Biochemical Journal

Research article

Variation in the organization and subunit composition of the mammalian pyruvate dehydrogenase complex E2/E3BP core assembly

Swetha Vijayakrishnan, Philip Callow, Margaret A. Nutley, Donna P. McGow, David Gilbert, Peter Kropholler, Alan Cooper, Olwyn Byron, J. Gordon Lindsay


Crucial to glucose homoeostasis in humans, the hPDC (human pyruvate dehydrogenase complex) is a massive molecular machine comprising multiple copies of three distinct enzymes (E1–E3) and an accessory subunit, E3BP (E3-binding protein). Its icosahedral E2/E3BP 60-meric ‘core’ provides the central structural and mechanistic framework ensuring favourable E1 and E3 positioning and enzyme co-operativity. Current core models indicate either a 48E2+12E3BP or a 40E2+20E3BP subunit composition. In the present study, we demonstrate clear differences in subunit content and organization between the recombinant hPDC core (rhPDC; 40E2+20E3BP), generated under defined conditions where E3BP is produced in excess, and its native bovine (48E2+12E3BP) counterpart. The results of the present study provide a rational basis for resolving apparent differences between previous models, both obtained using rhE2/E3BP core assemblies where no account was taken of relative E2 and E3BP expression levels. Mathematical modelling predicts that an ‘average’ 48E2+12E3BP core arrangement allows maximum flexibility in assembly, while providing the appropriate balance of bound E1 and E3 enzymes for optimal catalytic efficiency and regulatory fine-tuning. We also show that the rhE2/E3BP and bovine E2/E3BP cores bind E3s with a 2:1 stoichiometry, and propose that mammalian PDC comprises a heterogeneous population of assemblies incorporating a network of E3 (and possibly E1) cross-bridges above the core surface.

  • E3-binding stoichiometry
  • E2/E3BP core organization
  • isothermal titration calorimetry (ITC)
  • pyruvate dehydrogenase complex
  • small-angle neutron scattering (SANS)
  • variable substitution model


The hPDC [human mitochondrial PDC (pyruvate dehydrogenase complex)] is a large macromolecular machine (Mr, ~9.5 × 106), responsible for the conversion of pyruvate into acetyl-CoA and NADH, the key committed step in carbohydrate utilization in humans. It is assembled from multiple copies of four components: pyruvate decarboxylase (E1), dihydrolipoamide acetyltransferase (E2), dihydrolipoamide dehydrogenase (E3) and E3BP (E3-binding protein). Its E2 and E3BP subunits form a 60-meric pentagonal dodecahedral core to which 20–30 α2β2 E1 tetramers and 6–12 E3 homodimers bind tightly but non-covalently. Human E2 and E3BP have similar modular domain organizations comprising outwardly extended N-terminal LDs (lipoyl domains) (two on E2 and one on E3BP), an SBD (subunit-binding domain) and a large CTD (C-terminal domain) that mediates core assembly. Individual domains are joined by flexible linkers rich in alanine and proline allowing the peripherally located LDs to act as ‘swinging arms’, such that their lipoamide cofactors can visit all three active sites in turn during the catalytic cycle. PDC deficiencies have been implicated in a broad range of genetic, metabolic, autoimmune and neurodegenerative disorders [17].

PDCs from prokaryotes lack E3BP, and the crystal structure of the Bacillus stearothermophilus (now Geobacillus stearothermophilus) (PDB code 1B5S) E2-CTD core provides clear insights into its subunit organization, with the basic building blocks, namely E2 trimers, forming the 20 vertices of the dodecahedron [8]. The mammalian E2/E3BP core is also assembled from 20 trimeric units [9]. Interaction of E3BP with E2 occurs co-translationally and is mediated by their respective CTDs.

The structure and subunit composition of E2/E3BP is critical to the efficient functioning of the PDC. Initially it was reported that 12 E3BPs were situated on the 12 pentagonal faces of the E2 assembly in mammalian and yeast PDCs – the so-called ‘addition’ model of core organization (60E2+12E3BP) [10,11]. However, in subsequent 48E2+12E3BP and 40E2+20E3BP ‘substitution’ models, 12 or 20 E3BPs were proposed to replace an equivalent number of E2s within the 60-meric core structure [12,13]. AUC (analytical ultracentrifugation) studies, SAXS (small-angle X-ray scattering) and SANS (small-angle neutron scattering) solution structures combined with cryo-EM (cryo-electron microscopy) reconstructions of rhE2/E3BP (recombinant human E2/E3BP) assemblies reveal the presence of open pentagonal faces, strongly supporting the substitution model [1214].

Current substitution models vary in the trimer population forming the 60-meric core assembly. Thus the 48E2+12E3BP version would be assembled from a mixture of eight E2 homotrimers and twelve 2E2–1E3BP heterotrimers, whereas the 40E2+20E3BP core would be formed exclusively from the latter. The number and distribution of E3BPs within the core is a key issue, as overall geometry and subunit composition will influence core stability, catalytic efficiency and PDK (pyruvate dehydrogenase kinase)-mediated regulation [14].

The stoichiometry of E1 and E3 interactions with the central core has also proved controversial. In B. stearothermophilus PDC, E1 and E3 bind to E2-SBDs with 1:1 stoichiometries in a mutually exclusive fashion [1517], whereas, in eukaryotes, E1 binds to a specific E2-SBD and E3 binds exclusively to an equivalent E3BP-SBD in normal circumstances. Crystal structures and solution studies of bacterial E1–E2-SBD [18] and E3–E2-SBD [19] sub-complexes show that association of a second E2-SBD to either E1 or E3 is prevented by steric hindrance. The crystal structures of rhE3 (recombinant human E3) bound to E3BP-SBD [20,21] as well as ITC (isothermal titration calorimetry) studies of rhE2/E3BP–rhE3 and rhE3BP-SBD–rhE3 interactions [13] suggest a 1:1 stoichiometry [20]. In contrast, a SAXS solution structure of rhE3 complexed with an E3BP-didomain (E3BP-DD, LD plus SBD) indicates a 2:1 (E3BP-DD–E3) relationship [22].

Although considerable structural and biochemical information is available for the individual mammalian PDC components, relatively little is known about the arrangement of its icosahedral E2/E3BP core, including its precise subunit composition, overall organization and stoichiometry of binding to E1 and E3. In the present study, we use a range of biochemical and biophysical approaches to demonstrate clear differences in the subunit compositions of native bovine and rhE2/E3BP cores, and propose a new variable substitution model. Structural and geometric constraints that govern core organization and assembly have also been explored by mathematical modelling, revealing a plausible basis for favouring the 48E2+12E3BP model in situ. Moreover, we show that, at maximal occupancy, the number of E3 homodimers tethered to the assembled E2/E3BP 60-mer is equivalent to 50% of the E3BP subunits present, indicative of a 2E3BP–1E3 stoichiometry and suggesting the presence of a novel architectural feature, namely a network of E3 cross-bridges linking pairs of E3BPs above the core surface.

Part of this work was first presented at the 16th European/UK Analytical Ultracentrifuge Meeting ‘AUC and Hydro’ held at the Newcastle University, Newcastle upon Tyne, U.K., in September 2008.


Protein expression and purification

rhE3 (in pET14b) and rhE2/E3BP (in pET11/pET28b) were overexpressed in Escherichia coli BL21 star (DE3) cells (Invitrogen), grown in LB (Luria–Bertani) medium supplemented with the appropriate antibiotics. Cells were induced at an A600 of 0.6–0.8 with 1 mM IPTG (isopropyl β-D-thiogalactopyranoside) for 4 h at 30 °C (rhE3) or 16 h at 15 °C (rhE2/E3BP). rhE2/E3BP cultures were supplemented with 100 μg/ml lipoic acid. Cells were harvested (10000 g for 15 min at 4 °C) and overexpression was assessed by SDS/PAGE. Cell pellets [in 0.1 M NaCl, 0.01 M imidazole and 50 mM potassium phosphate buffer (pH 8.0) supplemented with EDTA-free protease inhibitors (Roche), DNase I (Sigma) and Halt protease inhibitor cocktail (Thermo Scientific)] were lysed by French press treatment and clarified by centrifugation (10000 g for 15 min at 4 °C). Proteins were purified by zinc-chelate chromatography on a BioCAD 700E Workstation by virtue of their N-terminal His tags. In the case of the purified rhE2/E3BP assembly, only E3BP-contained a His tag, so isolation of intact rhE2/E3BP core required co-integration of E2 and E3BP. Further purification involved either anion exchange on a high-capacity 20HQ column (rhE2/E3BP) for selective DNA removal (Applied Biosystems) and/or gel filtration (rhE3, rhE2/E3BP) on a Sephacryl S-300 column (Amersham). Excess unincorporated, monomeric E3BP was removed from the high Mr E2/E3BP core during the gel-filtration step. Purified enzymes in PEBS100 buffer [2 mM EDTA, 0.01% sodium azide, 0.1 M NaCl and 50 mM potassium phosphate (pH 7.5)] were analysed by SDS/PAGE. A molar absorption coefficient of 11300 M−1·cm−1 at 450 nm was used for FAD and 1951320 M−1·cm−1 at 280 nm calculated for the rhE2/E3BP assembly using ProtParam (

dE3 (deuterated E3) was overexpressed as follows: plasmid encoding rhE3 was transformed into BL21 (DE3) cells and overexpression was carried out in Enfors minimal medium (containing 85% 2H2O and hydrogenated glycerol as the carbon source) at 30 °C to an A600 of 15 using an Enfors fermentation system. dE3 was purified as described above for hydrogenated rhE3 and purity assessed by SDS/PAGE.

Bovine heart PDC was purified as described previously [23] with slight modifications. PDC was kept at 4 °C in 2 mM EDTA, 0.01% sodium azide, 0.5% Triton X-100 and 50 mM potassium phosphate buffer (pH 7.4). Subsequently, bovine E2/E3BP core (termed bE2/E3BP) was purified on discontinuous sucrose gradients [24] and its concentration measured using the biuret method.

Purified enzymes were exchanged into PEBS100 buffer (or PEBS100 with 0.5% Triton X-100) and used within 1–2 days.

SV (sedimentation velocity) AUC

SV experiments were performed using an Optima XL-I analytical ultracentrifuge (Beckman Coulter) and an An-50 Ti rotor. rhE2/E3BP complexes were maintained at 191.5 nM and the rhE3 concentration varied to achieve E3BP to E3 ratios from 4:1 (48E2/12E3BP to 3E3) to 1:4 (48E2/12E3BP to 48E3). Samples (360 μl) in PEBS100 buffer were loaded into 12 mm double-sector centrepieces and centrifuged at 20000 rev./min at 4 °C. Sedimentation profiles acquired at 1 min intervals with Rayleigh interference optics were analysed using c(s) analysis in SEDFIT [25]. The partial specific volumes of rhE2/E3BP and rhE3, calculated using SEDNTERP [26], were 0.744 and 0.739 ml/g respectively at 20 °C.

GFC (gel-filtration chromatography)

rhE2/E3BP–rhE3 samples were prepared by maintaining rhE2/E3BP at 0.42 mM and varying rhE3 levels to satisfy rhE2/E3BP to rhE3 ratios ranging from 3:1 (48E2/12E3BP to 4E3) to 1:3 (48E2/12E3BP to 36E3). Samples were analysed by Sephacryl S-300 gel filtration with uncomplexed rhE2/E3BP and rhE3 included as controls.


SANS was conducted on beamline D22 of the ILL (Institut Laue-Langevin), Grenoble, at detector distances of 4 and 14 m. Transmission and scattering data (buffer and sample) were recorded for 4 min and 15 min respectively at 4 °C in 1 mm quartz cuvettes, covering an overall momentum transfer (Q) range 0.0034<Q<0.143 Å−1 (1 Å=0.1 nm). The response of the two-dimensional area gas detector was calibrated against water. Raw data were analysed using the program GRASansP written by Charles Dewhurst, ILL ( The p(r) distance distribution, maximum dimension (Dmax) and radius of gyration (Rg) were obtained using the programs PRIMUS [27] and GNOM [28,29].

[14C]NEM (N-ethyl-1-[14C]maleimide) radiolabelling and counting

Purified PDC (40, 60 and 120 μg) and rhE2/E3BP (20 and 40 μg) complexed with rE3 (at stoichiometries of 2:1, 10:1, 20:1 and 40:1) suspended in PEBS100 buffer were radiolabelled with 0.65 mM [14C]NEM (33.2 mCi/mmol, PerkinElmer) and incubated for 30 min at room temperature (18 °C). All samples were subsequently mixed with NAD+ (0.5 mM) or NADH (1 mM) followed by incubation for 10 min at room temperature. Reactions were terminated by the addition of 0.05 M DTT (dithiothreitol) after which SDS-loading dye was added. Labelled samples were assessed by SDS/PAGE on 8% (w/v) slab gels (17 cm×15 cm×2 mm) at 400 V, 65 mA. Gels were stained and destained, and the radioactive bands were excised and dissolved overnight in SOLVABLE (PerkinElmer). Gel pieces from blank tracks were used to estimate background radioactivity. Solubilized samples were mixed with EcoScint A (National Diagnostics) and counted (10 min/sample) in a Beckman LS 6500 scintillation counter. Four bovine PDC and four rhE2/E3BP:E3 preparations were separately radiolabelled. Counts for each were determined in triplicate. Data were analysed statistically as follows [30]: Embedded Image Embedded Image where CE2 and CE3BP, and σE2 and σE3BP represent the [14C] counts and associated errors determined for radiolabelled rhE2 and rhE3BP respectively.


To eliminate background labelling of accessible cysteine residues as a possible source of error in the [14C]NEM radiolabelling study, fluorographic analysis of rhE2/E3BP–rhE3 was undertaken. Samples of radiolabelled rhE2/E3BP (6 μg) complexed with rhE3 at a stoichiometric ratio of 10:1 were run on SDS/PAGE (8% gels), then stained and fixed overnight. Fluorography was performed as described by Chamberlain [31]. Fixed gels were immersed in Amplify (GE Healthcare) and incubated for 30 min prior to drying under vacuum at 80 °C for 2 h. Fluorographs were stored at −80 °C and developed on X-ray film (Kodak) after 4–7 days.

Mathematical modelling

The algorithm for modelling the 60-meric core organization was based on its assembly from 20 trimeric units employing two important constraints: only one E3BP was allowed per trimer and E3BP was allowed to interact only with itself. On this basis, core subunit arrangements ranging from 60E2+0E3BP to a maximum of 40E2+20E3BP can be theoretically envisaged. Dodecahedral faces were parameterized by [X,Y,Z] (the number of pentagonal dodecahedral core faces that incorporate 3, 4 or 5 pairs of E2s respectively). The resulting set of simultaneous linear constraint equations was solved to obtain a list of solutions for the number of 3-, 4- and 5-E2 pair faces at each E2+E3BP subunit stoichiometry. The equations were implemented as a finite domain constraint logic program in Sicstus Prolog (, using its in-built constraint solver to generate the solutions.


Purified proteins were dialysed at 4 °C in PEBS100 buffer (rhE2/E3BP and rhE3) or PEBS100 plus 0.5% Triton X-100 (bE2/E3BP and rhE3). ITC measurements were conducted on a MicroCal VP-ITC titration microcalorimeter at 20 °C [32]. Analysis and curve fitting to the binding isotherms using MicroCal ORIGIN v7.0 gave values for the association constant (Ka), the molar ratio (n) and the change in enthalpy (ΔH) for protein–protein interactions.


Titration of the rhE2/E3BP core assembly with increasing amounts of rhE3 indicates saturation at a 2:1 stoichiometry

Initial GFC and AUC studies were conducted to assess the number of E3 dimers tethered to the rhE2/E3BP core at maximal occupancy as a means of assessing binding stoichiometry and E3BP content. The GFC profiles of a series of rhE2/E3BP–rhE3 mixtures (based on the 48E2+12E3BP model) covering ratios from 48E2/12E3BP to 4E3 (3:1) to 48E2/12E3BP to 36E3 (1:3) are shown in Figure 1(A). The Kd values for rhE3BP–rhE3 and E3BP-DD–E3 sub-complex formation are reported to be 6.3 nM and 35.7 nM respectively [24,33]. E3 and E3BP concentrations used in the present study were in the low micromolar range, ensuring tight and complete binding.

Figure 1 Binding stoichiometry of rhE3 to the rhE2/E3BP core assembly determined by GFC and AUC

(A) rhE2/E3BP–rhE3 samples, prepared by maintaining rhE2/E3BP at 421 nM and varying rhE3 levels to satisfy ratios ranging from 3:1 (48E2/12E3BP to 4E3) to 1:3 (48E2/12E3BP to 36E3), were analysed on a Sephacryl S-300 column (120 ml bed volume) with uncomplexed rhE2/E3BP and rhE3 as controls. Free rhE3 is observed from a ratio of 1:1 onwards, consistent with 2:1 E3BP–E3 binding. (B) For SV AUC, samples were maintained at 191.5 nM and the rhE3 concentration was varied to achieve ratios from 4:1 (48E2/12E3BP to 3E3) to 1:4 (48E2/12E3BP to 48E3). c(s) analysis of the data again indicates a 2:1 (E3BP–E3) relationship as the low s peak corresponding to free E3 is apparent only at rhE2/E3BP to rhE3 ratios of 1:1 and below.

Void volume peaks (38 ml) for the high Mr rhE2/E3BP core uncomplexed and complexed with rhE3 were apparent at all ratios, whereas free rhE3 (Ve of 65 ml) was evident only in the range from 1:1 to 1:3, i.e. 12–36 E3 dimers per 48E2/12E3BP core. SDS/PAGE analysis confirmed the presence of free E3 at these ratios (Supplementary Figure S1 at These estimates of maximal E3 binding to assembled E2/E3BP core indicated that saturation was achieved prior to addition of 12 E3BPs per core and supported our previous data indicating 2:1 binding between a free (monomeric) truncated version of E3BP (LD plus adjacent SBD) and E3 [22].

SV AUC analysis was also conducted for a similar series of rhE2/E3BP–rhE3 mixtures, and resultant profiles were analysed using the c(s) model in SEDFIT [25] (Figure 1B). s20,w0 for free rhE3 was 5.8 S, in agreement with previous data [22]. An increase in sedimentation coefficient for the rhE2/E3BP core from 29.3 to 35.3 S was observed on rhE3 binding. Moreover, uncomplexed rhE3 became detectable at a 1:1 stoichiometry (12 E3 dimers per 48E2/12E3BP core), again consistent with 2:1 binding. GFC and AUC studies were designed on the basis of the 48E2+12E3BP substitution model; however, the presence of excess rhE3 at 12 E3 dimers per core is also consistent with a 40E2+20E3BP model with rhE3 saturation occurring at 10 E3s for a 2:1 stoichiometry in this case.

Ten rhE3s bind to the rhE2/E3BP core

To establish the binding stoichiometry and assess the rhE2/E3BP core composition more precisely, rhE2/E3BP was reconstituted with increasing amounts of dE3 and complex formation was monitored by SANS. Given the high molecular mass of the rhE2/E3BP core assembly (~3.5 MDa) compared with E3 (106 kDa), accurate titration measurements of hydrogenated E3 binding using SAXS would prove difficult. By exploiting neutron scattering, it was possible to increase the E3 signal by deuteration, markedly enhancing the sensitivity of stoichiometry measurements. Scattering curves for rhE2/E3BP and rhE2/E3BP–dE3 obtained at various levels of dE3 saturation are shown in Figure 2(A). Intensity increases, particularly in the low angle region, were clearly evident on dE3 addition, over a stoichiometry range from 0 (uncomplexed rhE2/E3BP) to 18 dE3 dimers per core. As expected, dE3 attachment to rhE2/E3BP resulted in an increased radius of gyration, Rg, and scattering intensity at zero angle, I(0) (Table 1). Moreover, dE3 addition yielded a maximum Rg of 162 Å (1 Å=0.1 nm) for the E2/E3BP–E3 sub-complex, in agreement with a previous report [12]. The I(0) value, normalized for concentration and plotted against the number of dE3s present per core is shown in Figure 2(B). Importantly, at ten dE3 dimers bound per core, the gradient of the line changes markedly, indicating saturation.

Figure 2 Binding stoichiometry of dE3 to rhE2/E3BP measured by SANS

(A) SANS curves were acquired for protonated rhE2/E3BP complexed with dE3 at ratios ranging from 0 to 18 E3 dimers per core. The rhE2/E3BP–dE3 complex was formed initially by maintaining rhE2/E3BP at 1.04 μM prior to dE3 addition at the required molarity to give an E2/E3BP+E3 subunit composition of 48/12+2. Thereafter, equivalent amounts of dE3 were added stepwise, covering the stoichiometry range from 48/12+2 to 48/12+18. Free rhE2/E3BP and dE3 were used as controls. (B) I(0) normalized for concentration and plotted against the number of dE3 per core indicates saturation at a ratio of 10:1 (dE3 to rhE2/E3BP).

View this table:
Table 1 Rg determined by SANS from rhE2/E3BP–dE3 samples

I(0) represents intensity at zero scattering angle.

The observed saturation value is close to the published value of 12:1 [12] and could be interpreted as evidence for 1:1 binding assuming the 48E2+12E3BP substitution model. On repeated measurement, however, saturation was always obtained at exactly ten dE3 dimers per rhE2/E3BP core. Furthermore, possible dE3 aggregation was eliminated as a source of error. Additionally, the Rg (34 Å) and Dmax (130 Å) values for dE3 (Supplementary Figure S2 at compared well with published data for hydrogenated E3 (38 Å and 130 Å), indicating no major structural changes on E3 deuteration [22]. A marked increase in particle diameter, Dmax, from 462 Å to 520 Å, was also evident from the rhE2/E3BP p(r) distribution function on dE3 saturation (Supplementary Figure S3 at

Since E3 interactions with the intact E2/E3BP scaffold will be governed by core geometry, such stoichiometry measurements are more physiologically relevant than previous analyses of E3 binding to monomeric E3BP-SBD or E3-DD constructs [13,1922]. However, a 1:1 stoichiometry is observed in crystal structures of E3-DD and E3-SBD sub-complexes [1921] where steric hindrance prevents access to a putative second binding site located near the two-fold axis of symmetry. It is possible that crystal packing could account for this key difference, implying that the E3 crystal structure differs significantly from its structure in solution. In this regard, the Dmax for protonated or deuterated E3 (determined by SAXS and SANS) is 130 Å [22] (Supplementary Figure S2) compared with 100 Å in the crystal structure, indicating that E3 adopts a more open and dynamic conformation in solution. Thus a plausible explanation for the SANS-based estimate of ten E3-binding sites per core is that it represents 2:1 binding to a 40E2+20E3BP assembly.

Distinctive subunit organization of the rhE2/E3BP and bE2/E3BP cores

As already indicated, there are currently two conflicting 48E2+12E3BP and 40E2+20E3BP substitution models for rhPDC core subunit content and organization in which 12 or 20 E3BPs are proposed to replace an equivalent number of E2s within the 60-meric assembly [12,13]. Importantly, evidence for both of these models was obtained with rhE2/E3BP cores produced using similar recombinant vectors, overexpression systems and protein purification regimes. However, no results were provided on the relative levels of E2 and E3BP expression in either case. Our present study relies on the presence of a His-tag on E3BP for the isolation of the integrated rhE2/E3BP core. Excess monomeric E3BP is always present after the initial purification step prior to its removal by gel filtration, ensuring maximal E3BP incorporation. Interestingly, however, previous studies on native bovine heart and yeast PDCs all report the presence of only 12 E3BP subunits per core (i.e. 48E2+12E3BP) [10,11]. Importantly, both models yield predictable and distinct numbers of bound E3s at saturation (Supplementary Table S1 at

In view of these conflicting reports, NADH-induced E2 and E3BP lipoamide group radiolabelling was performed to assess possible differences between native bovine (bE2/E3BP) and human (rhE2/E3BP) cores. Specific [14C]NEM modification of reduced E2 and E3BP lipoamide thiols is achieved in the presence of NADH and catalytic amounts of rhE3. The ratio of [14C]NEM incorporated into E2 (two LDs) and E3BP (one LD) subunits alters significantly depending on the model in question (Table 2). In particular, a marked difference in ratios is predicted for the 48E2+12E3BP (8:1) [12] and 40E2+20E3BP substitution models (4:1) [13], whereas potential variations in ratio for E2+E3BP core compositions in the range 44E2+16E3BP (5.5:1) to 40E2+20E3BP (4:1) are relatively minor (Table 2). Radiolabelling was conducted for triplicate samples of four bovine heart PDC and rhE2/E3BP–rhE3 preparations, to prevent uncertainties owing to batch variation or sample degradation.

View this table:
Table 2 Core stoichiometries and their theoretical 14C-radiolabelled E2 to E3BP ratios

Ratio analysis of radiolabelled subunits in rhE2/E3BP and native bE2/E3BP cores is shown in Table 3. Values for the bE2/E3BP core supported the 48E2+12E3BP model (8:1, Table 2), as reported previously [12], whereas rhE2/E3BP cores generally gave values (4:1) consistent with the 40E2+20E3BP model [13]. Occasional higher values were obtained in the latter case with 5.63, the highest ratio observed, equivalent to a 44E2+16E3BP composition. In summary, radiolabelling data suggested distinctive core models with average compositions of 48E2+12E3BP for bE2/E3BP, and approaching 40E2+20E3BP for rhE2/E3BP.

View this table:
Table 3 Determination of subunit composition of bovine and recombinant E2/E3BP cores via [14C]NEM radiolabelling

Subscripts 1–4 denote different protein preparations used for the experiment. Average±S.E.M. counts of triplicate samples are shown.

To eliminate background labelling of accessible cysteine residues as a possible source of error, fluorographic analysis of rhE2/E3BP–rhE3 (Supplementary Figure S4 at was conducted after modification with [14C]NEM in the presence of NAD+ or NADH to maintain the E2 and E3BP-linked lipoamide moieties in their oxidized or reduced states. Specific NADH-induced modification of E2 and E3BP was observed with no discernable labelling of complexed rhE3. However, radiolabelling of accessible E3 cysteine residues, possibly the active-site disulfide pair, was detectable with non-complexed E3 (Supplementary Figure S4).

Previous studies have established specific modification of E2 and E3BP-reduced lipoamide moieties in bovine PDC [34]. Under these conditions, estimates of the rhE2/E3BP core composition obtained by specific NADH-induced [14C]NEM labelling of E2 and E3BP lipoamide thiols are in agreement with the 40E2+20E3BP model. Brautigam et al. [13] provided several lines of evidence for a 40E2:20E3BP core stoichiometry using quantitative Coomassie Blue staining, AUC, native gel electrophoresis and ITC measurements. They also concluded that previous AUC analyses of maximal E1 and E3 binding to the high Mr E2/E3BP core leading to the development of the original 48E2+12 E3BP model by Hiromasa et al. [12] were technically flawed. However, neither group took account of the relative levels of E2 and E3BP expression in their systems. This is likely to prove a key factor in determining the final core composition, particularly as E2 itself is capable of self-assembly into a 60-meric icosahedron, whereas E3BP is largely monomeric (see Figure 4) and requires to be co-expressed with E2 to achieve core integration. In this context, it is apparent that generation of recombinant cores with various E2/E3BP subunit contents is a reflection of the relative E2 and E3BP expression levels.

A clear difference in subunit organization and composition between recombinant human and native bovine cores was also confirmed by ITC. Binding isotherms were generated by titrating rhE3 into the reaction cell containing rhE2/E3BP (Figure 3A) or bE2/E3BP (Figure 3B). On curve fitting, molar ratios (n, E3BP to E3) of 9.68±0.18 and 5.70±0.09 were obtained for rhE2/E3BP to rhE3 and bE2/E3BP to rhE3 respectively. The E3 titration profile for rhE2/E3BP and bE2/E3BP yielded Kd values with an upper limit of 7 nM and 10±5 nM respectively, consistent with previously reported values [24,33], although binding was too tight for accurate measurement. Three independent binding isotherms for rhE2/E3BP–rhE3 and bE2/E3BP–rhE3 complex formation yielded n values within 10% of 9.68 and 5.70 respectively, consistent with 2:1 E3BP to E3 binding molar ratios predicted (Supplementary Table S1 at for the recombinant 40E2+20E3BP and bovine 48E2+12E3BP models respectively. The recombinant E3–E2/E3BP complex molar ratio (9.68) was also consistent with our SANS data (ten E3 dimers per core).

Figure 3 Stoichiometry of binding of rhE3 to rhE2/E3BP and bE2/E3BP core assemblies determined by ITC

ITC measurements of rhE3 binding to (A) rhE2/E3BP and (B) bE2/E3BP cores are shown. Typically, a 1 μl pre-injection was followed by 28 injections (10 μl, at 3 min intervals) of 210 μM or 23 μM rhE3 into the reaction cell containing 1.4 ml of rhE2/E3BP (1.09 μM) or bE2/E3BP (0.34 μM) respectively. Corrections for heats of ligand and protein dilution were obtained by injecting rhE3 into buffer and buffer into protein respectively. The top panels represent changes in binding enthalpy that arise from titrating E3 into E2/E3BP. The bottom panels show the binding isotherm obtained by integrating the heat of binding at every injection, to which the best fit is obtained using ORIGIN version 7.0. Distinct molar ratios of 9.68 and 5.70 were obtained for rhE3 bound to rhE2/E3BP and bE2/E3BP cores respectively.

A problematic issue is that our present ITC and previous biochemical and biophysical analyses [22], all indicating a 2:1 stoichiometry of E3 binding to E3BP using either truncated E3BP constructs or the fully assembled E2/E3BP core, are in direct contrast with the findings of Brautigam et al. [13]. In our case, ITC evaluation of E3 binding to rhE2/E3BP shows saturation at n=10 E3 dimers per core, with a maximal Kd of 7 nM. In contrast, Brautigam et al. [13] report a much higher Kd (102 nM) that may account for their high n value of 18.8, implying 1:1 E3 binding to a 40E2+20E3BP assembly. In this context, it is noteworthy that PDC-deficient patients totally lacking E3BP retain partial PDC activity (10–20% of controls), suggesting that E3 may still interact with the E2 core to a limited extent in the absence of its normal binding component. In support of this idea, bovine PDC reconstituted in vitro with a homogeneous 60-meric E2 core is virtually inactive at stoichiometric E3 levels, whereas overall complex activity (approaching 40–50% of wild-type) can be restored with a 200-fold E3 excess, suggesting that E3 can interact weakly with the normal E1-binding site on E2 under these conditions [35]. Recent surface plasmon resonance data from our laboratory (G. Singh, A. E. Brown, S. D. Richman, Z. Cao, H. Al-Amodi, S. M. Kelly, D. McGow, O. Byron, G. K. Brown and J. G. Lindsay, unpublished work) confirm that E3 has a low-affinity interaction with a secondary binding site on E2 (Kd≈410 nM) that can lead to partial E1 displacement at markedly elevated E3 concentrations. This Kd value is only 4-fold higher than the Kd of 102 nM for E3–E3BP complex formation [13], and may account for the additional E3 binding required to achieve saturation in their experiments leading to an erroneously high estimate of E3 binding stoichiometry. In our case, E3 interacts with its primary E3BP-binding site with a much lower Kd (7 nM or less), i.e. with at least a 60-fold greater affinity than its weak secondary interaction with E2. It is difficult to assess the reasons underlying these marked differences in E3-binding affinity for the rhE2/E3BP core without a detailed knowledge of the sequences of the E2, E3BP and E3 constructs and recombinant expression system used by Brautigam et al. [13]. Inaccurate protein measurement is another possible source of error in these experiments. In our case, the concentration of purified E2/E3BP and E3 samples was estimated by two independent methods. For E2/E3BP, this involved A280 measurement based on the calculated molar absorption coefficient of the core assembly and biuret analysis. Possible DNA contamination of purified rhE2/E3BP was eliminated by use of an ion-exchange step. The E3 concentration was also estimated by the biuret method and by A450 measurement of FAD content after ensuring that the protein contained a full complement of cofactor.

E2/E3BP core composition: investigation by mathematical modelling

Mathematical modelling of possible E2/E3BP core arrangements was conducted by systematic variation of E2 homotrimer and 2E2–1E3BP heterotrimer populations that form the 60-meric icosahedron (Supplementary Table S2 at Two fundamental constraints were applied to the modelling process: first, only one E3BP was allowed per heterotrimer, and secondly, E3BPs of neighbouring heterotrimers could interact only with each other. The first constraint recognizes that E3BP is itself incapable of core formation with no evidence for the existence of 1E2–2E3BP heterotrimers. Thus, despite high rhE3BP expression relative to E2, purified rhE2/E3BP always contains E2 as the major species, as judged by SDS/PAGE and radiolabelling (Supplementary Figures S1 and S4). Justification for the second constraint is that E3BP lacks key residues involved in the hydrophobic ‘ball and socket’ connection that mediates inter-trimer contacts within the core [12]. It has been suggested that E3BP could self-associate on the core surface [12], and in our earlier AUC studies [14], purified rhE3BP showed little or no capacity to form higher oligomeric states. As shown in Figure 4, non-specific cross-linking of purified rhE3BP with glutaraldehyde revealed a weak tendency to form dimers, although the majority of rhE3BP remained monomeric. Successful cross-linking of rhE3, a naturally occurring homodimer, served as a positive control.

Figure 4 Glutaraldehyde cross-linking of E3BP

SDS/PAGE of full-length rhE3BP before (−) or after (+) chemically cross-linking with 2% (v/v) glutaraldehyde indicates weak covalent dimerization. Homodimeric rhE3 was used as a control. Molecular masses of markers (lane Mw) are shown in kDa. Monomeric rhE3BP runs at a lower apparent subunit molecular mass after glutaraldehyde treatment because intramolecular cross-links produce a more compact species with a higher mobility in the gel.

Several solutions of core organization for each model were obtained by mathematical modelling (Table 4). Each represents an E2 and E3BP arrangement that satisfies a particular core composition ranging from 60E2+0E3BP to 40E2+20E3BP. Final states were parameterized by [X,Y,Z] (the number of pentagonal dodecahedral core faces that incorporate 3, 4 or 5 pairs of E2s respectively). The number of possible final states follows a near Gaussian distribution (Figure 5A). Importantly, whereas the 40E2+20E3BP model has only two final states (one symmetric and one asymmetric; Figure 5B), the number of possible states peaks at seven for the 48E2+12E3BP core (Figure 5C), with each state showing a distinctive distribution of E3BPs around the core surface. Thus a 48E2+12E3BP subunit stoichiometry may be advantageous in that it ensures the greatest number of routes to, and final arrangements of, the core structure, enabling rapid and efficient production in vivo. Moreover, the relative abundance and distribution of E2 and E3BP subunits will have a direct effect on the number and location of associated E1 and E3 enzymes. Thus, from a functional and regulatory standpoint, a 48E2+12E3BP core composition may be favoured in that it recruits the optimal balance of cognate E1s and E3s for achieving optimal catalytic efficiency and regulatory fine-tuning. E1 normally catalyses the rate-limiting step in the reaction sequence, so alterations in bound E1 levels will directly affect the overall catalytic rate in most circumstances. Interestingly, a reconstituted bE2/E3BP core containing only 35% of the normal E3BP complement supports a correspondingly reduced level of PDC activity, indicating that the ratio of bound E1s and E3s (as determined by the E2/E3BP core composition) is also important in regulating the overall rate of catalysis [35].

View this table:
Table 4 Plausible E2/E3BP core organizations, corresponding numbers of homo- and hetero-trimers and the number of final states/solutions
Figure 5 Mathematical modelling of E2/E3BP core organization

(A) Mathematical modelling of the final states for the different core models follows a near Gaussian distribution, peaking at the 48E2+12E3BP core composition. (B and C) Various arrangements of E2 (black) and E3BP (orange) on the core faces are shown. In square brackets above each diagram ([X,Y,Z]) are the numbers of core faces that incorporate 3, 4 or 5 (indicated by numbers on the faces) pairs of E2s respectively, for core models of (B) 40E2+20E3BP and (C) 48E2+12E3BP. The schematic is drawn as a 2D diagram looking along the five-fold axis of symmetry; the unseen face is denoted by the line extending from the bottom right of each model.


Variable subunit composition of the rhE2/E3BP core assembly

Based on current evidence, it is apparent that there is considerable potential for variation in both overall E2/E3BP core organization and subunit stoichiometry. In our case, rhE2 and E3BP co-expression is carried out under conditions where E3BP is produced in excess leading to formation of a 40E2+20E3BP core in support of the hypothesis that only one E3BP subunit can be incorporated per trimer. We also demonstrate for the first time that, in contrast with E2, rhE3BP was incapable of independent self-assembly into a high Mr 60-meric complex. Indeed, it is largely monomeric with a weak tendency to form dimers (Figure 4). In this context, we have previously reported that the stability of the rhE2/E3BP core was reduced in comparison with a homogeneous 60-meric E2 core, indicating that the E3BP content was likely to influence core ‘breathing’ and protein dynamics [14,37].

Advantages in maintaining a standard 48E2+12E3BP core structure in native eukaryotic PDCs

Although current evidence lends support to the 40E2+20E3BP model for the rhE2/E3BP core when E3BP is present in excess, we also show that the native bovine assembly has a distinctive 48E2+12E3BP subunit stoichiometry. Interestingly, in vivo, E3BP subunit levels per core have always been estimated at 12 in mammalian and yeast PDCs [10,11], so it will be essential in future to determine whether the 48E2+12E3BP core represents the standard archetypal assembly that has been widely adopted in eukaryotes or whether there are characteristic tissue- or species-specific alterations.

As a central 48E2+12E3BP core is composed of two types of trimeric units, namely E2 homotrimers and 2E2/1E3BP heterotrimers, E2 and E3BP synthesis must be tightly co-ordinated in vivo, with E3BP being produced in limiting amounts leading to formation of non-uniform cores averaging 12E3BPs per icosahedron. Electron microscopy studies of PDC have highlighted its inherent heterogeneity [3739], so it is also probable that a mixed population of E2/E3BP assemblies is generated in vivo such that the 48E2+12E3BP model represents an ‘average’ structure. As illustrated in Figure 5, even assuming that all cores have a uniform 48E2+12E3BP subunit composition, they can still be arranged in seven possible final states differing in the distribution of E2 and E3BP subunits around the core surface and consequently the spatial organization of its interacting E1 and E3 partners.

In summary, on the basis of limited stoichiometry data available to date, the 48E2+12E3BP model appears to represent the standard archetypal PDC core assembly found in eukaryotes, although there is undoubtedly potential for significant variation in the subunit content of this central structural framework. Mathematical modelling and biochemical data on E3BP-deficient core assemblies also give important insights into potential advantages of deploying a 48E2+12E3BP core composition. In future, it will be essential to determine whether there are species- or tissue-specific variations in the native E2/E3BP subunit content which has the potential for modulating overall PDC composition leading to subtle alterations in catalytic efficiency and regulatory fine-tuning.

Stoichiometry of E3 interaction with the recombinant human and native bovine E2/E3BP assemblies

Our current results using GFC, AUC, SANS and ITC confirm that ten and six E3s respectively bind to individual recombinant human and native bovine E2/E3BP cores at maximal occupancy. These data highlight their distinctive 40E2+20E3BP and 48E2+12E3BP compositions while confirming the 2:1 stoichiometry of E3BP–E3 binding to the surface of these large pentagonal dodecahedral assemblies. It should be noted, however, that final rhPDC core composition is governed by the relative levels of E2 and E3BP expression. Thus the production of a 40E2+20E3BP assembly in recombinant human systems merely reflects one possible scenario, namely where there is excess production of the E3BP subunit (which is never detected in situ as an independent non-integrated species) thereby ensuring its maximal incorporation into the assembling E2 core.

A schematic illustration of an archetypal 48E2+12E3BP core assembly is shown in Figure 6 with representative E2 and E3BP N-terminal segments, consisting of LDs and SBDs joined by flexible linkers, extending outwards from the core. E3BP content and distribution may vary considerably, so the core would be more accurately depicted as a heterogeneous population of closely related assemblies with an average 48E2+12E3BP composition. Representative E3 cross-bridges linking adjacent pairs of E3BPs are also illustrated. There is an additional possibility of E1 heterotetramers linking neighbouring E2-SBDs highlighting the existence of a network of cross-bridges on the core surface (Figure 6 and Supplementary Movie S1 at This architectural feature (i) has the potential to moderate the flexibility of the E2 and E3BP lipoyl ‘swinging arms’ and (ii) may facilitate PDK movements around the core. Interestingly, cryo-EM images of rhE2/E3BP reveal the appearance of prominent E3-induced ‘spikes’ located just above the surface of the CTD. The increased order in this region permits it to be properly resolved on cryo-EM reconstruction, and is consistent with E3 cross-bridge formation restricting movement of the flexible lipoyl ‘swinging arms’ in this region [14].

Figure 6 Cross-bridge formation in human PDC

Cylindrical domain model of hPDC, in which E1 heterotetramers (orange) form cross-bridges between E2 monomers (blue), and E3 dimers (green) form cross-bridges between E3BP monomers (grey). The LDs and SBDs of E2 (dark- and mid-blue respectively) and E3BP (dark- and mid-grey respectively) are connected by linkers (light blue and grey). The specific model illustrated corresponds to the [0,12,0] 48E2+12E3BP arrangement shown in Figure 4(C). Views are along the (A) two-fold, (B) three-fold and (C) five-fold axes of symmetry. The Figure was drawn using the Google SketchUp Tool (


Swetha Vijayakrishnan designed and performed the experiments, analysed the data and wrote the manuscript. Philip Callow carried out the SANS experiment and analysis. Matthew Nutley performed the ITC experiment and analysis. Donna McGow provided technical support. Mathematical modelling and data analysis was carried out by David Gilbert and Peter Kropholler. Alan Cooper provided conceptual advice on the ITC. Olwyn Byron and Gordon Lindsay supervised the project, designed the experiments and edited the manuscript before submission.


This work was partly supported by the Engineering and Physical Sciences Research Council (EPSRC) [grant numbers GR/R99393/01, EP/C015452/1]. S.V. was the recipient of a Wellcome Trust-funded Ph.D. studentship and a Wellcome Trust VIP award. J.G.L. acknowledges continued financial support from BBSRC (Biotechnology and Biological Sciences Research Council), U.K. and the UMDF (United Mitochondrial Disease Foundation), U.S.A.. This research has also benefited from the activities of the DLAB consortium funded by the EU under contracts HPRI-2001-50065 and RII3-CT-2003-505925.


We thank the staff of the ILL-EMBL Deuteration Laboratory for their assistance during sample preparation, the ILL for use of the D22 diffractometer and the EPSRC (Engineering and Physical Sciences Research Council). We thank Michael Haertlein, Peter Timmins and Trevor Forsyth for advice, support and encouragement.

Abbreviations: AUC, analytical ultracentrifugation; bE2/E3BP, bovine, E2/E3BP; [14C]NEM, N-ethyl-1-[14C]maleimide; cryo-EM, cryo-electron microscopy; CTD, C-terminal domain; DD, didomain; dE3, deuterated E3; E3BP, E3-binding protein; GFC, gel-filtration chromatography; ITC, isothermal titration calorimetry; LD, lipoyl domain; PDC, pyruvate dehydrogenase complex; hPDC, human PDC; PDK, pyruvate dehydrogenase kinase; rhE2/E3BP, recombinant human E2/E3BP; rhE3, recombinant human E3; SANS, small-angle neutron scattering; SAXS, small-angle X-ray scattering; SBD, subunit-binding domain; SV, sedimentation velocity


View Abstract