Article No. jmbi.1999.3391 available online at http://www.idealibrary.com on J. Mol. Biol. (2000) 295, 1023±1037

Protein-protein Interaction Revealed by NMR T2

Relaxation Experiments: The Lipoyl Domain and E1Component of the Pyruvate DehydrogenaseMultienzyme Complex of Bacillus stearothermophilus

Mark J. Howard, Hitesh J. Chauhan, Gonzalo J. DomingoChristopher Fuller and Richard N. Perham*

Cambridge Centre forMolecular RecognitionDepartment of BiochemistryUniversity of Cambridge, 80Tennis Court Road, CambridgeCB2 1GA, UK

The ®rst two authors have contriPresent address: M. J. Howard, M

Engineering, University Chemical LAbbreviations used: BCDH, bran

dimensional; DSS, 4,4-dimethyl-4-siE1b, beta subunit of E1p; E1p, pyrudihydrolipoyl dehydrogenase; HSQdehydrogenase; NOESY, nuclear Opyruvate dehydrogenase; T2 relaxatspectroscopy.

E-mail address of the correspond

0022-2836/00/041023±15 $35.00/0

T2 relaxation experiments in combination with chemical shift and site-directed mutagenesis data were used to identify sites involved in weakbut speci®c protein-protein interactions in the pyruvate dehydrogenasemultienzyme complex of Bacillus stearothermophilus. The pyruvate decar-boxylase component, a heterotetramer E1(a2b2), is responsible for the ®rstcommitted and irreversible catalytic step. The accompanying reductiveacetylation of the lipoyl group attached to the dihydrolipoyl acetyltrans-ferase (E2) component involves weak, transient but speci®c interactionsbetween E1 and the lipoyl domain of the E2 polypeptide chain. The inter-actions between the free lipoyl domain (9 kDa) and free E1a (41 kDa),E1b (35 kDa) and intact E1a2b2 (152 kDa) components, all the products ofgenes or sub-genes over-expressed in Escherichia coli, were investigatedusing heteronuclear 2D NMR spectroscopy. The experiments were con-ducted with uniformly 15N-labeled lipoyl domain and unlabeled E1 com-ponents. Major contact points on the lipoyl domain were identi®ed fromchanges in the backbone 15N spin-spin relaxation time in the presenceand absence of E1(a2b2) or its individual E1a or E1b components.Although the E1a subunit houses the sequence motif associated with theessential cofactor, thiamin diphosphate, recognition of the lipoyl domainwas distributed over sites in both E1a and E1b. A single point mutation(N40A) on the lipoyl domain signi®cantly reduces its ability to be reduc-tively acetylated by the cognate E1. None the less, the N40A mutantdomain appears to interact with E1 similarly to the wild-type domain.This suggests that the lipoyl group of the N40A lipoyl domain is notbeing presented to E1 in the correct orientation, owing perhaps to slightperturbations in the lipoyl domain structure, especially in the lipoyl-lysine b-turn region, as indicated by chemical shift data. Interaction withE1 and subsequent reductive acetylation are not necessarily coupled.

# 2000 Academic Press

Keywords: pyruvate dehydrogenase; lipoic acid; multienzyme complex;protein-protein interaction; T2 relaxation

*Corresponding author

buted equally to the work in this paper.RC Unit for Protein Function and Design, Cambridge Centre for Protein

aboratories, Lens®eld Road, Cambridge CB2 1EW, UK.ched chain 2-oxo acid dehydrogenase; BSA, bovine serum albumin; 2D, two-lapentane-1-sulfonate; E1, 2-oxo acid decarboxylase; E1a, alpha subunit of E1p;vate decarboxylase of PDH complex; E2p, dihydrolipoyl acetyltransferase; E3,C, heteronuclear single quantum correlation; OGDH, 2-oxoglutarateverhaÈuser effect spectroscopy; PAGE, polyacrylamide gel electrophoresis; PDH,ion, spin-spin relaxation; ThDP, thiamin diphosphate; TOCSY, total correlation

ing author: r.n.perham@bioc.cam.ac.uk

# 2000 Academic Press

1024 Interaction of Lipoyl Domain and E1

Introduction

The 2-oxo acid dehydrogenase multienzymecomplexes consist of multiple copies of three com-ponent enzymes that bring about the oxidativedecarboxylation of 2-oxo acids and the transfer ofthe resulting acyl group to coenzyme A. The threeenzymes of the pyruvate dehydrogenase (PDH)complex are pyruvate decarboxylase (E1p; EC1.2.4.1), dihydrolipoyl acetyltransferase (E2p; EC2.3.1.12) and dihydrolipoyl dehydrogenase (E3; EC1.8.1.4). The E1p component, a thiamin dipho-sphate-dependent enzyme, is responsible for theoxidative decarboxylation of the pyruvate and sub-sequent reductive acetylation of a lipoyl groupcovalently bound to the N6-amino group of a lysineresidue in the E2p component. E2p forms the struc-tural core (octahedral or icosahedral, depending onthe source) of the multienzyme complex and isalso responsible for transferring the acetyl groupto coenzyme A. The residual dihydrolipoyl moietyis reoxidized by E3 to regenerate its dithiolane ringto permit the next round of reductive acetylation.Comparable enzymes are found in the 2-oxogluta-rate dehydrogenase (OGDH) and branched chain2-oxo acid dehydrogenase (BCDH) complexes (forreviews, see Perham, 1991; Mattevi et al., 1992a;Patel et al., 1996; Berg & de Kok, 1997).

As in all icosahedral PDH complexes, the E1 ofBacillus stearothermophilus PDH complex consists oftwo E1a (41 kDa) and two E1b (35 kDa) polypep-tide chains assembled to form an active (a2b2)heterotetramer of Mr 152,000 (Lessard & Perham,1994). The E1a chain houses the conserved thiamindiphosphate-binding site (Stepp & Reed, 1985;Hawkins et al., 1989), whereas the E1b chain isresponsible for the tight non-covalent binding ofE1p to E2p (Wynn et al., 1992; Lessard & Perham,1995; Lessard et al., 1996). The B. stearothermophilusE2 chain consists of three domains separated by¯exible linker regions: an N-terminal lipoyl domain(9 kDa) which hosts the speci®c lysine residue(Lys42) to which the lipoyl group is attached; amuch smaller peripheral subunit-binding domain(4 kDa), responsible for the tight but non-covalentbinding of E1p and E3; and a C-terminal acetyl-transferase domain (28 kDa) responsible for assem-bly of the 60-mer icosahedral inner core andcatalytic transfer of the acetyl group to coenzymeA (Reed & Hackert, 1990; Perham, 1991; Berg & deKok, 1997).

The three-dimensional solution structures of theB. stearothermophilus PDH lipoyl domain (Dardelet al., 1993) and lipoyl domains from several other2-oxo acid dehydrogenase complexes (Green et al.,1995; Ricaud et al., 1996; Berg et al., 1996, 1997;Howard et al., 1998) have been solved by means ofNMR spectroscopy. In all instances, the domain isa ¯attened b-barrel formed from two four-strandedb-sheets with a 2-fold axis of quasi-symmetry, asshown in Figure 1. The lipoyl-lysine residue islocated in an exposed b-turn in one of the b-sheetsand this structural feature is essential for correct

post-translational modi®cation by the lipoylatingenzyme (Wallis & Perham, 1994). There are alsothree-dimensional solution structures for the per-ipheral subunit-binding domain of the B. stearother-mophilus PDH (Kalia et al., 1993) and Escherichia coliOGDH (Robien et al., 1992) complexes. Structuresof the octahedral (24-mer) cores of the PDH com-plex of Azotobacter vinelandii (Mattevi et al., 1992b)and the OGDH complex of E. coli (Knapp et al.,1998) and the icosahedral (60-mer) cores of B. stear-othermophilus and Enterococcus faecalis PDH com-plexes (Izard et al., 1999) have been determined bymeans of X-ray crystallography, as have structuresof several E3 dimers (Mattevi et al., 1991, 1992c,1993) including that of B. stearothermophilus E3complexed to the peripheral subunit-bindingdomain of E2p (Mande et al., 1996). A crystal struc-ture of the homologous E1(a2b2) component of thebranched chain 2-oxo acid dehydrogenase complexof Pseudomonas putida has only recently been pub-lished (ávarsson et al., 1999).

The lipoyl domain plays a key role in the cataly-tic mechanism, in particular in its interaction withthe E1 component. Reductive acetylation of thelipoyl group is the rate-limiting step of the overallPDH complex reaction (Danson et al., 1978; Cateet al., 1980; Berg et al., 1998). Free lipoic acid orlipoamide can function as substrate for the E2pand E3 but not the E1p of the E. coli PDH complex(Reed et al., 1958); E1p requires the lipoyl group tobe attached to a folded lipoyl domain, whichcauses kcat/Km to be raised by a factor of 104

(Graham et al., 1989). In addition, E1 is found to behighly speci®c for the lipoyl domain of its cognateE2 chain for both the PDH and OGDH complexesof E. coli (Graham et al., 1989) and the PDH com-plex of A. vinelandii (Schulze et al., 1992; Berg et al.,1998). This forms the molecular basis of an elegantsystem of substrate channeling in the complexes(Perham, 1991).

In the B. stearothermophilus PDH complex, theresidues that ¯ank the lipoyl-lysine residue at pos-ition 42 in the lipoyl domain are important for thereductive acetylation of the lipoyl group by the E1component (Wallis & Perham, 1994). The threeresidues form a DKA motif that is widely con-served in the lipoyl-lysine b-turn in lipoyl domainsfrom various PDH complexes (Dardel et al., 1993;Green et al., 1995). Another prominent feature ofthe lipoyl domain (Figure 1) is a surface loop (resi-dues 7 to 15 in B. stearothermophilus E2p), whichlies close in space to the lipoyl-lysine b-turn. Thenumber and type of amino acids within this loopvary between species (Wallis et al., 1996; Dardel etal., 1993) and it has no counterpart in the sym-metry-related half of the lipoyl domain (Dardelet al., 1993; Green et al., 1995). The loop appears tobe important for ef®cient reductive acetylation ofthe nearby lipoyl-lysine residue (Wallis et al., 1996)but it has been shown not to be the sole determi-nant of speci®city (Berg et al., 1998; Jones et al.,1999).

Figure 1. Three-dimensionalstructure of the lipoyl domain ofthe B. stearothermophilus PDH com-plex. The structure consists of a¯attened b-barrel formed by twoanti-parallel b-sheets (indicated indark blue and light blue) related bya 2-fold axis of quasi-symmetry,each sheet containing three majorand one minor strand (Dardel et al.,1993). The surface loop betweenstrands 1 and 2 is shown in redand the type-I b-turn betweenstrands 4 and 5 is shown in green.The lysine residue (K42) that car-ries the lipoyl group is shown inball-and-stick form in an arbitraryconformation. The diagram wascreated using the programMOLSCRIPT (Kraulis, 1991).

Interaction of Lipoyl Domain and E1 1025

Here, we show how the measurement of theNMR T2 relaxation rates of 15N-labeled B. stear-othermophilus lipoyl domain in the absence and pre-sence of E1(a2b2) can be used to probe the transientcontact surface of the lipoyl domain with the E1component. In agreement with earlier NMR exper-iments based on chemical shifts (Wallis et al., 1996),the importance of residues in the surface loop adja-cent to the lipoyl-lysine turn is apparent, but othercontact regions are indicated that were notrevealed by analysis of chemical shifts alone. Wedemonstrate that the lipoyl domain interacts withthe individual E1a and E1b components,suggesting that its recognition site in E1 lies at theinterface between the E1a and E1b subunits in theE1(a2b2) heterotetramer. Further, the diminishedability of the N40A mutant lipoyl domain to bereductively acetylated (Wallis et al., 1996) is shownhere not to be due to any signi®cant disruption ofthe interaction between the lipoyl domain and E1,but more likely to be the result of a small confor-mational change in the b-turn region of the domainin which the lipoyl-lysine residue is displayed.

Results

Theoretical background

The interaction between the lipoyl domain andits partner E1 is known to be weak with an equili-brium dissociation constant, Ks, in the millimolarrange (Graham & Perham, 1990) despite an appar-ent Km in the micromolar range for the lipoyldomain as a substrate for E1p in the E. coli PDHcomplex (Graham et al., 1989). However, inter-actions of the lipoyl domain with E1 should beobservable by means of NMR spectroscopythrough changes in 15N spin-spin relaxation rateconstants (T2) for the backbone nitrogen atoms onthe lipoyl domain that occur in the presence of E1.Obtaining meaningful 15N T2 rate constants relies

on the two interacting components being in slowexchange on the NMR T2 relaxation time-scale. Ifthe system is in intermediate or fast exchange withrespect to T2 relaxation, lipoyl domain resonanceswill be broadened beyond detection because ofexchange broadening owing to the much smallerand dominant T2 of the larger El component. Pre-vious studies (Wallis et al., 1996) showed that cer-tain residues of the lipoyl domain are in fastexchange with respect to chemical shift but, asmost lipoyl domain resonances are still observed, itis reasonable to assume that the system is in slowexchange with respect to T2. As the NMR exper-iment is observing a dynamic process, the 15N T2

values for the lipoyl domain in the presence of E1will be in¯uenced by chemical exchange and eachT2 will be modi®ed according to the relationship(Dwek, 1973; Jardetsky & Roberts, 1981; Lian &Roberts, 1993):

1

T2 interaction� 1

T2 free� 1

tlip�1�

where T2 free is the T2 of the lipoyl domain on itsown, T2 interaction is the measured T2 in the presenceof a second component, and tlip is the lifetime ofthe lipoyl domain species in the free state in thepresence of the second component. The applica-bility of this general equation to the interaction ofthe lipoyl domain with E1 was supported by theobservation that an increased concentration of E1led to a fall in the average value of T2 across thedomain backbone, commensurate with the shorterlifetime of the free lipoyl domain (data not shown).

On the basis of this equation, an interaction ofthe lipoyl domain with El might be expected toreduce all the observable T2 values to the sameextent, governed by the experimental conditionsthat de®ne tlip. In practice, we have observed thatcertain amino acid residues undergo greater

1026 Interaction of Lipoyl Domain and E1

changes in 15N T2 values than those of the bulkbackbone. We attribute this to contact with E1.

Since the completion of our work describedbelow, Wagner and colleagues have also reportedthe identi®cation of contact residues on a smallprotein interacting with a much larger one usingNMR T2 relaxation measurements (Matsuo et al.,1999). The theoretical principles of NMR two-sitechemical exchange have been studied extensively(Abragam, 1961; Rogers & Woodbrey, 1962;SundstroÈm, 1982) and the origin of effects observedin the 15N NMR spectrum of a small proteindomain that interacts with a large macromoleculehave been described. Rogers & Woodbrey (1962)derived simple expressions for the calculation ofNMR line shapes for two-site exchange:

v / ÿP 1� t

pB

T2A� pA

T2B

� �� ��QR

� �P2 � R2

�2�

where:

dn � nA ÿ nB

�n � 0:5�nA � nB� ÿ n

P � t1

T2A � T2Bÿ 4p2�n2 � p2�dn�2

� �� pA

T2A� pB

T2B

Q � t�2p�nÿ pdn�pA ÿ pB��

R � 2p�n 1� t1

T2A� 1

T2B

� �� �

� pdnt1

T2Bÿ 1

T2A

� �� pdn�pA ÿ pB�

t � pA

kB� pB

kA

For simplicity it can be considered that the twostates A and B de®ne a small protein domain inthe free state and bound to a large macromolecule,respectively. Consequently, T2A, T2B de®ne theeffective transverse relaxation times; pA, pB de®nethe populations; nA, nB de®ne the chemical shiftfrequencies for the domain in the free (A) andbound (B) state, respectively; and kA and kB arethe association (kon) and dissociation (koff) rateconstants, respectively.

Two distinct situations can be considered for asystem involving a small isotopically labeled pro-tein interacting with a large macromolecule. First,where chemical shift differences upon binding arevery small (550 Hz), line-broadening effectsowing to chemical shift become negligible andonly occur from the relaxation rate of the residuein the bound state. Alternatively, where chemicalshift differences upon binding are large (>500 Hz),

the line broadening and hence observed T2 aredependent on the exchange rate in addition to therelaxation rate of the residue in the bound state.Thus, NMR signals of residues that form bindingcontacts associated with large chemical shift per-turbations have two sources of exchange broaden-ing, as opposed to one for those residues that donot undergo environmental changes upon binding.Such exchange broadening will be evident via 15NNMR T2 relaxation measurements such as thosepresented here. Further exchange broadening ofcontact residues can also be achieved via secondaryprocesses of conformational change at the contactsite. This would create an effective three-siteexchange mechanism for these residues, which canbe described by modi®ed three-site Blochequations (SundstroÈm, 1982). The analysis of suchexpressions is somewhat similar to those for two-site exchange summarized above and describedelsewhere (Matsuo et al., 1999), except that eachcontact residue would now be de®ned in twostates with small T2 values and one state with stan-dard T2 values. As before, residues that undergolarge environmental changes on binding andconformational exchange will experience the great-est line-broadening effects and hence exhibit thesmallest T2 values.

Data analysis

15NH chemical shifts for the B. stearothermophiluslipoyl domain uniformly labeled with 15N weretaken from those previously recorded in the struc-ture determination (Dardel et al., 1993). Changescaused by temperature and pH differences weremeasured using a suite of HSQC-NOESY andHSQC-TOCSY experiments at pH 7.0 and at tem-peratures of 298, 308 and 318 K.

The preferred method for comparing spin-spin relaxation data between the free and interact-ing lipoyl domain is based on normalizedratios. The normalized ratio is de®ned as [T2 free/T2 interaction] ÿ 1. Thus a normalized ratio of zero fora particular residue indicates no change in T2

between free and interacting domain, and a nor-malized ratio of 1.0 indicates a T2 which duringinteraction is half that observed in the free state. T2

values were derived as outlined in Materials andMethods.

The normalized ratios of 15N spin-spin relaxationtimes for the B. stearothermophilus E2p lipoyldomain in the presence of E1a, E1b and E1(a2b2)are shown as bar charts in Figure 2(a) to (c). Nor-malized spin-spin relaxation ratios derived from acontrol experiment of lipoyl domain mixed withbovine serum albumin (BSA), chosen as a proteinwith which it does not speci®cally interact, areshown in Figure 2(d).

The threshold value used to identify major con-tact points in each interaction experiment wasde®ned as the average normalized T2 ratio forthe system plus two standard deviations (2 sn ÿ 1).The average normalized T2 ratio and the standard

Interaction of Lipoyl Domain and E1 1027

deviation for each experiment were calculatedusing all normalized ratios with values less than1.0, on the premise that normalized ratios >1.0 areindicative of a contact point and that theirinclusion would bias the average to a higher value.The threshold values obtained using this de®nitionare summarized in Table 1 and shown as brokenlines in Figure 2(a) to (d) (see also Figure 6).Normalized T2 ratios greater than the de®nedthreshold value were taken as identifying aminoacid residues at points of signi®cant contactbetween the two proteins (Figure 3). The differencein the average normalized T2 ratios betweenexperiments is most probably a re¯ection of thedifferent lifetimes of the lipoyl domain in separateexperiments.

In each interaction experiment, a small numberof resonances were no longer detectable. Thesewere derived from residues Leu6, Gly79 andAsn82 in the E1a experiment; Gly12, Gly79and Asn82 in the E1b experiment; and Gly79and Asn82 in the E1a2b2 experiment and theN40A/E1a2b2 experiment. As these resonancesare clearly detectable in the spectrum of the freelipoyl domain and remain so in the mixturewith BSA, they must be experiencing a fast con-formational change with respect to T2 in the pre-sence of the various E1 components and thusbecoming line-broadened beyond detection.Therefore, it is likely that these residues aremajor contact points.

Estimates of the van der Waals surface area ofthe lipoyl domain involved in the interaction withE1(a2b2), E1a and E1b were made for the struc-tured region (residues 1-80) of the domain usingNaccess (Hubbard & Thornton, 1993). These®gures do not include the unstructured C-terminalregion, which is the beginning of the ¯exible inter-domain linker region in E2 (Dardel et al., 1993).

Interaction of the lipoyl domain with E1aaa

As indicated in Figure 2(a), seven major contactpoints were detected between the lipoyl domainand free E1a: residues Leu6, Ile9, Glu11, Val51,Thr74, Gly79 and Asn82. Of these residues, Leu6and Val51 are also major contact points in theinteraction with E1(a2b2); and Ile9, Glu11, Gly79and Asn82 are common contact points for theinteraction with E1a, E1b and E1a2b2 (Figure 3).The residues interacting with E1a all map onto oneface of the three-dimensional structure of the lipoyl

Table 1. Statistical values for the normalized T2 data calcula

Mean

E1a/lipoyl domain 0.215E1b/lipoyl domain 0.386E1a2b2/lipoyl domain 0.567E1a2b2/N40A lipoyl domain 0.465BSA/lipoyl domain 0.127

domain (Figure 4(a)), and comprise 8 % of the sur-face area of the domain.

Interaction of the lipoyl domain with E1bbb

As indicated in Figure 2(b), 12 residues (Ile9,Gly10, Glu11, Gly12, Gly26, Gln39, Val45, Leu59,Gln70, Gly79, Asn82 and Phe85) were identi®ed asmajor contact points for the lipoyl domain withfree E1b. The majority of these residues makingcontact with E1b coincide with those interactingwith E1a2b2, with the exceptions of Gly10, Gly12,Gly26 and Gln70 (Figure 3). Nine of the 12 residueslie on the same face of the domain (Ile9, Gly10,Glu11, Gly12, Val45, Gln70, Gly79, Asn82 andPhe85), with the other three (Gly26, Gln39 andLeu59) close to the ``periphery'' of the oppositeside (Figure 4(b)). Ideally this interaction exper-iment should be performed with the dimer of E1b,since this is its natural state in intact E1a2b2.However, free E1b has been shown to exist as adistribution of monomers, dimers, tetramers andhexamers, depending on the protein concentration(Lessard & Perham, 1994). Under the experimentalconditions used in the present work, E1b will existpredominantly as a tetramer. Thus it is possiblethat some contacts with the lipoyl domain are arte-factual, owing to the higher quaternary structureof E1b. Residues identi®ed as likely contact pointswith E1b comprise 11 % of the surface area of thelipoyl domain.

Interaction of the lipoyl domain with E1(aaa2bbb2)

As indicated in Figure 2(c), 13 residues (Leu6,Ile9, Glu11, Gln39, Val45, Glu46, Val51, Lys52,Leu59, Gly79, Asn82, Thr84 and Phe85) were ident-i®ed as major contact points for the lipoyl domainwith E1a2b2. Of these, Glu46, Lys52 and Thr84 areexclusive to the E1a2b2 interaction; four residues(Gln39, Val45, Leu59 and Phe85) are common tothe E1b interaction only; and Leu6 and Val51 arecommon to the interaction with E1a only (Figure 3).The interaction appears to extend over one face ofthe lipoyl domain, with residues Gln39 and Leu59being contacts close to the periphery of the oppo-site side of the molecule (Figure 4(c)). Interestingly,residues in the ¯exible, C-terminal tail of the lipoyldomain (Asn82, Thr84 and Phe85) appear to beinvolved in the interaction with E1a2b2. The resi-dues identi®ed as likely major contact points withE1a2b2 comprise 14 % of the surface area of thelipoyl domain.

ted as described in the text

Standard deviation Threshold

0.195 0.6050.221 0.8280.164 0.8950.155 0.7750.145 0.417

Figure 2 (legend opposite)

1028 Interaction of Lipoyl Domain and E1

Figure 2. Interaction of the lipoyl domain with E1. Bar charts of normalized T2 ratio against residue number for theB. stearothermophilus lipoyl domain in the presence of (a) E1a, (b) E1b, (c) E1a2b2 and (d) BSA. The threshold valuederived as de®ned in the text is indicated as a broken line on the charts. Bars shown in dark grey are for thoseresidues that have HN-N correlations that are line-broadened beyond detection.

Interaction of Lipoyl Domain and E1 1029

Figure 3. Amino acid sequence of the B. stearothermophilus lipoyl domain indicating all contact residues with E1a,E1b and E1a2b2. Residues in red are those which are inferred to make major contacts. The horizontal arrows representthe eight b-strands in the structured region of the domain. Strands 1, 3, 6 and 8 form one b-sheet; strands 2, 4, 5 and7 form the other. The lipoyl-lysine residue is at position 42.

1030 Interaction of Lipoyl Domain and E1

Interaction of the lipoyl domain with BSA

In the presence of BSA an overall drop in T2 wasexperienced by the lipoyl domain, but this wasrelatively small compared with the falls in T2

observed in the presence of the various E1 com-ponents (Figure 2(d)). Further, the changes in T2

were much more limited and distributed over theentire amino acid sequence, with the exception ofIle9 and Glu11. This may be due to the resonancesassociated with these two residues overlapping inthe 2D NMR spectra, causing complications whenmeasuring peak intensities. This point must be con-sidered with all 15N T2 data sets.

Comparison of 15N spin-spin relaxation timedata and chemical shift data for the interactionof the lipoyl domain with E1(aaa2bbb2)

In earlier NMR experiments, changes in chemicalshift were used to identify contact points betweenE1a2b2 and the lipoyl domain (Wallis et al., 1996).In those experiments, residues 7-15 comprising theentire surface loop between b-strands 1 and 2 wereobserved to undergo signi®cant chemical shifts,whereas we ®nd that only residues 9 and 11 revealchanges in T2. This could be due to the resonancesassociated with these two residues overlapping inthe 2D NMR spectra (see above). Likewise, in theregion of the exposed b-turn that houses the lipoyl-lysine residue (residues 41-43), where Asn40 andAsp41 experience chemical shifts in the presence ofE1a2b2 (Wallis et al., 1996), residues Gln39 andVal45 (which are closest to this b-turn) display achange in T2. Similarly, residues in the ¯exibleregion at the C-terminal end of the lipoyl domainshow changes in T2 in the presence of E1a2b2,whereas they exhibited no detectable change in

chemical shift under comparable conditions (Walliset al., 1996).

Interaction of the N40A mutant lipoyl domainwith E1(aaa2bbb2)

A single replacement of Asn40 with alanine(N40A) in the B. stearothermophilus lipoyl domainsigni®cantly reduces the ability of E1 to recognizethe mutant domain as a substrate for reductiveacetylation (Wallis et al. 1996). 15N-labeled N40Alipoyl domain was prepared in order to determinethe reason for this loss in activity. A comparisonbetween the 15N-1H HSQC spectra of the nativelipoyl domain and the N40A mutant domainshowed that the single point mutation caused sig-ni®cant changes in chemical shift for several resi-dues (Figure 5), particularly in the lipoyl-lysine b-turn (residues 38 to 44) and in the ¯exible loopregion (residues 11 to 19). However, the N40Alipoyl domain interacted with E1(a2b2) very simi-larly to the wild-type lipoyl domain, but with anoverall lower average normalized T2 ratio(Figure 6). These results suggest that the inter-action between the mutant N40A lipoyl domainand E1 is essentially normal, but that it does notlead to ef®cient reductive acetylation of the pro-tein-bound lipoyl group. This is probably the resultof a small conformational change, especially in thelocal structure of the lipoyl-lysine b-turn, whichprevents the lipoyl-lysine side-chain from beingpresented correctly.

Discussion

The 15N T2 values for individual residues in theB. stearothermophilus lipoyl domain in the presenceand absence of E1a, E1b and E1(a2b2) have been

Figure 4. Space-®lling diagrams of the B. stearothermophilus lipoyl domain highlighting all residues which de®ne acontact point with (a) E1a, (b) E1b or (c) E1a2b2. The color coding is the same as in Figure 3. The left-hand sideshows the domain from the front (same orientation as in Figure 1); the right-hand side shows the domain from theback (rotated by 180 � about a horizontal axis). The diagrams were created using MOLMOL (Koradi et al., 1996).

Interaction of Lipoyl Domain and E1 1031

determined, as well as those for the N40A lipoyldomain in the presence of E1(a2b2). Residues thatdisplayed changes in normalized 15N T2 ratiosgreater than an arbitrary, high threshold valuewere identi®ed as potential major contact points.As a control, the experiment was repeated withBSA in place of E1 or E1a and E1b, there being

no speci®c interaction between it and the lipoyldomain. Whereas relatively large changes in theT2 ratios were observed for individual residues ofthe lipoyl domain in the presence of E1a, E1band E1(a2b2), only a small but widespread dropin T2 values was noted for the lipoyl domain inthe presence of BSA. The relatively small changes

Figure 5. Changes in chemical shifts on the B. stearothermophilus lipoyl domain as a result of the single pointmutation, N40A. (a) Bar chart of changes in chemical shift �av (ppm) determined from the 15N-1H HSQC spectra ofthe N40A and wild-type lipoyl domains.

�av ������������������������������������������f��HN�2 � ��N=7�2g

qwhere �HN and �N correspond to the chemical shifts of amide protons and nitrogen, respectively. The averagechange in chemical shift (0.09 ppm) is indicated by the horizontal broken line. (b) Space-®lling diagram of the wild-type lipoyl domain with residues exhibiting chemical shift changes above the average highlighted in red. Themutated residue, Asn40, is indicated in yellow. The orientation of the lipoyl domain is the same as for Figure 1.

1032 Interaction of Lipoyl Domain and E1

in 15N T2 induced by BSA are most likely due toa difference in sample viscosity.

The results give an insight into the molecularbasis of recognition of the lipoyl domain by E1. In

the E1a2b2 heterotetramer, the sequence motifinvolved in the binding of thiamin diphosphateessential for decarboxylation, is in the E1a subunit(Stepp & Reed, 1985; Hawkins et al., 1989). The

Figure 6. Interaction of the N40A lipoyl domain with E1. Bar chart of normalized T2 ratio against residue numberfor the B. stearothermophilus lipoyl domain in the presence of E1a2b2. Bars are shown for the interaction of E1 with theN40A mutant (dark grey) and the wild-type (light grey) lipoyl domain for comparison. The threshold value derivedas de®ned in the text is indicated as a dotted line for the wild-type lipoyl domain, and as a broken line for the N40Amutant.

Interaction of Lipoyl Domain and E1 1033

E1b subunit is responsible for the non-covalent buttight binding of E1 to the E2 core (Wynn et al.,1992; Lessard & Perham, 1995). The NMR exper-iments described above indicate that the lipoyldomain makes speci®c contact with both the E1aand the E1b subunits in the a2b2 heterotetramer.Thirteen residues on the surface of the lipoyldomain were identi®ed as potential contact siteswith E1(a2b2), namely, Leu6, Ile9, Glu11, Gln39,Val45, Glu46, Val51, Lys52, Leu59, Gly79, Asn82,Thr84 and Phe85. Six of these (Leu6, Ile9, Glu11,Val51, Gly79 and Asn82) were also identi®ed ascontact sites with E1a and eight (Ile9, Glu11,Gln39, Val45, Leu59, Gly79, Asn82 and Phe85)with E1b. Differences in the contacts for E1a2b2

compared with the separate E1a and E1b com-ponents, may re¯ect conformational changes thatthe individual subunits undergo on association toform the E1a2b2 heterotetramer. In the case of E1b,there may also be ``unnatural'' contact sites arisingfrom the interaction of the lipoyl domain with apredominantly tetrameric form of E1b instead ofthe dimeric species. Nonetheless, it is reasonable toconclude that the lipoyl domain is recognized bysites on the E1a and E1b subunits which, given therelative sizes of the lipoyl domain and E1, couldwell lie on the ab-subunit interface in the E1a2b2

heterotetramer. These results are entirely consistentwith the new structure of the homologous E1(a2b2)of P. putida branched chain 2-oxo acid dehydrogen-

ase complex, which reveals the four subunits in atightly-packed arrangement, with the b2-dimersandwiched between the two a-subunits. Eachactive site is at the bottom of a 20 AÊ -long funnel-shaped channel, lined by residues from both thea- and b-subunits (ávarsson et al., 1999). Dockingexperiments of E1 and the lipoyl domain, informedby our NMR data, can now be attempted tounravel the mechanism of the interaction and itslinkage to reductive acetylation.

Several features in the NMR data describedabove indicate that the lipoyl domain interactsmore strongly with E1a2b2 than with the individ-ual E1a or E1b subunits, and that the larger contri-bution to this molecular recognition process maybe made by E1b. This is implied by the relativemean change in T2 for the three interaction exper-iments (Table 1), the number of residues on thelipoyl domain interacting exclusively with eachcomponent and with E1 (Figure 3), and the relativesurface area involved in the interaction in the threedifferent systems (8 %, 11 % and 14 % of the lipoyldomain with E1a, E1b and E1a2b2, respectively).

A comparison of the 15N spin-spin relaxationtime data (see above) with the previously pub-lished chemical shift and mutagenesis data (Walliset al., 1996) highlights the need for complementaryexperimental approaches in attempting to delineatemolecular recognition surfaces. Although thechemical shift data might be taken to imply that

1034 Interaction of Lipoyl Domain and E1

the whole surface loop (residues 7-15) forms a rec-ognition determinant, site-directed mutagenesis ofresidue E15, for example, led to no loss in reduc-tive acetylation by E1 (Wallis et al., 1996). This isconsistent with the new T2 data, which indicate nocontact with E1 beyond residue Glu11. Likewise,the T2 changes highlight Gln39 as the contact pointclosest to Asn40 Nd2, a site in the residue precedingthe DKA motif in the lipoyl-lysine b-turn thatdisplays a signi®cant chemical shift on interactionwith E1. From the work presented above, it is clearthat the surface loop makes important contactswith E1 but is not alone in de®ning the recog-nition. Other evidence leading to the same con-clusion has been obtained for the lipoyl domain ofthe A. vinelandii (Berg et al., 1998) and E. coli (Joneset al., 1999) 2-oxo acid dehydrogenase complexes.

A single point mutation, N40A, signi®cantlyreduces the ability of the lipoyl domain to act as asubstrate for reductive acetylation (Wallis et al.,1996). The 15N-1H HSQC spectrum of the N40Amutant domain exhibits signi®cant changes inchemical shift for several residues (Figure 5),suggesting that a conformational change accompa-nies the N40A mutation. However, the N40Amutant lipoyl domain interacts with E1 in a similarmanner to the wild-type lipoyl domain, as judgedby NMR spectroscopy (Figure 6). Inspection of thesolution structure of the wild-type lipoyl domain(Dardel et al., 1993) shows that Asn40 may beinvolved in two intramolecular interactions: ahydrophobic interaction with Ile13 of the nearbysurface loop, and a hydrogen bond with Asp41.This residue may thus form a structural linkbetween two major regions on the lipoyl domain.It would appear that protein-protein interactionand subsequent reductive acetylation are notnecessarily coupled.

Finally, it should be noted that NMR exper-iments such as those described above can indicatecontact points between two proteins, but that it isnot safe to conclude that all such contacts are cru-cial for speci®city in recognition and/or any sub-sequent biological effect. Other studies, includingdirected mutagenesis will be required to dissect thepart played by individual residues, as used byWallis et al. (1996) and others (Jones et al., 1999).Nonetheless, the mapping of the contact site offersan important starting point in such work andshould help characterize point mutations in the E1components of patients with pyruvate dehydro-genase de®ciency (Tripatara et al., 1999). Moreover,the T2 experiments described above can beexpected to ®nd applications in the analysis ofother transient interactions.

Materials and Methods

Chemicals

Milli-Q2 and deionized water was used for all exper-iments unless otherwise stated. For most purposes,reagents of analytical grade or the purest available were

employed. All chemicals were obtained from SigmaChemical Co. Ltd, Poole, UK unless otherwise stated.Ampicillin was from Beecham Research Laboratories,Brentford, UK. Kanamycin sulfate was from Calbiochem-Novabiochem Ltd, Nottingham, UK. Tryptone was fromUnipath Ltd, Basingstoke, UK. Yeast extract was fromBeta Lab, West Molesey, UK. EDTA was from GibcoBRL Life Technologies Ltd, Paisley, UK. Celtone1

(15N-labeled) was from Martek Biosciences Corporation,Columbia, USA.

Bacterial strains and plasmids

E. coli host strain TG1 recO [K12, D(lac-proAB), supE,thi, hsdD5, recO::Tn5 Kanr/F'traD36, proA�B�, lacIq,lacZDM15] was from Amersham International. PlasmidspKBstE1a and pKBstE1b expressing genes encoding theE1a and E1b subunits, respectively, of the E1 componentof the B. stearothermophilus PDH complex have beendescribed previously (Lessard & Perham, 1994). E. colihost strain BL21 (DE3) [Fÿ, ompT, hsdSa, (raÿ, maÿ), gal,dcm, (DE3)] (Studier & Moffatt, 1986) was from NovagenInc., Madison, USA. The plasmid pET11ThDD encodesthe thrombin-cleavable di-domain (ThDD); thiscomprises the ®rst 171 amino acid residues of theB. stearothermophilus E2p chain, with a thrombin-cleavagesite inserted in the linker region between the lipoyldomain and the peripheral subunit-binding domain(Wallis et al., 1996). The amino acid sequence of thelipoyl domain released from ThDD by thrombincleavage is shown in Figure 3.

Bacterial growth media

Media used to grow bacterial liquid cultures were LBmedium (10 g/l tryptone, 5 g/l yeast extract, 5 g/l NaCland 1 ml/l 1 M NaOH) and 2 � TY medium (16 g/ltryptone, 10 g/l yeast extract and 5 g/l NaCl). 15N-label-ing of proteins was achieved by growing bacterial cul-tures in a modi®ed K-Mops (3-(N-morpholino)-propanesulfonic acid) minimal medium (Neidhardt et al.,1974). The medium contained 40 mM K-Mops (pH 7.4),4 mM tricine (pH 7.4), 30 mM FeSO4, 0.84 mM Na2SO4,0.5 mM CaCl2, 1.5 mM MgCl2, 50 mM NaCl, 1 % Celtone(15N-labeled) and trace micronutrients. This wassterilized by autoclaving before adding the remainingingredients which were sterile ®ltered: 5.3 mM K2HPO4,0.4 % (w/v) glucose, 10 mM 15NH4Cl and 100 mg/mlampicillin.

Purification of E1(aaa2bbb2), E1aaa and E1bbb

B. stearothermophilus E1p and its constituent subunits(E1a and E1b) were puri®ed from E. coli cells trans-formed with plasmids pKBstE1a and pKBstE1b, asdescribed by Lessard & Perham (1994).

Purification of 15N-labeled lipoyl domain

E. coli strain BL21 (DE3) cells transformed withpET11ThDD were grown in modi®ed K-Mops minimalmedium. Puri®cation of the 15N-labeled lipoyl domain ofE2p from the ThDD was carried out as described byWallis et al. (1996).

Interaction of Lipoyl Domain and E1 1035

Purification of 15N-labeled N40A mutantlipoyl domain

The N40A mutant lipoyl domain is encoded in theplasmid pBSTNAV(N40A) described by Wallis et al.(1996). E. coli strain TG1 recO cells transformed withpBSTNAV(N40A) were grown in modi®ed K-Mopsminimal medium. Puri®cation of the 15N-labeled N40Amutant lipoyl domain of E2p from the ThDD was carriedout as described by Wallis et al. (1996).

NMR spectroscopy

All samples were prepared in 20 mM potassiumphosphate buffer (pH 7) containing 20 mM DSS(4,4-dimethyl-4-silapentane-1-sulfonate) and 10 % 2H2O.The ®nal protein concentrations were 0.48 mM for thelipoyl domain and 0.08 mM for E1a2b2, E1a, E1b or BSA,in a ®nal volume of 0.6 ml.

All T2 measurements were carried out on a BrukerAM-500 spectrometer operating at 500 MHz 1Hfrequency. 15N T2 values were obtained using inversedetection experiments described by Barbato et al. (1992)which employed a spin echo sequence for the determi-nation of T2.

2D NMR spectra were recorded at 298 K, andadditional data were collected at 308 K and 318 K toresolve chemical shift ambiguity or overlap with the sol-vent. Data processing was carried out using the AZARApackage (W. Boucher, unpublished). All NMR spectrawere acquired in the phase-sensitive mode with quadra-ture detection in the F2 dimension and hypercomplexStates-TPPI method in F1 (Marion et al., 1989). The NMRcarrier frequency was placed on the 1H2O resonance,which was suppressed by presaturation, except for theHMQC experiments where suppression was achievedusing jump-return pulses. The receiver reference phaseand the delay between the opening of the receiver gateand acquisition of the ®rst data point were optimized toobtain a ¯at baseline (Marion & Bax, 1989; Hoult et al.,1983). Published pulse sequences were used to obtain theHSQC-NOESY, HSQC-TOCSY (Norwood et al., 1990),HMQC (Bax et al., 1990) and 15N T2 (Barbato et al., 1992)data, with delay values of 8.48, 25.44, 76.32, 127.2, 178.08and 228.96 ms. In the directly acquired dimension, thespectral width and acquisition time were 8064.52 Hz and0.128 second, respectively. In the indirectly acquireddimension, the spectral width and acquisition time were1450 Hz and 0.044 second for all 15N T2 experiments.

Acknowledgments

We are grateful to the Biotechnology and BiologicalSciences Research Council for a research grant (toR.N.P.) and the award of a BBSRC Special Research Stu-dentship (to H.J.C.). The core facilities of the CambridgeCentre for Molecular Recognition are supported by theBBSRC and The Wellcome Trust. We thank Drs R. W.Broadhurst and D. Neuhaus for stimulating discussionsregarding the use of T2 values to study weak proteininteractions.

References

Abragam, A. (1961). Principles of Nuclear Magnetism,Oxford University Press, New York.

ávarsson, A., Seger, K., Turley, S., Sokatch, J. R. & Hol,W. G. J. (1999). Crystal structure of 2-oxoisovaleratedehydrogenase and the architecture of 2-oxo aciddehydrogenase multienzyme complexes. NatureStruct. Biol. 6, 785-792.

Barbato, G., Ikura, M., Kay, L. E., Pastor, R. W. & Bax,A. (1992). Backbone dynamics of calmodulin stu-died by N-15 relaxation using inverse detected 2-dimensional NMR spectroscopy - the central helix is¯exible. Biochemistry, 31, 5269-5278.

Bax, A., Ikura, M., Kay, L. E., Torchia, D. A. &Tschudin, R. (1990). Comparison of different modesof 2-dimensional reverse-correlation NMR for thestudy of proteins. J. Magn. Reson. 86, 304-318.

Berg, A. & de Kok, A. (1997). 2-oxo acid dehydrogenasemultienzyme complexes. The central role of thelipoyl domain. Biol. Chem. 378, 617-634.

Berg, A., Vervoort, J. & de Kok, A. (1996). Solutionstructure of the lipoyl domain of the 2-oxoglutaratedehydrogenase complex from Azotobacter vinelandii.J. Mol. Biol. 261, 432-442.

Berg, A., Vervoort, J. & de Kok, A. (1997). Three-dimen-sional structure in solution of the N-terminal lipoyldomain of the pyruvate dehydrogenase complexfrom Azotobacter vinelandii. Eur. J. Biochem. 244, 352-360.

Berg, A., Westphal, A. H., Bosma, H. J. & de Kok, A.(1998). Kinetics and speci®city of reductive acyla-tion of wild-type and mutated lipoyl domains of 2-oxo-acid dehydrogenase complexes from Azotobactervinelandii. Eur. J. Biochem. 252, 45-50.

Cate, R. L., Roche, T. E. & Davis, L. C. (1980). Rapidintersite transfer of acetyl groups and movement ofpyruvate dehydrogenase components in the kidneypyruvate dehydrogenase complex. J. Biol. Chem.255, 7556-7562.

Danson, M. J., Fersht, A. R. & Perham, R. N. (1978).Rapid intramolecular coupling of active sites in thepyruvate dehydrogenase complex of Escherichia coli:mechanism for rate enhancement in a multimericstructure. Proc. Natl Acad. Sci. USA, 75, 5386-5390.

Dardel, F., Davis, A. L., Laue, E. D. & Perham, R. N.(1993). Three-dimensional structure of the lipoyldomain from Bacillus stearothermophilus pyruvatedehydrogenase multienzyme complex. J. Mol. Biol.229, 1037-1048.

Dwek, R. A. (1973). Nuclear Magnetic Resonance(N.M.R.). In Biochemistry, chapt. 2, Clarendon Press,Oxford, UK.

Graham, L. D. & Perham, R. N. (1990). Interactions oflipoyl domains with the Elp subunits of the pyru-vate dehydrogenase multienzyme complex fromEscherichia coli. FEBS Letters, 262, 241-244.

Graham, L. D., Packman, L. C. & Perham, R. N. (1989).Kinetics and speci®city of reductive acylation oflipoyl domains from 2-oxo acid dehydrogenasemultienzyme complexes. Biochemistry, 28, 1574-1581.

Green, J. D. F., Laue, E. D., Perham, R. N., Ali, S. T. &Guest, J. R. (1995). Three-dimensional structure of alipoyl domain from the dihydrolipoyl acetyltrans-ferase component of the pyruvate dehydrogenasemultienzyme complex of Escherichia coli. J. Mol. Biol.248, 328-343.

Hawkins, C. F., Borges, A. & Perham, R. N. (1989).A common structural motif in thiamin pyro-phosphate binding enzymes. FEBS Letters, 255, 77-82.

1036 Interaction of Lipoyl Domain and E1

Hoult, D. I., Chen, C. N., Eden, H. & Eden, M. (1983).Elimination of baseline artifacts in spectra and theirintegrals. J. Magn. Reson. 51, 110-117.

Howard, M. J., Fuller, C., Broadhurst, R. W., Perham,R. N., Tang, J. G., Quinn, J., Diamond, A. G. &Yeaman, S. J. (1998). Three-dimensional structure ofthe major autoantigen in primary biliary cirrhosis.Gastroenterology, 115, 139-146.

Hubbard, S. J. & Thornton, J. M. (1993). NACCESSComputer Program, Department of Biochemistry andMolecular Biology, University College London, UK.

Izard, T., ávarsson, A., Allen, M. D., Westphal, A. H.,Perham, R. N., de Kok, A. & Hol, W. G. J. (1999).Principles of quasi-equivalence and Euclidean geo-metry govern the assembly of cubic and dodecahe-dral cores of pyruvate dehydrogenase complexes.Proc. Natl Acad. Sci. USA, 96, 1240-1245.

Jardetsky, O. & Roberts, G. C. K. (1981). NMR in Mol-ecular Biology, chapt. 4, Academic Press, New York.

Jones, D. D., Horne, H. J., Reche, P. A. & Perham, R. N.(1999). Structural determinants of post-translationalmodi®cation and catalytic speci®city for the lipoyldomains of the pyruvate dehydrogenase multien-zyme complex of Escherichia coli. J. Mol. Biol. 295,289-306.

Kalia, Y. N., Brocklehurst, S. M., Hipps, D. S., Appella,E., Sakaguchi, K. & Perham, R. N. (1993). The highresolution structure of the peripheral subunit bind-ing domain of dihydrolipoamide acetyltransferasefrom the pyruvate dehydrogenase multienzymecomplex of Bacillus stearothermophilus. J. Mol. Biol.230, 323-341.

Knapp, J. E., Mitchell, D. T., Yazdi, M. A., Ernst, S. R.,Reed, L. J. & Hackert, M. L. (1998). Crystal structureof the truncated cubic core component of the Escher-ichia coli 2-oxoglutarate dehydrogenase multien-zyme complex. J. Mol. Biol. 280, 655-668.

Koradi, R., Billeter, M. & WuÈ thrich, K. (1996).MOLMOL: a program for display and analysis ofmacromolecular structures. J. Mol. Graph. 14, 51-55.

Kraulis, P. J. (1991). MOLSCRIPT: a program to produceboth detailed and schematic plots of protein struc-tures. J. Appl. Crystallog. 24, 946-950.

Lessard, I. A. D. & Perham, R. N. (1994). Expression inEscherichia coli of genes encoding the E1a and E1bsubunits of the pyruvate dehydrogenase complex ofBacillus stearothermophilus and assembly of a func-tional E1 component (a2b2) in vitro. J. Biol. Chem.269, 10378-10383.

Lessard, I. A. D. & Perham, R. N. (1995). Interaction ofcomponent enzymes with the peripheral subunit-binding domain of the pyruvate dehydrogenasemultienzyme complex of Bacillus stearothermophilus:stoichiometry and speci®city in self assembly. Bio-chem. J. 306, 727-733.

Lessard, I. A. D., Fuller, C. & Perham, R. N. (1996).Competitive interaction of component enzymeswith the peripheral subunit-binding domain of thepyruvate dehydrogenase multienzyme complex ofBacillus stearothermophilus: kinetic analysis using sur-face plasmon resonance detection. Biochemistry, 35,16863-16870.

Lian, L. & Roberts, G. C. K. (1993). NMR of Macromol-ecules, chapt. 6, IRL Press, Oxford, UK.

Mande, S. S., Sarfaty, S., Allen, M. D., Perham, R. N. &Hol, W. G. J. (1996). Protein-protein interactions inthe pyruvate dehydrogenase multienzyme complex:dihydrolipoamide dehydrogenase complexed with

the binding domain of dihydrolipoamide acetyl-transferase. Structure, 4, 277-286.

Marion, D. & Bax, A. (1989). Baseline correction of 2-DFT NMR-spectra using a simple linear predictionextrapolation of the time-domain data. J. Magn.Reson. 83, 205-211.

Marion, D., Ikura, M., Tschudin, R. & Bax, A. (1989).Rapid recording of 2-D NMR-spectra without phasecycling - application to the study of hydrogen-exchange in proteins. J. Magn. Reson. 85, 393-399.

Matsuo, H., Walters, K. J., Teruya, K., Tanaka, T.,Gassner, G. T., Lippard, S. J., Kyogoku, Y. &Wagner, G. (1999). Identi®cation by NMR spec-troscopy of residues at contact surfaces in large,slowly exchanging macromolecular complexes.J. Am. Chem. Soc. 121, 9903-9904.

Mattevi, A., Schierbeek, A. J. & Hol, W. G. J. (1991).Re®ned crystal structure of lipoamide dehydrogen-ase from Azotobacter vinelandii at 2.2 AÊ resolution: acomparison with the structure of glutathionereductase. J. Mol. Biol. 220, 975-994.

Mattevi, A., de Kok, A. & Perham, R. N. (1992a). Thepyruvate dehydrogenase multienzyme complex.Curr. Opin. Struct. Biol. 2, 877-887.

Mattevi, A., Obmolova, G., Schulze, E., Kalk, K. H.,Westphal, A. H., de Kok, A. & Hol, W. G. J.(1992b). Atomic structure of the cubic core of thepyruvate dehydrogenase multienzyme complex.Science, 255, 1544-1550.

Mattevi, A., Obmolova, G., Sokatch, J. R., Betzel, C. &Hol, W. G. J. (1992c). The re®ned crystal structureof Pseudomonas putida lipoamide dehydrogenasecomplexed with NAD� at 2.45 Angstrom resolution.Proteins: Struct. Funct. Genet. 13, 336-351.

Mattevi, A., Obmolova, G., Kalk, K. H., Westphal, A. H.,de Kok, A. & Hol, W. G. J. (1993). Re®ned crystal-structure of the catalytic domain of dihydrolipoyltransacetylase (E2p) from Azotobacter vinelandii at2.6 Angstrom resolution. J. Mol. Biol. 230, 1183-1199.

Neidhardt, F. C., Bloch, P. L. & Smith, D. F. (1974).Culture medium for enterobacteria. J. Bacteriol. 119,736-747.

Norwood, T. J., Boyd, J., Heritage, J. E., Soffe, N. &Campbell, I. D. (1990). Comparison of techniquesfor H-1-detected heteronuclear H-1 N-15 spec-troscopy. J. Magn. Reson. 87, 488-501.

Patel, M. S., Roche, T. E. & Harris, R. A. (1996). Alpha-Keto Acid Dehydrogenase Complexes, BirkhaÈuserVerlag, Basel.

Perham, R. N. (1991). Domains, motifs, and linkers in 2-oxo acid dehydrogenase multienzyme complexes: aparadigm in the design of a multifunctionalenzyme. Biochemistry, 30, 8501-8512.

Reed, L. J. & Hackert, M. L. (1990). Structure-function-relationships in dihydrolipoamide acyltransferases.J. Biol. Chem. 265, 8971-8974.

Reed, L. J., Koike, M., Levitch, M. E. & Leach, F. R.(1958). Studies on the nature and reactions of pro-tein bound lipoic acid. J. Biol. Chem. 232, 143-148.

Ricaud, P. M., Howard, M. J., Roberts, E. L., Broadhurst,R. W. & Perham, R. N. (1996). Three-dimensionalstructure of the lipoyl domain from the dihydroli-poyl succinyltransferase component of the 2-oxoglu-tarate dehydrogenase multienzyme complex ofEscherichia coli. J. Mol. Biol. 264, 179-190.

Robien, M. A., Clore, G. M., Omichinski, J. G., Perham,R. N., Appella, E., Sakaguchi, K. & Gronenborn,A. M. (1992). Three-dimensional solution structureof the E3 binding domain of the dihydrolipoamide

Interaction of Lipoyl Domain and E1 1037

succinyltransferase core from the 2-oxoglutaratedehydrogenase multienzyme complex of Escherichiacoli. Biochemistry, 31, 3463-3471.

Rogers, M. T. & Woodbury, J. C. (1962). A proton mag-netic resonance study of hindered internal rotationin some substituted N,N-dimethylamides. J. Phys.Chem. 66, 540-546.

Schulze, E., Westphal, A. H., Veeger, C. & de Kok, A.(1992). Reconstitution of pyruvate dehydrogenasemultienzyme complexes based on chimeric corestructures from Azotobacter vinelandii and Escherichiacoli. Eur. J. Biochem. 206, 427-435.

Stepp, L. R. & Reed, L. J. (1985). Active-site modi®cationof mammalian pyruvate dehydrogenase by pyri-doxal 50-phosphate. Biochemistry, 24, 7187-7191.

Studier, F. W. & Moffatt, B. A. (1986). Use of bacterio-phage T7 RNA polymerase to direct selective high-level expression of cloned genes. J. Mol. Biol. 189,113-130.

SundstroÈm, J. (1982). Dynamic NMR Spectroscopy, Aca-demic Press, London.

Tripatara, A., Korotchkina, L. G. & Patel, M. S. (1999).Characterization of point mutations in patients withpyruvate dehydrogenase de®ciency. Role of meth-ionine-181, proline-188 and arginine-349 in thealpha subunit. Arch. Biochem. Biophys. 367, 39-50.

Wallis, N. G. & Perham, R. N. (1994). Structural depen-dence of post-translational modi®cation and reduc-tive acetylation of the lipoyl domain of thepyruvate dehydrogenase multienzyme complex.J. Mol. Biol. 236, 209-216.

Wallis, N. G., Allen, M. D., Broadhurst, R. W., Lessard,I. A. D. & Perham, R. N. (1996). Recognition of asurface loop of the lipoyl domain underlies sub-strate channeling in the pyruvate dehydrogenasemultienzyme complex. J. Mol. Biol. 263, 463-474.

Wynn, R. M., Chuang, J. L., Davie, J. R., Fisher, C. W.,Hale, M. A., Cox, R. P. & Chuang, D. T. (1992).Cloning and expression in Escherichia coli of matureE1 beta subunit of bovine mitochondrial branched-chain alpha-keto acid dehydrogenase complex -mapping of the E1 beta binding region on E2. J. Biol.Chem. 267, 1881-1887.

Edited by P. E. Wright

(Received 16 August 1999; received in revised form 11 November 1999; accepted 12 November 1999)