Structural and Biochemical Basis for Novel Mutations in Homozygous Israeli Maple Syrup Urine Disease Patients A PROPOSED MECHANISM FOR THE THIAMIN-RESPONSIVE PHENOTYPE*

Maple syrup urine disease (MSUD) results from mutations affecting different subunits of the mitochondrial branched-chain (cid:1) -ketoacid dehydrogenase complex. In this study, we identified seven novel mutations in MSUD patients from Israel. These include C219W- (cid:1) (TGC to TGG) in the E1 (cid:1) subunit; H156Y- (cid:2) (CAT to TAT), V69G- (cid:2) (GTT to GGT), IVS 9 del[ (cid:3) 7: (cid:3) 4], and 1109 ins 8bp (exon 10) in the E1 (cid:2) subunit; and H391R (CAC to CGC) and S133stop (TCA to TGA) affecting the E2 subunit of the branched-chain (cid:1) -ketoacid dehydrogenase complex. Re-combinant E1 proteins carrying the C219W- (cid:1) or H156Y- (cid:2) mutation show no catalytic activity with defective subunit assembly and reduced binding affinity for cofactor thiamin diphosphate. The mutant E1 harboring the V69G- (cid:2)

The metabolic block in the BCKD complex results in the inability of MSUD patients to degrade BCKA. The elevated BCKA levels produce severe clinical consequences including oftenfatal ketoacidosis, mental retardation, and neurological impairment. There are presently five known MSUD clinical phenotypes (i.e. classic, intermediate, intermittent, thiaminresponsive, and E3-deficient forms) (2). The classic form with the most severe phenotype, was originally reported by Menke et al. in 1954 (3) and manifests within the first 2 weeks of life by poor feeding, lethargy, seizures, coma, and death if left untreated. The classic form accounts for 75% of MSUD patients (1). Intermediate MSUD is associated with progressive mental retardation and developmental delay without a history of catastrophic illness. An intermittent form of MSUD has normal levels of BCAA as well as normal intelligence and development until a stress such as infection precipitates metabolic decompensation with ketoacidosis. Thiamin-responsive MSUD is similar to the intermediate or intermittent phenotype but responds to pharmacological doses of thiamin with normalization of the BCAA levels (2). The E3-deficient MSUD is caused by defects in the dihydrolipamide dehydrogenase (E3) component of the BCKD complex, which is common to that of the pyruvate and ␣-ketoglutarate dehydrogenase complexes. Patients with E3 deficiency display a combined dysfunction of the three ␣-ketoacid dehydrogenase complexes (4). Two prevailing mutations (G229C and Y335stop) in the E3 gene have been identified in the Ashkenazi Jewish community in Israel (5). Most of the Israeli patients who were homozygous for the G229C E3 mutation presented in early childhood or later with recurrent episodes of vomiting, encephalopathy, and prolonged prothrombin time, occasionally associated with lactic acidosis and ketoacidosis. Patients who were compound-heterozygous for these two mutations presented neonatally with more severe sequelae.
The mammalian BCKD multienzyme complex is a 4 ϫ 10 6dalton metabolic machine organized around a cubic core comprising 24 lipoate-bearing dihydrolipoyl transacylase (E2) subunits, to which multiple copies of branched-chain ␣-ketoacid decarboxylase (E1), E3, a specific kinase, and a specific phosphatase are attached through ionic interactions. The kinase and the phosphatase are responsible for the regulation of BCKD complex by a reversible phosphorylation (inactivation)/ dephosphorylation (activation) cycle (6). The E1 component is a thiamin diphosphate (ThDP)-dependent enzyme consisting of two E1␣ and two E1␤ subunits. The E3 component is a homodimeric flavoprotein. There are in total six genetic loci that encode subunits of the BCKD complex. Mutations in the four different catalytic subunits (E1␣, E1␤, E2, and E3) have been described in MSUD patients (2). Genetic subtypes of MSUD have been proposed to indicate the altered subunit in the BCKD complex (2). These include type IA MSUD affecting the E1␣ subunit, type IB affecting the E1␤ subunit, type II affecting the E2 subunit, and type III affecting the E3 subunit. Types IV and V, which have not been reported, are reserved for MSUD in which the kinase and the phosphatase, respectively, are affected. Except for type II and III MSUD, which are linked to the thiamin-responsive (2,7,8) and E3-deficient phenotypes (4), respectively, a tight correlation between a specific genetic subtype and a particular clinical phenotype of MSUD has not been demonstrated.
Reaction steps catalyzed by the three enzyme components, based largely on those elucidated for the related pyruvate dehydrogenase complex (9), are as follows. R™CO™COOH  The E1 component binds ThDP and catalyzes a ThDP-mediated decarboxylation of ␣-ketoacids (Reaction 2), and subsequent reduction of the lipoyl moiety, which is covalently attached to E2 (Reaction 3). The lipoyl-bearing domain (LBD) carrying the S-acyldihydrolipoamide serves as a "swinging arm" (10) to transfer the acyl group from E1 to the E2 active site giving rise to acyl-CoA (Reaction 4). Finally, the E3 component with a tightly bound FAD moiety reoxidizes the dihydrolipoyl residue on E2 (Reaction 5) with NAD ϩ as the ultimate electron acceptor (Reaction 6). The overall reaction is the production of branched-chain acyl-CoA, CO 2 , and NADH from BCKAs (Reaction 1).
We previously described two novel missense mutations in MSUD patients from the non-Jewish Druze kindred in Israel (11). The incidence of MSUD in Israel is relatively high, presumably as a result of consanguinity (12). Toward the goal of providing a wider spectrum of MSUD mutations in this region, we continue to analyze cell cultures derived from Israeli patients. In the present study, we report seven additional MSUD mutations in the Israeli population, which affect the E1␣ (type IA), E1␤ (type IB), and E2 (type II) subunits of the BCKD complex. Among them are novel type IA and type IB homozygous missense mutations that impede subunit assembly of the E1 component and nullify ThDP-mediated decarboxylation of BCKAs. A type II homozygous point mutation H391R in the E2 subunit involves the key catalytic residue that functions as a general base in the acyltransfer reaction catalyzed by E2 (13,14). A biochemical mechanism is proposed to explain the unequivocal thiamin-responsive phenotype presented by the Israeli type II MSUD patients. The genetic and biochemical information presented here provides new insights into struc-ture and function of the human BCKD complex, and will facilitate DNA-based detection of these MSUD alleles in the Israeli population.

EXPERIMENTAL PROCEDURES
Cell Lines and Cell Cultures-Blood samples (15 ml) were withdrawn from classic MSUD patients B.R, W.A., D.M., K.Y., T.G., and C.R. as well as Druze thiamin-responsive patients H.C. and H.S., all from Israel. Lymphoblasts were prepared from blood samples by infection with the Epstein-Barr virus, and lymphoblastoid cell cultures were grown and assayed for the decarboxylation of ␣-keto[1-14 C]isovalerate by intact cells as described previously (15).
Production of Wild-type and Mutant Proteins-The pTrcHisB expression plasmid (Invitrogen) for N-terminally His 6 -tagged wild-type E1 was constructed as described previously (16). The same plasmids carrying MSUD mutations were produced using the QuikChange™ sitedirected mutagenesis kit from Stratagene (La Jolla, CA) according to the manufacturer's instructions. Wild-type and mutant E1 proteins were expressed in Escherichia coli by co-transformation with the pGroESL plasmid over expressing chaperonins GroEL and GroES, as described previously (17,18). His 6 -tagged wild-type and mutant E1 were purified with Ni 2ϩ -nitrilotriacetic acid resin and Resource Q and Superdex-200 columns (17) and used for biochemical studies without removal of the His 6 tag. The E2 24-mer was expressed in XL1-Blue cells and lipoylated in vitro with bacterial lipoyl ligase as described previously (19). C-terminal His 6 -tagged E2 domain constructs LBD2 (residues 1-99) containing the lipoyl domain and the C-terminal linker and the di-domain (DD) (residues 1-167) containing the LBD and subunitbinding domain (SBD) were prepared as described previously (20). Lipoylated LBD2 (lip-LBD2) and DD (lip-DD) were prepared as described above. To produce the SBD, a tobacco-etch virus protease site (LENLYFQ2S) with a nucleotide sequence of 5Ј-ctcgagaatctttattttcaatca-3Ј was inserted into the linker between the lipoyl-bearing and subunit binding domains. The purified C-terminally His 6 -tagged DD was digested with the tobacco-etch virus protease, and the SBD was extracted by Ni 2ϩ -nitrilotriacetic acid resin.
Assays for Activities of Individual E1 and E2 Components-The decarboxylation catalyzed by the isolated E1 component (Reaction 2) was assayed with ␣-ketoisovalerate as a substrate in the presence of an artificial electron acceptor 2,6-dichlorophenolindophenol (DCPIP) (Reaction 7) as described previously (21).
The assay mixture contained 100 mM potassium phosphate, pH 7.5, 2.0 mM MgCl 2 , 0.2 mM ThDP, and 0.1 mM DCPIP. The rate of decarboxylation at 30°C was measured by monitoring the reduction of the dye at 600 nm. The acyltransferase activity (Reaction 4) of E2 was assayed using a model reaction (Reaction 8) with [1-14 C]isovaleryl-CoA and dihydrolipoamide (Lip(SH) 2 ) as substrates.
Western Blotting of Cell Lysates-Homogenates from cultured lymphoblasts were subjected to SDS-PAGE, followed by transfer to Immobilon-P membranes (Millipore Corp.). After blocking with a phosphatebuffered saline containing 5% (w/v) nonfat dry milk, these membranes were probed with either anti-E2 or anti-E1 (with titers against both E1␣ and E1␤ subunits) antibodies each at a 1:500 dilution. Unbound antibodies were removed by washing the membranes with the same buffer containing 0.05% (v/v) Tween 20 and 0.05% (v/v) Nonidet P-40. The bound antibodies were detected with 125 I-protein A as described previously (15).
Binding Measurements Based on Tryptophan Fluorescence Quenching-Steady-state fluorescence quenching upon ThDP binding (23) to wild-type and mutant E1 was measured using a Perkin-Elmer Life Sciences LS50 B luminescence spectrometer as described previously (24). Fluorescence intensities were recorded at 25°C at an excitation wavelength of 290 nm and an emission wavelength of 335 nm. Slit widths were set at 5 nm for both excitation and emission. A 290-nm cut-off emission filter was installed to reduce light scattering effects. Protein concentrations for E1 (A 280 ϭ 1.14 mg Ϫ1 ml⅐cm Ϫ1 ) and ThDP (A 235 ϭ 11,300 M Ϫ1 cm Ϫ1 , pH Ͼ7.0) were determined spectrophotometrically. The concentration for all protein samples was 0.23 M (as heterotetramers) in 50 mM potassium phosphate buffer, pH 7.5, 200 mM KCl, and 1 mM MgCl 2 . The binding data were fitted by nonlinear regression using the program KaleidaGraph (Synergy Software, Essex Junction, VT) according to Equation 1 describing a bimolecular reaction (25), where ⌬F represents the fluorescence change corrected for dilution and inner filter effects (26), F 0 is the fluorescence intensity prior to the addition of ThDP, ⌬F max is the maximal fluorescence change, K d is the dissociation constant, and [ThDP] is the concentration of ThDP in the cuvette. The parameters determined by the fitting procedure were ⌬F max and K d .

Treatment of Mutant E1 Proteins with Natural Osmolyte
Trimethylamine N-Oxide-Wild type or mutant E1 proteins (200 g/ml) were incubated with different concentrations of trimethylamine N-oxide (TMAO) at 23°C for 16 h in buffer A that contained 50 mM potassium phosphate, pH 7.5, 100 mM KCl, 2 mM ThDP, 2 mM MgCl 2 , 10 mM dithiothreitol, 0.6 M lysine, and a mixture of protease inhibitors (Roche Applied Science, Indianapolis, IN). Activity of the BCKD complex reconstituted with TMAO-treated wild-type or mutant E1 was assayed as described above.

RESULTS
Clinical Phenotypes-B.R., W.A., D.M., and K.Y. of Israeli origin manifested the classic MSUD phenotype as described above. T.G. and C.R. were from the Druze kindred in Israel, who suffered also from the classic MSUD phenotype with a turbulent neonatal presentation. Both exhibited mild to moderate developmental retardation with leukodystrophy in brain MRI. Furthermore, both patients had recurrent encephalopathic episodes during viral infections or suspected dietary noncompliance. T.G. underwent hemodialysis at age 2 and 4, which may have also resulted from dietary noncompliance.
H.C. is the third child of first generation Arabs. He was delivered uneventfully after a normal pregnancy. The parents reported irritability of the child since the first week of life. Brain MRI reveals leukodystrophy, which in combination with elevated levels of BCAAs led to the diagnosis of MSUD. The patient underwent hemodialysis and was placed on a MSUD formula supplemented with thiamin at the dosage of 100 mg/ day. The BCAA levels decreased to normal after 3 days. On follow-ups, the plasma BCAA levels remained in the normal range (50 -100 nmol/ml) for the next 4 years. The patient consumed 4 times the leucine levels that were allowed for MSUD patients, but during acute infections his BCAA and alloisoleucine levels rose only slightly above the normal range. At age 5, he had spastic cerebral palsy and his IQ was 80. H.S. was also born uneventfully after a normal pregnancy. Parents of H.S. and H.C. were first-degree cousins; therefore, dietary restriction was instituted for H.S after birth. She was diagnosed with MSUD at 2 days of age, and thiamin supplement (100 mg/day) was added to the restricted diet. At 1 week of age, H.S. was discharged and given breast milk at home. Three days later, she was readmitted with vomiting, apathy, and encephalopathy with the plasma leucine level at 1,800 nmol/ml, and underwent hemodialysis. Since this episode, the regiment of restricted diet with thiamin supplements was reinstated and maintained. At 18 months of age, H.C. showed normal growth and development.
Identification of the Affected BCKD Subunit in MSUD Patients-Since the human BCKD complex consisted of six different subunits, it was necessary for mutational analysis to identify which subunit was affected in the MSUD patient of interest. Homogenates of cultured lymphoblasts from the eight Israeli patients were subjected to Western blotting using antibodies against E1 (both E1␣ and E1␤ subunits) or E2 as a probe. Cells from the classic patient B.R. showed reduced levels of both the E1␣ and the E1␤ subunit, with the level of the E2 subunit in the normal range. Levels of both E1␣ and E1␤ subunits were below the detection limit in cells from classic patients W.A., D.M., and K.Y. The data strongly suggested that either the E1␣ or the E1␤ locus was affected in these Israeli MSUD patients. Individual E1 subunits are not stable in cells, and a mutation that impedes the stability of either the E1␣ or E1␤ subunit usually results in reduced levels or the absence of both subunits in cells from MSUD patients (7). Levels of both E1 subunits were normal in thiaminresponsive patients H.C. and H.S. as well as classic patients T.G. and C.R. In contrast, the E2 protein level was reduced in H.S. and H.C. compared with the wild type and was absent in T.G. and C.R., implicating that the E2 locus was affected in these MSUD patients.
Novel MSUD Mutations in the BCKD Complex-Samples of total RNA isolated from cells of MSUD patients were amplified by reverse transcriptase-PCR using primers corresponding to the cDNA sequence of the affected subunit. Nucleotide sequencing identified mutations in different subunits of the BCKD complex. All mutations in mRNA were confirmed by sequencing of the corresponding gene of the human BCKD complex in these patients. Table I shows seven novel MSUD mutations (types IA, IB, and II) identified in the eight Israeli patients investigated. A homozygous C to G type IA missense mutation C219W-␣ (TGC 3 TGG) is present in classic patient B.R. Two homozygous type IB mutations, H156Y-␤ (CAT 3 TAT) and V69G-␤ (GTT 3 GGT), occur in classic patients W.A. and D.M., respectively. Another classic patient, K.Y., harbors two compound heterozygous type IB alleles: a 4-base pair deletion in intron 9 (IVS9 del[Ϫ7:Ϫ4], resulting in the deletion of entire exon 10, and an 8-base pair insertion in exon 10 (1,109 ins 8bp) that causes a frameshift beginning at residue Glu 304 -␤. Two type II missense mutations were also detected. The homozygous H391R substitution (CAC 3 CGC) in the E2 subunit is present in related thiamin-response patients H.C. and H.S. The homozygous S133stop (TCA 3 TGA) in the E2 subunit is detected in classic patients T.G. and C.R.
Activity Measurement of MSUD Mutant Proteins-The E1 component of the human BCKD complex carrying one of the above MSUD missense mutations was expressed in E. coli co-transformed with the pGroESL plasmid overexpressing chaperonins GroEL and GroES. The yield for the mutant E1 protein carrying the C219W-␣ mutation was 87% of the wild type. The H156Y-␤ mutant E1 protein was poorly expressed with a yield at 18% of the wild-type. The V69G-␤ mutant E1 was not expressed in the soluble fraction of transformed BL-21 cells. Mutant E2 harboring the H391R missense mutation was as efficiently expressed as the wild-type without the co-transformation of chaperonins GroEL and GroES and existed as a stable homo-24-mer as determined by gel filtration.
The activity of the BCKD complex (Reaction 1) reconstituted with stoichiometric amounts of E1, E2, and E3 components was measured using ␣-keto[1-14 C]isovalerate as a substrate. The wild-type human BCKD complex when reconstituted with normal E1, E2 and E3 exhibits a decarboxylation rate of 196 min Ϫ1 . A substitution of the wild-type E1 with a mutant carrying the C219W-␣ or H156Y-␤ mutation does not produce BCKD activity in the presence of normal E2 and E3. Direct measurements of E1 component activity using 2,6-dichlorophenolindophenol as an electron acceptor (Reaction 7) confirmed the absence of E1 activity in the C219W-␣ and H156Y-␤ E1 variants (data not shown). Similarly, no BCKD activity was reconstituted when the wild-type E2 was replaced by a mutant containing the H391R alteration. The activity of the E2 component alone, based on the transfer of the radiolabeled acyl moiety from [1-14 C]isovaleryl-CoA to dihydrolipoamide (Reaction 8), was measured. The wild-type E2 showed a specific activity of 1.97 nmol/min/mg, whereas the mutant E2 harboring the H391R substitution exhibited no acyltransferase activity. The above data establish that the C219W-␣ and H156Y-␤ substitutions are the cause of E1-deficient or type IA MSUD, whereas the H391R-E2 mutation is the culprit for E2-deficient or type II MSUD. The remaining three mutations in Table I (i.e. intron 9 (IVS9 del[Ϫ7:Ϫ4]) and exon 10 (1,109 ins 8bp) type IB gene alterations and S133stop type II subunit truncation) are obligatory null mutants and therefore also responsible for MSUD.
Assembly State of Wild-type and Mutant E1 Proteins-We have shown previously that the folded E1␣ and E1␤ subunits assemble into a heterodimeric intermediate that subsequently dimerizes to form a functional native E1 heterotetramer (16,27). To dissect the effect of these novel MSUD mutations on E1 assembly, purified wild-type and mutant E1 proteins were subjected to a 10 -30% sucrose density gradient centrifugation, and collected fractions were analyzed by SDS-PAGE coupled with Coomassie Blue staining. Both ␣ and ␤ subunits of wild-type E1 migrate as a single species consistent with a 170-kDa heterotetramer (Fig. 1). The mutant E1 carrying the C219W-␣ substitution exists as two major species, one corresponding to the heterotetramer and the other migrating as a soluble high molecular weight form. A small amount of heterodimeric species that sediments close to the top of the gradient is also present in this mutant. The H156Y-␤ mutant E1 occurs in solution in multiple species ranging from heterodimers to heterotetramers to the higher molecular weight form.
Affinity of Mutant E1 for Cofactor ThDP-Effects of the above MSUD mutations on the ability of mutant E1 to bind cofactor ThDP were studied. The heterotetrameric fractions of wild-type and mutant E1 proteins separated on a FPLC Superdex-200 column were collected and used for cofactor binding studies. The heterodimeric species does not bind ThDP, since the cofactor-binding pocket is formed at the interface between two heterodimers. K d values for ThDP independent of the E1 concentration were determined by measuring tryptophan fluorescence quenching upon cofactor binding. Wild-type E1 shows a K d value of 1.52 M (Table II)  M were obtained for mutant E1 carrying the C219W-␣ and H156Y-␤ mutations, respectively. Results of these binding measurements indicate that the above type IA and type IB MSUD mutations severely reduce the affinity of mutant E1 for cofactor ThDP. Activation of Mutant E1 by Natural Osmolyte TMAO-We have shown previously that TMAO, a naturally occurring osmolyte, promotes assembly of the native E1 heterotetramer from otherwise assembly-incompetent heterodimers carrying MSUD mutations (28). To determine whether the present missense mutations affect subunit interactions, wild-type and mutant E1 proteins were incubated with 0 -2 M concentrations of TMAO. The osmolyte does not significantly affect wild-type E1 when assayed for activity of the reconstituted BCKD complex in the presence of E2 and E3 (Fig. 2). However, the mutant E1 containing the H156R-␤ type IB MSUD mutation was activated by TMAO from null to 6% of the wild-type BCKD activity, with the optimum activation occurring at 1.2 M TMAO concentration. In contrast, essentially no BCKD activity was recovered with the C219W-␣ variant incubated with the same concentration range of TMAO.
Activation of E1 Activity by Wild-type and Mutant E2 Proteins-Type II MSUD patients with the H391R mutation carry a normal E1 component of the BCKD complex. Therefore, the effect of wild-type and mutant E2 components on E1 activity was investigated. Individual and combined E2 domains (Fig.  3A) were also constructed and used in the same study. The basal E1 activity using the dye DCPIP as an artificial electron acceptor (Reaction 2) is independent of lip-E2 and E3 components, whereas BCKD activity (Reaction 1) requires the presence of the three catalytic components. As shown in Fig. 3B, the E1 component alone catalyzes the decarboxylation of substrate ␣-ketoisovalerate at a rate of 11.9 min Ϫ1 . lip-LBD2 at a 1:1 molar ratio to E1 has no effect on E1 activity. The addition of SBD resulted in a 2.4-fold increase in E1 activity. A similar 3-fold increase is obtained in the presence of lip-DD that contains both lip-LBD and SBD. Activation of E1 activity by 3.9and 3.5-fold, respectively, was obtained with lip-E2 and H391R lip-E2. The results indicate that the binding of E1 to SBD confers increased E1 activity, with the maximal activation obtained with a full-length E2 harboring the SBD sequence. The H391R mutation in the inner core domain of E2 is also capable of imparting a similar degree of stimulation to the wild-type E2 on E1 activity. DISCUSSION In the present study, we were able to produce sufficient quantities of mutant E1 proteins for biochemical characteriza-

FIG. 2. Activation of mutant E1 carrying MSUD mutations by trimethylamine N-oxide.
Wild-type and mutant E1 proteins (200 g/ml) were incubated with increasing concentrations of TMAO for 16 h in the presence of 0.5 M lysine. BCKD activity was measured spectrophotometrically using the reconstituted assay (Reaction 1). Both H156Y-␤ and C219W-␣ mutants exhibit no BCKD activity in the absence of TMAO. Activation of mutant H156Y-␤ (q) E1 in the presence of TMAO is expressed as percentage of wild-type BCKD activity, which remains relatively constant at 196 min Ϫ1 over the TMAO concentration range. No activation by TMAO was obtained with the mutant C219W-␣ (E) E1.

FIG. 3. Activation of E1 activity by wild-type and mutant E2
protein/domains. A, three separate domains of the full-length E2 chain. The wild-type E2 chain (residues 1-421) contains (from N to C termini) an LBD, an SBD, and an inner core domain that are linked by flexible hinge regions. The inner core domain confers the 24-meric structure for the full-length E2. Lys 44 (K44) is the lipoylation site, and His 391 (H391) is the catalytic residue that serves as a general base in the acyltransferase reaction (Reaction 4). B, E1 activity was measured with purified wild-type E1 (15 g) in the absence and presence of wild-type and mutant E2 protein/domains according to Reaction 7. The assays were carried out at nonsaturating ThDP concentration (10 M). The reduction DCPIP was monitored spectrophotometrically at 600 nm. The molar ratio of E1 to E2 domain was 1:1, and the molar ratio of E1 to full-length E2 was 12:1 at the fixed E1 concentration. The basal activity for E1 alone was 11.9 min Ϫ1 . The activation of E1 activity by wild-type and H391R E2 protein/domain is indicated by the -fold increase over the basal activity. tion using the chaperonin-augmented expression system (17,18), with the exception of the V69G-␤ mutation. The cotransformation with chaperonins GroEL and GroES is essential for proper folding and assembly of the native wild-type E1 heterotetramer in E. coli. The yield for the H156Y-␤ mutation was 20% of the wild type, and the yield for the C219W-␣ mutation was comparable with the wild type. The inability to express the V69G-␤ E1 protein suggests that this mutation causes a misfolding of the E1␤ subunit in E. coli, which cannot be corrected by the overexpression of chaperonins GroEL and GroES. The wild-type E2 24-mer can be readily expressed in XL1-Blue E. coli cells without cotransformation with the chaperonins, indicating that the intrinsic chaperonins are sufficient for the folding and assembly of E2. The H391R E2 mutation does not appear to affect folding and assembly of E2, since the yield of this mutant is comparable with that of the wild-type E2.
The crystal structure of human E1 was previously determined at 2.7-Å resolution (29). The overall structure is organized in a tetrahedral arrangement with each lobe roughly corresponding to one of the four subunits. The two active sites that accommodate cofactor ThDP are at the interfaces of ␣-␤Ј and ␣Ј-␤ subunits. Novel structural K ϩ ion-binding pockets are present in both ␣ and ␤ subunits. Our laboratory has recently refined the human E1 structure to 1.8-Å resolution (24). This high resolution structure is used to provide a structural basis for the present novel MSUD mutations in Israeli patients. As shown in Fig. 4A, the Mn 2ϩ ion in the ThDP-binding pocket coordinates to side chains of Asn 222 -␣ and Glu 193 -␣ and the main-chain carbonyl oxygen atom of Tyr 224 -␣ (not shown) as well as two phosphate oxygen atoms of cofactor ThDP. The Cys 219 -␣ residue is in the immediate vicinity of the ThDPbinding pocket. The replacement of Cys 219 -␣ by a large bulky Trp residue in the C219W-␣ mutation is likely to produce a clash with Glu 193 -␣ and Asn 222 -␣, thereby disrupting the octahedral coordination of the Mn 2ϩ ion. This explains the significant reduction in the affinity of this mutant for cofactor ThDP (Table II). The strained conformation surrounding the ThDPbinding site probably accounts for the loss of E1 activity in the C219W-␣ variant. However, the binding of ThDP cannot correct this structural defect, as indicated by the absence of activity for the BCKD complex reconstituted with this mutant and measured at a saturating cofactor concentration. Since this mutation is away from the subunit interface, the structural basis for the appearance of high molecular weight species in addition to the heterotetramer in this E1 mutant is not immediately clear.
The H156Y-␤ mutation occurs at a tight interface between ␤ and ␤Ј subunits along a pseudo-2-fold symmetry (Fig. 4B). In the wild-type structure, His 156 -␤ in the ␤ subunit is hydrogenbonded to Thr 284 -␤Ј on the opposite ␤Ј subunit. A substitution of the His 156 -␤ by a tyrosine residue disrupts this hydrogen bond and therefore impedes the ␤-␤Ј subunit interactions required for heterotetrameric assembly. This results in the formation of soluble heterodimer and soluble high molecular weight aggregates in addition to the heterotetramer. The defective subunit interactions leading to improper assembly apparently accounts for loss of E1 activity in this mutant. The assembly defect is confirmed by the restoration of the H156Y-␤ mutant BCKD activity from null to 6% of the wild-type by incubating with the naturally occurring osmolyte TMAO (Fig.  2). We have shown previously that TMAO forces the assembly of otherwise permanently locked inactive heterodimers caused by MSUD mutations into a partially active heterotetramer (28). TMAO promotes protein folding by preferential hydration of the exposed peptide backbone of an unfolded protein (31). This entropically unfavorable situation promotes the folding of the unfolded protein into a native conformation. The H156Y-␤ mutation is 15-20 Å away from the ThDP-binding site, and therefore the impaired ␤-␤Ј subunit interactions may also have an adverse effect on the cofactor binding. This is confirmed by the 2-order of magnitude increase in the binding constant (K d ) for ThDP over the wild-type E1. It is of interest that a second mutation involving His 156 -␤ was previously described in the Japanese population in which this residue is replaced by an arginine (32). The introduction of a large charged residue in the H156R-␤ mutation may have more severe consequences than the H156Y-␤ mutation, because the arginine residues may form an ion pair with Glu 151 -␤Ј on the 2-fold symmetry-related ␤Ј subunit (not shown). This aberrant interaction could interfere with the normal hydrogen bond formed between Glu 151 -␤Ј and Trp 227 -␤Ј in the same subunit, causing further instability in the E1 heterotetramer (29).
The V69G-␤ alteration occurs on strand c in the interior of the ␤ subunit (Fig. 4C). Strand c precedes helix 3 in the E1 active site, which carries the invariant Glu 76 -␤ that serves as the electron-withdrawing residue during ThDP-mediated decarboxylation. The Val 69 -␤ residue coordinates to Ile 43 -␤ on strand b, Leu 60 -␤ on helix 2, and Val 48 -␤ on the loop between strand b and helix 2 through hydrophobic interactions. The replacement of a valine by a glycine residue may disrupt these interactions as well as produce a kink in strand c. These combined effects could impart instability in the structure surrounding helix 3 that carries the invariant catalytic residue Glu 76 -␤. The net result is the misfolding of the ␤ subunit, which explains the inability to express the mutant E1 carrying the V69G-␤ substitution.
We proposed previously that the invariant His 391 in E2 is the essential catalytic residue that functions as a general base for the E2-mediated transacylation reaction (13,14). A replacement of His 391 by an asparagine, glutamine, or alanine results in a nearly complete inactivation of E2 activity. The structure of the E2 component of human BCKD complex has not been determined. Based on the available structure of the cognate transacetylase of the pyruvate dehydrogenase complex from Azotobacter vinelandii (33), a simplified reaction scheme involving His 391 and the conserved Ser 338 is shown in Fig. 5. The N-3 atom on the imidazole ring of His 391 on E2 abstracts a proton from the reactive thiol group of acyl-CoA. The nucleophilic sulfur atom, strengthened by the proton abstraction, then attacks the carbonyl carbon of S-acyldihydrolipoamide on the lipoyl-bearing domain of the same subunit, forming the putative tetrahedral intermediate. The free energy of the tran-sition state is significantly lowered by Ser 338 that confers a hydrogen bond to the charged transition state. The acyltransfer reaction is completed following the breakdown of the tetrahedral intermediate. It is unclear whether the formation or the breakdown step of the tetrahedral intermediate is rate-limiting in the E2-catalyzed acyltransfer reaction. A substitution of arginine for His 391 in the H391R E2 mutation in the Israeli MSUD patients prevents the formation of the tetrahedral intermediate, thereby thwarting the E2-mediated acyltransferase reaction.
Israeli MSUD patients are important ethnic groups for mutational studies because of the relatively high incidence of homozygosity associated with consanguinity (12). In this study, we show that among the eight Israeli MSUD patients studied, six are homozygously affected, each carrying one of the four novel point mutations (Table I). The only compound-heterozygous type IB patient (K.Y.) carries two mutant alleles in the E1␤ locus, one with a deletion in intron 9 and the other an insertion in exon 10, resulting in an obligatory null phenotype. The availability of homozygous MSUD patients will allow further studies of the correlation between a particular new mutation and one of the five established clinical phenotypes (2). The four homozygous patients carrying different types IA and IB missense mutations and the fifth compound-heterozygous patient with the two null alleles ubiquitously show the severe classic MSUD phenotype. The results are consistent with the prevailing view that a deficiency in the E1 component of the BCKD complex often manifests the classic phenotype. The exceptions are some of the missense mutations that do not significantly impede structure and function of the E1 component. Patients with this type of point mutations usually show milder variant forms of MSUD (34).
On the other hand, type II MSUD patients in the present study (H.C. and H.S.) carrying the homozygous H391R mutation in the full-length E2 subunit exhibit an unequivocal thiamin-responsive phenotype. These two patients presented with neonatal encephalopathy, which was completely eradicated by thiamin supplements at 100 mg/day while on a restrictive diet. On this regimen, the patients have been asymptomatic, and their plasma BCAA levels remained in the normal range for the next 4 years. The overriding feature of these two Israeli thiamin-responsive patients are 1) the presence of a obligatory wild-type E1 and a full-length mutant E2 protein and 2) the display of significant residual decarboxylation activity by intact cells incubated with ␣-keto[1-14 C]isovalerate. These criteria are met by an additional four established thiamin-responsive patients that we studied previously (8) including the first patient (WG-34) described by Scriver et al. (35), who was found to contain one allele expressing a full-length F215C mutant E2 (36) (Table  III). In contrast, the nine type II patients that manifest the classic MSUD phenotype including the two in this study do not contain at least one allele that produces a full-length mutant E2 protein. These null mutations, as defined by the absence of a full-length E2 protein, are accompanied by the complete absence of residual decarboxylation activity in intact cells derived from these patients.  6. A biochemical model for the thiamin-responsive MSUD phenotype. A, basal E1 activity in null type II (E2-deficient) MSUD. In the absence of E2 due to a null or truncated (e.g. S133stop) mutation, E1 is dephosphorylated, since the BCKD kinase is inactive. Thiamin supplements convert the E1 active site from partially occupied by cofactor ThDP to fully occupied. The thin curved arrow denotes the basal E1 activity that decarboxylates BCKAs to branched-chain fatty acids (Reaction 7), with DCPIP replaced by an unidentified naturally occurring electron acceptor. The basal E1 activity at 11.9 min Ϫ1 is 6% of the oxidative decarboxylation catalyzed by the BCKD complex (Reaction 1). In vivo, the basal E1 activity is too low to significantly degrade BCKAs, resulting in the classic MSUD phenotype. B, augmented E1 activity in the presence of a mutant (e.g. H391R) 24-meric E2. Binding of BCKD kinase to the lip-LBD domain of the mutant full-length E2 renders the kinase active. This results in a partially phosphorylated E1 (indicated by circled P) bound to the SBD domain. Thiamin supplements result in complete occupation of the E1 active site by the cofactor and the complete dephosphorylation of E1. More significantly, conformational changes in the E1 active site induced by binding to SBD result in a 4-fold increase in the rate of BCKA decarboxylation as indicated by the thick curved arrow. The stimulated E1 activity may account for significant residual activity of BCKA decarboxylation in thiamin-responsive MSUD cells. The dotted circle in the E2 inner core domain depicts the E2 active site that contains the H391R type II MSUD mutation.
Thiamin-responsive MSUD is unique among vitamin-treatable metabolic disorders in that the E1 component binds ThDP derived from thiamin but does not carry disease-causing mutations (37). In light of the apparent requirement for the presence of a full-length mutant E2 and the data from E1 activity measurements, a mechanism for the thiamin-responsive phenotype can be proposed (Fig. 6). In the E2 null mutations such as S133stop, the E1 component alone is capable of decarboxylating BCKAs according to Reaction 7 with the production of branched-chain fatty acids using a naturally occurring oxidant(s) as electron acceptor (Fig. 6A). In the absence of a full-length E2, the BCKD kinase is inactive, resulting in a completely dephosphorylated E1 in type II MSUD cells (38). Thiamin supplements result in a full occupation of the cofactor in the active site required for E1 activity. Without thiamin enrichment in the diet, the E1 active site of the BCKD complex is only partially saturated with ThDP in vivo (39). The basal E1 activity measured with DCPIP as an artificial electron acceptor is at 11.9 min Ϫ1 (Fig. 3). This decarboxylation rate is only 6% of that catalyzed by the BCKD complex (Reaction 1) and therefore is unable to efficiently dispose of the BCKAs in the cell, resulting in the classic MSUD phenotype.
In the presence of a mutant full-length E2 (e.g. H391R), the kinase bound to the LBD domain of the mutant E2 is active. This renders the wild-type E1 bound to the SBD domain of the mutant E2 partially phosphorylated (Fig. 6B). The thiamin treatment results in 1) the inactivation of the kinase and thus a fully dephosphorylated and functional E1 (40) and 2) the complete occupation of ThDP in the E1 active site as discussed above. The binding of the dephosphorylated holo-E1 to the SBD domain of H391R full-length E2 results in a 4-fold increase in E1 activity over the basal level in the absence of E2 (Fig. 3). The stimulation of basal E1 activity can be explained, in part, by an increased affinity of E1 for ThDP, which is induced by conformational changes in the active site upon E1 binding to the SBD domain of the fulllength mutant E2. We propose that this augmented E1 activity accounts for the significant residual decarboxylation activity in thiamin-responsive MSUD cells (Table III). Moreover, the presence of at least one allele expressing the fulllength mutant E2 is diagnostic for the thiamin-responsive phenotype. The H391R mutant E2 does not exhibit E2 activity as measured by Reaction 3. Thus, a mutant BCKD complex reconstituted with this E2 variant is not capable of catalyzing the oxidative decarboxylation of BCKAs according to Reaction 1. It is conceivable, however, that a BCKD complex reconstituted with a partially active mutant E2 (e.g. the full-length F215C mutant E2 in thiamin-responsive MSUD patient WG-34) (36) can still catalyze a partial oxidative decarboxylation of BCKAs according to Reaction 1. This explains the high (30 -40%) residual decarboxylation activity obtained with WG-34 cells, despite the presence of only one full-length E2-expressing allele (Table III).