Assembly and full functionality of recombinantly expressed dihydrolipoyl acetyltransferase component of the human pyruvate dehydrogenase complex.

The dihydrolipoyl acetyltransferase (E2) component of mammalian pyruvate dehydrogenase complex (PDC) consists of 60 COOH-terminal domains as an inner assemblage and sequentially via linker regions an exterior pyruvate dehydrogenase (E1) binding domain and two lipoyl domains. Mature human E2, expressed in a protease-deficient Escherichia coli strain at 27°, was prepared in a highly purified form. Purified E2 had a high acetyltransferase activity, was well lipoylated based on its acetylation, and bound a large complement of bovine E1. Electron micrographs demonstrated that the inner core was assembled in the expected pentagonal dodecahedron shape with E1 binding around the inner core periphery. With saturating E1 and excess dihydrolipoyl dehydrogenase (E3) but no E3-binding protein (E3BP), the recombinant E2 supported the overall PDC reaction at 4% of the rate of bovine E2·E3BP subcomplex. The lipoates of assembled human E2 or its free bilipoyl domain region were reduced by E3 at rates proportional to the lipoyl domain concentration, but those of the E2·E3BP were rapidly used in a concentration-independent manner consistent with bound E3 rapidly using a set of lipoyl domains localized nearby. Given this restriction and the need for E3BP for high PDC activity, directed channeling of reducing equivalents to bound E3 must be very efficient in the complex. The recombinant E2 oligomer increased E1 kinase activity by up to 4-fold and, in a Ca2+-dependent process, increased phospho-E1 phosphatase activity more than 15-fold. Thus the E2 assemblage fully provides the molecular intervention whereby a single E2-bound kinase or phosphatase molecule rapidly phosphorylate or dephosphorylate, respectively, many E2-bound E1. Thus, we prepared properly assembled, fully functional human E2 that mediated enhanced regulatory enzyme activities but, lacking E3BP, supported low PDC activity.

The dihydrolipoyl acetyltransferase (E2) component of mammalian pyruvate dehydrogenase complex (PDC) consists of 60 COOH-terminal domains as an inner assemblage and sequentially via linker regions an exterior pyruvate dehydrogenase (E1) binding domain and two lipoyl domains. Mature human E2, expressed in a protease-deficient Escherichia coli strain at 27°, was prepared in a highly purified form. Purified E2 had a high acetyltransferase activity, was well lipoylated based on its acetylation, and bound a large complement of bovine E1. Electron micrographs demonstrated that the inner core was assembled in the expected pentagonal dodecahedron shape with E1 binding around the inner core periphery.
With saturating E1 and excess dihydrolipoyl dehydrogenase (E3) but no E3-binding protein (E3BP), the recombinant E2 supported the overall PDC reaction at 4% of the rate of bovine E2⅐E3BP subcomplex. The lipoates of assembled human E2 or its free bilipoyl domain region were reduced by E3 at rates proportional to the lipoyl domain concentration, but those of the E2⅐E3BP were rapidly used in a concentration-independent manner consistent with bound E3 rapidly using a set of lipoyl domains localized nearby. Given this restriction and the need for E3BP for high PDC activity, directed channeling of reducing equivalents to bound E3 must be very efficient in the complex.
The recombinant E2 oligomer increased E1 kinase activity by up to 4-fold and, in a Ca 2؉ -dependent process, increased phospho-E1 phosphatase activity more than 15-fold. Thus the E2 assemblage fully provides the molecular intervention whereby a single E2-bound kinase or phosphatase molecule rapidly phosphorylate or dephosphorylate, respectively, many E2-bound E1. Thus, we prepared properly assembled, fully functional human E2 that mediated enhanced regulatory enzyme activities but, lacking E3BP, supported low PDC activity.
Pyruvate dehydrogenase complexes from various sources are among the largest enzyme systems that serve strategic roles in metabolism (1,2). The mammalian complex has a highly organized structure in which the dihydrolipoyl transacetylase (E2) 1 component has a central role in the organization and integrated chemical reactions of the complex, and supports enhanced functioning of dedicated kinase and phosphatase components (3)(4)(5)(6)(7)(8). The other components of the mammalian complex required for the overall reaction are: the pyruvate dehydrogenase (E1) component, an ␣ 2 ␤ 2 tetramer present at 20 -30 copies (1); the dihydrolipoyl dehydrogenase (E3), a homodimer present in about 6 copies (9); and the E3-binding protein (E3BP) estimated at 6 -12 copies (10,11). Dedicated and highly regulated kinase and phosphatase components control the conversion of E1 between an active (nonphosphorylated) form, E1a, and an inactive (phosphorylated) form, E1b.
Mammalian PDC-E2 subunits have four flexibly connected domains and form the core of the complex in which 60 of its COOH-terminal inner (I) domains assemble into a dodecahedron-shaped structure with its other three domains extending out around this porous surface. The icosahedral I 60 inner core anchors the E3BP (12)(13)(14) and carries out the transacetylase reaction (15), probably via functional trimer units as established for the octahedral inner core of Azotobacter vinelandii PDC-E2 (16,17). Trimer units were also shown to be important in the assembly of bovine heart E2 subunits dissociated to nonfunctional monomers in 4 M guanidinium chloride (18).
The outer globular domains of E2 consist of two lipoate-bearing domains (L1 and L2), followed by a small E1 binding (B) domain. The connecting hinge regions are 20 -30 residues in length, are enriched in Ala and Pro residues (19,20), and are highly mobile but stiffer than random coil structures (5,7). Three-dimensional structures have been derived for examples of each of E2's domain classes from bacterial PDC-E2s (16,(21)(22)(23). However, the domains in mammalian E2 have significant differences both in structure and function (3)(4)(5)(6). For instance, the mammalian PDC-E2 lipoyl domains appear to be larger and at least L2 has several specialized roles (see below). The small B domains of eukaryotic PDC-E2s only bind E1, whereas the structurally related domains in other PDC E2s bind only E3 (e.g. Escherichia coli PDC; Refs. 5 and 24) or bind both E3 and E1 (e.g. Bacillus stearothermophilus PDC; Refs. 5, 25, and 26). The I domains of eukaryotic PDC-E2s are apparently unique in binding the E3BP (3-7). 2 E3BP is also composed of three linker-connected domains (an inner domain that binds to the I domain of E2, an E3-binding domain, and a lipoyl domain) (3,4,13,27).
Prokaryotic lipoyl domains (22,23) are flattened ␤ barrel structures with the lipoate attached to a specific lysine residue located in tight turn protruding from one end. Building on this framework, larger mammalian PDC-E2 lipoyl domains have added structure at their COOH-terminal ends that may contribute to performing their unique roles. High mobility of the connecting hinge regions allows the lipoyl domains to be delivered to the E1, E2, and E3 components where the prosthetic group extends into active site channels (8). Specific interaction of the lipoyl domain with E1 is essential for efficient E1 catalysis and probably aids the E2 and E3 reactions (5,28,29). 3,4 The L2 domain binds the E1a kinase (30) and the E1b phosphatase (31). Binding of a kinase dimer probably engages two L2 domains plus their lipoyl prosthetic groups (30), while the binding of the phosphatase to L2 requires Ca 2ϩ (31). Furthermore, the L2 domain has a direct role in mediating regulatory stimulations of the kinase by the PDC products, NADH and acetyl-CoA (32), and the marked enhancement of phosphatase activity by the Ca 2ϩ -facilitated binding to L2 constitutes an important means of second messenger activation (33,34).
Although recombinant lipoyl domain structures of human E2 have proved useful for determining binding sites and characterizing component reactions, some catalytic and regulatory processes require a higher level of structural complexity. For instance, markedly activated functioning of the kinase and the phosphatase, when bound to the complex, is not supported by these isolated domains and apparently requires the full oligomeric E2 structure, which can bind multiple copies of the E1 substrate as well as a regulatory enzyme and can facilitate a sequence of steps needed to produce enhanced activity. The capacity of the mammalian E2 core to support the overall PDC reaction has been previously evaluated only when a major portion of E3BP has been removed from the E2 core under highly chaotropic conditions (35,36), making uncertain the status of the residual E2. Recombinant expression of assembled E2 would overcome these problems and allow development of altered structures in subsequent work.
Successful expression and recovery of assembled recombinant E2 could not be expected on the basis of previous studies, in which bacterial (E. coli and A. vinelandii) or Saccharomyces cerevisiae PDC-E2s were recovered only upon co-expression in E. coli of the cognate E1 or E1 ϩ E3 (13,37,38). The successful expression of assembled but lipoyl-deficient E2 core of the bovine branched-chain ␣-keto acid dehydrogenase complex was encouraging (39), as well as evidence for the independent folding and assembly of unfolded bovine E2 subunits (18). Here, we have developed conditions for expressing fully assembled mature human E2 in E. coli and for obtaining highly purified preparations that have the expected structure and known catalytic functions. We have evaluated how well this E3BPdeficient E2 supports the PDC reaction, E3 catalysis, and the enhanced catalytic reactions of the kinase and the phosphatase.

EXPERIMENTAL PROCEDURES
Materials-Bovine kidney PDC (40), E1 component, E2⅐E3BP-kinase subcomplex (41), and the recombinantly expressed bilipoyl domain region of human E2 (42) were prepared as described previously. Porcine heart E3 was from Boehringer Mannheim. The original pBTA vector that coded for all of mature human E2 as well as a portion of its leader sequence was kindly provided by M. E. Gershwin (University of California, Davis). The expression vector pSE420 and its recommended E. coli host strain, Top10, were from Invitrogen; and the host E. coli strain BL21(DE3), lacking the lon protease and ompT outer membrane protease, was from Novagen. PEG-8000 was from J.T. Baker Co., and Pluronic F-68 was from BASF Corp. Other chemicals and reagents were as described by Liu et al. (42).
Vector Construction and Expression-The plasmid pBTA harboring the human E2 cDNA insert was cut with restriction enzymes SfaNI and BamHI to produce a DNA fragment lacking the coding region for the leader sequence and first 7 amino acids of mature E2. Two phosphorylated synthetic oligonucleotides (sequences in Fig. 1) were hybridized to produce a small double-stranded DNA fragment that coded for this remaining portion of mature E2 preceded by the ATG start codon and containing compatible splicing sequences for a NcoI site at the 5Ј-end of the coding sequence and a SfaNI at its 3Ј-terminus. This synthetic structure, the pBTA restriction fragment, and the pSE420 vector digested with NcoI and BamHI were ligated to produce the expression vector coding for mature E2 preceded by a start codon (Fig. 1). The ligated plasmid was transformed into E. coli strains by electroporation using Transfector 300 BTX under the conditions recommended in the instrument manual. The expression of E2 was first confirmed by dot blots and then Western blots using a mixture of monoclonal antibodies 150.2 and 157.2 (42), which react with the first and second lipoyl domain, respectively. Clones were designated pShE2 (pS for pSE420 vector, h for human, E2 for coding for full-sized mature E2). The DNA sequencing was conducted using a series of synthetic primers and the dideoxynucleotide sequencing procedure using the Sequenase Version 2.0 kit from U. S. Biochemical Corp.
Experiments to optimize expression used Western blot analyses employing not only the lipoyl domain-specific monoclonal antibodies (above) but also a polyclonal antibody that reacts with the inner domain of E2. With different host strains, induction at different stages of growth, and expression for variable periods, Western blots were used to detect production and cleavage of E2 when intact bacterial cells were extracted with hot SDS-sample buffer, soluble and particulate fractions were prepared from different host strains, and clarified lysates were incubated for various times.
Purification of the Recombinant E2 Oligomer-All steps were performed at 4°C; PEG and (NH 4 ) 2 SO 4 fractionations involved continuous stirring with centrifugation conducted 20 min after additions of these precipitants were completed. E2 activity was measured in a rapid continuous assay (below) with corrections in the case of crude extracts for E. coli PDC-E2 activity made assuming the same ratio of E2 to PDC activity as measured in the nontransformed host strain. (The expression of human E2 had no effect on the E. coli PDC activity.) Cells (12-14 g) harvested from 1.5 liters were thawed and resuspended at 12.5% w/v in 50 mM potassium phosphate (pH 7.2), containing 0.5 mM EDTA, 1 g/ml aprotinin and leupeptin (buffer A) and immediately ruptured using Vibra-cell high intensity ultrasonic processor (Sonics and Materials Inc.) equipped with a 0.5-inch probe at a setting of 4 using 10 1-min bursts. Particulate material was removed by centrifugation at 15,000 ϫ g for 20 min. PEG-8000 (50% w/v) was added dropwise to the clarified lysate to 8% (v/v), and precipitated protein, removed by centrifugation, was discarded. An additional 8% (v/v) was added followed by centrifugation. This 8 -16% PEG pellet was dissolved in buffer A. The protein concentration was diluted to 2-3 mg/ml, and solid (NH 4 ) 2 SO 4 was added to 25% saturation. The E2-containing precipitate was removed by centrifugation and dissolved in 4 ml of buffer A containing 0.2 mg/ml Pluronic F-68. After 12 h, the fraction was clarified by centrifugation and applied to 2.7 ϫ 75-cm bed volume Sephacryl S-400 column equilibrated with 40 mM sodium phosphate, 0.5 mM EDTA, 1 g/ml aprotinin, and leupeptin, and 0.2 mg/ml Pluronic F-68 (buffer B). Fractions (4 ml) were assayed, and those containing high E2 activity were clarified by centrifugation at 15,000 ϫ g for 20 min and then simultaneously analyzed by SDS-PAGE (43) with silver staining (44) and concentrated 10 -12 fold by centrifugation in Centricon 10 units. Concentrated fractions were then stored in aliquots at Ϫ80°C.
The purified E2 preparations were analyzed to identify cleavage products, establish the NH 2 -terminal sequences of all E2 bands, and determine the proportion of intact E2 subunits. Putative intact E2 and cleaved E2-polypeptides, initially identified by immunoblotting were blotted onto ProBlott, stained with 0.1% Coomassie Blue R250 in 1% acetic acid, 40% methanol for 1 min, and destained in 50% methanol, and then NH 2 -terminal sequencing was conducted as described previously (42). To estimate the proportion of E2-protein and relative levels of intact versus cleaved E2, quantitative area scanning of silver-stained patterns was performed following SDS-PAGE separation of 0.4, 0.8, 1.6, and 2.4 g of purified human E2 and 0.6, 1.2, 2.0, and 3.2 g of bovine E2⅐E3BP loaded into lanes. Alternatively, 0.48, 0.96, 1.6, 2.4 and 4.0 g of human E2 and 0.8, 1.2, 2.4, 5.0 g of bovine E2⅐E3BP were analyzed in Coomassie Brilliant Blue R-250-stained patterns. Band densities were evaluated by scanning with an Epson ES-800c (Model 2) scanner, acquiring area density data with Adobe Photoshop program, and analyzing the data using NIH Image software. Each series gave a linear increase in area with the amount loaded (data not shown).
E2 Activity Assays-Routine E2 activity assays monitored acetyldihydrolipoamide production by recording the ␦A 232 nm with time (45). Reaction mixtures contained 30 mM Tris⅐HCl (pH 7.4), 1 mM dihydrolipoamide, 1 mM acetyl-phosphate, 5.0 M CoA (prepared as concentrate 1:1 with cysteine), and 1 unit of phosphotransacetylase in a 1-ml cuvette at 30°C. After the absorbance became constant due to formation of acetyl-CoA (ϳ10 s), the E2 source was added and increased absorbance at 232 nm measured. The activities are recorded simply as absorbance/min because the precise extinction coefficient of the immediate 8-acetyl-dihydrolipoamide product (about 4 A/mM⅐cm) cannot be determined (shuffling of acetyl groups in acetylated dihydrolipoamide leads to formation of a 3 to 1 mixture of the 8-and 6-acetyl-dihydrolipoamide and significant amounts of the biacetyl derivative at high levels of acetylation; see Refs. 46 and 47).
[1-14 C]Acetyl-dihydrolipoamide formation was assayed by procedure of Butterworth et al. (48). Reactions were initiated at 25°C by addition of [1-14 C]acetyl-CoA (8 ϫ 10 5 cpm/mol) to yield 0.5 mM concentration in a reaction mixture containing 25 mM potassium phosphate (pH 7.5), 0.5 mM dihydrolipoamide, 0.05 mM EDTA, 4 g of E2 in a 0.5-ml reaction mixture. After 120 s, 1 ml of benzene was added and vortexed for 20 s to stop the reaction and to extract the radiolabeled acetyldihydrolipoamide formed. Two 0.2-ml aliquots of the benzene layer were withdrawn and counted; duplicate assays were within Ϯ6%.
Acetylation Capacity-To acetylate all lipoyl groups with [1-14 C]acetyl-CoA, reductive acetylation by the reverse E2 and E3 reactions was coupled to conversion of CoA formed to succinyl-CoA as described previously (31,49,50). Reactions were carried out at 30°C in 20 mM potassium phosphate (pH 7.5) containing 0.2 mM EDTA, 0.25 mM NADH and NAD ϩ , 0.4 mM ␣-ketoglutarate, 50 M Ca 2ϩ , 2 g of ␣-ketoglutarate dehydrogenase complex, and the indicated level of E2 source. Reactions were initiated by addition of acetyl-CoA (8 cpm/pmol), and after 60 s reactions were terminated and protein-bound acetyl groups determined.
PDC and E3 Assays-The capacity of an E2 source (4 -6 g) to support the overall PDC reaction was evaluated by incubating E1 (2 g/g of E2) and variable levels of E3 for 60 s at 30°in 15-21 l and then adding two-thirds of the volume to 1 ml of standard assay mixture (51) and measuring the rate of NADH production at 340 nm. E3 activity was measured by the cyclic E3 reduction of lipoates and their rapid chemical reoxidation (42). 3 The E3 content of a E2⅐E3BP subcomplex preparation was reduced to 0.6 E3 dimers/subcomplex by a second passage through the standard resolution procedure (41) that removes E1 and most E3, followed by pelleting of the subcomplex through a sucrose step gradient containing 2.5 M urea and 0.35 M NaCl. The lipoyl domain sources (E2⅐E3BP, recombinant E2, or bilipoyl domain construct) were added at 80 -320 pmol to a 200-l reaction mixture containing 50 mM sodium phosphate (pH 7.4), 0.3 mM NADH, 0.1 mM NAD ϩ , 0.2 mM 5,5Ј-dithiobis-(2-nitrobenzoate), and 1 mM EDTA. Reactions were initiated by addition of 0.1 g of E3. 5-Thio-2-nitrobenzoate production was monitored with a UVmax microplate reader at 405 nm.
Electron Microscopy-Purified recombinant human E2 with or without bovine E1 was dialyzed against Buffer B to remove Pluronic F-68. Samples were prepared and negatively stained with 0.5% uranyl acetate (52), and micrographs acquired with a Philips EM420T transmission electron microscope at a nominal magnification ϫ50,000 using 1000 e/nm 2 (53).
Sucrose Gradient Analyses-To evaluate removal of residual non-E2 bands and test E1 binding to E2, fractionation of E2 and E2 ϩ E1 was evaluated using a modification of the step gradient procedure of Powers-Greenwood (35). 25 g samples of purified E2 with or without 20 g of purified bovine E1 were incubated for 60 min at 4°C in 70 l of 50 mM sodium phosphate (pH 7.2) containing 0.2 mM EDTA, 0.5 mM dithiothreitol, 1 g/ml leupeptin and aprotinin. These samples were layered over a three-step gradient (50 l each of 7.5, 10, and 15% sucrose in the same buffer) in 5 ϫ 20-mm ultraclear tubes (Beckman) and centrifuged at 130,000 ϫ g for 150 min at 4°C. Fractions were withdrawn as upper 120 l (S1), middle 50 l (S2), and bottom 50 l (S3); and the pellets were dissolved in 50 l of the same buffer. SDS-PAGE analysis (43) with silver staining (44) was conducted as described in Fig. 4.
[ 32 P]Phosphate release from PDCb or E1b was measured at 30°C in 30-l reaction mixtures containing 20 mM Mops-K buffer (pH 7.3), 0.4 mM dithiothreitol, 0.6 mM EDTA, 1.0 mM EGTA, 1.2 mM CaCl 2 , and the indicated level of E2 oligomer (56,57). E1b phosphatase was added for 120 s and then activity initiated by addition of MgCl 2 to 10 mM. After 120 s, reactions were terminated by addition of 50 l of 20% trichloroacetic acid and precipitated protein was pelleted after at least 30 min of incubation at 4°C by centrifugation for 4 min in a microcentrifuge at 8000 ϫ g. Aliquots (28 l) were withdrawn and radioactivity determined by counting in scintillation fluid. All kinase and phosphatase assays were conducted in duplicate or triplicate; absolute deviations are shown.

Construction and Sequencing of E2-coding Expression Vector and Development of Optimal Expression Conditions-
The initial expression vector that was provided coded for all of mature E2, 16 amino acids of the presequence (location of a convenient EcoRI restriction site), and a small terminal portion of ␤-galactosidase (58). The expression product accumulated only in inclusion bodies and lacked E2 activity before and after a variety of solubilization procedures but was fully antigenic in Western blots using polyclonal and monoclonal antibodies. 5 The construction of the expression vector, as outlined in Fig. 1, removed the leader sequence and adjoined a start codon at the front of the coding region for mature E2. The pSE420 vector was chosen for its bacteriophage T 7 gene 10 element that enhances translation of eukaryotic genes. Cell lysates from pShE2 transformed colonies gave strong responses in dot blots using lipoyl domain-specific monoclonal antibodies and contained soluble E2 activity well beyond that of E. coli PDC (data not shown).
The pShE2 plasmid was sequenced not only at the ligated 5Ј-end but throughout most of the E2-encoding insert, including all regions in which the sequences of Coppel et al. (19) and of Thekkumkara et al. (20) differed. Sequencing confirmed that the desired construct was obtained and established a sequence for this placental-cDNA derived construct was identical to that reported for the human liver E2 (20), establishing that previously reported differences were due to sequencing errors.
Immunoblot analysis revealed substantial full-sized E2 (as expected with a mobility in SDS-PAGE somewhat faster than bovine E2; Ref. 59) along with degradation products, which were not reduced by inclusion of a variety of protease inhibitors in cell lysates. Furthermore, there was no time-dependent change in the pattern of intact versus partially cleaved E2 when cell lysates were incubated at 25°C in buffer A and the same pattern of E2 bands was observed when cells were disrupted in 2% SDS-sample buffer by immediate insertion in boiling water (data not shown). Co-expression of groELS chaperonin proteins did not reduce E2 cleavage or increase the portion of E2 fractionating into the supernatant fraction. Thus, if chaperonins are required to aid the folding or assembly of E2, the endogenous levels in E. coli are sufficient. Using pShE2 transformed into the protease-deficient BL21 strain, a greater proportion of full-sized E2 was found in the soluble fraction from cell lysates than with the Top10. Using the BL21 host, changing the IPTG induction from stationary to mid-log phase, and lowering the temperature of expression to 27°C gave a higher level of soluble E2, significantly increased the proportion of intact E2, and reduced the number of E2-derived bands arising from the limited proteolysis. With a correction for the low level of E. coli PDC E2, the transacetylase activity indicated that recombinant E2 constituted about 3% of the soluble protein after 8 h of post-induction growth.
Purification of Recombinant E2-Using the steps detailed under "Experimental Procedures," the recombinant E2 was purified to near homogeneity. The initial PEG fractionation step left Ͼ99% of the contaminating E. coli PDC in solution, but a significant portion of the bacterial ␣-ketoglutarate dehydrogenase complex co-precipitated with the E2 in the 8 -16% cut. Human E2 (intact E2 60 has calculated M r of 3.57 ϫ 10 6 ) eluted prior to the similarly sized ␣-ketoglutarate dehydrogenase complex in gel-filtration chromatography, demonstrating that E2 subunits assembled into a large aggregate. 6 The most prominent protein (M r Ӎ 42,000 in SDS-PAGE) not separated from E2 by the gel filtration, as well as the E. coli ␣-ketoglutarate dehydrogenase complex, were removed by fractionation with ammonium sulfate prior to the size exclusion step; both contaminants stayed in solution when E2 was precipitated.
Gel-filtration chromatography then removed low molecular weight contaminants, followed by a clarifying centrifugation step, which reduced or removed additional contaminants. A variety of other procedures either failed to allow E2 to be recovered (ion exchange or hydroxylapatite chromatography) or failed to significantly improve the purification (additional PEG fractionations, sucrose gradient centrifugation). Fig. 2 shows the change in SDS-PAGE profile, and Table I presents information on the recovery and change in specific activity at each stage in the purification. Intact E2 is the major band in the SDS-PAGE profile of the purified product. Two other bands (labeled in the SDS-PAGE pattern of Fig. 3 as E2Ј and E2 I ) were identified by Western blotting as cleavage prod- 6 The E2 eluted in a broad peak from Sephacryl S 400 prior to a sharp peak for the ␣-ketoglutarate dehydrogenase complex. Since the latter complex has a similar mass (M r Ӎ 2.8 ϫ 10 6 versus 3.6 ϫ 10 6 for unclipped E2 oligomer) and both structures have high frictional coefficients, this suggests that the E2 oligomer has a tendency to form aggregates. The inclusion of Pluronic F-68 (2 mg/ml) reduced but did not eliminate peak broadening due to decreased aggregation and minimized peak trailing due to interaction with the gel matrix. ucts of E2 retaining the oligomer forming inner domain. The other minor contaminants all changed in their ratio to mature E2 during the purification process. NH 2 -terminal sequencing gave SLPPHQKV, AKILVAEG, and VVPPTGPG for bands identified as E2, E2Ј, and E2 I , respectively, establishing that the upper band is mature E2 with start codon Met removed and that E2Ј and E2 I resulted from cleavage following Met-59 (located in L1 domain) and Leu-318 (located in the third linker region just prior to the inner domain of E2), respectively.
Densitometric area scanning of silver-stained bands of SDS-PAGE patterns of E2 loaded over a 6-fold range (see "Experimental Procedures") gave estimates that intact E2 constituted about 70 Ϯ 5% of the protein and that the two E2 cleavage products contributed about 15 Ϯ 5% of the protein. These estimates were made assuming that staining is proportional to protein. On a molar basis, this suggests that about 75% of the E2 protein consists of full-sized E2 subunits. Somewhat higher relative staining intensity of the ϳ38-kDa contaminant was observed when Coomassie Brilliant Blue-stained patterns were similarly analyzed, but that analysis gave roughly the same relative molar proportion of intact versus cleaved E2. Corrected specific activities and acetylation levels (below) would be about 9% higher if the proportion of E2 protein was estimated based on Coomassie staining rather than silver staining.
Binding of Bovine E1 and Morphology of Human E2- Fig. 3 shows the SDS-PAGE patterns for three supernatant fractions and one pellet fraction derived from microsucrose-gradient fractionation of E1, human E2, and the combination. Alone, E1 is concentrated at the top and E2 toward the bottom of the gradient, consistent with E1 existing as a tetramer and E2 as a large oligomer. In combination, virtually all the E1 fractionated with the recombinant E2, establishing the B domain is fully functional in binding E1. The most prominent contaminant in the E2 preparation (band position between E1 subunits) must be in a large aggregate that does not interact with E2, since it sedimented more rapidly whether or not E1 was present.
The upper panel of Fig. 4 shows a portion of a field of negatively stained human E2. As previously presented and modeled (53,60), only the inner core is observable and presents images characteristic of an icosahedrally symmetric pentagonal dodecahedron. Typical views at or almost down the 5-fold axis are common (left center) but examples of images viewed approximately along the 2-fold (lower image; upper right) and 3-fold axis (top upper right) are apparent. The lower panel shows images with a high (left) and a lower (right) level of E1 bound around the periphery of the inner core, further supporting effective E1 binding.
Lipoyl Content and Catalytic Activity-To evaluate the lipoate content, we compared the extents of acetylation of our preparations of human and bovine E2. Using [1-14 C]acetyl-CoA and CoA removal to drive the acetylation reaction, 75% as high an extent of acetylation of the recombinant E2 as the E2⅐E3BP subcomplex was observed (Table II, third column). This falls within experimental error of the value of 80 Ϯ 7% as high of acetylation, calculated using the proportion of bands estimated from gel scanning (silver stain) and accounting for single site acetylation of E3BP in bovine E2⅐E3BP and E2Ј (truncated in the first lipoyl domain) in the recombinant E2 oligomer. Acetylation of E2Ј was confirmed by autoradiography (54) after SDS-PAGE separation of subunits (data not shown). This agreement indicates that all or nearly all lipoyl domains are lipoylated.
The specific activity (␦A 232 nm ⅐min Ϫ1 ⅐mg Ϫ1 ) of the purified E2 oligomer was 58% higher than that of a recent bovine E2⅐E3BP  subcomplex (Table II). The other transacetylase assay measuring [ 14 C]acetyl-dihydrolipoamide, which uses a 20-fold higher acetyl-CoA concentration, also gave a higher activity for the recombinant E2 than for E2⅐E3BP subcomplex (Table II, second column). We estimate the E2 specific activity of the E2 oligomer (following corrections due to non-E2 protein (decrease) and due to the presence of active, truncated forms of cleaved E2 (increase)) to be about 5.0 mol⅐min Ϫ1 ⅐mg Ϫ1 . Participation in the Overall PDC and E3 Reactions-The capacity to support the overall PDC reaction was compared for E2 oligomer relative to E2⅐E3BP subcomplex when both were reconstituted with excess E1 and the E3 components (Table II, last column). The E2 oligomer supported a low but significant rate of the overall reaction in the absence of E3BP. Even when a 4-fold higher level of E3 was added to the E2 oligomer, there was no increase in reconstituted PDC activity (data not shown). We have not been able to prepare an active form of E3BP in the absence of E2. Retention of about 4% activity is consistent with previous studies with bovine E2 (35) from which E3BP was removed, whereas with S. cerevisiae E2 no activity in the overall reaction was observed in the absence of E3BP (61). Although not as extreme as in the case of the yeast PDC, the retention of such a low activity indicates that the overall PDC reaction is facilitated by molecular mechanisms that efficiently channel reducing equivalents to E3BP-bound E3 (cf. "Discussion").
Liu et al. (42) found that the individual lipoyl domains of E2, prepared by recombinant techniques, were effectively utilized by the E3 component in a cyclic form of the reverse reaction involving NADH-dependent reduction by E3, followed by rapid chemical reoxidation to the disulfide by reaction with DTNB. Table III shows E3 reaction rates using 0.4 -1.6 M intact E2 subunits of the E2⅐E3BP subcomplex or recombinant oligomer and of the bilipoyl domain portion of E2 (L1⅐H1⅐L2). With the recombinant sources (lines 1 and 2, Table III), concentrationdependent increases in rates fit E3 using these lipoyl domain sources as a substrate with a K m above this concentration range. Indeed, increasing the level of a free lipoyl domain yields a maximal specific activity that is much higher. 3 The somewhat higher rates with the free bilipoyl domain structure than with the E2 oligomer may reflect more efficient delivery of lipoyl domains to the E3 active site with the freely diffusing structure than when lipoyl domains are grouped in close proximity within an E2 structure that E3 cannot bind.
Results from varying the E2⅐E3BP subcomplex were very different (Table III). The subcomplex contained a low level of endogenous E3 even after additional resolution steps to reduce the E3 level. Both this endogenous E3 (line 3) and added E3 (line 4) gave a fixed and nearly identical specific activity as the subcomplex level was increased. This implies that E3 bound to E3BP is not responding to variation in the E2 concentration but only to the fixed stoichiometry of lipoyl domains surrounding it within the E2⅐E3BP subcomplex. Addition of E3 to the E2⅐E3BP to levels exceeding 30% of its E3 binding capacity lowered E3 specific activities (data not shown). This outcome suggests an increased competition for localized lipoates as more E3 is bound in the subcomplex. The results are consistent with E3, localized by binding to E3BP, encountering a fixed availability of lipoyl domains to introduce reducing equivalents.
Capacity of the Recombinant E2 Oligomer to Support Kinase and Phosphatase Function-Bovine kidney E1a kinase binds to the inner (L2) domain of E2 (30); binding increases the rate of phosphorylation of E1a severalfold by lowering the K m for E1a 20-fold and increasing the V m of the kinase 2-fold (51). In the resolution of bovine kidney PDC that produces the E2⅐E3BP subcomplex and E1, a major portion of the kinase fractionates with the subcomplex and a small portion with E1. The activation state of the dilute kinase that fractionates with E1 is enhanced with the level of E2. With 2.9 M E1, 6 g of the recombinant E2 oligomer gave somewhat more than a 2-fold increase and 19 g provided Ͼ4-fold increase in this kinase activity (Fig. 5). In contrast, no increase in kinase specific activity was found when E2⅐E3BP-kinase subcomplex was varied to provide a similar range of E2 levels. 7 The activation of b This recombinant bilipoyl domain construct contains amino acids 1-233 of sequence of human E2 and was prepared as described previously (31).
c The level of bound E3 was estimated to be 0.015 g/g subcomplex by matching staining intensities of SDS-PAGE analysis on several concentrations of subcomplex and several levels of E3. E3BP also contributes a lipoyl domain, and its concentration is 10 -20% of the concentration of E2 subunits (so E3BP contributes 5-10% of lipoyl domains in subcomplex).
d Deviations are the sum of the deviation range in assays with and without added E3 to E2⅐E3BP. the kinase associated with E1 but not that associated with E2⅐E3BP as the concentration of recombinant or subcomplex E2 was varied in the same range may reflect enrichment of different kinase isozymes (63)(64)(65) with E1 or with E2⅐E3BP during the resolution of bovine PDC. The Ca 2ϩ -dependent E2 enhancement of the Mg 2ϩ -requiring E1b phosphatase activity is pronounced, with activation in the presence of saturating Mg 2ϩ primarily resulting from a decrease in the K m of the phosphatase for E1b (66). Fig. 6 shows that, in 25-l reaction mixtures, the recombinant E2 oligomer markedly increased phosphatase activity with as little as 2 g of E2 giving a 10-fold increase in activity and 8 g of E2 facilitating up to a 16-fold increase in phosphatase. Relatively low E1b (1.52 M) was included, which maximizes E2 activation in these assays. From a reciprocal plot of activity to E2 subunit concentration, half-maximal activation of the phosphatase is estimated to occur at 0.74 Ϯ 0.1 E2 subunit concentration, consistent with tight binding of the phosphatase to E2.
When 8 g of phosphorylated PDCb and 4 g of E1b (3.8 g of bovine E2, 1.88 M E1b in total) were present, phosphatase activity was increased by further addition of 8 g of recombinant E2 by 2-fold over the enhancement by bovine E2 (right series, Fig. 6). Comparing the absolute activities for a change from 3.8 g of bovine E2 to a total of 7.8 g of E2 (3.8 g of bovine plus 4 g of human E2) to that for a change from 4 g to 8 g of human E2, the initial activity is 2-fold higher with 4 g of human E2 than with the bovine E2 and the final activity is still 41% higher with 8 g of human E2 than with the combination of human and bovine. These data suggest human E2 is more effective than bovine E2 in activating the phosphatase. It remains to be determined whether that is due to interference of other components (E3BP or E3), damage to bovine E2 during preparation of the complex, or an innately greater capacity of the human E2 to activate the bovine phosphatase in dephosphorylating bovine E1b. Regardless, these data further support high functionality of the recombinant human E2. DISCUSSION E2 components have many roles in the organization, operation, and regulation of ␣-keto acid dehydrogenase complexes. Because E2 subunits invariably have modular structures with independently folded domains set off from each other by mobile linker regions, the domain-specific roles can be fruitfully diagnosed by recombinant DNA approaches as exemplified by using individual domains or groups of domains in native or mutated forms in studies of E. coli (e.g. Refs. 37 and 67-69), B. stearothermophilus (e.g. Refs. 21, 22, and 24 -26), A vinelandii (16,17,70), and S. cerevisiae PDC-E2s (e.g. Refs. 13 and 61). However, as noted in the Introduction, some roles require assembled E2 structures. Guest and colleagues have characterized a variety of assembled E. coli PDC-E2 structures for lipoyl domain roles and found the three lipoyl domains have equivalent capacities in supporting the overall PDC reaction (67)(68)(69). As in the case of S. cerevisiae PDC (27), production of mammalian E2 free of E3BP is important for distinguishing the contribution of these two lipoyl-bearing components in catalytic and regulatory processes. Mammalian PDC is regulated by phosphorylation and dephosphorylation, but that is not the case in the well characterized bacterial PDCs, and yeast PDC has only recently been shown to produce a kinase and phosphatase under specialized growth conditions (71).
We have successfully expressed and prepared in a highly purified state the assembled E2 core of human PDC without coexpression of other components of the complex. Beyond deleting the upstream coding region (including a partial leader sequence), key conditions were expression of E2 in a proteasedeficient host (E. coli BL21), lipoate supplementation, and induction of expression in mid-log growth phase at a reduced temperature of 27°C. The latter two conditions minimized but did not completely eliminate proteolytic cleavage and allowed recovery of an assembled structure with about 75% of the subunits intact. Sequencing of NH 2 termini identified a truncated structure, labeled E2Ј, whose sequence began after Met-59 located in the first domain; E2Ј may have been produced by the Met-59 codon serving as an internal start site. There is a good Shine-Dalgarno sequence (GGAGGAG) located 6 nucleotides upstream of the Met codon. This Met residue aligns with Leu, Val, and Ile in other PDC-E2 lipoyl domains and corresponds to a buried amino acid in the established three-dimensional structures of bacterial domains (22,23), suggesting that production of E2Ј by proteolysis might require cleavage to occur prior to folding of the lipoyl domain. A second site of cleavage is located near the end of the third linker region. By whatever means, the E2Ј and E2 I structures were created prior to rupture of E. coli cells since they were observed in immunoblots when cells were extracted with immediate heating in SDS-sample buffer. Residual contaminants in E2 preparations are large structures that do fractionate somewhat differently in sucrose gradients (e.g. Fig. 3), and there is no indication that these E. coli proteins interact with human E2.
Although not fully intact, the recombinant E2 exhibited high activity in the acetyltransferase reaction, bound the E1 component, and supported the activated function of the kinase and phosphatase. The observable inner core has the expected dodecahedron structure. Supporting the independent folding of lipoyl domains, internal disruption of the terminal lipoyl domain in E2Ј did not prevent folding and lipoylation of the inner lipoyl domain since it could be effectively acetylated by the E1 (data not shown) or E2 reactions.
Cryoelectron electron microscopy and quasielastic light scattering measurements indicate that hydrated mammalian PDC has a diameter of ϳ500 Å, that the E1 and E3 components are tethered with gaps between them and the inner core, that there is considerable solvent entrapped at the surface of the complex (53,72). This solvent-rich environment can support domain and component movement, a capacity eminently important for the operation of active site coupling in catalytic processes and for regulatory interconversion of the E1 component. The high kinase and phosphatase activities supported by the recombinant E2, lacking E3BP, establish that it has the full capacity to facilitate the molecular processes whereby individual kinase and phosphatase dimers rapidly encounter many molecules of their E2-bound E1a and E1b substrates. In the case of the tightly bound kinase, this is proposed to involve a kinase dimer moving across the surface of the complex by a "hand over hand" mechanism that allows it to find and phosphorylate bound E1 without losing its grip on the complex (30). Rapid dissociation and reassociation of the less tightly bound phosphatase in a Ca 2ϩ -regulated process are thought to account for the dephosphorylation of E1b (31,51,66). Our results suggest that the recombinant human E2 is significantly more effective than bovine E2⅐E3BP in enhancing phosphatase activity. Although many explanations are possible (cf. "Results"), testing the interesting prospect that E3BP has an inhibitory role will require preparation of human E2⅐E3BP.
In agreement with an earlier preparation (35) in which bovine E2 was freed of E3BP (then designated as protein X), the recombinant E2 oligomer was able to support about 4% of the PDC activity obtained with the E2⅐E3BP subcomplex. These results differ from those obtained with yeast PDC-E2, prepared in the absence of the E3BP, which failed to support any PDC activity even in the presence of high levels of E3 (61). Nevertheless, both outcomes demonstrate that the binding of E3 by the E3BP component must greatly enhance the capacity to feed reducing equivalents through E3 in the PDC reaction. This suggests unique organizational features allow limited E3 dimers, tethered at the surface of the complex, to function efficiently. The results in Table III indicate E3BP-bound E3 has access to a fixed number of lipoyl domains based on the unchanging specific activity of low levels of E3 (endogenous or added) with variation of the level of E2⅐E3BP subcomplex. The lipoyl domains of E2 and E3BP are concentrated at about 2 mM at the surface of the E2 core (53,72) and are probably anchored in ways that limit their orientations. These conditions may not only facilitate reducing equivalents being directly channeled to E3 but also aid a rapid thiol-disulfide interchange between lipoates. The latter may be coupled to electrons being fed to E3 with some degree of preference via the lipoate on E3BP's lipoyl domain (8,73). The E3 component must operate at Ͼ280 mol/ min/mg of E3 to give the observed specific activity of 20 mol/ min/mg of bovine heart PDC in the overall reaction of the complex (bound E3 is maximally 7% of the protein; Ref. 9). Given this high rate, it seemed likely that reaction of 0.2 mM DTNB with reduced lipoates is probably limiting when E3 was confined by binding E3BP. In support of this possibility, halving the DTNB concentration to 0.1 mM reduced the E3 reaction rates of E3 associated with E2⅐E3BP structures by ϳ40% (data not shown). Thus, a rapid E3 reaction followed by a ratelimiting chemical reoxidation of lipoates is indicated.
The production of assembled recombinant E2 that functions in all the roles of mammalian PDC-E2 creates abundant opportunities for further studies in which recombinant techniques are used to dissect catalytic, binding, and regulatory roles of this central component of mammalian PDC. Furthermore, new insights and questions are raised by the differences that we have observed between bovine E2⅐E3BP and human E2 in supporting catalytic processes and in their relative capacities to facilitate high E1b phosphatase activity.