J. Mol. Biol. (199‘1) 226. 867-882

Crystal Structure of Escherichia coli Malate Dehydrogenase A Complex of the fipoenzyme and Citrate at 1437A Resolution Michael D. Hall’, David G. Levitt2 and Leonard J. Banaszak’?

Tiniversity

‘Department of Biochemistry and ‘Department of Ph,ysiology of Minnesota, Minneapolix, MN 5.5455,

(Received

11 lkcember

79.91; nwepted 2 Apd

CT.A.A. 19.92)

The crystal structure of malate dehydrogenase from Escherichia coli has been determined with a resulting R-factor of @187 for X-ray data from 8.0 to 1.87 8. Molecular replacement, using the partially refined structure of porcine mitochondrial malate dehydrogenase as a probe, provided initial phases. The structure of this prokaryotic enzyme is closely homologous with the mitochondrial enzyme but, somewhat less similar to cytosolic malate dehydrogenase from eukaryotes. However. all three enzymes are dimeric and form the subunit’-subunit interface through similar surface regions. A citrate ion, found in the active sit,e, helps define the residues involved in substrate binding and catalysis. Two arginine residues, R81 and R153, interacting with the citrate are believed to confer substrate specificity. The hydroxyl of the citrate is hydrogen-bonded to a histidine, H177, and similar int8eractions could be assigned to a bound malate or oxaloacetate. Histidine 177 is also hydrogen-bonded to an aspartate, D150, to form a classic His. Asp pair. Studies of the act,ivr site cavity indicate that the bound citrate would occupy part of the site needed for t’he rornzymr. In a model building study, the cofactor. NTAD, was placed into the coenzyme site which exists when the citrat’e was converted to malate and crystallographic water molecules removed. This hypotheGca1 model of a ternary complex was energy minimized for comparison with the structure of the binary complex of porcine cytosolic malatr E. coli malate dehydrogenase. Many residues involved in cofactor binding in the minimized dehydrogenase structure are homologous to coenzymr binding residues in cytosolic malate dehydrogenase. In the energy minimized structure of the ternary complex, the C-4 atom of NAD is in van der Waals’ contact wit,h the (J-3 atom of the malatr. A catalytic cycle involves hydride transfer between these two at,oms.

Krywords:

malate

dehydrogenase:

citric

1. Introduction Malat,e dehydrogenase (MDHasef) catalyzes the reversible oxidat’ion of malate to oxaloacetate using NAT> as a cofactor. Tn prokaryotes there is only one form of t,he enzyme. Escherichia coli MDHase (eMDHase) is a homodimer in solution and is caomprised of 31% amino acid residues per subunit

t Author addressed.

to whom all rorrespondenre

should

he

1 Abbreviations used: MDHase, malate dehydrogenase: rMDHase. malate dehydrogenase from E. coli; mMDHaae, mitochondrial malate dehydrogenase: cMDHase. cytosolic malate dehydrogenase: LDHase, lactate

dehydrogenase:

r.m.s..

root-mean-square.

acid cycle: citrate:

dehydrogenase

(M(~Alister-Henn rt al., 1987: Sutherland & McAlister-Henn. 1985). Tn eukaryotic cells, at least two forms of t,hr enzyme can be found. One form is a principal enzvme of the citric acid cycle operating within the mttochondria. The other is found in the cytosol where it participates in the malatelaspartate shuttle. This shuttle exchanges reducing equivalents across the mitochondrial membranes in t’he form of malate/oxaloacetate rather t)han as NAD/NADH. A third isozyme can be found in the glyoxysomes of yeast, where it’ functions to convert malate that is produoed from glyoxylate in the glyoxylate cycle (Minard B McAlister-Henn, 1991). The crystal structures of cytoplasmic and mitochondrial malate dehydrogenases. both from porcine heart, have been determined (R&rick &

E. coli m-pig

E. coli m-pig

10 30 40 20 MKVAVLGAAGGIGQALALLLKTQLPSGSELSLYDIAPVTPGVAVDLSHIP A******XS***X*P*SX***NS-*LV*R*T*~*IA.H*****A*****E / -----aB----/ I--f$BT-I /-----aC--.---/ l-f=-I 60 90 70 80 TAVKIKGFSGEDATP.ALEGADWLISAGVRRKPGMDRSDLFNVNAGIVK *RATV**YL*PEQL*DC*KXC*""V*P***P***P**~**T~D****T**T**A I-aC'-1

I-W-I

m-pig

m-pig

m-pig

130

120

140

T*TAAC*QH**.D*M*C**S”“““S*IP*T**S*IP*T***F**H***NP**I*****

1----VW&

1BE I

F---m-

(

I-VI

I-

190 160 170 180 LDIIRSNTFVAELKGKQPGEVEVPVIGGHSGVTILPLLSQV.PGVSFTEQ ***V*A*A******XLD*AR*S**XX***A"R**A*K**I**I**CT*K*D*pQD

------a2F---E. coli

I--------aD&m

NLVQQVAKTCPK.ACIGIITNPVNTTVAIAAEVLKKAGVYDKNKLFGVTT ---------

E. coii

J-loop--j

16Dl

110 E. co/i

50

1

I-WI

I --

I --BH-I

240 230 210 220 EVADLTKRIQNAGTEWEAKAGGGSATLSMGQAAARFGLSLVRALQGEQG QLST**G***E***X**KX***A*******A~*G***VF***D*MN*KE* ---alG/a2G--------

250

(---------a3G-------I

/

260

270

280

290

WECAYVEGDGNYAR.. FFSQPLLLGKNGVEERKSIGTLSAFEQNALEGM CPY**T******K*I*KNLG**KI*P**EKMIAEA m--pig ****SF*KSQETD.. I__---_---[--@K---I I-P--I I PM- I

E, co/i

300 E. coli m-pig

310

LDTLKKDIALGQEFVNK.. IPE**AS*KK*E***KNMK -aH-------------j

Figure 1. The amino acid seyuenw of riClI)Hasr aligned with m?vll)Hase fiwm porcine heart. Sornendaturr for secondary structure description follows the syst,em adopted fhr LDHasrx and ?uIDHasrs (Rossmann et rrl.. 197:5). Elements of secondary struct,ure are indicated in the typiwl dehydrogenase manner with p indicaatine a p-strand and 2 signifying an a-helix. The secondary st,ructurr assignments were made according t,o rnlw of Kahsch & Sander (IW1). The ilmino acid sequence is numbered as in eMDHase. rn-pig. mlfDHase f’rom powine heart.

Banaszak, 1986; Birktoft et al., 19896). eMI)Haw has S9?;, identity with t’he mitochondrial form, as shown in the sequence alignment in Figure 1 (Honka et nl., 1990). Greater similarities are found among functionally significant residues. Residues involved in subunit interactions are 74y, identical and those involved in cofactor binding a,re (iSo, identical. when conservative substitutions are (‘orisidered, the overall homology between the mitochondrial enzyme and eM DHase approaches 7.i0,,, kleast mitoc:hondrial MMDHase (m?;lDHasr) has 4X0,, identity with eMl)Hase (Thompson rt ~rl., 1988). Severtheless, neither porcine mMl>Haw nor eMl)Hase has more t’han 20°, identity with t)he cytoplasmic enzyme. One other bacterial MI)Hasr. has been sequenced Thwrnus $nvus. from (Sishiyarna rt rl,l.. 1986). Interestingly rnough. it about do”, identity wit,h Gther shows only cMI)Hase or porcine mitochondrial M DHase. I~ut has X2 O, w&h porcine cytosolic, MDHasr / 0 identity (CM DHase). In eukarvotrs. mM DHase is heliered to exist a,s a wmplex with fumarase and cit.rats synthase in the wll (Kewkmans B Kanarek. 1981; Robinson of n/.. 1987). The teleological basis for complex formation is based on the fact) t,hat the standard free energy for the ,MDHasc reaction is unfavorable in the direction of the operation of the citric acid cbycole. I’nder t,he Jl I)Hasr rc>a&ion is standard conditions.

fa.vorrd in thr dirwtion of malate formation. It is Mieved that t’he formation of a citric ;lc*itl c.ycale or enzyme complex drives the cy(*Ie in metaholon the direction of oxaloacetate formation hy facilitat,ing the transfer of oxa1oacetat.e from MDHase to citrate synt)hase (Beeckmans & Kanarrk, 1981). Such weak enzyme-rnzymc complexes are t bought t.o occur in other pathways (&we. 19X7). For example, mJlDHase has ulso been shown to f’orm a c*omples \vith the mitochondrial form of aspartat,tL perhaps to facilitatcb t hr operaaminotratrsferase, tion of the rr~alate;~~slJartatf~ stlutt.lrb (Hee(:kmans & Kanarrk. 19%I ). In t,erms of catalytic mechanism. some M J)Hasrs Studies with mMIII)Hase have uniyuc properties. have shown that this enzyme is allosterically regurt al.. 19X2). lated (Teleadi et (a/.. 1973: Mullinax High cwncentrat~ions of mala,te stimulattx the product,ion of oxaloacetate. while high c~oncent,rations of oxaloacetate inhiM the reaction (Kernst,ein et al.. 197X: I)att,a rt nl.. 198fi: Fahien et nl.. t 988). (‘itrate also affects M DHase activity by stimulating oxaloawtate format,ion. ,L\ll three effect~trrs bind to t.he same fmtat,ive allosteric site (Mullinax 4 trl.. 19X2). (‘rystal structures of a number of’dehydrogenases have led t’o the conclusion that a similar open cow formation of’ the dinucleotide cwenzyme. SAD, is associated wit,h the bound form. The stereospecifit

869

Structure of E. coli Malate Dehydrogenase transfer of the hydride ion to either face of the nicotinamide ring depends on the syn versus anti orientation of the glycosidic bond leading to type A and type B dehydrogenases. The closest structural homology exists between subclasses. Both MDHases and lactate dehydrogenase (LDHase) are in the same subclass and are structurally homologous in both the NAD binding domain and the catalytic domain. The structural similarit,y between the MDHases and LDHases is presumably related t’o the similarity in the substrates; both catalyze the interconversion of a-hydroxyacids to u-ketoacids. Although the amino acid sequences of LDHase are only about 20% homologous with those of eMDHase and mMDHase, it is believed LDHases utilize t,he same catalyt,ic mechanism to reversibly oxidize lactate to pyruvate. Comparisons between LDHase and MDHase indicate that they have similar catalytic residues. Malate differs from lactate in that it is a dicarboxylic acid as indicated below:

cofactor in a homologous ture of eMDHase.

I

(“3H I 2

(“ooI HO(“H

YooHOC2-(‘5-(‘600-

I

I

C3H 2

C3Hz I

(“‘w-

malatr

&)Olactate

citrate

The carbon atoms are numbered as they will be referred to in the ensuing text. Here, we describe the refined structure of eMDHase at 1.87 ,% resolution. The crystalline enzyme contains a citrate ion bound in the active sit,e. Citrate is very similar in structure to malate as is shown above, and mimics malate in the active site without itself bring oxidized. More import)ant than the structural similarity, the interactions of citrate with residues in the active site clearly indicate that citrate is a substrate analog and point to all of the residues involved in the binding of substrate and some that are involved in catalytic function. The active site region of the MDHase-citrate complex resembles the structure of the active sit’e region of the t’ernary complex of porcine LDHase with S-lactate, covalent,ly linked to NAD, serving as a substrate analog (Grau et al., 1981). However, the eMDHase structure shows with greater clarity the residues specific to malate binding and to catalysis. Th e greatest degree of structural similarity among t,he dehydrogenases can be found in the NAD binding domain. In general, the dinucleotidr coenzyme binds to this domain in an open or extended conformat’ion. MDHases and LDHasen both belong to i.he type A group of dehydrogenases in which the sterrospecific transfer of the hydride takes place on {‘he A side of the nicotinamide ring. due to the anti orientat.ion of the glycosidic bond. The structural similarity between eMDHase and cMDHase. along w&h analysis of the cofactor binding cavity adjacent to the active site in eMDHasr. led us t’o attrmpt the docking of the

in the crystal

strut>-

2. Materials and Methods (a) Crystallization. Expression and purification of eMDHase. as well as cryst,allization conditions and X-ray diffraction data collected on eMDHase crystals have been described (Hall et al., 1991). Briefly, monoclinic C2 crystals were grown by vapor diffusion in 10 mlvl-sodium citrate buffer (pH 5.7) and polyethylene glycol. The unit cell dimensions are a = 116.9 !I. b = 42.9 -4, c = 841 .%, fi = 130.2”. There is I monomeric subunit in t,he asymmetric unit and 128,427 observations of 26,366 unique reflections were collected on a Siemens/lGcollet area detector from 4 ctrystais with an Rmergeof 5.996. The data set used in this study consists of 99”/,, of the data to 2.0 !I and 71% to 1.87 .$ resolution, with the upper limit, of reflections extending t,o l.‘i.5 8.

(b) (“OW I HOC”H

cavity

Molecular

rephceme7d

Molecular replacement techniques from the XPLOR program package were used to obtain initial phases (Brunger, 1990: Brunger & Krukowski. 1990). The first step was to perform a rot,ation function to orient the homologous structure or probe in the eMDHase unit cell. The partially refined structure of porcine heart mMDHase with an R-factor of 025 at 2.5 w resolution was used as the probe model. The range of the data used to calculate the Patt,erson maps for bot)h eMDHase and mMDHase was from 7.0 to 3.0 x resolution. Out of’ the 198 unique rotation function peaks detected. a single peak at 3140 above background was found. with the next highest peak at 1.92a. A model consisting of conserved residues from the probe. mMDHasr. but wit’h alaninr residues subs& tuted for all non-identical residues was also used in the rotation function and confirmed the init>ial solution. The unambiguous rotation function solution was then used for a rigid body refinement without the need for Patterson correlation refinement) as implement,ed in SPLOR. The next step was a translation search t.o find the Il,cation of t,he newly oriented probe structure in the unit cell. The original complete m1lDHasr model and data from 7.0 to 3.0 A$ were also used for the translation t’unc*tion. The translation search yielded an unequivocal solution at 8.3~ over background to give an init)ial X-fac%or of 0.483 from 8.0 to 2..5A resolution.

For refinement. the start,ing model c.onsist,rd of the conservtbd residues in mMDHase and alanine residues at the other positions. Xt positions where the residues are homologous. t.he smaller of t.he 2 f’rom cLit.her the eJlDHasr or the mJlDHase sequemae was chosen. Deletions and insertions in actcxortl with the fi. co/i MDHasr sequent Were carried out heforr beginning refinement The 1st’ step was rigid body rrfinetnrnt of the single subunit comprising the asymmetric unit. The E-factor decreased t,o 047; for data from 8.0 t.o 2.0 A. Sext. in order to more expeditiously correct major errors in the initial model. simulated annealing with YPLOK was applied to data rsnging from 8-O to PO 4 above 20. ;\ starting temperature of 4000 K with a tinlestep of

X70

.M. I). Null et al.

0-00025 ps gave optimal results ba,sed on Ii-factor decrease. Weighting of the ?( -ray contribution t,o the target or effective energy was determined hy the sbandard SPLOR protocol. After earh simulated annealing ru11. 2F0- I”‘~and YO- FC maps were inspected. and the model adjusted using a version of FRODO (,Jones. 1978) moditied by Dr Christian Cambilleau (IIniversit,y of hlarseillt~. Franc?). As the electron density became more int,erpretabie. thr correct side-chain atoms for eMDHase were added to the model. When no major errors remained in the structure, at an R-fact,or of 0.272, least-squares refinement with X PT,OR without molecular dynamics was used. ITnac:caounted-for electron density in the act,ive site region. similar in form to a citrate, was fitted wit,h a model of t,he Atrate ion. The presence of numerous hydrogen-bonding interact,ions between the citrate molecule and catalytic residues of t,he rMDHase further confirmed the existencne of this substrate analog in the structure. Wat,er molecules were eventually inserted into the structure in regions of electron densit,y having values ofa,t least P.T,a on fob- FC maps and 1.5a on PF,- Fc maps, providrd the water molecules were not less than a hvdrogrtl-bonding distance of 2.8 ip to neighboring atoms. -4fter all the csorrecatside-chains for rMl>Hase were in place and most of the water molecules were inserted, restrained temperature factor refinement was also carried out. The final st.ages of refinement involved interactive (*y&s of least-squares refinement. temperature factor refinement. watrbr insertion and model adjustment. Refinement was caonsidered finished when no correct,ablr errors were visi hle in the model. no further improvement in grometry should br made. and the R-factor would not clec*rrasr despitcl cahanges made to the (*o-ordinates.

An x,41) molecule with co-ordinates obtained from the 2.5 A resolution crystal structure of the cMDHasePKAI) binary complex (Birktoft rt al.. 1989a.6) was inserted with interactive graphics into t,he structure of eMDHase. In order to accommodate the KAD. atoms that were protruding into the ?r’Al) binding cavity were removed from the citrate molecule to create a malate moleculr. Water molecules occupying the NAD binding cavity w-err also removed. The hypothetical ternary complex of e~lI>HasePKADPmalate was submitted to energy minimization with XPLOR t,o optimize the fit of the cofactor and substrate in the protein. The energy minimization program was t.he XE’LOR positional refinement program without, the terms for X-ra>- energy and with thr sidrcahain cahargrs present. 1\ c,ontrol calculation for the eMDHase+SAI) studies was cbarried out t)y minimizing the cMDHase-NAI) structurr as IveIl. However. since this structure was refined using clonventiorml methods (Birktoft Pt al.. 1989a.b). t’hr c>lJjHase+P\‘AT) stxucturr was first. refined wit,h X F’LOR.. then minimized, to permit an unbiased comparison of final refined co-ordinat’es with energy minimized coordinates. Refinement was carried out in 2 parts. First. it B-as necessary to refine the structure with no hydrogen atoms present for 200 cycles. All X-ray data from 60 to 2.5 A were used. Hydrogen atoms were then added to the cMDHase- Ku’AD model and 200 more cycles were performed. The resulting co-ordinates were then used for energy minimization as it was carried out with eMDHaseNAD. Eo water molecules, however, were removed and thr sulfat,e ion in thr active site was retained.

3. Results and Discussion The conventional crystallographic K-factor. for eMIIHase is 18.7!& for all X-ray measurements from 8.0 to 1.75 !I resolution. When only data with IF,\ above ‘20 is used, t.hr R-factor is 18.5”,,. Xn analysis of t.he data based on a I,uzza,tti plot indicates that t,hc est,imatcd root,-mean-sqlral,~~ (r.m.s. error in atomic vo-ordinates is 0.2 A. The r.m.s. deviation of the geomet,ry terms from the c*anonical values as defined in SPIAIR is 0.015 A~1for bond lengt.hs. ‘t-73’ for bond angles. 2441’ for tlihedral irnprope’ angles. angles and 1.13” for Ramachandran plots of the 4 and II/ angles show t.hat all residues, except for most glycines. have angles within t,he acceptable range. Nineteen out of 36 glycine residues appear outside the l~ounds for the other amino acid residues and these include: (li. G27, G57. G69, C:78, (334. (:136, (:175. (:17ti. (:191, G219. GPBl. G244. (:%47. (:256. (:25x. c:271. (:274 rnodf~l and (:282. The refined c,rgstallogrrtI)~lic, includes 312 amino acid residues kno\vn from thr chemical sequence, one bound cit,rat,r rnolrc~ule and 94 water molecules. The (*o-ordinates will 1~ deposited in the Protein I)ata Bank (ac*caessiolb number for rMDHasr is I(:M13: that for t.hth et~erg> minimized ternary c*omplex is I CXF:). The electron density ~vas interpretable 1hroughout the map. Even in the earl>, maps, it was relatively easy to identify side-(*hains not) present. itI the initial model. [‘sing t,hr final caalculat,t:d phase angles, the agreement of the model with the rlertron density is ex‘ellent. as is shown in Figure 2(a) and (b). Thr hole in the tyrosine ring and the detail along the segment of polJ-peptide chain was characteristic of most of the final electron density map. I tI Figure 2(b), the etertron drnsity assoc,iated with the bound citrate ion is shown. The map permitted accurate posit,ioning of all three cbarhoxylate groups, two of which are the same as would be present in t,hr subst,rat,e malate. The only exceptions were the lack of clear electron density asso&ted with several side-chains located on the surface of t,he protein and near the last, two (I-terminal residues. The average temperature fact or fats the ent,ircs molecule is I X-8 AZ, t’ollrctivr t,erriprrat~urf~ t’acS0r.s for mairl-cbhain and side-(*hair] atoms are plott’ed iI1 Figure 3. Xs is typically the case. the temperatjure factors for main-chain atoms art: generally less than side-chain atoms. Also as rxpec%ed. there is (sorrelation bet,ween the mairr-chain and side-chain values. There are. nrvert,heless. four regions which appear to have correlated high temperature fac*tors. From Figure 3. t,his includes residues 55 t,o 70. 160 to 170. 255 to 265 and 27.5 to 285. The segment. including residues 55 to 70 is t.he outrrmost strand of the sixmembered dinucteotide binding domain, and is locat,ed on the surface of 1hr m&c-ule. K8esidurs 160 to 170 belong t 0 a loop (Wnnectinp the catjalytic* and nucleot,ide binding domains. Oddly, they comprise part of the subunit-subunit int,erfacar. Residues 255 to 265 chomprise t>he turn region i hat caorrnec+s two

Structure of E. coli Malate

nehydrogenase

871

W Figure 2. Segment’s of the final electron density map. (a) The stereo drawing shows a small segment of’ the electron density map obtained with csalculated phases using the final refined co-ordinates. A few side-chains are numbered as in Fig. 1. (11)The stereo drawing shows a close view of the citrate inn in the active sitme.

p-strands but is solvent exposed. The elevated temperature fact,ors for the segment 275 to 285, however, appea,r to be the most explainable. It is an extended random coil, for the most part not hydrogen-bonded t’o other segments of the polypeptide chain and lies on the surface of the molecule at a point farthest from the dimer dyad. Apart from the 160 to 170 segment, all these regions of higher temperature factors correspond t’o homologous regions of higher temperature factors in the refined cMDHase structure. Thtb single citrat,e molecule in t#he active site appears relatively fixed with an average temperature factor of 1 l-7 X2.

(t))

lVvcondary

Ntructure

The secondary structure present in eMDHase, on the basis of the met,hods of Kabsch & Sander (1983}.

is visible in cartoon form in Figure 4(a) or in st’ereo in Figure 4(b). A t,otal of 11 P-strands forming three p-sheets are present. The first’ six B-strands ‘in the protein are parallel to each other and connected by a-helical segments to form a single p-sheet. This is the dinucleotide binding domain of eMDHase and is homologous with the same region in c_MDHase and mMDHase. Tn Figure 4(a), the six strands st.art at the bottom right and continue diagonally upward to the left. A second P-sheet (II) is made up of t’wo antiparallel P-strands. BG (172 t,o 176) and PH (181 to 1871, shown in Figure 4(b), with a degree of righthanded twist that is somewhat greater than that seen in sheet, I. The segment connecting the two st’rands is a bulged three residue hairpin t’ype I turn, that bends sharply away from the sheet and points into the interior of t,he molecule (Richardson & Richardson, 1989). Between sheet I and sheet IT lies

50

_

.--__-.__-.--~~..--.

.~~~.

. __

-._-~_.-.-_-

.~

i--_--_-

Backbone 6

I 40

I 20

I 60

I 100

SiO

Side-chain

_---____--___-.--.

50

_

50

i

-____.-.

i

:

.-__.--~~_-.I_

Side-chain

i

.

I 240

; “’ 260

i

:

..I

” 2AO Residue’number

~-

i;j.

;I1 Backbone

240

----7-

i *

3Ao

I 320

Figure 3. The mean residue temperature factors for crystalline eMDHase. The temperature factors obtained from thts refinement are plotted against residue number. In each row, the averaged temperature factor for the backbone atoms are plotted on the bottom half and the averaged for the side-chain atoms are plotted on the upper half.

another sheet (III), The first’ half of sheet III consists of two antiparallel strands that twist approximately 90” to form the second half. In the second half, however, a third strand, PM, also antiparallel in direction, interacts with the two other strands. Therefore, P-sheet III includes /?K (249 to 259), /?I, (264 to 271) and PM (276 to 280). For the most part, p-strands are in the interior of the molecule and are not exposed to solvent,; only the turns connecting the B-strands protrude significantly into solvent, except for the /?G-/?H turn mentioned above. However, the p-strands /?C and /?M lie on the outer surface of t,he molecule and are in direct contact with the solvent. The location of /IM correlates with the elevated temperature factors seen for this area. Three long a-helices have distinct bends in them. In each case, at the site of the bend, a helical hydrogen bond is disrupted and is mediated instead by a water molecule (Blundell et al.. 1983). The helical bends, including residues 87 to 109, 195 to 216 and 284 to 312, are visible in Figure 4(b). The aDE helix, which constitutes a part of the connection between strands BD and BE in b-sheet I, bends about 45” at N93. The missing helical H-bond is

replaced by an intermediate water molecule (W319. B = 23.9 A’) linking the carbonyl group of N92 with the peptide nitrogen atom of G96. Water molecules with B values in the same range as those of protein atoms (18.8 A*) suggest well occupied sites. Water molecules with H values greater than 35 are probably poorly occupied sit,es. The alG/crBG helix bends at N207 about 50”. Although it appears as one long bent helix, it has been divided into two segments, ctlG and cr2G, in accordance with the nomenclature that has been used previously for dehydrogenases (Rossmann et al., 1975). It is unusual in that it is bent in a concave manner with respect to the rest of the subunit. In the ctlG/a2G bend, the carbonyl groups from three residues do not have normal helical hydrogen bonds but. instead, interact with water molecules. The carbonyl group of R205 forms a hydrogen bond with W381 (B = 23.7 A*), the carbonyl group of 1206 with W317 (B = 7.7 A*) and finally the carbonyl group of Q207 with W318 (B = 12.4 A*); which is, in turn, hydrogen bonded to the peptide nitrogen of G210. The final and longest secondary structure element in the molecule is the ctH helix at the C terminus,

Structure

of E. coli Malate

Dehydrogennse

873

C-TERMINAL

SUBUNIT INTERFACE

Figure 4. The conformation of eMDHase. (a) The portions of‘ the model represented by arrows are p-strands, and coiled ribbons signify a-helices. Segments of undesignated secondary structure, usually random coils, are indicated by the rope-like segment’s. Classification of secondary structure is based primarily on the configuration of the main-chain as described by G$/IJangles and on the hydrogen bonding pattern observed. The beginning and end of an a-helix or /?-strand are the first and last residues making the appropriate hydrogen bonds between peptide carboxp and amide groups. Regions not conforming to any known secondary structure element are called random coil. (b) The stereo drawing is an a-carbon representation of the eMDHase structure. Both subunits of the dimer are shown with the top one in roughly the same orientation as (a). Every 20th a-carbon is numbered to the right of the atom. The positions of the hound citrate ions are shown with ball and st’ick models.

which bends 31” beginning at E293. The carbonyl group of E293 forms a hydrogen bond with a water molecule (W363. B = 36-Z 8’). An exception occurs at G294, where the carbonyl group is pointing into solvent with no visible bound water at hydrogen bonding distance. However, the carbonyl oxygen

atom of G294 is only 3.4 A from W363 mentioned above. Finally, the carbonyl group of M295 forms a 3,,, type hydrogen bond with T298, while 1,296 forms a normal helical H-bond with L300. The helix then continues for 12 additional residues. Each of the three bent, helices mentioned above are on the

t)he most t~xtt~risivc Irytlroyt~~i k)olltlirq irltc~r;rt~t ions. I’artieularly int.errsting is t,hr int~t~r;lt%iorl 11t.twt~t*n H4X and ILl53. Thr atom NI)l of’ NW otI OIIP subunit is within WL\ of RI53 NE OII tilt, othcdr subunit ant1 il. water rnolrt*ulr~ mrdiates Ir~~drogt~n bonds hetnrrn H4X X1)1 and RI.53 NHl. IX 154 is a t*riticA residue in 1tie rc~tqgiition and t)indirlg of substrate. Solution studies with f)ort+ntL mhl 1)Haxt~ have concluded that a histidine residue is rt*st~onsiblr for 1he acid dissociat.iori of t hr dirnt,r ilt pH 54 (I3leile rt (11.. 1977). I’rt~tor~atiorr of t tift imitl;~xolf~ ring should trad to a positivr)ly t4lwrg:cd hist itlinc~ located close to the positivelv t~harged argiriint, anti destabilize the subunit inter&es of both ~11 I)Hil~st~ and m~ll)Hast~. h2utation of I)43 to asparagirrtl irl yeast mXl J)Hase. t,he homologous rrtsiduts to I)45 in r~Ml)Hasr. resuItjs in almcjst taomplrte Ioss 01’ t’nzy matit. activity (St.efiari f4 r/l.. 1990). The inttirat.lions IW~WtWI an activtb 5it.e rrsidut~ atitl rtGtiuc.> on ariothr~r subunil lend tac.it suppol? to thtl itlt’il 01’ nit~dulatit~rr of taatalytit. acti\-ity in M I)H;lst3 virr dimerization of subunits (Kirktoft cut u/.. l!tX!kr,h). a%(i anti x3(: haw ant sitit, in t*c)litat.t \\.it h tht~ substratr binding site ant1 t hr oppositt~ side ti)rrtiirig part of’ t hr tlirnr>r interfatstk. The transition f’rorrr t htb monomeric* to dimeric state might also t~fli~,*t t’ittaIytita rates if these helical stymrnts move cturitlg surface of the monomer and dimer: each has between 20 to 25 amino acid residues with the bend ot*curri np six to eight residues in from the N-tt,rminal pnd. The bending simply appears to make the helix conform to t,he contour of the inner t~)rt~ of the molecule. alt,hough the bending ma,?- be t*onvrx or t~onc‘avc~ with respect to the core subunit. Tht~ bend always seems to require the present”‘ ot well-occupied water ont’ or more relatively molecules interacting with the free taarbonyl oxygen atom.

As in other known MDHases. eMl)Hasr is a homodimer in solution. Although t,he asymrnctrit* unit contains one monomeric subunit 1 tight packing tjrtween monomers is found around a tyvstallographic Y-fold symrnetjry axis. Looking at this t tia,l t tie same region of the latt,ice. it is apparent structural elements found in the setzondary eM JIHase intclrfacc are roughly present in cMl>Hase (lSirkt,of% et 01.. 1989a,6). &It, unlike c~IDHase. t’he two subunits in the rMDHase dimrr are related 1)~ crystallographic symmetry. The overall view of the dirner interface in Figure 4 shows t’wo pairs of five helices involved in subunit-subunit interactions. ~13 intera& with XX: through polar and non-polar conta,cts and wit,h ctH of the other subunit through CCC fits int,o a triangular non-polar int,eractions. pocket created by the helices a2F. ad(: and LXX. int,eracting with the amino-t,ermina,l portion of ct2F and the carboxyt,erminal port,ion of (~3: through polar and non-polar contacts. All of the polar subunit-subunit interactions are listed in Table I. 1145 and H48 on CCCappear t,o have

i~.ssOc~iatic)~l.

In addit ion to direct hydrogen bonds I)c~tuet~Il residues. solvent -mediated hydrogen I)o,rtls I)liry a significalit role in the subunil-sut)ilrlit intrrfat7’. LVater ~riolt~t~ules 31-l (II= 184A2). Xl (I{= 1 I.9 AZ). 370 (I,’ = 32.2 X2) and 385 (IZ = :Wo .1’) Rl’f’ involved in a c*omplrx arra!. of’ hytlrogf~n honcls. A\s (‘an

ht. st~t~~l itI Ta.bIta

1. int*ludtd

ill

this

;Lr’r’il>. aw

hydrogen bonds dirrt~tly I)ctwetan polar group)* on boOi subunits and bf~t~wc~rri l~olar groups 011 tjottl subunits but mediated 1)~. onr’ or more’ \vitttAr molecult+. Thtb network aIs. U731-4 11’311’ (721) art’ t ht. c,losest atonrs 10 t trta t~r\.stallt)~~.;t~)hic~ and molec*utar 2-fold axis. as is ‘ilhrstr;ittYl in Figur.t> 5. Hvdrophohit~ tY~ntat~ts n1ust also t*ont rihutf, sign&t:antly to the stabilitl, of’ the tlirnt~r. crH. UC’, x%F and 2%: all have sign;ficant nurnberh of’ non polar rrsitlues. thr, majority of fvhich app(‘ar to point into the crknter of the dimer intrrfic~~. This creates a hydrophobic L)ochket along most of’ t’hth P~fold axis of the dimrr. Valillri alld It~ucint~ make uf) t,he bulk of thest, rtGdurs. A view tlour~ Itw 2-t’oltl axis of thtb tlirnrr. as is stio\vn in Figure -t(h). indcatrs a sizable holr that ih. nonetheless. somr~what~ illusory. ‘l’tlsts with ii :) X f)rohe reveal that the Cluster of h~drophobit~ residues in this rt@on prevent, anything 3 !I or larger in diameter from passing through the hole.

x7.5

Structure of E. coli Malate Dehzydrogenase

229 CL8

229 CLN 429 MR

p

p

Figure 5. \&‘atrr bridges at the subunit interface. The stereo drawing represents some of the symmetrical residues and water molecules involved in hydrogen bonds connecting t,he 2 subunits. The black dot in the middle designates the crystallographic P-fold axis. For purposes of co-ordinate numbering. symmetry related residues harr different identification. W321 and W7121 are equivalent, as are W38.5 and \\‘792. W314 and Wi21. Similarly. amino a(%1 rrsidues T2% and T/29. Q229 and Q636 and Q23 and Q430 are eyuivalent.

Each subunit of rMDHase, like other dehydrogenases. can he divided into two domains that’ are functionally distinct but visually intert.wined. The first’ 150 residues of the protein form the six b-strands of sheet I, connected by five cc-helices in a typical /3@ arrangement twisted to the right that is called the SAD hinding domain. In cMDHase and LDHase, essentially all of the atoms interacting with thr cocnzyme XAI)’ are part of this domain. With a few exceptions, the same is true of eMDHasr and a docking experiment describing the putative binding site will br described c~oenzymr subsequently. The topology of ,/-sheet) T is identical to that found in other JIDHases: + lx, +1x, -3x. -1x. -1x (Richardson. 1977). Most of t,he crossover caonnections are made lvith helices. The first tu;o helicscs. ccl< and XC’. are found on one side of t,he sheet. while the rrlnaining three. ctC”, crI)E and ctlF. are locat8ed on the opposite side. On the basis of these characteristics. t’ht> NAD binding domain has heen divided into two subdomains that in themselves form SI ructural units: so-called nucleotidr binding domains (Rossmann et crl., 1975). Tn 011t nucleotide binding subdomain elMI)Hasr. contains t,he first, three fi-strands of the sheet: a loop crosses bacak over the first’ three st’rands and c~~~~r~ect~s the rest of sheet I, which is the other subdomain. A va,riation of the characteristic amino ttS -GXGXXGis present acid seqnencar -GSXCXC:-, but only in the first nucleoWr binding domain. The second half of thtx molecule includes the catalytic. residues H177 and D150. which form a His..+ pait, si1nilar to that found in porcine

c&I DHase (Hirktoft & Banaszak. 1983). One of the arpinine residues involved in substrate binding, R153. is in the catalyt’ic domain. although the other. R81, is present in the nucleotide binding domain. Much of the struct’ure in the catalytic domain will be discussed in sect’ion (d). h&w. The degree of similarity between the structures of eMl)Hase and porcine mMDHase is very high on both the trrtia,ry and quaternary level. ITsing a preliminary set of co-ordinates for mMDHase, the r.m.s. deviation for the main-chains of each molecule is 0.76 &&. Roth molecules have identical secondary structure elemrnt)s including the three large a-helices with noticeable bends. Fifteen of the 19 glycine residues that show unusual 4 and $ angles in eSII>Hase are conserved in mMT)Hase. Minor differences can be found. of (‘ourse. in stretches csontaining surface loops and turns. I)43 and HlCi in mMDHase correspond to D-L.5 and HG3 in eMDHase in t,hat they also tnake important subunit-subunit! intera&ions. However. there is no equivalent residue in eYIDHasr t,o E-t8 in mMDHase that forms ext,ensive hydrogen bonds with at least two other residues at t’hr ditne1. i11tcrfac.e (our unpublished results). Porcine heart cMDHasr. on thp other hand, differ? from eNDHase to a greater extent t,han tnMDHase. with a r.m.s. devjation in hackbone atoms of l+% ,A. The secondary structure is similar except for an additional B-strand. called PJ. found it1 cMDHase from residues 204 to POT. In rMDHase, this is replaced by t,he a(:’ helix. C’onformational differences can also be observed in the area between ztZ and j313 in eMDHase. where t#here is no loop corresponding t,o t’he one found in c*MDHasr. and in

876

I)150

.I/. 1,. llrrll et al.

--

0111

HlTi

-..-__

-__-_-.

_~...._...-_.

-.

SD1

I)150 0122~~ HlTi .c’I)I H 177 SE?.

(‘IT 02 011 RX7 NI’ 1 ” (‘IT 02 RX7 NH%. (‘IT 011 Itlb4 SH2’..(‘IT O-11 Itli’i. . 1‘VHI . ..(‘IT 04.) RX1 Sl5~“(‘ITOl” I:81 NH1 ” (‘IT 01 1 h-1 I!) SIX?. ‘. (‘IT 012 \V:W (‘IT 042 lv’ir’l * .h ” (‘IT 01‘1

HI77 Sl!X~..(‘IT

the part of t)he structure from residues 241 to 257 spanning ct3G and /?K. Residues involved in subunit interactions in cMDHase are different from those found in eMDHase and have been discussed in detail (Kirktoft ef ab., 1989a,h). (d) Thr actiw

sitr

The close structural homology hetwern the MDHases and LDHases suggest similarities in the active sites and the catalytic mechanism (Clarke et nl., 1987; Wilks et al., 1988). However, prior to this about the st,ereostaudy. the only information c+emistrv of substrate binding in the MDHases had heen der:ved from the position of a bound sulfate ion in the crystal structure of cMDHase (Birktoft et nl.. 1989a;b). ,Many of the remaining details of dicar-

Thr sl)cf*itic*ity of hiruling of t ht. tlic~;~rbo;~lif* subst,ratfas of J1I)Hasrs involves two argmlrrc~ residues. RX 1 and RI XI. as shown in Figure 6. 11 third a,rginine residue. RX7. is near the (‘-2 hydroxyl group of the hound citrate a,nd is visible to thl, ktli analog in Figure 6, The int.rractions of the snbstrat,e between each of thcx arginirrtx residues. RX1 and RI.%. and 01th c~arhos~latt~ groups of thr (ait rat(l arch somewhat clift‘rrcwt. I< 153 USPS both of t hr tt~rrrrinal guanidino nitrogen at,onls. while KS1 IIS’s NE anal only onr of the terminal guaniclino nitrogen atomti. Additionally. a somewhat weaker hydrogen l~)nd is found bet ween N1 19 NI)% and 04% of citrattl. Other aspects of the acative sit.? st’ereochtmist r> are also shown in Figurr ti ittld ill’ that t.hr panforma&n of the domain in t,he eMDHase struc’turr is extn>mely similar t,o that of the SXI) binding domain in cMDHase and dogfish I,I)Hasr. In all cases there are significant numbers of non-polar residues lining the cavity. However. the cY1DHase structure contains bound NAD molrc&~s. one per monomeric subunit, thus indicating that the conformation of t.he domain changes little af’ttxr the cofactor binds. To facilitate visualization of the caavity in the NA r> binding domain, the cavities were modelled as filled spaces. The cavities calculated from the crystal co-ordinates (Levitt, & Banaszak, 1991) without bound water molecules art’ shown in the present-e of citrate and malate. see Figure i(b) and (c). respectively. Figure 7(a) shows the c*avity in the overall rrystallographic struct.ure. The cavity search procedure finds a large number of indentations and small ctavitirs, but the ones illustrated in Figure 7(a) and (b) are adjacent to the sub&rate binding site and agree with the location of the cor~~vme binding site based on the &DHase homolog\-. sate that, the available coenzyme cavity in the presence of (titrate is considerably smaller than in the presence of malate (Fig. 7(b) ~t~rsus (c)). ‘Thtl volume of the cornzyme cavity is decreased from 509 to 451 .A3 when citrate is substituted for malate brcaause the additional arm, C-5 and carboxylatr (‘-6 of citrate. protrudes into the NAD cavity. These addit,ional atoms of cit,rate. not present in malat,e or oxaloacetate. appear to have no interactions with the eMDHase, nor do they appear to interfere with the interactions of the part of citrate t,ha.t mimics substrate. The C-6 carbosylate group makes three hydrogen bonds to water molecules. Determination of the cavit); volume and shape in thr region of the XAD bindmg site was done after removal of thesr and all other bound wat,er molecules (Levitt & Banaszak, 1991). ;\s can be seen in Figure 7(a). the appearance of the cavit,y suggest.ed an extended conformation for the XAD in agreement with the structure of X.41) hound to other tlehydrogenases.

(f)

A hypothetical

ternary

romplr.~

,lilore detailed information on the coenzyme conformation and protein-cofactor interactions was sought, by fit.ting an NAD molecule into the cavity

(bl

fc)

Figure 7. Spw filfing (*:ivity with citrate in thi middle diagram represents wntwt with the cavit,y c*itrste has been c~hanged

model of’ the NAI) vavit)~. (a) 7’hv tjop st~errodiagram represents an o~~arall \-iew of tlw KA41) active site in a bac~kbone model of’ the protein. All filled-in areas belong to the (:a~i&y. (1)) The a dose-up view of the X’AD binding wvity wit,h citrate in the active sit)e. All side-c:hnins ~II are shown. (c) The bot,tom diagram represents the same feabres shown in (1)): ho\rrver. t’hr to rrpresent a malatr molrculc.

and submitt,ing the structure to erwrgg rninimization. The NAD molecule. the cwordinates of which were obt’ained from the crystal structure of the cMDHase-KAD binary complex at 2.5 A resolution. was placed into the cofactor binding cavity of

eMDHase until an optimal fit 11.v visual inspec*t.ion was obtainw~. The (Atrate already prcwnt in the active site was converted to a mltlate molecule by removal of atoms (‘-5. (‘-6. U-61 and O-62 that prot’rude into t.hr c9fa,ctor hindinp cavity. Five

Structure

of IX. coli Malate

x79

Dehydrogenase

153 ARC

4

87 ARC

81 ARC

OP

OP

177 HIS

87

A

ARC

81 ARG

02R

Figure 8. Hypothetical model of P;AD in the XAD with malatr as found in the energy to obtain this structure are described more labeled. Hydrogen bonds found between the broken line drpic+s the trajectory of hydride

the active site of rMDHase. The stereodrawing shows the conformation of minimized structure of the eMDHase--NSD-malate complex. X&hods used fully in the text. Atoms involved in hydrogen bonding t,o the protein are malate and active site residues are shown by broken lines. 4 single heavy ion transfer during catalysis.

water molecules. also in the cavity, were removed from the co-ordinate list. The eMDHase-malateNAD ternary complex was then submitted to energy minimization with no atoms restrained. The structure was minimized well before 200 cycles at substantially lower energy than the starting structure. The r.m.s. deviation from the cryst’allographic co-ordinates of eMDHase is 0.45 ,L. Without the const’raint of X-ray data, the geometry, of course. improved significantly: r.m.s. deviation is 0.008 AA for bonds and 2.56” for angles. The overall conformation of the NAD in the energy minimized eMDHase-NAD complex differs little from that seen in the crystallographic cMDHase structure. It is shown in Figure 8. Note t,hat the nicotinamide and adenosine rings remained planar and both glycosidic bonds are anti. Both riboses have 2’.endo puckering. As can be seen in Figure 8, the largest difference occurs at the 3’ oxygen atom on the nicotinamide ribose. In the crystal structure of cMDHase. it is in an axial while in the energy minimized orientation. eMDHase complex, it appears more equatorial. The only identifiable reason for this change in the eMDHase structure is a hydrogen bond that, appears to occur between the carbonyl oxygen atom of WC1 and the 3’WH of the nicotinamide ribose. After the energy minimization, t’he adenosine and pyrophosphate moieties remain similar in conformation to the corresponding atoms of NAD in t,he cMDHase cryst’al st’ructure as well as in the minimized cMDHase structure. In eMDHase, however,

there is a definite shift of the plane of the nicotinamide ring so that the C-4 position of the ring is approximately O.‘i,? w closer to the malate than it was in the st’arting structure. A possible cause for this diffference could be due to the presence of a sulfatta ion in the cMDHase struct’ure rather than a substrate analog. Another contributing fact’or could be the difference in loop positions. closed in eMDHase but open in cMDHase. There was no change in the position of the loop in either eMDHase or cMDHase during minimization. By comparing Figures 6 and 8. one can see t)hat the position of the substrate and the active site residues changes little when the cofactor is added. All the hydrogen bonding int’eractions found in t)he binary complex are preserved in the hypothetical t,ernary complex. One difference detected. however, is that the imidazole ring of H177 moves approximately 0.5 L& toward the P1’4D, presumably t’o allow hydrogen bonding with 0-1X of the carboxyamide group. Malate rotates slightly t’o accommodate the cofact)or and the interact’ion with the histidine. In speculation about the catalytic mechanism, it is of interest that the C-4 of the nicotinamide ring moves to within 3.4 A of the C-3 of malat’e. It is between these two atoms that hydride transfer must occur during a catalytic cycle and the proposed trajectory is marked by a heavy broken line in Figure 8. Although the ordered mechanism states t’hat the cofactor binds before substrate, there appear to be no steric constraints t#o prevent the binding of substrate first.

N3A ObR. 03K:

I)‘&1 ()I)” (:I:% N INI Ol)l I)41 01)”

OPl.4~ OF%\:

\V.‘i’l%fi El4 s \VATfil

OPIN. ows:

(Xl&02f.J N”N.

OILY.

For t’he sake of comparison and to reaffirm t,he energy minimization approach described above. the original cMDHase-SAD binary complex was also refined with XPLOR. combining X-ra)- data and t’he XPLOR’ energy terms. Next, it was energ) minimized without X-ray dat’a, as in the case of eMDHase-NAD-malate. described above. The X-ray R-factor from the XPLOR positional retinement stage is 0.178 compared to 0167 using TST as et al., 1989a,b). The described previously (Birktoft r.m.s. deviation of the XPLOR refined structure from the original (Birktoft rt ab., 1989n.b) is 0.36 .% for both subunits. The energy minimization without .X-ray dat’a gives a structure that has an r.rn.s. deviation from the XPLOR refined structure of 0.48 A. The SAD molecules of the original cbMI)Hase--NAD binary complex, the XPLOR refined and the energy mmlmized structures are all highly superimposable, essent,ially ident’ical. The XPLOR refined and then energy minimized structure differs from the original in only one respect. the 3’ oxygen on the nicotinamide ribose is shifted from an axial to a more equatorial posit’ion as was found in the minimized eMDHase-NAD structure. One can conclude t,hat further refinement of the original least-syuares refined cMDHase-NAD structure with XPLOR essentially made it more similar to t,he minimized e)IDHase-NAD st’ructure. The energ). rninimization of t,he c,MDHase holo-structure appears to be a reasonable cont,rol for the predicted eMDHas+ NACmalate comples.

Table 3 lists the major hydrogen bonding interactions present between the protein and the eoenzyme in the energy minimized structures of eMDHase and GDHase. In both cases, backbone amide groups and carbonyl oxygen atoms plaq major roles in forming polar bonds to the ribose and phosphate oxygen atoms. Homologous residues are involved in similar interactions with NAD. For instance. 1)34 in eMDHase and 1141 in cMDHase both bind to 0-BR and 0-3R, and Nl 19 in eMDHaae and S130 in c&lDHase bind to Cl-SQ. The caat’alytic histidine anchors the nicotinamide by binding to t’he carboxyamide oxygen atom of the nicotinamide moiety. Water interactions are also abundant. As in the original crystal structure of the et nl.. 1989n,b). where the c,\ilDHase dimer (Birktoft interactions between the enzyme and the (sofactor are not ident,ical in the two subunits. different intera&ions caan be seen in the energy minimized c:MDHase-N.41) subunits as well. As can be seen in Table 3, the interactions found in the eMDHa,se9AD complex are shared with either subunit A or 13. or both in the &DHase-NAD crystal structure. The slight variation in binding modes of NA4D to different enzymes. or to separate subunits of the same enzyme (c?olDHase). leaves open the question of a single “caorrec+” mode of binding.

4. Conclusions The family of MDHase structures now includes three members, a proka,ryotic form and two eukary-

XXI

Structurr of E. coli Malate Deh,ydrogenase otic enzymes that have unusual evolutionary and structural relationships (McAlister-Henn. 1988). The fortuitous binding of citrate in the active site and the high resolut,ion of the st’ructure provided a unique opportunity to view catalytic residues int,eracting with a substrate analog and to observe the interactions between subunits. The four additional atoms contained in citrate as opposed to the substrate malate, do not interact directly with the protein. Studies of the nearby coenzyme cavity indicate that they protrude into t,he Z?AD binding site. When these atoms are removed. it is possible to dock XAD into a site that corresponds to the coenzyme binding site in After docking, t’he NAD-rMDHase cMDHase. complex can he energy minimized without any dramatic alterat’ions in the enzyme or the coenzyme conformation. This hypothetical model then contains the co-ordinates of the enzyme, coenzyme and the substrate. I>-malate. A control experiment with cMDHase produced little change in the crystallographic: co-ordinat,es. Nonetheless, it is necessary to recognize that the eMDHase-NXD-malate sterpochemical arrangement is only a hypothetical model of the eM DHase active site. In t)he active site. the (‘-2 hydroxJ;l of malate. by virtue of its hydrogen bond interactlon with H177, is in the correct posit)ion for proton removal by the ternay His. Asp pair. In the energy minimized complex. the Hl77. I>150 pair is intact while a hydrogen bond is formed bet*ween H177 and the oxygen atom of the rarboxyamide group of the NAD. This suggests that H177 is not only responsible for the proton removal from t,he hydroxvl group of malate, but may also help hold the nicot%inamide ring in posit,ion. During the catalytic reaction, t,he (‘-4 position of the nicotinamide ring is thought to awept a hydride ion from the substrate C-2. In the cwergy minimizc,d structure, these two carbon atoms

are virtually in van der Waals‘ cont#act. This suggests that the dist,anccb a hydride ion must movr is small. \Z’e are in thtk pro(*ess of preparing crystals of thtl rMl>Hase-N.41) complex. However. on the basis of the c*rystallographir data present,ed here, it is unlikel~~ that this form will include bound cit’rate. due to the parCal obstruction of thth coenzymr binding csavity. Structures of mutants of eMDHase should br very useful in elucidating further details about various properties of the molec~ule. partichularl\- t.hr a(4 irrl site. \Ye thank J)r IAre M(~Allist~er-Henn for the E. coli strain cwntaining the Ml)Hase used in these studies. \Ve are

gratt4’ui f’or thr hrl~) of IIianne Stockhausrn who purified the enzyme in the early stages of the 1)rojrlxt and to Et1 Hoeffnrr for overseering the crystallographic computing system. Wr itrt’ (q>ecGdly indrbt,ed to I)r Kill Gleason for the use of thus partially refined mitochondrial rnalatr drhydrogrnase cwordinates. which were critical during the molt~cular rrl,lawment stage. We also acknowledge his continued help in the ~4~2 of dynamics calculations during rrfinrmrnt. The studies were supported by grants from

XSF (DMB 8941746) the Supewomputer Institute.

and

the

Minnesota

References Abad-Zapatero. C‘.. Griffith, J. I’.. Sussman. .J. 1,. B Rossmann. 11. G. (1987). Refined crystal structure of dogfish M, apo-lact’ate dehydrogenase. .J. AMol. Biol. 198. 445-46i. Bewkmans. S. & Kanarek. L. (1981). Demonstration of physical interactions between consecutive enzymes of the citric acid cycle and of the aspartate-malate shuttle. Eur. J. &ochem. 117. 527--X35. Bernstein. L. H., Grisham. 31. H.. C’ole. Ii. I). & Evrrsr. *J. (I 978). Substrate inhibition of the mitoc*hondrial and cytoplasmic malate dehydrogenasrs .J. 12iol. Chew 253. 8697~ 87Ot. Rirktoft. ,J, ,J. & Banaszak. L. .J. (1983). The presence of a histidine-aspartic acid pair in the ac+ivr site of Y-h\-droxvacid dehvdrogenases. .J. Hiol. ( ‘hem. 258. 17i-4x2. L Hirktoft. .J. ,J.. Fu. %.. (‘arnahan. CG. E.. Rhodes. (i.. Rotlericuk. S. 1,. & Banaszak. L. .I. (l989*). (‘omparison of the molecular struc~tures of cytoplasmic. and mitochondrial malate drhydrogenasrs. lliochent. Sot. l’mns. 17. X~l-:~O4. liirktoft. J. .J.. Rhodes. G. & Banaszak. I,. .I (1989b). Refined crystal strrrct,ure of c~ytoplasmica malat’e tlehydrogenasr at 2.5 .A resolution. /1iwh~n/istry, 28, 606.i- 608 I t3teilr. I). 11.. Schulz. R. A.. (iregory. E. 31. B Harrison. ,J. H. (197-i). Tnvestigation of the subunit interactions in malate dehydrogenase. J. Hiol. C’hrm. 252. 755-758. Klundetl. T.. Barlow. I>.. Borkakoti. S. d Thornton. .I. (19X:3). Solvent-induced distortions and the cwrvat,ure of r-helixes. Satuw (London), 306. S-%7. Brunger. A. ‘I’. (1990). S-PLOK Nn~~rr/. Version 2. I Yale l’nivrrsity Press. X;rw Harrrr I’.S..A. Brunger A T. KT Krukowski. A. (1990). Slow-waling protowls for wystatlographiv rrfinenwllt hy simulated antwaling. dcfrr C’r,ysfa/lo~q~~..w/ .4, 46. ~5tK-iOY. 9. 1

(‘tarke. .4. R.. \Vi,@y. 1). IS.. (‘hia. LY. S.. t3arstow. I).. .1tkinson. T. & Holhrook,

.I. .J. (1986). Site-directed

mutagenesis rpvrals role of mobitr argitlint> rrasiduein lactate drhydrogenase catalysis. .Vr~tjcrc / l,ondor/), 324. 6!19-502. (‘Iarke. .I. Ii.. Smith. (‘. tJ.. Hart. K. \V.. \I’ilks. H. Il.. (‘hia. \V. S.. Lee. T. V.. Rirktoft. .J .I.. Ranaszak. I,. .l.. Barstow, I). :I., Atkinson. T. h IL. ‘brook. .J. .I. (1987). Rational construction of a L’-hvtlroxvacid with IlP\V dehvdrogenasr suhstratfx qprciicity. IZ&hm I1iophy.s. Rvs. (‘ornr,//ir~. 148. 1.1;--23, (‘larke. A. R.. \Vilks. H. 12.. Harstov. I) .I.. .\tkinson, T.. (‘hia. \Y. S. & Holhrook. ,J. .J. !l!X-+X). .\n invwtigation of the caontrihution matte I,>- thr c~;~rhoxylate gr‘c,up of an active site histidinr~~~~l,artat~~ c~,uplr to hif~ding and c,atalysis in Iac,tattt tlc,llvlfrc,een~rsr. Ki~rhr,,ristr,//. 27. I617 IWP. I)atta. .\.. Jlrrz. ,I. 91. & Spivrv. H. 0. (I!M). Substl,ate c~hannc~lingof osaloacetatr ‘in solid-statta c~c~mJ~leses of malatr drhytlrogrnase and citrate uynthase. ,1. Hiol. (‘hen,. 260. I .iOo8- I:501“. Fahirn. I,. A.. Ktniot,ek. E. H.. Mac~l)onald. JI .I.. Fihlch, 13.& llilka. Jl. (1988). Regulation of inaiatv tlehytlrogeru~r a(*tivity by glutamate. c,ltr’ntr. ;c-krtoglutarate. and multienzyme intrrac~tion. .I /,‘iol. (‘h~t)~. 263. 10687 106!#7. (:rali. ly. )I.. Trotnmrr. \I’. E. & Rossmann >I. (:. (19*1).

Structure of the wtivr trrnary t~omplw of pig heart lat~t,atv tlehydrogrnaw with S-ltrc-SAl) * at 2.7 .\ resolution. .I. Yol. Hiol. 151. 18!1&3Cb7. Hall. M I).. I,tlvitt. I). C:., PIl(~:~listrr-t-leI1I1. I.. & Ilanaszak. I,. .I. (1991). Purification and (3rystallization ot’ recombinant Eacherichiu coli drhy~rogrnasr, .J. Mol. Hiol. 220. 551-553. Honka, E.. Fabrg. S.. Kiermann. T., Palm. P. B Hensrl. R. (1990). Propwtitls and primary srquencr of thr r,-malate dehydrogenase from thr extremely thrrmophilitarchaet)at%erium Nrthnnothwntccs jkvirlu.v. Ear. ,I. Hiochrm. 188. lii!3MX~1. Kahsch. \V. B Sandrr. (‘. (1983). IXctionary of protrin swondary strutsturr: pattrrn recognition of H-bonded and gromrtritA feat,ures. Biopolym~rs, 22. %:577 26X.

,lonrs. 7’. .A. (1978). I1 graphics modrl building merit system for macromolwulrs. (‘r?q.stalloyr.

and rrtinr.J. .-I pg.

11. %8-272.

Levitt. I). G. & Hanaszak, 1,. .I. (1991). l’O~‘KET: a wmputer graphics method for idmtiflving and displaying protrin cavities and ththir surrounding amino acids. .J. Nol. Graph,ic.s in the press. McAlistrr-Henn, L. (1988). Evolut~ionary relat~ionships 13. among the malate dehydrogrnases. TIBS. 178-181. JIcAlister-Henn. I,.. Blaber. M.. Kradshaw. It. A. & Xist.0. S. .I. (1987). (‘omplet,e nwleotide sequenrr of the ~scl~urirhia coli genr encoding malate drhydrop~naw. Sucl. .-lcids Kes. 15. 4993. Minartl. K. I. & Mt,Alister-Henn. 1,. (1991). Isolat,ion. nucleotide sequtww analysis, and disrupt,ion of thr .IIDH% gene from Saccharonyces cerrr+sine: rvidrnw for three isozymes of yeast malab dehydrogrnase. NOI.

f’rll.

Bid.

11. 37Ok380.

Mullinax. T. It., Mock. J. K.. MlcEvilv. A. ,I. B Harrison. .I. H. (1982). Regulation of m~t~ochondrial malate titlhydrogenasr. J. Kiol. C’huu. 257. 13233G13239. Nishiyama. %I., Matsuhara. K.. Yamamoto. K.. Tijima. S., Ortzumi, T. & I+ppu, T. (1986). Xucleotide srqurnce of tht) malate drhydrogenase grnr of Thernws ,jlar~us and its mutation tlirect~ing an incwasr in enzytnt’ atativit>.. ,I. Hiol. (‘hem. 261. 14178~~~11183. Richardson. ,J. S. (1977). B-Sheet topology and tht Saturn (London). 268. relatedness of proteins. 495-500.

Robinsotj. .I. IS.. ,Jr. Inn~an. I,.. Sumrgi. Imic I’rrss. Ntaa \‘ork. Silvrrstrin. E. & Sulebt~lr. G, (1969). (‘at~al~?it~ tne~liatiisn~ of pig hrart, mitochondrial tnalat,e drhytlrog,rrnas~ stud&d t)y kinrtiw at equilibrium. Jliochrmistr,q, 8. 2543 -25.3). i’irrrr. 1’. .1. (1987). (‘omplrxt~s of srqwntial tnt~talwlit. 56, 89-I 24. enzymes. .4 nn~. IZuv. Riochwr. St&an. .J. S.. Minard. I