Impaired Assembly of E1 Decarboxylase of the Branched-chain α-Ketoacid Dehydrogenase Complex in Type IA Maple Syrup Urine Disease*

The E1 decarboxylase component of the human branched-chain ketoacid dehydrogenase complex comprises two E1α (45.5 kDa) and two E1β (37.5 kDa) subunits forming an α2β2 tetramer. In patients with type IA maple syrup urine disease, the E1α subunit is affected, resulting in the loss of E1 and branched-chain ketoacid dehydrogenase catalytic activities. To study the effect of human E1α missense mutations on E1 subunit assembly, we have developed a pulse-chase labeling protocol based on efficient expression and assembly of human (His)6-E1α and untagged E1β subunits inEscherichia coli in the presence of overexpressed chaperonins GroEL and GroES. Assembly of the two35S-labeled E1 subunits was indicated by their co-extraction with Ni2+-nitrilotriacetic acid resin. The nine E1α maple syrup urine disease mutants studied showed aberrant kinetics of assembly with normal E1β in the 2-h chase compared with the wild type and can be classified into four categories of normal (N222S-α and R220W-α), moderately slow (G245R-α), slow (G204S-α, A240P-α, F364C-α, Y368C-α, and Y393N-α), and no (T265R-α) assembly. Prolonged induction in E. coli grown in the YTGK medium or lowering of induction temperature from 37 to 28 °C (in the case of T265R-α), however, resulted in the production of mutant E1 proteins. Separation of purified E1 proteins by sucrose density gradient centrifugation showed that the wild-type E1 existed entirely as α2β2 tetramers. In contrast, a subset of E1α missense mutations caused the occurrence of exclusive αβ dimers (Y393N-α and F364C-α) or of both α2β2 tetramers and lower molecular weight species (Y368C-α and T265R-α). Thermal denaturation at 50 °C indicated that mutant E1 proteins aggregated more rapidly than wild type (rate constant, 0.19 min−1), with the T265R-α mutant E1 most severely affected (rate constant, 4.45 min−1). The results establish that the human E1α mutations in the putative thiamine pyrophosphate-binding pocket that are studied, with the exception of G204S-α, have no effect on E1 subunit assembly. The T265R-α mutation adversely impacts both E1α folding and subunit interactions. The mutations involving the C-terminal aromatic residues impede both the kinetics of subunit assembly and the formation of the native α2β2 structure.

The E1 decarboxylase component of the human branched-chain ketoacid dehydrogenase complex comprises two E1␣ (45.5 kDa) and two E1␤ (37.5 kDa) subunits forming an ␣ 2 ␤ 2 tetramer. In patients with type IA maple syrup urine disease, the E1␣ subunit is affected, resulting in the loss of E1 and branched-chain ketoacid dehydrogenase catalytic activities. To study the effect of human E1␣ missense mutations on E1 subunit assembly, we have developed a pulse-chase labeling protocol based on efficient expression and assembly of human (His) 6 -E1␣ and untagged E1␤ subunits in Escherichia coli in the presence of overexpressed chaperonins GroEL and GroES. Assembly of the two 35 S-labeled E1 subunits was indicated by their co-extraction with Ni 2؉nitrilotriacetic acid resin. The nine E1␣ maple syrup urine disease mutants studied showed aberrant kinetics of assembly with normal E1␤ in the 2-h chase compared with the wild type and can be classified into four categories of normal (N222S-␣ and R220W-␣), moderately slow (G245R-␣), slow (G204S-␣, A240P-␣, F364C-␣, Y368C-␣, and Y393N-␣), and no (T265R-␣) assembly. Prolonged induction in E. coli grown in the YTGK medium or lowering of induction temperature from 37 to 28°C (in the case of T265R-␣), however, resulted in the production of mutant E1 proteins. Separation of purified E1 proteins by sucrose density gradient centrifugation showed that the wild-type E1 existed entirely as ␣ 2 ␤ 2 tetramers. In contrast, a subset of E1␣ missense mutations caused the occurrence of exclusive ␣␤ dimers (Y393N-␣ and F364C-␣) or of both ␣ 2 ␤ 2 tetramers and lower molecular weight species (Y368C-␣ and T265R-␣). Thermal denaturation at 50°C indicated that mutant E1 proteins aggregated more rapidly than wild type (rate constant, 0.19 min ؊1 ), with the T265R-␣ mutant E1 most severely affected (rate constant, 4.45 min ؊1 ). The results establish that the human E1␣ mutations in the putative thiamine pyrophosphate-binding pocket that are studied, with the exception of G204S-␣, have no effect on E1 subunit assembly. The T265R-␣ mutation adversely impacts both E1␣ folding and subunit interactions. The mutations involving the C-terminal aromatic residues impede both the kinetics of subunit assembly and the formation of the native ␣ 2 ␤ 2 structure.
In patients with maple syrup urine disease (MSUD) or branched-chain ketoaciduria, the activity of the BCKD complex is deficient. This leads to clinical manifestations including often fatal ketoacidosis, neurological derangements, and mental retardation (1). The molecular genetics of MSUD are heterogeneous as mutations in the E1␣, E1␤, E2, and E3 genes have been described (1,6). Based on the locus affected, genetic subtypes of MSUD have been proposed, with type IA referring to mutations in the E1␣ gene, type IB to the E1␤ gene, type II to the E2 gene, and type III to the E3 gene (1). It has been suggested that certain type IA MSUD missense mutations, for example Y393N-␣ (7) and Y368C-␣ (8), may impede the assembly of mutant E1␣ with normal E1␤ subunit, resulting in the degradation of E1 subunits in patient's cells.
We have recently established that chaperonins GroEL and GroES are essential for efficient folding and assembly of the E1 tetramer in Escherichia coli (9) and the E2 24-mer in vitro (10).
To gain insight into the biochemical basis of the apparently impaired assembly of E1 in type IA MSUD, we have co-expressed both mature mutant E1␣ and normal E1␤ in E. coli co-transformed with a second plasmid overproducing chaperonins GroEL and GroES. Pulse-chase labeling of both E1 subunits was carried out to measure the kinetics of assembly of the mutant E1␣ with normal E1␤ in the bacterial cell. The results showed a marked reduction in the rate of E1 assembly in certain E1␣ mutants compared with normal. It was also found that a subset of E1␣ mutations affect the assembly state of mutant E1 after the prolonged induction in E. coli. Thermostability and protease digestion studies further indicated these slowly assembled mutant E1 proteins had less stable conformations than the wild type. These results define the residues that are critical for subunit interactions and stability of E1 and have implications for understanding chaperonin-mediated biogenesis of hetero-oligomeric structures.  (11).
Construction of pHisT-E1 Prokaryotic Expression Vector-The 5Ј portion of the mature human E1␣ cDNA sequence was amplified from the pMAL-c2-hE1␣ expression vector (12) using an internal 22-mer antisense primer with sequence 5Ј-GTAACAGATATCGACCCTGTT-3Ј, and a 49-mer sense primer with sequence 5Ј-GGCTCTAGACTCGAGAA-TCTTTATTTtcaatcatctctggatgacaagc-3Ј to yield a 601-bp product. The sense primer adds exogenous sequence (shown in uppercase) to the 5Ј terminus of the mature E1␣ open reading frame (shown in lowercase). This exogenous sequence includes an XbaI restriction site (shown in bold), followed by sequence encoding the first six amino acids specific for the tobacco etch virus (TEV) protease cleavage (shown underlined). The seventh required amino acid for the TEV protease cleavage is supplied by the N-terminal serine of the mature E1␣ sequence.
To generate the pHisT-E1␣ expression vector, the 601-bp amplification product was cut with XbaI and NarI to yield a 454-bp fragment encoding the TEV cleavage site and the 5Ј portion of the E1␣ open reading frame. The pEBO-hbE1␣ expression vector (7) was cleaved with NarI and XhoI to yield a 1064-bp fragment that encoded the 3Ј portion of the mature E1␣ open-reading frame. Both fragments were ligated into the pTrcHisB expression vector (Invitrogen, Carlsbad, CA) digested with NheI and XhoI to yield the pHisT-E1␣ expression vector. To generate the pHisT-E1 expression vector, a BamHI/ScaI fragment (2,555 bp) comprising the trc promoter and the mature human E1␤ open reading frame was isolated from expression vector pKK-hE1␤ (13) and ligated into the corresponding sites in the host pHisT-E1␣ expression vector, yielding the pHisT-E1 expression vector. Mutant E1␣ variants of the pHisT-E1 expression vectors were constructed identically, except for the substitution of 1064-bp NarI/XhoI fragments isolated from variant pEBO-hbE1␣ plasmids (7) harboring the desired E1␣ mutation.
Expression and Purification of Recombinant E1 Proteins-E. coli strain CG712 (ES ts ) and the plasmid pGroESL, which overexpresses chaperonins GroEL and GroES (14), were kind gifts of Dr. Anthony Gatenby of DuPont Experimental Station, Wilmington, DE. CG712 cells co-transformed with GroESL and the wild-type or mutant pHisT-E1 plasmids were grown as an overnight culture at 37°C in YTGK medium. The medium was modified from the 2ϫ YT medium (15), and contained the following per liter: 16 g of yeast extract, 10 g of bacto-tryptone, 5 g of NaCl, 10 ml of glycerol, and 0.75 g of KCl; 100 mg of ampicillin and 12.5 mg of tetracycline. Antibiotics were added to maintain the expression plasmid (Amp r ) and the FЈ plasmid (Tet r ) that carried the lacI transcriptional repressor gene. The overnight culture was diluted 3:1000 into 1 liter of YTGK medium and grown with shaker aeration at 37°C to a measured A 550 of 0.6. The expression of (His) 6 -E1 was induced with 1 mM IPTG at 37°C for 15-20 h.
Pulse-Chase Labeling to Determine Kinetics of E1␣ and E1␤ Subunit Assembly-CG712 cells were transformed with the pGroESL plasmid and pHisT-hE1 expression vectors carrying the normal mature E1␤ cDNA and either normal or mutant His-tagged mature E1␣ cDNA. Cells were grown at 42°C to an A 595 of 0.8 in C2 broth minimal media. The C2 broth minimal medium was modified from the low sulfate and low amino acid content C broth medium described by Guzman-Verduzco and Kupersztoch (18) and contained the following per liter: 2 g of NH 4 Cl, 6 g of Na 2 HPO 4 ⅐7H 2 O, 3 g of KH 2 PO 4 , 3 g of NaCl, 6 g of yeast extract, 40 mol of TPP, 50 mg of carbenicillin, and 50 mg of chloramphenicol. Cells were pelleted and resuspended in one-fifth original volume of the same media without antibiotics and allowed to recover with shaking for 5 min at 37°C. Cells were subsequently induced with 2 mM IPTG for 5 min, pulsed with 50 Ci/ml [ 35 S]Cys/[ 35 S]Met (ICN Radiochemicals, Costa Mesa, CA) for 1 min, and chased with 3 volumes of the same media (without antibiotics) supplemented with 8 mg/ml each of non-radioactive L-cysteine and L-methionine. At specified time points following the chase, cell samples (1.8 ml) were taken and quickly frozen in liquid N 2 . Thawed samples were lysed by sonication, and supernatants after microcentrifugation were treated batchwise with an excess (15 l) of Ni 2ϩ -NTA resin. The resin was washed three times (total volume: 2.4 ml) with 15 mM imidazole in 100 mM KP i , pH 7.5, containing 2 mM MgCl 2 , 0.1 mM EDTA, 0.1 mM EGTA, 0.2 mM TPP, and 2 mM ␤-mercaptoethanol. Bound (His) 6 -tagged E1␣ and assembled untagged E1␤ polypeptides were eluted with 30 l of Laemmli SDS sample buffer (19) containing 50 mM EDTA. Eluted labeled polypeptides were analyzed by SDS-PAGE, and autoradiograms were obtained by storage phosphorimaging.
Analysis of Assembly State by Sucrose Density Gradient Centrifugation-Wild-type and mutant E1 proteins were fractionated on a 10 -25% sucrose density gradient (10 ml). Gradients were poured in 50 mM potassium phosphate, pH 7.5, 250 mM KCl, and 0.2 mM EDTA. Proteins (25-200 g) were applied to each gradient following elution from Ni 2ϩ -NTA resin and fractionation on Sephacryl S-100 HR column. The gradients were run at 41,000 rpm (210,000 ϫ g) in a Beckman SW-41 rotor for 18 h. Gradients were fractionated (700 l/fraction) from top to bottom. Each fraction was treated with 10 l of 12.5 mg/ml deoxycholate and 700 l of 15% trichloroacetic acid. Samples were incubated on ice for 20 min and then spun in a microcentrifuge for 15 min. Each pellet was washed with 700 l of acetone (80%), followed by a 5-min microcentrifuge spin and evacuation of the supernatant by vacuum. Pellets were resuspended in 10 l of H 2 O and 5 l of SDS-PAGE sample buffer, and the entire volume was applied to the gels for analysis by SDS-PAGE.
Thermal Denaturation of Normal and Mutant E1 Proteins-Thermal aggregation was monitored by measuring absorbance at 360 nm versus time in a Gilford response spectrophotometer equipped with a Peltier heating device as described previously (20). Wild-type or mutant E1 proteins (1.2 M, final concentration) were added to a buffer preheated to 50°C, which contained 50 mM KP i , 250 mM KCl, 0.5 mM ␤-mercaptoethanol, 0.2 mM EDTA, and 10% glycerol at pH 7.5 in a final volume of 0.5 ml. The temperature of samples in glass cuvettes (2 mm in width and 10 mm in light path length) was measured using a small-bead thermocouple. The effects of cofactors on thermal denaturation and aggregation was studied by adding 2 mM TPP and 1 mM MgCl 2 to the incubation mixture.
Thermal denaturation curves were analyzed as follows. The aggregation for each sample was allowed to proceed until no additional absorbance changes were detected. The final absorbance was taken as the 100% denaturation point. All the earlier data points relative to the final end point were used to calculate percentage of the protein that remained soluble. The log of these values were plotted against the incubation time, and the slopes of the lines were determined. The plots appeared to be pseudo-first order decays. The rate constant, k obsd , was calculated from these slopes using CA-Cricket Graph III version 1.01 for the Macintosh.
Expression and Purification of Normal (His) 6 -Human E1-In our earlier study of chaperonin-augmented expression of mammalian E1, a maltose-binding protein (MBP) ligand was fused to the N terminus of the E1␣ subunit. The presence of the MBP sequence increases the solubility of MBP-E1 and facilitates its purification by amylose resin affinity chromatography (12). In the present study, the MBP ligand is replaced with a (His) 6 -tag, which is linked to the N terminus of the mature E1␣ subunit through a TEV-protease recognition sequence (LEN-LYFQ). Co-expression of (His) 6 -E1␣ and untagged E1␤ (the pHis T-E1 plasmid) was carried out in an E. coli CG712 host, which contained a second plasmid GroESL that overproduced chaperonins GroEL and GroES. The cells grown in the YTGK medium were heat-shocked at 42°C for 4 h, followed by induction at 37°C with IPTG for 16 h. The cell lysate was purified by Ni 2ϩ -NTA column chromatography. Fig. 1 (lower panel) shows the elution profile with an imidazole gradient. The upper panel shows the SDS-PAGE profile of column fractions stained with Coomassie Blue. The appearance of the (His) 6 -E1 tetramer coincides with the presence of E1 activity as measured by a spectrophotometric assay with 2,6-dichlorophenol indophenol as an electron acceptor. The earlier fractions corresponding to the GroEL-E1␣-E1␤ complex may represent a folding intermediate (see "Discussion").
Measurements of E1 Assembly Kinetics by Pulse-Chase Labeling-Our previous data suggested that the Y393N-␣ mutation in MSUD impedes the assembly of the mutant E1␣ subunit with normal E1␤, resulting in the preferential degradation of the latter subunit in patient's cells (7). In the present study, we set out to measure the kinetics of assembly of normal and the nine MSUD mutant E1␣ subunits, including Y393N-␣, with normal E1␤ by pulse-chase labeling. The method was based on the efficient expression of (His) 6 -E1 in the presence of excess chaperonins GroEL and GroES as described above. CG712 cells co-expressing (His) 6 -E1␣, untagged E1␤, GroEL, and GroES were grown in the C2 broth minimal medium and heat-shocked at 42°C for 4 h, followed by induction with IPTG for 5 min at 37°C. The cells were pulsed with [ 35 S]cysteine/[ 35 S]methionine for 1 min and then chased with unlabeled amino acids from 2 to 120 min. Lysates prepared from cells harvested at different times were purified by Ni 2ϩ -NTA affinity chromatography. The eluted radiolabeled polypeptides were separated by SDS-PAGE, and autoradiograms were obtained by storage phosphorimaging. Since the E1␤ subunit was untagged, the co-purification of this subunit with the (His) 6 -E1␣ subunit by Ni 2ϩ -NTA indicated assembly of the two polypeptides synthesized during the 1-min pulse. Fig. 2 shows that the assembly of normal E1␤ with normal (His) 6 -E1␣ occurs as early as 10 min in the chase and reaches a plateau at 30 min. Similar results were obtained with N222S-␣ (second panel from the top) and R220W-␣ (data not shown). These two mutations are in the category of a normal assembly with the E1␤ subunit. A second group of mutations represented by the G245R-␣ showed that significant assembly with E1␤ did not occur until 30 min into the chase and plateaued at 60 min. Mutations that produce this sluggish assembly kinetics belong to the category of moderately slow assembly. A third group of E1␣ mutations comprising G204S-␣, A240P-␣, F364C-␣, Y368C-␣ and Y393N-␣ did not generate detectable assembly with normal E1␤ during the 2-h chase. This is indicated by the absence of the E1␤ subunit in the autoradiogram. This group of mutations are classified as slow assembly, as they produce the assembled mutant E1 only after a prolonged growth of the transformed cells at 37°C for 16 h (see below). The fourth category of no assembly is represented by the T265R-␣ subunit, which is not soluble when cells are grown at 37°C, as indicated by the rapid disappearance of the mutant E1␣ subunit in the chase. This resulted in a complete absence of assembly with the normal E1␤ subunit, even after a prolonged growth at 37°C. However, a significant amount of the assembled T265R-␣ E1 was produced, when the induction temperature was lowered from 37 to 28°C (see below).
Measurements of Total and Soluble E1␣ and E1␤ Subunits-The levels of total recombinant E1␣ and E1␤ polypeptides in E. coli cells were measured by Western blotting. Cells co-transformed with pHisT-E1 and pGroESL plasmids and grown in the C2 broth minimal medium were heat-shocked as described above and induced with IPTG for 12 h. Cells were harvested and total lysates prepared by sonication followed by solubilization in an SDS-PAGE sample buffer. After SDS-PAGE, the samples were subjected to Western blotting using the antibody to E1␣ or E1␤ as a probe. Fig. 3A shows that the levels of normal and mutant E1␣ are comparable. More significantly, the levels of normal E1␤ in cells expressing normal or mutant E1␣ are relatively constant. The data rule out the possibility that the reduced level or absence of normal E1␤ assembled with the mutant E1␣ subunits (Fig. 2) is caused by aberrant E1␤ expression. Fig. 3B shows Western blotting of total soluble E1␣ (normal or mutant) and E1␤ (normal) subunits in E. coli cells when co-expressed at 37°C. Five (G204S-␣, R220W-␣, N222S-␣, A240P-␣, and G245R-␣) of the nine E1␣ mutants studied are associated with wild-type levels of soluble E1␣ and E1␤ subunits. In contrast, the level of T265R-␣ is markedly reduced with a concomitant near-absence of the normal E1␤ subunit. The results support the conclusion drawn from pulsechase labeling (Fig. 2), which indicates that the mutant T265R-␣ is largely insoluble at 37°C and fails to assemble with the normal E1␤ subunit. The unassembled E1␤ subunit, while expressed at a normal rate (Fig. 3A), became aggregated and was removed from the supernatant after centrifugation. As for the F364C-␣, Y368C-␣, and Y393N-␣ mutants (Fig. 3B) 2. Kinetics of normal and aberrant E1 subunit assembly analyzed by pulse-chase labeling in E. coli. CG712 cells were co-transformed with the pGroESL plasmid and the pHisT-E1 plasmid carrying the (His) 6 -tagged wild-type or mutant E1␣ cDNA and the untagged normal E1␤ cDNA. Cells grown in the C2 broth minimal medium were induced for expression of E1 subunits with IPTG for 5 min and pulsed with [ 35 S]cysteine/[ 35 S]methionine (50 Ci/ml) for 1 min. The cells were then chased with 8 mg/ml non-radioactive cysteine and methionine for up to 120 min. Lysates prepared from cells harvested at different chase times were mixed batchwise with Ni 2ϩ -NTA resin. 35 S-Labeled (His) 6 -E1␣ subunits assembled with untagged E1␤ subunits or GroEL were eluted with the SDS sample buffer (19). The eluted radiolabeled polypeptides were analyzed by SDS-PAGE, and autoradiograms were obtained by storage phosphorimaging. The aberrant assembly of mutant (His) 6 -E1␣ subunits with the normal E1␤ subunit is classified into four categories based on the kinetics of appearance of the co-purified E1␤ subunit: normal assembly (10 -20 min), moderately slow assembly (30 -40 min), slow assembly (Ͼ2 h), and no assembly (indefinite time).

FIG. 3. Western blot analysis of the levels of normal and mutant E1 subunits expressed in E. coli.
The expression of the normal or mutant (His) 6 -E1␣ subunit and the normal E1␤ subunit in CG712 (ES ts ) cells co-transformed with the wild-type or mutant pHisT-E1 plasmid and the pGroESL plasmid was induced with IPTG in the C2 broth minimal medium at 37°C for 12 h. Crude lysates were prepared by sonication, and the supernatant was obtained by removing the pellet after centrifugation for 2 h at 29,000 ϫ g. Samples of both crude lysates and the supernatant were subjected to SDS-PAGE, and separated polypeptides were electroblotted to polyvinylidene difluoride membranes. The blots were probed with anti-E1␣, stripped of the antibody, and reprobed with anti-E1␤. A, crude lysates. B, supernatants.
Proteins-The co-purification of (His) 6 -E1␣ with untagged E1␤ cannot discern the assembly state of assembled E1 subunits. To address this question, E. coli cells in the C2 broth medium which expressed wild-type E1 subunits were grown for 20 h. E1 subunits in cells harvested at different induction times were purified by Ni 2ϩ -NTA resin and subsequently subjected to size fractionation on a TSK-G3000SW XL column by HPLC. The molecular weight of assembled E1 subunits species is inversely proportional to the retention time on HPLC. Fig. 4A shows that (His) 6 -E1␣ and untagged E1␤ subunits form mostly ␣␤ dimers with molecular mass in the 80-kDa range at the 1-h induction time. The Ni 2ϩ -NTA extractable ␣␤ dimers are the predominant species with the appearance of a minor species of ␣ 2 ␤ 2 tetramers (165 kDa) when induced for 2 h. Conversely, at the 3and 20-h induction times the major species is ␣ 2 ␤ 2 tetramers and the minor species ␣␤ dimers. The data indicate that (His) 6 -E1␣ and E1␤ subunits initially form the ␣␤ dimers which later slowly dimerize to produce ␣ 2 ␤ 2 tetramers. The fractions collected at different retention times from HPLC were analyzed for E1 activity using an assay mixture containing added excess lipoylated recombinant E2 and recombinant E3, which allowed one to measure reconstituted BCKD activity. This radiochemical assay for BCKD activity is 15-fold more sensitive than the one measuring the E1 component activity alone. As shown in Fig. 4B, the fraction collected at the 9-min retention time, which corresponds to the ␣ 2 ␤ 2 tetramer in the 20-h induction lysate, contains the peak enzyme activity. The fractions corresponding to ␣␤ dimers with a retention time of 10.3 min in 2and 3 h-induction lysates do not have enzyme activity.
Despite the slow assembly of mutant E1␣ subunits with normal E1␤ as determined by pulse-chase labeling, prolonged induction with IPTG in E. coli grown in the YTGK medium resulted in the production of mutant E1 proteins. Bacterial lysates prepared from E. coli cells after the 16-h induction at 37 or 28°C were purified by Ni 2ϩ -NTA column, followed by gel filtration on Sephacryl S-100 column. The purified wild-type and mutant E1 proteins were subjected to sucrose density gradient centrifugation to analyze their subunit assembly state. Fig. 5 depicts the sedimentation profiles of E1 proteins as determined by SDS-PAGE analysis of gradient fractions after the centrifugation. The wild-type E1 protein induced at 37°C migrated as a tetrameric species of 165 kDa with the ␣ 2 ␤ 2 structure (fractions 7-9). The trace amounts of E1␣ and E1␤ subunits at the bottom of the gradient were the result of slight aggregation that occurred during the centrifugation. In contrast, the Y393N-␣ mutant E1 expressed at 37°C migrated as an ␣␤ dimeric species with a molecular mass of 83 kDa (fractions 3-6). The mutant E1 with the F364C-␣ mutation also occurred entirely as ␣␤ dimers in sucrose density gradient centrifugation (data not shown). Interestingly, the mutant FIG. 4. Assembled wild-type E1 species during 20 h induction in the C2 broth minimal medium. CG712 E. coli cells co-transformed with the pHis T-E1 and the pGroESL plasmids were induced for the expression of wild-type (His) 6 -E1␣ and E1␤ subunits with IPTG in the C2 broth minimal medium. Cells were harvested at different times during the 20 h induction. Assembled E1 subunits were extracted from the cell lysate with Ni 2ϩ -NTA resin. Purified E1 species in 50 mM potassium phosphate, pH 7.5, 0.2 mM EDTA, and 250 mM KCl were subjected to size fractionation by HPLC on a TSK-G3000SW XL column. Fractions collected at different retention times were analyzed by SDS-PAGE and Coomassie Blue staining. E1 activity in eluted fractions was assayed using a reconstituted system with addition of recombinant E2 and recombinant E3. Radiolabeled ␣-keto-[1- T265R-␣, when expressed at 28°C, was able to remain soluble and assemble with the normal E1␤ subunit. The sedimentation profile indicated that T265R-␣ mutant E1 migrated predominantly as tetramers, although lesser amounts of lower molecular weight species were present and sedimented in early gradient fractions. Similarly, the mutant E1 bearing the Y368C-␣ mutation sedimented as approximately equal amounts of both tetramers and lower molecular weight species when expressed at 37°C. The mutant E1 proteins carrying the remaining E1␣ mutations (R220W-␣, N222S-␣, G245R-␣, and A240P-␣) were present only as tetramers (Table I). The assembly state of the above wild-type and mutant E1 proteins was confirmed by the elution profiles from TSK-G3000SW XL sizing column on HPLC and Sephacryl S-100 column (data not shown).
Activity Levels and Stability of Wild-type and Mutant E1 Proteins-Wild-type and mutant E1 proteins produced in E. coli grown on the YTGK medium were purified by Ni 2ϩ -NTA affinity column and gel filtration on Sephacryl S-100 column. The E1 activity of normal and mutant proteins was assayed by the radiochemical method with 2,6-dichlorophenol indophenol as an electron acceptor. As shown in Table I, only the mutant E1 carrying N222S-␣ or G245R-␣ has residual E1 activity (1.37 and 2.66% of normal activity, respectively). The mutant E1 proteins containing each of the remaining seven E1␣ mutations do not have detectable E1 catalytic activity.
The thermal stability of the above purified normal and mutant E1 proteins was studied by heat denaturation at 50°C. Light scattering at 360 nm as a result of protein aggregation was monitored. The fraction of soluble proteins was calculated from the progress curve and expressed as log % values versus the incubation time. Fig. 6 shows that the wild-type E1 at 1.21 M in the presence of 2 mM TPP is most stable with a denaturation rate constant of 0.08 min Ϫ1 . Similar denaturation rate constants were obtained at 0.97, 0.48, 0.24, and 0.12 M concentrations of E1 (data not shown). The data support the thesis that the aggregation of E1 is caused by heat-induced conformational changes rather than high concentrations of E1. In the absence of TPP, the denaturation rate constant of wild-type E1 increased to 0.19 min Ϫ1 . TPP had no effect on the thermal stability of mutant E1 proteins. Among the nine mutant E1 proteins studied, the one containing N222S-␣ was more stable than wild-type E1 (in the absence of TPP) with a denaturation rate constant of 0.14 min Ϫ1 . The mutant E1 protein carrying the remaining E1␣ mutations are less stable than the wild-type E1 (ϮTPP). The mutant E1 that harbors the T265R-␣ mutation is the least stable with a denaturation rate constant of 4.45 min Ϫ1 .
The susceptibility of wild-type E1 tetramers and Y393N-␣ E1 dimers to proteolysis was also studied. The E1 proteins were incubated with different concentrations of trypsin (protein/ trypsin ϭ 250 to 25,000:1, w/w) at 0°C for 20 min, followed by termination of the digestion with 10 mM PMSF. Fig. 7 shows that Y393N-␣ dimers are markedly more susceptible to the tryptic digestion than wild-type tetramers as analyzed by SDS-PAGE. DISCUSSION The major focus of this investigation is to characterize the effect of human E1␣ mutations in type IA MSUD on the assembly and stability of mutant E1 proteins. For these studies,

TABLE I
Assembly state and specific activities of wild-type and mutant E1 proteins Wild-type and mutant E1 proteins were expressed at 37°C in E. coli grown on the YTGK medium, except for the T265R-␣E1, which was expressed at 28°C. Recombinant proteins were purified by Ni 2ϩ -NTA affinity column and gel filtration on Sephacryl S-100 column. The purified wild-type and mutant E1 proteins were analyzed for the subunit assembly state by sucrose density gradient centrifugation as described in the legend to Fig. 5. E1 activity was assayed using the radiochemical assay with 2,6-dichlorophenol indophenol as an electron acceptor.

FIG. 5. Assembly state of wild-type and mutant E1 proteins as analyzed by sucrose density centrifugation.
Wild-type and mutant E1 carrying different mutations in the E1 subunits were expressed in co-transformed CG712 cells and purified by Ni 2ϩ -NTA extraction, followed by gel filtration on Sephacryl S-100 HR column. Purified wild-type and mutant E1 protein were fractionated on 10 -25% 10 ml of sucrose density gradient, which was spun at 210,000 ϫ g in a Beckman SW-41 rotor for 18 h. Fractions (0.7 ml each) were collected from top to bottom and analyzed by SDS-PAGE and Coomassie Blue staining. The molecular mass markers used were as follows: bovine serum albumin (68 kDa), E3 (110 kDa), E1 tetramers (165 kDa), and GroEL (840 kDa).
we have developed efficient bacterial expression systems for folding and assembly of E1 ␣ 2 ␤ 2 tetramers. We showed previously that co-expression of mature MBP-E1␣ and E1␤ sequences of the human BCKD complex in the same E. coli cells is essential for MBP-E1 assembly; however, the yield was very low (12). Co-transformation of a second plasmid that overexpressed GroEL and GroES into the same E. coli cell resulted in a 500-fold increase in the yield of active MBP-E1 tetramers (9). In the present study, a (His) 6 affinity tag is fused to the mature E1␣ N terminus through a TEV protease recognition site. Cotransformation with the pGroESL plasmid was also found necessary and sufficient for productive folding and assembly of (His) 6 -E1. The results argue against the suggestion that the dependence of human E1 on chaperonins for a high yield is due to the presence of the MBP sequence in the chimeric E1␣ polypeptide (23). Our recent in vitro refolding results indicate that the reconstitution of untagged E1, MBP-E1, and (His) 6 -E1 show the same chaperonin-dependent kinetics. 2 The findings further established that productive folding and assembly of mature human E1 have an absolute requirement for enrichment for chaperonins GroEL and GroES and are not affected by the presence of affinity tags. Thus, the pulse-chase labeling protocol developed in this study provides the first approximation of the rate of E1 subunit assembly under optimal conditions through the augmentation of bacterial chaperonins that are homologue of mitochondrial chaperonins Hsp60 and Hsp10, respectively.
The wild-type and mutant human E1␣ and the wild-type E1␤ subunits are expressed at relatively equal efficiencies and are stable, as indicated by Western blotting of the total crude lysates prepared 12 h after induction (Fig. 3A). This allows one to follow the fates of E1␣ and E1␤ subunits synthesized during the 1-min window of pulse-labeling. The presence of the (His) 6 tag in the wild-type and mutant E1␣ subunits facilitates the isolation of the 35 S-labeled subunit. One can measure the kinetics of the E1␤ assembly with the wild-type and mutant E1␣ subunits by the co-purification of the untagged E1␤ with (His) 6 -E1␣ with the Ni 2ϩ -NTA resin as a function of time. The equally strong signals of E1␣ and E1␤ subunits during the 2-h chase ( Fig. 2A) indicate that both subunits are efficiently synthesized. The total numbers of cysteine and methionine residues in the E1␣ and E1␤ subunits are similar, which are 16 and 15, respectively. The autoradiogram of the co-purified E1 subunits as separated by SDS-PAGE cannot discern the assembly state of the associated subunits. However, size fractionation of the pulse-chase labeled products by HPLC show that the wild-type E1␣ and E1␤ assemble during the 2-h chase occur predominantly as inactive ␣␤ dimers, which are later converted to active ␣ 2 ␤ 2 tetramers (Fig. 4, A and B). It is noteworthy that a significant amount of the wild-type ␣␤ dimeric intermediate was observed only when E. coli cells were grown on the C2 broth minimal medium. When the bacterial cells were cultured on the YTGK medium, wild-type E1 was expressed predominantly as ␣ 2 ␤ 2 tetramers (Fig. 5) with little or no accumulation of ␣␤ dimers during the 16-h induction period (data no shown). The factors responsible for the apparent effects of culture media on the accumulation of the wild-type dimeric intermediate during E1 assembly are currently unknown. The major differences between the two bacterial culture media lie in the fact that the C2 minimal medium is low in the content of SO 4 2Ϫ and amino acids when compared with the YTGK medium (18). Possible effects of these ingredients on the dimerization of wild-type ␣␤ dimers are under investigation.
It is of interest that a weak GroEL signal co-purifies with wild-type (His) 6 -E1␣ and E1␤ at 120 min into the chase ( Fig.  2A). This apparent ternary complex is also observed in the early fractions of the Imidazole gradient during purification of wild-type (His) 6 -E1 (Fig. 1). The GroEL-E1␣-E1␤ ternary complex is a productive intermediate at a later step of the chaperonin-mediated assembly of E1 ␣ 2 ␤ 2 tetramers. 2 Only a FIG. 6. Rates of thermal aggregation of wild-type and mutant E1 proteins at 50°C. Purified wild-type and mutant E1 proteins (100 g) were added to 0.5 ml of 50 mM KP i , pH 7.5, 250 mM KCl, 10% glycerol, 0.5 mM ␤-mercaptoethanol, and 0.2 mM EDTA preheated to 50°C in a 2 ϫ 10-mm cuvette. Aggregation at 50°C was monitored by light scattering at 360 nm wavelength as a function of time. The plateau of the progressive curve is considered 100% aggregation and was used to calculate percent soluble E1 at a given time point. The slope of percent soluble E1 versus time was used to calculate the rate constant, k obsd . q, wild type (ϩTPP); E, wild type (no addition); ϫ, G204S; f, R220W; Ⅺ, N222S; OE, A240P; ‚, T265R; ƒ, G245R; , Y393N. FIG. 7. Susceptibility of wild-type E1 tetramer and Y393N-␣ E1 dimers to tryptic digestion. Purified wild-type E1 tetramers or Y393N-␣ E1 dimers at 50 g/0.1 ml were incubated with trypsin for 20 min on ice at the trypsin-to-E1 ratio (w/w) indicated. The digestion was terminated by addition of PMSF (10 mM final concentration). The sample was analyzed by SDS-PAGE and Coomassie Blue staining.
GroEL-E1␣ binary complex is observed in F364C-␣ because the assembly of E1␤ with mutant E1␣ did not occur within the 2-h chase. The weak and sub-stoichiometric signal of GroEL relative to E1 subunits is a result of isotopic dilution by the overabundance of unlabeled GroEL in E. coli.
The assembled mutant ␣␤ dimers were produced in E. coli grown on the YTGK medium after the 16-h induction with "slow assembly" E1␣ mutants including Y393N-␣ and F364C-␣.
The results indicate that this group of mutations not only reduces the rate of the assembly of the mutant E1␣ with normal E1␤ but also prevents conversion of the dimeric assembly intermediate into the stable ␣ 2 ␤ 2 structure of wild-type E1. The production of these mutant ␣␤ dimers apparently is not affected by growth media, as the expression of mutant E1 carrying these mutations in E. coli grown in the C2 broth minimal medium also resulted in the expression of exclusive dimers. In vitro reconstitution of the 6 M urea-denatured mutant Y393N-␣ E1 in the presence of chaperonins GroEL and GroES also resulted exclusively in mutant ␣␤ dimers. 2 The unstable Y393N-␣ dimeric intermediate as demonstrated by its propensity for thermal aggregation and proteolytic digestion compared with the wild-type tetramer explains the markedly reduced levels E1␣ and E1␤ subunits in cells from Mennonite MSUD patients homozygous for the Y393N-␣ mutation (24,25). The present study establishes that C-terminal aromatic residues (F364-␣, Y368-␣, and Y393-␣) in the E1␣ subunit are crucial for proper E1 assembly. The important roles of the C terminus in subunit assembly and protein interactions have been demonstrated. For example, the C-terminal 25 amino acid residues of the herpes simplex virus type 1 UL26.5 protein are required for the assembly of the icosahedral capsid shell (26). In the case of the E2 core of the related pyruvate dehydrogenase complex from Azotobacter vinelandii, the C-terminal residues 632-637 comprise a 3 10 -like helix (H6) which acts as a "hydrophobic knob" that fits into a "hole" in the 2-fold related subunit to produce the 24-mer cubic assembly (27). Introduction of a polyhistidine tag into the C terminus of BCKD-E2, which is highly homologous to the bacterial pyruvate dehydrogenase-E2, results in the formation of stable trimers instead of the native 24-mer structure (data not shown).
The T265R-␣ subunit, when expressed at 37°C, was largely insoluble even in the presence of excess chaperonins GroEL and GroES. This was indicated by the rapid disappearance of the mutant E1␣ signal in the soluble fraction during the 2-h chase ( Fig. 2A); however, the level of the T265R-␣ subunits in total crude lysates was comparable to that of wild type (Fig.  3A). These results strongly suggest that the T265R-␣ residue is important for proper folding of the E1␣ subunit. It is also of interest that lowering of the induction temperature from 37 to 28°C resulted in the production of a significant amount of assembled mutant E1 protein carrying the T265R-␣ mutation. The yield of the mutant E1 was 5 mg/liter culture at 28°C compared with 20 -40 mg/liter culture for the wild-type E1 expressed at 37°C (data not shown). The finding is consistent with the thesis that lowering the expression temperature slows the folding kinetics of the nascent peptide, thereby reducing the probability of the off-pathway folding reactions, as demonstrated by the expression of rabbit muscle phosphorylase (28) and the human E1␤ (13). However, the assembled T265R-␣ mutant E1 has a grossly altered conformation, which renders it very unstable as indicated by its most rapid thermal aggregation at 50°C among the nine E1␣ mutants studied. This unstable conformation is also manifested by the apparent dissociation of the mutant tetramers to lower molecular weight species as detected by sucrose density gradient centrifugation. The current data indicate that the T265R-␣ residue also plays a key role in subunit interactions and are consistent with the location of this residue at the putative subunit-interaction site conserved between BCKD and pyruvate dehydrogenase E1 proteins (29).
The crystal structure of the E1 ␣ 2 ␤ 2 has not been solved, but structures are known for the related TPP-dependent proteins transketolase (30) and pyruvate decarboxylase (31) from Saccharomyces cerevisiae and pyruvate oxidase from Lactobacillus plantarum (32). The transketolase is a homodimer, whereas the human E1 is a tetramer made up of two non-identical subunits. Sequence alignment between the two enzymes shows that the highly conserved TPP-binding pocket in E1 is composed of residues from both E1␣ and E1␤ subunits (33,34). Aromatic residues from E1␤ form a hydrophobic pocket to accommodate the pyrimidium and thiazolium rings of cofactor TPP. On the other hand, the highly conserved TPP-binding motif GDG(X) 22-28 NN, which was first described by Hawkins et al. (35) and is essential for binding the pyrophosphate moiety, is located in the E1␣ subunit (Fig. 8). It should be mentioned that a D440E mutation introduced via mutagenesis into this motif in pyruvate decarboxylase from Zymomonas mobilis yielded a homodimeric enzyme with reduced affinity for TPP, in contrast to the wild-type enzyme which exists as a homotetramer (36). It was proposed that deficient TPP binding may have caused a failure in the conversion of the mutant dimeric forms into native tetramers. However, the occurrence of ␣␤ dimers in the mutant E1 carrying Y393N-␣ and F364C-␣ substitutions is likely through a different mechanism, since these residues are not involved in TPP binding and are located in the C-terminal region. G204S-␣, R220W-␣, and N222S-␣ that are affected in type IA MSUD are residues within the TPP-binding motif (Fig. 8). Specifically, N222S-␣ aligns with Asn-187 in the yeast transketolase and provides a ligand to this pentameric coordination involving the Mg 2ϩ cation. The N222S-␣ mutation conceivably disrupts the pentameric coordination, resulting in the inability of E1 to bind the pyrophosphate moiety of TPP and the loss of E1 catalytic function. However, the N222S-␣ mutation is without effect on the assembly of the mutant E1␣ with normal E1␤, as determined by pulse-chase labeling (Fig. 2A). The R220W-␣ mutation, which is also located in the pyrophosphate moiety binding site, also has no adverse effect on E1 subunit assembly. In contrast, G204S-␣ mutation, which is presumably located at the interface between the two non-identical subunits of E1, based on the yeast transketolase structure (37), impedes the assembly of the mutant E1␣ subunit with the normal E1␤ unit.