Roles of active site and novel K+ ion-binding site residues in human mitochondrial branched-chain alpha-ketoacid decarboxylase/dehydrogenase.

The human mitochondrial branched-chain alpha-ketoacid decarboxylase/dehydrogenase (BCKD) is a heterotetrameric (alpha(2)beta(2)) thiamine diphosphate (TDP)-dependent enzyme. The recently solved human BCKD structure at 2.7 A showed that the two TDP-binding pockets are located at the interfaces between alpha and beta' subunits and between alpha' and beta subunits. In the present study, we show that the E76A-beta' mutation results in complete inactivation of BCKD. The result supports the catalytic role of the invariant Glu-76-beta' residue in increasing basicity of the N-4' amino group during the proton abstraction from the C-2 atom on the thiazolium ring. A substitution of His-146-beta' with Ala also renders the enzyme completely inactive. The data are consistent with binding of the alpha-ketoacid substrate by this residue based on the Pseudomonas BCKD structure. Alterations in Asn-222-alpha, Tyr-224-alpha, or Glu-193-alpha, which coordinates to the Mg(2+) ion, result in an inactive enzyme (E193A-alpha) or a mutant BCKD with markedly higher K(m) for TDP and a reduced level of the bound cofactor (Y224A-alpha and N222S-alpha). Arg-114-alpha, Arg-220-alpha, and His-291-alpha interact with TDP by directly binding to phosphate oxygens of the cofactor. We show that natural mutations of these residues in maple syrup urine disease (MSUD) patients (R114W-alpha and R220W-alpha) or site-directed mutagenesis (H291A-alpha) also result in an inactive or partially active enzyme, respectively. Another MSUD mutation (T166M-alpha), which affects one of the residues that coordinate to the K(+) ion on the alpha subunit, also causes inactivation of the enzyme and an attenuated ability to bind TDP. In addition, fluorescence measurements establish that Trp-136-beta in human BCKD is the residue quenched by TDP binding. Thus, our results define the functional roles of key amino acid residues in human BCKD and provide a structural basis for MSUD.

The human mitochondrial branched-chain ␣-ketoacid decarboxylase/dehydrogenase (BCKD) 1 is a thiamine diphosphate (TDP)-dependent enzyme, which catalyzes the oxidative decarboxylation of branched-chain ␣-ketoacids derived from leucine, isoleucine, and valine (1,2). BCKD is the E1 component of the macromolecular BCKD complex (M r ϭ 4 ϫ 10 6 ), which comprises multiple copies of BCKD or E1, dihydrolipoyl acyltransferase (E2), dihydrolipoamide dehydrogenase (E3), a specific kinase, and a specific phosphatase (3). Human BCKD is a heterotetrameric protein consisting of two ␣ (M r ϭ 45,500) and two ␤ (M r ϭ 37,500) subunits. In the oxidative decarboxylation reaction, BCKD or the E1 component catalyzes the TDP-mediated decarboxylation of a branched-chain ␣-ketoacid to produce an enamine-TDP, with concomitant reduction of the lipoamide moiety covalently attached to E2. This is followed by the reaction of enamine-TDP with the reduced dihydrolipoamide, which results in an S-acyldihydrolipoamide. E2 catalyzes the transfer of the acyl group from the tetrahedral intermediate to CoA to give rise to a branched-chain acyl-CoA. The nonconjugated dihydrolipoamide is subsequently reoxidized by E3, which is a flavoprotein, with NAD ϩ as the ultimate electron acceptor. The last step resets the cycle for oxidative decarboxylation of the branched-chain ␣-ketoacid. Mammalian BCKD is regulated by a reversible phosphorylation (inactivation)/dephosphorylation (activation) cycle under different hormonal and dietary stimuli (4). In patients with heritable maple syrup urine disease (MSUD), activity of the BCKD complex is deficient. This results in the accumulation of ␣-ketoacids with severe clinical consequences including often fatal ketoacidosis, neurological derangement and mental retardation (2).
To understand the structure and function of BCKD as well as the biochemical basis for MSUD, we have recently solved the three-dimensional structure of the human enzyme at 2.7-Å resolution (5). The human BCKD structure reveals the presence of two TDP-binding pockets, with each formed at the interface between the ␣ and ␤Ј subunits or between the ␣Ј and ␤ subunits (5). This topology is similar to the BCKD from Pseudomonas putida (6) and is equivalent to the homodimeric transketolase from Saccharomyces cerevisiae, where its two TDP-binding sites are formed at the head-to-tail interfaces between the two identical subunits (7). In contrast, crystal structures of pyruvate decarboxylases from Saccharomyces uvarum (8), S. cerevisiae (9) and Zymomonas mobilis (10), pyruvate oxidase from Lactobacillus plantarum (11), and benzoylformate decarboxylase from P. putida (12) indicate the presence of four TDP-binding sites in each homotetramer. The cofactor binding fold in these TDP-dependent enzymes is located at the interface between the ␣ domain of one subunit and the ␥ domain of the other. The physiological significance in doubling of the TDP-binding sites compared with transketolase and BCKD is not apparent at present. In addition, two novel K ϩ ion-binding sites in human BCKD, one on the ␣ subunit and one on the ␤ subunit, were identified. MSUD mutations that affect residues in the TDP-binding and the K ϩ ion-binding sites or residues involved in subunit assembly of BCKD have been described. These findings have provided a basis for understanding structure and function of human BCKD as well as how these properties are affected by naturally occurring MSUD mutations.
In this report, we carried out site-directed mutagenesis on residues in the TDP-binding and K ϩ ion-binding sites as well as residues that are affected in MSUD patients. Both the wildtype and mutant BCKD proteins were expressed in Escherichia coli in the presence of cotransformed chaperonins GroEL and GroES. These properly assembled heterotetramers were characterized for their kinetic parameters and their ability to bind cofactor TDP. The results confirm the roles of key amino acid residues in catalysis as well as cofactor and metal ion binding as implicated by the crystal structure of human BCKD. Moreover, the study depicts the conservation and divergence among TDP-dependent enzymes, which catalyze the cofactor-mediated oxidative or nonoxidative decarboxylation reaction. 6 -tagged BCKD-The Altered Site in vitro mutagenesis system (Promega, Madison, WI) was used to introduce desired mutations into the cDNA for the ␣ or the ␤ subunit. Detailed protocols for the mutant vector construction and subsequent mutagenesis were described previously (13). Briefly, oligonucleotides for the desired mutations and the ␤-lactamase repair primer were annealed to the single-stranded form of pAlter-␣ or pAlter-␤ vector. After the second strand synthesis and two rounds of ampicillin selection, clones harboring the correct mutations were isolated for plasmid preparation. DNA segments containing the mutations were used for cassette replacements of the expression vector pHis-TEV-E1 for wild-type BCKD (13).

Construction of Expression Plasmids for Mutant His
Expression and Purification of Human His 6 -tagged BCKD-The recombinant His 6 -tagged BCKD heterotetramer was efficiently expressed in E. coli strain CG-712 (ES ts ) by cotransformation of the pGroESL plasmid overproducing chaperonins GroEL and GroES as described previously (14,15). Wild-type and mutant His 6 -tagged BCKD heterotetramers were isolated from cell lysates using a Ni 2ϩ -NTA-derivatized Sepharose CL-6B column (Qiagen, Chatsworth, CA) as described previously (16). BCKD proteins were further purified on a Superdex-200 gel filtration column (2.6 ϫ 60 cm) connected to an FPLC system from Amersham Pharmacia Biotech. The column buffer consisted of 50 mM potassium phosphate, pH 7.5, 250 mM KCl, 10% (v/v) glycerol, 5 mM dithioerythritol, 1 mM benzamidine, and 1 mM phenylmethylsulfonyl fluoride. BCKD activity during purification was assayed spectrophotometrically (see below). Protein concentrations were determined using the Coomassie Plus protein reagent from Pierce with absorbance read at 595 nm. Alternatively, during enzyme purification, protein concentrations were determined by the direct measurement of absorbance at 280 nm using a calculated molar extinction coefficient of 1.15 cm Ϫ1 mg Ϫ1 ml for the ␣ 2 ␤ 2 heterotetramer.
Assays for BCKD Activity and Kinetic Studies-To determine K m for TDP, a radiochemical assay based on activity of the reconstituted BCKD complex was used (15). The rate of decarboxylation of 0.2 mM ␣-keto[1-14 C]isovalerate ([1-14 C]KIV) by BCKD in the presence of an excess of E2 and E3 was measured at varying concentrations of TDP. Double reciprocal plots were used to determine K m and V max values for cofactor TDP. For the determination of kinetic parameters for substrate KIV, a spectrophotometric assay, also according to the reconstituted BCKD complex activity, was employed. The assay mixture contained 50 mM potassium phosphate, pH 7.5, 100 mM NaCl, 3 mM NAD ϩ , 0.4 mM CoA, 2 mM MgCl 2 , 2 mM dithiothreitol, 0.1% Triton X-100, 400 M TDP, 7 nM lipoylated recombinant bovine E2, and 0.4 M recombinant human E3 (17). The reduction of NAD ϩ absorbance at 340 nm at different substrate concentrations was used to derive the k obs value. The plots observed the pseudo-first order decay. The double-reciprocal plots of k obs values versus KIV concentrations were used to determine K m and k cat values for KIV.
Measurements of Amounts of TDP Bound to Wild-type and Mutant BCKD-The purified recombinant human BCKD was essentially devoid of TDP. The residual bound TDP was removed by exhaustive dialysis in the presence of 0.2 mM EDTA. The wild-type or mutant apo-BCKD (500 g each) was incubated for 30 min with 150 M Mg-TDP at 4°C. The holo-BCKD was re-extracted with Ni 2ϩ -NTA resin and eluted with 100 mM imidazole. The BCKD-bound TDP was measured after oxidation to its fluorescent derivative thiochrome diphosphate in the presence of ferricyanide. The thiochrome diphosphate derivative was measured in a PerkinElmer Life Sciences model LS50B luminescence spectrometer The inverted V-shaped conformation of cofactor TDP is stabilized by stacking of the aminopyrimidine ring against the side chain of Tyr-102-␤Ј from the ␤Ј subunit (in greenish yellow) and the side chain of Leu-164-␣ from the ␣ subunit (in magenta). The invariant Glu-76-␤Ј important for cofactor activation coordinates to the N-1Ј atom of the aminopyrimidine ring (3.4 Å apart). A ketoacid substrate analog (in gray) labeled isocaproate is covalently modeled into the side chain of His-146-␤Ј, based on the crystal structure of BCKD from Pseudomonas putida (6). The carboxylate group of the inhibitor interacts with the N-4Ј amino group of TDP (separated by a distance of 4.3 Å). The side chain of Ser-162-␣ also coordinates to the N-4Ј amino group (3.0 Å apart) to position the cofactor in the correct conformation. Residue Ser-292-␣ is phosphorylation site 1 of human BCKD. The diphosphate moiety of TDP is stabilized, in part, by an octahedral coordination of the Mg 2ϩ ion. Two of the amino acid ligands Glu-193-␣ and Asn-222-␣ in this coordination are shown. Side chains of Arg-114-␣, Arg-220-␣, and His-291-␣, are, in turn, in direct contact with the distal phosphate oxygens, whereas the side chains of Gln-112-␣ and Tyr-113-␣ (not shown) interact with the proximal phosphate oxygens of the diphosphate moiety of TDP.
Other Methods-Emission spectra over a range of 300 -400-nm wavelengths for bound TDP were obtained using the luminescence spectrometer at an excitation wavelength of 280 nm as described previously (19). The quenching of tryptophan fluorescence by Mg-TDP was studied by adding increments of Mg-TDP to the cuvette containing apo-BCKD. The range of final Mg-TDP concentrations studied was 5-400 M. Cycles of Mg-TDP addition and fluorescence measurements were repeated until the fluorescence quenched by the TDP bound to BCKD reached a plateau. Circular dichroism measurements were carried out on an AVIV (Lakewood, NJ) model 62 DS spectrometer.

RESULTS AND DISCUSSION
Catalytic Residues for the TDP-mediated Decarboxylation-In human BCKD, residues that form each Mg-TDP cofactor binding pocket are derived from two separate subunits. Fig. 1 shows that residues from both the ␣ and ␤Ј subunits are involved in binding to TDP, and the Mg 2ϩ ion is held in an octahedral coordination. This topology maintains TDP in a strained "V-shaped" conformation having torsion angles ⌽ T ϭ 100°, and ⌽ P ϭ Ϫ71°to facilitate the proton extraction from the C-2 carbon (5). Most of these residues are conserved among TDP-dependent enzymes including Glu-193-␣, Gly-194-␣ (not shown), and Asn-222-␣ in the TDP-binding motif GDGX 26 -28 NN (20) and the invariant catalytic residue Glu-76-␤Ј. To establish the functional role of these active site residues, site-directed mutagenesis was carried out. His 6 -tagged wildtype and mutant BCKD carrying a single amino acid substitution were expressed in E. coli by cotransformation with chaperonins GroEL and GroES. The proteins were purified by Ni 2ϩ -NTA extraction followed by FPLC gel filtration on a HiLoad Superdex 200 column. The mutant BCKD used for this study uniformly formed heterotetramers, and no gross conformational changes were detected by circular dichroism spectroscopy (data not shown). The highly purified wild-type and mutant BCKD were used to determine their kinetic parameters and abilities to bind cofactor TDP.
As shown in Fig. 1, Glu-76-␤Ј is directly bound to the N-1Ј atom of the aminopyrimidine ring at a distance of 3.4 Å, whereas the main-chain carbonyl group of Ser-162-␣ coordinates to the N-4Ј group on the opposite side of the ring, with the two entities being 3.0 Å apart. The electron withdrawal caused by the interaction between Glu-76-␤Ј and the N-1Ј atom increases the basicity of the N-4Ј amino group through tautomerization. This interaction is essential for efficient proton abstraction from the C-2 atom of the thiazolium ring of TDP, giving rise to the reactive ylide (21). The nucleophilic attack of the ␣-ketoacid substrate at the reactive C-2 carbon of the ylide results in a tetrahedral adduct. Decarboxylation of this intermediate produces a 2-␣-carbanion or enamine-TDP, which is stabilized by a neutral resonance. This mechanism for TDPmediated decarboxylation is entirely conserved (22). In the case of human BCKD, the decarboxylated acyl moiety on the enamine-TDP is transferred to the E2 dithiolane ring to produce the tetrahedral adduct S-acyldihydrolipoamide. In the present study, substitution of Glu-76-␤Ј with Ala (E76A-␤, Table I) renders BCKD completely inactive. The data establish Glu-76-␤Ј as an essential catalytic residue. On the other hand, the interaction between the carbonyl group of Ser-162-␣ and the N-4Ј amino group is required to orient the latter for proton abstraction from the C-2 atom of the thiazolium ring. Since it involves a main-chain carbonyl group, a replacement with Ala is without effect on the catalytic efficiency for the TDP, although the same parameter for substrate KIV is significantly reduced.
His-146-␤Ј aligns with His-131-␤Ј in BCKD from P. putida. The latter residue was shown to form an adduct with a substrate analog, ␣-chloroisocaproate, in the crystal structure of the Pseudomonas enzyme (6). It has been proposed that binding of the substrate analog to His-131-␤Ј of the bacterial BCKD mimics a natural ␣-ketoacid substrate. The complete absence of enzyme activity in H146A-␤Ј (Table I) is consistent with the role of this residue in binding substrate KIV in human BCKD. The "isocaproate" moiety modeled into the His-146-␤Ј residue of human BCKD shows that the carboxylate oxygen is 4.3 Å away from the N-4Ј amino group of TDP (Fig. 1). The data support the possible role of His-146-␤Ј in positioning a native ketoacid substrate (e.g. KIV) for the TDP-mediated decarboxylation. This point needs to be confirmed by co-crystallization of human BCKD with ␣-chloroisocaproate, which has not been successful to date. Recently, the structure of the Desulfovibrio africanus pyruvate:ferredoxin oxidoreductase in complex with pyruvate (23) demonstrates that a carboxylate oxygen from pyruvate interacts directly with the N-4Ј amino group of TDP. Together, the data strongly imply that the polarization of the substrate carboxylate moiety by the N-4Ј amino group of the cofactor is necessary for activation of the substrate.
Ser-292-␣ and Ser-302-␣ are site 1 and site 2 for phosphorylation of human BCKD by the specific kinase. Phosphorylation at site 1 results in inactivation of BCKD, whereas phosphorylation at site 2 is silent. Introduction of a negatively charged Asp residue in the S292D-␣ mutant produces the same inactivation effect as phosphorylation at this residue (Table I). This result confirms an earlier study in which replacement of site 1 Ser with Glu in rat BCKD also results in an inactive enzyme (24). As shown in Fig. 1, Ser-292-␣ is positioned directly above the C-2 atoms of the thiazolium ring of TDP at an 8-Å distance. The introduction of a negatively charged phosphate group or amino acid residue is likely to interfere with the proton abstraction from the C-2 atom by the basic amino group at N-4Ј position of the aminopyrimidine ring. Moreover, Ser-292-␣ is 9 Å away from His-146-␤Ј (Fig. 1). The presence of negative charges by phosphorylation or site-directed mutagenesis may also disrupt binding of the ␣-ketoacid substrate as suggested in earlier studies (24,25).
The side chain of Tyr-102-␤Ј is packed against one side of the aminopyrimidine ring of the cofactor with the side chain of Leu-164-␣ approaching the other side of the ring, wedging in between the two rings of the cofactor (Fig. 1). The flanking of both sides of the aminopyrimidine ring by these two hydrophobic residues is important to orient the cofactor in the strained V conformation. Substitution of either residue with an Ala has adverse effects in the catalytic efficiency of the mutant enzymes (Table I). However, L164A-␣ is markedly more severely affected than Y102A-␤Ј. The data suggest that Leu-164-␣ is more important in providing the hydrophobic environment for the aminopyrimidine and possibly the thiazolium ring of the cofactor. The amounts of TDP bound to His 6 -tagged wild-type and mutant BCKD were measured by incubating the enzymes with 150 M of Mg-TDP. Following Ni 2ϩ -NTA extraction, the bound TDP was determined in the form of thiochrome pyrophosphate (18). This method allowed for end point measurements of bound TDP in BCKD incubated at saturating concentrations of the cofactor. The results showed that each mole of wild-type BCKD binds ϳ2 mol of TDP ( Fig. 2A). The data confirm the presence of two cofactor-binding sites in the BCKD heterotetramer. It is noteworthy that the active site residues involved in catalysis or substrate binding have little effect on TDP binding. This notion is supported by the slight decrease in TDP binding as observed with the E76A-␤Ј, S162A-␣, H146A-␤Ј, and S292D-␣ mutants ( Fig. 2A). In the S162A-␣ mutant, the near wild-type K m for TDP agrees well with a TDP-binding stoichiometry that is also similar to wild-type. In contrast, the L164A-␣ mutant binds only a trace amount of TDP, which is consistent with an over 400-fold increase in K m for TDP relative to the wild-type. The data indicate that Leu-164-␣ that wedges in between the two aromatic rings of the cofactor is important for TDP binding. It is of interest that TDP binding in the Y102A-␤Ј mutant is not as severely affected as the L164A-␣. The results confirm that Tyr-102-␤Ј, which is across the aminopyrimidine ring planar from Leu-164-␣, is less critical for the cofactor binding.
Active Site Residues That Coordinate to the Mg 2ϩ Ion-As shown in Fig. 1, the Mg 2ϩ ion is octahedrally coordinated between the cofactor phosphates, the carbonyl of Tyr-224-␣ (not shown) and the side chains of Asn-222-␣ and Glu-193-␣. A substitution of Glu-193-␣ with Ala results in complete loss of BCKD activity (Table II). An E193K-␣ mutation recently identified in an MSUD patient also renders the enzyme inactive (data not shown). The N222S-␣ mutation, which was identified in an MSUD patient, results in marked increases in K m for KIV and TDP, whereas the k cat values are one half of the wild-type. These data, taken together, indicate that the carbonyl groups in the side chains of Glu-193-␣ and Asn-222-␣ are critical for Mg 2ϩ ion binding. The hydroxyl group of the Ser side chain in the N222S-␣ mutant is probably a poorer ligand than Asn for the Mg 2ϩ ion. The O ␥ of a Ser residue would not be able to extend as far toward the Mg 2ϩ ion as the O ␦1 of the Asn residue; this is likely to decrease affinity for the Mg 2ϩ ion at this site. The less attenuated effect in affinity for KIV and TDP in the Y224A-␣ mutant, compared with the above two mutants is consistent with the fact that the main-chain carbonyl group of Tyr-224-␣ coordinates to the metal ion. The critical roles of Glu-193-␣ and Asn-222-␣ side chains in the metal ion binding is supported by the markedly diminished levels of bound TDP in the E193A-␣, N222A-␣, and Y224A-␣ mutants (Fig. 2B). The Y224A-␣ mutant also shows a markedly decreased ability to FIG. 2. Stoichiometry of bound TDP in wild-type and mutant BCKD. Wildtype or mutant apo-BCKD (His 6 -tagged) at 500 g each was incubated with 150 M Mg-TDP for 30 min at 4°C. The holoenzymes were extracted with Ni 2ϩ -NTA resin, and eluted proteins were oxidized with ferricyanide. The thiochrome-TDP released from the enzymes was measured by fluorescence emission at 430 nm following an excitation at 375 nm. The results are expressed as mol of TDP bound per mol of BCKD heterotetramers. A, mutants affecting catalytic residues required for the TDP-mediated decarboxylation. B, mutants affecting residues that coordinate to the Mg 2ϩ ion of TDP. C, mutants affecting residues that interact with diphosphate oxygens of TDP. D, mutants affecting residues in the K ϩ ion-binding site on the ␣ subunit.
bind TDP despite an only moderate increase in K m for the cofactor.
Residues That Interact with Diphosphate Oxygens of TDP-Gln-112-␣ and Tyr-113-␣ are hydrogen-bonded to the proximal and the terminal phosphate oxygens of TDP, respectively (5). Substitution of Gln-112-␣ with an Ala residue has marginal effects in catalytic efficiency of BCKD as well as K m values for KIV and TDP (Table III). The data suggest that either the side chain of the Gln-112-␣ is not an essential ligand to the Mg 2ϩ ion or that a neighboring residue(s) can fill in this function. The introduction of an Ala residue into the Tyr-113-␣ position results in more than 50-and 100-fold increases in K m for KIV and TDP, indicating that Tyr-113-␣ is an important ligand to the terminal phosphate oxygen. Remarkably, a replacement of Arg-114-␣, which also coordinates to the same terminal phosphate oxygen (Fig. 1), with an Ala residue results in a completely inactive BCKD. The R114W-␣ mutation, which occurs in MSUD patients, also renders the enzyme completely inactive. As for Arg-220-␣, which coordinates to another distal phosphate oxygen, when changed to Ala, Lys or Trp (an MSUD mutation) also results in complete loss in enzyme activity. The combined results strongly suggest that the ionic interactions between positively charged Arg-114-␣ or Arg-220-␣ and the negatively charged phosphate oxygens are critical in maintaining the conformational integrity of the diphosphate group of TDP (Fig. 1). In the case of R220K-␣, the positively charged Lys residue is pointing away from the distal phosphate oxygen in the BCKD structure. Therefore, the ionic interactions between the Lys side chain and the phosphate oxygen cannot occur.
His-291-␣ coordinates to another distal phosphate oxygen in the TDP-binding pocket (Fig. 1). Substitution of this residue with Ala results in a trace amount of BCKD activity (Table III). A previous report showed that the same mutation in the rat BCKD is associated with complete absence of activity (26); however, a spectrophotometric assay was used in the latter study, which is less sensitive than the radiochemical assay employed here. The combined results are thus consistent with His-291-␣ as an essential ligand to the diphosphate group of TDP. It is noteworthy that the K m for TDP with the His-291-␣ mutant is increased by 40-fold over the wild-type (Table III). This is consistent with about 15% of TDP bound to this mutant compared with the wild-type (Fig.  2C). Modifications of the three other residues Arg-114-␣, Arg-220-␣, and Gln-112-␣ that coordinate to the diphosphate oxygens invariably yield marginal binding of TDP (Fig. 2C). The data

Roles of Amino Acid Residues in Human BCKD
strongly suggest that the amino acid ligands to the diphosphate moiety of TDP are as important as the ligands to the aminopyrimidine ring in maintaining the V conformation of the cofactor. It should be mentioned that in addition to being a ligand for the TDP diphosphate group, His-291-␣ has been implicated to function as a catalytic residue (6). As described above, BCKD functions as both a decarboxylase and a dehydrogenase during the oxidative decarboxylation of ␣-ketoacids. His-291-␣ is proposed to serve as a proton donor during the reduction of the disulfide bond of the E2-attached lipoamide, which occurs during acyltransfer from the enamine-TDP to E2. In the H291A-␣ mutant, the reduction of lipoamide on E2 by the dehydrogenase activity of BCKD may be disrupted. The additional role of His-291-␣ as a proton donor to the E2-attached lipoamide will require further studies.
Residues in the Novel K ϩ Ion-binding Site of the ␣ Subunit-The crystal structure of human BCKD disclosed that the ␣ and the ␤ subunits each has a distinct K ϩ ion binding fold that has not been described previously in any TDP-dependent enzyme (5). Fig. 3 shows that the K ϩ ion on the ␣ subunit stabilizes a loop containing residues Ser-161-␣, Thr-166-␣, and Gln-167-␣, which are directly involved in ligating to the metal ion through their side chains. The structural integrity of this loop structure is essential for ordering Ser-162-␣ and Leu-164-␣ for interactions of these residues with the cofactor TDP as described above. An MSUD mutation, T166M-␣, is likely to abolish the coordination of the Thr side chain to the K ϩ ion, resulting in a disorder of this loop with a concomitant disruption of TDP binding. This accounts for the loss of both enzyme activity (Table IV) and the inability of the mutant enzyme to bind TDP (Fig. 2D). It is of interest that the replacement of Thr-166-␣ with an Ala has little effect on the catalytic efficiency of BCKD. It is possible that the side chain of Ala, which is shorter than that of Met, does not aberrantly protrude into the K ϩ ion binding pocket. As a result, the octahedral coordination is not severely impaired by this substitution. This is reflected by near wild-type catalytic efficiency of the T166A-␣ mutant when as-sayed at saturating TDP concentrations (Table IV). However, the binding affinity of the mutant enzyme for the cofactor is still reduced as indicated by an elevated K m for TDP (Table IV) and by a significant decrease in the amount of the bound cofactor compared with the wild-type (Fig. 2).
Identification of the Trp Residue Quenched by Bound TDP-One of the characteristics associated with TDP-dependent enzymes involves quenching of tryptophan fluorescence upon binding of the cofactor (19). This property has been used as a means to determine the level of the cofactor binding by TDPdependent enzymes (22,27). To identify the Trp residue quenched by BCKD-bound TDP, we replaced Trp-136-␤, which is the Trp residue closest to the aminopyrimidine ring of the cofactor at 15-18 Å, with a Phe. As a control, a distal Trp-309-␣ was also converted to a Phe. The W136F-␤ mutant BCKD showed K m values for KIV and TDP, which are similar to those for the wild-type (data not shown). Therefore, Trp-136-␤ appears to be nonessential for efficient binding of both the substrate and the cofactor. The apoenzymes of the wild type, W136F-␤, and W309F-␣ were excited at 280 nm. The fluorescence emission spectra of the wild-type and W309F-␣ apo-BCKD were similar (Fig. 4). The maximal fluorescence emissions at 340 nm for both proteins are progressively quenched when titrated with increasing concentrations of Mg-TDP. In contrast, the maximal emission at 340 nm of the W136F-␤ mutant enzyme is not quenched over the same concentration range of Mg-TDP. The result established that Trp-136-␤ is the residue that is quenched upon TDP binding. In pyruvate decarboxylase from Z. mobilis, the Trp-487 was shown to also be quenched by the bound TDP (28). The crystal structure of this bacterial pyruvate decarboxylase shows that Trp-487 is 18 Å away from the aminopyrimidine of a chemically modified cofactor, 2-(1-hydroxyethyl)-TDP (10). Thus, the Trp residues in human BCKD and the bacterial pyruvate decarboxylase, which are quenched upon TDP binding, are in similar conformations relative to the cofactor. In the related pyruvate dehydrogenase, Trp-135-␤ was reported to be the residue that is quenched in

Roles of Amino Acid Residues in Human BCKD
response to TDP binding using chemical modification and magnetic circular dichroism methods (29). Although the structure of the pyruvate dehydrogenase has not yet been determined, Trp-135-␤ in this enzyme aligns with the Trp-136-␤ in human BCKD. The data show the conservation of the quenchable Trp residue in TDP-dependent enzymes.
Conclusion-The present study defines the functional roles of active site residues in human BCKD, which catalyzes the oxidative decarboxylation of branched-chain ␣-ketoacids. The results depict a high degree of conservation in the structure and function of the TDP binding fold in TDP-dependent enzymes. These enzymes include those that catalyze the TDPmediated oxidative decarboxylation, such as human BCKD (5), and the TDP-mediated nonoxidative decarboxylation, such as yeast pyruvate decarboxylase (8) and benzoylformate decarboxylate (12), as well as those that promote the two-carbon transfer, for example, yeast transketolase (7,30). Human BCKD is distinct among TDP-dependent enzymes in that the enzyme is both a decarboxylase and dehydrogenase. The enamine-TDP intermediate becomes a substrate for the acyltransfer reaction during the BCKD-mediated reduction of lipoamide covalently attached to the E2 core of the BCKD complex. The mechanism for the BCKD-catalyzed acyltransfer reaction remains to be elucidated at the structure level.
The human BCKD is tightly regulated by reversible phosphorylation/dephosphorylation (4). Residue Ser-292-␣, which is responsible for phosphorylation and the resultant inactivation of BCKD, is located inside the TDP-binding pocket. From a structural standpoint, it is unclear how the specific BCKD kinase enters this TDP binding fold to phosphorylate Ser-292-␣. However, the phosphorylation efficiency of BCKD is markedly enhanced when the enzyme is in complex with the E2 acyltransferase (31)(32)(33). It can, therefore, be speculated that binding of BCKD to E2 may induce conformational changes that render the TDP-binding pocket more accessible to the kinase. These questions will be addressed by solving the structure of human BCKD that in complex with the binding domain of E2.
The utilization of the octahedrally coordinated K ϩ ion site to stabilize an essential loop structure in the TDP binding fold has not been described in other TDP-dependent enzymes. Both BCKD and pyruvate dehydrogenase require high concentrations of the K ϩ ion to stabilize enzyme activity (16,31,34). It is very likely that the novel K ϩ -binding sites are conserved in the decarboxylase/dehydrogenase components of ␣-ketoacid dehydrogenase complexes. The presence of K ϩ ion-binding sites explains the dependence of high K ϩ , but not Na ϩ , ion concentrations on the inhibitions of BCKD phosphorylation by TDP (31). A survey of the literature shows that the Pseudomonas dialkylglycine decarboxylase also uses the octahedral K ϩ ion coordination to maintain the conformation of a loop structure in the cofactor pyridoxal phosphate binding fold (35,36). This topology appears to be conserved among certain cofactor-dependent decarboxylases.