Cloning and Expression of the Mammalian Cytosolic Branched Chain Aminotransferase Isoenzyme*

The cDNA for the rat cytosolic branched chain amino- transferase (BCAT c ) has been cloned. The BCAT c cDNA encodes a polypeptide of 410 amino acids with a calculated molecular mass of 46.0 kDa. By Northern blot anal- ysis, BCAT c message of approximately 2.7 kilobases was readily detected in rat brain, but was absent from liver, a rat hepatoma cell line, kidney, and skeletal muscle. When expressed in COS-1 cells, the enzyme is immunologically indistinguishable from the native enzyme found in rat brain cytosol. Comparison of the rat BCAT c sequence with available data bases identified the Esch- erichia coli (and Salmonella typhimurium ) branched chain aminotransferase (BCAT) and revealed a Hae- mophilus influenzae BCAT, a yeast BCAT, which is hypothesized to be a mitochondrial form of the enzyme, and the murine BCAT c (protein ECA39). Calculated molecular masses for the complete proteins are 33.9 kDa, 37.9 kDa, 42.9 kDa, and 43.6 kDa, respectively. The rat BCAT c sequence was 84% identical with murine BCAT c , 45% identical with yeast, 33% identical with H. influenzae , 27% identical with the E. coli and S. typhimurium BCAT, and 22% identical with the evolutionary related D -amino acid aminotransferase ( D -AAT) (Tanizawa,

In mammals, the branched chain aminotransferase (BCAT) 1 catalyzes the first reaction in the catabolism of the essential branched chain amino acids leucine, isoleucine, and valine. Our laboratory has shown that there are two mammalian BCAT, a cytosolic and mitochondrial isoenzyme, BCAT c and BCAT m , respectively (Wallin et al., 1990). The rat BCAT c is found in brain, where it is the predominant isoenzyme, and in placenta and ovary (Hutson, 1988;Hall et al., 1993;Ichihara, 1985). The second isoenzyme, BCAT m , is found in almost all tissues of the body (Hutson et al., 1992). It has been shown that BCAT m is a bifunctional protein catalyzing transport of branched chain ␣-keto acids as well as transamination while BCAT c exhibits only transaminase activity .
Unlike mammals, bacteria appear to have a single BCAT. The two mammalian BCAT isoenzymes are similar to the Escherichia coli and Salmonella typhimurium BCAT with respect to substrate preference and kinetic properties (Inoue et al., 1988;Lee-Peng et al., 1979;Hall et al., 1993); however, the mammalian BCAT appear to be distinct from their ancestral progenitor in both size and quaternary structure. The E. coli and S. typhimurium enzymes are hexamers composed of identical 33.9-kDa subunits (Lipscomb et al., 1974). The mammalian BCAT that have been purified are larger than 40 kDa, and available data suggest that the functional unit of BCAT c is a homodimer while the functional unit of BCAT m may be a monomer (Wallin et al., 1990;Hall et al., 1993;Ichihara, 1985). No BCAT has been purified yet from lower eukaryotes such as yeast.
Based on primary sequence comparisons with other aminotransferases, it was hypothesized that the prokaryotic BCAT and D-AAT are members of a unique evolutionary subclass of aminotransferases (Mehta et al. 1993;Tanizawa et al., 1989). Recently, Grishin et al. (1995) have placed these aminotransferases in a separate folding class (fold type IV) and identified 4-amino-4-deoxychorismate lyase and murine protein ECA39 as additional members of this group. Because of its role in providing D-amino acids for bacterial cell wall synthesis, D-AAT has been a target enzyme for development of novel antimicrobial agents (Sugio et al., 1995). The crystal structure of D-AAT, which has just been solved, has revealed that the folding pattern of this protein is completely different from those of other known PLP enzymes (Sugio et al., 1995). As discussed by Sugio et al. (1995), alignment of the bacterial BCAT and D-AAT amino acid sequences suggests that the folding pattern of the bacterial BCAT resembles that of D-AAT (Sugio et al., 1995). Understanding the relationship between the mammalian and the bacterial BCAT has been limited to date by the lack of cDNAs for the mammalian enzymes which has precluded structural analysis.
In this study, the cDNA for the rat cytosolic isoenzyme, BCAT c , has been isolated. The deduced amino acid sequence revealed a 410-amino acid protein with a calculated molecular mass of 46,045 daltons. Sequence comparisons with available data bases identified two additional full-length eukaryotic sequences including the murine BCAT c (protein ECA39) and a yeast BCAT as well as a bacterial BCAT from Haemophilus influenzae. The deduced amino acid sequence for rat BCAT c shares considerable amino acid sequence identity with the murine and yeast sequences and small, but significant, identity with the bacterial BCAT and D-AAT. Amino acid sequence alignment of BCAT c and D-AAT firmly places the eukaryotic enzymes in fold type IV (Grishin et al., 1995) and shows that the evolutionary conservation of key residues involved in cofactor binding as well as provides clues to amino acid residues that may be involved in substrate recognition.

MATERIALS AND METHODS
Amino Acid Sequence of Peptides from BCAT c -Approximately 1 nmol of BCAT c was purified from rat brain as described previously . Purified BCAT c was precipitated, washed with ethanol, and solubilized in 8 M urea by heating at 37°C for 15 min. The urea was diluted to 2 M with 50 mM ammonium bicarbonate buffer, and the protein was digested with 1 mol % endopeptidase Lys-C. The resulting peptides were separated by reverse-phase HPLC on an Applied Biosystems Model 140B HPLC system using an Aquapore RP-300 C 8 column (2.1 ϫ 100 mm) (Millipore). Peaks from this separation were isolated and sequenced on an Applied Biosystems Model 4075A Sequencer by the Protein Analysis Core Laboratory of the Comprehensive Cancer Center of Wake Forest University. Trypsin digestion, peptide separation, and amino acid sequence analysis of purified BCAT c (0.5 nmol) were performed at the Harvard Microchemistry Facility (Cambridge, MA).
Synthesis of a Rat BCAT c Probe-The polymerase chain reaction was used to obtain a rat BCAT c cDNA probe. First strand cDNA was synthesized from 2 g of rat brain poly(A) ϩ RNA using murine Moloney virus reverse transcriptase according to the manufacturer's instructions (Superscript Kit; Life Technologies Inc.). For the polymerase chain reaction, oligonucleotide primers were designed based on the amino acid sequence obtained from the purified rat BCAT c and a mouse candidate BCAT c cDNA, ECA39 (accession number D21652) (Niwa et al., 1990). The sense oligonucleotide primer (5Ј-TCGTTTACTGACCA-CATGCTG-3Ј) and antisense primer (5Ј-TTCATTGTGCCTACTTCAGT-TAT-3Ј) corresponded to peptides Cyto 38 and Cyto 44 from the rat BCAT c (see Table I) and amino acids 52-58 and 255-262 of the ECA39 (see murine BCAT c in Fig. 5). Amplification was performed under the following conditions for 35 cycles: denaturation at 94°C (1 min), annealing at 60°C (1 min), and extension at 72°C (2 min). A product of the expected size (630 base pairs) was excised from an 8% (w/v) polyacrylamide gel, isolated by electroelution, and subcloned into a pT7Blue(R) vector (Novagen; Madison, WI). The insert was sequenced by the dideoxy method (Sambrook et al., 1989).
cDNA Library Screening-A rat brain gt10 cDNA library (Clontech) was screened by plaque hybridization (Ausubel et al., 1989) using the polymerase chain reaction-derived rat BCAT c cDNA as a probe. Duplicate filters were incubated for 4 h at 42°C in prehybridization buffer containing 50% (v/v) formamide, 5 ϫ SSPE (1 ϫ SSPE ϭ 150 mM NaCl, 10 mM NaH 2 PO 4 , 1 mM EDTA), 5 ϫ Denhardt's solution, 0.1% (w/v) SDS, and 100 g/ml denatured salmon sperm DNA. This was followed by hybridization for 16 h at 42°C in the same solution containing 1 ϫ 10 6 cpm/ml 32 P-labeled probe. After hybridization, the filters were washed twice for 30 min each in 0.2 ϫ SSC (1 ϫ SSC ϭ 150 mM NaCl, 15 mM sodium citrate), 0.1% (w/v) SDS at room temperature and then at 55°C. Filters were exposed overnight to Fuji RX film at Ϫ70°C with an intensifying screen. After screening approximately 1 ϫ 10 5 phage, eight positive clones were identified and plaque-purified. Plate lysate DNA (Xu, 1986) was purified and subcloned into pBluescript KS II for restriction enzyme mapping and DNA sequencing. A full-length BCAT c cDNA was sequenced on both strands using pBluescript-specific or BCAT c -specific oligonucleotides. Sequence analyses were performed using the Genetics Computer Group Sequence Analysis Software Package Version 8.0 (Devereux et al., 1984).
Isolation of RNA and Northern Blot Analysis-Total cellular RNA was isolated from whole rat brain using a lithium chloride centrifugation procedure (Duguid et al., 1988) and poly(A) ϩ RNA isolated using oligo(dT)-cellulose from Pharmacia LKB Biotechnology Inc. (Uppsala, Sweden). Rat liver, hepatoma (H4-II-E-C3), skeletal muscle, hippocampus, cortex, and cerebellum total RNA were isolated using the guanidinium isothiocyanate-CsCl centrifugation procedure (Chirgwin et al., 1979). Rat kidney total RNA was purchased from Clontech. For Northern blot analysis, total RNA (10 g) was fractionated on 1.2% (w/v) agarose gels containing 2.2 M formaldehyde, transferred to a Nytran membrane (Schleicher and Schuell), and fixed by UV irradiation (Ausubel et al., 1989). The filter was incubated in prehybridization buffer for 4 h at 42°C and hybridized for 12 h in the same solution containing 1 ϫ 10 6 cpm/ml of 32 P-labeled BCAT c cDNA probe. The filter was washed twice for 30 min each in 0.2 ϫ SSPE, 0.1% (w/v) SDS at room temperature and then at 65°C.
DNA Transfection and BCAT c Activity-The rat BCAT c cDNA (nucleotides 1-1370) was subcloned into a mammalian cell expression vector, pCMV5 (Andersson et al., 1989). After ligating the BCAT c cDNA into EcoRI-digested pCMV5, the plasmid was characterized by restriction mapping and dideoxynucleotide sequencing. For transfection, the plasmids were isolated using the Wizard Maxiprep DNA purification procedure (Promega). On day 0, 1.5 ϫ 10 6 COS-1 cells per 100-mm dish were plated in Dulbecco's modified Eagle's medium containing 4500 mg/liter D-glucose, 10% (v/v) fetal calf serum, 100 units/ml penicillin, and 100 g/ml streptomycin. On day 1, COS-1 cells were transfected with 5 g of either pCMV5:BCAT c or pCMV2:␤gal (Briggs et al., 1993) by the DEAE-dextran method (Wong et al., 1994). On day 4, cells were washed three times with phosphate-buffered saline, harvested using ice-cold phosphate-buffered saline containing inhibitor mixture (1 mM EDTA, 1 mM EGTA, 1 mM diisopropyl fluorophosphate, 5 mM benzamidine hydrochloride, and 10 g/ml leupeptin) and 1 mM dithiothreitol. The cells were lysed in a hypotonic solution (25 mM Hepes, 0.4% CHAPS, pH 7.5) containing protease inhibitor mixture and dithiothreitol and sonicated for 2 min at 50% duty cycle using a Branson Model 250 sonifier. The extract was centrifuged at 14,000 ϫ g for 30 min, and the supernatant was assayed for BCAT activity. BCAT activity was measured at 37°C in 25 mM potassium phosphate buffer, pH 7.8, as described Wallin et al., 1990). One unit of enzyme activity was defined as 1 mol of valine formed per min per mg of cell protein at 37°C.
Immunoblotting, Immunoprecipitation, and Tissue Fractionation-SDS-PAGE was performed according to Laemmli (1970) using 10% (w/v) acrylamide gels. Prior to electrophoresis, samples were boiled for 2 min in the presence of 1% (w/v) SDS and 5% (v/v) ␤-mercaptoethanol. For immunoblotting, proteins were transferred to Immobilon P membranes (Millipore) as described previously (Wallin et al., 1990). The filter was blocked with 5% (w/v) bovine serum albumin (Fraction V, Sigma) and incubated with rabbit anti-rat BCAT c serum (1:500 dilution) . Immunoreactive protein bands were visualized using horseradish peroxidase-labeled goat anti-rabbit antibody according to the manufacturer's instructions (Bio-Rad). For immunoprecipitation, rat brain cytosol (150 l) or cell extracts (20 l) were adjusted to 0.5 M NaCl and mixed with 20 l of preimmune or isoenzyme-specific anti-BCAT serum in a total volume of 280 l. Immune complexes were isolated using Staphylococcus aureus Protein A (Pansorbin) as described . Rat brain cytosols from whole rat brain and isolated cerebellum, cortex, and hippocampus were prepared as described . Protein was determined in cytosol fractions and COS-1 cell extracts using the DC Protein Assay (Bio-Rad).

RESULTS
N-terminal amino acid sequence could not be obtained from purified BCAT c ; therefore, two separate protein samples were subjected to Lys-C endopeptidase and trypsin digestion, respectively. The amino acid sequence of nine internal peptides is shown in Table I. When the amino acid sequence of these peptides was analyzed, a significant number of the peptides (Cyto 38, 54, 24, 44, 45, and 74; see Table I) were found to align with a previously identified mouse cDNA, ECA39, which was an overexpressed mRNA found in the murine PCC4 Aza1 teratocarcinoma cell line (Niwa et al., 1990). The rat BCAT c tryptic peptide sequence Cyto 74 and two overlapping endopeptidase Lys-C peptides, Cyto 44 and Cyto 45, showed 83% identity over a 29-amino acid region at the predicted carboxyl terminus of the ECA39 protein. However, the rat BCAT c peptide sequence extended an additional 17 amino acids past the end of the mouse-predicted coding sequence. Further analysis of the ECA39 sequence revealed that upon shifting the reading frame by the addition of a single nucleotide residue (guanosine) at position 1018, the predicted coding sequence would now align with the three BCAT c peptide sequences. The amino acid identity was now greater than 90%, and all of the BCAT c peptides, except Cyto 48, could be aligned with the mouse sequence. If a second nucleotide was inserted in the codon at position 1262 to 1263, the deduced ECA39 sequence would now align with Cyto 48 and place Cyto 48 at the C terminus of BCAT c . This sequence alignment and expression of ECA39 in brain and kidney (Niwa et al., 1990) strongly suggested that ECA39 is the mouse homolog of BCAT c and explains why Grishin et al. (1995) placed it in the same folding class as the E. coli BCAT.
Using the nucleotide sequence of ECA39 and peptide sequence derived from the purified rat BCAT c , nondegenerate oligonucleotide primers were designed and used in the polymerase chain reaction with rat brain cDNA. A single product of the expected size (630 base pairs) was obtained. Partial sequence analysis of the product revealed at least 80% nucleotide identity with the mouse candidate BCAT c . This product was used to screen a rat brain gt10 cDNA library (approximately 1 ϫ 10 5 plaques). Eight positives clones were identified. Six of the original eight plaques were positive by polymerase chain reaction, and two of these clones were plaque-purified and subcloned into pBluescript.
The cDNA and deduced amino acid sequence for BCAT c are shown in Fig. 1. The predicted initiator methionine lies in an appropriate consensus for initiation of translation (Kozak, 1987) and is preceded by a 5Ј-untranslated region of 62 nucleotides. The translation termination codon is followed by a 3Јuntranslated region of 72 nucleotides that did not extend to the poly(A) tail. The rat cDNA encodes a protein of 410 amino acids with a calculated molecular mass of 46,045 daltons which is in excellent agreement with the molecular mass of 47 kDa determined by SDS-PAGE . The translated BCAT c sequence contains an alanine residue immediately following the N-terminal methionine. The presence of a modified Nterminal alanine would be consistent with reports characterizing the processing of amino termini of cytosolic eukaryotic proteins (Flinta et al., 1986;Huang et al., 1987) which have shown that when alanine is adjacent to the N-terminal methionine, removal of the initiator methionine residue and modification of the exposed alanine occurs with high frequency. The location of the peptide sequences obtained from proteolytic digestion of purified BCAT c in the translated sequence are indicated in Table I. A comparison of the amino acid sequences determined chemically with those determined from the cDNA clone reveals 100% agreement with the tryptic peptide sequences (Cyto 74 and 48) and Cyto peptides 24, 41, and 54. The mismatches in Cyto 44 (His-283) and Cyto 45 (His-312) were correctly determined in Cyto 74 as glycine. Chemically, gluta-mine was observed at position 344 in Cyto 27, and only one serine instead of two was observed at position 70 in Cyto 38.
BCAT c is known to have a unique tissue distribution in the rat. The enzyme is found in brain and is notably absent in other tissues with the exception of low activity in ovary and placenta (Hutson, 1988;Ichihara, 1985;Hall et al., 1993). To compare BCAT c expression with the known tissue distribution of BCAT c activity and protein, Northern blot analysis was performed with total RNA from whole rat brain (see Fig. 2, lane B). Since the distribution of BCAT c throughout the brain is not known, both BCAT c expression and cytosol activity were examined in the cerebellum, cortex, and hippocampus (see Fig. 2, lanes marked C, H, and CE). A single mRNA transcript of approximately 2.7 kb was found in all brain regions. The additional length of the mRNA as compared to the length of the cDNA most likely represents an additional 3Ј-untranslated region. The putative murine BCAT c is also encoded by a larger mRNA with a 1107-nucleotide 3Ј-untranslated region. BCAT c activity was found in all brain regions. The ubiquitous expression of BCAT c in different brain regions is consistent with reports showing branched chain transamination in astrocyte and neuronal cultures (Hertz et al., 1987;Yudkoff et al., 1994). What has not yet been determined is the exact distribution of BCAT isoenzymes in different cell types in the brain. Fig. 3 shows Northern blot analysis performed with total RNA from a rat liver hepatoma (H4-II-E-C3) cell line, rat kidney, liver, and skeletal muscle. Brain RNA is shown for reference. Consistent with previous reports on the tissue distribution of rat BCAT c (Hutson, 1988;Ichihara, 1985;, no hybridization was observed with RNA from these tissues, even after lower stringency hybridization conditions. Stripping and reprobing the blots with a probe for glyceraldehyde-3-phosphate dehydrogenase (Ercolani et al., 1988) confirmed the load and integrity of the RNA (Figs. 2 and 3, lower panels).
To demonstrate functional expression of BCAT c , BCAT c cDNA was ligated into an expression vector (pBCAT c ) and transfected into COS-1 cells. The expressed BCAT c protein was visualized by immunoblotting using an antibody raised against purified rat BCAT c ) (see Fig. 4A). Cell extracts were also assayed for BCAT activity and immunoprecipitated with BCAT c antiserum (see Fig. 4B). Lane 5 (pBCAT c ) contains the pBCAT c -transfected COS-1 cell extract, and lane 1 extract from mock-transfected COS-1 cells (p␤Gal). Lanes 2-4 contain rat brain cytosol preparations (Brain Cytosol) as a reference. A band of about 47 kDa was readily detected in the rat brain cytosols and in the pBCAT c -transfected COS-1 cell extract. The appearance of the lower band in lane 5 varied according to the composition of the inhibitor mixture which indicated that the lower band resulted from proteolytic degradation. The same observation was made when the enzyme was purified from rat brain cytosol which suggests that the protein is very sensitive to proteolysis . COS-1 cell BCAT specific activity was 26-fold higher (range 20-to 100-fold in four separate transfection experiments) than in brain cytosol and essentially all BCAT activity was neutralized by the BCAT c antiserum (see Fig. 4B, pBCAT c ). Neither BCAT c antigen (see Fig.  4A, lane 1) nor significant BCAT activity (see Fig. 4B, p␤Gal) was observed in the extract from the p␤Gal-transfected cells.
The relative amino acid substrate specificity of BCAT c expressed in COS-1 cells with KIC as amino acceptor was also examined. The amino acid preference in descending order was isoleucine, leucine, valine, norvaline, glutamate, and norleucine, which is essentially the same as has been reported previously for purified BCAT c . Phenylalanine and alanine were not transaminated (Յ1% of control value with leucine as amino acceptor). In addition to the murine ECA39 cDNA, a search of the available protein and nucleic acid data bases revealed that the rat BCAT c protein sequence has significant similarity to a yeast sequence, and partial sequences from the nematode Caenorhabditis elegans, the plant Arabidopsis thaliana, several human fragments as well as the bacterial BCAT from E. coli and S. typhimurium and a gene product from H. influenzae Rd (Fleischmann et al., 1995). A multiple sequence alignment of the rat, mouse, yeast, and bacterial BCAT sequences is shown in Fig. 5. Since only 7 of 308 residues from the genetically FIG. 1. Nucleotide and predicted amino acid sequence of the rat BCAT c . The nucleotides and amino acids are numbered on the right of the sequence. Amino acid residue 1 is the alanine residue immediately following the predicted initiator methionine; an asterisk denotes the stop codon. The underlined nucleotide residues indicate the location of primers used in the polymerase chain reaction to generate a rat BCAT c probe.

FIG. 2. Northern blot hybridization of RNA from rat brain.
In the upper panel, total RNA (10 g) from whole brain (B), cerebellum (Ce), cortex (C), and hippocampus (H) was subjected to electrophoresis, blotted, and hybridized using the uniformly 32 P-labeled BCAT c cDNA probe (BCAT c ) as described under "Materials and Methods." The washed filter was exposed to Fuji RX film with an intensifying screen for 2 days at Ϫ70°C. The migration of RNA standards run in an adjacent lane are indicated. Branched chain aminotransferase activity was measured in cytosol fractions as described under "Materials and Methods." Aminotransferase activity is shown relative to the activity in whole brain cytosol. Activity in whole brain cytosol was 51.7 Ϯ 2.9 milliunits/mg of protein determined in four separate cytosol preparations. In the lower panel, the filter was stripped and rehybridized with a uniformly 32 P-labeled glyceraldehyde-3-phosphate dehydrogenase probe (GAPD), washed, and exposed to Fuji RX film for 2 days. similar E. coli (Ilv_E) and S. typhimurium (Ilv_S) are different, and these differences do not affect the consensus sequence, only the E. coli sequence was used in the alignment. The deduced amino acid sequence from the H. influenzae Rd BCAT, which is only 31% identical with the other two bacterial sequences, was included as a separate entry in the alignment. Rat BCAT c is significantly similar to the eukaryotic BCAT exhibiting 84% and 45% identity to the murine and yeast sequences, respectively. The rat sequence exhibits 27% identity with the E. coli BCAT and 33% identity with the H. influenzae gene product. Sequence identity with D-AAT from B. subtilis was 22%. On the other hand, comparison of the BCAT c sequence with the rat and porcine cytosolic aspartate aminotransferases revealed only 18% and 16% sequence identity, respectively.
Several important results have emerged from the determination of the complete amino acid sequence of rat BCAT c . All the eukaryotic enzymes have molecular masses greater than 40 kDa compared to just under 34 kDa for the E. coli and S. typhimurium enzymes. The calculated molecular mass for the H. influenzae BCAT is intermediate at 37.9 kDa. Without exception, the additional amino acid sequence is at the N-terminal region of the proteins, and this is the most variable region in these protein sequences (see Fig. 5). The consensus sequence for the eukaryotic enzymes reflects the high degree of sequence conservation with 208 residues in the consensus sequence (not shown). This includes a sequence of 13 amino acids starting at BCAT c Leu-71 immediately preceding the N terminus of the E. coli BCAT in Fig. 5. When the bacterial BCAT are included, 57 amino acids remain in the consensus sequence. The PLP binding site of rat brain BCAT c is most likely Lys-246 with the conserved sequence KXGXNY found in all BCAT sequences including the mitochondrial isoenzyme. 2 Other regions showing relatively high sequence identity in all BCAT enzymes include the sequence of 14 amino acids beginning at BCAT c Leu-112 and the sequence of 12 amino acids beginning at Gly-356. BCAT c Phe-135, Trp-92, and Trp-405 are also conserved in all BCAT sequences.

DISCUSSION
This manuscript presents the first cloning of a mammalian BCAT. The tissue distribution of the BCAT c mRNA is consistent with the known distribution of the cytosolic BCAT isoenzyme determined by measurements of activity and immunological techniques. The protein, when expressed in COS-1 cells, is immunologically indistinguishable from the enzyme in rat brain cytosol, and the amino acid substrate preferences were the same as the enzyme purified from rat brain cytosol .
Two additional complete eukaryotic BCAT sequences have been identified. The ECA39 cDNA (Niwa et al., 1990), once two nucleotide changes are made, appears to code for the murine BCAT c with a calculated molecular mass of 42.9 kDa. The other sequence is a yeast BCAT with a calculated molecular mass of 43.6 kDa for the complete protein. Very little is known about the yeast BCAT including whether or not these lower eukaryotes contain both isoenzymes. However, the calculated isoelectric point for the yeast enzyme of 9.59 is considerably higher than the calculated values of 5.6 and 5.25 for the rat and mouse cytosolic enzymes, respectively. Hartmann and Christen (1991) observed that, with few exceptions, the pI of the mitochondrial isoenzyme has a more basic isoelectric point than the cytosolic form of the same enzyme. If the yeast enzyme is a mitochondrial protein, then it is likely to have an N-terminal targeting sequence. The alignment of the mammalian BCAT with the yeast sequence becomes unambiguous at BCAT c -51 (Yeast BCAT-25) and Leu-71 in BCAT c (Yeast BCAT-45) is the start of a highly conserved sequence (LVFG in Fig. 5). Mitochondrial targeting signals are generally from 10 -70 amino acids, can form an amphipathic helix composed of positively charged and hydrophobic residues and generally lack acidic amino acids (Hartl et al., 1989). An arginine is usually found at the Ϫ2 position from the mitochondrial processing peptidase cleavage site (Hartl et al., 1989). Secondary sequence analysis shows potential helical structure in this region, and an ␣ helix would be amphipathic. Only one acidic residue, an aspartic acid, is found in the first 34 amino acids. Based on the alignment shown in Fig. 5 and assuming an arginine at the Ϫ2 position, cleavage could occur in the yeast BCAT between residues 16 and 17 resulting in a mature protein of 377 amino acids with a calculated molecular mass of 41 kDa. Two other potential cleavage sites preceding BCAT c Leu-71 appear less likely since they occur after BCAT c -51.
Cloning of the first mammalian BCAT, rat BCAT c , and identification of two additional full-length eukaryotic sequences, is the first step in developing a molecular model for the eukaryotic BCAT. Sequence similarity between the E. coli BCAT and D-AAT led to the hypothesis that these two enzymes evolved from a common ancestral gene (Tanizawa et al., 1989). The enzymes are 25% shorter than the average aminotransferase length of approximately 400 amino acids, and alignment of these enzymes with other aminotransferases indicates that the small domain as defined in aspartate aminotransferase is 2 R. Bledsoe, P. A. Dawson, and S. M. Hutson, unpublished results.
FIG. 3. Northern blot hybridization of RNA from selected rat tissues. Ten micrograms of total RNA from whole brain (B), rat hepatoma cell line H4-II-E-C3 (He), liver (L), kidney (K), and mixed skeletal muscle (M) were subjected to electrophoresis and hybridized with the uniformly 32 P-labeled BCAT c cDNA probe (BCAT c ) as described under "Materials and Methods" and in the legend to Fig. 2. In the lower panel, the stripped filter was rehybridized with a glyceraldehyde-3-phosphate dehydrogenase probe (GAPD).

FIG. 4. Expression of rat BCAT c in pBCAT c -transfected COS-1 cells.
A, cell extracts (60 g) from pCMV-␤-galactosidase-transfected COS-1 cells (lane 1, p␤Gal), pBCAT c -transfected COS-1 cells (lane 5, pBCAT c ), and whole rat brain cytosol (100 g) (lanes 2-4, Brain Cytosol) were subjected to SDS-PAGE and immunoblotted as described under "Materials and Methods." Protein molecular mass standards are labeled Std. B, branched chain aminotransferase activity was measured in cell extracts and brain cytosol before (Control) and after immunoprecipitation with BCAT c antiserum (Ab BCAT c ). NM, not measured. either absent or of different structure. Another unique property of these enzymes concerns the stereochemistry of the hydrogen transfer which occurs during catalysis on the re face of the cofactor while in other classes of aminotransferases proton transfer occurs from the si face (Yoshimura et al., 1993). The crystal structure of the pyridoxamine-form of thermostable D-AAT has been solved, and it is clear that the structure is distinct from the other PLP-containing enzyme structures that have been reported (Sugio et al., 1995). Fig. 6 presents a GAP alignment of rat BCAT c and D-AAT. Structural analysis of D-AAT shows that, unlike other aminotransferases, the N-terminal domain (D-AAT residues 1-120) FIG. 5. Multiple sequence alignment of eukaryotic and prokaryotic branched chain aminotransferases (BCAT). The deduced amino acid sequence of the rat BCAT c was aligned with mouse ECA39 (Mouse BCAT c ), the S. cerevisiae twt1 gene (Yeast BCAT), H. influenzae Rd (H. infl BCAT) and Ilve_E (E. coli BCAT) using the Multiple Sequence Alignment software of the Genetics Computer Group Sequence Analysis Software Package, version 8.0 (Devereux et al., 1984). The BCAT accession numbers are: D21652 (murine), X78961 (yeast), X02413 (E. coli), and HI1193 (H. influenzae Rd) (Fleischmann et al., 1995). The active site lysine which forms the internal aldimine with the pyridoxal phosphate cofactor is indicated with a solid circle. The consensus sequence was determined using a plurality of 5.0. and the C-terminal domain (D-AAT residues 121-282) are comprised of very different secondary and tertiary structures, and the pyridoxal phosphate binding site is located mainly at the interface of the two domains of the same monomer rather than being shared between the two monomers as in aspartate aminotransferases (Sugio et al., 1995). Only two residues, Arg-98* and His-100* from the other subunit, extend to the active site and PLP-binding region. Secondary structure predictions of the overlapping regions of rat BCAT c and D-AAT show a similar pattern, suggesting that the two proteins have similar folding patterns. Therefore, the mammalian BCAT can be classified as belonging in the folding group IV described by Grishin et al. (1995).
Since the components necessary for transamination are clearly present in the shortest bacterial BCAT, the function of the N-terminal region of the eukaryotic enzymes is not known. Except for the conserved sequence around BCAT c -71, the Nterminal portions of the eukaryotic BCAT are the most variable regions of these sequences (see Fig. 5). Secondary structure predictions for the N terminus of BCAT c are predominantly ␣-helix, suggesting an ordered structure. Since the quaternary structure of rat BCAT c and the E. coli BCAT appears to be different, dimer versus hexamer, the possibility exists that the N-terminal extension may influence quaternary structure. (The H. influenzae Rd BCAT appears to be of intermediate size, but the protein's quaternary structure is not known.) Another possibility is that the small highly conserved portion of this region (BCAT c -71 to -82) serves a function analogous to the extended amino segment of the aspartate aminotransferases and stabilizes the dimer (Fukumoto et al., 1991). The N terminus may also play a role in domain movement (Kirsch et al., 1984).
In the transamination reaction the catalytic lysine has an important role in determining the stereochemistry of the proton transfer (Kirsch et al., 1984;Sugio et al., 1995). In Laspartate aminotransferase, the lysine faces the side of the substrate where the ␣-proton is located (see Fig. 7A). The lysine is the general base involved in abstraction of the proton from the ␣-carbon and reprotonation of the aldehydic carbon of the coenzyme to yield a ketimine intermediate. The lysine then returns the proton onto the pro-S side of the substrate so that the L-configuration is maintained when the reaction proceeds FIG. 6. Alignment of rat cytosolic branched chain aminotransferase (BCAT c ) with D-amino acid aminotransferase (D-AAT). The deduced amino acid sequences of the rat BCAT c and thermostable Bacillus D-AAT were aligned using the GAP analysis software (Genetics Computer Group Sequence Analysis Software Package, Version 8.0) (Devereux et al., 1984). A Bestfit analysis did not give a significantly different alignment. A single manual adjustment was made at D-AAT-215 to preserve the ␣-helix in D-AAT (Sugio et al., 1995). Structurally important residues in D-AAT, conserved residues in folding class IV proteins, and residues that are also conserved in the BCAT c sequence are boxed. The asterisk denotes that Arg-98* and His-100* are from the opposite subunit in the D-AAT dimer. Residues discussed in relation to ␣-carboxylate binding are indicated by arrows.
in the reverse direction. The lysine approaches the coenzyme from the si face which is the side facing the protein. In D-AAT and E. coli BCAT, the proton is added to or abstracted from the C4Ј atom of the coenzyme-imine or external aldimine intermediate on the re face (Yoshimura et al., 1993). The crystal structure of D-AAT shows that the cofactor is bound so that the side chain of the catalytic lysine (D-AAT Lys-145) and the ␣-proton point toward the protein as in L-aspartate aminotransferase (see Fig. 7B). Key active site residues involved in cofactor binding in D-AAT are conserved in the bacterial BCAT suggesting a similar cofactor orientation (Fig. 7C). The GAP alignment shown in Fig. 6 indicates that these structural features are conserved in rat BCAT c and also in the other BCAT (see Fig. 5).
The key residues in D-AAT that are also found in BCAT c include the active site lysine (D-AAT Lys-145), D-AAT Glu-177, which interacts with the pyridoxal nitrogen, and the residues which interact with the phosphate group, D-AAT Ile-204, Thr-205, Thr-241, and Arg-50 (Sugio et al., 1995) (see boxed residues in Fig. 6). Since it is the main chain amide of D-AAT Ile-204 that forms the hydrogen bond with a phosphate oxygen, substitution of a Val (BCAT c -313) in this position should not perturb the structure significantly. The helix dipole (N terminus of helix D-AAT 203-216) which interacts with the negative charge on the phosphate may also be conserved in the BCAT (see Fig. 6). Secondary structure predictions for the equivalent residues in BCAT c , residues 312-325, suggest that BCAT c and D-AAT would have similar secondary structure in this region. D-AAT Leu-201, which is about 3.9 Å from the C4Ј of PLP, is also conserved in BCAT c (Leu-310 in Fig. 6) and the other BCAT sequences (see Fig. 5). The D-AAT L201A and L201W mutants show anomalous kinetics, and the enzymes appear to form an inactive pyridoxamine form during catalysis (Kishimoto et al., 1995). In addition, two other residues conserved in fold class IV enzymes, D-AAT Gln-181 and Phe-183, are also found in BCAT c . Therefore, it seems reasonable to conclude that orientation of the PLP cofactor in the eukaryotic BCAT is the same as in D-AAT (see Fig. 7, B and C).
Two elements of the PLP binding site in D-AAT may be altered in the BCAT enzymes. Tyr-31 in D-AAT which appears to be in a position to bind to the phenolic oxygen (O3Ј) and imine nitrogen atom of PLP is replaced by phenylalanine in all BCAT (see Figs. 5 and 6). The equivalent tyrosine in L-aspartate aminotransferase plays an important role in stabilization of intermediates during transamination (Kirsch et al., 1984;Arnone et al., 1985). Whether or not the tyrosine in the adjacent conserved motif beginning at BCAT c Leu-112 (LHYxxx-(L,v,c)FEG) would be in a position to hydrogen-bond the phenolic oxygen remains to be established. A second difference in D-AAT and aspartate aminotransferase is in the residues that form a parallel stacking interaction with the pyridoxal ring. The tryptophan residue that serves this function in aspartate aminotransferase is absent in D-AAT. Instead, the side chain hydroxyls of a loop of three serine residues (D-AAT Ser-179 to Ser-181) hydrogen-bonded to the carboxylate group of D-AAT Glu-166 form a barrier on one side of the pyridoxal ring. None of these residues appears to be conserved in any of the BCAT (see Figs. 5 and 6). Although several tryptophan residues are conserved in the eukaryotic BCAT sequences, only BCAT c Trp-92 and Trp-405 are conserved in all BCAT. These two tryptophans would presumably be found in the BCAT small and large domains, respectively. Thus, it would appear that the BCAT have evolved a different structural solution in this region of the protein.
D-AAT accepts a broad range of D-amino acids. On the other hand, the mammalian BCAT are selective for L-branched chain amino acids and their straight chain analogs, but exhibit no activity with alanine or aromatic amino acids Ichihara, 1985). As shown in Fig. 7, B and C, the orientation of the ␣-hydrogen dictates that the ␣-carboxyl and side chain of an L-amino acid in the BCAT active site must be in the opposite orientation from a D-amino acid in the substrate binding pocket. Two elements of the substrate binding site include a "carboxylate trap" which is positively charged and a "side-chain pocket" to discriminate substrate species (Kirsch et al., 1984;Arnone et al., 1985). In D-AAT, the substrate ␣-carboxylate is thought to interact with Arg-98* (see Fig. 6) of the other subunit dimer. No arginine is present in the BCAT consensus sequence around the D-AAT Arg-98. This difference is predict- able, because of the opposite orientation of an L-amino acid and D-amino acid in the active site (see Fig. 7, B and C); hence, this region of the BCAT is probably involved in side chain recognition. Modeling of E. coli BCAT on the D-AAT structure that is described, but not shown in Sugio et al. (1995), led to the proposal that either the Arg present in the E. coli sequence equivalent to residue D-AAT-35 or D-AAT-88 could be located on the proper side of the active site for ␣-carboxylate binding in the bacterial enzyme. Examination of the alignment of BCAT c with D-AAT in Fig. 6 and the consensus sequence in Fig. 5 does not reveal a conserved arginine residue at either position in the BCAT sequences although four arginines are found in the consensus sequence shown in Fig. 5. However, there is a conserved arginine at BCAT c -187 (equivalent to D-AAT-83) and an arginine that is found in all BCAT sequences except the E. coli enzyme at BCAT c -126 (D-AAT-38) (see Fig. 6). Mutagenesis will be required to test these candidate arginines.
In summary, cloning of the eukaryotic BCAT has confirmed placement of the BCAT in the same folding group as D-AAT and the bacterial BCAT (Grishin et al., 1995). The mammalian proteins, however, have acquired additional amino acid sequence during evolution, a structural change that has not been observed in the other aminotransferase subclasses (Mehta et al., 1993). Preliminary modeling of the mammalian BCAT on the D-AAT structure (data not shown) has provided some insights into regions that may be important in the structure of the BCAT active site which can be tested by site-directed mutagenesis.