Interaction of the E2 and E3 components of the pyruvate dehydrogenase multienzyme complex of Bacillus stearothermophilus Use of a truncated protein domain in NMR spectroscopy Mark D. Allen, R. William Broadhurst, Robert G. Solomon and Richard N. Perham Cambridge Centre for Molecular Recognition, Department of Biochemistry, University of Cambridge, UK The 2-oxo acid dehydrogenase complexes consist of multiple copies of three distinct enzymes that together catalyse the oxidative decarboxylation of 2-oxo acids, in the presence of thiamine diphosphate (ThDP), coen- zyme A (CoA), Mg 2+ and NAD + , to generate CO 2 and the corresponding acyl-CoA. The complexes are assembled around an oligomeric [octahedral (24-mer) or icosahedral (60-mer)] dihydrolipoyl acyltransferase (E2) core to which multiple copies of the relevant 2-oxo acid decarboxylase (E1) and dihydrolipoyl dehy- drogenase (E3) bind tightly, but noncovalently, to form the intact multienzyme complexes [1]. The pyru- vate dehydrogenase (PDH) complex has a pivotal role in most organisms, catalysing the irreversible reaction that links the glycolytic pathway and the tricarboxylic acid cycle. Pyruvate is oxidatively decarboxylated to acetyl-CoA, which can be either broken down further in the tricarboxylic acid cycle or used as an important Keywords pyruvate dehydrogenase; protein–protein interaction; NMR spectroscopy; multienzyme complex; protein domains Correspondence R.N. Perham, Department of Biochemistry, University of Cambridge, Sanger Building, Old Addenbrooke’s Site, 80 Tennis Court Road, Cambridge CB2 1GA, UK Fax: +44 1223 338707 Tel: +44 1223 338635 E-mail: r.n.perham@joh.cam.ac.uk (Received 19 July 2004, revised 27 September 2004, accepted 28 September 2004) doi:10.1111/j.1432-1033.2004.04405.x A 15 N-labelled peripheral-subunit binding domain (PSBD) of the dihydro- lipoyl acetyltransferase (E2p) and the dimer of a solubilized interface domain (E3int) derived from the dihydrolipoyl dehydrogenase (E3) were used to investigate the basis of the interaction of E2p with E3 in the assem- bly of the pyruvate dehydrogenase multienzyme complex of Bacillus stearo- thermophilus. Thirteen of the 55 amino acids in the PSBD show significant changes in either or both of the 15 N and 1 H amide chemical shifts when the PSBD forms a 1 : 1 complex with E3int. All of the 13 amino acids reside near the N-terminus of helix I of PSBD or in the loop region between helix II and helix III. 15 N backbone dynamics experiments on PSBD indicate that the structured region extends from Val129 to Ala168, with limited structure present in residues Asn126 to Arg128. The presence of structure in the region before helix I was confirmed by a refinement of the NMR structure of uncomplexed PSBD. Comparison of the crystal structure of the PSBD bound to E3 [Mande SS, Sarfaty S, Allen MD, Perham RN & Hol WGJ (1996) Structure 4, 277–286] with the solution structure of uncomplexed PSBD described here indicates that the PSBD undergoes almost no conformational change upon binding to E3. These studies exemplify and validate the novel use of a solubilized, truncated pro- tein domain in overcoming the limitations of high molecular mass on NMR spectroscopy. Abbreviations E1, pyruvate decarboxylase; E2, dihydrolipoyl acetyltransferase; E3, dihydrolipoyl dehydrogenase; E3int, dimer of a solubilized interface domain; PDH, pyruvate dehydrogenase; PSBD, peripheral subunit-binding domain; ThDD, thrombin-cleavable di-domain; ThDP, thiamine diphosphate. FEBS Journal 272 (2005) 259–268 ª 2004 FEBS 259 metabolite in the synthesis of fatty acids, cholesterol, steroids in eukaryotes and N-acetyl-derived carbo- hydrates. The structural and mechanistic core of the 2-oxo acid dehydrogenase complexes is provided by the E2 component, each chain of which is composed of three independent domains. At the N-terminus are 1–3 tan- demly repeated lipoyl domains, followed by a peri- pheral subunit-binding domain (PSBD) responsible for binding E3 in the majority of organisms. In icosahe- dral complexes, the PSBD is also involved in the bind- ing of E1 [1,2]. The catalytic (acyltransferase) core domain, which assembles to form the octahedral or icosahedral inner core of the complexes, is proposed to bind E1 in the octahedral complexes [2,3], and is loca- ted at the C-terminus. Each of the individual domains is separated by long and flexible linker regions, which make possible active site coupling by allowing for large movements of the lipoyl domain(s) [1,4–6]. To date it has proved impossible to obtain crystals of an intact PDH complex. However, several three- dimensional structures have been determined for the individual domains of E2 chains from both icosahedral and octahedral complexes: lipoyl domain structures from Bacillus stearothermophilus, Escherichia coli and Azotobacter vinelandii PDH complexes [7–9], and E. coli and A. vinelandii 2-oxoglutarate dehydrogenase complexes [10,11]; PSBD structures from E. coli 2-oxo- glutarate dehydrogenase and B. stearothermophilus PDH complexes [12,13]; the octahedral acyltransferase core from A. vinelandii PDH [14] and E. coli 2-oxo- glutarate dehydrogenase [15] complexes and the icosa- hedral core from the B. stearothermophilus PDH complex [16]. Additionally, E3 and E1 structures from a number of sources have been solved by X-ray crys- tallography [17–24], and it has been possible to obtain a crystal structure of a B. stearothermophilus E3–PSBD complex formed between a lipoyl domain-PSBD di-domain and an E3 dimer [24]. Most recently, mod- els for the overall structures of the assembled PDH complexes, of some 10 MDa in molecular mass, have been proposed based on cryoelectron microscopy data [25,26]. An E3 dimer can bind only one PSBD domain of E2 [1,27]. This is due to steric hindrance, the associ- ation with one PSBD close to the twofold axis of E3 preventing the association of a second PSBD [28,29]. The interaction occurs at the C 2 -axis of symmetry in the E3 dimer, chiefly with the interface domain of E3, which is highly conserved and generates the majority of contacts across the dimer interface. A surface loop region of the PSBD appears to undergo a conforma- tional change when PSBD binds to the E3 dimer, as judged by the observed differences between the struc- tures obtained by NMR spectroscopy for the free PSBD [13] and X-ray crystallography for the E3–PSBD complex [28]. This paper describes the interaction of a 15 N-labelled PSBD (residues 119–171) of the B. stearothermophilus E2p polypeptide chain with the dimer of a protein domain (E3int) representing the interface domain (resi- dues 343–470) of B. stearothermophilus E3. The use of the engineered E3int domain (27 kDa as the dimer) was introduced because the high molecular mass (112 kDa) of the intact E3 dimer limited the applica- tion of NMR spectroscopy. Chemical shift differences between the backbone resonances of the uncomplexed PSBD and PSBD bound to E3int indicate that several amino acids near the N-terminus of helix I of the PSBD are at or near the E3-binding site, confirming and extending the earlier crystal structure [28]. The backbone dynamics of PSBD were also investigated, as the rigidity or otherwise of the proposed E3-binding site is crucial to a proper understanding of protein– protein interactions within the multienzyme complex. Further, an improved solution structure of PSBD was subsequently determined and compared with that of the PSBD in the crystal structure of the PSBD–E3 complex [28], thereby allowing a better estimate of the extent of molecular rearrangement that accompanies binding to E3. Results and Discussion Interaction of PSBD with the intact E3 dimer E3 (0.1 nmol of dimer) was mixed with various amounts of PSBD before being subjected to nondena- turing polyacrylamide gel electrophoresis, as described elsewhere [30]. Saturation of binding, as evidenced by the appearance of free PSBD in the Coomassie-stained gel, was found to occur when 0.1 nmol of E3 dimer was mixed with 0.1 nmol of PSBD. As expected, there- fore, free PSBD interacts with B. stearothermophilus E3 in a 1 : 1 stoichiometry identical to that observed previously with the PSBD as part of the lipoyl-PSBD di-domain [30]. The changes in backbone amide 15 N and 1 H chem- ical shifts upon mixing PSBD with E3 were slight (maximum chemical shift changes are 0.213 p.p.m. and 0.029 p.p.m. for 15 N and 1 H shifts, respectively; results not shown). Significant chemical shift changes (greater than 0.10 and 0.02 p.p.m. in the 15 N and 1 H dimen- sions, respectively) were observed for Asn126, Arg127, Ala131, Gly156 and Glu161. However, for all but the eight and two residues in the N- and C-terminal Assembly of pyruvate dehydrogenase complex M. D. Allen et al. 260 FEBS Journal 272 (2005) 259–268 ª 2004 FEBS regions, respectively, linewidths in the PSBD–E3 dimer (10 : 1) mixture were much broader than those found for PSBD alone, greatly diminishing the information that could be derived from the spectrum. This is pre- sumably due to the high molecular mass (115 kDa) and correspondingly long rotational correlation time of the PSBD–E3 complex. Interaction of PSBD with E3int To overcome these problems connected with the high molecular mass of the PSBD–E3 complex, the inter- action of PSBD with a solubilized interface domain (E3int) of B. stearothermophilus E3 was studied. The E3int domain comprises the C-terminal portion, resi- dues 343–470, of the B. stearothermophilus E3 chain and, based on the crystal structure of E3, contains a majority of the intersubunit contacts in the E3 dimer [1,18,28]. It has proved possible to generate a dimer- ic form of the corresponding interface domain of the homologous glutathione reductase, a flavoprotein like E3, from E. coli. To achieve this, hydrophobic pat- ches on the surface of the domain were altered to display charged or hydrophilic residues in key posi- tions, thereby rendering the domain soluble but not prone to aggregation beyond the desired dimer stage [31]. A similar programme of mutation was therefore undertaken on a subgene encoding the interface domain of the B. stearothermophilus E3. The final version of the solubilized E3int domain contains seven mutations of surface hydrophobic residues (I384D, V352N, L391A, I465G, I384D, V352N and D443N) chosen to render the domain more soluble than its wild-type counterpart, but not to interfere with dimerization. The E3int domain forms a dimer in the same way as E3, as determined by gel filtra- tion (data not shown), and can be expected to inter- act with PSBD in essentially the same way as the intact E3 dimer [29,30]. However, the E3int dimer has a molecular mass of only 27 kDa compared with 110 kDa for E3. Linewidths for all but the 12 and terminal residues in the N- and C-terminal regions of PSBD (which do not form part of the structured region) were found to be narrower than those observed on complexation with E3. This reflects the smaller size and shorter rotational correlation time of the PSBD–E3int com- plex (molecular mass 32 kDa). However, linewidths were still broader than those in the spectra of PSBD alone. The changes in backbone amide 15 N and 1 H chem- ical shifts upon mixing PSBD with E3int are shown in Fig. 1A,B, respectively. The chemical shift changes between the free PSBD and PSBD–E3int spectra are more pronounced (maximum changes are 0.500 p.p.m. and 0.065 p.p.m. for 15 N and 1 H shifts, respectively) than those observed between free PSBD and PSBD–E3 spectra. Significant changes in chemical shift were observed for Ala123, Asn126, Arg127, Arg128, Val129, Ile130, Ala131, Met132, Val135, Arg136, Lys137, Lys154 and Arg157. The location of residues undergo- ing large chemical shift changes upon interaction with E3 and E3int are mapped on to the three-dimensional structure of PSBD in Fig. 1C (where changes greater than 0.10 p.p.m. or 0.02 p.p.m. in the 15 N and 1 H dimensions, respectively, are indicated by grey spheres). The major locations include residues 126–137 straddling the start of helix I (residues 134–142) and residues Lys154, Gly156 and Arg157 in the loop (L2) between helix III (residues 145–149) and helix II (resi- dues 159–168). As shown in Fig. 1C, all the affected residues lie relatively close in space, with both the region near the N-terminus of helix I of the domain and the three residues in loop L2 contributing to the E3-binding site. Backbone dynamics of PSBD The region (Met132 to Ser134) before the start of helix I of the PSBD structure was reported previously [13] to be largely unstructured owing to an absence of detectable long-range NOEs in the NMR spectrum. The availability of uniformly 15 N-labelled PSBD enabled us now to analyse the backbone dynamics using a steady-state [ 1 H]- 15 N NOE experiment [32,33] in 20 mm potassium phosphate buffer, pH 6.5, at 298 K. The calculated values of g for PSBD plotted against residue number are shown in Fig. 2. The plot indicates that the structured region extends from Val129 to Ala168, but residues Asn126, Arg127 and Arg128 do appear to be partly mobile. Structure determination of PSBD Because the backbone dynamics experiment revealed that residues 117–125 and 171 of the PSBD are highly flexible in solution, only residues 126–170 were inclu- ded in the structure calculations. The final set of struc- tures of PSBD contained 841 unambiguous conformational constraints (summarized in Table 1). The final ensemble comprised 25 structures, all of which had no distance violations > 0.25 A ˚ , no dihed- ral angle violations > 5 °, and very small deviations from ideal covalent geometry. Most of the residues are restricted to favoured or additionally allowed regions of (/,w) space with, of the nonglycine residues, only M. D. Allen et al. Assembly of pyruvate dehydrogenase complex FEBS Journal 272 (2005) 259–268 ª 2004 FEBS 261 Lys154 consistently falling into disallowed regions. Figure 3 illustrates the ensemble and secondary struc- ture regions of PSBD. All statistical calculations were carried out on the structured region from Val129 to Ala168. Outside this region the torsion angles of / and w deviated rapidly away from their mean values. The average root mean square (rms) deviation from the mean coordinates for backbone nuclei over the structured region from Val129 to Ala168 is 0.35 A ˚ (Table 1). Backbone fractional H- N- NOE enhancement 1 15 -4.0 -3.6 -3.2 -2.8 -2.4 -2.0 -1.6 -1.2 -0.8 -0.4 0.0 120 125 130 135 140 145 150 155 160 165 170 Residue number Fig. 2. Backbone dynamics of PSBD. Plot of the backbone steady state { 1 H}- 15 N NOE enhancement, g, for PSBD as a function of residue number. -0.06 -0.04 -0.02 0.00 0.02 0.04 0.06 K154 R127 V129 I130 A131 M132 V135 K137 B Residue Number H- chemical shift change (ppm) 1 120 125 130 135 140 145 150 155 160 165 170 -0.5 -0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 0.5 Residue Number N126 V129 I130 R136 R157 A C A123 N- chemical shift change (ppm) 15 120 125 130 135 140 145 150 155 160 165 170 Lys137 Arg136 Val135 Met132 Ala131 Ile130 Val129 Arg 128 Arg127 Asn126 Lys154 Gly156 Arg157 Glu161 N C Fig. 1. NMR spectroscopy of the interaction of PSBD and E3int. Plots of the (A) 15 N and (B) 1 H N chemical shift changes between the [ 15 N, 1 H]-HSQC spectra of free PSBD and the PSBD–E3int dimer (10 : 1) mixture as a function of residue number. (C) Schematic MOLSCRIPT [46] drawing of the three-dimensional structure of PSBD; residues undergoing significant chemical shift changes (> 0.10 p.p.m. or > 0.02 p.p.m. in the 15 Nor 1 H dimensions, respectively) are illustrated as grey spheres. The molecule is coloured from the N- to the C-termi- nus following the colours of the visible spectrum (violet for N-terminus and red for the C-terminus). Assembly of pyruvate dehydrogenase complex M. D. Allen et al. 262 FEBS Journal 272 (2005) 259–268 ª 2004 FEBS Description of the three-dimensional structure of PSBD The refined NMR structure of PSBD is consistent with the two previously reported E3-binding domain struc- tures: that of synthetic peptides representing the PSBD of the dihydrosuccinyltransferase chain of the 2-oxoglutarate dehydrogenase complex of E. coli [12] and of the dihydroacetyltransferase chain of the PDH complex of B. stearothermophilus [13]. The domain consists of two a-helices comprising residues Ser134 to Lys142 (helix I) and Leu159 to Leu168 (helix II), and a short 3 10 -helix Asp145 to Val149 (helix III), with loops connecting the helices. The N-terminal region (Val129 to Pro133) is relatively well-defined and possesses a hydrogen bond between Val158 H N and Ile130 C¢. This hydrogen bond was not observed for the construct used by Kalia et al. [13] and undoubtedly the inclusion of this restraint in our struc- ture calculations permitted this region to adopt a more rigid conformation. The region is further stabilized by a number of hydrophobic contacts between Val129, Ala131, Val135 and Ile146. The loop L2 between the 3 10 -helix and helix II contains several residues with amide protons exhibiting reduced rates of exchange with the solvent. Analysis of the initial structures allowed hydrogen bond acceptors for each of these residues to be identified. The structure also appears to be stabilized by hydrogen bonds between Tyr138 O g H g and Asp164 O d2 (although this was not included as a restraint in the structure calculations). Structural comparisons The three-dimensional structure of the PSBD from the E2 chain of the B. stearothermophilus PDH complex has been determined before, by NMR spectroscopy of the free PSBD [13] and X-ray crystallography of the E3–PSBD complex [28]. All three structures now avail- able are very similar with respect to the arrangement of structural motifs and loops, with the exception of loop L2 where differences exist. Previously it was noted that upon superposition of the NMR structure and crystal structure, the tip of loop L2 appeared to Table 1. Summary of constraints and statistics for the 20 accepted structures of B. stearothermphils PSBD domain. Structural constraints Intra-residue 341 Sequential 178 Medium-range ( 2 £ |i-j| £ 4 ) 119 Long-range ( |i-j| > 4 ) 137 Dihedral angle constraints 22 Distance constraints for 22 hydrogen bonds 44 Total 841 Statistics for accepted structures Statistics parameter (± SD) Rms deviation for distance constraints 0.006 A ˚ ± 0.001A ˚ Rms deviation for dihedral constraints 0.02 ° ± 0.01 ° Mean X-PLOR energy term (kcalÆmol )1 ± SD) E (overall) 40.0 ± 3.0 E (van der Waals) 11.3 ± 1.4 E (distance constraints) 2.5 ± 0.5 E (dihedral and TALOS constraints) 0.003 ± 0.002 Rms deviations from the ideal geometry (± SD) Bond lengths 0.0011 A ˚ ± 0.0001 A ˚ Bond angles 0.35 ° ± 0.01 ° Improper angles 0.16 ° ± 0.01 ° Average atomic rmsd from the mean structure (± SD) Residues 129–168 (N, Ca, C atoms) 0.33 A ˚ ±0.09A ˚ Residues 129–168 (all heavy atoms) 0.33 A ˚ ±0.09A ˚ Ser134 Lys142 Asp145 Ala168 Leu159 Val149 Fig. 3. Solution structure of PSBD from NMR spectroscopy. Superposition of back- bone traces of the 25 accepted structures over residues 129–168, and a MOLSCRIPT [46] representation of the PSBD structure in the same orientation. The residues defining the secondary structural elements are labeled. M. D. Allen et al. Assembly of pyruvate dehydrogenase complex FEBS Journal 272 (2005) 259–268 ª 2004 FEBS 263 move by 9.2 A ˚ [28]. It was suggested that the differ- ence might be due to changes taking place in loop L2 upon binding to the E3 dimer or might reflect an intrinsic flexibility of loop L2. The results from the backbone dynamics experiment described above, how- ever, reveal that the loop is inherently rigid on the sub- nanosecond time scale. Together with the absence of significant chemical shift changes for more residues in loop L2 when the PSBD binds to E3int, this suggests that the loop does not undergo significant structural rearrangement upon binding to E3. Moreover, super- positioning shows that the crystal [28] and present NMR (see above) structures of PSBD are virtually identical. Comparative Ramachandran plots of the PSBD NMR and crystal structures reveal only minor differences in loop L2, which are confined to residues Gly153, Lys154 and Asn155. The differences observed between the previous NMR structure [13] and that determined here (see above) are probably due to the use of uniformly 15 N-labelled PSBD and the inclusion of additional resi- dues in the construct at the N-terminus of the domain. In particular, the absence of residues Arg127 and Arg128 from the earlier construct [13] may have affec- ted the stability of the N-terminal region, thereby pre- venting it from adopting the defined conformation observed in the crystal and new NMR structures. The different conformations of loop L2 between the two NMR structures can also be attributed to the different N-terminal regions, as the lack of the observed hydro- gen bond constraint between Val158 H N and Ile130 C¢ would have significantly affected the earlier calcula- tions [13]. The remaining slight differences between the crystal structure and the new NMR structure are prob- ably due to the relatively low number of NOEs observed between Lys154 and Asn155 in the loop L2. The site of interaction between E3 and PSBD The chemical shift changes observed on formation of a complex between PSBD and E3int are likely to be due principally to direct contact between the two proteins, together with some changes in the conformation of PSBD upon binding. Figure 1 shows large chemical shift changes for Asn126, Arg127, Val129, Ile130, Ala131, Met132, Val135, Arg136, Lys137, Lys154, Gly156 and Arg157, all of which are located close in three-dimen- sional space. The crystal structure of the E3-lipoyl- PSBD di-domain complex has already been determined [28]. The hydrophobic residues Val129, Ile130, Ala131, Met132, Val135 and basic residues Arg127, Arg128, Arg136, Lys137, Lys154 and Arg157 implicated by these NMR studies are arranged in the crystal structure such that the hydrophobic regions on PSBD and E3 form multiple van der Waals contacts, while the acidic resi- dues of E3 and basic residues of PSBD generate multiple salt-bridges. The results obtained by means of NMR spectroscopy are now wholly consistent with the struc- ture of the PSBD–E3 complex determined by X-ray crystallography. This rules out any doubt that the crys- tal structure might not be a valid representation of the interaction in solution and indicates that the interaction between PSBD and E3, though very tight (K d  10 )9 m [1,30]), is of the direct ‘lock-and-key’ kind rather than an induced fit. As a corollary, it is clear that the approach we have been developing, of creating a soluble construct containing the interface region of the E3 dimer [31,34], has made it possible to overcome the molecular mass limitation in using NMR spectroscopy to study protein structure and protein–protein interaction. This augurs well for future studies, for example by transverse relaxation compensated NMR spectroscopy. Experimental procedures Materials Bacteriological media were from Difco (Detroit, MI, USA). The pBSTNAVDD vector carrying the dihydrolipoyl ace- tyltransferase gene that encodes residues 1–171 of B. stearo- thermophilus E2p was generated earlier [35]. Plasmid pET11d and E. coli host strain BL21(DE3) [(F – , ompT, hsdS B (r b – ,m b – ), gal, dcm (DE3)] were obtained from Nov- agen Inc (Madison, WI, USA). Construction of expression vector pET11ThDD Construction of plasmid pET11ThDD encoding residues 1–171 of B. stearothermophilus E2p with a thrombin-cleavage site (LVPRGS) in place of Ala118 proceeded via double overlapping PCR mutagenesis using standard techniques [36]. Plasmid pBSTNAVDD [35] was used as the template DNA. The PCR product was digested with NcoI and BamH1, purified by means of agarose gel electrophoresis and ligated into vector pET11d previously digested with NcoI and BamH1 and treated with calf intestinal alkaline phospha- tase. The resulting vector encodes the sequence of a di-domain with a thrombin-cleavable linker region (ThDD). The DNA was fully sequenced to ensure its fidelity. Expression and purification of 15 N-labelled PSBD E. coli strain BL21(DE3) cells transformed with pET11ThDD were grown to an A 600 of 1.5 in K-Mops minimal medium [37] containing 10 mm 15 N-NH 4 Cl before being induced by the addition of isopropyl thio-b-d-gal- Assembly of pyruvate dehydrogenase complex M. D. Allen et al. 264 FEBS Journal 272 (2005) 259–268 ª 2004 FEBS actoside (1 mm final concentration). The cells were harves- ted after 3 h of induction, resuspended in 50 mm Tris ⁄ HCl buffer, pH 7.5, containing 1 mm EDTA, 1 mm phenyl- methanesulfonyl fluoride and 0.02% (w ⁄ v) sodium azide, and disrupted in a French press. Cell debris was removed by centrifugation and the supernatant was fractionally pre- cipitated with ammonium sulphate. The protein that was precipitated between 35 and 80% saturation was dialysed overnight into 50 mm potassium phosphate, pH 7.0. ThDD was purified by consecutive cation-exchange and anion- exchange chromatography using Hi-load TM -S and Mono TM - Q columns, respectively [38]. Purified ThDD (10 mgÆmL )1 ) was treated with thrombin, 40 UÆmL )1 final concentration, in 20 mm potassium phosphate buffer, pH 7.0, at 37 °C for 6 h, to cleave the linker between the lipoyl domain and PSBD. PSBD released in this way was purified by cation- exchange chromatography and its purity checked by SDS ⁄ PAGE [38]. Generation of the E3int dimer B. stearothermophilus E3 was purified from an over-expres- sion system in E. coli described previously [39]. The dimeric interface domain, comprising residues 343–470 of B. stearo- thermophilus E3 [34] was prepared in essentially the same way as the dimeric interface domain was excised from the homologous E. coli glutathione reductase and rendered more soluble by appropriate mutations on its freshly exposed hydrophobic surfaces [31]. To achieve this for the B. stearothermophilus E3 interface domain, the following amino acid exchanges were made to its freshly exposed hydrophobic surfaces: four residues making hydrophobic contacts with the FAD domain (L389D, A390S, L391A and I465G) and three residues abutting the NAD or central domains (I348D, V352N and I443N). The interface domain with these seven changes was designed to retain one of the two C 2 axes of symmetry present in the intact wild-type E3 dimer. NMR spectroscopy For 2D NMR spectroscopy, samples of 15 N-labelled PSBD (5 mm) and unlabelled PSBD (8 mm)in20mm potassium phosphate buffer, pH 6.5 (90% H 2 O ⁄ 10% D 2 O, v ⁄ v) were used. Two-dimensional [ 1 H] 15 N-HSQC spectra, used to detect backbone and side-chain amide resonances which slowly exchange with D 2 O, were recorded in 20 mm potas- sium phosphate buffer, pH 6.5, in 99.996% (v ⁄ v) D 2 O. NMR spectra were recorded on a Bruker AM-500 spectro- meter (500.13 MHz for 1 H and 50.68 MHz for 15 N) at 298 K. Mixing times were 60 ms and 150 ms for the TOCSY and NOESY experiments, respectively. Proton and nitrogen chemical shifts were determined relative to sodium 2,2-di- methyl-2-silapentane-5-sulfonate and liquid ammonium, respectively. Sequential assignment of the cross-peaks was achieved using interresidue NOE connectivities by standard 2D NMR procedures [40,41]. Stereospecific assignments of H b resonances were determined by the use of cross-peak intensities in the 2D NOESY and 2D TOCSY [40]. The back- bone dynamics of PSBD were investigated using steady-state { 1 H}- 15 N nuclear Overhauser enhancement (NOE) experi- ments [32,33]. Values of the { 1 H}- 15 N NOE were determined according to the formula g ¼ (I ¢ – I ref ) ⁄ I ref , where I ¢ is the intensity of a cross-peak in an experiment with 3 s broad- band 1 H presaturation and I ref is the intensity in a reference spectrum recorded without presaturation. NMR spectroscopic studies of protein–protein interaction For 2D NMR spectroscopic studies of protein–protein interaction, samples of 15 N-labelled PSBD (1 mm)in 20 mm potassium phosphate buffer, pH 6.5 (90% H 2 O ⁄ 10% D 2 O, v ⁄ v), were used. B. stearothermophilus E3 dimer (final concentrations of either 0.25 mm and 0.1 mm) was added to PSBD samples to study the interaction of PSBD with E3. B. stearothermophilus E3int dimer (final concentration 0.1 mm) was added to PSBD samples to study the interaction of PSBD with E3int. Distance constraints A set of distance constraints was derived from NOESY spectra recorded in H 2 O and D 2 O with mixing times of 150 ms. Each NOESY spectrum was integrated according to the cross-peak strengths and calibrated by comparison with NOE connectivities obtained for standard interresidue distances within an a-helix. After calibration, the NOE con- straints were classified into four categories: 1.8–2.8, 1.8–3.5, 1.8–4.75 and 2.5–6.0 A ˚ . The distance constraints in the ini- tial ensemble of structure calculations were derived only from NOESY cross-peaks that could be unambiguously assigned on the basis of chemical shift alone. Structure calculation and refinement Structure refinement proceeded in an iterative manner in which distance constraints were added or modified follow- ing analysis of the previous ensemble of structures. The vic- inal proton coupling constant ( 3 J aN ) was determined by the use of a series of 2D J-modulated ( 15 N- 1 H)-COSY spectra [42]. Comparison of the coupling constant with an experi- mentally derived Karplus curve [43,44] enabled the torsion angle, /, to be estimated when the coupling constant was greater than 7.1 Hz. In the final round of structure calculations, hydrogen bond constraints were included for a number of backbone H N groups whose signals were observed to change slowly when the sample buffer was exchanged for D 2 O. For M. D. Allen et al. Assembly of pyruvate dehydrogenase complex FEBS Journal 272 (2005) 259–268 ª 2004 FEBS 265 hydrogen bond partners, two distance constraints were used where the distance (D) H-O (A) corresponded to 1.8–2.1 A ˚ and (D) N-O (A) to 2.8–3.2 A ˚ . The stereospecific assignments of H b resonances determined from NOESY and TOCSY spectra were confirmed by analysing the initial ensemble of structures. Additional stereospecific assignments were iden- tified for resolved resonances when the side-chain atoms were sufficiently well-defined in the ensemble of structures. Each round of assignment was followed by a set of struc- ture calculations with the structural constraints including the stereospecific assignment, and confirmed when the resulting structures did not show any distance violations greater than 0.25 A ˚ . The 3D structure of PSBD was calcu- lated from 841 experimental constraints using cns version 1.1 [45]. Twenty structures were calculated from an exten- ded conformation using torsion angle dynamics in a standard simulated annealing protocol. Stereospecific assignments of prochiral protons were used, where avail- able; otherwise r )6 averaging was used over all equivalent protons. Structures were accepted where no distance vio- lation was greater than 0.25 A ˚ and no dihedral angle viola- tions > 5°. The final coordinates have been deposited in the Protein Data Bank (PDB accession no. 1w3d). Acknowledgements We thank the Biotechnology and Biological Sciences Research Council (BBSRC) for the award of a research grant (to RNP) and a Research Studentship (to MDA). The core facilities of the Cambridge Centre for Molecu- lar Recognition were funded by the BBSRC and The Wellcome Trust. We are grateful to Mr C Fuller for skilled technical assistance and to Dr ARC Raine for his help with the structure calculation. References 1 Perham RN (2000) Swinging arms and swinging domains in multifunctional enzymes: catalytic machines for multi- step reactions. Annu Rev Biochem 69, 963–1006. 2 Packman LC, Borges A & Perham RN (1988) Amino acid sequence analysis of the lipoyl and peripheral subunit-binding domains in the lipoate acetyltransferase component of the pyruvate dehydrogenase complex from Bacillus stearothermophilus. Biochem J 252, 79–86. 3 Hanemaaijer R, Janssen A, De Kok A & Veeger C (1989) The dihydrolipoyl transacetylase component of the pyruvate dehydrogenase complex from Azotobacter vinelandii. Eur J Biochem 174, 593–599. 4 Texter FL, Radford SE, Laue ED, Perham RN, Miles JS & Guest JR (1988) Site-directed mutagenesis and 1 H-NMR spectroscopy of an interdomain segment in the pyruvate dehydrogenase multienzyme complex of Escherichia coli. Biochemistry 27, 289–296. 5 Radford SE, Laue ED, Perham RN, Martin SR & Appella E (1989) Conformational flexibility and folding of synthetic peptides representing an interdomain seg- ment of polypeptide chain in the pyruvate dehydrogen- ase multienzyme complex of Escherichia coli. J Biol Chem 264, 767–775. 6 Green JDF, Perham RN, Ullrich SJ & Appella E (1992) Conformational studies of the inter-domain linker pep- tides in the dihydrolipoyl acetyltransferase component of the pyruvate dehydrogenase multienzyme complex of Escherichia coli. J Biol Chem 267, 23484–23488. 7 Dardel F, Davis AL, Laue ED & Perham RN (1993) The three-dimensional structure of the lipoyl domain from Bacillus stearothermophilus pyruvate dehydrogen- ase multienzyme complex. J Mol Biol 229, 1037–1048. 8 Jones DD, Stott KM, Howard MJ & Perham RN (2000) Restricted motion of the lipoyl-lysine swinging arm in the pyruvate dehydrogenase complex of Escheri- chia coli. Biochemistry 39, 8448–8459. 9 Berg A, Vervoort J & de Kok A (1997) Three-dimen- sional structure of the N-terminal lipoyl domain of the pyruvate dehydrogenase component from Azotobacter vinelandii. Eur J Biochem 244, 352–360. 10 Ricaud PM, Howard MJ, Roberts EL, Broadhurst RW & Perham RN (1996) Three-dimensional structure of the lipoyl domain from the dihydrolipoyl succinyltrans- ferase component of the 2-oxoglutarate dehydrogenase multienzyme complex of Escherichia coli. J Mol Biol 264, 179–190. 11 Berg A, Vervoort J & de Kok A (1996) Solution struc- ture of the lipoyl domain of the 2-oxoglutarate dehydro- genase component from Azotobacter vinelandii. J Mol Biol 263, 432–442. 12 Robien MA, Clore GM, Omichinski JG, Perham RN, Appella E, Sakaguchi K & Gronenborn AM (1992) Three-dimensional solution structure of the E3-binding domain of the dihydrolipoamide succinyltransferase core from the 2-oxoglutarate dehydrogenase multienzyme complex of Escherichia coli. Biochemistry 31, 3463–3471. 13 Kalia YN, Brocklehurst SM, Hipps DS, Appella E, Sakaguchi K & Perham RN (1993) The high-resolution structure of the peripheral subunit-binding domain of dihydrolipoamide acetyltransferase from the pyruvate dehydrogenase multienzyme complex of Bacillus stearo- thermophilus. J Mol Biol 230, 323–341. 14 Mattevi A, Obmolova G, Schulze E, Kalk KH, West- phal AH, de Kok A & Hol WGJ (1992) Atomic struc- ture of the cubic core of the pyruvate dehydrogenase multienzyme complex. Science 255, 1544–1550. 15 Knapp JE, Mitchell DT, Yazdi MA, Ernst SR, Reed LJ & Hackert ML (1998) Crystal structure of the truncated cubic core component of the Escherichia coli 2-oxogluta- rate dehydrogenase multienzyme complex. J Mol Biol 280, 655–668. Assembly of pyruvate dehydrogenase complex M. D. Allen et al. 266 FEBS Journal 272 (2005) 259–268 ª 2004 FEBS 16 Izard T, Aevarsson A, Allen MD, Westphal AH, Per- ham RN, de Kok A & Hol WGJ (1999) Principles of quasi-equivalence and Euclidean geometry govern the assembly of cubic and dodecahedral cores of pyruvate dehydrogenase complexes. Proc Natl Acad Sci USA 96, 1240–1245. 17 Takenaka A, Kizawa K, Hata T, Sato S, Misaka E-J, Tamura C & Sasida Y (1988) X-ray study of baker’s yeast lipoamide dehydrogenase at 4.5 A ˚ resolution by molecular replacement method. J Biochem (Tokyo) 103, 463–469. 18 Schierbeek AJ, Swarte MBA, Dijkstra BW, Vriend G, Read RJ, Hol WGJ & Drenth J (1989) X-ray structure of lipoamide dehydrogenase from A. vinelandii deter- mined by a combination of molecular and isomorphous replacement techniques. J Mol Biol 206, 365–380. 19 Mattevi A, Obmolova G, Kalk KH, van Berkel WJH & Hol WGJ (1993) Three-dimensional structure of lipoa- mide dehydrogenase from Pseudomonas fluorescens at 2.8 A ˚ resolution. J Mol Biol 230, 1200–1215. 20 Mattevi A, Obmolova G, Kalk KH, van Berkel WJH & Hol WGJ (1992) The refined crystal structure of Pseudo- monas putida lipoamide dehydrogenase complexed with NAD + at 2.45 A ˚ resolution. Proteins Struct Funct Genet 13, 336–351. 21 Aevarsson A, Seger K, Turley S, Sokatch JR & Hol WGJ (1999) Crystal structure of 2-oxoisovalerate dehydrogen- ase and the architecture of 2-oxo acid dehydrogenase multienzyme complexes. Nat Struct Biol 6, 785–792. 22 Aevarsson A, Chuang JL, Wynn RM, Turley S, Chuang DT & Hol WGJ (2000) Crystal structure of human branched-chain alpha-ketoacid dehydrogenase and the molecular basis of multienzyme complex deficiency in maple syrup urine disease. Structure Fold Des 8, 277–291. 23 Arjunan P, Nemeria N, Brunskill A, Chandrasekhar K, Sax M, Yan Y, Jordan F, Guest JR & Furey W (2002) Structure of the pyruvate dehydrogenase multienzyme complex E1 component from Escherichia coli at 1.85 A ˚ resolution. Biochem 41, 5213–5221. 24 Ciszak EM, Korotchkina LG, Dominiak PM, Sidhu S & Patel MS (2003) Structural basis for flip-flop action of thiamin pyrophosphate-dependent enzymes revealed by human pyruvate dehydrogenase. J Biol Chem 278, 21240–21246. 25 Milne JL, Shi D, Rosenthal PB, Sunshine JS, Domingo GJ, Wu X, Brooks BR, Perham RN, Henderson R & Subramaniam S (2002) Molecular architecture and mechanism of an icosahedral pyruvate dehydrogenase complex: a multifunctional catalytic machine. EMBO J 21, 5587–5598. 26 Zhou ZH, McCarthy DB, O’Connor CM, Reed LJ & Stoops JK (2001) The remarkable structural and functional organization of the eukaryotic pyruvate dehydrogenase complexes. Proc Natl Acad Sci USA 98, 14802–14807. 27 Maeng C-Y, Yazdi MA, Nui X-D, Lee HY & Reed LJ (1994) Expression, purification and characterisation of the dihydrolipoamide dehydrogenase-binding pro- tein of the pyruvate dehydrogenase complex from Saccharomyces cerevisiae. Biochemistry 33, 13801– 13807. 28 Mande SS, Sarfaty S, Allen MD, Perham RN & Hol WGJ (1996) Protein–protein interactions in the pyruvate dehydrogenase multienzyme complex: dihydrolipoamide dehydrogenase complexed with the binding domain of di- hydrolipoamide acetyltransferase. Structure 4, 277–286. 29 Hipps DS, Packman LC, Allen MD, Fuller C, Sakag- uchi K, Appella E & Perham RN (1994) The periph- eral subunit-binding domain of the dihydrolipoyl acetyltransferase component of the pyruvate dehydro- genase complex of Bacillus stearothermophilus: pre- paration and characterization of its binding to the dihydrolipoyl dehydrogenase component. Biochem J 297, 137–143. 30 Lessard IAD & Perham RN (1995) Interaction of com- ponent enzymes with the peripheral subunit-binding domain of the pyruvate dehydrogenase multienzyme complex of Bacillus stearothermophilus. Biochem J 283, 665–671. 31 Leistler B & Perham RN (1994) Solubilizing buried domains of proteins: a self–assembling interface domain from glutathione reductase. Biochemistry 33, 2773–2781. 32 Kay LE, Torchia DA & Bax A (1989) Backbone dynamics of proteins as studied by 15 N inverse detected heteronuclear NMR spectroscopy: Application to sta- phylococcal nuclease. Biochemistry 28, 8972–8979. 33 Barbato G, Ikura M, Kay LE, Pastor RW & Bax A (1992) Backbone dynamics of calmodulin studied by 15 N relaxation using inverse detected two-dimensional NMR spectroscopy: The central helix is flexible. Bio- chemistry 31, 5269–5278. 34 Perham RN, Leistler B, Solomon RG & Guptasarma P (1996) Protein engineering of domains in flavoprotein disulphide oxidoreductases: contributions to folding and assembly. Biochem Soc Trans 24, 61–66. 35 Hipps DS & Perham RN (1992) Expression in Escheri- chia coli of a sub-gene encoding the lipoyl and periph- eral subunit-binding domains of the dihydrolipoamide acetyltransferase component of the pyruvate dehydro- genase complex of Bacillus stearothermophilus. Biochem J 283, 665–671. 36 Sambrook J, Fritsch EF & Maniatis T (1989) Molecular Cloning: a Laboratory Manual (Nolan. C, ed.) Volumes 1–3, 2nd edn. Cold Spring Harbor Laboratory Press, New York. 37 Neidhardt FC, Bloch PL & Smith DF (1974) Culture medium for enterobacteria. J Bacteriol 119, 736–747. M. D. Allen et al. Assembly of pyruvate dehydrogenase complex FEBS Journal 272 (2005) 259–268 ª 2004 FEBS 267 38 Wallis NG, Allen MD, Broadhurst RW, Lessard IAD & Perham RN (1996) Recognition of a surface loop of the lipoyl domain underlies substrate channelling in the pyruvate dehydrogenase multienzyme complex. J Mol Biol 263, 463–474. 39 Lessard IA, Domingo GJ, Borges A & Perham RN (1998) Expression of genes encoding the E2 and E3 components of the Bacillus stearothermophilus pyruvate dehydrogenase complex and the stoichiometry of subu- nit interaction in assembly in vitro. Eur J Biochem 258, 491–501. 40 Wu ¨ thrich K (1986) NMR of Proteins and Nucleic Acids. J. Wiley, New York. 41 Englander SW & Wand AJ (1987) Main-chain-directed strategy for the assignment of 1 H NMR spectra of pro- teins. Biochemistry 26, 5953–5958. 42 Billeter M, Neri D, Otting G, Qian YQ & Wu ¨ thrich K (1992) Precise vicinal coupling-constants 3 J Na in proteins from non-linear fits of J-modulated ( 15 N- 1 H)-COSY experiments. J Biomol NMR 2, 257–274. 43 Karplus M (1959) Contact electron-spin coupling of nuclear magnetic moments. J Chem Phys 30, 11–15. 44 Karplus M (1963) Vicinal proton coupling in nuclear magnetic resonance. J Am Chem Soc 85, 2870–2971. 45 Nilges M, Clore GM & Gronenborn AM (1988) Deter- mination of the three-dimensional structures of proteins from inter-proton distance data by dynamic simulated annealing from a random array of atoms. FEBS Lett 239, 129–136. 46 Kraulis PJ (1991) MOLSCRIPT: a program to produce both detailed and schematic plots of protein structures. J Appl Crystallog 24, 946–950. Assembly of pyruvate dehydrogenase complex M. D. Allen et al. 268 FEBS Journal 272 (2005) 259–268 ª 2004 FEBS . Interaction of the E2 and E3 components of the pyruvate dehydrogenase multienzyme complex of Bacillus stearothermophilus Use of a truncated. Expression of genes encoding the E2 and E3 components of the Bacillus stearothermophilus pyruvate dehydrogenase complex and the stoichiometry of subu- nit interaction