Interactions of the peripheral subunit-binding domain of the dihydrolipoyl acetyltransferase component in the assembly of the pyruvate dehydrogenase multienzyme complex of Bacillus stearothermophilus Hyo-Il Jung 1 , Alan Cooper 2 and Richard N. Perham 1 1 Cambridge Centre for Molecular Recognition, Department of Biochemistry, University of Cambridge, UK; 2 Department of Chemistry, University of Glasgow, UK The enzymes pyruvate decarboxylase (E1) and dihydro- lipoyl dehydrogenase (E3) bind tightly but in a mutually exclusive manner to the peripheral subunit-binding domain (PSBD) of dihydrolipoyl acetyltransferase in the pyruvate dehydrogenase multienzyme complex of Bacillus stearo- thermophilus. The use of directed mutagenesis, surface plasmon resonance detection and isothermal titration microcalorimetry revealed that several positively charged residues of the PSBD, most notably Arg135, play an important part in the interaction with both E1 and E3, whereas Met131 makes a significant contribution to the binding of E1 only. This indicates that the binding sites for E1 and E3 on the PSBD are overlapping but probably significantly different, and that additional hydrophobic interactions may be involved in binding E1 compared with E3. Arg135 of the PSBD was also replaced with cysteine (R135C), which was then modified chemically by alkylation with increasingly large aliphatic groups (R135C -methyl, -ethyl, -propyl and -butyl). The pattern of changes in the values of DG°, DH° and TDS° that were found to accom- pany the interaction with the variant PSBDs differed between E1 and E3 despite the similarities in the free ener- gies of their binding to the wild-type. The importance of a positive charge on the side-chain at position 135 for the interaction of the PSBD with E3 and E1 was apparent, although lysine was found to be an imperfect substitute for arginine. The results offer further evidence of entropy– enthalpy compensation (Ôthermodynamic homeostasisÕ) ) a feature of systems involving a multiplicity of weak inter- actions. Keywords: pyruvate dehydrogenase multienzyme complex; surface plasmon resonance; isothermal titration micro- calorimetry; protein–protein interaction; thermodynamics. The oxidative decarboxylation of pyruvate in most cells and organisms is carried out by enzymes of the pyruvate dehydrogenase (PDH) multienzyme complex, which con- sists of pyruvate decarboxylase (E1; EC 1.2.4.1), dihydrolipoyl acetyltransferase (E2; EC 2.3.1.12) and dihydrolipoyl dehydrogenase (E3; EC 1.8.1.4). The three enzymes are noncovalently, but tightly, assembled into a highly organized multifunctional catalytic machine ([1–4] and references therein). The E2 chain of the PDH complex of Bacillus stearo- thermophilus is composed of three major folding units: a lipoyl domain (LD,  80 residues), a peripheral subunit- binding domain (PSBD,  35 residues) and an acetyltrans- ferase inner-core catalytic domain (CD,  250 residues), which are joined together by long and flexible linker segments ( 25–40 residues) rich in alanine, proline and charged amino acids [2,5]. It is the acetyltransferase CD that aggregates to form an inner core of icosahedral (60-mer) symmetry [4,6]. The PSBD, located between the LD and CD, is one of the smallest known globular protein domains that lacks disulfide bridges or stabilizing metal ions. According to the three-dimensional solution [7] and crystal [8] structures of the binding domain, it has a compact fold stabilized mainly by hydrophobic interactions, and consists of two almost parallel short a-helices (H1 and H2), a turn of distorted 3 10 -helix, and loops L1 and L2 joining these structural elements. The main function of the PSBD is to attach both E1 and E3 to the icosahedral E2 core [9,10]. The binding stoichio- metry and the kinetic and thermodynamic parameters of the interactions have been investigated by means of non- denaturing gel electrophoresis, surface plasmon resonance Correspondence to 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 333667, Tel.: + 44 1223 333663, E-mail: r.n.perham@bioc.cam.ac.uk Abbreviations: CD, catalytic domain; E1, pyruvate decarboxylase; E2, dihydrolipoyl acetyltransferase; E3, dihydrolipoyl dehydrogenase; ITC, isothermal titration microcalorimetry; LD, lipoyl domain; PDH, pyruvate dehydrogenase; PSBD, peripheral subunit-binding domain; SPR, surface plasmon resonance; ThDD, thrombin-cleavable di-domain. Enzymes: pyruvate decarboxylase (EC 1.2.4.1); dihydrolipoyl acetyl- transferase (EC 2.3.1.12); dihydrolipoyl dehydrogenase (EC 1.8.1.4). (Received 18 July 2003, revised 11 September 2003, accepted 19 September 2003) Eur. J. Biochem. 270, 4488–4496 (2003) Ó FEBS 2003 doi:10.1046/j.1432-1033.2003.03842.x (SPR) analysis and isothermal titration microcalorimetry (ITC) [11–13]. One B. stearothermophilus PSBD (i.e. in effect one E2 chain) is capable of binding either one E3 dimer or one E1 (a 2 b 2 ) heterotetramer, but not both simultaneously. The dissociation constant (K d )forthe complex formed with E3 (5.8 · 10 )10 M ) was found to be almost twofold higher than that for the complex with E1 (3.2 · 10 )10 M ). Although the PSBD has such a strong affinity, formation of the complex does not appear to cause any major conformational change near the active site of either E1 or E3 [8,10]. The association of the PSBD with E3 at 25 °Cis characterized by a small, unfavourable enthalpy change (DH° ¼ +2.2 kcalÆmol )1 ) and a large, positive entropy change (TDS° ¼ +14.8 kcalÆmol )1 ), whereas that with E1 is accompanied by a favourable enthalpy change (DH° ¼ )8.4 kcalÆmol )1 ) and a less positive entropy change (TDS° ¼ +4.5 kcalÆmol )1 ), in both instances with marked DC p effects, as described in detail elsewhere [12]. The replacement of Arg135 in the PSBD with alanine (R135A) causes a significant decrease in the binding affinity of PSBD for E3 [13]. Indeed, Arg135 plays a central role in the binding energetics; the R135A mutation is associated with more favourable enthalpy changes and less positive entropy changes in E3 binding. Such detailed information on the interaction with E1 has thus far been unavailable. The R135A mutation results in the loss of both the charged guanidino group and most of the hydrophobic aliphatic side-chain. To date our investigations have quan- tified the overall effects of the R135A mutation on the kinetics and energetics of the binding to E3 [13] but had provided no evidence as to the relative parts played by the ionized guanidino group and the aliphatic side-chain of the arginine residue. In this paper we extend our studies to the interaction of the PSBD with E1. The binding sites on the PSBD for E1 and E3 may overlap but the interactions are clearly distinguishable. We have also been able to separate the thermodynamic contributions made by the guanidino group and the aliphatic side-chain of the arginine residue, using a technique of cysteine engineering [14]. Materials and methods Materials All reagents used were of analytical grade unless otherwise stated. The sources of all restriction endonucleases, fine chemicals, bacterial strains and media, plasmids and antibiotics have been listed elsewhere [12,15]. Design of mutagenic oligonucleotides The following mutagenic oligonucleotides, designed to convert the indicated wild-type amino acid residues into alanine, leucine, methionine and cysteine respectively, were used: 5¢-end primer (forward), 5¢-GATAACAATTCC CCTCTAGAAA-3¢;3¢-end primer (reverse), 5¢-GCGGG ATATCCGGATATAGT-3¢; R135L (forward), 5¢-ATG CCGTCCGTG CTCAAGTATGC-3¢; R135L (reverse), 5¢- CGCGCATACTT GAGCACGGACGGCAT-3¢; R135M (forward), 5¢-ATGCCGTCCGTG ATGAAGTATGC-3¢; R135M (reverse), 5¢-CGCGCATACTT CATCACGGAC GGCAT-3¢; R135K (forward), 5¢-GCCATGCCGTCCGT G AAGAAGTATGCGCGCGAAAAA-3¢; R135K (reverse), 5¢-TTTCGCGCGCATACTT CTTCACGGACGGCA-3¢; R135C (forward), 5¢-ATGCCGTCCGTG TGCAAGTA TGC-3¢; R135C (reverse), 5¢-CGCGCATACTT GCACAC GGACGGCAT-3¢. The altered codons are underlined. The mutants were constructed in plasmid pET11ThDD as described elsewhere [12]. The various forms of plasmid pET11ThDD encoding the M131A, R135A, K136A, R139A, R146A, K153A and R156A mutants have been described previously [13]. Expression of genes and purification of proteins Plasmid pET11ThDD carries a subgene encoding residues 1-170 of the B. stearothermophilus E2chainwithathrom- bin-cleavage site in the linker region between the LD and PSBD [16]. The mutant plasmids were overexpressed in Escherichia coli BL21(DE3) cells grown at 37 °Cin Luria–Bertani medium supplemented with ampicillin. After induction with isopropyl thio-b- D -galactosidase (final concentration, 1 m M ) for 2 h, mutant forms of the thrombin-cleavable di-domain (ThDD) were purified as described by Jung et al. [13]. The LD in the ThDD was lipoylated (on Lys42) by treatment with E. coli lipoate protein ligase in the presence of ATP and lipoic acid [13]. Recombinant B. stearothermophilus E1 and E3 were puri- fied from genes overexpressed in E. coli cells as described elsewhere [10,17]. Chemical modification of the R135C ThDD The newly introduced cysteine residue at position 135 in the purified mutant R135C ThDD was chemically modified by exposure to alkyl iodides, essentially as described by Hasan and Leatherbarrow [14]. Iodomethane, iodoethane, iodo- propane and iodobutane [2 M solutions in 96% (v/v) ethanol] were added to the mutant R135C protein ( 25 mg) in 0.1 M Tris/HCl (pH 8.2), containing 10 m M dithiothreitol under nitrogen, to a final concentration of 70 m M ,andthemixture was incubated for 6 h at 25 °C in the dark. Each reaction tube was shaken several times at regular intervals. The extent of reaction with each iodoalkane was estimated by subjecting samples of the protein to electrospray mass spectrometry (ESI-MS) intermittently in a Micromass Quattro-LC mass spectrometer. When full alkylation was achieved, the reac- tion mixture was loaded onto a Mono Q TM high-perform- ance anion-exchange column equilibrated with buffer solution A [20 m M potassium phosphate, pH 7.0, and 0.02% (w/v) sodium azide] to remove excess reagent. The modified protein was subsequently eluted by applying a linear gradient (15–75%) of buffer B [20 m M potassium phosphate, pH 7.0, 1 M NaCl and 0.02% (w/v) sodium azide]. Fractions containing the chemically modified R135C protein were pooled, dialysed exhaustively against water and concentrated by Centriprep TM filtration. Analysis of the binding affinity of mutant PSBDs with E1 and E3 Interaction of the PSBD in each ThDD with E1 and E3 was analysed by means of SPR detection in a BIAcore Ó FEBS 2003 Thermodynamics of protein complex assembly (Eur. J. Biochem. 270) 4489 instrument, as described elsewhere [11,13], and by ITC measurements carried out in a Microcal calorimeter, as described by Jung et al. [12,13]. General protein chemical techniques Non-denaturing PAGE, SDS/PAGE and amino acid analysis were carried out as described elsewhere [10,11]. Results Possible overlap between the binding sites on the PSBD for E1 and E3 The small but highly compact PSBD ( 35 amino acid residues) of the B. stearothermophilus E2 chain binds either E1 or E3 but not both simultaneously. Although the small size of the PSBD is sufficient to explain the mutual exclusivity, there is no evidence as to how the E1 binding site on the PSBD relates to that for E3. The crystal structure of B. stearothermophilus E3 complexed with the PSBD has been determined to 2.6 A ˚ resolution [8]. Residues Arg135, Arg139 and Arg156 of the PSBD are involved in an Ôelectrostatic zipperÕ with AspB344 and GluB431 in the E3 dimer, and residues Ser133 and Lys136 of the PSBD make interactions with monomer A of E3 (Fig. 1A). Alanine- scanning mutagenesis confirmed that the positively charged residues at positions 135, 136, 139 and 156 in the PSBD play a vital part in the interaction with E3 and highlighted the particular importance of Arg135 [13]. To assess their contribution, if any, to the binding of E1, these four positively charged residues (Arg135, Lys136, Arg139 and Arg156) plus two neutral residues (Met131 and Ser133) and two other positively charged residues (Arg146 and Lys153) were targeted for alanine-scanning mutagenesis. As shown in Fig. 1B, all these residues are located on the helix 1 and loop 2 regions of the PSBD and are fully solvent-exposed. Binding of mutant PSBDs to E1 All the mutant forms of the ThDD (a di-domain comprising theLDplusPSBDofB. stearothermophilus E2 [8,15]) generated from subgenes overexpressed in E. coli (BL21) cells, were purified as described elsewhere [13]. They are known to be folded correctly [13]. The ThDD binds to E1 and E3 by virtue of the PSBD without interference from the N-terminal LD [11]. The capacity of mutant PSBDs to bind to wild-type E1 was investigated by means of nondenaturing polyacryl- amide gel electrophoresis. The mutant ThDDs were incu- bated with wild-type E1 in 20 m M potassium phosphate buffer, pH 7.0, for 5 min at room temperature, and the mixtures were then submitted to polyacrylamide gel electrophoresis under nondenaturing conditions (Fig. 2A). In every case, the E1 band was found to be retarded, indicating that all of the mutant PSBDs bound tightly to E1. However, such gel electrophoresis, while serving to show that none of the mutations in the PSBD prevented tight binding to E1, gives no quantitative information on the strength of the interaction. SPR analysis was therefore carried out to assess rate and dissociation constants for the interaction of mutant PSBDs with the wild-type E1 (a 2 b 2 ) tetramer. The mutant ThDDs were immobilized on a BIAcore CM5 sensor chip by attachment of the lipoylated LD in each instance, leaving the PSBD free to interact with E1, as described elsewhere [13]. The BIAcore sensor chip contained four flow cells in each of which a different ThDD could be immobilized independently. The wild-type ThDD was Fig. 1. Structure of the PSBD of B. stearothermophilus E2 and its interaction with E3. (A) Details of the interaction of the PSBD with E3 [8]. (B) Structure of the PSBD in a ribbon representation. The amino acid residues chosen for replacement by mutagenesis and chemical modification are highlighted, and portrayed in the ball-and-stick representation. The figure was produced using MOLSCRIPT [22]. 4490 H I. Jung et al.(Eur. J. Biochem. 270) Ó FEBS 2003 immobilized in the second flow cell as an internal reference, and the first flow cell was used as a blank. Representative SPR profiles for association of the wild-type E1 to the PSBD of immobilized wild-type and mutant ThDDs are shown in Fig. 2B. SPR response is measured in resonance units (RU). Experimental variation was minimized by expressing all the kinetic data as relative ratios in which the SPR response for a mutant domain immobilized in a given chip was divided by the response for the wild-type domain immobilized in the same chip. For the mutants M131A and R135A, the association of E1 reached a steady-state immediately and generated rectangular-shaped binding curves, which is a good indica- tor of weak binding (k off is high). In such circumstances, a large error may be incurred by the simultaneous measure- ment of k on and k off values with the typical fitting method in the BIAEVALUATION TM software (Pharmacia). Therefore, the equilibrium measurement method (Pharmacia Biosensor AB, Application note 301) was used, as described elsewhere [13,18]. The kinetic parameters for other mutants (S133A, K136A, R139A, R146A, K153A and R156A) were meas- ured using the BIAEVALUATION TM software, also as described elsewhere [11]. The results are summarized in Table 1. Alanine substitution at positions Met131 and Arg135 of the PSBD was found to lower the binding affinity for E1 by almost 140-fold, indicating that these residues are of major importance in the formation of the complex (Table 1). Given that the nonpolar residue, Met131, is not involved in the interaction with E3 [13], this observation strongly suggests that the E1-PSBD interaction differs significantly from the E3-PSBD interaction. In addition to Arg135, three positively charged residues (Lys136, Arg139 and Arg156) arealsoinvolvedintheinteractionwithE3[13].However, only the R156A mutation, and to a much lesser extent the R139A mutation, displayed major effects on the binding to E1 (Table 1). Fig. 2. Interaction of the PSBD and E1 ana- lysed by means of nondenaturing PAGE and SPR detection. (A) Wild-type and mutant ThDDs (each 100 pmol) were incubated with E1 (100 pmol of a 2 b 2 heterotetramer) in 20 m M potassium phosphate buffer, pH 7.0, at 25 °C, and samples of the mixtures were subjected to nondenaturing PAGE. Lane 1, Arg135 wild-type; lane 2, M131A; lane 3, S133A; lane 4, R135A; lane 5, K136A; lane 6, R139A; lane 7, R146A; lane 8, K153A; lane 9, R156A. (B) SPR sensorgrams of E1 binding to wild-type and mutant ThDDs. Wild-type E1 (a 2 b 2 ) was injected onto the sensor surface using a series of increasing concentrations (12.5, 25, 50, 100 and 200 n M ) and represen- tative sensorgrams at 50 n M are shown. The start of the association and dissociation phases are indicated by the arrows a and d, respect- ively. Wt, wild-type. Table 1. Kinetic and thermodynamic parameters for the interaction of PSBD mutants with E1. Kinetic data were determined by SPR analysis. All measurements were made in HBS buffer (10 m M Hepes, 150 m M NaCl, 3.4 m M EDTA), pH 7.4 at 25 °C, as in Materials and methods. The kinetic parameters for the binding of the wild-type PSBD to E1 are k on ¼ 3.27 · 10 6 M )1 Æs )1 , k off ¼ 1.06 · 10 )3 s )1 , K d ¼ 3.24 · 10 )10 M )1 [11]. The standard errors on kinetic parameters were less than 5% except for mutants R135A and M131A (£ 10%). DG° ¼ –RTlnK d ,whereR¼ 1.987 calÆmol )1 ÆK )1 and T ¼ 298K. NM, not measurable by ITC; mut, mutant; wt, wild-type. PSBD k on (mut)/ k on (wt) k off (mut)/ k off (wt) K d (mut)/ K d (wt) DG° (kcalÆmol )1 ) Arg135 (wt) 1.0 1.0 1.0 )12.9 M131A NM NM 142 a )10.0 S133A 1.0 1.0 1.0 )12.9 R135A NM NM 140 a )10.0 K136A 1.0 3.2 3.2 )12.2 R139A 1.0 6.1 6.1 )11.8 R146A 1.0 1.9 1.9 )12.6 K153A 1.0 2.1 2.1 )12.5 R156A 1.0 18.7 18.7 )11.2 a Obtained by equilibrium measurement method. Ó FEBS 2003 Thermodynamics of protein complex assembly (Eur. J. Biochem. 270) 4491 For the other amino acid residues tested (Ser133, Arg146 and Lys153), the sensorgrams for the mutants were essentially identical to that of the wild-type PSBD and no significant effect on the binding to E1 was detected (Table 1). In summary, values of the dissociation constant (K d ) for the mutant PSBDs interacting with E1 were in the order M131A ‡ R135A  R156A > R139A > (K136A, K153A, R146A, S133A). Indeed, the last four residues in this list contributed virtually nothing to the interaction. Chemical modifications of Arg135 in the ThDD In order to dissect the particular contributions of the positively charged guanidino group and the aliphatic side- chain of Arg135 of the PSBD, this Arg residue in a recombinant di-domain (i.e. LD plus PSBD), was first replaced with cysteine by site-directed mutagenesis. The newly introduced thiol group was then chemically modified by alkylation (i.e. R135C -methyl, -ethyl, -propyl and -butyl), as described in Materials and methods. The thiol group of cysteine residues reacts rapidly with alkyl halides, such as methyl, ethyl, propyl and butyl iodides, to give the corresponding stable alkyl derivatives: As the PSBD contains no cysteine residue in its primary structure, Cys135 created by mutagenesis (R135C) is a unique site for the subsequent chemical modification. The LD has one cysteine residue at position 37, but this residue is completely buried inside the protein [19]. Thus, when the wild-type ThDD was treated with alkyl iodides, no chemical modification could be detected by mass spectrometry (data not shown). Likewise, when the alkylated forms of the mutant ThDD (R135C) were examined, only one alkyl group was found to have been added in each instance (data not shown). Binding constants for Arg135 variants of the PSBD From Tables 2 and 3, the dissociation constants for the interaction of the PSBD with E3 and E1, determined by means of ITC or SPR, do not seem to be influenced greatly by the incremental changes in the length of the aliphatic side-chain of the Arg135 variants (especially bearing in mind the experimental uncertainties in determining K d values in this region by ITC). The R135K mutant behaved most like the wild-type. This suggests that electrostatic interactions are the dominant factor in the binding affinity. In each instance, when the dissociation constants for each mutant at 25 °Cand37°C were compared, no major difference was observed. Taken together, these data support the view that a Table 2. Thermodynamic parameters for the interaction of E3 with PSBD variants having different side-chain lengths at position 135. The numbers in parentheses are the standard errors from repeated measurements. NM, not measurable by ITC; ND, not determined; wt, wild-type. Variant Side-chain K d a (25 °C) ( · 10 )8 M ) K d a (37 °C) ( · 10 )8 M ) DH(25 °C) (kcalÆmol )1 ) DH(37 °C) (kcalÆmol )1 ) DC p (calÆmol )1 ÆK )1 ) R135A (–CH 3 ) 2.0 (7.0 b ) 1.6 )2.6 (0.1) )8.4 (0.1) )483 R135C-methyl (–CH 2 SCH 3 ) 1.2 2.0 )2.2 (0.4) )7.4 (0.5) )433 R135C-ethyl (–CH 2 SCH 2 CH 3 ) 1.6 1.2 )3.7 (0.2) )8.9 (0.5) )433 R135C-propyl (–CH 2 SCH 2 CH 2 CH 3 ) 1.5 1.3 )2.8 (0.4) )8.2 (0.5) )450 R135C-butyl (–CH 2 SCH 2 CH 2 CH 2 CH 3 ) 1.2 1.5 )2.8 (0.2) )8.0 (0.4) )433 R135M (–CH 2 CH 2 SCH 3 ) 2.3 1.3 )2.7 (0.4) )7.2 (0.5) )375 R135L (–CH 2 CH(CH 3 )(CH 3 )) NM 1.9 0 )4.6 (0.6) )383 R135K (–CH 2 CH 2 CH 2 CH 2 NH 3 + ) 0.6 b ND 0 )4.0 (0.7) )333 Arg135 (wt) (–CH 2 CH 2 CH 2 NHC(¼ NH)NH 3 + ) 0.06 b ND +2.2 (0.1) )1.8 (0.3) )316 c a Determined by means of ITC. b Determined by means of SPR. c Taken from previous results [12,13]. Table 3. Thermodynamic parameters for the interaction of E1 with PSBD variants having different side-chain lengths at position 135. The numbers in parentheses are the standard errors from repeated measurements. NM, not measurable by ITC; ND, not determined; wt, wild-type. Variant Side-chain K d a (25 °C) ( · 10 )8 M ) K d a (37 °C) ( · 10 )8 M ) DH(25 °C) (kcalÆmol )1 ) DH(37 °C) (kcalÆmol )1 ) DC p (calÆmol )1 ÆK )1 ) R135A (–CH 3 ) 1.1 (4.5 b )ND )6.6 (0.2) )13.1 )542 R135C-methyl (–CH 2 SCH 3 ) 1.5 1.3 )5.9 (0.5) )11.1 (0.4) )433 R135C-ethyl (–CH 2 SCH 2 CH 3 ) 1.9 1.0 )4.5 (0.5) )8.0 (0.6) )292 R135C-propyl (–CH 2 SCH 2 CH 2 CH 3 ) 2.2 3.1 )3.1 (0.3) )8.1 (0.2) )417 R135C-butyl (–CH 2 SCH 2 CH 2 CH 2 CH 3 ) 1.8 7.1 )3.9 (0.6) )8.3 (0.1) )367 R135M (–CH 2 CH 2 SCH 3 ) 4.5 ND )2.6 (0.2) ) 7.5 )408 R135L (–CH 2 CH(CH 3 )(CH 3 )) 2.5 2.2 )3.8 (0.0) )8.6 (0.0) )400 R135K (–CH 2 CH 2 CH 2 CH 2 NH 3 + ) 0.4 b NM )2.6 (0.1) )7.7 (0.1) )425 Arg135 (wt) (–CH 2 CH 2 CH 2 NHC(¼ NH)NH 3 + ) 0.03 b NM )8.4 (0.1) )14.3 (0.4) )470 c a Determined by means of ITC. b Determined by means of SPR. c Taken from previous results [12,13]. 4492 H I. Jung et al.(Eur. J. Biochem. 270) Ó FEBS 2003 specific electrostatic interaction between Arg135 and negat- ively charged side-chains contributed by E3 [13] and E1 (see above) has a major influence on the binding affinity. Thermodynamic changes on binding Arg135 variants of the PSBD ITC experiments were performed at 25 °Cand37°Cto study the interaction of the variant ThDDs described above with E3 or E1, and the associated thermodynamic changes (DH°, DS° and DC p ) are compared in Tables 2,3 and 4. In all cases, the enthalpies and entropies of binding showed significant temperature-dependence (large negative DC p ), as described previously for the E3-PSBD interaction [12], but with enthalpy–entropy compensation resulting in relatively smaller changes in free energies of binding with temperature. Consequently, the absolute values of DH° and DS° under specific conditions are difficult to interpret, but the relative changes in these parameters observed for the variants under otherwise similar experimental conditions might be more tractable. For the binding of the PSBD variants to E3, the major changes in both standard Gibbs free energy of binding (DG°) and its component parts (DH° and TDS°) compared with the wild-type are associated with the loss of the positive charge on Arg135 (Table 2). In all cases, binding is predominantly entropy-driven (positive TDS°) with only a relatively small enthalpic component, which may be exothermic, endother- mic or even athermal (Table 4), depending on conditions. With the exception of the R135L mutant (for which ITC data could not be obtained), all the PSBD variants with an uncharged side-chain at position 135 displayed similar DH° and TDS° values regardless of the side-chain length. Com- pared with the wild-type PSBD, the reduction in binding free energy (DDG°) of around 2 kcalÆmol )1 can be seen to arise from a more favourable DH° (more exothermic by  5kcalÆmol )1 ) offset partly by less favourable reductions in TDS° of around 7 kcalÆmol )1 . The R135K mutant, which carries a positive charge at position 135, although slightly different from Arg135, shows roughly inter- mediate changes. With E1 the changes appear to follow a different pattern. Although the binding free energies are very similar to those seen with E3, again with a DDG° of  +2 kcalÆmol )1 compared with the wild-type PSBD, the separate DH° and TDS° contributions show much greater variability with respect to changes in the side-chain at position 135. In particular, and in contrast with E3, removal of the positive charge results in a less exothermic interaction, albeit to a lesser extent as the side-chain length is reduced. These unfavourable enthalpy changes are offset partially by more favourable (more positive) TDS° contributions, with a similar general trend to smaller effects associated with shorter side-chains. This difference in pattern is also seen in the heat capacity data. For the binding of PSBD variants to E3, the DC p values are consistently more negative than that found with the wild-type, and the effect is generally larger the shorter the length of the side-chain at position 135. In contrast, with the possible exception of the R135A mutant, DC p for binding of the PSBD to E1 is consistently less negative than for the wild-type. (Absolute values of DC p should be treated with caution because they are derived from measurements at only two temperatures.) Discussion The PSBD of the E2 chain in the B. stearothermophilus PDH complex is responsible for binding both the E1 and E3 components to the multimeric (60-mer) E2 core. One PSBD is capable of binding one E1 tetramer (a 2 b 2 ) or one E3 dimer but not both simultaneously [9–11]. The crystal structure of E3 bound to the PSBD [8], together with a combination of thermodynamic analyses and site-directed mutagenesis stud- ies [12,13], has provided detailed information about the binding interface in the E3-PSBD complex. Electrostatic attractions constitute the driving force for complex forma- tion and Arg135 of the PSBD is a key residue. Important subsidiary roles are played by Arg139 and Arg156. Although we lack a structure for the E1-PSBD complex, it is now clear from the results described above (Table 1) that the binding site on the PSBD for E1 has something in common with that for E3, not least in the major importance of Arg135 and the lesser importance of Arg156. However, Met131 is also exceptionally important in the interaction with E1, the M131A mutation causing a decrease in binding Table 4. Comparison of thermodynamic parameters for the interaction of E3 and E1 with PSBD variants having different side-chain lengths at position 135. All the values are at 25 °C. The units of DG°, DH° and TDS° are kcalÆmol )1 . DG° wascalculatedas-RTlnK d ,whereR ¼ 1.987 calÆmol )1 ÆK )1 , T ¼ 298K and all K d values are from Tables 2 and 3. TDS° was calculated as DH°–DG°.ValuesofDH° were determined by means of ITC. ND, not determined (Table 2); wt, wild-type. Variant Side-chain E3 E1 DG° DH° TDS° DG° DH° TDS° R135A (–CH 3 ) )10.5 )2.6 +7.9 )10.9 )6.6 +4.3 R135C-methyl (–CH 2 SCH 3 ) )10.8 )2.2 +8.6 )10.7 )5.9 +4.8 R135C-ethyl (–CH 2 SCH 2 CH 3 ) )10.6 )3.7 +6.9 )10.5 )4.5 +6.0 R135C-propyl (–CH 2 SCH 2 CH 2 CH 3 ) )10.7 )2.8 +7.9 )10.4 )3.1 +7.3 R135C-butyl (–CH 2 SCH 2 CH 2 CH 2 CH 3 ) )10.8 )2.8 +8.0 )10.6 )3.9 +6.7 R135M (–CH 2 CH 2 SCH 3 ) )10.4 )2.7 +7.7 )10.0 )2.6 +7.4 R135L (–CH 2 CH(CH 3 )(CH 3 )) ND 0 ND )10.4 )3.8 +6.6 R135K (–CH 2 CH 2 CH 2 CH 2 NH 3 + ) )11.2 0 +11.2 )11.5 )2.6 +8.9 Arg135 (wt) (–CH 2 CH 2 CH 2 NH 2 C(¼ NH)NH 3 + ) )12.6 +2.2 +14.8 )12.9 )8.4 +4.5 Ó FEBS 2003 Thermodynamics of protein complex assembly (Eur. J. Biochem. 270) 4493 affinity similar to that observed for R135A. The M131A mutation has no effect on the binding of E3 [13]. Thus, the binding sites for E1 and E3, although at least overlapping, are also likely to be significantly different. Moreover, the involvement of Met131 suggests that hydrophobic inter- actions make a large contribution to the binding of E1, in contrast with the electrostatic interactions that dominate E3 binding. These results are consistent with the earlier speculation, based on the favourable enthalpy change and modest entropy change, that a mixture of electrostatic and hydrophobic interactions drives the E1-PSBD complex formation [12]. With the exception of Met131, the residues identified as important for the binding to E1 (Arg135, Arg156) and to E3 (Arg135, Arg139, Arg156) are located in helix 1 and loop 2 of the PSBD (Fig. 1B). The residue Met131 is located at the C-terminal end of the linker region between the LD and the PSBD, just before the beginning of helix 1 in the PSBD [6]. Preceding Met131 in the linker sequence, there is a series of hydrophobic residues (e.g. Val128, Ile129 and Ala130) as well as charged residues (Arg126 and Arg127). To explore their participation, Arg126, Arg127, Val128 and Ile129 were replaced by alanine and the effects on the affinity of PSBD for E1 were monitored by SPR analysis (data not shown). None of these mutations changed the binding constants from that of the wild-type PSBD. Thus, we can rule out the possibility that the linker region in general interacts significantly with E1; Met131 appears to be the residue principally involved. The differences between the binding sites for E1 and E3 on the PSBD are further highlighted by the changes in the thermodynamic parameters in response to mutation and chemical modification at position Arg135. As listed in Tables 2 and 3, and replotted for convenience in Fig. 3 in terms of DH° values, the enthalpy changes varied comparatively little with the different side-chains introduced at this position until a positive charge was included. In the case of E3, the value changed to zero (R135K) or became positive (Arg135, wild-type). With E1, the R135K mutation again led to a less negative DH°, but the interaction with the wild-type domain was substantially the most exothermic. The importance of the positive charge at Arg135 for the interaction with E3 (Table 2) and E1 (Tables 1 and 3) is abundantly clear. The thermodynamic parameters for the interaction of E3 ([12,13] and above) and E1 ([12] and above) with the various forms of the PSBD are summarized in Table 4. For E3 these can be interpreted in detail, in terms of the origin of the entropy-driven interaction with the PSBD, and from a comparison with the X-ray crystal structure of the E3-PSBD complex [12,13]. The discrepancy between the measured value of DC p for this interaction and that calculated from the buried polar and nonpolar surfaces in the interface has been discussed previously [12]. Once again, the particular importance of the positive charge on the side- chain at position 135 in the PSBD is evident from the data in Table 2. We are not in a position to carry out an equivalently detailed analysis for the E1-PSBD interaction, as we lack a crystal structure on which to base the calculations of buried surface area. However, it is clear from the data in Table 3 that including a positive charge on the side-chain at position 135 is not associated with a large change in DC p ; the one major curiosity is associated with R135C-ethyl and R135A, which cause DC p to change significantly in positive and negative directions, respectively. This remains to be explained. In mutagenesis studies, lysine is usually regarded as the best substitute for arginine in conserving the positively charged side-chain and many binding proteins appear to be unaffected by interchanging these residues. However, the dissociation equilibrium constant (K d ) of the R135K mutant for binding E3 (Table 2) and E1 (Table 3) was observed to Fig. 3. Effect of side-chain length and charge on the binding enthalpy for the interaction of PSBD variants with E3 and E1. All measure- ments were made in HBS buffer (10 m M Hepes, 150 m M NaCl, 3.4 m M EDTA), pH 7.4 at 25 °C. The side-chain lengths at position 135 were calculated from the sum of the indi- vidual bond lengths using values of 1.54 A ˚ and 1.82 A ˚ for C-C and C-S bonds, respect- ively. Terminal C-H and S-H bond lengths are not included. Enthalpy changes for the inter- actions with R135K and wild-type Arg135 PSBDs are highlighted. Solid line, interaction with E3; broken line, interaction with E1. 4494 H I. Jung et al.(Eur. J. Biochem. 270) Ó FEBS 2003 be about 10-fold higher than that of the wild-type PSBD. In fact, the side-chains of lysine and arginine are significantly different, both in length and in the point (lysine) and delocalized (arginine) positive charge. Unlike lysine, argi- nine is capable of multiple types of interaction; it has the ability to form both simple and bidentate ionic interactions with carboxyl groups and a hydrogen bond network with up to five hydrogen bonds. These properties account for the favoured occurrence of arginine residues at binding inter- faces. Indeed, in a database analysis of the contributions of individual amino acid residues to protein–protein inter- actions [20], arginine appears with a frequency of more than 10% (21% for tryptophan, 13.3% for arginine and 12.3% for tyrosine residues). This is borne out in the importance of arginine residues in the PSBD for the interactions with E1 andE3detailedabove. A striking feature of the thermodynamic data described here is that, although mutations and other side-chain replacements at Arg135 of the PSBD give rise to very similar changes in the free energies of binding to E1 and E3, the underlying pattern of enthalpy, entropy and heat capacity changes is significantly different in each case. This is in line with a general pattern of entropy–enthalpy compensation or Ôthermodynamic homeostasisÕ that is coming to be seen as commonplace in (macro)molecular interactions involving a multiplicity of weak interactions [21]. The structural or molecular basis for these differences is harder to understand. Electrostatic interactions appear to dominate the binding of E3 [8], whereas hydrophobic interactions may make a more substantial contribution to the binding of E1 (see above). Nonetheless, the positive charge of the side-chain of one particular residue, Arg135, in the PSBD of E2 assumes a particular significance in the thermodynamics, although not the kinetics, of both these competitive bimolecular inter- actions. There are useful lessons to be learned here for the rational design of protein–protein binding interfaces, which will be amplified when a crystal structure of the E1-PSBD complex is determined. Acknowledgements This work was supported in part by a research grant (to R.N.P.) from the Biotechnology and Biological Sciences Research Council. We are grateful to the BBSRC and the Wellcome Trust for their support of the core facilities in the Cambridge Centre for Molecular Recognition, and to the BBSRC and the Engineering and Physical Sciences Research Council for funding the biological microcalorimetry facilities in the University of Glasgow. We thank the Cambridge Overseas Trust, the Department of Biochemistry and St John’s College, Cambridge for financial support to H.I.J. References 1. deKok,A.,Hengveld,A.F.,Martin,A.&Westphal,A.H.(1998) The pyruvate dehydrogenase multi-enzyme complex from Gram- negative bacteria. Biochim. Biophys. Acta 1385, 353–366. 2. Perham, R.N. (2000) Swinging arms and swinging domains in multifunctional enzymes: catalytic machines for multistep reac- tions. Annu.Rev.Biochem.69, 961–1004. 3. Zhou, H.Z., McCarthy, D.B., O’Connor, C.M., Reed, L.J. & Stoops, J.K. (2001) The remarkable structural and functional organization of the eukaryotic pyruvate dehydrogenase complex. Proc.NatlAcad.Sci.USA98, 14802–14807. 4. Milne, J.L.S., Shi, D., Rosenthal, P.B., Sunshine, J.S., Domingo, G.J., Wu, X., Brooks, B.R., Perham, R.N., Henderson, R. & Subramaniam, S. (2002) Molecular architecture and mechanism of an icosahedral pyruvate dehydrogenase complex: a multi- functional catalytic machine. EMBO J. 21, 1–12. 5. Borges, A., Hawkins, C.F., Packman, L.C. & Perham, R.N. (1990) Cloning and sequence analysis of the genes encoding the dihydrolipoyl acetyltransferase and dihydrolipoamide dehydro- genase components of the pyruvate dehydrogenase multienzyme complex of Bacillus stearothermophilus. Eur. J. Biochem. 194, 95–102. 6. Izard, T., Ævarsson, A., Allen, M.D., Westphal, A.H., Perham, R.N., de Kok, A. & Hol, W.G.J. (1999) Priciples of quasi- equivalence and Euclidean geometry govern the assembly of cubic and dodecahedral cores of pyruvate dehydrogenase complexes. Proc.NatlAcad.Sci.USA96, 1240–1245. 7. Kalia, Y.N., Brocklehurst, S.M., Hipps, D.S., Appella, E., Sak- aguchi, K. & Perham, R.N. (1993) The high resolution structure of the peripheral subunit-binding domain of dihydrolipoamide acetyltransferase from the pyruvate dehydrogenase multienzyme complex of Bacillus stearothermophilus. J. Mol. Biol. 230, 323–341. 8. Mande, S.S., Sarfaty, S., Allen, M.D., Perham, R.N. & Hol, W.G.J. (1996) Protein–protein interactions in the pyruvate dehy- drogenase multienzyme complex: dihydrolipoamide dehydroge- nase complexed with the binding domain of dihydrolipoamide acetyltransferase. Structure 4, 277–286. 9. Hipps, D.S., Packman, L.C., Allen, M.D., Fuller, C., Sakaguchi, K.,Appella,E.&Perham,R.N.(1994)Theperipheralsubunit- binding domain of the dihydrolipoyl acetyltransferase component of the pyruvate dehydrogenase complex of Bacillus stearo- thermophilus: preparation and characterization of its binding to the dihydrolipoyl dehydrogenase component. Biochem. J. 297, 137–143. 10. Lessard, I.A.D. & Perham, R.N. (1995) Interaction of component enzymes with the peripheral subunit-binding domain of the pyruvate dehydrogenase multienzyme complex of Bacillus stearo- thermophilus: stoichiometry and specificity in self-assembly. Bio- chem. J. 306, 727–773. 11. Lessard, I.A.D., Fuller, C. & Perham, R.N. (1996) Competitive interaction of component enzymes with the peripheral subunit- binding domain of the pyruvate dehydrogenase multienzyme complex of Bacillus stearothermophilus: kinetic analysis using surface plasmon resonance detection. Biochemistry 35, 16863– 16870. 12. Jung, H.I., Bowden, S.J., Cooper, A. & Perham, R.N. (2002) Thermodynamic analysis of the binding of component enzymes in the assembly of the pyruvate dehydrogenase multienzyme complex of Bacillus stearothermophilus. Protein Sci. 11, 1091–1100. 13. Jung, H.I., Cooper, A. & Perham, R.N. (2002) Identification of key amino acid residues in the assembly of enzyme into the pyruvate dehydrogenase complex of Bacillus stearo- thermophilus: a kinetic and thermodynamic analysis. Biochemistry 41, 10446–10453. 14. Hasan, Z. & Leatherbarrow, R.J. (1998) A study of the specificity of barley chymotrypsin inhibitor 2 by cysteine engineering of the P1 residue. Biochim. Biophys. Acta 1384, 325–334. 15. Fries, M., Chauhan, H.J., Domingo, G.J., Jung, H I. & Perham, R.N. (2003) Site-directed mutagenesis of a loop at the active site of E1 (a 2 b 2 ) of the pyruvate dehydrogenase complex. A possible common sequence motif. Eur. J. Biochem. 270, 861–870. 16. Wallis, N.G., Allen, M.D., Broadhurst, R.W., Lessard, I.A.D. & Perham, R.N. (1996) Recognition of a surface loop of the lipoyl domain underlies substrate channelling in the pyruvate dehy- drogenase multienzyme complex. J. Mol. Biol. 263, 463–474. 17. Lessard, I.A.D., Domingo, G.J., Borges, A. & Perham, R.N. (1998) Expression of genes encoding the E2 and E3 components of Ó FEBS 2003 Thermodynamics of protein complex assembly (Eur. J. Biochem. 270) 4495 the Bacillus stearothermophilus pyruvate dehydrogenase complex and the stoichiometry of subunit interaction in assembly in vitro. Eur. J. Biochem. 258, 491–501. 18. Karlsson, R., Michaelsson, A. & Mattsson, L. (1991) Kinetic analysis of monoclonal antibody–antigen interactions with a new biosensor based analytical system. J. Immunol. Methods 145, 229–240. 19. Dardel, F., Davis, A.L., Laue, E.D. & Perham, R.N. (1993) Three- dimensional structure of the lipoyl domain from Bacillus stearo- thermophilus pyruvate dehydrogenase multienzyme complex. J. Mol. Biol. 229, 1037–1048. 20. Bogan, A.A. & Thorn, K.S. (1998) Anatomy of hot spots in protein interfaces. J. Mol. Biol. 280,1–9. 21. Cooper,A.,Johnson,C.M.,Lakey,J.H.&No ¨ llmann, M. (2001) Heat does not come in different colours: entropy–enthalpy com- pensation, free energy windows, quantum confinement, pressure perturbation calorimetry, solvation and the multiple causes of heat capacity effects in biomolecular interactions. Biophys. Chem. 93, 215–230. 22. Kraulis, P. (1991) MolScript, a program to produce both detailed and schematic plots of protein structures. J. Appl. Crystallog. 24, 946–950. 4496 H I. Jung et al.(Eur. J. Biochem. 270) Ó FEBS 2003 . Interactions of the peripheral subunit-binding domain of the dihydrolipoyl acetyltransferase component in the assembly of the pyruvate dehydrogenase multienzyme complex of Bacillus stearothermophilus Hyo-Il. Competitive interaction of component enzymes with the peripheral subunit- binding domain of the pyruvate dehydrogenase multienzyme complex of Bacillus stearothermophilus: kinetic analysis using surface. Perham, R.N. (2002) Thermodynamic analysis of the binding of component enzymes in the assembly of the pyruvate dehydrogenase multienzyme complex of Bacillus stearothermophilus. Protein Sci. 11, 1091–1100. 13.