Overexpression and Characterization of the Human Mitochondrial and Cytosolic Branched-chain Aminotransferases*

,

Branched-chain aminotransferases (BCAT) 1 catalyze reversible transamination of the short chain aliphatic branched-chain amino acids, leucine, isoleucine, and valine, to their respective ␣-keto acids. Although transamination of branched-chain amino acids in animals and microorganisms was observed in the 1950s (for a review, see Ref. 1), it was not until 1966 that Ichihara and Koyama (2) and Taylor and Jenkins (3) each reported independently that these reactions were catalyzed by a single enzyme.
L-Leucine ϩ ␣-ketoglutarate^␣-Ketoisocaproate ϩ L-glutamate L-Valine ϩ ␣-ketoglutarate^␣-Ketoisovalerate ϩ L-glutamate L-Isoleucine ϩ ␣-ketoglutarate^␣-Keto-␤-methylvalerate ϩ L-glutamate  Bacteria generally contain a single BCAT (4 -6), but in mammals it has been established that there are two BCAT isoenzymes, a mitochondrial (BCATm) and a cytosolic (BCATc) form (7). In humans and rodents, BCATm is found in most tissues (1,8). In contrast, BCATc is found almost exclusively in the brain (1,9). Each rat isoenzyme has been purified (9,10), and the cDNA sequences of the rat (11,12) as well as the human (12,13) BCATc and BCATm are now available. Recently, it has been reported that the mammalian BCATc gene may be regulated by the c-myc oncogene product (14) and that the BCAT may be a target for the anticonvulsant drug and nonmetabolizable leucine analog, gabapentin (15). Furthermore, in yeast, deletion of both BCAT genes results in severe growth retardation even after supplementation with branched-chain amino acids, suggesting that these enzymes perform an essential function in the cell (16). Nevertheless, the functional significance and structure of the individual eukaryotic BCAT isoenzymes is not yet known.
Another feature of the BCAT is their evolutionary relationship to the bacterial enzyme D-amino acid aminotransferase (17). Based on primary sequence comparisons, the BCAT and D-amino acid aminotransferase, which have opposite stereospecificity (Lversus D-amino acids), and another bacterial enzyme, 4-amino-4-deoxychorismate lyase, were placed in a separate folding class (fold type IV) (11,12,18). With a few exceptions, other known aminotransferases fall within the fold type I or L-aspartate aminotransferase superfamily. A unique feature of the fold type IV PLP-dependent enzymes is that the proton is added to or abstracted from the C4Ј atom of the coenzyme-imine or external aldimine intermediate on the re face instead of the si face of the PLP cofactor (19). The crystal structure of the PMP form of D-amino acid aminotransferase has been solved by Sugio et al. (20), and the structure of the Escherichia coli BCAT in the pyridoxal form has just been reported by Okada et al. (21). The folding pattern of both enzymes is not only different from that of other known PLPdependent enzymes but also from that of other known proteins (20). The E. coli BCAT enzyme consists of six identical 34-kDa subunits that are an assembly of three dimer units around a three-fold axis (21,22). On the other hand, the mammalian BCAT do not appear to be hexamers and have subunit molecular masses ranging from 41 to 46 kDa (11,12). Thus, cloning and overexpression of both human BCAT isoenzymes has provided the tools necessary to develop a molecular model of the mammalian BCAT that may impact on our current understanding of this unusual class of PLP-dependent enzymes.

MATERIALS AND METHODS
Plasmids, Bacterial Strains, Enzymes, and Chemicals-Plasmids pT7Blue(R) and pET-28a were purchased from Novagen (Madison, WI). The pTrcHis vector was obtained from Invitrogen (Carlsbad, CA). Bacterial strains used were E. coli BL21DE3 and DH5-␣. Restriction endonucleases and other DNA modification enzymes were purchased from Promega, Inc. Nucleotide sequencing was performed using the Sequenase version 2.0 7-deaza-dGTP sequencing kit (U.S. Biochemical Corp.).
Construction and Expression of phBCATm, phBCATc, and phB-CATc2-For human BCATm, the sense primer 5Ј-AGCCATATGGC-CTCCTCCAGTTTCAAG-3Ј contained an NdeI restriction site. The last 18 nucleotides of the sense primer corresponded to the first 18 bases encoding the mature BCATm protein (12). The antisense primer 5Ј-GAGTGTCGCAACCACAT-3Ј corresponded to nucleotides 1316 -1332 of the full-length human cDNA (12). For increased accuracy, a polymerase mixture of Taq and Pwo polymerases (Boehringer Mannheim) was used. The amplified product was cloned into the pT7 vector. The insert was sequenced on both strands, and the fidelity was verified by comparison with the original human BCATm cDNA clone. The purified cDNA was ligated into the pET-28a vector cut with NdeI. Following transformation, plasmid DNA isolated from single colonies was screened for correct orientation of the cDNA insert, and then the construct was used to transform BL21DE3 cells. Cells were grown to an A 600 between 0.6 and 0.9, and expression was induced with 1 mM isopropyl-␤-D-thiogalactopyranoside. After 8 h, expression was approximately 8 mg of BCATm/liter of culture with a typical recovery of 6 mg of purified protein/liter of E. coli.
A similar strategy was used to construct the phBCATc expression vector. Using the cDNA sequence in Ref. 13 (accession number U21551), a sense primer was designed that contained an NheI site immediately preceding the codon for the initiator methionine followed by the nucleotides encoding the first 5 amino acids of the protein, 5Ј-TATGGCTAG-CATGGATTGCAGTAACGGA-3Ј. The antisense primer 5Ј-TCAGGAT-AGCACAATTGTC-3Ј corresponded to nucleotides 1137-1155. Human brain cDNA was used in the polymerase chain reaction, and the amplified product was cloned into the pT7 vector and sequenced. The purified cDNA was ligated into the pTrcHis vector cut with NheI and HindIII. The phBCATc2 plasmid was engineered by ligating the full-length BCATc sequence into the pET-28a vector cut with SalI and NheI. BL21DE3 cells were grown and induced as described for phBCATm, and after 5-7 h, expression was 2 and 8 mg of BCATc/liter with phB-CATc and phBCATc2, respectively. Typically, 50% of the expressed protein was recovered after purification.
Protein Purification-Nickel-NTA resin (Qiagen, Chatsworth, CA) was used according to the manufacturer's instructions to purify the histidine-tagged recombinant proteins. For BCATm, cells harvested from 2 liters of culture were resuspended in extraction buffer containing 0.1 M sodium phosphate, pH 8.0, 0.01 M Tris-HCl, and 5 mM ␤-mercaptoethanol. The mixture was sonicated for 10 1-min intervals at 70% duty cycle using a Branson model 250 sonifier. The extract was centrifuged for 8 min at 7800 ϫ g at 4°C. The supernatant was saved, and the pellet was resuspended in extraction buffer containing 4 M urea and sonicated as before. After centrifugation, the second supernatant was combined with the first supernatant to which urea had been added to 4 M. The total extract was incubated with 7 ml of 50% nickel-NTA resin equilibrated in the extraction buffer containing urea for 1 h at 4°C with gentle stirring. The mixture was loaded into a column, and the resin was washed with the extraction buffer containing urea. The column was then washed with buffer A containing 0.01 M Tris, pH 7.5 (HCl), 20% glycerol, 150 mM NaCl, 5 mM ␤-mercaptoethanol followed by buffer B, which contained 0.1 M sodium phosphate (pH 6.0), 0.01 M Tris-HCl, 750 mM NaCl, 10% glycerol. The column was washed further with buffer B containing 50 mM imidazole. The protein was eluted with buffer B containing 350 mM imidazole. The histidine tag was removed by digestion with thrombin (100 NIH units) in 50 mM Tris (pH 7.5), 150 mM NaCl for 1 h at 25°C. The final purification step was hydrophobic interaction chromatography as described in Ref. 10 with the following modifications. The BCATm-containing fraction from the nickel-NTA column was incubated with 5 mM ␣-ketoisocaproate in 100 mM potassium phosphate at pH 7.5 on ice for 5 min before loading onto the column, and the elution gradient began at 35% saturated ammonium sulfate. The concentration of BCATm was determined from the absorbance at 280 nm using the extinction coefficient of 67,600 M Ϫ1 cm Ϫ1 per monomer.
For human BCATc in either the pTrcHis or pET-28a vector, extraction and nickel-NTA purification steps were identical to BCATm, except that urea was not added to any buffers. Following nickel-NTA chromatography, purification was carried out using hydroxyapatite chromatography. The BCATc containing fraction from the nickel-NTA column was loaded onto the hydroxyapatite column in 10 mM potassium phosphate, pH 7.0. Human BCATc was eluted with 200 mM potassium phosphate, pH 7.0. The final purification step was ion exchange chromatography using a Mono-Q column (Pharmacia Biotech Inc.) as described in Ref. 10 except that the buffer was 10 mM potassium phosphate, pH 8.0. The histidine tag remained on the recombinant protein when expressed from the pTrcHis vector, but when the pET-28a construct was used, the added histidines were cleaved with thrombin as described for BCATm. The concentration of BCATc with and without the histidine tag was determined from the absorbance at 280 nm using the extinction coefficients of 80,000 M Ϫ1 cm Ϫ1 and 86,300 M Ϫ1 cm Ϫ1 per monomer, respectively.
Storage Conditions-Human BCATm was extremely sensitive to oxidation and required the presence of 100 mM DTT to remain fully active. If samples were purged with argon every other day, the enzyme retained significant activity for periods of up to 2 weeks at 4°C. When stored in 50% glycerol at Ϫ20°C with argon purging, BCATm was stable for as long as 4 months. BCATc was not as sensitive to oxidation and could be stored in the presence of 10 mM DTT without argon purging. BCATc was stable for about 2 weeks at 4°C and for 3-4 months at Ϫ20°C in 50% glycerol.
Amino Acid Analysis-The extinction coefficient for each isoenzyme was calculated from the amino acid composition of each protein. The N terminus of each recombinant protein was verified by amino acid sequence analysis. Both composition analysis and N-terminal sequencing were performed by the Protein Analysis Core Laboratory of the Comprehensive Cancer Center of Wake Forest University.
Spectrophotometric Measurements-Absorption spectra were taken with a Beckman DU 640 spectrophotometer. CD measurements were carried out with a JASCO J-720 spectropolarimeter equipped with a variable temperature accessory. The enzymes were converted to the PLP and PMP forms by incubation with the appropriate substrate followed by dialysis. The instrument was calibrated with ammonium-D-camphorsulfonate. CD spectra in the near UV region were measured in a 1-cm quartz cylindrical cuvette at a protein concentration of 1 mg/ml. In the far UV region, CD spectra were acquired using a protein concentration of 0.5 mg/ml in 5 mM potassium phosphate buffer and a 0.05-cm path length. The final spectra were the average of four accumulations. The CD spectra of the PLP and PMP forms of BCATm and BCATc were analyzed using the convex constraint algorithm described in Refs. 23 and 24. In all cases, the targeted protein CD spectrum was added to the appropriate data set, which contained 30 proteins, and the enlarged spectral set (30 ϩ 1) was analyzed by CDANAL software (Jasco). Based on the quality of the fit, the number of pure components was set either at three or four, and the contribution of each element was obtained from the conformational weight matrix. Sigma values varied from 0.8 to 1.5.
Sulfhydryl Group Titrations-5-15 nmol of protein were dissolved in 1 ml of buffer containing 50 mM Tris-HCl (pH 7), 1 mM EDTA, and 6 M guanidine hydrochloride. The reaction was initiated by the addition of 200 l of 10 mM DTNB at room temperature, and the absorbance at 412 nm was monitored for 15 min. The amount of free thiol was calculated from the liberated 2-nitro-5-thiobenzoate anion using a molar extinction coefficient of 14,150 (25). The same procedure was repeated in the absence of denaturant to determine the number of cysteines accessible to solvent. Cysteine titrations were also performed in the presence of either 10 mM ␣-ketoisocaproate or isoleucine.
Branched-chain Aminotransferase Assay and Steady State Kinetics-Branched-chain aminotransferase activity was measured at 37°C in 25 mM potassium phosphate buffer, pH 7.8, using 1 mM ␣-keto [1-14 C] isovalerate and 12 mM isoleucine as described (9,10). A unit of enzyme activity was defined as 1 mol of valine formed per min at 37°C. In the kinetic studies, bovine serum albumin was omitted, and the assay contained 2 mM DTT and 12.5 mM EDTA.
Reaction rates with the branched-chain amino acid/␣-ketoglutarate pairs were determined from the formation of glutamate, which was assayed fluorimetrically by a slight modification of the method de-scribed by Williamson and Corkey (26). Briefly, excess ␣-ketoglutarate was removed from the neutralized sample by the addition of either 10 mM (BCATm) or 15 mM (BCATc) H 2 O 2 . Samples were incubated at room temperature for 5-15 min before an aliquot (250 l) was added to 1 ml of assay buffer. NADH fluorescence was determined with an excitation wavelength of 340 nm using a FOCI Mark I spectrofluorimeter (Farrand Optical Co., New York). The kinetic parameters of the amino acid reactions were determined by holding the concentration of ␣-ketoglutarate constant at 5 mM for BCATm and 10 mM for BCATc. For the ␣-keto acid/glutamate pairs, reaction rates were determined from the formation of 1-14 C-labeled branched-chain amino acid as described previously (7,10). The glutamate concentration was kept constant at 100 mM as the high salt concentration affected the kinetic parameters. The data were fit to the following velocity equation.
The K m of ␣-ketoglutarate or glutamate was first estimated by varying the concentration of each branched-chain amino or ␣-keto acid at approximately 5 times K m . For each substrate pair, data were collected from six or seven concentrations of amino acid or ␣-keto acid. Kinetic parameters are the means and standard errors of 3-5 separate determinations from two different BCATm and BCATc preparations. In separate experiments, the apparent K m for ␣-ketoglutarate and glutamate was determined by varying the concentrations of both the amino acid and ␣-keto acid substrates simultaneously. The data were fit to the initial velocity rate equation pertaining to a ping-pong kinetic mechanism (27) via nonlinear regression in the approach outlined by Cleland (28).
In this equation, the parameters K a and K b represent the Michaelis constants for substrates A (amino acid) and B (␣-ketoacid), respectively.
Analytical Ultracentrifugation-The equilibrium sedimentation experiments were performed in the Analytical Ultracentrifugation Facility at Wake Forest University. Using a Beckman Optima XL-A analytical ultracentrifuge, BCATm and BCATc in 25 mM Tris (pH 7.5), 150 mM KCl were centrifuged at 7500, 9000, 11,000, 14,000, and 40,000 rpm. Data were gathered after 12 and 14 h at each speed. The data were examined to verify that equilibrium was established. Partial specific volumes were calculated based on the amino acid composition of each protein as described by Laue et al. (29). Data analysis was performed using Origin software provided by Beckman.
Miscellaneous Procedures-SDS-PAGE was carried out according to Laemmli (30) in 10% gels as described in Wallin et al. (10).

RESULTS AND DISCUSSION
Expression and Purification-The human BCATm and BCATc have been cloned (12,13). Mature BCATm consists of 365 amino acids with a calculated molecular mass of 41,300 daltons, while BCATc consists of 385 amino acids with a calculated molecular mass of 42,800 daltons. A comparison of the primary sequences of these proteins reveals that they are 58% identical to each other (see Fig. 1) and 74% (BCATc) and 82% (BCATm) identical to their respective rat enzymes. Identity is found throughout the overlapping primary sequences including the C terminus. The most significant differences in primary sequences are found in the N-terminal portion (BCATc, residues 1-50) of the proteins.
In initial attempts to overexpress BCATm in a variety of expression plasmids, low expression, inactive, or insoluble protein was observed. The BCATm cDNA was then cloned into the pET-28a vector, which introduced a six-histidine residue tag onto the N terminus of the protein. High levels of expression of active protein were obtained, but the majority of the protein was still found in inclusion bodies. In addition, the soluble protein would not bind to the nickel-NTA resin. This suggested that the N terminus of the BCATm fusion protein was somewhat buried, making the histidine tag inaccessible to the resin. Upon complete denaturation in 8 M urea, BCATm bound to the resin; however, active enzyme could not be recovered either by gradual refolding of the protein while it remained on the resin or by dialysis versus urea-free buffer. In 4 M urea, the tertiary or quaternary structure of the protein was altered enough to permit binding of the N-terminal histidine tag to the resin. The far UV CD spectrum of BCATm (data not shown) showed no detectable change upon increasing urea concentrations from 0 through 4 M, thus indicating that the secondary structure of BCATm remained intact. In 4 M urea, the specific activity of the enzyme was reduced about 50% but could be restored after gradual removal of the denaturant. The use of 4 M urea also increased the yield of recombinant protein. Specific activity of purified BCATm after thrombin cleavage was 88 Ϯ 6 units/mg of protein (n ϭ 3). The N-terminal sequence of the recombinant protein carries the additional amino acids gshm (these additional residues are listed throughout in lowercase italic type) as verified by amino acid sequencing (see Fig. 1). SDS-PAGE of the purified recombinant BCATm protein is shown in Fig. 2A.
For BCATc, initial attempts with expression plasmids driven by the highly processive T7 RNA polymerase were problematic, because the protein was found primarily in inclusion bodies. With the lower level of expression found with the pTrcHis construct, the solubility of the recombinant protein was increased. Recently, we have expressed BCATc using the pET-28a vector/T7 expression system and removed the histidine tag. Although some protein was still found in inclusion bodies, the high yield in the pET vector/T7 system permitted purification of sufficient protein for kinetic and structural analysis. SDS-PAGE of the purified recombinant BCATc proteins is shown in Fig. 2, B and C. Unlike BCATm, recombinant BCATc readily bound to the nickel-NTA resin under native conditions, indicating that the N terminus of BCATc was exposed to the solvent, and urea addition (4 M) did not improve recovery. Thus, the behavior of the recombinant proteins on the nickel-NTA resin suggests that in the N-terminal portion of the proteins, the tertiary structure of BCATm and BCATc is different. Specific activity of BCATc, with or without the histidine tag, was 124 Ϯ 9 units/mg of protein (n ϭ 5). Recombinant BCATc expressed using the pTrcHis construct (phBCATc) contained an additional 14 amino acid residues (mggshhhhhhgmas) at the N terminus of the recombinant protein that were contributed by the vector sequence. The protein expressed using the pET-28a vector construct (phBCATc2) contained six additional amino acids after thrombin cleavage. Both nucleotide sequencing of the phBCATc2 construct and N-terminal amino acid sequencing of the recombinant protein showed that the Nterminal extension was gshmac (see Fig. 1). The N-terminal sequence of the recombinant protein with the histidine tag (phBCATc) was also confirmed by direct amino acid sequence analysis. Preliminary experiments showed that the physical and kinetic properties of the two recombinant enzymes were similar, and both enzymes were used in subsequent experiments.
Sedimentation Equilibrium-Upon size exclusion chromatography, purified rat BCATm behaved anomalously and appeared to be a monomer, while rat BCATc was clearly a homodimer (9,10). To determine unequivocally the subunit composition of the human isoenzymes, equilibrium sedimentation experiments were performed with the recombinant proteins. After the data were subjected to curve fitting analysis, BCATm and BCATc showed average molecular weights of 81,500 Ϯ 2510 and 83,900 Ϯ 4800, respectively. These values are approximately twice the calculated monomer molecular weights of the recombinant proteins after thrombin cleavage, 41,700 (BCATm) and 43,400 (BCATc), identifying both enzymes as dimers in solution.
Spectral Analysis of Recombinant Human BCAT Isoenzymes-The absorption spectra of the recombinant enzymes at pH 7.5 are shown in Fig. 3. In addition to the peak at 280 nm, two peaks characteristic of bound cofactor are observed at 416 and 326 nm, indicating the pyridoxal and pyridoxamine forms, respectively. In the visible range, the spectra for each isoenzyme are essentially superimposable. The higher absorption of BCATc at 280 nm probably reflects the higher tyrosine and tryptophan content of this protein.  Fig. 4 shows the far and near UV CD spectra of the proteins. The far UV CD spectra of both BCATc and BCATm in their PLP and PMP forms have similar features with minima at 219 and 210 nm and a maximum at 193 nm (see Fig. 4A). ␣-Helical structure is normally characterized by the presence of two minima at 222 and 208 -210 nm and a maximum at 191-193 nm (31); therefore, ␣-helices are present in both of these pro-teins. ␤-Sheets are also present because a minimum occurs at 219 nm, which is close to the minimum (216 -218 nm) normally observed for ␤-forms (31). The similarity of the far UV CD spectra of the two proteins indicates that their secondary structure content is similar. In addition, secondary structure estimates calculated using the far UV CD spectra of the PLP form of the enzymes suggested an ␣-helix content of 38% for BCATm and 35% for BCATc. Predicted ␤-sheet was on the order of 20 -24%. As shown in Fig. 4, no major change occurred in the global structure of the proteins upon conversion of the PLP form to the PMP form, and predicted ␣-helix content was similar for the PMP forms (33% for BCATm and 36% for BCATc). The crystal structure of the PLP form of the smaller E. coli BCAT (34 kDa) revealed a secondary structure content of 39% ␣-helix and 41% ␤-sheet (21). The similarity of the predicted ␣-helix content in the human BCAT proteins and observed ␣-helix content in the bacterial enzyme structure is also consistent with the hypothesis that all of these forms of the enzymes have the same basic folded structure.
The near UV CD spectra of the PLP form of both BCATm and BCATc were dominated by a band at 421 nm (see Fig. 4B). This peak resulted from the presence of the pyridoxal group in an asymmetric environment. Despite the similarity of this peak, there was a major difference in the spectrum of BCATc compared with BCATm around 280 nm. While BCATc showed a peak at 285 nm with a positive ellipticity, BCATm showed a peak at 270 nm with a negative value. Upon conversion of the PLP form to the PMP form, the peak at 421 nm disappeared for both proteins. For BCATm, this was concomitant with the appearance of a weaker band at 326 nm as well as a major increase in the intensity of the band at 277 nm. When the PLP form of BCATc was converted to the PMP form, the positive ellipticity of the peak at 285 nm was changed to a negative value. Since the far UV CD spectra of the PLP form of BCATm and BCATc were very similar to their respective PMP forms (see Fig. 4A), the conversion of PLP to PMP did not seem to affect the global structure of the proteins. Consequently, the changes observed around 280 nm could be assigned to an al- tered asymmetric environment of the aromatic residues in the active site. Furthermore, based on the difference in intensity and sign of the ellipticity of peaks around 280 nm, the structure of the active site in BCATc appears different from BCATm.
DTNB Titration-Like their rat counterparts (9), human BCATm, unlike human BCATc, must be stored in a reducing environment, and all four enzymes require the addition of DTT to the assay for maximal activity, whereas the bacterial protein from Salmonella typhimurium, which contains three cysteine residues, is not sensitive to sulfhydryl reagents (32). The requirement for the presence of a reducing agent indicates a unique property of the mammalian aminotransferases, i.e. a relationship between the state of reduction of the cysteines in the proteins and activity of the enzymes. Therefore, to evaluate whether modification of sulfhydryl groups within these proteins altered their catalytic capacity, DTNB titrations were performed on both human enzymes.
Based on the primary sequence, BCATm contains six cysteine residues/monomer (see Fig. 1). As shown in Table I, two -SH groups/monomer were titrated with native enzyme. Upon denaturation, 5.7 -SH groups were titrated, indicating that no disulfide bonds were present. However, upon storage of BCATm for a few days in the absence of DTT, the number of titratable -SH groups decreased to 3.7, suggesting that a disulfide bond was gradually being formed. This process was accompanied by a decrease in the specific activity of the enzyme. Fig.  5 shows an SDS-PAGE of BCATm stored in the absence of DTT.
Lane A shows that a more compact structure gradually forms upon storage without a reducing agent. As shown in lane B, the addition of DTT prior to loading the sample on the gel led to the disappearance of the lower band; thus, this faster migrating band results from the formation of a disulfide bond. This is consistent with the observation that the number of titratable cysteines decreases over time in the absence of DTT.
The cytosolic isoenzyme contains 10 cysteines/monomer, yet only one -SH group was titrated in the native form of the histidine-tagged BCATc (see Table I). The phBCATc2 construct added an extra cysteine residue immediately preceding the N-terminal methionone (see Fig. 1). An additional -SH group was titrated in the native form after removal of the histidine tag, suggesting that this cysteine was accessible to DTNB. In the denatured form, 4.6 (histidine-tagged) and 6.6 (without histidine tag) -SH groups were seen, indicating that at least two disulfide bonds are present in BCATc. This conclusion is supported by the electrophoretic movement of the protein (see Fig. 5, lanes C and  D). In the absence of DTT, BCATc migrated faster than when DTT was present. This increased mobility under nonreducing conditions suggests that the enzyme maintains a more compact structure due to the presence of disulfide bonds (33).
The activity of rat BCATm is known to be inhibited by sulfhydryl reagents (9), so the effect of DTNB on activity of the human enzymes was investigated. Spectrophotometric titration of BCATm with a 100-fold molar excess of DTNB occurred rapidly, and human BCATm activity was inhibited 90 -100% after titration of both accessible -SH groups. By reducing the ratio of DTNB to enzyme to 5, the reaction could be slowed such that essentially one -SH group was titrated after 6 min. To determine the effect of this titration on BCATm activity, the protein was incubated with DTNB for 6 min in assay buffer without DTT. Then substrate was added, and activity was  a The extra -SH group titrated in this protein comes from phBCATc2 sequence that remains on the protein following removal of the histidine tag (see Fig. 1 followed for 20 min. Without DTT, the enzyme does not have maximal activity, so a sample under identical assay conditions without DTNB was used as a control. As shown in Fig. 6, BCATm activity was decreased 40% after incubation with DTNB. Therefore, it appears that labeling of one -SH group by DTNB is not enough to fully inactivate BCATm. When substrate was added along with DTNB in the spectrophotometric assay, no -SH groups were titrated. Further, after incubation with DTNB in the presence of substrate, BCATm activity was unaltered (see Fig. 6), indicating that substrate protects against labeling and inhibition. Therefore, the modified -SH group may lie in or near the active site of BCATm, and studies are now in progress to verify its location. As shown in Fig. 6, titration of the single accessible -SH group in the histidinetagged BCATc did not affect enzyme activity. Similar results were obtained with BCATc without the histidine tag (data not shown). This suggests that the modified residue is away from the active site. Since it is known that the catalytic mechanism of transamination does not require a cysteine residue, one could speculate that the loss of activity seen with BCATm is due to steric hindrance from the introduction of the bulky DTNB molecule in the active site. Another possibility is that this effect results from a conformational change in the enzyme. Possible candidate cysteines in BCATm are the two cysteine that are not conserved in the BCATc sequence (Cys 108 and Cys 168 , see Fig. 1). Ultimately, however, understanding the effects of sulfhydryl reagents on catalytic function and protein stability (see below) will require isolation of the derivatized peptides and knowledge of the crystal structure of both proteins.
The Effect of Ni 2ϩ Cations on the Activity of BCATc and BCATm-Affinity chromatography using nickel-NTA resin is based on the interaction of the histidine tag in the protein with nickel cations chelated to the resin. After elution from this column, human BCATm had low specific activity, which did not increase even after the second purification step of hydrophobic interaction chromatography. However, when EDTA was added to the assay, the specific activity increased to about 75 mol/ min/mg of protein or higher. Further, the addition of Ni 2ϩ cations to the assay led to 67% inactivation of the purified enzyme, while subsequent addition of EDTA restored activity completely. On the other hand, when DTT or ␤-mercaptoethanol was omitted from the assay, added Ni 2ϩ had no inhibitory effect on BCATm activity. One possible explanation for these results is that one or more disulfide bonds are reduced by either DTT or ␤-mercaptoethanol, allowing interaction of Ni 2ϩ with cysteine residues within the active site. This seems unlikely, because in freshly purified BCATm, titration with DTNB suggests that no disulfide bonds are present. Alternatively, one can hypothesize that empty coordinates of Ni 2ϩ cations bound in or near the active site of BCATm interact with the thiol groups of DTT or ␤-mercaptoethanol, leading to either steric hindrance or a conformational change.
Chemical Denaturation of BCATc and BCATm-Since BCATc contains two or three disulfide bonds, one expects to see a higher stability for this protein compared with BCATm; therefore, chemical denaturation experiments were performed to compare the two proteins. Fig. 7 shows the urea denaturation of BCATc and BCATm as monitored by CD spectroscopy. Human BCATm remains partially folded up to a urea concentration of 8 M, indicating high structural stability. On the other hand, the denaturation curve of BCATc appeared as a relatively broad transition with a midpoint of 4.1 M urea. Surprisingly, this suggests that the structural stability of BCATc is lower than BCATm. It is generally accepted that disulfide bonds can make a substantial contribution to the stability of proteins (33); however, for BCATc, it seems that this is not the case. Moreover, the broadness of the transition indicates that unfolding of BCATc is less cooperative than BCATm. Since chemical denaturation using urea or guanidine hydrochloride is irreversible, the free energy of denaturation could not be calculated. Nonetheless, the data suggest that there are differences in the stability of the secondary structure of the two proteins.
Kinetic Parameters-The steady state kinetics of BCATm and BCATc were examined, and the calculated Michaelis-Menten parameters are summarized in Table II. Transamination of branched-chain amino acids with ␣-ketoglutarate exhibited kinetics similar to what has been reported for the rat enzymes. Values calculated for the K m for leucine and isoleucine were around 1.0 mM or less with a significantly higher K m observed with valine. Differences between the two isoenzymes were also FIG. 6. The effect of incubation with DTNB on the activity of BCATm and histidine-tagged BCATc. Enzymes were preincubated with a 5-fold molar excess of DTNB for either 6 min (BCATm) or 12 min (BCATc). Substrate was added, and product formation was followed for 20 min. Values are expressed as a percentage of the control, which contained no DTNB, after 20 min. Product formation was also measured after preincubation with substrate and DTNB added simultaneously. Open bar, control sample; filled bar, DTNB only; hatched bar, substrate and DTNB. In a separate spectrophotometric titration assay, it was determined that after 6-min (BCATm) and 12-min (BCATc) incubation with a 5-fold excess of DTNB, one -SH group was titrated in either enzyme. found. First, the K m values for leucine and valine were lower for BCATc than for BCATm. Observed k cat values suggested that the mitochondrial enzyme turns over faster using leucine than isoleucine and valine, but k cat /K m values were higher for isoleucine with both enzymes, suggesting preference for this branched-chain amino acid. For reamination of the branchedchain ␣-keto acids, generally K m values for the ␣-ketoacids were about 2-fold lower for BCATc than for BCATm. With both isoenzymes, the K m for KIC was similar with either fixed substrate (glutamate or isoleucine), whereas the K m for KIV paired with glutamate was 2.5-fold higher than when the fixed substrate was isoleucine.
The possibility that the observed differences in kinetic parameters with KIV/Glu and KIV/Ile were a result of differences in ionic strength was investigated by adding KCl to the ␣-ketoacid/Ile assay buffer (see Table II). The addition of 150 mM KCl raised the apparent K m values for KIV and KIC 2-fold or more. Smaller effects were observed on k cat , resulting in a decrease in k cat /K m for both isoenzymes. The effect of added KCl on branched-chain amino acid deamination was also examined in a single set of experiments. Increases of 40 -50% in the K m values for Leu and Ile were found with BCATm. Both K m and k cat were increased with BCATc (data not shown). While the molecular basis for the effects of K ϩ and/or Cl Ϫ on the kinetic parameters of the BCAT isoenzymes is not yet understood, there is precedence for both cation and anion effects on PLP-dependent enzymes (35)(36)(37)(38)(39). The structural basis for the cation stimulation of tryptophan synthase (35), dialkylglycine decarboxylase (40), tryptophanase (41), and tyrosine phenol-lyase (42) have been investigated using x-ray crystallography, and cation binding sites have been defined. In both tryptophanase (43,44) and tyrosine phenol-lyase (45) from E. coli, cations like K ϩ were shown to induce and stabilize active conformations of these enzymes. With aspartate aminotransferase, at alkaline pH, it was shown that chloride anions (38) as well as dicarboxylic acids (46) can act as competitive inhibitors, possibly by ion pairing with positively charged residues in the active site that serve to bind the ␣-carboxylate group of the substrate.
In the absence or presence of KCl, k cat and k cat /K m values for the deamination of the branched-chain amino acids were higher for BCATc than BCATm, suggesting that BCATc turns over 2-5 times faster than BCATm. For reamination, generally k cat values between the two enzymes are similar; however, the markedly lower K m values for ␣-keto acids found with BCATc result in calculated k cat /K m values that are approximately 2-fold higher for KIC and 3-4-fold higher for KIV than found with BCATm. An equilibrium constant of 1.4 Ϯ 0.1 was also determined with BCATm, and this value agrees quite well with the value of 1.7 originally reported by Taylor and Jenkins (3).
Substrate Specificity-The amino acid preference of BCATc and BCATm was determined by examining the relative rate of transamination of [1-14 C]KIC with the branched-chain amino acids, branched-chain amino acid analogs, and other amino acids. Rates relative to leucine are presented in Table III. As expected, branched-chain amino acids were clearly the preferred substrates. Glutamate was a better substrate for BCATc than BCATm, 82% of control versus 38%, respectively. This difference could be attributed to a lower K m for glutamate for BCATc (13 mM) versus BCATm (24 mM). Human BCATc appeared to accept the five-carbon straight chain analog, norvaline, twice as readily as BCATm. Depending on the isoenzyme, L-alloisoleucine was transaminated at about 50 -80% of the rate of the natural substrate, L-threoisoleucine. D-Isoleucine, methionine, the aromatic amino acids, glutamine, alanine, and aspartate were not accepted by either enzyme. The range of ␣-keto acid substrates is shown in Table IV. Rates are shown relative to KIC and [1-14 C]leucine. The pattern of ␣-keto acids accepted by the human isoenzymes is virtually identical. As expected, the hydroxy acids of leucine (␣-hydroxyisocaproate) and valine (␣-hydroxyisovalerate) were not substrates (see Table IV). The straight chain ␣-keto acids were good substrates, with the five-carbon ␣-ketovalerate preferred over the six-carbon ␣-ketocaproate. With five carbons, ␣-ketovalerate has the same chain length as isoleucine and leucine but without branching. ␣-Ketobutyrate was transaminated poorly compared with KIV, showing that, when the carbon length was 4, branching increased affinity for the enzyme. Both ␣-ketocaproate and ␣-keto-␥-methiobutyrate have a carbon length of 6, but ␣-keto-␥-methiobutyrate has a sulfur atom substituted at the fifth carbon position. ␣-Ketocaproate was favored over the ␣-keto acid of methionine, suggesting that the introduction of the bulky sulfur atom decreases the affinity of the enzyme. Pyruvate and phenylpyruvate were not accepted by either human isoenzyme. This pattern of amino and ␣-keto acid preferences is very similar to what has been reported for the rat BCAT isoenzymes (9). Thus, although differences in apparent substrate affinity and rates of transamination between BCATm and BCATc suggest localized variations in active site architecture, overall structure appears conserved within the mammalian BCAT. CONCLUSIONS This paper presents the first overexpression and characterization of the human branched-chain aminotransferase isoenzymes. Our data show that the human BCAT share the basic characteristics of other known BCAT with respect to substrate specificity (1). Physical comparison of the recombinant proteins has also revealed subtle kinetic and physical differences in the two proteins with regard to stability, tertiary structure of the N-terminal regions of the proteins, and structure of the active sites. Although the structure of the E. coli protein has provided insight into the structure of the human BCAT, neither the E. coli nor the D-alanine aminotransferase structure appears to have sufficiently high sequence homology for use as a model in molecular replacement. 2 Thus, understanding the molecular basis for the differences between the BCATc and BCATm pro-teins and their bacterial counterparts will require knowledge of the crystal structure of the mammalian BCAT. These experiments are currently in progress.