus, long-term supplementation could contribute to higher cumulative calorie intake; lack of Fgf21 might impede expenditure of these extra calories, worsening metabolic health. Another possibility is that Fgf21 could regulate betaine metabolism and function; lack of Fgf21 would impede the beneficial effects exerted by long-term betaine supplementation by altering its metabolism.

Additional mechanisms are likely to contribute to betaine-induced improvements in systemic and hepatic metabolism. Betaine is a crucial methyl group donor in the methionine cycle, converting homocysteine into methionine and dimethylglycine in a reaction catalyzed by the enzyme Bhmt. Methionine is sequentially converted into S-adenosylmethionine, and dimethylglycine is further metabolized via the dimethylglycine dehydrogenase to glycine. It is interesting that betaine supplementation increased hepatic content not only of its direct product dimethylglycine, but also of S-adenosylmethionine and its demethylation product S-adenosylhomocysteine (52). Thus, while static assessment of the metabolome does not allow assessment of net cycle flux, these and other downstream metabolites of betaine may be important modulators of betaine effects.

Results from Bhmt^−/− mice indicate additional complexity (53). These mice show increased liver fat, possibly due to impaired methionine cycle activity and phosphatidylcholine biosynthesis. Surprisingly, Bhmt^−/− mice are leaner and have improved glucose homeostasis, increased energy expenditure, and higher respiratory exchange ratio, resembling the phenotypic responses to betaine. Interestingly, lack of Bhmt also increases betaine accumulation and circulating Fgf21, strongly supporting a role for betaine itself, rather than solely its methyl donor capacity, in increasing Fgf21 levels and improving whole-body metabolism. Further studies will be required to evaluate flux through one-carbon metabolic pathways and to identify the contribution of specific metabolites to betaine-mediated improvements in both hepatic lipid metabolism and Fgf21 secretion.

Another unique feature of betaine is its zwitterion structure; thus, osmolyte functions may also contribute to betaine-mediated improvements in metabolic health. Finally, betaine may also reduce oxidative stress, as demonstrated in multiple tissues and experimental paradigms (23,24), potentially related to betaine-mediated increases in glutathione biosynthesis and antioxidant responses. In agreement, we found that betaine reduces the ratio of oxidized/reduced glutathione. Reductions in oxidative stress may also be linked to increased complete fatty acid oxidation and AMP kinase activation, as observed with a variety of methyl donors (13,35,54).

The robust effects of betaine supplementation in rodents have prompted human clinical studies focusing on fatty liver and weight loss. Betaine administration to obese humans with nonalcoholic fatty liver disease protected against worsening steatosis and improved hepatic inflammation but did not fully reverse disease (55). In another study, betaine administration to obese individuals on a hypocaloric diet did not enhance weight loss (56). However, these studies were performed on subjects with a preexisting condition and were not designed to examine glucose homeostasis. By contrast, new human data, presented by Walford et al. (6), indicate that betaine levels are not only linked to diabetes risk but also increased in response to lifestyle interventions that successfully reduce diabetes incidence.

Taken together, our human and animal model data support a potential use for long-term betaine supplementation as part of a comprehensive lifestyle intervention aimed at metabolic disease prevention. Future clinical studies will be required to establish whether increased intake of betaine-rich foods and/or long-term betaine supplementation will be a safe and effective strategy to improve metabolic health and prevent type 2 diabetes in at-risk humans.

Funding and Duality of Interest. A.E. and M.-E.P. were supported by an unrestricted 3ARP grant from Ajinomoto, Inc., and a mentor-based grant from the American Diabetes Association (7-12-MN-66). A.B.G. was supported by a grant from the American Diabetes Association (7-13-CE-17). This work was supported by Seventh Framework Programme Marie Curie FP7-PEOPLE-2011-CIG, grants from the Spanish Government (Ministerio de Economía y Competitividad) (RYC2010-06789 and SAF2011-28502), and European Foundation for the Study of Diabetes/Lilly Diabetes grants (to C.L.); Ministerio de Economía y Competitividad grants RYC-2009-05158 and BFU2011-24679 (to P.M.G.-R.); and the Seventh Framework Programme FP7-BetaBat project (to F.V.). R.E.G. was supported by National Institutes of Health (NIH) grant DK-081572. A.E., A.G.-F., J.M.D., and M.-E.P. were also supported by NIH grant DK-036836. The National Mouse Metabolic Phenotyping Center at the University of Massachusetts was supported by NIH grant U24-DK-093000. No other potential conflicts of interest were reported.

Author Contributions. A.E., L.M.-G., M.-E.P., and C.L. designed the experiments, analyzed data, and wrote the manuscript. A.B.G. recruited and characterized metabolic phenotypes in the human cohort. F.R.-A. and F.V. provided Fgf21^−/− mice and contributed to metabolic phenotyping. V.D.N. performed islet morphometry and together with S.R. assisted with metabolic phenotyping. A.G.-F. and P.M.G.-R. performed high-resolution respirometry. E.L. performed mouse care and contributed to physiologic analysis. J.M.D. performed statistical analysis on metabolomics data. W.G. and R.E.G. performed metabolomics analyses on human and mouse samples, respectively. J.K.K. performed insulin sensitivity analysis. T.B. performed quantitative assay for one-carbon metabolites. All authors assisted in manuscript writing and editing. M.-E.P. and C.L. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.