Novel Binding Motif and New Flexibility Revealed by Structural Analyses of a Pyruvate Dehydrogenase-Dihydrolipoyl Acetyltransferase Subcomplex from the Escherichia coli Pyruvate Dehydrogenase Multienzyme Complex*

Background: The E. coli pyruvate dehydrogenase complex containing E1p, E2p, and E3 components converts pyruvate to acetyl-CoA. Results: The E1p-E2p subcomplex revealed a novel binding motif relative to the only other known example. Conclusion: Tethering of E1p to E2p depends on the α2 or α2β2 assembly of E1p. Significance: The new tethering mode should affect overall complex assembly in all such complexes from Gram-negative bacteria. The Escherichia coli pyruvate dehydrogenase multienzyme complex contains multiple copies of three enzymatic components, E1p, E2p, and E3, that sequentially carry out distinct steps in the overall reaction converting pyruvate to acetyl-CoA. Efficient functioning requires the enzymatic components to assemble into a large complex, the integrity of which is maintained by tethering of the displaced, peripheral E1p and E3 components to the E2p core through non-covalent binding. We here report the crystal structure of a subcomplex between E1p and an E2p didomain containing a hybrid lipoyl domain along with the peripheral subunit-binding domain responsible for tethering to the core. In the structure, a region at the N terminus of each subunit in the E1p homodimer previously unseen due to crystallographic disorder was observed, revealing a new folding motif involved in E1p-E2p didomain interactions, and an additional, unexpected, flexibility was discovered in the E1p-E2p didomain subcomplex, both of which probably have consequences in the overall multienzyme complex assembly. This represents the first structure of an E1p-E2p didomain subcomplex involving a homodimeric E1p, and the results may be applicable to a large range of complexes with homodimeric E1 components. Results of HD exchange mass spectrometric experiments using the intact, wild type 3-lipoyl E2p and E1p are consistent with the crystallographic data obtained from the E1p-E2p didomain subcomplex as well as with other biochemical and NMR data reported from our groups, confirming that our findings are applicable to the entire E1p-E2p assembly.

The pyruvate dehydrogenase multienzyme complex (PDHc) 4 catalyzes the oxidative decarboxylation of pyruvate, producing acetyl-coenzyme A (acetyl-CoA) for the citric acid cycle and other metabolic pathways (1,2). PDHcs are large complexes minimally composed of multiple copies of three enzymatic components: the thiamin diphosphate (ThDP)-dependent pyruvate dehydrogenase (E1p; EC 1.2.4.1); dihydrolipoamide acetyltransferase (E2p; EC 2.3.1.12) containing lipoyl domains with a covalently amidated lipoyl group; and dihydrolipoamide dehydrogenase (E3; EC 1.8.1.4) binding FAD and NAD ϩ , with 24 copies of E1p (mass 99,474 Da), 24 copies of E2p (mass 65,959 Da), and 12 copies of E3 (mass 50,554 Da) present in the Escherichia coli complex. In PDHcs, the enzymes are assembled in a highly symmetrical arrangement around a central core built mainly from catalytic domains in the E2p components, with the peripheral E1p and E3 components displaced and separated from the core by large distances but tethered to it by flexible linkers non-covalently bound to them (3). For the E1p component, there are three types known: ␣ 2 homodimeric and ␣ 2 ␤ 2 heterotetrameric, both in prokaryotes, and a ␣ 2 ␤ 2 heterotetra-meric E1p subject to regulation via phosphorylation in eukaryotes. There is a correlation between the E1p component nature and the overall multienzyme complex architecture, with homodimeric ␣ 2 E1p typically found in 60-subunit, ϳ4.5-MDa PDHc complexes having overall octahedral symmetry but containing a cubic core, whereas heterotetrameric ␣ 2 ␤ 2 E1ps are usually found in larger, roughly 120-subunit, ϳ10-MDa PDHc complexes having overall icosahedral symmetry with a dodecahedral core. In both architectures, catalytic sites in the E1p decarboxylate pyruvate, those in E2p generate the key metabolic product acetyl-CoA, and those in E3 restore initial redox conditions in E2p, enabling cycling while generating NADH. Each E2p has a flexible, multidomain structure. Proceeding from the N terminus, the E. coli E2p consists of three tandem lipoyl domains (LD; named LD1, LD2, and LD3, ϳ80 amino acids each), a peripheral subunit-binding domain (PSBD; ϳ45 amino acids), and a larger, core-forming C-terminal catalytic acetyltransferase domain (ϳ250 amino acids), with each domain linked to the next by a flexible 25-30-residue segment rich in alanine, proline, and charged amino acids (4,5). The PSBDs are responsible for non-covalent tethering of the peripheral E1p and E3 components to the E2p core, thus maintaining the overall complex integrity, whereas lipoamides (lipoylated lysine residues) on the LDs are responsible for shuttling intermediates between the E1p, E2p, and E3 catalytic sites via a "swinging arm" substrate channeling mechanism (1). A schematic diagram of the overall process is depicted in Fig. 1, including representations of the E2p domain structures and of the E2p didomain.
The homodimeric E. coli E1p catalyzes the rate-limiting step in the overall PDHc reaction (6), whereas structural and biochemical studies revealed the important regions and residues involved (7)(8)(9)(10)(11)(12). Two residues, Glu-571 and His-407, present in the E1p active site play important roles for the ThDP binding: reaction activation for Glu-571 and communication between E1p and E2p-lipoyl domains for His-407 (13). The residue Glu-571 is a highly conserved glutamate in virtually all ThDP-dependent enzymes (the known exception being glyoxylate carboligase) and assists in accessing the rare and critical 1Ј,4Јiminopyrimidyl-ThDP tautomer on the enzyme (14). The H407A E1p variant clearly indicated the importance of residue His-407 in postdecarboxylation steps (13).
In all previous x-ray structures of E. coli E1p, including the apoenzymes (ThDP-and Mg 2ϩ -free) and holoenzymes as well as variants with active center substitutions and complexes with the inhibitor thiamin 2-thiathiazolone diphosphate, and the reaction intermediate analogue 2-phosphonolactyl-ThDP, the N-terminal region residues 1-55 were totally disordered and not locatable (7)(8)(9)(10)(11)(12). Here and throughout this work we use the terms "ordered" and "disordered" only in the crystallographic sense (i.e. convincingly recognizable in electron density maps or not, respectively). In either case these terms imply nothing about whether the region involved is "folded" or "unfolded." Biochemical studies of E. coli and Azotobacter vinelandii E1ps, however, indicated a role for the N-terminal regions in binding to their E2p components, whereas N-terminal deletion variants revealed that not one but both N-terminal regions of the two E1p subunits constituting a functional dimer are required for overall activity of the multienzyme complex (15)(16)(17)(18). It has also been suggested that the N-terminal region of the E. coli 2-oxoglutarate dehydrogenase component (E1o) interacts with its E2o component because, upon deletion of the first 77 N-terminal residues of E1o, it failed to assemble with its E2o (19). Only one E1-E2 subcomplex has been structurally determined to date, that from an icosahedral, Bacillus stearothermophilus PDHc containing a heterotetrameric (␣ 2 ␤ 2 ) E1p in complex with its PSBD. That structure revealed that the PSBD of E2p was bound to the C-terminal region of the E1p ␤ subunits via electrostatic interactions (20).
The amino acid sequences of E1p from Gram-negative bacteria contain a large percentage of highly conserved, acidic amino acid residues in their N-terminal regions. In addition, spectroscopic, biochemical, and kinetic studies on E. coli PDHc suggest that the negatively charged residues Asp-7, Asp-9, Glu-12, and Asp-15 in the N-terminal region of E1p are important for overall activity and for interaction with its E2p (15). It was thus postulated that these negatively charged residues interact with the positively charged residues present in the PSBD of E2p to facilitate E1p to E2p core tethering within the assembled complex. In addition, extensive biochemical and NMR studies on interaction of E. coli E1p with E2p-derived proteins revealed that the entire N-terminal region of E1p is involved in the interaction with E2p.
In order to obtain structural information regarding E1p and E2p assembly for a member of the octahedral PDHc family containing homodimeric (␣ 2 ) E1ps, we have determined and analyzed the crystal structure of E. coli E1p in complex with an N-terminal E2p didomain. This didomain consists of a single hybrid lipoyl domain (LD h ), comprising residues 1-46 of LD1 and 251-289 of LD3, and the PSBD along with linker residues. The E2p with this LD h is denoted as 1-lip E2p, whereas the wild type intact E2p is denoted as 3-lip E2p (Fig. 1). The 1-lip E2p displays similar kinetic behavior as the 3-lip E2p according to the NADH (overall complex) assay, indicating the LD h is functionally competent (48).
Our results lead to the following novel conclusions. (a) The crystal structure of the E1p-E2p didomain subcomplex reveals a disorder-to-order transformation in the N-terminal region of E1p (as compared with the E1p structure by itself) enabling visualization of a new, domain-swapped, four-helix bundle. (b) Binding of E1p (␣ 2 ) to PSBD in the E2p didomain differs substantially in both location and nature relative to that seen in the other known example, the complex of PSBD with an ␣ 2 ␤ 2 E1p from B. stearothermophilus. (c) Structural data on the E1p-E2p didomain are consistent with HD exchange mass spectrometric (HDX-MS) studies of the loci of interaction of the E1p with 3-lip E2p and with biochemical studies reported earlier. (d) The novel assembly reported here has never been observed before in E1 structures from any source and should be common to other PDHcs and related complexes from Gram-negative bacteria. (e) The newly identified E1p-E2p tethering interactions probably affect the overall assembly of the entire PDHc. through the BsaI and XhoI restriction sites. The nucleotide sequence was confirmed by Genewiz, Inc. DNA encoding His 6tagged E1p (E1p-pSUMO) was introduced into Rosetta (DE3) cells (Novagen), and cells were grown in 100 ml of LB medium supplemented with 40 g/ml kanamycin and 34 g/ml chloramphenicol for 16 h at 30°C. 40 ml of the overnight culture was inoculated into 2 liter of the LB medium supplemented with 40 g/ml kanamycin and 34 g/ml chloramphenicol. The cells were grown to an A 600 of 0.8, and then isopropyl 1-thio-␤-Dgalactopyranoside (1 mM) was added, and cells were grown for an additional 4 h at 37°C. Cells were washed with 0.10 M sodium phosphate (pH 7.2) containing 0.15 M NaCl (1ϫ PBS) and stored at Ϫ80°C. The harvested cells (10 g) were thawed and were resuspended in 25 mM HEPES (pH 7.5) containing 0.2 M KCl, 2 mM ␤-mercaptoethanol, 640 g/ml benzamidine⅐HCl, 0.6 g/ml leupeptin, and 2.7 g/ml pepstatin (Buffer A). Cells were treated with lysozyme (0.5 mg/ml) at 4°C for 20 min and were disrupted by sonication. The clarified lysate was loaded onto a 5-ml His Trap HP column (GE Healthcare) at 1 ml/min. The column was washed with 20 column volumes of Buffer A containing 5 mM imidazole. Next the column was washed with 5 column volumes of Buffer A with 25 mM imidazole, and the final wash was 5 column volumes of Buffer A with 50 mM imidazole. E1p was eluted with Buffer A with 400 mM imidazole.
For crystallization, His 6 -tagged E1p was treated as follows. 20 l of (10 units/l) Ulp-1 protease (Invitrogen) was added to His 6 -tagged E1p, and the reaction mixture was dialyzed against Buffer A for 16 h to cleave the His 6 tag. To remove the His 6 tag, the mixture was loaded onto a 5-ml HisTrap HP column (GE Healthcare). The flow-through E1p free of His 6 tag was concentrated using a Vivaspin 20, 100,000 molecular weight cut-off concentrating unit (GE Healthcare) to a concentration of E1p of 25 mg/ml. Construction of plasmid and expression and purification of E2p didomain (15) and His 6 -tagged E2p didomain (21) were reported earlier. The didomain, starting from the N terminus, contains a single hybrid lipoyl domain LD h , linker, and the PSBD, totaling 190 amino acids, and was engineered from 1-lip E2p, where the LD h has the N-terminal sequence of the first and C-terminal sequence of the third lipoyl domain in the wild type 3-lip E2p (Fig. 1).
Expression and Purification of 3-lip E2p-Wild type 3-lip E2p was expressed and purified in a manner similar to 1-lip E2p (22). The pCA24N plasmid encoding 3-lip E2p was obtained from the National BioResource Project, National Institute of Genetics, Japan. The AG1 cells were grown at 37°C in LB medium supplemented with 30 g/ml chloramphenicol and 0.30 mM lipoic acid. The expression of 3-lip E2p was induced by isopropyl 1-thio-␤-D-galactopyranoside (0.5 mM) for 4 h at 37°C. The purification protocol for 3-lip E2p was similar to that for 1-lip E2p (22).
Assembly of the E1p-E2p Didomain Subcomplex-8 mg of E1p and 4 mg of E2p didomain were mixed together and were incubated for 1 h at 4°C. The reaction mixture was injected into a Superdex G200 16/60 column (GE Healthcare) equilibrated with 20 mM HEPES (pH 7.0), containing 120 mM KCl, 2 mM ␤-mercaptoethanol, and 0.02% sodium azide (Buffer B). Fractions of 1.5 ml were collected at a flow rate of 1.5 ml/min and were analyzed by SDS-PAGE. Fractions containing the E1p-E2p didomain subcomplex were pooled together and were concentrated using a Vivaspin 20 concentrating unit (100,000 molecular weight cut-off) (GE Healthcare). The concentration of KCl was reduced to 60 mM, and samples were stored at Ϫ80°C.
Lipoylation of the His 6 Tag E2p Didomain-Lipoylation of the E2p didomain was as reported recently (21). Briefly, to representation of the covalently amidated lipoyl group (in its initial state) present in E2p lipoyl (LD) domains; it sequentially enters into and transfers intermediates between catalytic sites. C, schematic representing the swinging arm substrate channeling sequence of events taking place in the PDHc complex during a catalytic cycle. PSBD binding is required to maintain overall complex integrity, and in some of the E2p subunits, its PSBD is bound to E1p, whereas in others, it is bound to E3. Although a useful depiction, within the multienzyme complexes, the situation is actually much more complicated because there are many copies present of each component shown; E1p and E3 are both dimers containing two active sites, E2p is a trimer containing three active sites (and swinging arms), and each flexible linker is long enough to reach more than just the nearest E1p, E2p-core, and E3 catalytic sites. ensure full lipoylation of the E2p didomain, the lipoylation reaction was conducted in vitro using E. coli lipoyl protein ligase (10 M), 1.2 mM ATP, 1.2 mM MgCl 2 , 0.6 mM lipoic acid, and 60 -150 M E2p didomain in 50 mM NH 4 HCO 3 (pH 7.0), as reported earlier (23,24). Lipoylation was confirmed by ESI-MS.
Reductive Acetylation of the His 6 -tagged E2p Didomain-In steady state experiments, the His 6 -tagged E2p didomain (40 M) and E1p (0.2-0.4 M) in 50 mM NH 4 HCO 3 (pH 7.0) containing MgCl 2 (4 mM) and ThDP (0.40 mM) in syringe A was mixed with pyruvate (2 mM) in syringe B, and the reaction was stopped at times 0.2-16 s by the addition of 83 l of quench solution containing 50% methanol and 2% formic acid in a rapid chemical quench apparatus (Kintek RQF-3 model, Kintek Corp.). Samples were diluted to result in a concentration of E2p didomain of 1-2 M in 50% methanol, 0.2% formic acid and were analyzed for the relative concentrations of the acetylated and unacetylated His 6 -tagged E2p didomain by ESI.
In a single-turnover experiment, the E2p didomain (30 M) and E1p (30 M active center concentration) in 50 mM NH 4 HCO 3 (pH 7.0) containing ThDP (0.4 mM) and MgCl 2 (2.0 mM) in syringe A were mixed at 25°C with pyruvate (2 mM) in the same buffer in syringe B, and the reaction was stopped at times 0.005-1.0 s by the addition of quench solution containing 50% methanol and 2% formic acid in a rapid chemical quench instrument according to our recently published protocol (6). Samples of E2p didomain were diluted to 1-2 M concentration and were analyzed by ESI-MS for the presence of the acetylated and unacetylated E2p didomain. Data were fitted to a single exponential Equation 1 or double exponential Equation 2.

Crystallization and Data Collection for E1p-E2p Didomain
Subcomplex-The final concentration of the E1p-E2p didomain subcomplex (His tag removed from E1p, no in vitro E2p didomain lipoylation) used for crystallization was 22 mg/ml in 20 mM HEPES (pH 7.0), 40 mM KCl, 2 mM ␤-mercaptoethanol, and 0.02% sodium azide. Crystals were obtained by the sitting drop vapor diffusion method at 22°C. The best crystals were obtained with a precipitant/reservoir solution having 20% PEG 3350, 0.2% NaN 3 , and 0.2 M ammonium tartrate dibasic buffer (pH 6.35). Drops were 4 -6 l consisting of equal parts of reservoir and protein solution. Crystals appeared in 3-5 days and typically grew to 0.15 ϫ 0.20 ϫ 0.30 mm in size over 4 -6 weeks. They were then briefly soaked in 30% PEG 3350 prior to flashcooling in liquid nitrogen. Low temperature (Ϫ180°C) diffraction data were collected at a synchrotron (SERCAT, Sector 22-BM, Advanced Photon Source, Argonne National Laboratory) to 2.8 Å resolution due to rapid decay, although higher resolution was initially (and only briefly) observed. Attempts to acquire the higher resolution data initially observed included the use of high flux minibeams, helical scans, segmented scans, searching for hot spots, and merging of data from multiple crystals, but none of these methods provided higher resolution data that scaled/merged with acceptable statistics. The data were processed with the d*TREK package (25). The crystals are orthorhombic with a ϭ 210.94, b ϭ 326.84, c ϭ 77.21Å in space group P2 1 2 1 2. The Mathews coefficient V m is 3.04 Å 3 /Da, based on 16 E1p subunits/cell (26). Crystal parameters and data collection statistics are given in Table 1.
Structure Determination and Refinement-The molecular replacement method was used for initial structure determination of the E1p-E2p didomain subcomplex, starting with the E1p structure (7) as a search model in the program PHASER (27). The single subunit search model included the 801 amino acids previously reported in a monomer (i.e. without the missing residues 1-55, 401-413, and 541-557, as well as the cofactors). The results clearly indicated four subunits (i.e. two E1p dimers) in the crystallographic asymmetric unit. Following rigid-body refinement, simulated annealing was performed with the PHENIX program (28) without imposing any non-crystallographic symmetry. 2F o Ϫ F c composite, simulated annealing, omit maps were then calculated and revealed several segments of electron density corresponding to ␣-helices close to the N-terminal region of the two dimers, but they were of poor quality. Electron density for the cofactors ThDP and Mg 2ϩ was also diffuse, and they were not included. Several polyalanine ␣-helix models (total of 12 ␣-helixes) were then built into the new densities using Coot (29). Subsequent refinement/model building iterations allowed for unambiguous tracing of the N-terminal E1p residues 5-55 and residues 122-167 corresponding to the PSBD of E2p (Fig. 3A). In the final map, there is no clear electron density for the E2p lipoyl domain region and most of the linker to the PSBD. This is presumably due to disorder because gel analysis of washed and dissolved crystals indicated that the full E1p and E2p didomain are present, and we note that the lipoyl domain has never been observed in any crystal structure complex with E1 or E3 components by anyone before. For the E1p, the N-terminal residues 1-4 and active site loop residues 401-405 and 543-556 were completely disordered, as has been the case in all other E. coli E1p structures in the absence of intermediate analogues (7,10). The observed residues, therefore, included 406 -413, 541-542, and 557 at the ends of regions previously disordered in the absence of intermediate analogues and, most importantly, the N-terminal residues 5-55, which had never been observed before. Only the observed residues were included in the final refinement. The final model, refined with the program BUSTER (30), contains two homodimers of E1p, two copies of PSBD, and 1293 water molecules. The working and free R values are 0.198 and 0.233, respectively; further statistics regarding the quality of the model are given in Table 1. The geometry/stereochemistry of the final model was also validated by the program MolProbity (31) with a MolProbity score of 1.68 and percentile of 100%. The electron density did not reveal the presence of the LD h but did show the PSBD. Graphical representations of protein models were generated with the program RIBBONS (32).

Methods Used for MS-detected Hydrogen/Deuterium Exchange Studies of the E1p-3-lip E2p
Interaction Loci-The HDX-MS analysis was conducted as described (21,33). The interaction of proteins used in this study (E1p and 3-lip E2p) was analyzed with HDX-MS, both in the free state and in the complexed state. Prior to deuterium labeling, three individual protein stock solutions (80 M E1p, 80 M 3-lip E2p, and a mixture of 80 M E1p and 3-lip E2p, all in aqueous buffer containing 10 mM KH 2 PO 4 (pH 7.0), 50 mM KCl, 0.2 mM ThDP, and 1 mM MgCl 2 , were allowed to equilibrate for 1 h at 25°C. The deuterium labeling reaction was initiated by diluting 15 l of protein stock solutions into 285 l of labeling buffer (10 mM KH 2 PO 4 (pH 7.0), 50 mM KCl, 0.2 mM ThDP, 1 mM MgCl 2 , 99.9% D 2 O), followed by incubation at 25°C. At selected time points ranging from 20 s to 1000 min, 30-l aliquots from the labeling reaction were rapidly quenched by 30 l of ice-cold quench buffer (0.2 M KH 2 PO 4 , pH 2.6). The samples were immediately frozen in liquid nitrogen and stored at Ϫ80°C before analysis. Non-deuterated samples were generated following the same procedure except that protein samples were diluted into aqueous buffer and incubated for 1 min followed by the quench process. The frozen deuterated sample was rapidly thawed and loaded with an ice-cold syringe into a 20-l sample loop inside the refrigeration system. The protein sample (ϳ40 pmol) was carried by a 0.3-ml min Ϫ1 digestion flow of 0.1% (v/v) formic acid into an immobilized pepsin column (Poroszyme Immobilized Pepsin Cartridge, 2.1 ϫ 30 mm; Applied Biosystems) and digested at 15°C for 20 s. The resultant peptides were immediately cooled down to 0°C through a heat exchanger and were concentrated and desalted on a peptide trap (Michrom peptide MacroTrap; 3 ϫ 8 mm). The peptides were eluted and separated in 15 min through a reversed-phase C18 HPLC column (Agilent Poroshell 300SB-C18, 2.1 ϫ 75 mm) at a flow rate of 0.2 ml min Ϫ1 with a 0°C 2-40% acetonitrile gradient containing 0.1% (v/v) formic acid. ESI-FT-MS measurements began 5 min after the initiation of the elution process and lasted 10 min. The time from initiation of digestion to elution of the last peptide was less than 20 min. The mass spectrometer settings were as follows: ESI ϩ mode; capillary, 4500 V; spray shield, 4000 V; drying gas temperature, 190°C; mass acquisition range, 400 -2000 m/z; scan rate, 0.5 scans/s (see Table  2 for E1p and Table 3 for 3-lip E2p). All experiments were run in triplicate.
Bruker Daltonics DataAnalysis version 4.0 was used for spectrum analysis and data treatment. Peptides were identified from non-deuterated samples by a customized program, DXgest, which matches experimental peptide mass with theoretically generated peptic peptide mass by using statistical data (34) for the pepsin cleavage pattern under hydrogen/deuterium exchange conditions. Mass tolerance was set at Ͻ1.0 ppm for accurate identification. Hydrogen/deuterium exchange data for each individual peptide at various time points were processed using the program HX-Express (35). No back-exchange correction was needed for purposes of comparative analysis. The number of exchangeable backbone amides was calculated as described (36,37). Butterfly and difference plots were produced by Origin (OriginLab, Northampton, MA) and Microsoft Excel (38).

RESULTS AND DISCUSSION
Evidence for Reactivity of E2p Didomain from Kinetic Studies-Before interpretation of the structural data, we wished to ascertain that the E2p didomain is functionally competent. The functional interaction of E1p with E2p didomain was confirmed in the model reductive acetylation reaction by direct measurement of the masses of acetylated and unacetylated E2p didomains detected at different times of incubation with E1p and pyruvate. , for reductive acetylation) of 78 s Ϫ1 was calculated from the initial slope of this curve (Fig. 2). In a pre-steadystate experiment, k r values of 96 and 134 s Ϫ1 were reported by us (6). The calculated rate constants are similar to the k cat of 95 s Ϫ1 determined in the overall reaction of NADH production by PDHc reconstituted from E1p, E2p (the entire E2 subunit), and E3, indicating functional competence of the E2p didomain used in this model reaction.
Additional evidence of the reactivity of the E2p didomain is that an H407A substitution in E1p, produced a significant reduction of the reaction rate (k r ϭ 0.04 Ϯ 0.01 s Ϫ1 ) in the model reaction, which correlates well with the significant reduction of the overall activity of NADH production (0.15% activity remaining) by PDHc reconstituted from H407A E1p, 1-lip E2p, and E3 (14). These kinetic data clearly indicate that the E2p didomain employed is functionally active and can be reductively acetylated by E1p and pyruvate with a rate constant that is comparable with k cat for the overall PDHc reaction. Our groups recently published HDX-MS studies comparing the interaction of E1p with E2p didomain and 3-lip E2p, revealing that both E2p proteins react with similar regions of E1p (21).
We also demonstrated that NMR methods are capable of providing sequence-specific interaction of the E1p with the E2p  (21). These points are important to keep in mind as we consider the x-ray structure.
Overall Description of the E1p-E2p Didomain Subcomplex Crystal Structure-The structure of the E1p in the E1p-E2p didomain subcomplex is generally very similar to the previously reported crystal structure of E1p by itself (7). Two independent E1p homodimers are present in the asymmetric unit of the crystal, and they are referred to here as AB and CD. The asymmetric unit also was found to contain two copies of the PSBD (named E and F), with one bound to each E1p dimer. The overall structural folds are essentially the same for the two PSBDs, and each binds to its respective E1p dimer in a similar manner at corresponding dimeric interfaces. Accordingly, the results discussed here apply to both subcomplexes. In the E1p, a small number of N-terminal residues (residues 1-6 and 47-55) and several residues in surface loops (residues 401-404 and 543-556) are disordered in all four subunits and were not included in the final model. In addition, the active site residues 261-269 in E1p sub- with the diffraction data, and 96% of the 3526 residues fall in the most favored Ramachandran regions. By comparing the E1p structure in the subcomplex including the E2p didomain with the E1p structures by themselves published earlier (7)(8)(9)(10)(11)(12), no significant perturbations are found in the previously observed regions of E1p (residues 56 -886) caused upon complexation with E2p didomain because the r.m.s. deviation is 0.59 Å for ␣ carbons in the corresponding regions. As was often the case, in the subcomplex, electron density for the residues in surface loops 401-404 and 543-556 at the entrance to the active site was missing in all four E1p subunits. Also in both dimers, in one of the active sites (B in AB dimer and D in CD dimer), residues 262-269 close to the ThDP site were disordered, whereas in the other active site, these residues were ordered, revealing a structural asymmetry of the two active sites. However, in the current E1p-E2p didomain subcomplex, the most important observation is the well defined electron density present for N-terminal residues 5-55, which had never been observed in any of the E1p structures reported by us earlier (7)(8)(9)(10)(11)(12). In each subunit within an E1p dimer, near its N terminus, this region folds into a helix-turn-helix motif, which interlocks with its counterpart from the "other" subunit to form a domain-swapped, four-helix bundle. Each helix-turnhelix motif is connected to the remainder of its E1p subunit by extended loop residues 46 -55. The two helices present in each helix-turn-helix motif are defined here as N-␣h1 (residues 10 -27) and N-␣h2 (residues 30 -45). The four-helix bundle serves as a recognition site for binding of E2p PSBD to E1p, as one PSBD binds to each bundle formed by an E1p dimer, as shown in Fig. 3, A and B. Although the four-helix bundles (and associated PSBDs) in the two dimers present in an asymmetric unit are nearly identical structurally, the orientations of the

E1-E2 Tethering in the Pyruvate Dehydrogenase Complex
OCTOBER 24, 2014 • VOLUME 289 • NUMBER 43 bundles/PSBDs relative to the rest of the E1p structure differ profoundly. To adopt the same orientation with respect to the main E1p structure (residues 56 -886), a rotation of ϳ64°is required for one dimer's bundle to align with the others, as is shown in Fig. 3D. This feature is attributed to flexibility in the regions of residues 46 -55 connecting each helix-turn-helix motif in the bundle to the rest of its covalently bound subunit and introduces a substantial amount of asymmetry into the otherwise non-crystallographic 2-fold-related character of the overall E1p dimers.
Overall Description of the E2p Didomain-In the crystal structure of the E1p-E2p didomain subcomplex, the electron density did not reveal the presence of the LD h but did show the PSBD. The PSBD was found to be composed of 46 residues starting from 122-167. It contains two parallel ␣-helices comprising residues 125-136 (H1) and 152-166 (H2), a distorted short helix consisting of residues 138 -142, and loop residues 143-151 joining the two larger, parallel ␣-helix segments. The helices H1 and H2 have extensive hydrophobic interactions involving residues Ile-128, Leu-131, and Phe-135 from helix H1 with residues Val-156, Tyr-159, and Val-160 from helix H2. A single PSBD binds to each E1p homodimer and is situated near the highly localized, pseudo-2-fold symmetry axis relating corresponding elements of the four-helix bundle. The overall structures of the two PSBDs are nearly identical and very similar to those of other structures reported earlier associated with different E3 species (39 -43) and a heterotetrameric ␣ 2 ␤ 2 E1p subcomplex with PSBD (20). The r.m.s. deviation obtained when comparing 46 C␣ atoms of the two PSBD structures is 0.43 Å. For both subcomplexes in the asymmetric unit, the H1 helices of the PSBD residues 125-130 intimately contact the four-helix bundle in the respective E1p dimers. Helix H2, comprising residues 152-162, does not interact with E1p but may be involved in extensive hydrophobic interaction with residues in helix H1. The orientation of the PSBD with respect to its fourhelix bundle binding partner is not influenced by the different orientations of the four-helix bundles themselves with respect to the main E1p dimers, as was shown in Fig. 3C. Amino acid sequences for the N-terminal part of E. coli E1p and the PSBD are given in Table 4.
The E1p-PSBD Binding Interface-The association of E1p with PSBD buries roughly 1540 Å 2 surface area, which is about 39% of the total PSBD accessible surface area. The PSBD binds across the local 2-fold axis present in each four-helix bundle, thereby contacting residues in both subunits of the E1p homodimer. The subcomplex is mainly stabilized by electrostatic and hydrophobic interactions between the PSBD and residues from the interface formed by the N-␣h1 helices of two E1p subunits in the bundle. Surface residues in the PSBD are largely electropositive, and complementary surface residues from E1p are mainly electronegative. PSBD basic residues Arg-129, Arg-130, and Arg-133 from helix H1 as well as residues Arg-147, Lys-148, and Arg-150 from the loop region are involved in salt bridges with E1p acidic residues Asp-9, Glu-12, Asp-15, Glu-26, and Glu-27, as well as interacting with Gln-18 and Ser-22 from the N-␣h1 helix of both E1p subunits. Interestingly, salt bridges involving helix H1 residues Arg-129, Arg-130, and Arg-133 are located at one end of the E1p-PSBD interface, and salt bridges for loop residues Arg-147, Lys-148, and Arg-150 are located in the opposite end of the binding surface.
Hydrophobic interactions are also an important component of the E1p-PSBD interface and are located between the two salt bridge regions. Pro-126 from the PSBD fits snugly into a hydrophobic pocket formed at the E1p subunit interface composed of Trp-16, Ala-19, and Val-23 from the N-␣h1 helix of both subunits. The residue Trp-16 from each E1p subunit packs on opposite sides around the ring of Pro-129 from the PSBD, and this proline is present in all E2p components. In addition to the salt bridges and hydrophobic interactions, there are two PSBD residues, His-123 and Thr-125, involved in hydrogen bonding interactions with E1p residues Ser-22 and Asp-15, respectively. Key binding interactions between the PSBD and the E1p N-terminal four-helix bundle are shown in Fig. 4A.
X-ray Results Are Consistent with Biochemical and NMR Studies-The current structural analysis of the E. coli ␣ 2 E1p-E2p didomain subcomplex is consistent with earlier biochemical studies reported for the E. coli PDHc (15). The E1p substitutions of negatively charged N-terminal N-␣h1 helix residues Asp-7, Asp-9, Glu-12, Asp-15, and Glu-27 displayed low overall activity and impaired interaction with the E2p component. Kinetic and binding studies established that the N-terminal region of E1p interacts with the E2p PSBD region and that this interaction precedes reductive acetylation. Sequence-specific NMR assignments enabled us to carry out 15 N-1 H HSQC experiments, revealing that upon complexation of E1p with the E2p didomain (similar to that used in the x-ray structure), the resonances for Trp-16, Gly-47, Gln-18, Gln-33, and Gln-38 were all reduced/absent compared with the spectrum of E1p by itself, indicating that the entire N-terminal region of E1p is affected by interaction with the E2p didomain (15). Complexation of E1p with either E2p didomain or a synthetic PSBD peptide, but not with independently expressed LD h , led to reduction/absence of the side chain NH resonance of Trp-16, the NH resonance of Gly-47, and side chain resonances for Gln-18, Gln-33, and Gln-38, clearly indicating that PSBD, but not the LD h in the E2p didomain, is essential for the E1p-E2p interaction (15). None of the N-terminal substitutions affected the E1p-specific activity (reflecting the ability to decarboxylate pyruvate), indicating that these substitutions prevented binding of E1p to the PSBD of E2p. The results show that the PSBD is the most important E2p domain for assembly with E1p, whereas the lipoyl domain provides the E2p recognition element for the catalytic cycle (reductive acetylation by E1p) (44,45). The fact that the site-directed mutagenesis study of negatively charged N-terminal residues resulted in diminished binding agrees well with the crystal structure analysis of the E1p-E2p didomain subcomplex. The crystal structure studies explain the results of the solution studies and clearly establish the antiparallel helix interface formation for the E1p N-terminal region N-␣h1 helix residues and hydrophobic and electrostatic environments for protein-protein association.
Similarities and Differences in the E1p-E2p Binding Mode for Related Enzymes-Solution studies of the related enzyme ␣ 2 E1p from A. vinelandii also indicated interaction of N-terminal residues with its E2p PSBD (17)(18). Kinetic and binding studies indicated that the A. vinelandii E1p N-terminal residues Asp-  (20) reveals both similarities and differences in the binding interface. The overall folds and secondary structure elements are similar for the two PSBDs, and residues from the H1 helices and immediate loop regions of the binding domains interact with their respective E1p dimers. Sequence identity of the two PSBDs is 46%, and comparable C␣ atoms have r.m.s. deviations of 0.6 Å. In E1p-PSBD complexes, positively charged Arg and Lys residues from the PSBD interact with negatively charged residues from the corresponding E1p dimer interface. These basic residues are located in the same position in the H1 helix and loop region of the two PSBDs, although the sequence identity is low in the rest of the protein. The binding site interfaces in the two E1p structures, however, are significantly different. For the ␣ 2 E1p-E2p didomain subcomplex, the binding site is located in the N-terminal domains of the E1p dimer, and its interface is formed by a four-helix bundle interleaving each The key interactions involve two regions of electrostatic contacts and a region of hydrophobic contacts in between. B, hybrid drawing of the B. stearothermophilus PSBD-E1p interaction region at the junction of its PSBD and ␤ subunits in the E1p heterotetramer (20). Key interactions involve electrostatic contacts flanking a region of hydrophobic interactions. In both A and B, different E1p subunits are colored green and yellow, and the PSBD is colored purple. C, ribbon drawing (32) showing the relationship between homodimeric (E. coli) and heterotetrameric (B. stearothermophilus) E1p complexes with their respective PSBDs and the large displacement of the PSBD binding site. The structures were placed in a common orientation by least squares alignment starting with active site residues and extended outward by including other structurally matching residues (neglecting the PSBDs). The images in B and the right side of C were constructed from Protein Data Bank entry 1W85.

E1-E2 Tethering in the Pyruvate Dehydrogenase Complex
subunit's helix-turn-helix motif. This four-helix bundle is only observed in the presence of PSBD; otherwise, it is totally disordered, including the loop region residues (residues 46 -55) linking it to the remaining E1p structure. In stark contrast, for the ␣ 2 ␤ 2 E1p-PSBD subcomplex, the binding region is located in the C-terminal domains of ␤-subunits and is ϳ116 Å away from that in the ␣ 2 E1p-PSBD (Fig. 4C).
Previous structural analysis indicated the association of PSBD does not cause any appreciable conformational changes in the PSBD binding region or in other parts of the ␣ 2 ␤ 2 E1p heterotetramer structure (20). This is also the case for the ␣ 2 E1p dimer except for the ordering induced in the N-terminal four-helix bundle-binding region itself. The antiparallel helices in the ␣ 2 E1p dimer (N-␣h2 from two subunits) form a hydrophobic pocket with interior residues Ala-32, Leu-35, Ile-36, Leu-39, Leu-40, and Ala-43 stabilizing the dimeric association of the four-helix bundle, but a similar hydrophobic environment is missing in the ␣ 2 ␤ 2 E1p heterotetramer subcomplex. The protein-protein association buries close to 1540Å 2 surface area in the ␣ 2 E1p-PSBD subcomplex and 1100Å 2 for the ␣ 2 ␤ 2 E1p-PSBD subcomplex, suggesting stronger PSBD binding with the ␣ 2 E1p dimer. Surprisingly, in both E1p (␣ 2 and ␣ 2 ␤ 2 ) binding interfaces, residues involved in interaction with the PSBDs maintain a strong, pseudo-2-fold relationship despite the asymmetric nature of the PSBD.
Except in multienzyme complexes of higher eukaryotes, PSBDs in each E2p component bind both E1p and E3 components in the complex. In higher eukaryotes, E1p is bound only to the binding domain of E2p, and E3 is bound only to the subunitbinding domain of the E2p-like, E3-binding protein. Recently reported structures of several E3 components in complex with PSBD, the structure of the ␣ 2 ␤ 2 E1p-PSBD subcomplex, and the current ␣ 2 E1p-E2p didomain subcomplex indicate that the overall fold and secondary structural elements are similar for PSBDs of E2p regardless of species, with pairwise comparisons of C␣ atoms from all known PSBD structures having r.m.s. deviations between 0.6 and 1.5 Å. The structures also reveal the binding interface for PSBD in the E3 dimer to be located in the same region (interface domain) for E3 components from different species (39 -43). In contrast, the contact surface for the PSBD is situated in markedly different regions for ␣ 2 E1p and ␣ 2 ␤ 2 E1p. Compared with the ␣ 2 ␤ 2 E1p binding interface region, the PSBD binding N-terminal four-helix bundle in ␣ 2 E1p is more flexible with respect to the main E1p structure (residues 56 -886), and this flexibility may be common to all Gram-negative bacteria (such as Vibrio cholerate, Haemophilus influenzae, A. vinelandii, and Neisseria meningitidis) ␣ 2 E1ps. Homodimeric ␣ 2 E1ps typically have more than 90 additional residues than ␣ 2 ␤ 2 E1ps, and sequence alignment indicates they are located at the N-terminal region, which is where PSBD binding occurs. It appears likely then that most, if not all, ␣ 2 E1p class enzymes will involve PSBD binding in the manner demonstrated in this report and possibly induce similar disorderto-order transformations in the N-terminal segments.
Evidence for Interaction Loci between E1p and 3-lip E2p from HDX-MS-For these studies, intact, wild type E1p and 3-lip E2p components were used. This method enabled us to work with the intact proteins, creating an environment more similar to that found in the entire complex. The E1p-3-lip E2p interaction was profiled at peptide resolution by a comparative HDX-MS analysis with 62 peptides originating from E1p (98% sequence coverage) and 29 peptides originating from 3-lip E2p (83% sequence coverage) (Tables 2 and 3). Both the butterfly plot and difference plot show a snapshot of changes in the deuterium uptake pattern of either E1p or 3-lip E2p in the presence and absence of the other component, as seen in Fig. 5.
Upon complexation of E1p with 3-lip E2p, peptides originating from two regions of E1p were observed with significant reduction in deuterium uptake. (a) Peptides from the N-terminal region of E1p (residues 1-55), according to the crystal structure, are located in the interface bundle involved in the interaction with PSBD ( Fig. 5A and Table 2), an interaction expected to be conserved across all homodimeric E1p-PSBD structures. This region is disordered in the crystal structures of all noncomplexed E1ps (7)(8)(9)(10)(11)(12).  7) involved in ThDP binding as well as the two loops (inner active center loop 401-413 and outer active center loop 541-557), which were previously shown (9,46) to undergo a disorder to order transformation required for E1p-E2p active center communication according to our x-ray results. Residues 262-277 are in the loop containing Asn-258, Asn-260, and Gln-262 involved in Mg 2ϩ coordination. Residues 92-110 encompass Ser-109 and His-106 for ThDP binding. Residues 317-347, forming a distorted helix (according to x-ray in this paper), showed an immediate 1.6-Da reduction of deuterium uptake at 20 s, and the value persisted, suggesting hindrance at certain residues in this region. These findings suggest a cooperative conformational change between the ThDP-binding active center and the open and close movement of the loops in the presence of 3-lip E2p.
As to PSBD binding to E1p, almost the entire PSBD was involved. Residues in both of the peptides 320 AENDAYVHATPLIRRLARE 338 and 343 LAKVKGTGRKGRILREDVQA 362 showed decreased deuterium uptake upon complexation with E1p (corresponding to residues 116 -134 and 139 -158, respectively, in the E2p didomain), whereas residues 339 FGVNL 343 (corresponding to residues 135-139 in the didomain) were not involved.
For the first time, an attempt was undertaken to analyze the binding of the wild type 3-lip E2p to E1p rather than with any truncated version of 1-lip E2p. The first HDX-MS experiment applied to such subcomplexes demonstrates that notwithstanding the accepted aggregation of intact E2p components, our high resolution method is capable of elucidating important information about protein-protein association in solution. Because the current crystal structure analysis did not reveal lipoyl domain-E1p active site interactions due to disorder, these solution study results, although consistent with the PSBD-E1p N-terminal structural results, also provide useful information about general, E1p-E2p interactions remote from the PSBD binding site.
Conclusions-Within the E. coli PDHc multienzyme complex, the ThDP-dependent E1p catalyzes the decarboxylation of pyruvate and subsequently the reductive acetylation of a lipoamide group bound to a LD domain in the E2p component. Recognition and binding of E2p with E1p is important for maintaining structural integrity of the overall complex and for effective substrate channeling between the E1p and E2p active centers (47). To investigate E1p-E2p assembly, we determined and analyzed the crystal structure of a subcomplex between the E1p and an N-terminal didomain of the E2p and carried out HDX-MS solution studies on the interaction of E1p and E2p. From the results, the most important conclusions are now presented.  Whereas in the structure of the B. stearothermophilus E1p-PSBD subcomplex (␣ 2 ␤ 2 E1p), the ␤ domain of the C-terminal region of E1p interacts with the PSBD (20), in the E. coli E1p-E2p didomain subcomplex, it is the N-terminal regions of E1p that are tethered to the PSBD. This constitutes a shift of binding sites to opposite ends of the E1p dimers and represents a displacement of roughly 116 Å for the sites, as is shown in Fig. 4C. This shift has significant ramifications in assembly of the intact PDHc complexes because the E1p dimers will almost certainly be oriented to have their PSBD binding sites pointing at and near the E2p cores from which the PSBDs originate, given the length of the E2p core-to-PSBD linker. The result would be a ϳ180°change in orientation for the E1p components relative to that in the B. stearothermophilus PDHc complex (20). Using the now available, proper model would be crucial for correctly orienting and positioning E. coli and related E1ps in cryo-EM (or small angle x-ray scattering) density maps of their entire complexes at the lower resolutions usually obtainable.
The observed structures of PSBDs from E. coli and B. stearothermophilus PDHcs are themselves very similar, although some sequence differences exist; thus, one expects similarities in their respective E1p binding sites because they should be complementary. There is, however, little sequence homology between the two E1ps, with that from E. coli being a member of the homodimeric ␣ 2 class and that from B. stearothermophilus being a member of the heterotetrameric ␣ 2 ␤ 2 class. Given the sequence dissimilarity and fold differences associated with greatly displaced binding sites and assembly architectures, how then, can similar PSBDs bind to the different sites? A comparison of the two PSBD binding sites in Fig. 4, A and B, shows that despite the sequence and structural differences, a common pattern does exist. In both cases, the key contacts on the PSBDs are made by two positively charged regions flanking a small, hydrophobic region centered on a proline residue, with two negatively charged regions flanking a hydrophobic pocket on the E1p dimer into which the proline residue inserts. In E. coli E1p, the hydrophobic pocket is made from 4 residues, Trp-16 and Ala-19 from both of the subunits at the dimer interface, whereas in B. stearothermophilus, it is also made from 4 residues, Phe-324 and Ile-281 from both of the ␤ subunits at their dimer interface. It is also noteworthy that in one of the key electrostatic interactions across the binding interface, the negatively charged residue Asp-9 in E. coli E1p is replaced by the C-terminal carboxylate of Phe-324 from the ␤ subunit of B. stearothermophilus E1p. These sequence differences, along with the fact that the interactions involve multiple subunits, and that in one case a main chain atom substitutes for a side chain atom, would very likely hinder, if not render impossible, identification of these binding sites as being common by bioinformatic methods.
In all previously determined crystal structures of ␣ 2 E1p, the N-terminal region 1-55 has been disordered and was never visualized (7)(8)(9)(10)(11)(12). Therefore, we conclude that the disorder-toorder transformation observed in the current study involved stabilization of the N-terminal region through the presence of its PSBD binding partner. The presence of multiple, non-identical, copies of the E1p-E2p subcomplex in the current crystal structure provides insight into possible functional roles for the flexible N-terminal residues as well as raising a new question. The ability of strikingly different complex conformations/orientations to maintain binding to the E2p via the same PSBDfour-helix bundle interactions while preserving the main E1p dimer structure (Fig. 3, C and D) would allow for additional flexibility in the overall PDHc complex. This flexibility lies in the residues 46 -55 of each subunit connecting the bundle to the rest of the E1p dimer and could conceivably facilitate substantial motion of the main dimer structure relative to the core, perhaps during assembly or even catalysis. A question that arises, however, is whether the N-terminal four-helix bundle is always intact and was never previously observed simply because flexibility in the connecting residue 46 -55 regions caused it to be disordered as a rigid group, or whether the participating N termini are themselves disordered and do not associate to form the bundle until PSBD is present.
A definitive answer would be provided by a crystal structure of the E1p dimer enabling visualization of the complete, N-terminal region in the absence of PSBD, but this has been unobtainable so far. However, we do have partial answers from our own NMR data regarding flexibility of the N-terminal E1p region (15) and analysis of the new data in this paper. The 1 H-15 N HSQC spectrum of 15 N-enriched E1p displayed Ͼ85 spots in the N-H chemical shift region, shown by selective sequence-specific assignments to correspond to amino acid residues 1-55, 401-413, and 541-557. These are residues for which there was no assignable electron density in the uncomplexed x-ray structures (the N-terminal region and the inner and outer active center loops, respectively). These represent only 10% of the 886 amino acids in the E1p, which in the homodimer has an M r of 200,000. The ability to observe the N-terminal residues but not the bulk of the others suggests that there is significant mobility in this region. Also, the indole NH 1 H chemical shift of 10.12 ppm for Trp-16, a key residue in the interaction with PSBD, as discussed above, is precisely the chemical shift for this proton when not participating in either secondary or tertiary interactions, pointing to an unstructured region, also confirmed by other assigned residues (15). Upon the addition of E2p didomain, similar to the one used for the x-ray structure, or of PSBD by itself to the E1p, the resonances corresponding to the N-terminal residues became too broad to observe, again suggesting reduced mobility (i.e. transition to an ordered state, as found in the x-ray study of the complex).
From the new data, as shown in Fig. 6, A and B, the largest changes revealed by HD exchange upon complexation occur in the E1p peptides 18 -23 and 24 -39, which essentially correspond to the two helices comprising residues 13-26 and 31-41, respectively, involved in bundle formation. The key to deducing information about the likely nature of the initial disordered state from the data is to recognize that in each of the E1p subunits, the first of these helices, residues 13-26, is also involved in all of the binding contacts with PSBD, whereas the second helix, residues 31-41, interacts only with the "first" helices, as shown in Fig. 4A, and makes no contacts with PSBD. From the structure then, backbone amide protection from exchange in the 31-41 helices is provided only by the 13-26 helices when the four-helix bundle is intact. Because the HD exchange data show added protection for the 31-41 helices upon complexation with PSBD, the results suggest that the four-helix bundle also is intact only with PSBD added, thereby accounting for this new protection. If the bundle had been preformed, one would expect no change in amide accessibility within the 31-41 helices upon PSBD binding because their residues do not contact it, and the amide-protecting interactions with the 13-26 helices would not be new because they would already have been present. Accordingly, when one considers the topology of the fourhelix bundle in the structure, its observed mode of binding to PSBD, the individual HD exchange results for the two helices involved in bundle formation, the NMR data indicating ordering of the E1p N terminus upon PSBD binding, and the NMR data showing the resonance for Trp-16 in the center of the bundle occurring precisely at the position expected for a free (unbound) Trp in the absence of PSBD, the data seem to point to bundle formation occurring upon PSBD binding and not before.
The structural results clearly show that each E1p dimer binds to a single PSBD, and it is worthwhile to consider the overall ramifications of this stoichiometry on the intact multienzyme complex. In the PDHc, there are 24 E2p subunits present, whose catalytic domains constitute the cubic core, meaning that there are also 24 PSBDs (and swinging arms). The PDHc also contains six E3 dimers, with each known to bind a single PSBD, and 12 E1p dimers, which we now know to each bind a single PSBD. Thus, only 18 of the PSBDs emanating from the E2p core are accounted for in peripheral subunit binding. What are the remaining six PSBDs doing? Are they merely redundant, or are they involved in some as yet undiscovered function? This question will require further study.
From activity measurements and HDX-MS data on E. coli E1p with intact wild type E2p, we conclude that the crystallographic analyses using the E2p didomain and subsequent interpretation were justified. The E1p-PSBD interactions were consistent with binding of PSBD at the E1p N terminus, and some information regarding general interactions near the active site, not observed in the crystal structure due to disorder, can be obtained from such studies.
The conclusions drawn above were based on the new structural and biochemical results in combination with preexisting knowledge. The structural results for the E1p-E2p didomain subcomplex represent the first example of E1p-E2p binding from a PDHc containing homodimeric ␣ 2 E1p, typically found in prokaryotes. Because homodimeric E1ps are also correlated with overall, octahedral PDHc assemblies having cubic cores, the results are likely to be applicable to most, if not all, other members of this architecture.