Deletion of Betaine-Homocysteine S-Methyltransferase in Mice Perturbs Choline and 1-Carbon Metabolism, Resulting in Fatty Liver and Hepatocellular Carcinomas*

Betaine-homocysteine S-methyltransferase (BHMT) uses betaine to catalyze the conversion of homocysteine (Hcy) to methionine. There are common genetic polymorphisms in the BHMT gene in humans that can alter its enzymatic activity. We generated the first Bhmt−/− mouse to model the functional effects of mutations that result in reduced BHMT activity. Deletion of Bhmt resulted in a 6-fold increase (p < 0.01) in hepatic and an 8-fold increase (p < 0.01) in plasma total Hcy concentrations. Deletion of Bhmt resulted in a 43% reduction in hepatic S-adenosylmethionine (AdoMet) (p < 0.01) and a 3-fold increase in hepatic S-adenosylhomocysteine (AdoHcy) (p < 0.01) concentrations, resulting in a 75% reduction in methylation potential (AdoMet:AdoHcy) (p < 0.01). Bhmt−/− mice accumulated betaine in most tissues, including a 21-fold increase in the liver concentration compared with wild type (WT) (p < 0.01). These mice had lower concentrations of choline, phosphocholine, glycerophosphocholine, phosphatidylcholine, and sphingomyelin in several tissues. At 5 weeks of age, Bhmt−/− mice had 36% lower total hepatic phospholipid concentrations and a 6-fold increase in hepatic triacyglycerol concentrations compared with WT (p < 0.01), which was due to a decrease in the secretion of very low density lipoproteins. At 1 year of age, 64% of Bhmt−/− mice had visible hepatic tumors. Histopathological analysis revealed that Bhmt−/− mice developed hepatocellular carcinoma or carcinoma precursors. These results indicate that BHMT has an important role in Hcy, choline, and one-carbon homeostasis. A lack of Bhmt also affects susceptibility to fatty liver and hepatocellular carcinoma. We suggest that functional polymorphisms in BHMT that significantly reduce activity may have similar effects in humans.

donor (supplied by methylenetetrahydrofolate reductase) (9). Alternatively, Hcy is removed by the trans-sulfuration pathway, forming cystathionine and cysteine catalyzed by the enzymes cystathionine-␤-synthase (CBS) and cystathionine-␥-lyase (C␥L), respectively (9). Both enzymes require vitamin B 6 ; CBS also requires AdoMet as the allosteric activator (9). All the above pathways might be perturbed with the deletion of Bhmt. Here, we report on the metabolic phenotype of the first Bhmt knock-out mouse created.

EXPERIMENTAL PROCEDURES
Generation of Bhmt Ϫ/Ϫ Mice-The gene targeting vector was designed to remove exons 6 and 7, the zinc-binding domains of the gene (Fig. 1A). The targeting vector was constructed by placing a MC1-neomycin (Neo) selectable marker flanked by 2 flippase (FLP) recognition target (FRT) sites and 1 locus of X-over P1 (loxP) site ϳ330 bp downstream of exon 7. A second loxP site was introduced ϳ240 bp upstream of exon 6. The vector was electroporated into E14Tg2A embryonic stem cells. Targeted cells were screened by PCR, confirmed by Southern blot analysis, and injected into blastocysts derived from mouse strain C57Bl/6 to create Bhmt transmitting chimeras. Neo deleted mice were generated by crossing the transmitting chimeras with Flpe deleters (number 003800, Jackson Laboratory, Bar Harbor, ME), which expressed Flpe recombinase that recognized FRT sites and deleted the Neo cassette. Bhmt knockout mice were generated by crossing the Neo deleted mice with Cre deleters (number 003724, Jackson Laboratory), which expressed Cre recombinase that recognized loxP sites and deleted exons 6 and 7. Bhmt knock-out mice were continuously backcrossed to C57Bl/6 mice. Generations F1 to F4 were used in this study.
Reproducibility, Viability, Body Weight, Body Length, and Life Span-Pairs of Bhmt ϩ/ϩ , Bhmt ϩ/Ϫ , and Bhmt Ϫ/Ϫ parents were bred to assess their reproductive potential. Sizes, genotype distribution, and gender distribution of litters were recorded. Body weights and lengths (nose to hip, nose to tail) of Bhmt ϩ/ϩ , Bhmt ϩ/Ϫ , and Bhmt Ϫ/Ϫ mice were measured from age 3.5 to 20 weeks old using a scale and a ruler, respectively. Eleven Bhmt ϩ/ϩ and 14 Bhmt Ϫ/Ϫ mice were maintained over the course of a year to determine 1-year survival rates.
Western Blot and Enzymatic Assay-Mice were anesthetized by inhalation of isoflurane (Hospira, Lake Forest, IL). Tissues were harvested, weighted, snap frozen, pulverized under liquid nitrogen, and stored at Ϫ80°C until used. Western blot analysis of BHMT was performed in liver homogenate as previously described (10). For all enzymatic assays, pulverized liver from 5-week-old mice were homogenized in buffer using a motorized tissue homogenizer (Talboys Engineering Corporation, Montrose, PA). BHMT activity was measured as the conversion of [ 14 C]betaine to [ 14 C]methionine and [ 14 C]dimethylglycine as previously described (11). CHDH activity was measured as the conversion of [ 14 C]choline to [ 14 C]betaine as previously described (12) using high pressure liquid chromatography (HPLC) on a Varian ProStar HPLC system (PS-210, Varian Inc., Palo Alto, CA) with a Pecosphere Silica column, 4.6 ϫ 83 mm (PerkinElmer Life Sciences) and a Berthold LB506 C-1 radiodetector (Oak Ridge, TN). GNMT activity was measured as the conversion of [ 3 H]AdoMet to [ 3 H]methylglycine using the charcoal absorption method as previously described (13). PEMT activity was measured as the conversion of [ 3 H]AdoMet to [ 3 H]PtdCho as previously described (14,15). Protein concentrations were measured using the Bradford Assay (16).
Choline and Homocysteine Metabolites-Tissues from 5-week-old animals were harvested as described above. Blood was collected via cardiac puncture, and plasma was isolated by centrifugation at 3,000 ϫ g for 5 min at room temperature. The concentrations of choline metabolites were measured by liquid chromatography-electrospray ionization-isotope dilution mass spectrometry as previously described (17). The concentration of AdoMet and S-adenosylhomocysteine (AdoHcy) were measured using the Varian ProStar HPLC system as previously described (18,19) with an Ultrasphere ODS 5-m C18 column, 4.6 ϫ 25 cm (number 235329, Fullerton, CA), and a UV-visible detector (model 118, Gilson, Middleton, WI). Plasma total Hcy (tHcy) and cysteine were measured after derivatizing 50 l of plasma with 7-fluorobenzofurazan-4-sulfonic acid. The derivatized products were measured as previously described (20) using the Varian ProStar HPLC system with a Microsorb-MV 5-m C18 column, 4.6 mm ϫ 25 cm (Varian) and a Prostar model 360 fluorescence detector (Varian). Mercaptopropinylglycine (10 M) was used as an internal standard. Hepatic tHcy, cystathionine, methionine, cysteine, methylglycine, dimethylglycine, and glycine were measured as previously described (21) using capillary stable isotope dilution gas chromatography/ mass spectrometry (Hewlett-Packard Co., Palo Alto, CA). Plasma total folate was measured using Lactobacillus casei assay as previously described (22).
Clinical Analysis-Concentrations of plasma urea nitrogen (BUN) and creatinine as well as activities of lactate dehydrogenase (LDH), alanine transaminase (ALT), and creatinine kinase (CK) were 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. Urine-specific gravity was measured using a refractometer (AO Instrument Co., Buffalo, NY) at the Department of Laboratory Animal Medicine Veterinary and Technical Services Facility, UNC, Chapel Hill. Plasma triacylglycerol (TAG), ␤-hydroxybutyrate (Stanbio, Boerme, TX), cholesterol, high density lipoprotein cholesterol (HDL-C), nonesterified fatty acids (NEFA), and glucose (Wako, Richmond, VA) were measured colorimetrically per the manufacturer's instructions.
Triacylglycerol, Phospholipids, and Cholesterol-PtdCho, PtdEtn, phosphatidylserine (PtdSer), and phosphatidylinositol (PtdIns) were extracted from the liver of 5-week-old mice using the Bligh and Dyer method (23), and isolated by thin layer chromatography (TLC) (24). Concentrations were determined using a phosphate assay (25). TAG was extracted from liver samples using the Folch method (26) and measured colorimetrically as described above.
In Vivo VLDL Secretion Rate-5-week-old animals were fasted for 4 h. 500 mg/kg body weight of Triton WR1339 (as a 7.5% solution dissolved in PBS) was injected retro-orbitally. Blood samples were taken using heparinized capillary tubes retro-orbitally at 2, 30, 60, and 120 min after Triton WR1339 injection for TAG measurement as an indicator of the secretion rate of the very low density lipoprotein (VLDL).
Tissue Histology-Liver was harvested and fixed in 4% paraformaldehyde, 0.2% gluteraldehyde for 48 h. Liver was processed, paraffin embedded, sectioned at 5 m, and stained with hemotoxylin and eosin as previously described (27). For tumor analysis, whole liver was sectioned serially, and multiple sections per sample were examined by 2 veterinary pathologists from the Department of Laboratory Animal Medicine, UNC, Chapel Hill. High resolution images were collected using a Zeiss Axiovision A1 upright microscope.
␥-Glutamyltransferase 1 (␥GT1)-␥GT1 levels were determined using ELISA (Cedarlane, Burlington, NC). Frozen liver samples from 1-year-old mice were homogenized in ice-cold PBS at pH 7.0 containing 10% protease inhibitors (Sigma) using an ultrasonic cell disruptor. The resulting suspension was centrifuged and protein levels were determined by the Lowry method (28). Equal amounts of proteins were loaded onto the ELISA plate according to the manufacturer's protocol. The optical density was read at 450 nm on a Synergy 2 microplate reader (Biotek, VT).
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 mean Ϯ S.E.

RESULTS
Generation and Validation of Bhmt Knock-out Mice-Bhmt Ϫ/Ϫ mice were generated using a cre/loxP system to delete exons 6 and 7, which contain the zinc-binding domain essential for BHMT activity. The mouse genotype was confirmed by PCR using genomic DNA (Fig. 1B). Bhmt Ϫ/Ϫ mice were also validated by assaying hepatic BHMT activity (Fig. 1C) and Western blot (Fig. 1D). Deletion of 1 allele of Bhmt resulted in a 46% reduction of activity, whereas deletion of both alleles resulted in the absence of activity. BHMT protein is predominantly found in mouse liver and was absent in Bhmt Ϫ/Ϫ liver.
Bhmt Ϫ/Ϫ Mice Had Reduced Body Weight at Early Age-Body length and weight of pups were monitored from 3.5 to 20 weeks of age. There was no significant difference in body length FIGURE 1. Confirmation of Bhmt ؊/؊ mice. A, Bhmt chimeric mice were generated using a gene targeting vector that removed exons 6 and 7 of the gene as described under "Experimental Procedures." loxP, locus of X-over P1; frt, flippase recognition target; neo, neomycin cassette; flp, flippase recombination enzyme; cre, cyclization recombination enzyme. B, PCR analysis of genomic DNA isolated from Bhmt ϩ/ϩ , Bhmt ϩ/Ϫ , and Bhmt Ϫ/Ϫ mice. The PCR product of the WT allele was 1600 bp, whereas the knock-out allele was 545 bp. C, hepatic BHMT activity was measured using a radiometric assay in Bhmt ϩ/ϩ (black bar), Bhmt ϩ/Ϫ (hatched bar), and Bhmt Ϫ/Ϫ (white bar) mice. Data are presented as mean Ϯ S.E., n ϭ 4 -8 per group. Different letters differ significantly (p Ͻ 0.01) by analysis of variance and Tukey-Kramer HSD tests. D, BHMT protein in liver from Bhmt ϩ/ϩ and Bhmt Ϫ/Ϫ mice were probed by Western blot analysis. The size of the BHMT protein is 45 kDa. E, body weights of Bhmt ϩ/ϩ (black square), Bhmt ϩ/Ϫ (gray square), and Bhmt Ϫ/Ϫ (white square) mice were measured using a scale from age 3.5 to 20 weeks old. Data are presented as mean Ϯ S.E., n ϭ 10 -25 per group. *, p Ͻ 0.05, different from Bhmt ϩ/ϩ by Student's t test.

TABLE 2 Deletion of Bhmt results in reduced methylation potential and increased homocyteine concentrations
Tissues were harvested from 5-week-old Bhmt ϩ/ϩ and Bhmt Ϫ/Ϫ mice. Hepatic homocysteine metabolites were quantified by capillary stable isotope dilution gas chromatography/mass spectrometry. Hepatic AdoMet, AdoHcy, and plasma total homocysteine and cysteine were quantified by high pressure liquid chromatography. Plasma total folate was quantified by a microbial (L. casei) assay. Data are presented as mean Ϯ S.E., n ϭ 5-6 per group.

Organ
Metabolites  cholesterol and HDL-C concentrations compared with Bhmt ϩ/ϩ mice.

Bhmt Deletion Resulted in Preneoplastic, Neoplastic, and Regenerative Lesions in the Mouse Liver-Eleven Bhmt ϩ/ϩ and 14
Bhmt Ϫ/Ϫ mice were maintained over the course of a year to determine 1-year survival rates. At time of sacrifice, 64% (9 of 14) of Bhmt Ϫ/Ϫ mice and 0% (0 of 11) of Bhmt ϩ/ϩ mice had visible hepatic tumors (Fig. 4A). Tumors observed in Bhmt Ϫ/Ϫ mice were solitary or multifocal, found in multiple liver lobes. The average liver weight of Bhmt Ϫ/Ϫ mice was heavier than that of Bhmt ϩ/ϩ mice (6.22 Ϯ 2.11% BW in Bhmt Ϫ/Ϫ versus 4.36 Ϯ 0.26% BW in Bhmt ϩ/ϩ ; p Ͻ 0.001), attributable to tumor mass. A subset of the liver samples (4 of 11 Bhmt ϩ/ϩ livers, 6 of 9 Bhmt Ϫ/Ϫ livers with tumors, and 4 of 5 Bhmt Ϫ/Ϫ livers without tumors) were serial-sectioned entirely, stained with H&E, and send to two veterinary pathologists for diagnosis. Microscopic analysis depicted a series of histological abnormalities from preneoplastic lesions (acidophilic/eosinophilic cell foci) to frank neoplasia visible only in the Bhmt Ϫ/Ϫ mice (Table 4 and   Fig. 4B). Hepatocellular adenomas were found in 30% (3 of 10) of the analyzed Bhmt Ϫ/Ϫ livers and none were found in the control animals. These benign lesions were represented by single or multiple nodules, sharply demarcated from the surrounding tissues, with a mild to moderate compression border and loss of lobular structure. Hepatocellular carcinomas (HCC) were found in 50% (5 of 10) of the analyzed Bhmt Ϫ/Ϫ livers. HCC were represented by nodules of atypical, megalohepatocytes that were vacuolated or contained hyaline droplets. Indicative of malignancy, these lesions were organized in cords that were 5-6 cells wide, had poor or nonexisting interface with the surrounding tissues, and compressed mildly the adjacent normal cords. The portal triads and sinusoids were poorly or nondistinguishable in most of the nodules. Formation of heptocellular adenomas and HCC in Bhmt Ϫ/Ϫ mice were accompanied by macro-and microvesicular steatosis, and stellate (Ito) cell and/or oval (bililary) cell hyperplasia (Table 4). We did not study Bhmt ϩ/Ϫ mouse survival or carcinogenesis. Bhmt Deletion Resulted in Elevated ␥ Glutamyltransferase 1 in Livers-␥GT1 is a highly sensitive marker of chronic liver disease, neoplasia, and HCC (29). Bhmt Ϫ/Ϫ mice had hepatic ␥GT1 concentrations 3.3-fold higher than that of the control mice (0.815 Ϯ 0.049 ng/mg of tissue in Bhmt Ϫ/Ϫ versus 0.245 Ϯ 0.085 ng/mg of tissue in Bhmt ϩ/ϩ ; p Ͻ 0.01).
Changes in Choline and One-carbon Metabolites at 5 Weeks Persisted at 1-Year-old Bhmt Ϫ/Ϫ Mice-Choline deficiency and reduced methylation potential have been hypothesized as mechanisms underlying development of HCC. Compared with wild type mice, Bhmt Ϫ/Ϫ mice had a 91% reduction in hepatic methylation potential (AdoMet:AdoHcy) compared with Bhmt ϩ/ϩ mice (p Ͻ 0.001) at 1 year (Table 5), which was greater than the 76% reduction seen at 5 weeks of age ( Table 2). The reduced methylation potential was due to reduced hepatic AdoMet (by 50%; p Ͻ 0.01) and to elevated AdoHcy (by 5.5fold; p Ͻ 0.001) concentrations in Bhmt Ϫ/Ϫ mice compared with Bhmt ϩ/ϩ mice. Compared with wild type littermates, Bhmt Ϫ/Ϫ mice had a 30-fold increase in hepatic betaine, a 73% reduction in GPCho, a 64% reduction in PCho, and an 18.8% reduction in PtdCho concentrations. At 1 year old, Bhmt Ϫ/Ϫ mice still had an elevated plasma tHcy concentration (by 6.9fold; p Ͻ 0.01), but no longer had decreased plasma cysteine concentration.

DISCUSSION
To our knowledge, this is the first report of the successful creation of Bhmt Ϫ/Ϫ mice. Deletion of the Bhmt gene resulted in the complete loss of BHMT activity in liver, the organ where nearly all of activity resides in mice. Heterozygous mice had 53% of the hepatic BHMT activity measured in Bhmt ϩ/ϩ , suggesting that the Bhmt gene is biallelically expressed. Mating pairs of Bhmt ϩ/ϩ , Bhmt ϩ/Ϫ , and Bhmt Ϫ/Ϫ produced offspring with a normal number of pups and evenly distributed genders, suggesting that BHMT is not essential for reproduction. This is interesting because mice in which choline dehydrogenase (Chdh; form betaine from choline, immediately upstream from Bhmt in choline pathway, Fig. 5) was deleted are infertile (30). Fetuses from heterozygous (Bhmt ϩ/Ϫ ) mating pairs were born at the expected Mendelian ratio; they were viable and survived to at least 1 year of age, suggesting that BHMT is not essential for embryonic development or for the viability of mice.
Although a Bhmt knock-out mouse has not been previously created, when S-(␣-carboxybutyl)-DL-homocysteine (CBHcy; a potent and specific inhibitor of BHMT) was administered intraperitoneally to wild type mice (31) or fed to rats (32), it caused an elevation of plasma Hcy concentrations and a reduction in the hepatic AdoMet:AdoHcy ratio. Our current study confirms the finding that inhibiting or ablating flux through BHMT affects one carbon metabolism. The changes are summarized in Fig. 5. As expected, Bhmt Ϫ/Ϫ mice had significantly higher con- TABLE 4 Pathology in livers from a subset of 1-year-old Bhmt ؉/؉ and Bhmt ؊/؊ mice A subset of 1-year-old Bhmt ϩ/ϩ (4 out of 11, none had visible tumors) and Bhmt Ϫ/Ϫ mouse liver (6 out of 9 that had visible tumors, 4 out of 5 that did not have visible tumors) were serial-sectioned, H&E stained, and send to 2 veterenary pathologists for diagnosis. Dots indicated the changes observed in a particular sample.

Genotype
Visible tumor HCA HCC Fat accumulation Cellular alteration Hyperplasia centrations of the substrates for this enzyme (betaine and Hcy) in various tissues as compared with Bhmt ϩ/ϩ mice. Our data show that the transsulfuration and methionine synthase pathways do not have the capacity to remove excess tHcy that accumulates when BHMT is not present, indicating that BHMT has a critical role in Hcy homeostasis. Bhmt Ϫ/Ϫ mice also had a significant reduction in hepatic dimethylglycine concentrations, one of the products of BHMT. Methionine, the other reaction product, was not affected by the gene deletion, probably because dietary methionine and protein stores of methionine are so large that the loss of BHMT-catalyzed biosynthesis does not alter overall concentrations of this amino acid. However, AdoMet concentration, which is formed from methionine, was reduced in livers of Bhmt Ϫ/Ϫ mice, whereas AdoHcy concentrations were elevated. AdoMet is used by GNMT to form methylglycine from glycine (33) (Fig. 5). Methylglycine is also formed from dimethylglycine, the product of the BHMT catalyzed reaction. Because Bhmt Ϫ/Ϫ mice do not accumulate dimethylglycine, the elevated methylglycine observed in Bhmt Ϫ/Ϫ mice likely comes from the action of GNMT. GNMT is allosterically inhibited by methyl-THF (13), which serves as the methyl donor for methionine synthase, a parallel pathway to BHMT that catalyzes the conversion of Hcy to methionine (9). It is possible that the Bhmt Ϫ/Ϫ mice rely heavily on the use of methyl-THF to form the needed methionine. Hence, GNMT is released from methyl-THF inhibition, resulting in increased methylglycine, reduced AdoMet and increased AdoHcy concentrations that we observed. This hypothesis is supported by elevated hepatic GNMT activity in Bhmt Ϫ/Ϫ mice (Fig. 2), as well as the reduction in plasma concentrations of total folate (the predominant form contributing to plasma total folate is methyl-THF) ( Table 2). AdoHcy is cleaved by AdoHcy hydrolase, producing Hcy and adenosine. This reaction is reversible and thermodynamically favors the synthesis of AdoHcy (34); this should occur in the presence of the high tHcy concentrations present in the Bhmt Ϫ/Ϫ mice.
Despite high plasma tHcy concentrations, Bhmt Ϫ/Ϫ mice had significantly lower plasma cysteine concentrations; we observed no differences in hepatic cysteine concentrations. Normally, high Hcy would be expected to result in increased cystathionine and cysteine concentrations. For example, methylenetetrahydrofolate reductase-deficient mice had elevated plasma Hcy, accompanied by elevated plasma cysteine (35). Supplemental betaine has been shown to increase the activity of liver CBS (36), thus the accumulation of this metabolite as well as tHcy in Bhmt Ϫ/Ϫ mice would be expected to increase concentrations of cysteine. However, AdoMet is an allosteric activator of CBS (37), and we suggest that reduced hepatic AdoMet concentrations inhibit the transsulfuration pathway. The flux of cysteine from liver to plasma must be rapid, perhaps explaining why plasma and not hepatic cysteine concentrations are altered.
Deletion of Bhmt results in a substantial accumulation of betaine in most tissues except testis. Betaine is either used as a methyl donor (which is blocked by Bhmt deletion) or as an osmolyte that regulates cellular volume and fluid balance. The accumulation of betaine in Bhmt Ϫ/Ϫ mice may affect osmoregulation. Tests of kidney function did not reveal any malfunction (Table 3). However, Bhmt Ϫ/Ϫ mice had reduced urinary specific gravity compared with that of Bhmt ϩ/ϩ mice (Table 3), suggesting reduced reabsorption of water by the kidney tubule. This finding is puzzling because high betaine should result in increased reabsorption of water and increased urine osmolality.
Surprisingly, the deletion of Bhmt does not result in accumulation of choline, the precursor of betaine. Choline is either oxidized to form betaine by the enzyme CHDH or it is used to form PtdCho. Because Bhmt Ϫ/Ϫ mice cannot utilize betaine, we expected to see product inhibition of CHDH resulting in increased choline availability for use in PtdCho synthesis. We observed the opposite, with lower choline and PtdCho concentrations in a number of tissues. Bhmt Ϫ/Ϫ mice had elevated hepatic expression of Chdh (data not shown) and hepatic CHDH activity (Fig. 2), suggesting that this gene is induced by some factor in Bhmt Ϫ/Ϫ mice, possibly explaining increased use of choline to form betaine. The regulatory mechanisms for such an effect are not known. At 5 weeks of age, Bhmt Ϫ/Ϫ mice developed a fatty liver, and had elevated plasma ALT activity (a measure of hepatic damage). The fatty liver is the result of decreased synthesis of PtdCho, which is essential for the secretion of VLDL from liver (38). These findings are consistent with the literature. Rodents and humans fed a choline-deficient diet develop fatty liver (39,40), and have elevated plasma ALT activity because choline deficiency induces hepatocytes apoptosis (41). Consistent with these observations in Bhmt Ϫ/Ϫ mice, others have reported administration of betaine or the induction of BHMT activity prevents fatty liver in Mthfr Ϫ/Ϫ mice (35), in nonalcoholic fatty liver (42), and in alcohol-induced fatty livers (43,44) in rodents. Together, these data suggest the importance of BHMT activity in modulating hepatic steatosis. In addition, Bhmt Ϫ/Ϫ mice had reduced plasma total cholesterol and HDL-C. The role of Bhmt deletion in cholesterol and lipoprotein homeostasis warrants further studies.
At 1 year of age, 64% of Bhmt Ϫ/Ϫ mice developed hepatic tumors. Histopathology confirmed that these tumors were either hepatocellular adenomas or carcinomas. Even those that did not have visible tumors had a higher incidence of fat accumulation, cellular alteration, and hyperplasia when compared with Bhmt ϩ/ϩ livers. Histological modifications were accompanied by elevation in ␥GT1 levels, a highly sensitive marker of HCC. In rodents, choline deficiency results in a higher incidence of spontaneous hepatocellular carcinomas (45). Several mechanisms are suggested for these carcinogenic effects. Choline deficiency increases lipid peroxidation in the liver (46), which is a source of free radicals in the nucleus that could modify DNA and cause carcinogenesis. Choline deficiency perturbs protein kinase C signaling, resulting in altered cell proliferation signals and cell apoptosis and eventually in carcinogenesis (40). Methyl deficiency may also mediate the mechanisms that underlie the etiology of cancers. AdoMet levels control cell growth, and a persistent decrease in AdoMet concentrations leads to the development of HCC as observed in the methionine adenosyltranferase 1A knock-out mouse model (cannot make AdoMet from Met) (48). Methyl deficiency may also alter epigenetic machinery by hypomethylating some genes but paradoxically hypermethylating specific genes (tumor suppressor genes) and consequently increasing the recruitment of methylbinding proteins to the CpG islands. Consequently, the expression of tumor suppressor genes responsible for DNA repair (BRCA1, hMLH1), cell cycle regulation (p15, p16), carcinogenesis metabolism (GSTP1), hormonal response (RAR␤2), apoptosis (DAPK, APAF-1), and cell adherence (CDH1, CDH3) (49 -51) are silenced, leading to tumor development. All the above mechanisms are potentially the underlying causes of HCC observed in Bhmt Ϫ/Ϫ mice because these mice are essentially choline and methyl-group deficient. We plan to investigate each of these hypotheses in ongoing work in our laboratory.
Hepatocellular carcinoma is the fifth most common type of cancer worldwide (52). In mouse models harboring genetic defects, HCC occurs spontaneously in a very limited number of cases (47). For the first time, we present evidence that Bhmt gene dysfunction links a metabolic defect to spontaneous neoplastic transformation in rodent livers. Bhmt Ϫ/Ϫ mice model marks the upper boundary for an effect of a lost-of-function gene variant in humans. Our studies in the Bhmt Ϫ/Ϫ mice suggest that humans with lost-of-function mutation in the BHMT gene may have important consequences, including hyperhomocysteinemia, altered choline metabolites, fatty liver, and hepatocellular carcinomas. It would be interesting to explore whether enhanced BHMT activity would reverse these symptoms.