Mouse Betaine-Homocysteine S-Methyltransferase Deficiency Reduces Body Fat via Increasing Energy Expenditure and Impairing Lipid Synthesis and Enhancing Glucose Oxidation in White Adipose Tissue*

Background: Mice with the gene encoding betaine-homocysteine S-methyltransferase deleted (Bhmt−/−) have absent hepatic BHMT activity and reduced fat mass. Results: Bhmt−/− mice have increased energy expenditure, reduced fat synthesis, and enhanced glucose oxidation in white adipose tissue. Conclusion: BHMT plays a role in energy homeostasis. Significance: Liver BHMT activity affects adipose tissue metabolism. Betaine-homocysteine S-methyltransferase (BHMT) catalyzes the synthesis of methionine from homocysteine. In our initial report, we observed a reduced body weight in Bhmt−/− mice. We initiated this study to investigate the potential role of BHMT in energy metabolism. Compared with the controls (Bhmt+/+), Bhmt−/− mice had less fat mass, smaller adipocytes, and better glucose and insulin sensitivities. Compared with the controls, Bhmt−/− mice had increased energy expenditure, with no changes in food intake, fat uptake or absorption, or in locomotor activity. The reduced adiposity in Bhmt−/− mice was not due to hyperthermogenesis. Bhmt−/− mice failed to maintain a normal body temperature upon cold exposure because of limited fuel supplies. In vivo and ex vivo tests showed that Bhmt−/− mice had normal lipolytic function. The rate of 14C-labeled fatty acid incorporated into [14C]triacylglycerol was the same in Bhmt+/+ and Bhmt−/− gonadal fat depots (GWAT), but it was 62% lower in Bhmt−/− inguinal fat depots (IWAT) compared with that of Bhmt+/+ mice. The rate of 14C-labeled fatty acid oxidation was the same in both GWAT and IWAT from Bhmt+/+ and Bhmt−/− mice. At basal level, Bhmt−/− GWAT had the same [14C]glucose oxidation as did the controls. When stimulated with insulin, Bhmt−/− GWAT oxidized 2.4-fold more glucose than did the controls. Compared with the controls, the rate of [14C]glucose oxidation was 2.4- and 1.8-fold higher, respectively, in Bhmt−/− IWAT without or with insulin stimulus. Our results show for the first time a role for BHMT in energy homeostasis.

Betaine-homocysteine S-methyltransferase (BHMT) 2 is an enzyme predominantly found in the liver in rodents (1), and it plays an essential role in regulating one-carbon metabolism. BHMT catalyzes the formation of the amino acid methionine from homocysteine using the choline metabolite betaine as the methyl donor. BHMT deficiency leads to elevated betaine and homocysteine concentrations and reduced choline concentration (2). BHMT also influences hepatic lipid accumulation via altering phosphatidylcholine (PtdCho) concentration. BHMT exerts this effect through two mechanisms. First, choline, the precursor of betaine, can alternatively be used to make PtdCho via the Kennedy pathway. Second, methionine, the end product of BHMT, is the precursor of S-adenosylmethionine, which is required for the synthesis of PtdCho from phosphatidylethanolamine by the enzyme phosphatidylethanolamine methyltransferase. BHMT overexpression increases PtdCho synthesis, leading to reduced hepatic lipid accumulation (3), whereas BHMT deficiency leads to fatty liver (2). Although the role of BHMT in hepatic lipid metabolism has been elucidated, there is no previous evidence that BHMT plays a role in whole body energy metabolism and adiposity.
We observed that Bhmt-deficient (Bhmt Ϫ/Ϫ ) mice had a lower body weight from 5 to 9 weeks of age compared with wild type littermates (2). Magnetic resonance image (MRI) analysis and dissection of adipose fat depots revealed that the reduced * This work was supported, in whole or in part, by National Institutes of Princeton, NJ) to 7-week-old mice after a 6-or 4-h fast, respectively. Blood glucose was measured at base line, 15,30,60, and 120 min by tail nick using a calibrated glucometer (One Touch Ultra, LifeScan, Inc., Milpitas, CA).
Plasma Metabolites-Plasma insulin (Ultra-sensitive Mouse Insulin ELISA kit, Crystal Chem Inc., Downers Grove, IL), total thyroxine (T 4 ), total triiodothyronine (T 3 ) (Alpha Diagnostic International, San Antonio, TX), and growth hormone (Millipore) levels were determined using commercially available ELISA kits according to the manufacturers' protocols. Plasma fibroblast growth factor 21 (FGF21) (Phoenix Pharmaceuticals Inc., Burlingame, CA) and glucagon (Millipore) were determined using RIA kits according to the manufacturers' protocols. Plasma creatine kinase was measured using an automatic chemical analyzer (Johnson and Johnson VT250, Rochester, NY) at the Animal Clinical Chemistry and Gene Expression Facility, University of North Carolina, Chapel Hill. Plasma TAG, glycerol (Sigma), and nonesterified fatty acids (NEFA) (Wako, Richmond, VA) were measured colorimetrically per the manufacturers' instructions. Plasma glucose was measured using a calibrated glucometer (One Touch Ultra).
Tissue Metabolites-Hepatic free glucose and glycogen were measured using an adopted acid hydrolysis method (6). Liver was collected from mice after a 4-h fast. Briefly, 40 mg of pulverized liver was homogenized in 1 ml of 1 N HCl. Half of the homogenate was neutralized with 0.5 ml of 1 M NaOH (nonhydrolyzed group). The other half was incubated at 95°C for 90 min before being neutralized with NaOH (hydrolyzed group). Glucosyl units released in both groups were measured using a colorimetrical glucose kit (Wako). The nonhydrolyzed group represented the free glucose measurement, while the difference between the hydrolyzed and the nonhydrolyzed groups represented the glycogen measurement. For bile acids and steroids, liver and gonadal fat pads were collected from 5-week-old mice after a 4-h fast. Tissues were sent to Metabolon (Research Triangle Park, NC), where the metabolites were measured using gas chromatography/mass spectrometry (GC/MS) and liquid chromatography/mass spectrometry (LC/MS) platforms as described previously (7)(8)(9)(10). Data were extracted and analyzed by Metabolon (7)(8)(9)(10). Briefly, raw data counts for each compound were corrected for variation resulting from instrument inter-day tuning differences and then rescaled to median equal 1. Data were expressed as fold change relative to wild type controls.
Reverse Transcription-PCR-Total RNA was isolated from tissue using RNeasy mini kit (Qiagen, Valencia, CA), and cDNA was synthesized using a high capacity cDNA reverse transcription kit (Applied Biosystems, Carlsbad, CA). cDNA was amplified by real time PCR using SsoFast EvaGreen Supermix (Bio-Rad) with primers specific to each gene of interest. Results were normalized to the housekeeping genes and calculated relative to the controls by the 2 Ϫ⌬⌬CT method.
Lipolysis-For in vivo lipolysis, mice were injected intraperitoneally with lipolytic stimuli (10 mg isoproterenol/1 kg BW or 1 mg CL316243/1 kg BW) (Sigma). Plasma was collected retroorbitally at base line and 1.5 h after injection. Plasma glycerol (Sigma) and NEFA (Wako) were measured colorimetrically, and glucose was measured using a glucometer (One Touch Ultra). Gonadal and inguinal adipose explants were used for ex vivo lipolysis measurement as described previously (11,12). Briefly, pieces of fat explants (20 mg) were incubated in 0.5 ml of Krebs Ringer Buffer (KRB) (12 mM HEPES, 121 mM NaCl, 4.9 mM KCl, 1.2 mM MgSO 4 , and 0.33 mM CaCl 2 ) supplemented with 3.5% fatty acid-free bovine serum albumin (FFF-BSA) and 0.1% glucose with or without 10 M isoproterenol at 37°C. Twenty five l of buffer was collected at various time points to measure the release of glycerol colorimetrically (Sigma).
Adipocyte Isolation-Mature adipocytes were isolated from gonadal or inguinal fat depots from 12-16 week-old mice and used for incorporation and oxidation studies (13). Briefly, gonadal or inguinal fat depots were removed under sterile conditions, minced, and incubated at 37°C with shaking for 1 h in KRB supplemented with 1% FFF-BSA, 2.5 mM glucose, and 200 nM adenosine, containing 1 mg/ml collagenase I. The digested tissues were filtered through a sterile 250-m mesh and washed three times with supplemented KRB. Adipocytes were diluted with supplemented KRB to reach 10% (by volume) adipocyte solution and were maintained for 30 min in a polypropylene tube before the start of the assay. Because of the small fat mass in Bhmt Ϫ/Ϫ mice, the same fat depots from two Bhmt Ϫ/Ϫ mice were pooled together to get enough adipocytes per sample.
Fatty Acid Incorporation and Oxidation-For fatty acid incorporation into TAG, 500 l of 10% adipocyte solution were incubated with a 125-l reaction mixture containing 15 M [1-14 C]oleate (50 Ci/500 l; PerkinElmer Life Sciences) and 200 M unlabeled oleate complexed to BSA for 2 h at 37°C. [ 14 C]TAG was extracted (14), isolated by thin layer chromatography, and detected and quantified with a Bioscan AR2000 Image System (Bioscan Inc., Washington, D. C.). The chromatoplate was developed in chloroform/methanol/ammonium hydroxide (65:25:4; v/v/v) to 8 cm from the top. After the residual solvents evaporated, the plate was rerun in heptane/isopropyl ether/acetic acid (60:40:4; v/v/v) to the top of the plate. For oxidation, 250 l of 10% adipocytes solution were incubated with a 125-l reaction mixture containing either 15 M [1-14 C]oleate and 200 M unlabeled oleate or 4 M [ 14 C]glucose (50 Ci/500 l; PerkinElmer Life Sciences) and 4 mM unlabeled glucose (in KRB) for 1.5 h at 37°C. The reaction mixture was incubated in a 13-ml polypropylene tube that contained a center well filled with folded filter paper. The tube was sealed with a rubber stopper. After 1.5 h of incubation, the reaction mixture was acidified with 150 l of 70% perchloric acid, and 300 l of 1 M NaOH was added to the center well. The sealed tubes were incubated at RT for an additional 45 min. The center well (containing [ 14 C]CO 2 trapped by base) was then collected and placed into scintillation fluid (ScintiSafe 30% Mixture, Fisher) and counted. The acidified reaction mixture (ϳ400 l) was incubated overnight at 4°C with 200 l of 20% BSA. The mixture was centrifuged at 13,200 rpm for 20 min, and an aliquot of the supernatant was counted to determine 14 C-labeled acidsoluble metabolites (ASM). [ 14 C]CO 2 represents complete fatty acid or glucose oxidation, and 14 C-labeled ASM represents incomplete fatty acid oxidation. For some experiments, 0.01 unit of insulin (Novo Nordisk Inc., Princeton, NJ) was added. For hepatic FA oxidation experiments, fresh liver was excised, minced, and homogenated in a buffer (250 mM sucrose, 1 mM EDTA, 10 mM Tris-HCl, 2 mM ATP) to reach a 1:50 dilution. Forty l of liver homogenate was used to measure FA oxidation as described above.
Tissue Histology-Adipose tissues were collected from 7-and 12-week-old mice and fixed in 4% paraformaldehyde in 0.1 M phosphate buffer for 24 h. Tissues were processed, paraffinembedded, sectioned at 5 m, and stained with hematoxylin and eosin as described previously (15). For adipocyte size analysis, three images (ϫ200) per sample were taken, and the average adipocyte cell areas were analyzed using AxioVision 4.8 software (Zeiss).
Betaine Supplementation Study-7-Week-old Bhmt ϩ/ϩ mice were fed either 0 or 5% betaine (Sigma)-supplemented water for 2 weeks. Livers were collected for the measurement of Fgf21 expression as described above.
Statistics-Statistical differences were determined using analysis of variance, Tukey-Kramer HSD, and Student's t test (JMP Version 6.0; SAS Institute, Cary, NC) and reported as means Ϯ S.E. For the indirect calorimetry data analysis, the total area under the curve was calculated per measure per mouse using SAS. Statistical differences between the groups were determined using Student's t test (for mean comparison) and Mann-Whitney test (for median comparison).

BHMT Deficiency Resulted in Reduced Fat Mass and Smaller
Adipocytes-We previously reported that Bhmt Ϫ/Ϫ mice gained less body weight between 5 and 9 weeks of age than did their wild type littermates (p Ͻ 0.05) (2). To determine whether the reduced weight gain of Bhmt Ϫ/Ϫ mice was associated with reduced fat or lean mass, we measured body composition of these mice using MRI scan analysis. Bhmt Ϫ/Ϫ mice had 20 and 30% less fat mass than did Bhmt ϩ/ϩ mice at 7 and 14 weeks of age, respectively (p Ͻ 0.05), with no significant difference in lean mass (Fig. 1A). The difference in fat mass was confirmed by collecting the individual fat pads. The two major white adipose tissues (WAT), gonadal (GWAT) and inguinal (IWAT), of Bhmt Ϫ/Ϫ mice were significantly smaller than those of Bhmt ϩ/ϩ mice at all ages (Fig. 1B). The disparity in fat mass was exacerbated as mice aged, such that by 48 weeks of age, Bhmt Ϫ/Ϫ mice had a fat mass that was 41% that of Bhmt ϩ/ϩ mice. To determine whether the reduced fat mass in Bhmt Ϫ/Ϫ mice was due to fewer fat cells and/or smaller fat cells, the sizes of individual adipocytes in GWAT and IWAT of Bhmt ϩ/ϩ and Bhmt Ϫ/Ϫ mice were analyzed. Histological images showed that both GWAT and IWAT from Bhmt Ϫ/Ϫ mice contained a larger number of smaller adipocytes (Fig. 1C). We observed a 43% reduction in adipocyte size in GWAT in both 7-and 12-weekold Bhmt Ϫ/Ϫ mice compared with those of Bhmt ϩ/ϩ mice (p Ͻ 0.05) (Fig. 1D). We observed a 69% (p Ͻ 0.05) and an 80% (p Ͻ 0.001) reduction in adipocyte size in IWAT in 7-and 12-week old Bhmt Ϫ/Ϫ mice, respectively, compared with those of Bhmt ϩ/ϩ mice (Fig. 1D). These data suggested that Bhmt Ϫ/Ϫ mice likely had normal adipocyte differentiation and that the Bhmt Ϫ/Ϫ adipocytes either stored less and/or used more lipids than did Bhmt ϩ/ϩ adipocytes. Bhmt is predominantly expressed in rodent liver. Western blotting confirmed that BHMT protein was specific to the liver and was absent in the adipose tissue from both Bhmt ϩ/ϩ and Bhmt Ϫ/Ϫ mice (Fig. 1E), suggesting that the reduced fat mass in Bhmt Ϫ/Ϫ mice was due to an indirect effect of the lack of hepatic BHMT activity.
BHMT Deficiency Resulted in Enhanced Glucose Tolerance and Insulin Sensitivity-Changes in adiposity are often associated with alterations in glucose and insulin homeostasis. Bhmt Ϫ/Ϫ mice had normal basal blood glucose levels ( Fig. 1, F and G). Upon intraperitoneal injection of either glucose (Fig.  1F) or insulin (Fig. 1G), Bhmt Ϫ/Ϫ mice exhibited a faster glucose removal rate than did wild type controls. Bhmt Ϫ/Ϫ mice also had a basal plasma insulin concentration that was 50% that of the controls (p Ͻ 0.05) ( Table 2). These data indicated that Bhmt Ϫ/Ϫ mice had enhanced insulin sensitivity and glucose tolerance.
Bhmt Ϫ/Ϫ Mice Had Increased Energy Expenditure-To investigate the mechanisms that elicit the reduced fat mass in Bhmt Ϫ/Ϫ mice, elements of energy metabolism, including food intake, lipid uptake and absorption, body temperature, energy expenditure and activity, were determined. Bhmt ϩ/ϩ and Bhmt Ϫ/Ϫ mice consumed similar amounts of food per day ( Fig.  2A). Fat load test (Fig. 2B) and fecal lipid content (Fig. 2C), tests for lipid uptake and absorption, respectively, displayed no differences between Bhmt ϩ/ϩ and Bhmt Ϫ/Ϫ mice. Bhmt ϩ/ϩ and Bhmt Ϫ/Ϫ mice had similar rectal temperatures (Fig. 2D). Compared with wild type controls, Bhmt Ϫ/Ϫ mice had increased O 2 consumption ( Fig. 2E and Table 1) and increased CO 2 release ( Fig. 2F) throughout the light and dark phases (Table 1), indicating increased energy expenditure. Locomotor activity was not increased in Bhmt Ϫ/Ϫ mice (Fig. 2H), suggesting that the increase in energy expenditure was independent of activity. The RER, an indicator of metabolic fuel preference, was not significantly different between Bhmt ϩ/ϩ and Bhmt Ϫ/Ϫ mice (Fig. 2G), although there was a trend of increasing RER during the light phase (p ϭ 0.053) ( Table 1).
Thermogenesis Was Not Increased in Bhmt Ϫ/Ϫ Mice-We next investigated whether the reduced adiposity observed in Bhmt Ϫ/Ϫ mice was due to increased thermogenesis. We predicted that Bhmt Ϫ/Ϫ mice would maintain a higher body temperature during a cold challenge if they had increased thermogenesis. After 6 h of cold exposure, the wild type controls dropped body temperature by 7.4% (from 38.18 Ϯ 0.14°C to 35.36 Ϯ 0.07°C), whereas Bhmt Ϫ/Ϫ mice dropped body temperature by 28.9% (from 37.92 Ϯ 0.09°C to 26.98 Ϯ 2.15°C) (Fig. 3A). Mice were removed from the cold room when body temperature was less than 28°C. Brown adipose tissue (BAT) is a major site of nonshivering adaptive thermogenesis. We found no significant difference in the morphology or the weight of interscapular BAT between wild type and knock-out mice (data not shown). Real time PCR analysis revealed normal thermogenic gene responses in Bhmt Ϫ/Ϫ BAT tissue during cold exposure, with elevated Pgc1 (peroxisome proliferator activated receptor ␥ coactivator), Lpl (lipoprotein lipase), Ucp1 (uncoupling protein 1), and Mcpt1 (muscle-type carnitine palmitoyltransferase 1) expression levels similar to those of Bhmt ϩ/ϩ mice ( Fig. 3B), indicating that adrenergic signaling was intact and similar between genotypes. Bhmt ϩ/ϩ and Bhmt Ϫ/Ϫ mice had similar BAT TAG concentration before and after cold exposure, suggesting similar fuel storage and usage (Fig. 3C). Mice were also able to maintain body temperature using exogenous fuels derived from organs, such as white adipose tissue and liver via lipolysis and gluconeogenesis, and through adaptive thermogenesis involving shivering in skeletal muscle to increase energy output. Upon cold exposure, Bhmt Ϫ/Ϫ mice had significantly lower plasma glucose (by 46%, p Ͻ 0.01), glycerol (by 46%, p Ͻ 0.05), and NEFA (by 17%, p Ͻ 0.05) concentrations compared with those of wild type controls (Fig. 3, D-F), suggesting a restricted supply of exogenous fuels to maintain a normal body temperature. Bhmt Ϫ/Ϫ mice also had a significantly elevated plasma creatine kinase activity (by 1.76fold, p Ͻ 0.05), a marker for muscle breakdown, perhaps because of increased shivering (Fig. 3G).
BHMT Deficiency Resulted in Altered Biochemical Measurements in Plasma-We examined whether genotypic differences in energy balance were accompanied by changes in plasma bio-FIGURE 1. Bhmt ؊/؊ mice had reduced adiposity and better insulin and glucose sensitivities. A, body mass was measured in 7-and 14-week-old Bhmt ϩ/ϩ (black bar) and Bhmt Ϫ/Ϫ (white bar) mice using magnetic resonance images. *, p Ͻ 0.05, different from Bhmt ϩ/ϩ mice by Student's t test; n ϭ 12 and 6 per group for 7-and 14-week-old mice respectively. B, two major fat pads, GWAT and IWAT, were harvested from Bhmt ϩ/ϩ (black square) and Bhmt Ϫ/Ϫ (white square) mice at various ages. *, p Ͻ 0.01, different from Bhmt ϩ/ϩ of the same age by Student's t test; n ϭ 10 -16 per group for 5-12week-old mice and n ϭ 5 per group for 20 -48-week-old mice. C, morphology of GWAT and IWAT from 12-week-old mice was shown by hematoxylin-eosin staining. Scale bar, 50 m. D, adipocyte cell area of GWAT and IWAT from 7and 12-week-old Bhmt ϩ/ϩ (black bar) and Bhmt Ϫ/Ϫ (white bar) mice was analyzed using Zeiss AxioVision software. *, p Ͻ 0.05; **, p Ͻ 0.01, different from Bhmt ϩ/ϩ mice of the same age by Student's t test; n ϭ 3-5 per group. E, BHMT protein in liver and adipose tissue from Bhmt ϩ/ϩ and Bhmt Ϫ/Ϫ mice were probed by Western blot analysis. The size of BHMT protein is 45 kDa. F and G, glucose tolerance test (F) and insulin tolerance test (G) were performed in 7-week-old Bhmt ϩ/ϩ (black square) and Bhmt Ϫ/Ϫ (white square) mice via intraperitoneal injection of either glucose or insulin and measuring tail blood glucose at various time points. *, p Ͻ 0.05, different from Bhmt ϩ/ϩ by Student's t test; n ϭ 7 per group. Data are presented as mean Ϯ S.E. chemical measurements. Basal plasma glycerol, NEFA, and glucose concentrations were similar for the two genotypes (Fig. 3, D-F). Basal plasma insulin concentration was 50% lower in Bhmt Ϫ/Ϫ mice than in controls (p Ͻ 0.05) ( Table 2), suggesting a better insulin sensitivity. Plasma growth hormone and glucagon concentrations were not different between genotypes ( Table 2). Because of the prominent role of thyroid hormone in controlling metabolic rate, we investigated whether increased thyroid hormone concentration contributed to the hypermeta-bolic phenotype in Bhmt Ϫ/Ϫ mice. Indeed, Bhmt Ϫ/Ϫ mice had a 33% increase in plasma T 4 (the major form of thyroid hormone in blood) concentration compared with that of wild type controls (p Ͻ 0.05) ( Table 2). Plasma T 3 concentrations were not different between genotypes. Concentration of fibroblast growth factor 21 (FGF21), a recently discovered metabolic and glucose regulator, was found to be 1.75-fold higher in Bhmt Ϫ/Ϫ plasma than in Bhmt ϩ/ϩ plasma (p Ͻ 0.01) ( Table 2). The increase in plasma FGF21 concentration observed in Bhmt Ϫ/Ϫ FIGURE 2. Bhmt ؊/؊ mice had increased energy expenditure. A, food intake was measured in 7-week-old Bhmt ϩ/ϩ (black bar) and Bhmt Ϫ/Ϫ (white bar) mice using an indirect calorimetry; n ϭ 8 per group. B, fat tolerance test was performed in 7-week-old Bhmt ϩ/ϩ (black square) and Bhmt Ϫ/Ϫ (white square) mice via oral gavage of olive oil and measuring plasma TAG at various time points; n ϭ 3 per group. C, fecal fat from Bhmt ϩ/ϩ (black bar) and Bhmt Ϫ/Ϫ (white bar) mice was extracted and measured using a colorimetric assay; n ϭ 7 per group. D, rectal temperature of 7-week-old Bhmt ϩ/ϩ (black bar) and Bhmt Ϫ/Ϫ (white bar) mice was measured using a rectal probe thermometer at room temperature; n ϭ 6 -7 per group. E-H, oxygen consumption (E), carbon dioxide release (F), respiratory exchange ratio (CO 2 release/O 2 used) (G), and locomotor activity (H) were measured in 7-week-old Bhmt ϩ/ϩ (black) and Bhmt Ϫ/Ϫ (white) using an indirect calorimetry. Data were obtained after mice being acclimated to the chambers for 24 h and adjusted to lean body mass; n ϭ 8 per group. Statistical analysis was as shown in Table 1. Data are presented as mean Ϯ S.E. VO 2 , oxygen consumption; VCO 2 , carbon dioxide release.

TABLE 1 Indirect calorimetry data analysis
Oxygen consumption, carbon dioxide release, respiratory exchange ratio, and activity were measured in 7-week-old Bhmt ϩ/ϩ and Bhmt Ϫ/Ϫ mice using an indirect calorimetry; n ϭ 8 per group. Data for light and dark cycles were obtained after mice were acclimated to the chambers for 24 h. VO 2 and VCO 2 were expressed based on either the body weight or the lean body mass (LBM). Area under the curve (AUC) on each measure was analyzed followed by t test and Mann-Whitney test. The unit of the AUC is X min, where X is the unit for each measure. Data are presented as mean Ϯ S.E. VO 2 , oxygen consumption; VCO 2 , carbon dioxide release; MW test, Mann-Whitney test.  (16 -19). Hepatic gene expression analysis revealed that Bhmt Ϫ/Ϫ mouse liver did not have altered expressions of Ppar␣ and Chrebp but had increased expressions of Ppar␥ (by 3.5fold, p Ͻ 0.05) as compared with the controls (Table 2). BHMT Deficiency Resulted in Increased Bile Acids in Liver and Adipose Tissue-A novel concept indicating a signaling role of bile acids in the control of energy metabolism has emerged (20,21). Increasing bile acid pool size in mice increased energy expenditure (21), while decreasing bile acid pool size reduced energy expenditure, resulting in weight gain and insulin resistance (20). Because Bhmt Ϫ/Ϫ mice had reduced plasma cholesterol (the precursor of bile acids) (2), we investigated whether Bhmt deletion altered the concentrations of bile acids. Cholic acid and chenodeoxycholic acid are the primary bile acids and are often conjugated with taurine (as taurocholate or taurochenodeoxycholate) or glycine (as glycocholate or glycochenodeoxycholate) forming bile salts. Compared with wild type littermates, Bhmt deletion resulted in 9.50-and 3.41fold increases, respectively, in cholate (p Ͻ 0.01) and taurocholate (p Ͻ 0.05) in adipose tissue and in 2.33-and 1.62-fold increases, respectively, in taurocholate (p Ͻ 0.05) and glycocholate (p Ͻ 0.01) in liver (Table 3). Cholesterol is hydrolyzed to form 7-hydroxycholesterol, the precursor of bile acids. Compared with the controls, Bhmt Ϫ/Ϫ mice had reduced cholesterol (by 20%, p Ͻ 0.05) and increased 7-␤-hydroxycholesterol (by 2.98-fold, p Ͻ 0.05) concentrations in the liver and increased 7-␣-hydroxycholesterol (by 2.17-fold, p Ͻ 0.01) concentrations in the adipose tissue. These data suggest that Bhmt deletion enhanced the synthesis of bile acids from cholesterol and that the bile acids may contribute to the increased energy expenditure observed in Bhmt Ϫ/Ϫ mice.
Bhmt Ϫ/Ϫ Mice Did Not Have Altered Lipolytic Rate-To investigate whether the markedly decreased adiposity observed in Bhmt Ϫ/Ϫ mice was due to increased lipolysis, we measured lipolytic rates in both in vivo and ex vivo models where lipolytic stimuli were given, and glycerol released was measured as the indicator of lipolytic rate. Plasma TAG and glucose were also measured in the in vivo model to determine the whole body response to adrenergic stimuli. During the in vivo experiment, when CL316243 (␤ 3 -adrenergic specific lipolytic stimulus) was administered, Bhmt ϩ/ϩ and Bhmt Ϫ/Ϫ mice had similar levels of plasma TAG, glycerol, and glucose release (Fig. 4, A-C). When isoproterenol (nonspecific lipolytic stimulus) was administered, Bhmt ϩ/ϩ and Bhmt Ϫ/Ϫ mice had similar levels of plasma TAG and glycerol (Fig. 4, A and B) but had a 50% decreased plasma glucose concentration compared with wild type controls (p Ͻ 0.05) (Fig. 4C), indicating a deficiency in hepatic glucose storage or generation. Measurements of hepatic glucose and glycogen concentrations revealed that Bhmt Ϫ/Ϫ mice had reduced hepatic glucose (by 22%, p Ͻ 0.05) and glycogen (by 17%, p Ͻ 0.05) concentrations compared with those of Bhmt ϩ/ϩ mice ( Table 2). These data perhaps explain why Bhmt Ϫ/Ϫ mice had lower plasma glucose concentration during cold exposure and nonspecific lipolysis tests. Lipolysis was further analyzed in isolated GWAT and IWAT explants from Bhmt ϩ/ϩ and Bhmt Ϫ/Ϫ mice. The explants from Bhmt Ϫ/Ϫ mice released similar amounts of glycerol as did those from Bhmt ϩ/ϩ FIGURE 3. Bhmt ؊/؊ mice were cold-sensitive due to lack of exogenous fuels. Seven-week-old mice were exposed to room temperature or 4°C, and tissues collected after 6 h of exposure. A, body temperature of Bhmt ϩ/ϩ (black square) and Bhmt Ϫ/Ϫ (white square) mice exposed to 4°C was measured using a rectal probe thermometer at different time points. *, p Ͻ 0.05; **, p Ͻ 0.01, different from Bhmt ϩ/ϩ mice at the same time point by Student's t test; n ϭ 13-14 per group. B, expression of genes involved in thermogenesis in the BAT of Bhmt ϩ/ϩ (black bar) and Bhmt Ϫ/Ϫ (white bar) mice at room temperature (n ϭ 3-4 per group) or 4°C (n ϭ 7-8 per group) exposure. C, TAG in BAT from room temperature (n ϭ 3-6 per group) or 4°C (n ϭ 13-14 per group) exposed Bhmt ϩ/ϩ (black bar) and Bhmt Ϫ/Ϫ (white bar) mice were extracted and measured colorimetrically. D-G, plasma glucose (D), glycerol (E), NEFA (F), and creatine kinase (CK) (G) from room temperature (n ϭ 3-6 per group) or 4°C (n ϭ 13-14 per group) exposed Bhmt ϩ/ϩ (black bar) and Bhmt Ϫ/Ϫ (white bar) mice were measured colorimetrically. *, p Ͻ 0.05, different from Bhmt ϩ/ϩ mice under the same condition by Student's t test. Data are presented as mean Ϯ S.E. BAT, brown adipose tissue; NEFA, nonesterified fatty acid; CK, creatine kinase; PGC1␣, ppar␥ coactivator 1␣; LPL, lipoprotein lipase; UCP1, uncoupling protein 1; MCPT1, muscle carnitine palmitoyl transferase 1.

TABLE 2 Metabolites and gene expression in Bhmt ؉/؉ and Bhmt ؊/؊ plasma and liver
Plasma and liver were collected from 7-week-old Bhmt ϩ/ϩ and Bhmt Ϫ/Ϫ mice after 4 h of fasting unless otherwise indicated. Plasma metabolites were measured using commercially available kits; n ϭ 8 -17 per group. Gene expression was measured by real time PCR, and expressed relatively to the controls; n ϭ 6 -10 per group. Glucose and glycogen were measured using an acid hydrolysis method; n ϭ 6 per group. mice under basal or stimulated conditions (Fig. 4, D and E), suggesting normal lipolytic machinery in Bhmt Ϫ/Ϫ mice.
Betaine Supplementation Increased Hepatic Fgf21 Expression-Betaine is commonly used in the livestock and poultry industries to generate leaner meats (22), and FGF21 is a novel metabolic regulator (23). Bhmt ablation resulted in a significant accumulation of betaine in many tissues (2), and in increased hepatic Fgf21 expression as well as circulating FGF21 protein ( Table 2). To investigate the potential interaction between TABLE 3 Bile acids and sterols in Bhmt ؉/؉ and Bhmt ؊/؊ liver and adipose tissue Liver and gonadal adipose tissue were collected from 5-week-old Bhmt ϩ/ϩ and Bhmt Ϫ/Ϫ mice after 4 h of fasting; n ϭ 6 per group. The amount and the relative abundance of metabolites were measured and analyzed by liquid and gas chromatography by mass spectrometry. Data are presented as mean Ϯ S.E.

Metabolites
Bhmt  betaine and FGF21, wild type mice were fed either 0 or 5% betaine supplemented water for 2 weeks. A 5% betaine supplemented water was selected based on our experience, that this concentration did not alter food or water intake in mice (data not shown). Interestingly, at 2 weeks, 5% betaine water feeding increased Fgf21 expression by almost 7-fold in wild type liver (p Ͻ 0.05) (Fig. 6).

DISCUSSION
BHMT is one of the most abundant proteins in the liver, representing 0.6 -1.6% of the total protein (24). However, its known functions are limited to its roles in choline and onecarbon metabolism. Our current studies establish a unique role for BHMT in energy homeostasis.
Previously, we found that Bhmt Ϫ/Ϫ mice had reduced body weight from 5 to 9 weeks of age compared with wild type littermates, although the difference in body weight disappeared after 9 weeks (2). MRI and dissection of fat pads revealed that the reduced body weight in Bhmt Ϫ/Ϫ mice was due to lower fat mass. Bhmt Ϫ/Ϫ mice had smaller fat pad mass that persisted through at least 1 year of age compared with the controls. Bhmt Ϫ/Ϫ mice were able to maintain body weight in light of the reduced fat mass after 9 weeks of age perhaps because they had heavier livers that made up for the difference in fat mass weight (2). For example, at 12 weeks of age, Bhmt Ϫ/Ϫ mice had an average liver weight 409 mg heavier than that of Bhmt ϩ/ϩ mice (1.85 Ϯ 0.08 g in Bhmt Ϫ/Ϫ mice versus 1.44 Ϯ 0.06 g in Bhmt ϩ/ϩ mice), which made up for the difference in fat pad mass (511 Ϯ 39 mg in Bhmt Ϫ/Ϫ mice versus 815 Ϯ 74 mg in Bhmt ϩ/ϩ mice). BHMT is a liver-specific enzyme in rodents (1). Because BHMT protein was absent in normal adipose tissue, this observation indicated that the reduced fat mass in Bhmt Ϫ/Ϫ mice was due to an indirect effect of the lack of hepatic BHMT activity. How could lack of an hepatic enzyme result in reduced fat mass? One possibility is diminished lipid transport from the liver to adipose tissue via very low density lipoprotein (VLDL). Our previous work showed that Bhmt Ϫ/Ϫ mice had reduced hepatic PtdCho concentration, resulting in decreased VLDL secretion and increased hepatic fat accumulation (2). However, the amount of fat missing in fat pads exceeds the excess amount of fat accumulated in the liver. In addition, adipose should be able to store lipid obtained from the diet instead of depending solely on hepatic lipid sources. Thus, decreased VLDL secretion is unlikely to account for all the fat absence. Homeostasis of energy stores is regulated by mechanisms that include modulation of metabolic rate and thermogenesis, control of metabolite fluxes among various organs, and modulation of fuel synthesis and usage within adipose tissue. Examination of each of these parameters in Bhmt Ϫ/Ϫ mice showed that multiple factors might contribute to the reduced adiposity observed in this mouse model.
Bhmt Ϫ/Ϫ mice had normal food intake, fat clearance rate, and fecal fat content, indicating that the reduced adiposity was not due to reduced energy intake or fat malabsorption. Bhmt Ϫ/Ϫ mice had increased rates of O 2 consumption and CO 2 FIGURE 5. Bhmt ؊/؊ mouse adipocytes had altered TAG synthesis and glucose oxidation. A, mature adipocytes were isolated from Bhmt ϩ/ϩ (black bar) and Bhmt Ϫ/Ϫ (white bar) mice GWAT and IWAT and used to determine the rate of labeled fatty acid incorporated to triacylglycerol. **, p Ͻ 0.01, different from Bhmt ϩ/ϩ mice by Student's t test; n ϭ 4 -5 per group. B, mature adipocytes were isolated from Bhmt ϩ/ϩ and Bhmt Ϫ/Ϫ mice GWAT and IWAT and used to determine the rate of 14 C 18:1 oxidation: CO 2 (gray bar), ASM (white bar); n ϭ 10 -13 per group. C and D, mature adipocytes were isolated from Bhmt ϩ/ϩ (black bar) and Bhmt Ϫ/Ϫ (white bar) mice GWAT (C) and IWAT (D) and used to determine the rate of glucose oxidation with or without insulin. *, p Ͻ 0.05; **, p Ͻ 0.01 different from Bhmt ϩ/ϩ mice under the same treatment by Student's t test; n ϭ 7-10 per group without insulin, n ϭ 4 -7 per group with insulin. E and F, liver homogenates from Bhmt ϩ/ϩ (black bar) and Bhmt Ϫ/Ϫ (white bar) mice were used to determine the rate of 14 C 18:1 oxidation. *, p Ͻ 0.05, different from Bhmt ϩ/ϩ mice by Student's t test n ϭ 5-7 per group Data are presented as mean Ϯ S.E. FIGURE 6. Betaine supplementation increased hepatic Fgf21 expression. 7-Week-old Bhmt ϩ/ϩ mice were fed either a 0 or a 5% betaine supplemented water for 2 weeks. Livers were collected for the measurement of Fgf21 expression using real time PCR. Expression was expressed relatively to the controls; n ϭ 6 -8 per group. *, p Ͻ 0.05, different from 0% group by Student's t test. Data are presented as mean Ϯ S.E. release, suggesting that the reduced fat mass was partially due to increased energy utilization. Because locomotor activity did not differ between genotypes, the increased energy expenditure exhibited by Bhmt Ϫ/Ϫ mice was not due to increased physical activity. We investigated whether the reduced adiposity was due to increased thermogenesis. Upon cold exposure, Bhmt Ϫ/Ϫ mice failed to maintain their body temperature. Bhmt Ϫ/Ϫ mice exhibited normal induction of genes involved in mitochondrial biogenesis, uncoupling, and fatty acid oxidation in BAT. Furthermore, Bhmt Ϫ/Ϫ mice had normal BAT TAG storage and functional lipolytic machinery in peripheral tissues. However, during prolonged cold exposure, Bhmt Ϫ/Ϫ mice had lower plasma glucose, glycerol, and NEFA and higher plasma creatine kinase (marker of muscle breakdown). These data indicated that Bhmt Ϫ/Ϫ mice were not hyperthermogenic. In fact, these animals were cold-sensitive due to lack of exogenous fuel and had to depend on shivering thermogenesis in an attempt to maintain body temperature. The reduced plasma glucose concentration in Bhmt Ϫ/Ϫ mice during cold exposure was probably because they had less hepatic glucose stored. The reduced plasma glycerol and NEFA concentrations in Bhmt Ϫ/Ϫ mice during cold exposure were probably because they had less TAG stored in the adipose tissue to be released during energy needs. It was also possible that the Bhmt Ϫ/Ϫ mouse liver was taking up glycerol more efficiently than did Bhmt ϩ/ϩ for glucose synthesis to make up for the lack of hepatic glucose.
Within the adipose tissue, we observed that Bhmt Ϫ/Ϫ IWAT, but not GWAT, made 62% less TAG than did the controls. Although Bhmt Ϫ/Ϫ mice had normal lipolytic and FA oxidation machinery in both fat depots, the mice might have less TAG stored at a time when fuel was needed. We suspected that Bhmt Ϫ/Ϫ mice might utilize more glucose instead of FA as fuel. Indeed, we observed increased glucose oxidation in Bhmt Ϫ/Ϫ GWAT with insulin stimulus, and increased glucose oxidation in Bhmt Ϫ/Ϫ IWAT with or without insulin stimulus. This hypothesis was further supported by the enhanced glucose tolerance and insulin sensitivity observed in Bhmt Ϫ/Ϫ mice, the reduced plasma glucose during cold exposure, ␤-adrenergic stimulation, 16 h of fasting, and the reduced hepatic glucose storage. Enhanced glucose oxidation in adipose tissue and reduced hepatic glucose storage could potentially limit the availability of acetyl-CoA for forming fatty acids and the availability of the glycerol backbone for forming TAG, contributing to reduced adiposity. These data also suggested that the altered fuel metabolism in Bhmt Ϫ/Ϫ mice was not limited to adipose tissue but also occurred in the liver. Bhmt Ϫ/Ϫ mice had fatty liver and lower hepatic glucose concentrations, suggesting the possibility of FA utilization impairment. Indeed, we found Bhmt Ϫ/Ϫ mouse liver burned less FA as compared with Bhmt ϩ/ϩ mouse liver.
It is intriguing that the observed effects were more pronounced in IWAT as compared with GWAT in Bhmt Ϫ/Ϫ mice. The size of 12-week-old Bhmt Ϫ/Ϫ GWAT adipocytes was 57% that of controls, although IWAT adipocytes size was 20% the size of that in controls. Adipocytes from Bhmt Ϫ/Ϫ IWAT synthesized 62% less TAG and oxidized 2.4-fold more glucose than did control IWAT, although no changes were detected in adipocytes from GWAT. It is commonly accepted that visceral (GWAT) and subcutaneous (IWAT) fat pads are different. GWAT is considered to be the "pure WAT," whereas IWAT is a "convertible WAT" that may convert to BAT (31). Compared with visceral fat, subcutaneous fat cells express more thyroid hormone receptors (32) and mitochondrial Ucp1 (33,34), have more efficient insulin signaling and Glut4 (35), and are more responsive to Ppar␥ thiazolidinedione ligand (36,37). When subcutaneous fat was transplanted into the visceral cavity of mice, it reduced body weight and fat mass and enhanced insulin and glucose sensitivities (38), suggesting an intrinsic difference between the two fat pads. These differences in GWAT and IWAT may explain why effects in Bhmt Ϫ/Ϫ IWAT are more pronounced. Other investigators also report metabolic differences between GWAT and IWAT (39).
Bhmt Ϫ/Ϫ mice had reduced adiposity due to increased energy expenditure, enhanced glucose oxidation in WAT, and reduced TAG synthesis in WAT. The next question was how Bhmt deficiency could exert these differences, and we identified several candidate mechanisms that may be responsible. Coinciding with increased energy expenditure, Bhmt Ϫ/Ϫ mice had elevated plasma T 4 (the major form of thyroid hormone in the blood), a well known energy metabolism regulator. Plasma T 4 is converted to T 3 , the active form of thyroid hormone, with tissues. Unfortunately, we were unable to measure the conversion of T 4 to T 3 and T 3 concentration within tissues. Another potential candidate that may be responsible for these metabolic changes in Bhmt Ϫ/Ϫ mice is FGF21. Bhmt Ϫ/Ϫ mice overexpressed hepatic Fgf21 and had elevated plasma FGF21 concentration. FGF21 is a novel metabolic regulator that is preferentially expressed in the liver (23). FGF21 increases glucose uptake (40,41) and decreases intracellular TAG content in adipocytes (40); it lowers plasma glucose and TAG when administered to diabetic mice (41); and it increases energy expenditure and improves insulin sensitivity in diet-induced obese mice (42,43). Another potential candidate is bile acid. Recent studies suggested a role of bile acids in glucose and energy homeostasis. Mice fed a cholic acid (a bile acid)-supplemented diet had increased energy expenditure, preventing obesity and insulin resistance (21), although mice with reduced bile acids had decreased energy expenditure (20). Bhmt deletion resulted in increased cholate and taurocholate in adipose tissue and in increased taurocholate and glycocholate in liver. Bhmt Ϫ/Ϫ mice had reduced hepatic and serum cholesterol (2). This was perhaps due to an increased conversion of cholesterol to cholic acid. Indeed, the concentration of 7-hydroxycholesterol, the precursor to bile acid from cholesterol, was significantly elevated in both liver and adipose in Bhmt Ϫ/Ϫ mice.
Thyroid hormone, FGF21, and bile acids all induce similar metabolic changes and were all altered in Bhmt Ϫ/Ϫ mice. In fact, a growing body of evidence suggests an interaction among bile acids, thyroid hormone, and FGF21. Thyroid hormone induced cholesterol 7␣-hydroxylase (CYP7A1), the rate-limiting enzyme in bile acid synthesis from cholesterol (44,45), and there was an inverse relationship between thyroid hormone level and serum cholesterol (46). Thyroid hormone also induced hepatic Fgf21 expression in mice (47). Bile acids enhanced the conversion of T 4 to T 3 within tissues, resulting in increased energy expenditure in mice (21). Both bile acids and FGF21 concentrations were elevated in patients with nonalcoholic fatty liver (48). These studies suggest an interaction among these factors and Bhmt deficiency, resulting in reduced adiposity observed in Bhmt Ϫ/Ϫ mice.
Aside from the increased bile acids, thyroid hormone, and FGF21, deletion of hepatic Bhmt may cause reduced adiposity via changes in choline metabolites. As mentioned previously, BHMT is a major regulator of choline metabolism. Deletion of Bhmt resulted in altered choline metabolites, including a 21-and 2-fold increase in hepatic and adipose tissue betaine concentrations and an 82 and 53% decrease in hepatic and adipose tissue choline concentrations (2). Recently, studies suggested that manipulation of dietary choline or alteration of pathways in choline metabolism may influence energy expenditure and adiposity. Mice fed a methionine-and choline-deficient diet were hypermetabolic, lost weight (49), and had better insulin sensitivity and glucose tolerance (50). Mice with the Pemt gene (making PtdCho from phosphatidylethanolamine) deleted had normal metabolic parameters on a regular chow diet compared with wild type controls; however, when fed a high fat diet, they had increased oxygen consumption and RER, accumulated less body fat, and had increased glucose sensitivity (51). The lack of weight gain in Pemt Ϫ/Ϫ mice disappeared when mice were supplemented with choline. These observations suggest a potential role of choline in modulating energy metabolism. Similar to methionine-and choline-deficient diet fed mice and the Pemt Ϫ/Ϫ mice, Bhmt Ϫ/Ϫ mice had reduced choline and PtdCho concentrations, enhanced glucose and insulin sensitivities, increased energy expenditure, and reduced adiposity.
Bhmt ablation also resulted in a significant accumulation of betaine, which serves as a methyl donor and osmolyte. Betaine is commercially available as a feed additive; it is widely dis-cussed as a "carcass modifier" because of its lipotropic and growth-promoting effects, generating leaner meat (22). Dietary betaine supplementation resulted in reduced abdominal fat in poultry and reduced carcass fat (by 10 -18%) in pigs (22). The exact mechanism by which betaine modulates carcass quality is still unclear but has been suggested to involve interactions of betaine with growth factors and thyroid hormones (52-60). We did not observe increased growth hormone but did observe enhanced plasma T 4 and FGF21 concentrations and consequently hypermetabolism in Bhmt Ϫ/Ϫ mice. Betaine may also have a role as a glucose sensitizer. Betaine supplementation enhanced insulin sensitivity in diet-induced obese mice and increased insulin signaling pathway in isolated adipocytes (61). We found that 2 weeks of 5% betaine feeding increased Fgf21 expression by 7-fold in wild type liver. How betaine increases Fgf21 expression and whether FGF21 is responsible for the reduced adiposity phenotype in Bhmt Ϫ/Ϫ mice warrant more studies.
To summarize (Fig. 7), Bhmt deletion resulted in reduced adiposity, enhanced insulin sensitivity, and glucose tolerance. The reduction in fat mass in Bhmt-deficient mice could be attributed to increased whole body energy expenditure, decreased mobilization of lipids from the liver (VLDL) to be stored in adipose tissues, decreased TAG synthesis in white adipose tissue, and increased glucose oxidation in white adipose tissue. Deletion of Bhmt resulted in a variety of biochemical abnormalities, including elevated thyroid hormone, bile salts, FGF21 and betaine concentrations, and reduced choline and PtdCho concentrations. Many of these changes have been related to elevated energy metabolism, reduced adiposity, and enhanced glucose tolerance and insulin sensitivity. How these parameters interact with each other warrants further studies. This study suggests a role of Bhmt in regulating energy metabolism and adiposity. Overview of metabolic disturbances in Bhmt ؊/؊ mice. Increased energy expenditure, decreased lipid mobilization from the liver to adipose tissue, decreased TAG synthesis, and enhanced glucose oxidation together resulted in reduced fat mass in Bhmt Ϫ/Ϫ mice. Bhmt deletion increased plasma FGF21 and T 4 concentrations and increased betaine and bile acids concentrations while decreased choline concentration in both liver and adipose tissue. All of these changes had been associated with increased energy expenditure, enhanced insulin sensitivity, and reduced adiposity, and might be interrelated with one another. BHMT, betaine-homocysteine S-methyltransferase; Cho, choline; Bet, betaine; FGF21, fibroblast growth factor 21; Chol, cholesterol; 7 hydroxychol, 7-hydroxycholesterol; BAs, bile acids; Glu, glucose; Gly, glycogen; G-6-P, glucose 6-phosphate.