Purification, Kinetic Properties, and cDNA Cloning of Mammalian Betaine-Homocysteine Methyltransferase*

Porcine liver betaine-homocysteine methyltrans- ferase (BHMT; EC 2.1.1.5) was purified to homogeneity, and the Michaelis constants for betaine, dimethylaceto- thetin, and L -homocysteine are 23, 155, and 32 (cid:109) M , respectively. The maximum rate of catalysis is 47-fold greater using dimethylacetothetin as a methyl donor compared with betaine. Partial amino acid sequence of porcine BHMT was obtained, and inosine-containing redundant oligonucleotide primers were used to amplify an 815-base pair sequence of the porcine cDNA by polymerase chain reaction (PCR). Nondegenerate oligonu- cleotide primers based on the porcine cDNA were synthesized and used to isolate a 463-base pair fragment of the human cDNA by PCR. The human PCR DNA product was then used to screen a cDNA library by plaque hy- bridization, and cDNAs encoding human BHMT were isolated. The primary structure of the human cDNA is reported here, and the open reading frame encodes a 406-residue protein of M r 44,969. The deduced amino acid sequence of human BHMT shows limited homology to bacterial vitamin B 12 -dependent methionine syn- thases (EC 2.1.1.13). A plasmid containing the human BHMT cDNA fused in frame to the N terminus of (cid:98) -ga-lactosidase was transformed into Escherichia coli , and transformants expressed BHMT activity, an activity that is absent from wild type E. coli .

There is considerable interest in the regulation of homocysteine (Hcy) 1 metabolism since even moderate elevations in plasma total homocyst(e)ine (pHcy) have been established as an independent risk factor for the development of arteriosclerotic vascular disease (1). The enzyme betaine-homocysteine methyltransferase (BHMT; EC 2.1.1.5) may have an important role in the modulation of pHcy. BHMT is a cytosolic enzyme that catalyzes the conversion of betaine and Hcy to dimethylglycine and methionine, respectively (2). This reaction also is required for the irreversible oxidation of choline. The only other enzyme known to methylate Hcy in mammalian cells is the folate/vitamin B 12 -dependent methionine synthase (MS; EC 2.1. 1.13).
Large oral doses of betaine have been shown to be an effective treatment for homocystinuria due to deficiencies of cystathionine ␤-synthase (EC 4.2.1.22) and methylenetetrahydrofolate reductase (EC 1.7.99.5) and inborn errors of cobalamin metabolism (3,4). The level of pHcy in these individuals decreases significantly with betaine treatment, and the incidence of thromboembolism is significantly reduced. The efficacy of betaine treatment is due, at least in part, to increased methylation of Hcy by the BHMT-catalyzed reaction. Betaine treatment, however, usually does not lower pHcy levels to within the normal range, and the moderately elevated levels that remain are highly correlated with vascular disease. BHMT is therefore a target for the treatment of homocystinuria in that better methyl donors may have more potent pHcy-lowering effects. Furthermore, BHMT activity has been shown to vary with dietary and hormonal treatments in rats (5-7) and dietary treatments in chickens. 2 Manipulations of these parameters could improve current therapies for some forms of homocystinuria.
A genetic defect in BHMT could result in hyperhomocyst-(e)inemia, and perhaps homocystinuria, since it has been reported that this enzyme is responsible for up to 50% of the Hcy methylation capacity in liver (8). This possibility has been difficult to determine, since fibroblasts and lymphocytes, tissues normally assayed to detect enzyme defects, do not express BHMT activity (9). The lack of a cDNA encoding BHMT has precluded investigations into the molecular mechanisms responsible for dietary and hormonally induced changes in BHMT activity. A human cDNA encoding BHMT would aid in the search for defects in its gene.
This report describes the purification and kinetic constants of porcine liver BHMT as well as the molecular cloning of a partial porcine cDNA and a near full-length human cDNA encoding this enzyme. A fusion construct of the human cDNA results in BHMT activity in Escherichia coli extracts, an activity absent from wild type bacteria. A comparison of the deduced amino acid sequence of human BHMT with that of E. coli MS indicates regions that may be involved in Hcy binding and catalysis of methyl transfer.

EXPERIMENTAL PROCEDURES
Materials-[methyl- 14 C]Choline chloride (55 mCi/mmol; 200 Ci/ml), [␣-32 P]dCTP (3000 Ci/mmol; 10 mCi/ml), and random primer DNA labeling kit were obtained from Amersham Corp. Betaine hydrochlo-* This work was supported by the Illinois Agricultural Experiment Station (60-0305), University of Illinois Research Board, American Heart Association, Illinois Affiliate (GB-10 gar), and the Future Leader Award from the International Life Science Institute, North America.The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Preliminary accounts of some of these studies were presented at Experimental Biology in Georgia (1995 Bacteria, Bacteriophage, and Plasmids-E. coli strains XL1-Blue MRFЈ and SOLR were obtained from Stratagene, and strain C600Hfl was purchased from Clontech. Porcine and human liver cDNA libraries were purchased from Clontech (gt10) and Stratagene ( ZAP XR), respectively. Libraries were titered according to manufacturers' protocols. PCR fragments were ligated into plasmid pCRII and transformed into E. coli strain INV␣FЈ (Invitrogen). Human cDNAs were isolated in pBluescript SK(Ϫ) using the automatic subcloning feature of ZAP.
Synthesis of Substrates-[methyl-14 C]betaine was prepared from [methyl-14 C]choline using choline oxidase, and labeled betaine was separated from residual choline and betaine aldehyde by chromatography on Amberlite CG-50. These procedures have been previously described (10,11).
Dimethylacetothetin (DMAT) chloride was synthesized by reacting equimolar amounts of methylsulfide and chloroacetic acid by the method of Maw (12). The solid product was recrystallized several times from ethanol-ether. Radiolabeled [ 14 C]DMAT chloride was prepared by reacting 250 mol of methylsulfide with equimolar [1-14 C]iodoacetic acid (50 Ci) in a minimum volume of ethanol. The reaction was carried out in a capped vial at 37°C for 36 h. The product mixture was diluted with 70% ethanol and applied to a Dowex 50W-X8 (Hϩ; 100 -200 mesh) column (0.5 ϫ 10 cm) equilibrated with 70% ethanol. The column was washed with 70% ethanol followed by water alone. DMAT chloride was eluted by a linear gradient (60 ml) of aqueous sodium chloride (100 -750 mM).
Labeled substrates were analyzed by two-dimensional thin layer chromatography (Silica Gel G) using the solvent system described by Speed and Richardson (11) for the metabolites of choline oxidation. Subsequent autoradiography displayed one spot for each substrate, verifying radiopurity.
BHMT Assay-BHMT activity was measured as described by Finkelstein and Mudd (13) with several modifications. DL-Hcy was prepared fresh by a procedure described by Duerre and Miller (14). Hcy thiolactone hydrochloride (15.4 mg) was dissolved in 400 l of 2 N sodium hydroxide. The solution was allowed to sit for 5 min at room temperature (23°C). The reaction was then neutralized by the addition of 600 l of a saturated solution of monopotassium phosphate and used immediately in the BHMT assay. The standard assay contained 5 mM DL-Hcy, 2 mM betaine (0.05-0.1 Ci), and 50 mM Tris-HCl (pH 7.5). The final reaction volume was 0.5 ml. Reaction tubes were capped with rubber stoppers and kept in ice water until the assay was started by transferring the tubes into a 37°C water bath. Following a 1-2-h incubation, samples were chilled in ice water, and 2.5 ml of ice-cold water was added. The samples were applied to a Dowex 1-X4 (OH-; 200 -400 mesh) column (0.6 ϫ 3.75 cm), and the unreacted betaine (or DMAT) was washed from the column with water (3 ϫ 5 ml). Dimethylglycine (or methylthioacetate) and methionine were eluted into scintillation vials with 3 ml of 1.5 N HCl. Seventeen milliliters of scintillation mixture (Ecolume, ICN) were added to each vial and counted. Blanks contained all of the reaction components except enzyme, and their values were subtracted from the sample values. All samples were assayed in duplicate.
Kinetic assays used L-Hcy instead of DL-Hcy, and substrate concentrations were varied when appropriate using fixed levels of either betaine (250 M) or L-Hcy (500 M). The kinetic assays also contained 250 M dithiothreitol and 50 mg/ml bovine serum albumin (fraction V). The ionic strength of each reaction was held constant and equal to the standard assay. The Michaelis and V max constants were estimated by plotting initial rate data according to the method of Hanes (15).
Purification of Porcine Liver BHMT-A typical purification can be seen in Table I. Livers from freshly slaughtered pigs were immediately placed on ice and transported to the laboratory, cut into pieces, and frozen at Ϫ80°C. Procedures completed prior to the heat treatment of the crude extract (described below) were carried out at 0°C, while subsequent steps were conducted at room temperature. All buffers were adjusted to the indicated pH at room temperature.
Liver (10 g) was partially thawed, minced with scissors, and suspended in 30 mM potassium phosphate buffer, pH 7.6 (30 ml), containing 10 mM 2-mercaptoethanol and 2 mM EDTA. The suspension was homogenized for 20 s using a Polytron homogenizer. The mixture was then centrifuged at 18,000 ϫ g for 1 h. The supernatant was decanted through a double layer of cheesecloth to give the crude extract.
The crude extract (25 ml) was placed in a 125-ml Erlenmeyer flask and heated at 70°C for 10 min with constant swirling. Precipitated protein and cellular debris were removed by centrifugation at 15,000 ϫ g for 20 min. The supernatant was decanted to give fraction 2 enzyme. Fraction 2 (17 ml) was applied to a hydroxyapatite column (2.5 ϫ 2.5 cm) equilibrated with 30 mM potassium phosphate buffer, pH 7.6, containing 10 mM 2-mercaptoethanol and 2 mM EDTA. The column was washed with 50 ml of 45 mM potassium phosphate buffer, pH 7.6, containing 10 mM 2-mercaptoethanol and 2 mM EDTA. BHMT was step eluted from the column with 50 ml of 90 mM potassium phosphate buffer, pH 7.6, containing 5 mM 2-mercaptoethanol and 1 mM EDTA. The eluent was dialyzed against two changes (4 liters) of 10 mM Tris-HCl buffer, pH 8.7, containing 5 mM 2-mercaptoethanol and 500 M EDTA for 12 h to give fraction 3.
Fraction 3 enzyme was applied to a DEAE-cellulose column (1 ϫ 20 cm) equilibrated with 10 mM Tris-HCl buffer, pH 8.7, containing 5 mM 2-mercaptoethanol and 500 M EDTA. The column was washed with the equilibration buffer (100 ml), and enzyme was eluted with a linear gradient (200 ml) of KCl (0-500 mM) in the same buffer. Fractions containing BHMT activity were pooled to give fraction 4.
The protein concentration of each fraction was measured by the procedure of Bradford (16) using bovine serum albumin as standard. BHMT purity was estimated by polyacrylamide gel electrophoresis in sodium dodecyl sulfate using a 5% stacking gel and a 12% separation gel. The discontinuous buffer system of Laemmli (17) was used in a miniprotein II slab gel apparatus (Bio-Rad). Protein was visualized by a Coomassie staining procedure.
N-terminal Sequencing of Porcine Liver BHMT and Derived Peptides-BHMT (4 g) was subjected to sodium dodecyl sulfate polyacrylamide gel electrophoresis as described above. The protein was blotted onto a polyvinylidene diflouride membrane using a Tris-glycine transfer buffer in a mini-trans-blott cell (Bio-Rad) following the manufacturer's suggested protocol. The protein was visualized by Coomassie staining, and the band was excised and stored at Ϫ20°C until sequenced. Amino acid sequence was determined using an Applied Biosystems 477A protein sequencer at the Biotechnology Center of University of Illinois, Urbana.
The following procedure was used to generate peptide fragments of porcine BHMT. An aliquot of protein (500 g) was dialyzed against 5 mM ammonium bicarbonate (1 liter) and lyophilized to dryness. The sample was resuspended in 200 l of 10% acetic acid and heated at 50°C for 72 h. The sample was neutralized with triethylamine acetate and lyophilized. A portion of the sample (25 g) was dissolved in sample buffer, and peptides were separated by sodium dodecyl sulfate polyacrylamide gel electrophoresis using a 4% stacking gel and a 15% separation gel. The Tris-Tricine buffer system of Schagger and von Jagow was used (18). The peptides were transferred onto a polyvinylidene diflouride membrane and visualized by Coomassie staining. The bands were excised and stored at Ϫ20°C until sequenced as described above.
Isolation of cDNAs-The following redundant inosine-containing oligonucleotide primers were designed from the N-terminal and internal amino acid sequence of porcine liver BHMT, respectively: 5Ј-GCIC- . These primers were used to isolate a portion of the BHMT-encoding porcine cDNA by PCR. The final concentrations of the PCR reaction mixture were as follows: 10 mM Tris-HCl (pH 8.3), 50 mM potassium chloride, 0.5 M each primer, 200 M each dNTP, 2.5 mM magnesium chloride, and 12.5 ϫ 10 6 gt10 bacteriophage containing porcine liver cDNA. Three units of Taq polymerase (Perkin-Elmer) was used in a final volume of 100 l. The first cycle used a 5-min denaturation step (97°C), a 1-min annealing step (50°C), and a 2-min elongation step (72°C). This sequence was followed by 28 cycles of 94°C, 50°C, and 72°C, for the denaturation, annealing, and elongation steps (1 min each), respectively. A final cycle used an elongation step of 10 min to complete any unfinished products. A single PCR product of about 800 base pairs was observed on a 1% agarose gel and ligated into pCRII (Invitrogen). Both strands of the insert were sequenced.
The following nonredundant oligonucleotide primers were synthesized based on nucleotides 205-227 and 646 -667 of the pig cDNA (see Fig. 2), respectively: 5Ј-GTCATGCAGACCTTCACCTTCTA-3Ј (sense) and 5Ј-TAATTGTGGGGTCAAAATGGCA-3Ј (antisense). These primers were used to isolate a portion of the BHMT-encoding human cDNA by PCR. The template for PCR was a human cDNA library in ZAP XR (7 ϫ 10 6 clones). PCR reactions were carried out as described above except that the annealing temperature was 55°C and there were a total of 35 cycles performed. One PCR product of approximately 500 base pairs was observed on a 1.2% agarose gel. This fragment was ligated into pCRII, and both strands of the insert were sequenced.
The human liver cDNA library (ZAP XR) was then plated, and the bacteriophage plaques were lifted onto nylon membranes (DuPont, Hybond-N). The membranes were screened as described by the manufacturer's protocol provided with the library. The hybridization probe was made from the human PCR fragment (described above) that was labeled with [␣-32 P]dCTP by a random primer extension procedure. A total of 2,200 plaques were screened.
Primers were synthesized at the Biotechnology Center of the University of Illinois, Urbana. DNA sequencing was performed using an Applied Biosystems 373A automated DNA sequencer at the same facility.

Purification and Kinetic Constants of Porcine Liver BHMT-
BHMT was purified by heat-treating a crude extract followed by chromatography on hydroxyapatite and DEAE-cellulose (Table I). This procedure results in a preparation that displays a single band of M r 45,000 on a sodium dodecyl sulfate-containing polyacrylamide gel (Fig. 1). The V max of porcine liver BHMT using the standard assay is very low (4 mol h Ϫ1 mg Ϫ1 ).
The kinetic constants of betaine, DMAT, and L-Hcy for the porcine liver enzyme were determined from initial rate data. The Michaelis constants were 23, 155, and 32 M for betaine, DMAT, and L-Hcy, respectively. BHMT displayed a 47-fold increase in V max and 7-fold increase in V max /K m when DMAT was the methyl donor relative to when betaine was used (Table  II). The V max of BHMT using the standard assay (4160 units/ mg) was about 2-fold higher than that obtained using the kinetic assay, whether L-Hcy (2036 units/mg) or betaine (1997 units/mg) was the varied substrate. The highest concentration of variable substrate used in all kinetic analyses ended up being slightly over 5 times the final K m determination. As with the rat enzyme (19), no indication of substrate inhibition was observed, and reducing agents had no effect on V max . The possibility of substrate activation at levels higher than 5 times the K m value reported here needs to be investigated.
Preliminary kinetic analysis indicated that the K m for DL-Hcy was exactly twice that of L-Hcy, suggesting that D-Hcy is not a substrate and has very low affinity for BHMT. Activity was then measured using 100 M D-Hcy and/or 100 M L-Hcy. The level of betaine was saturating (2 mM), and the ionic strength was held constant to the standard assay. The results indicated that porcine BHMT can methylate D-Hcy at about 15% the rate observed for L-Hcy. When BHMT was assayed with 100 M of both D-Hcy and L-Hcy, the rate of methionine production was 95% of that observed with L-Hcy alone. These data indicate that D-Hcy has very low affinity for the enzyme and that L-Hcy is preferentially methylated when a mixture of the enantiomers are used (not shown).
Porcine liver BHMT is inhibited by dimethylglycine and the demethylated product of DMAT, methylthioacetate, but not sarcosine (Table IV).
Amino Acid Sequencing of Porcine Liver BHMT-The Nterminal amino acid sequence of porcine BHMT was determined to be APVGDKKAKKGILERLNSGEV. Treatment with dilute acid (partially) cleaved the protein into five major and several minor peptides that could be separated on a polyacrylamide gel containing sodium dodecyl sulfate (not shown). This procedure has been reported to cleave at Asp-Pro sequences. Two peptides were sequenced and were subsequently discovered to be partially overlapping: XGKQGFIDLP and LPEFPF-GLEPRVATR. These porcine BHMT peptide sequences were found in the deduced amino acid sequence of the porcine cDNA (Fig. 2), and nearly identical sequences are present in the human enzyme (residues 2-22 and 257-273; Fig. 3). The first internal peptide sequence was preceded by Pro-Asp in the deduced amino acid sequences of both porcine and human cDNAs. The second internal peptide was preceded by an Asp residue that was part of an Asp-Leu-Pro sequence, also in the deduced sequences of both porcine and human cDNAs.
Nucleotide and Deduced Amino Acid Sequence of BHMT cDNAs-Redundant inosine-containing oligonucleotide primers, based on porcine amino acid sequence, were used to amplify an 815-base pair cDNA fragment encoding a portion of porcine BHMT by PCR (Fig. 3). The deduced amino acid sequence of this cDNA fragment contained all the amino acid residues that were identified from peptide sequencing. The   primers annealed to regions of the porcine cDNA corresponding to nucleotides 1-23 and 790 -815. Two nonredundant oligonucleotide primers, whose sequences were based on the porcine cDNA (Fig. 2), were used to screen a human cDNA library by PCR. These primers amplified a 463-base pair fragment of the human cDNA (Fig. 3). Two mismatches were identified in the human PCR fragment and the antisense (porcine-based) primer used in the reaction. These differences were still present in the near full-length human cDNA obtained from the library screen (see below), indicating species codon differences.
The human PCR fragment was 32 P-labeled by a random priming procedure and used to screen 2,200 bacteriophage plaques containing human cDNA. Following probe hybridization and autoradiography, 12 positive plaques were identified, and the phage was isolated (TG1-TG12). The cDNAcontaining plasmids were rescued from each phage using the automatic subcloning features of the ZAP system. All of the cDNA inserts had similar restriction maps, and the plasmid containing the largest insert (pTG9) was chosen for DNA sequencing.
The cDNA and deduced amino acid sequence of human BHMT (pTG9) is shown in Fig. 3. The insert has a poly(A) tail, but a consensus poly(A) signal sequence (AAUAAA) is not present. A similar sequence, AUUAAA (nucleotides 2395-2400), may serve this function. The open reading frame codes for a protein of 406 amino acids with a calculated M r of 44,969. The corresponding regions of porcine and human sequences shared 88 and 94% nucleotide and amino acid identity, respectively. The N-terminal amino acid sequence of the porcine protein begins with Ala, and this residue aligns to the second amino acid (Pro) in the deduced human sequence.
The 5Ј-untranslated region and open reading frame of pTG9 was in frame with the N-terminal portion of the ␤-galactosidase sequence present in the pBluescript (SKϪ) vector. Crude extracts of E. coli strain SOLR containing pTG9 expressed BHMT activity, whereas cells transformed with vector alone had no activity (Table III). These data confirm that the cDNA insert in pTG9 encodes human BHMT.
Homology of Human BHMT with Bacterial Vitamin B 12 -dependent MS-A comparison of the deduced amino acid sequence of human BHMT with other sequences using BLASTP revealed regions of homology with several bacterial vitamin B 12 -dependent MS proteins. Six amino acid alignments between human BHMT and E. coli MS can be seen in Fig. 4. The E. coli enzyme has been the most extensively characterized and is therefore the only one shown for clarity. MS is a large protein (approximately 1200 residues) composed of two distinct domains identified by trypsin sensitivity (20).
Fragments B and D align to the C-terminal domain of MS (38 kDa). Fragment B aligns to the N-terminal portion of this domain, an area that also resembles a methyltransferase from Clostridium thermoaceticum (21) that catalyzes a methyl transfer from 5-methyltetrahydrofolate to the cobalt center of a corrinoid/iron-sulfur protein (alignment not shown). Fragment D aligns to the C-terminal portion of this domain, an area that is responsible for binding S-adenosylmethionine (20), a cofactor required for the reductive activation of MS. This alignment is significant, since both S-adenosylmethionine and S-adenosylhomocysteine have been reported to inhibit BHMT activity (22,23).
Fragments A, C, E, and F align to the N-terminal domain of E. coli MS. The N-terminal domain (98 kDa) of E. coli MS is not as well characterized as the C-terminal domain. However, this domain must retain the determinants for Hcy and N 5 -methyltetrahydrofolate binding and also the residues required for methyl transfer, since it retains 70% of the activity of the holoenzyme (20). The N-terminal domain, however, can no longer be reductively activated by S-adenosylmethionine once the cobalamin is oxidized to the cob(II)alamin state. Although BHMT fragments C and E display some homology to E. coli MS, these alignments do not correspond to residues that are conserved among other MS proteins (not shown). Alignments A and F match the highly conserved regions of PXXXXXX-HXXXXEAGADXXETXTF and GGCCGTXPXHI in MS proteins and therefore could be significant for Hcy binding and catalysis of methyl transfer. DISCUSSION BHMT is found primarily in the liver and kidney of mammals (2,5,24,25), and it has been purified to homogeneity from rat (26,27), horse (28), and human (29) liver. A preparation of the enzyme from porcine liver has been reported that was estimated to be 90% pure (30). These reports consistently show that liver expresses very high levels of this enzyme, ranging from 0.6 to 1.6% of the total soluble protein in crude extracts, and that the enzyme is a hexamer of identical subunits of M r 45,000.
The porcine liver BHMT purification reported here enriched activity 47-fold, with a final yield of 16%. The preparation was judged to be homogenous after sodium dodecyl sulfate polyacrylamide gel electrophoresis yielding a single band of M r 45,000 (Fig. 1). The enrichment required to obtain homogenous enzyme varied from 47-to 98-fold. The low specific activity of the purified enzyme is similar to that reported for rat (27) and human (29) liver BHMT. The specific activity of purified BHMT is also similar to that of E. coli vitamin B 12 -independent methionine synthase (EC 2.1.1), another transferase that translocates a nitrogen-bound methyl group to Hcy (31).

FIG. 2. Nucleotide sequence of a porcine liver BHMT cDNA (partial) amplified by PCR and derived amino acid sequence.
Amino acid sequence obtained from porcine liver BHMT peptides are singly underlined. Amino acids that are different in the deduced amino acid sequence of human BHMT (Fig. 3) are shown entirely in lowercase letters and are italicized. Primers used to amplify a portion of human cDNA encoding BHMT are doubly underlined.
The K m estimates for betaine and L-Hcy are similar in magnitude to previous reports on the rat and human enzymes (19,27,29). The K m estimate for L-Hcy should be considered a maximum estimate. Although dithiothreitol was used to slow the oxidation of Hcy, no effort was made to correct for loss of substrate due to disulfide formation. Furthermore, the hydrox-ide-dependent conversion of Hcy thiolactone to Hcy is not quantitative (14).
The level of Hcy in rat and mouse liver is approximately 4 M (32); therefore, it is likely that liver Hcy levels are lower than the K m of Hcy for BHMT. As with Hcy, the intracellular concentration of betaine has been determined only for the rodent FIG. 3. Nucleotide sequence of a human liver BHMT cDNA and derived amino acid sequence. Region amplified by PCR using nonredundant primers based on porcine sequence are doubly underlined. (7,(33)(34)(35)(36). The concentration has been shown to be dependent upon dietary choline and betaine and generally ranges from 1 to 10 mol/g of liver. These levels are significantly higher than the K m of betaine for porcine BHMT (23 M) and the K m of betaine reported for the rat and human enzymes (48 and 100 M, respectively). These data indicate that rat liver BHMT is saturated with betaine, although this may not be true for all species. For example, choline oxidase, the first committed reaction for choline oxidation, is reportedly 17-60-fold higher in rat liver than human liver (37,38); therefore, humans probably have significantly lower levels of hepatic betaine. The fact that choline or betaine treatment for homocystinuria lowers pHcy supports the idea that BHMT is not normally saturated with betaine in human liver.
The higher V max obtained for the methylation of Hcy using DMAT relative to betaine is consistent with previous reports using the purified horse (28) and porcine (39) liver enzymes. These studies used saturating concentrations of both Hcy and methyl donor and measured methionine production by either colorimetric or microbiological assays. It was found that the methylation rate of Hcy using DMAT was 1-2 orders of magnitude greater than that observed using betaine. Comparison of the two substrates using crude rat liver preparations indicated that DMAT was 20 times more effective than betaine at stimulating the production of methionine (40 -42).
Dimethylglycine has been reported to be a potent inhibitor of porcine (39), rat (19), and human BHMT (29, 43), although a K i has never been reported. This report compared the relative affinities of dimethylglycine, methylthioacetate, and sarcosine for porcine BHMT (Table IV). Methylthioacetate, the demethylated product of DMAT, has a lower affinity for BHMT than dimethylglycine. Sarcosine, an isostere of methylthioacetate, displays no affinity for the enzyme under the assay conditions used. The lower affinity methylthioacetate has for BHMT partly explains the higher V max of BHMT when DMAT is used instead of betaine as methyl donor, since the maximum velocity of the reaction is in part a function of the off rates of the products.
Betaine is presently used as an effective treatment for vitamin-nonresponsive homocystinuria. This treatment significantly reduces pHcy and improves clinical prognosis. Two different mechanisms likely contribute to the pHcy-lowering effect of betaine. The first is the direct methylation of Hcy by the BHMT-catalyzed reaction. A second effect could be mediated by the subsequent oxidation of dimethylglycine to glycine, catalyzed by dimethylglycine dehydrogenase (EC 1.5.99.2) and sarcosine dehydrogenase (EC 1.5.99.1), respectively. These reactions introduce one carbon unit into the folate pool as N 5 ,N 10methylenetetrahydrofolate. This anapleurotic effect could enhance the folate-dependent pathway of Hcy methylation in some forms of homocystinuria. Although the use of betaine in the treatment of homocystinuria elicits a pHcy-lowering response, Allen et al. (43) have pointed out that this treatment usually does not lower pHcy to within the normal range (10 -15 M). The hyperhomocyst(e)inemia that persists, generally 30 -80 M, is significantly correlated with vascular disease.
The inability of betaine treatment to normalize pHcy in homocystinurics is likely related to the kinetic properties of the BHMT reaction. It has been shown that concomitant with a decrease in pHcy during betaine treatment, there is also a dramatic increase in plasma (43) and urinary (44) betaine and dimethylglycine. Up to 37-fold increases in plasma dimethylglycine (250 M) were observed in individuals receiving betaine for homocystinuria compared with a control population. These data indicate that the subsequent oxidation of dimethylglycine is insufficient and therefore accumulates in tissues. Kinetic studies using the purified porcine enzyme indicate that dimethylglycine inhibits BHMT activity uncompetitively when Hcy is varied at either subsaturating (25 M) or saturating (250 M) levels of betaine. It is likely that liver BHMT is inhibited by dimethylglycine in vivo.
This report indicates that DMAT is a more specific substrate for BHMT than betaine and that the demethylated product of DMAT, methylthioacetate, has lower affinity for BHMT than dimethylglycine. DMAT may be useful for the treatment of homocystinuria. Early work initiated by Maw and Du Vigneaud (45) indicated that DMAT can replace choline (or betaine) in rodent diets devoid of methionine but containing homocystine. DMAT supported growth and prevented fatty liver and renal hemorrhage. Methyl donor-deficient diets have been shown repeatedly to result in growth depression, fatty liver, and, when given to weaning animals, renal hemorrhage.   The subsequent metabolic fate of methylthioacetate is not completely understood. However, when rats were given 60 mg of DMAT by either diet or subcutaneous injection, over 50% of the sulfur from this compound was found as sulfate in the urine when measured after 24 h (46). There have been no reports on DMAT transport or renal clearance. Although previous studies were relatively short term (3 weeks), it seems that DMAT is nontoxic to animals at levels that might be efficacious in the treatment of homocystinuria. It should be pointed out that a homologue of DMAT, dimethylpropiothetin, is also a substrate for BHMT (28, 39 -41). The rate of methionine formation using dimethylpropiothetin was consistently shown to be between those of betaine and DMAT when assayed at saturating concentrations of substrates. Dimethylpropiothetin is a naturally occurring compound (47), and its demethylated product, methylthiopropionate, is an intermediate in the transamination pathway of methionine catabolism (48). Dimethylpropiothetin can also replace choline (or betaine) in rodent diets devoid of methionine but containing homocystine (45). Further studies on the metabolism and toxicity of DMAT and dimethylpropiothetin are warranted, since these compounds may improve the nutritional management of homocystinuria. The recent generation of transgenic mice deficient in cystathionine ␤-synthase will be useful for the comparison of betaine versus thetins as pHcy-lowering treatments for homocystinuria (49). An important objective of this research was to isolate a cDNA encoding a mammalian BHMT. Toward this end, porcine BHMT was sequenced, and oligonucleotide primers were designed and subsequently used to isolate a partial cDNA encoding porcine BHMT by PCR. This partial cDNA encoded 272 amino acids of porcine BHMT (Fig. 2). Nonredundant primers based on the porcine cDNA were then used to isolate a partial cDNA encoding human BHMT. This partial cDNA encoded 154 amino acids of human BHMT (Fig. 3). The human PCR fragment was used to screen a cDNA library by plaque hybridization, and 12 positive clones were isolated that had similar restriction maps. The BHMT cDNAs represented about 0.5% of the total number of clones in the library, consistent with the high expression of this protein in liver. The vector containing the longest cDNA insert (pTG9) was sequenced (Fig. 3), and the open reading frame encoded a 406-residue protein of M r 44,969. An alignment of the porcine and human deduced sequences displayed 94% amino acid identity over 272 amino acids (Fig.  2). The deduced human protein is similar in size to the porcine liver enzyme (Fig. 1) and the enzymes isolated from human (29) and rat (27) liver. The N terminus of the intact porcine protein begins with an Ala, and this residue aligns with the second amino acid of the deduced human sequence. Since BHMT is a cytosolic protein, it is unlikely that a large signaling peptide is posttranslationally removed, and therefore, the 2.4-kilobase pair human cDNA likely encodes the entire human protein. It is presently not known how much of the 5Ј-untranslated region of this cDNA may be missing, but a search of the freely accessible DNA data bases failed to retrieve additional sequence beyond that reported here.
In vitro simulation studies indicate that MS and BHMT contribute equally to the methylation of Hcy in rat liver (8), although in vivo studies indicate that MS and BHMT activities vary depending upon nutritional and hormonal status (5)(6)(7). Due to the very high abundance of BHMT in mammalian liver, it is difficult to hypothesize any dietary or physiological condition where BHMT would not significantly contribute to the methylation of Hcy. Mutations in the BHMT gene that diminish BHMT activity would therefore be predicted to cause some degree of hyperhomocyst(e)inemia, if not homocystinuria. A defect in this enzyme has not been reported, but its activity is absent from lymphocytes and fibroblasts (9), the usual sources of enzyme extracts used for clinical measurements. Dudman et al. (50) suggested that a deficiency of BHMT could lead to hyperhomocyst(e)inemia, but the fact that such a defect has not been observed could be due to the metabolism of Hcy, produced by the liver and kidney, by cystathionine ␤-synthase and MS in peripheral tissues. It is likely that the phenotype of BHMT deficiency will have to await the generation of transgenic mice deficient in this enzyme.
The possession of DNA sequence encoding BHMT provides the opportunity to screen for mutations in the BHMT gene. The effects of missense mutations on enzyme catalysis can easily be evaluated, since the enzyme can be expressed in E. coli. Furthermore, the specificity of the human enzyme toward thetin substrates can be evaluated, and more detailed studies on the mechanism of the BHMT reaction can proceed. The human cDNA will also allow studies concerning the regulation of expression of this enzyme by diet and hormones.