Human Mitochondrial Branched Chain Aminotransferase Isozyme

Mammalian branched chain aminotransferases (BCATs) have a unique CXXC center. Kinetic and structural studies of three CXXC center mutants (C315A, C318A, and C315A/C318A) of human mitochondrial (hBCATm) isozyme and the oxidized hBCATm enzyme (hBCATm-Ox) have been used to elucidate the role of this center in hBCATm catalysis. X-ray crystallography revealed that the CXXC motif, through its network of hydrogen bonds, plays a crucial role in orienting the substrate optimally for catalysis. In all structures, there were changes in the structure of the β-turn preceding the CXXC motif when compared with wild type protein. The N-terminal loop between residues 15 and 32 is flexible in the oxidized and mutant enzymes, the disorder greater in the oxidized protein. Disordering of the N-terminal loop disrupts the integrity of the side chain binding pocket, particularly for the branched chain side chain, less so for the dicarboxylate substrate side chain. The kinetic studies of the mutant and oxidized enzymes support the structural analysis. The kinetic results showed that the predominant effect of oxidation was on the second half-reaction rather than the first half-reaction. The oxidized enzyme was completely inactive, whereas the mutants showed limited activity. Model building of the second half-reaction substrate α-ketoisocaproate in the pyridoxamine 5′-phosphate-hBCATm structure suggests that disruption of the CXXC center results in altered substrate orientation and deprotonation of the amino group of pyridoxamine 5′-phosphate, which inhibits catalysis.

The reaction is accompanied by interconversion of the cofactor between the PLP and the pyridoxamine 5Ј-phosphate (PMP) forms (7)(8)(9)(10)(11)(12). In the first half-reaction, the PLP form of BCAT reacts with the branched chain amino acid, and the reaction proceeds through a Michaelis complex, an external aldimine, quinonoid intermediate, ketimine, and finally the PMP form of the BCAT and the branched chain ␣-keto acid product. In the Michaelis complex, the ␣-amino group of the substrate must be deprotonated to form a new Schiff base with PLP. The migration of the proton on the ␣-amino group of the substrate to the short contact pair between the ␣-carboxylate and the phosphate of PLP reduces the electrostatic repulsion between the pair and renders the ␣-amino group of the substrate deprotonated (6). The second half-reaction is the reverse of the first half-reaction.
A second characteristic of the mammalian BCATs is the presence of a consensus sequence CXXC motif (17)(18)(19). In mitochondrial hBCATm, these cysteines can participate in a thiol-thiolate interaction and are responsible for the redox sensitivity of the enzyme (1,17). The cysteines of the CXXC center (Cys 315 and Cys 318 ) can form a disulfide bond in the presence of an oxidizing agent. Cys 315 is the sensor, whereas Cys 318 acts as the "resolving cysteine," permitting reversible formation of a disulfide bond (19). hBCATm enzyme activity is affected by the reduction state of the CXXC center. The reduced protein is fully active (3,19), whereas the oxidized enzyme is inactive (18,19). hBCATc activity is less affected by oxidation of its CXXC center (20). Prokaryotic BCATs do not contain the CXXC structural motif (6).
The structural and kinetic basis for the influence of the CXXC center on hBCAT catalysis is not known. To understand the molecular basis for catalytic inactivation of hBCATm by oxidation of the CXXC center and the structural role of this motif in hBCATm catalysis, we have solved the x-ray crystallographic structure of oxidized mitochondrial hBCATm, the C315A, C318A, and C315A/C318A mutants. Structural changes include an altered hydrogen bonding pattern concomitant with changes in the net turn dipole moment at the beta turn preceding the CXXC motif, altered flexibility of the interdomain loop, a change in the conformation of one of the residues lining the substrate binding pocket, and changes in the mode of substrate binding as a consequence of a mutation. These structural changes result in kinetic inactivation of the second half-reaction due to improper positioning of the ␣-keto acid substrate in the active site.
Protein Purification-The overexpression and purification of the hBCATm proteins for crystallization studies were performed according to the method of Conway et al. (18). For kinetic studies, the procedure was modified. Briefly, the pET-28a expression vector containing the hBCATm cDNA was transformed into BL21 (DE3) cells. The E. coli were cultivated in LB medium containing 30 g/ml kanamycin and 100 M pyridoxine-HCl. The histidine-tagged proteins were extracted and purified using nickel-nitrilotriacetic acid resin (Qiagen, Chatsworth, CA). The histidine tag was removed by digestion with thrombin (100 National Institutes of Health units) at 30°C for 1 h followed by buffer exchange into the storage buffer (50 mM Tris-HCl at pH 7.5, 150 mM sodium chloride, 5 mM DTT, 1 mM EDTA, 1 mM ␣-ketoisocaproate, and 5 mM glucose). Purified hBCAT proteins were filtered using a 0.22-m filter and stored at Ϫ20°C in 30% glycerol in the storage buffer. The mutant proteins are as stable as wild type hBCATm (WT-hBCATm).
Radioactive hBCATm Assay-The radioactive hBCATm assay was performed as described previously (18). The activity of the enzyme was measured at 37°C in 25 mM potassium phosphate buffer at pH 7.8 with 5 mM DTT, using 1 mM ␣-keto[1-14 C]isovalerate and 12 mM isoleucine. A unit of enzyme activity was defined as 1 mol of valine formed/min at 37°C. The reported values are the average of three independent assays.
Preparation of the PMP Form of hBCATm-The purified enzyme is in the PLP form. To convert hBCATm to the PMPform, 100 nmol of purified hBCATm protein were exchanged into 50 mM NaHEPES (pH 8.0) containing 0.1 M KCl and 0.1 mM EDTA and incubated with 20 mM leucine for 15 min at 25°C. The leucine was removed using a PD-10 column as described by the manufacturer (Amersham Biosciences). The protocol was then repeated. Conversion to the PMP form was confirmed spectrophotometrically, and hBCATm activity was determined using a spectrophotometric assay (see below). The specific activity of PMP-hBCATm was the same as the specific activity of the PLP form of the enzyme.
Preparation of the Oxidized Form of hBCATm-Oxidation of hBCATm was conducted by incubation with hydrogen peroxide at 25°C. A 60 M solution of the enzyme was first exchanged into 50 mM KHEPES (pH 7.5) containing 0.1 M KCl and 0.1 mM EDTA using a PD-10 column (Amersham Biosciences). Enzyme (50 M) was incubated with 500 M hydrogen peroxide for 2 h. Then the enzyme reaction mix was exchanged into the same buffer without hydrogen peroxide. The activity of oxidized hBCATm was 1.5 units/mg of protein compared with the reduced hBCATm activity of 106 units/mg of protein (99.5% inhibition). Loss of the free cysteines as a result of disulfide bond formation was determined by 5,5Ј-dithiobis(2-nitrobenzoic acid (DTNB) titration (18).
Preparation of WT and Mutant Enzymes-Site-directed mutagenesis for the single mutants C315A and C318A and double mutant C315A/C318A were performed according to the QuikChange kit protocol. The pET-28a plasmid, carrying the human cDNA gene for BCATm, was used as the template for the desired mutations. Primer design, PCR, expression, and purification of the proteins were described previously (19). The recombinant protein contains four additional residues (GSHM) at the amino terminus. These residues were not well defined in the electron density map and are not included in the deposited structures. Mass spectroscopy was used to confirm the molecular masses of the mutants and the oxidized proteins.
Determination of Protein Concentration-The concentrations of WT and mutant hBCATm proteins were determined spectrophotometrically at 280 nm and at pH 8.0. Molar extinction coefficients used were ⑀ M ϭ 67,600 M Ϫ1 cm Ϫ1 for the PLP form of the enzyme and ⑀ M ϭ 67,500 M Ϫ1 cm Ϫ1 for the PMP form of the enzyme. The calculation was based on the contribution of tryptophan and tyrosine residues to the apparent molar absorption at 280 nm, and the estimated ⑀ M was 5800 M Ϫ1 cm Ϫ1 and 1450 M Ϫ1 cm Ϫ1 for tryptophan and tyrosine, respectively. The ⑀ M for the coenzyme moiety was determined from the absorption change upon binding of coenzyme to apoenzyme, which was 2000 M Ϫ1 cm Ϫ1 for PLP and 700 M Ϫ1 cm Ϫ1 for PMP, respectively.
Sulfur Chemistry and Thiol Oxidation-DTNB titration was performed to quantify the free thiol groups in hBCATm. The molar extinction coefficient of 13,600 M Ϫ1 cm Ϫ1 for 2-nitro-5thiobenzoate (TNB) was used for all calculations (19). A 20 mM DTNB stock solution was prepared, and reactions were performed in 50 mM HEPES, 0.1 mM EDTA, pH 7.5, for 20 min at room temperature. Final DTNB concentrations were always in 100-fold molar excess over the thiol concentration.
Kinetic Analysis-Fast reaction kinetics were performed using a stopped-flow SX.17MV spectrophotometer (Applied Photophysics, Leatherhead, UK). Enzyme (concentration 40 M after mixing) and varied concentrations of each substrate were reacted in 50 mM NaHEPES at pH 8.0, 0.1 M KCl, and 1 mM EDTA. Time-resolved spectra were collected using the SX.17MV system equipped with a photodiode array accessory and the XScan (version 1.0) controlling software (Applied Photophysics). The reaction temperature was set at 25°C using a circulating water bath to maintain a constant temperature in the water jackets surrounding the mixing syringes and the mixing cell. Enzyme (80 M) and substrate were placed in different syringes, and the reaction was initiated by injecting the fresh enzyme into a mixing chamber at a 1:1 ratio. The dead time was 2.0 ms under a pressure of 600 kilopascals. The absorption maximum was fixed depending on the observed transient spectral changes. Photomultiplier voltage was controlled automatically, and the times used to measure absorption changes were determined experimentally. The apparent rate constant (k app ) for the monoexponential absorption changes was calculated with the program provided with the instrument at 418 nm. For each sample, a minimum of three rate constants were collected to minimize the error. The nonlinear least square fitting of the collected data were performed, and the obtained k app values were fit to the following equation using IGOR Pro software (WaveMatrics Inc.) using nonlinear regression analysis, where k cat and K d are the turnover number and substrate dissociation constants for the "half" reactions, respectively, and [S] is the substrate concentration.
The direct titrations of the substrates with the enzymes were performed using a Beckman Coulter DU-800 spectrophotometer at 25°C equipped with a Beckman Coulter temperature controller. The enzyme was exchanged into 50 mM HEPES (pH 8.0, KOH) buffer containing 5 mM DTT using a PD-10 column, and the enzyme concentration was adjusted to 30 M. Substrates were added, and spectral changes were determined. The equilibrium constants (K eq ) for hBCATm substrate reactions were evaluated by monitoring the absorption maximum changes at 418 nm and fitting the values to the following equation by nonlinear regression analysis (21), where A o is the observed absorbance, A␣ and Ai are the fitted values of the initial and final absorbance, respectively, [S] is the substrate concentration, and [E] is the enzyme concentration. All calculations and analyses were performed using IGOR Pro software (WaveMatrics).
Crystallization, Soaking, and Data Collection-Crystals of the mutants of human recombinant hBCATm were obtained using the vapor diffusion method of hanging drops at room temperature similar to the wild type protein reported earlier (1,17). The drop consisted of 5 l of protein solution and 5 l of the reservoir. The drop was equilibrated against 1 ml of the reservoir. The protein solution contained 2.5 mg/ml mutant proteins in a solution of 50 mM HEPES (pH 7.0), 20 mM DTT, and 50 mM EDTA. Yellow crystals grew in grids set up with the wild type crystallization condition containing 22-30% polyethlylene glycol 1500, 100 mM HEPES (pH 6.9 -7.2), and 20 mM DTT. Some of the crystals were in the monoclinic form, space group P2 1 , and some in the orthorhombic form with space group P2 1 2 1 2 1 symmetry. There was a dimer in the asymmetric unit in both forms (Table 1).
For data collection, the crystals were first soaked in a solution of the crystallization condition having 10% glycerol added to it so as to get a glassy freeze. The crystals were picked up with a loop (Hampton research nylon loop, size 0.5 mm) and frozen directly in the liquid nitrogen cold stream. In order to see the complex with N-methylleucine, the crystals were soaked in the crystallization solution with an excess amount of the substrate mimic. The crystals did not change color from yellow to clear with time, indicating that N-methylleucine forms the Michaelis complex with the mutant proteins. In order to solve the structure of the oxidized protein, the protein crystals obtained under standard conditions were incubated with the standard crystallization buffer without DTT and with the addition of 3% hydrogen peroxide. The crystals were flash frozen as described above.
Structure Determination-The molecular replacement method was used to solve the mutant crystal structures with the earlier solved PLP form as the search probe in the program AMoRe (23), part of the CCP4 program suite (24). The program CNS (25) was then used for all further refinements. The slow cool protocol was used to minimize model bias, and O (26,27) was used for examination and manual adjustment of the structure during the refinement. Several cycles of positional refinement and isotropic B-factor refinement were performed after every model building cycle. Electron density maps confirmed the mutated side chains (Fig. 1). The pyridoxal phosphate was removed from the initial refinement cycles to get an unbiased view of the active site. Toward the end of the refinement, -weighted difference Fourier maps and 2F o Ϫ F c maps clearly showed the changes that had occurred at the active site as a result of the mutation. The cofactor is covalently linked to the active site lysine, Lys 202 , in all of the mutant structures. In the N-methylleucine soaked crystals of C315A enzyme (C315A-NML), the active site density maps indicated that the N-methylleucine has bound as a Michaelis complex in both the monomers.
In the oxidized enzyme (hBCATm-Ox), Cys 315 and Cys 318 are seen to be part of a disulfide (Fig. 1A). Cys 108 , which is on the surface of the enzyme and does not affect hBCATm activity (19), is overoxidized to the sulfonic acid. The N terminus loop, residues 15-32, is disordered in monomer A and cannot be traced. In monomer B, the conformation of Phe 30 has changed significantly. A HEPES molecule is stabilized with its cyclohexyl ring in the substrate binding pocket, and the sulfate end plugs the access to the active site. In the double mutant C315A/ C318A structure, an ␣-ketoisocaproate molecule is seen tightly bound as a Michaelis complex in the active site of one of the monomers (Fig. 5). ␣-Ketoisocaproate is present in the enzyme storage buffer. In the single mutant C318A, a HEPES molecule is seen bound on the surface of the protein. In monomer B, a sulfate ion occupies the substrate backbone carboxylate site. In monomer A, two acetate molecules are seen in the substrate backbone carboxylate and glutamate side chain distal carboxylate binding site. In the structure of the N-methylleucine-bound C315A mutant, an acetate molecule is seen bound in the latter site in both of its monomers. A HEPES molecule is seen on the surface of the protein in this mutant as well.
Water molecules were automatically located using CNS (25) and manually checked in O (26,27). Acetate, HEPES, and glycerol molecules were modeled manually into -weighted (2F o Ϫ F c ) maps. The refinement parameters with the R-factors and average B-factors of the various structures are listed in Table 1.
Analysis of the stereochemistry of both of the structures with PROCHECK (28) showed that 98.1% of the main chain atoms fall within the core and generously allowed regions of the Ram-achandran plot. Gln 316 has good electron density and is seen in the disallowed region of the Ramachandran map. The other residues that might be in the disallowed regions have poor density. None of these are close to the active site and part of surface loops.

Modeling of N-Methylleucine into PLP-WT hBCATm and ␣-Ketoisocaproate into Oxidized and Reduced PMP-hBCATm-
The coordinates of the PLP-hBCATm protein were used as the reference molecule (Protein Data Bank code 1EKF), and the N-methylleucine-bound C315A mutant of PLP-hBCATm was superimposed on it using the Swiss PDB viewer program. The transformed coordinates of the N-methylleucine molecule from the C315A mutant structure was appended into the PLP-hBCATm structure and used as the starting N-methylleucinebound PLP form WT-hBCATm model. Similarly, the coordinates of the oxidized PLP-hBCATm protein were used as the reference molecule, with the PMP-hBCATm structure (Protein Data Bank code 1KTA) and the Michaelis complex structure of ␣-ketoisocaproate-bound double mutant structure superimposed on it using the Swiss PDB viewer program. The transformed coordinates of the ␣-ketoisocaproate molecule from the double mutant structure and the PMP cofactor from 1KTA were appended into the oxidized protein and used as the starting oxidized ␣-ketoisocaproate-bound PMP form hBCATm model. For a model of ␣-ketoisocaproate bound to the reduced PMP-hBCATm structure (Protein Data Bank code 1KTA), the latter was used as the reference, and the Michaelis complex structure of ␣-ketoisocaproate and C315A/C318A was superimposed on it to get the transformed coordinates for the substrate. All of the models were subjected to four iterations and 300 cycles for each iteration of conjugate gradient energy minimization using the program CNS (25). The resulting model was used for further analysis.

Temperature Effects on hBCAT
Kinetics-In general, each parameter in a kinetic model is temperature-dependent, and that is why it is best to perform all of the kinetic studies at physiological temperature (37°C). However, hBCATm was partially denatured (PLP form of hBCATm) or converted to apoenzyme (PMP-form of hBCATm) at this temperature (data not shown) under the conditions of the spectrophotometric assay. In vivo it is likely that chaperones or other protein interactions protect hBCATm from thermal denaturation so that hBCATm does not work alone in the physiological system. There are several proteins that can associate with hBCATm in vitro. 5 Therefore, the kinetic analysis was conducted at 25°C.
Effect of Oxidation on Reaction Equilibrium Constants-Reduced hBCATm loses enzyme activity upon reaction with hydrogen peroxide, and loss of enzyme activity correlates with the formation of the intrasubunit disulfide bond between Cys 315 and Cys 318 (18,19). To determine the molecular basis for the effect of oxidation of the CXXC center on hBCATm activity, the reaction equilibrium constants (K eq ) for the enzyme-substrate reactions were determined using the spectral changes induced in the coenzyme by substrate association at pH 8.0 and at a fixed enzyme concentration (30 M). As shown in Fig. 2A, in the absence of substrate, there is an absorption maximum at 418 nm, which represents the PLP ketoenamine ( max ϭ 418 nm) form of hBCATm. Upon the addition of amino acid substrate, the absorbance peak at 418 nm decreased, and a peak appeared at 330 nm. The shift in max from 418 to 330 nm results from the disappearance of the PLP form of the cofactor and appearance of the PMP form ( max ϭ 330 nm) of the cofactor. With the PMP form of the enzyme, the order was reversed (i.e. upon the addition of ␣-keto acids, the PLP form of the cofactor PLP was formed with concomitant lowering of the PMP form) (Fig. 2B). Changes in the 418 nm absorbance were used for the substrate titration and kinetic analysis, because the 418 nm peak represents a single absorbing species (PLP ketoenamine), whereas the PMP maximum at 330 nm (PMP form) also contains the PLP enolimine species ( max ϭ 340 nm).  The calculated K eq values for the hBCATm substrate reactions are shown in Table 2. Isoleucine and leucine exhibit similar equilibrium constants (ϳ1) and show higher affinity to the reduced enzymes than the shorter branched chain amino acid valine. All branched chain amino acids have the higher reactivity than the dicarboxylic amino acid substrate glutamate.
The branched chain ␣-keto acids exhibited ϳ10 -15-fold greater reactivity with the active site of the reduced PMP form of the hBCATm isozyme than the branched chain amino acid substrates to the reduced PLP form of the enzyme ( Table 2). As observed with the amino acid substrates, the monocarboxylate branched chain ␣-keto acid substrates have lower equilibrium constants than the dicarboxylate substrate ␣-ketoglutarate. Differences in K eq values between glutamate and ␣-ketoglutarate were smaller than observed differences between branched chain amino acid and ␣-keto acid substrates. Based on calculated K eq values, reaction affinity of ␣-ketoglutarate to the reduced form of the hBCATm isozyme was ϳ7-fold higher than the reaction with glutamate.
The overall activity (sum of both half-reactions) of hBCATm determined by radioisotope assay showed that the oxidized enzyme is Ͼ95% inactive (18). Therefore, to determine if inactivation resulted from inhibition of one or both half-reactions, the kinetics of each half-reaction were investigated separately using the oxidized PLP and PMP forms of hBCATm. Oxidation of hBCATm increased the equilibrium constant of all amino acid substrates for the PLP form of the enzyme. Thus, whereas the addition of amino acids to the PLP form of oxidized hBCATm resulted in similar spectral changes as observed for the reduced form of enzyme, calculated substrate K eq values for leucine and isoleucine were ϳ9-fold higher than observed with the reduced enzyme ( Fig. 2C and Table 2). With valine and glutamate, calculated K eq values were 6.7 and 15.2 for the reduced form of the enzyme, respectively, and these values are both greater than 40 with oxidized enzyme. In contrast to the forward reaction, there was no reaction of ␣-keto acid substrate with the oxidized PMP form of hBCATm. No spectral change was observed upon the addition of ␣-keto acid substrates even at a concentration of 20 mM (Fig. 2D). Therefore, the spectroscopic assay shows that inactivity of oxidized hBCATm depends on the enzyme form (PLP-form or PMPform). The enzyme is less active in the first half-reaction and completely inactive in the second half-reaction.
There are several possible explanations for the observed inactivity of the oxidized PMP form of hBCATm. Either the ␣-keto substrate was unable to bind to the active site of the enzyme or the substrate could bind to the active site without any chemistry with PMP. To investigate this question, the oxidized enzyme was separated from the enzyme substrate reaction mixture. Then DTT was added to the enzyme, and spectroscopic changes were monitored. The spectrum of the PMP enzyme changed to the spectrum of PLP enzyme within 30 min (Fig. 3). Thus, ␣-keto acid substrate bound to the active site of the oxidized PMP form of hBCATm and remained bound during buffer exchange that removed unbound ␣-keto acid. Although bound, the results suggest that the substrate was not able to interact properly with the PMP cofactor for catalysis to occur until the addition of DTT, which reduced the disulfide bond between Cys 315 and Cys 318 . Reduction of the disulfide bond allows the active site to regain the proper conformation for chemistry to occur between the ␣-keto acid and the PMP form of the enzyme.

Pre-steady State Kinetic Analysis of the Half-reactions of hBCATm, the CXXC Mutant Enzymes, and Oxidized hBCATm-
To understand the role of the CXXC center in catalysis, we generated the C315A, C318A, and C315A/C318A mutant enzymes. Pre-steady state kinetic analysis was used to determine the effect of these mutations on the rate constants for both half-reactions. Spectrophotometric measurements were performed under single-turnover conditions, and the monoexponential absorption changes were monitored at 418 nm. The k app rate constants were fit using Equation 1 to obtain the k cat and K d values that are shown in Table 3. The pre-steady state kinetic constants are in good agreement with the previously  determined steady state kinetic constants of the reaction of reduced WT-hBCATm with amino acids (19). For the halfreaction with the PLP form and amino acid substrate, the K d values of the branched chain amino acids are increased by 5-10-fold, and k cat values are decreased by 3-4-fold upon oxidation of hBCATm (hBCATm-Ox) ( Table 3). Oxidation of hBCATm had less effect on enzyme specificity (k cat /K d ) for glutamate than branched chain amino acids ( Table 3). Mutation of the CXXC cysteines predominantly affected the kinetic efficiency of C315A and double mutant enzymes in catalyzing the first half-reaction with branched chain amino acid substrates. Substrate binding to the C315A mutant was generally weaker, with K d values ranging from Ͻ10% to 50% higher than observed with wild type enzyme. Substrate preferences (k cat /K d ) of the C318A were very similar to the substrate preference of WT-hBCATm. These data indicate that the amino acid substrates can bind properly to the mutant enzymes. The rate of catalysis (k cat ) with branched chain amino acids was ϳ5-fold lower in C315A and double mutant enzymes. Mutation of Cys 318 to alanine (C318A) had little effect on k cat and K d values for the amino acid substrates, since they were very similar to values for the WT enzyme. The mutations had less effect on the kinetics of the half-reaction with glutamate than with the branched chain amino acids, and changes were observed primarily in k cat values. For example, the specificity constants (k cat /K d ) were 8 -10-fold lower in C315A and double mutants for the reaction with branched chain amino acids, whereas this value was 2-fold lower in the mutants for the reaction with glutamate. The specificity constants of C318A mutant enzyme were similar to the WT-hBCATm values.
As expected from the spectroscopic properties of BCATm-Ox, the predominant effect of mutation of the CXXC center was on the second half-reaction rather than the first half-reaction. As shown in Table 3, K d values for branched chain ␣-keto acids were ϳ5-6-fold lower than the values for branched chain amino acids with the PLP form of the WT enzyme. The ␣-keto acids of leucine and isoleucine, ␣-ketoisocaproate and ␣-keto-␤-methylvalerate, had the highest affinity for the PMP enzyme when compared with the ␣-keto acids of valine and glutamate. The pre-steady state kinetic constants of the C318A mutant were less affected than the kinetic constants of the other two mutant enzymes. The k cat values of the C318A enzyme with the branched chain ␣-keto acids were decreased by ϳ2-3-fold, and K d values of the branched chain ␣-keto acids were increased by ϳ4 -5-fold compared with values for the WT enzyme. With the C315A and double mutant enzymes, decreases in the k cat values compared with WT-hBCATm were significantly greater (ϳ8 -9-fold). Calculated K d values of the branched chain ␣-keto acid substrates were increased more than 15-fold for the ␣-keto acids of leucine and isoleucine and ϳ8-fold with the ␣-keto acid of valine compared with WT-hBCATm. As observed with glutamate, the effect of mutation on the kinetics of the reaction with ␣-ketoglutarate was comparatively less than with branched chain ␣-keto acids. The specificity constant of the mutant enzymes for the reaction with branched chain ␣-keto acids was decreased by 10 -100-fold, whereas this value was decreased by only 50 -100% for the reaction with ␣-ketoglutarate. The dramatic effect of oxidation of the CXXC center in hBCATm on the second half-reaction and the effects of individual and double mutation of the CXXC cysteines on the presteady state kinetics suggest that the proper positioning of Cys 315 is important for the reaction of ␣-keto acid substrate with the PMP enzyme.
Structures of the C315A Mutant in the PLP (Protein Data Bank Code 2HGX) and Michaelis Complex with N-methylleucine (C315A-NML) (Protein Data Bank Code 2HG8)-The published structures of hBCATm and its reaction intermediates (1,17) and E. coli BCAT (16) (Protein Data Bank code 1I1L) provide the molecular information needed to describe the substrate orientation in each catalytic step. In the active site of both monomer A and B of the N-methylleucine-bound C315A mutant enzyme (hBCATm C315A-NML) structure (Fig. 4, A  and B) the active site Lys 202 is covalently bound to the PLP cofactor, and it has not been affected by the single mutation of Cys 315 . Ala 315 and Cys 318 , which immediately follow a ␤-turn (residues 311-314) are located just outside of the active site. As shown in the overlay of the active site of monomer B of hBCATm C315A-NML and monomer B of N-methylleucinebound WT-hBCATm modeled from the structure of WT-hBCATm (see "Experimental Procedures"), the N-methylleucine is not suitably positioned for the reaction to proceed (Fig. 4C). Similar results were obtained for monomer A (not shown).
All of the substrate binding residues in WT-hBCATm and the mutant are almost superimposable with a root mean square deviation value of 0.48 Å 2 . Mutation of Cys 315 to Ala does not affect the conformation of Cys 318 . Without the thiol/thiolate interaction, van der Waals interactions with Val 182 side chain and Met 241 hold the Cys 318 side chain in the same position observed in WT-hBCATm. There is a rotation of the peptide plane following Ala 315 compared with its position seen in the wild type structures. This rotation results in differences in the hydrogen bond interactions at the ␤-turns preceding and following Ala 315 (Cys 315 in WT-hBCATm). Three new hydrogen bonds are seen in the mutant protein (Fig. 4, A and B). First, the backbone amidic nitrogen at Gln 316 interacts with the carbonyl oxygen at Gly 312 . The backbone carbonyl oxygen at Cys 315 was hydrogen bonding with the backbone amide at Gly 312 in WT-hBCATm. Second, in the C315A mutant, the carbonyl oxygen at Ala 315 has formed a new hydrogen bond with the backbone amide at Val 317 . The side chain of Gln 316 , which was interacting with the side chain of Gln 353 and a water molecule in the wild type enzyme structures, has formed a new hydrogen bond with the backbone carbonyl oxygen at Thr 313 . The changes in the ␤-turn region preceding the CXXC motif seen here are representative of the cysteine mutant structures studied, C318A and C315A/C318A. The side chains of Thr 313 and Val 317 are seen in two possible conformations in monomer B. In the WT-hBCATm ketimine structure (1) and E. coli Michaelis complex structure (Protein Data Bank code 1I1L), Thr 313 plays a crucial role in stabilizing the carboxylate of the incoming substrate.
N-Methylleucine is displaced from the substrate binding pocket in monomer A (at a distance of 12.5 Å from the C4Ј atom of the PLP cofactor) and is bound close to the surface of the enzyme and has blocked the access to the active site (Fig. 4A). The backbone carboxylate interacts with the side chain of Gln 224 , and the side chain of the substrate mimic is in van der Waals contact with Phe 30 . The N-methyl group has hydrophobic interaction with Val 182 and Cys 318 . Thus, the methylleucine molecule has bound in the substrate binding pocket of the active site in monomer B (Fig. 4B), but its orientation is very different from that seen in the E. coli BCAT structure. The ␣-amino nitrogen is 6.8 Å away from the C4Ј of the cofactor and pointed so that it would have to rotate by ϳ90°in order to attack the cofactor C4Ј atom. The substrate ␣-carboxylate is actually in the L-glutamate side chain ␥-carboxylate binding site, hydrogen-bonded with Arg 143 (terminal guanidinium group), Val 155 * (backbone amide; the asterisk denotes a residue from opposite subunit), and Tyr 70 * (side chain terminal hydroxyl). The N-methylleucine side chain faces the C4Ј atom and is interacting with the side chains of Tyr 207 , Phe 75 , Thr 240 , and Val 155 *. The backbone N-methyl group of the substrate mimic is ori-ented toward the ␤-turn preceding the CXXC motif. Hence, in this monomer, the substrate orientation is not optimal for the reaction to form the geminal diamine with the PLP C4Ј atom.
Structure of the Michaelis Complex of C315A/C318A hBCATm with ␣-Ketoisocaproate (Protein Data Bank Code 2HDK)-Both Cys 315 and Cys 318 are mutated in the double mutant (Fig. 1C). The active site structure in monomer B and schematic for hydrogen bonding are shown in Figs. 5A and 6A, respectively. The changes seen in the hydrogen bond network at the ␤-turns preceding and following Cys 315 are similar to the C315A mutant structure. Electron density for an ␣-ketoisocaproate molecule is seen in monomer B but not in monomer A. The ␣-ketoisocaproate was not added during crystallization but was present in the storage buffer. The ␣-ketoisocaproate, although bound in the substrate binding pocket, is not oriented in a fashion similar to the substrate mimic in the E. coli wild type BCAT (16) or N-methylleucine modeled in wild type (Fig. 5B); nor does it correspond to the N-methylleucine in the bound C315A-NML structure (Fig. 4). The ␣-carboxylate of the ␣-ketoisocaproate is hydrogenbonding to the amides at Thr 313 and Ala 314 and a water molecule (Fig.  6A). The ␣-keto oxygen is at a hydrogen bonding distance from the side chain hydroxyl of Thr 240 . The leucine side chain has van der Waals contacts with residues Phe 30 , Phe 75 , and Val 155 (of monomer A). Bound in this disposition, the ␣-keto substrate would have to reorient significantly to undergo chemistry. Hence, we have another example where mutation of the cysteines has changed the hydrogen bonding at the ␤-turn and influenced substrate binding.
Structure of C318A Mutant in PLP Form (Protein Data Bank Code 2HGW)-Specific activity and kinetic properties of the C318A mutant were very similar to those of the WT enzyme (see Table 3). The  structure of this enzyme in the PLP form resembled the WT enzyme with several exceptions. In this mutant, the absence of the thiol/thiolate hydrogen bond between Cys 318 and Cys 315 , results in the side chain of Cys 315 reorienting in both monomers to another minimum energy conformer. The changes in the peptide backbone following Cys 315 are similar to that seen in the C315A mutant protein structure. The three new hydrogen bonds that were seen in the C315A structure are also seen in the C318A structure. Unlike in the C315A mutant, Thr 313 and Val 317 are ordered in both monomers of the C318A mutant. The differences that are seen between the kinetic properties of the C315A and C318A mutants could be attributed to the flexibility of the N terminus loop. Phe 30 , which is part of this loop, is crucial to the integrity of the substrate binding pocket, particularly for branched chain amino acids. The N terminus loop shows the least change in the C318A mutant and most flexibility in the Cys 315 single and double mutants.
Structure of the Oxidized Enzyme hBCATm-Ox (Protein Data Bank Code 2HHF)-Crystals of the PLP form of hBCATm-Ox were obtained under oxidizing conditions in the presence of hydrogen peroxide. Under these conditions, Cys 315 and Cys 318 form a disulfide bond (Figs. 1A and 7A), and a surface thiol, Cys 108 , has been oxidized to the sulfonic acid species. This confirms our previous results that suggested that Cys 108 is the cys-teine residue that reacts slowly with DTNB (19). Oxidation changes the hydrogen bonding pattern of the backbone atoms in one of the monomers so that it looks similar to the mutants (Fig. 6B). The second monomer has interactions resembling WT-hBCATm. The structure of the oxidized protein indicates that there is disruption of the substrate binding pocket upon oxidation. As seen in the ketimine intermediate structure of hBCATm (1), Phe 30 is part of the hydrophobic surface of the substrate binding pocket. It has van der Waals interactions with the substrate isoleucine side chain in this structure. It also participates in a perpendicular edge-to-face ring interaction with Tyr 141 . This interaction anchors the side chain of Tyr 141 in order to participate in hydrogen bonds with Arg 143 and a glutamate substrate side chain. These interactions were seen in a model of glutamate built into the substrate binding pocket of the enzyme (1). Residue Phe 30 is disordered in both monomers of the oxidized protein. In monomer A, the N terminus residues 15-32 cannot be modeled because of a lack of any traceable electron density. In monomer B, Phe 30 has reoriented significantly ( Fig. 7A) compared with its conformation in the WT-hBCATm. Although this change could affect binding constants, it alone does not explain the complete inhibition of enzyme activity.
Role of the Interdomain Loop-In the various structures of the WT-hBCATm enzyme studied so far, the interdomain loop (residues 171-181) is quite flexible. This is reflected in the high temperature factors and weaker electron density compared with the rest of the protein. Tyr 173 is part of the loop and interacts with Cys 315 in some of the structures through an S-H . . . hydrogen bond (1).
The loop is disordered in monomer A of the N-methylleucine-bound C315A mutant structure. In the C318A structure, the interdomain loop is disordered in both the monomers. In the double mutant, the interdomain loop is fairly ordered in both the monomers and is in a new conformation compared with the wild type enzyme (Fig. 8). In monomer A, Phe 174 and Pro 175 have moved significantly, facilitating atoms of the Phe 174 and Tyr 173 rings to participate in an edge-to-face interaction, constricting access to the active site (Fig. 8A). In monomer B, Tyr 173 of the interdomain loop has flipped its backbone conformation by 180°away from the active site pocket, widening the access to it (Fig. 8B). In monomer A of the oxidized protein, Phe 174 has flipped its side chain conformation by 180°com-  increase in the dipole moment at this turn. The interdomain loop residue Tyr 173 hydroxyl has a weak S-H . . . hydrogen bonding interaction with the sulfur of Cys 315 (Fig. 7B). Oxidation of the CXXC cysteines abolishes this interaction, and Tyr 173 flips away from the CXXC center. The N terminus loop between residues 15 and 32 is disordered in one of the monomers of the oxidized hBCATm protein (Fig. 7B). Phe 30 , which is part of a thiol loop, provides a crucial van der Waals surface for the substrate side chain and also participates in and edge-toface interaction with Tyr 141 , facilitating formation of a hydrogen bonding interaction with Arg 143 (a crucial residue that stabilizes the glutamate side chain). Phe 30 has significantly reoriented in a new conformation in the second monomer of the oxidized protein, weakening its interactions with the other residues as well as substrate binding. Oxidation causes a significant difference in substrate binding in hBCATm. The Phe 30 , Arg 143 , Tyr 207 , and Tyr 173 of the hydrophobic pocket are displaced and weaken the side chain binding of the substrate. The ␣-keto acid substrate in the PMP form is localized at an unfavorable location for catalysis (Fig. 7B). The ␣-keto group of the substrate forms a strong hydrogen bond with the OH group of Tyr 173 , orienting it for proper bond formation with the N4 of PMP. In the reduced PMP enzyme, the N4 of PMP is in a protonated form, which is crucial for catalysis. The interaction of N4 with the main chain oxygen of the CXXC motif stabilizes the protonation state of N4. N4 of PMP becomes protonated at a distance of 3.2 Å between N4 and O3Ј. However, PMP loses this interaction with the CXXC center in the oxidized protein. N4 is unable to be protonated at a distance of 3.7 Å between N4 and O3Ј. This can explain the inactivation of the oxidized enzyme. Reduction of the CXXC center by DTT regenerates the proper active site conformation, and catalysis proceeds (Fig. 3).
The CXXC center is only present in the mammalian BCATs (17)(18)(19). The x-ray structures of E. coli BCAT (6) show that residues 256 -262 superimpose well with the ␤-beta turn residues 312-318 in human BCAT. There are no cysteines in this stretch in E. coli BCAT. Ala 259 aligns with Cys 315 , and Thr 262 aligns with Cys 318 . The microdipoles at the first ␤-turn are aligned just as in the human enzyme. The peptide bond at Ala 259 is not aligned with the rest of the microdipoles in two of the three monomers in the asymmetric unit, although cysteines are not present. The energy against the torque in this case comes from a favorable hydrogen bond interaction between the Thr 262 side chain hydroxyl and the backbone carbonyl oxygen at Ala 259 (distance of 3.2 Å seen in monomer C and 3.3 Å seen in monomer B). In monomer A, this hydrogen bond is absent (distance of 4.1 Å), and the peptide dipole has rotated to partially align with the microdipoles. The , values of 82.1, Ϫ101.1 mean that the backbone is in the forbidden region of the Ramachandran map similar to our mutant structures. Since two subunits of the hexamer have aligned dipoles at Ala 259 , the catalytic efficiency of the Escherichia coli enzyme can be expected to be lower than its human counterpart for the transamination reaction (32).
In conclusion, the present study provides a structural and kinetic mechanism for the role of the CXXC center in hBCATm catalysis. The structures of the cysteine mutants (Cys 315 and Cys 318 ) of hBCATm show subtle conformational changes that probably impede proper substrate binding and reduce catalytic activity. The oxidized protein structure shows an additional structural alteration of the side chain binding pocket compared with the reduced protein. The integrity of the side chain binding pocket is paramount to correct substrate orientation in the second half of transamination.