Crystal structure analysis and molecular model of a complex of citrate synthase with oxaloacetate and S-acetonyl-coenzyme A

J. Mol. Biol.

(1984) 174, 205-219

Crystal Structure Analysis and Molecular Model of a Complex of Citrate Synthase with Oxaloacetate and S-Acetonyl-coenzyme A GEORG WIEGAND,

STEPHEN REMINGTON?, JOHANN DEISENHOFER AND ROBERT HUBER

Max-Plan&Institut D-8033 Martinsried (Received

fuer Biochemie bei Muenchen, Germany

20 September

1983)

The crystal structure of the complex of pig heart citrate synthase and oxaloacetate in the presence of the potent inhibitor S-acetonyl coenzyme A has been determined at a nominal resolution of 2.9 A by Patterson search techniques and refined by restrained crystallographic refinement. The complex crystallizes in the presence of polyvinylpyrrolidone in space group P4,2,2 with a = 101~5A and c = 224.6 8, with one dimeric molecule of molecular weight 100,000 in the asymmetric unit. The crystallographic R factor is 0.194 for the 14,332 unique reflections between 6.0 and 2.9 a resolution. The structures of two forms of citrate synthase in the presence and absence of product molecules have been determined recently and shown to differ in the relative arrangement of the large and small domains (“closed” and “open” forms). The third crystal form described here is also closed, but there is substantial rearrangement within the small domain relative to either of the other crystal forms. We conclude that this is a third structural state of the enzyme, and catalytic activity of the enzyme depends on structural changes during the course of the reaction affecting domain conformation also. The three structures are compared, and it is shown that the large domain is considerably more rigid than the small domain. The conformation of the small domain adapts to the ligand. The inhibitor, and the “coenzyme-A-binding segment” of the enzyme are disordered. No electron density is observed for the inhibitor, and only weak density for the coenzyme-A-binding segment. Electron density for oxaloacetate is well defined. It binds in a very similar manner to citrate.

1. Introduction Citrate synthase (EC 4.1.3.7) has a key position within the central metabolic pathway, the citric acid cycle. It catalyzes the condensation of acetyl-coenzyme A and oxaloacetic acid to form citrate, thus forming a new carbon-carbon bond. The determination of the crystal and molecular structures of two different form of citrate synthase has provided considerable insight into the structural and functional properties of this enzyme (Remington et al., 1982). One of these two t Present address: Institute of Molecular Biology, University

of Oregon, Eugene, OR 97403, U.S.A.

205 0022%2836/84/090205-15

$03.00/O

0 1984 Academic Press Inc. (London) Ltd.

2Nj

(:. W’IEGASD

ET AL.

forms, the tetragonal P4,2,2 form, represents the free enzyme (the open form), the monoclinic C2 form represents the complex with products citrate and coenzyme A bound to it (the closed form). The molecular structure is divided into two domains, a large domain of residues 1 to 274 and 381 to 437, and a small domain of residues 275 to 380. The large domain forms the site of interaction of the two subunits in the dimeric molecule. In the structure of the free enzyme the two domains are separated by a deep cleft. This cleft is closed in the C2 form? which includes bound products, by a movement of the small domain relative to the large domain. This movement can be described: to a good approximation, as a rigid body motion a,round a hinge a.xis passing through the molecule close to residue His274. The product molecules citrat.e and coenzyme A are buried wit,hin t.he closed form. Citrate is in contact with a number of arginine and histidine residues (see Fig. 15 of Remington et al., 1982). Coenzyme A binds to the polypeptide segment 313 to 320 with its adenine and its pantothenic acid portions and to arginine residues 46 and 146 and to trimethylysine 368 with its phosphate groups (see Table 9 of Remington et al., 1982). A wealth of functional properties of the enzyme have been determined, which are in accord with and explained by these struct’ural features. The relationship between structure and function has been discussed by Remington et al. (1982) and Huber & Bennett (1983). It was suggested that the open form represents the molecular conformation to allow substrate binding and product release. In the closed form catalysis proceeds in the forward direction, perhaps in distinct steps of condensation and hydrolysis, with citroyl-coenzyme A as intermediate (Eggerer & Remberger, 1963). There was a principal weakness in the arguments concerning structure, as the binding site of oxaloacetate had not been experimentally verified. We assumed on structural grounds that it coincides with the observed citrate binding site in the closed form. As circumstantial evidence it was observed that’ the crystals of the open form crack in the presence of oxaloacetate, suggesting domain closure upon binding. It was obviously highly desirable to determine the molecular conformation and the binding site of oxaloacetate experimentally as described below.

2. Materials (a) Chemical

and Methods characterization

Pig heart citrate synthase was from Boehringer, Tutzing; oxaloacetate from Merck. Darmstadt; polyvinylpyrrolidone and polyethylene glycol 4000 from Sigma, Muenchen. S-acetonyl-coenzyme A was prepared accordmg to Rubenstein & Dryer (1980). The amount of S-acetonyl-coenzyme A bound to the crystalline material was determined as follows: 3 to 4 crystals were taken from the buffer as used for the X-ray measurements and washed 3 times with fresh buffer, but omitting S-acetonyl-coenzyme A and oxaloacetate; after 5 to 10 min they were dissolved in 20 ~1 of 0.05 M-phosphat’e buffer. The solution was chromatographed on a 0.5 cm x 16 cm column filled with Sephadex G-50 superfine (Pharmacia, Uppsala). The effluent was monit’ored at 254 nm with a spectrophotometer (Knauer, Berlin). Under the same conditions calibration runs were performed with known amount,s of citrate synthase and S-acetonyl-coenzyme A.

CITRATE

SYNTHASE

COMPLEX

STRUCTURE

“07

(b) Crystallization A solution (100 ~1) of citrate synthase (10 mg/ml) was dialyzed against 0.1 iv-Tris . HCl buffer (pH 8.0). To this, 100 ~1 of a solution of oxaloacetate (4 x 10m3 M) and X-acetonylcoenzyme A (lo-’ M) was added, followed by 14 ~1 of 40% (w/v) polyvinylpyrrolidone solution buffered with Tris . HCl and 0.45 to 0.55 tir-potassium/sodium phosphate (pH 8.0). The enzyme crystallized in needles of about 0.05 mm diameter and 3 mm length. These crystals were transferred for data collection into a solution of Zoo/c (w/v) polyethylene and 10m3 Mglycol 4000 with 0.1 iv-Tris . HCl (pH 8.0), 4.1 x 10e4 M-oxaloacetate S-acetonyl-coenzyme A. The crystals belong to the tetragonal system with lattice constants a = b = 101.5 A; c = 224.6 8. The space group is P4,2,2 as revealed by the systematic extinctions and the final crystal structure. The size of the unit cell suggested one dimeric molecule in the asymmetric unit. (c) Data collection and evaluation Data were collected photographically using a rotation camera (Huber, Rimsting). Crystals were rotated around their c-axis and photographs, in 3 film packs, were taken at intervals of 3”. The films used were Kodak No-screen. The X-ray source was a Rigaku Denki rotating-anode generator operated at 2 kW with a nominal spot-size of 0.1 mm x 0.1 mm in projection. The X-ray beam was focussed on the film using a system with double glass mirrors (Brandeis University). The films were scanned on an Optronics photoscan P 100 and evaluated using the FILME program system (Schwager et al., 1975) modified by W. S. Bennett. The data collection statistics are given in Table 1. The data reduction was performed with the PROTEIN program system (St’eigemann, 1974). (d) Patterson search techniques The orientation and location of the molecule, and the space group were determined by Patterson search techniques (for a review, see Huber, 1969), using the 2 structures of citrate synthase determined previously (Remington et al., 1982). The orientation of the molecule was determined using the real-space option of the PROTEIN program system. Structure factors of the given model in space group Pl were calculated for a given model in a cell large enough to eliminate intermolecular vectors shorter than about 20 A. A Patterson map was then calculated, and the N grid points with the largest density values in a shell about the origin were selected (the peak set). A threshold was chosen such that N was 2000 to 6000. The program then reads in the asymmetric unit of the Patterson synthesis; and performs a product correlation of the Patterson synthesis and the rotated TABLE 1 Data statistics Measured unique Measurements 55,485 I, is the intensity sources of I,. t Rmerge= 1 c w h i

reflections 18,318

of the P4,2,2

form

(%)

Completeness~ (%)

(Shell)

10.7

63.3

11.0

Rmerge P

observed on the ith source, and (I,,) is the mean intensity

Resolution (d) 2.95-2.9 of the reflection for all

h $ Completeness and (Shell) are the ratio of the measured to the possible number of reflections for the nominal resolution of 2.9 d and for the last shell as indicated by resolution.

“Oti

G. WIEGBND

ET AL.

peak set in steps of typically 5” in the selected angular syst,em. summing :\’ points per orientation. Two angular systems were used in this investigation. In the first system (z, z’, y”), the model is rotated first about the z-axis, then about the new x-axis and finally about the resulting y-axis. This system has the advant,age that. the symmetry of the correlation calculation is easy to relate to the crystallographic symmetry. The polar angle system (definition of the polar angle system: kappa is the rotation angle around an axis, whose inclination towards the y-axis is psi. It makes an angle phi with the z-axis when projected into the zz plane) was used in a similar series of calculations to find the local 2-fold axis in the Patterson map. In this case; only the kappa = 180’ section was calculated, as we were interested only in the location of the 2-fold axis. Given the orientation of the molecule, the space group and the translation vector placing the molecule in the unit cell were determined in reciprocal space using the modified Tra,nslation Function described by Crowther & Blow (1967). using a program written by E. E. Lattman. By choosing the particular intermolecular vector for which the program is to search; it is possible to distinguish between a symmetry element that consists of a pure rotation and one that contains a translation element, e.g. a 2-fold axis from a 2-fold screw axis, in a single calculation. The Translation Function coefficientss generated by the program are the same in both cases, since they depend only on the rotational component of the symmetry operator, but the peak will be either on the Harker section at 0 or at l/2 in the example given. A distinction between the enantiomorphic space groups P4,2,2 and P4,2,2 can be made by calculation of the vector sets related by the 4-fold screw axis. These are then found on either z = l/4 or 3/4, depending on the sense of rotation. (e) Model building and crystallographic

rejnement

The electron density map and model were inspected on a computer graphics system consisting of a Vector General 3400 and a PDP 11/40 linked to a VAX 11/780. Alterations to the model were performed using the FRODO program (Jones; 1978). Both unaveraged and a.veraged 2F,,,-Fcalc “Sim-weighted” Fourier maps (Sim, 1959: Hendrickson & Lattman, 1970) were inspected. The PROTEIN program was used to average the Fourier maps about the non-crystallographic 2-fold axis, the position of which wars redetermined by the algorithm of Kabsch (1978) after each round. of crystal!ographic refinement from the co-ordinate sets of the 2 monomers. After the solution of the orientation and translation problem, the initial model was subjected to rigid-body reciprocal space refinement using the CORELS program (Sussman rt al., 1977). The entire dimeric molecule was first refined as an entire unit, then broken up into 4 domains, which were treated independently. Further crysta.llographic refinement was performed using the fast DERIV/EREF program system of Jack & Levitt (1978), as implemented in this laboratory. For a complete discussion of the use of this technique, see Remington et al. (1982). (i) Treatment of non-crystallographic

symmetry

A particular advantage of the DERIV/EREF technique is that the crystallographic term in the function minimized is calculated separately from the energy term. This fact can be used to enforce exact non-crystallographic symmetry by simply averaging the normal equations calculated by DERIV about the non-crystallographic symmetry axis, equivalent, t)o averaging the difference Fourier map but simpler to program. A single monomer can be then refined subject to both sets of observations, effectively reducing the number of variable parameters by a factor of 2. This technique was successfully used in the initial stages of the refinement. of the P4,2,2 crystal form and may be applied to a problem with any number of general transformations. At a later stage, the EREF program was modified (a trivial change) to permit the coordinates of one monomer to be restrained to the other, after transforming the second by the non-crystallographic symmetry operation. Given that the 2 monomers (Xl, x2) are

CITRATE

SYNTHASE

related by a general transformation

COMPLEX

209

STRUCTURE

of the form: Xl

= A*X2+T

where A is an orientation matrix and T a translation during crystallographic refinement then becomes:

vector. The function

minimized

E(Xl)+k*R+w*x

(Xl,-(A*X2,+T))‘, I where E(X1) is the energy associated with the co-ordinate set Xl; k is the relative scaling of the crystallographic residual, R, and w is the weight applied to the non-crystallographic symmetry term. The summation is over all atomic co-ordinates Xl;, X2, of the 2 monomers. This must of course be applied in 2 steps, one for each subunit. This allows one to weight arbitrarily adherence to non-crystallographic symmetry, but the problem of choosing k and w is not trivial.

3. Results and Discussion (a) Composition

of the crystalhe

material

As described in Materials and Methods the stoichiometry of X-acetonylcoenzyme A binding has been determined. The amount of protein in each of the seven experiments varied between 0.12 nM and O-58 nM. The molar amount of X-acetonyl-coenzyme A found per dimer varied between 1.3 and 2.0 (Table 2). These experiments indicate a high, if not full occupation of the two binding sites with the inhibitor in the citrate synthase dimer. S-acetonyl-coenzyme A is a competitive inhibitor of the binding of acetyl-coenzyme A to citrate synthase (Rubenstein & Dryer, 1980) with a Ki of 8.8 ymol/l. K, for acetyl-coenzyme A is 7 pmol/l. Therefore, we expected replacement of X-acetonyl-coenzyme A upon soaking the crystalline material with acetyl-coenzyme A. However, after about two hours cracks developed perpendicular to the c-axis when the crystals were soaked in a buffer (described in Materials and Methods) in which S-acetonylcoenzyme A was replaced by lop3 M-acetyl-coenzyme A. (b) Crystallography and rotation of zero-layer precession photographs From inspection photographs, Laue class 4/mmm was apparent. Systematic absences indicated

TABLE

S-acetonyl Protein

0.12 O-18 0.58 o-14 0.25 0.14 0.16

(nM)

coenzyme A or acetyl-coenzyme Substrate

(nM)

0.2 acetyl-CoA 0.28 acetyl-CoA 0.76 S-acetonyl-CoA

O-18S-acetonyl-CoA 0.39 S-acetonyl-CoA 0.29 S-acetonyl-CoA 0.21 S-acetonyl-CoA

2 A content of crystalline

material

Substrate/Protein 1.7 1.6 1.3 1.3 1.4 2 1.9

210

G. WIEGAND

ET AL,

space group P4,2,2 or P4,2,2, but we could not exclude possibilities like P&22 or P4,22, due to the weakness of the zero-layer photographs. Space group

either

P4,2,2 was assumed for subsequent calculations. The Patterson map was calculated on a 1.0 J! grid using data from 6.0 to 3.0 8. When data in a resolution range lower than 6.0 a were included in the calculation, large ripples about the origin were observed with a wavelength of about 7.0 or 8.0 A, indicating that there may be a problem with the very lo-w-angle data. ,< local 2-fold axis was detected using the Pat’terson search t’echnique (the “self-rotation function”) described in Materials and Methods. A number of trial self-rotation function calculations were necessary before the local 2-fold axis could be located. Contrary to previous experience wit’h calculations of this sort, the peak did not become apparent until more than 25% of the grid points in the asymmetric unit of the Patterson map were included in the calculetion. A peak with 72% of the height of the origin was obtained with a peak set consisting of 4000 points out of the 15,468 points in t’he shell 6.0 t’o 20.0 L%from the origin. This peak was at phi = 0, psi = 10” on the kappa = 180” (polar angles) section of the self-rotation function, indicating a 2-fold axis in the a,b plane Wed about 10” from the b-axis. A general three-dimensional search in a different angular system (z, x’, y”) yielded the same solution, the highest peak in the map was 5.5 sigma. Subsequent fine scans with a peak set of 6000 peaks indicated that the local 2-fold axis in the a,b plane tilted 8.0” from the b-axis. Model Patterson searches were conducted with three models in parallel, the final refined (R = 0.19 to 2.7 8) “0 p en -f arm” model transformed into the frame of reference of the “closed-form” model, and a truncated model derived from the closed-form model by removing the small domain, residues 276 to 380. All models were exact dimers generated on the basis of t’he crystallographic 2-fold axis. Structure factors were calculated in a cell with axes 120 a x 100 L% x 120 $ in space group P2, with the model dyad coinciding with the crystallographic dyad. This set of st,ructure factors was expanded to a triclinic set for calculation of model Patterson maps. In each case about 6100 grid points with density >200 were selected from the 32,892 points in the shell 5.0 to 20.0 d from the origin. A general three-dimensional search in the (2, x’, y”) angular system, calculated in steps of 5”, yielded a peak 4.70 sigma, the highest in the map, at (5, 0, 160”) for the complete closed-form model. The truncated model yielded a higher peak, 5.20 sigma at (IO, 0, 155”). The next highest peak was 3.00 sigma at (25, 25, 110”). A general search was not performed with the open-form model, but a fine scan yielded a large peak at (6, 0, 157”). A fine scan with the truncated model indicated a solution at (8, 0: 157”), in perfect agreement with the orientation of the local 2-fold axis found in the Patterson map. After orienting the dimer by the solution of the rotation function, both the correct space group and the translation correctly positioning the dimer model in the unit cell were solved using the translation function programs. Structure factors were calculated for the correctly oriented truncated dimer, with its center of mass at the origin, in a cell with dimensions 160 a x 160 L% x 160 A (the dimer is 80 a across). The Translation Function (TF) was calculated for three sets of intermolecular vectors. The first calculation eliminated the possibility of a 4, axis

CITRATE

SYNTHASE

COMPLEX

STRUCTURE

“11

along c. We calculated the TF for two molecules related by two applications of the 4-fold screw axis. The rotational component of the symmetry operator takes x, y, z into -x, -y, Z. Table 3, below, indicates which space groups can be distinguished by this calculation, and the expected location of the intermolecular peak. A large peak (10.4 sigma) was found on z = l/2 at (98, 10, 110) in a 100 x 100 x 220 grid unit cell and no peak on layer z = 0. This also gives us x = 49, y = 5 (plus or minus 112, or 50 grid units). The next calculation yielded the z component of the translation vector and eliminated the possibility of P4,22 or P4,22. This was done by considering the 2-fold (or 2,) symmetry operator along b. All of the possibilities below (Table 4) have the same rotational symmetry operator, which takes x, y, z into -x, y, --z. A large peak was found at y = l/2 at (48,50, 93). There was no peak on layer y = 0. The last calculation yielded the space group by considering one application of the 4-fold axis. The rotational operator takes x, y, z into y, -x, x. A large peak was found at (94, 4, 165) on layer z = 314 and none on layer x = l/4. Therefore, th.e space group is P4,2,2 and the translation plaeing the dimer in the unit cell is given by x = 49, y = 5, and x = 19. In Figure 1 we present the contoured Harker sections containing the above three peaks. The same peaks were present for the complete models of the open and closed forms of the enzyme, but reduced to about 2.0 sigma. This is quite disturbing, and indicates that the Translation Function is very sensitive to the presence of incorrect information, i.e. in this case the somewhat incorrect structure of the small domain, discussed in greater detail below. All t.hree of the models were placed in the P4,2,2 unit cell, and structure factors were calculated, R values between 6 and 3.5 A resolution were: open-form 0,411, closed-form 0*495, truncated model 0.388. These results suggested that the new form resembles the closed form more closely than the open form. Similar indications had come from the rotation-function calculations. Refinement TABLE 3 Molecule 1

Spacegroup P4,2,2 P4,2,2 I%,22

Molecule 2 --z, -y, -x,.-y, --z, -y.

Zf1/2 z+1/2 f

Harker vector 2x, 2Y> l/Z 22, 2Y, l/2 2x, 2Y, 0

TABLE 4 Molecule 1 2, y, x3 y, x, y, x, y.

z 2 7. z

Space group P4,2,2 P4,2,2 P4,22 P4,22

Molecule 2 1/2-x, 1/2+y, 1/4--z 1/2-x, 1/2+y, 314-Z --z, y> --z --2, y> --z

Harker vector 1/2+2x, 1/2+22,

l/Z, l/4+22 l/2, 3/4+22 2x, 0, 22 2x, 0, 22

-“I:!

G. WJEGAXD

ET At.

TABLE Molecule 1

Space group P4,2,2 P4,2,2

5, y3 f x, y> 7.

5

Molecule 2 1/2+y, 1/2+y,

l/2--2, 1/2-z,

3/4+z 1/4+2

___-

Harker vector -. .s--y+ l/2, y+z+ l/2, l/4 x-y+ I/2, y+x+ l/2, 3/4

-Y

r i X

a

FIG. l(a) to (c). Translation Function (a) Harker section z = 112; (b) Harker section y = l/J; (c) Harker section z = 3/4. The contour levels are in steps of 1 sigma from 2 sigma on.

proceeded by applying rigid body refinement (CORELS; Sussman et al., 1977) and using the truncated dimer model. Three cycles produced shifts in positional parameters of 0.11, -0.09, 0.09 .k and reduced R to 36.1%. These shifts were applied also to the closed model. R was 38.3% (G-3.5 !I data). A Fourier synthesis

CITRATE

SYNTHASE

COMPLEX

STRUCTURE

213

FIG. l(b).

confirmed that the closed form was an appropriate starting model for further refinement. Energy-restrained crystallographic refinement (EREF; Jack & Levitt, 1978) was applied. Data to 3.5 a were used first, later 3.0 il data were included. Fourier coefficients were weighted according to Sim (1959). The shift factor was 0.2. The R value dropped from 36.1 to 28.6% in seven cycles. The root-mean-square (r.m.s.)

r

0

FIG. l(c).

TABLE 6 Matrix of r.m.s. deviations of all atoms of the small and the large domains of the closed form (C2), the open form (P41) and the two independent subunits of the closed oxaloacetate form c2 c2 P41 Small domain

P43/1 P43/2

1.35 748 l-00 697 0.99 697

P41

P43/1

P43/2

0.72 2299

0.56 2336 0.77 2341

0.54 2320 0.75 2338 0.29 2360

1.87 662 1 G36 662

0.32 666

Large domain

CITRATE

SYNTHASE

COMPLEX

STRUCTURE

215

deviation of bond lengths from the equilibrium values was 0.016 A. At this stage the model was rebuilt using a graphics display operated with the FR’ODO program system (Jones, 1978). Thirteen cycles of EREF followed, with various experiments with averaged shifts. R was 23.1%. The model was rebuilt at this stage. After 17 cycles of EREF applying non-crystallographic constraints and restraint’s as described in Materials and Methods, followed by six cycles of refinement with no restraints on non-crystallographic symmetry the final R value was 19.4% (14,332 reflections from 6.0 to nominal 2.9 A resolution). Individual temperature factors were refined. These were averaged over atoms of an amino acid residue, separately for main-chain and side-chain atoms. After superimposing the two monomers in the asymmetric unit, the r.m.s. deviation for all atoms is 0.37 A, which is precisely the co-ordinate error calculated for this data set using the method of Luzzati (1952). The r.m.s. deviation from ideal bond lengths is 0.014 A, and from ideal bond angles is 2.7”. (c) Molecular conformation All three forms, the open (P41), the closed form (C2) and the P4,2,2 (P43) form, described here were compared by superimposing the domains individually. It is clear that the new form is closed with respect to the relative arrangement of the large and the small domains. From kinetic studies, Johansson & Pettersson (1977) had deduced that the interaction of citrate synthase with its substrates follows an ordered mechanism with oxaloacetate binding first. Acetyl-coenzyme A binding is greatly enhanced in the binary complex with oxaloacetate. Remington et al. (1982) had offered as an explanation that oxaloacetate induces the formation of a closed form in which the binding site of coenzyme A is developed, as shown by the C2 form. The structure analysis of the oxaloacetate complex described here proves this hypothesis. There are substantial differences in the conformation of the domains as shown by the matrix of r.m.s. deviations at all atomic positions of the three forms displayed in Table 6. In the Table, 62 is the closed C2 form, P41

FIG. 2. The small domains of the closed form (C2) and the closed form (P4,2,2) are optimally superimposed and shown in the stereo diagram. The conformations also differ substantially for internal residues as indicated specifically for some residues.

C. WIEGAND

fs’T AL.

(6) E‘rG. 3. The closed form (C2) and the open form (f’4,2,2) are compared by superimposing the 250275 segments. (a) The 274275 main-chain conformation is different. In (b) the 255-280 segments are superimposed and the same segment is plotted.

is the open P4,2,2 form, and P43 is the closed P4,2,2 form. The deviations of the large domains are generally smaller than that of the smaJ1 domains. The la,rgest differences are between the small domains of the I’43 (Z’4,212) form (P43/1, P43/2) and the P41 form, of nearly 2 8. The deviakions are distributed over the whole domain and are particularly evident for some internal aromatic side-chains. Similar observations are made when the small domains of P-13 and C2 are compared as shown by Phe347, Tyr354, Tyr330, Trp284 (Fig. 2; note that in this Figure the most similar pair of small domains is present,ed). Table 3 demonstrates that the small domain is soft and responds to the funct.ionai and environmental state of the enzyme, while the large domain appears to be relatively rigid. This is illustrated in another way by the following experiment. In an attempt to address

CITRATE

SYNTHASE

COMPLEX

STRUCTURE

217

(b)

FIG. 4(a) and density.

(b). Oxaloacetate

as bound

to the P4,2,2

form.

(a) Model;

(b) associated

electron

the question of the existence of a well-defined hinge for the domain motion, we have carefully examined the region of the chain near residue His274 in the open P41 form and the closed C2 form. It is clear that the two conformations abruptly deviate at this location essentially by an alteration of the 274-275 main-chain conformation angles. This is illustrated in Figure 3(a) and (b). In Figure 3(a) we have superimposed all atoms of residues 270-275 and plotted the region from 270 to 280. One can see that residues 270-274 superimpose remarkably well, and one might be tempted to assign the hinge motion to a concerted change (approaching

218

G.

WIEGAND

ET AL.

90”) in the angles of residues 274 and 275. If, however, we superimpose residues 276-280 and again plot the region from 270 to 280 (Fig. 3(b)), we see that the superimposed regions do not agree nearly as well as in Figure 3(a). Small changes take place all along the backbone of 276-280 and many side-chains are in substantially different conformations. The motion by which the molecule changes from the open to the closed form must be very complicat,ed indeed. His274 is a hinge with similar characteristics to those observed for the swit,ch peptide in immunoglobulins (see Fig. 7 of Marquart et al., 1980). It may be significant that, His274 is associated with product (citrate) and substrate (oxaloacetate) binding (see Fig. 4 of this paper and Fig. 15(a) and (b) of Remington et cr,E.,1982) and as pointed out in that paper His274 is in a “disallowed” conformation in t’he closed C2 form. (d) Oxaloacetate binding Figure 4 shows the elect’ron density and the model associated with the oxaloacetate bound to the P4,2,2 crystals. They may be compared with the product citrate bound to the C2 form (see Fig. 15(a) and (b) of Remington et al., 1982). Oxaloacetate and citrate are bound in a very similar manner. The association with His238, His320, Arg329, Arg401 and Arg1421”f shows only small changes. His274 and Asp375, which interact mainly with the si-carboxymethylene group of citrate, also show little change when oxaloacetate (which lacks the carboxymethylene group) is bound. Figure 4 demonstrates that there is unexplained electron density at the si-carboxymethylene position, which might represent bound solvent or part of the acetonyl group of S-acet’onyl-coenzyme A. The dissocation constant for oxaloacetate is below 0.01 rnx. 4-ii for citrate is 0.2 mM (Weidman & Drysdale, 1979). Citrate can bind to both crystal forms, P4,2,2 and C2 (Remington et al.: 1982). The binding to the C2 form seems substantially tighter, but citrate does not induce the conformational change from the open-form to the closed form of the crystals. Oxaloacetate, however; makes the open-form crystals crack. The substantial differences in interaction energy between citrate and oxaloacetate must be due to rather subtle struct.ural differences as the principal binding geometry is the same. As we have seen, there are numerous such differences distributed all over the small domain, in particular when the C2 form and the P4,2,2 form are compared, which have citrate and oxaloacetate bound, respectively. (e) Inhibitor

binding

There is no continuous and interpretable electron density for X-aeetonylcoenzyme A in the Fourier map. This is probably a consequence of conformational disorder as the chemical analysis showed a high occupation of the inhibitor in the crystalline material. The polypeptide loop of Va1314 to Pro316 shown in association with t,he adenine part of coenzyme A in the C2 form is also partly disordered in the P4,2,2 form. t Arg1421

is Arg421

of the other

monomer.

CITRATE

SYXTHASE

COMPLEX

STRUCTURE

219

The analysis of the C2 form with product citrate and coenzyme A bound has shown that the pantothenic acid arm of coenzyme A makes a number of hydrogen-bonding interactions with the enzyme and the adenine portion of coenzyme A, The si-carboxymethylene group of citrate interacts with His274, Asp375 and Asn373. These interactions are disrupted or disturbed when acetylcoenzyme A is replaced by the longer acetonyl-coenzyme A. Disorder is therefore not unexpected. REFEREPJCES Crowther, R. A. & Blow, D. M. (1967). Acta Crystallogr. 23, 544-548. Eggerer, H. & Remberger, U. (1963). Biochem. 2. 337, 2022223. Hendrickson, W. A. & Lattmann, E. E. (1970). Acta CrystaZZogr. sect. B, 26, 136143. Huber, R. (1969). In Crystallographic Computing, Proceedings of the 1969 International Summer School on Crystallographic Computing (Ahmed, F’. R., ed.), Munksgaard, Copenhagen. Huber, R. & Bennett, W. S. (1983). Biopolymers, 22, 261-279. Jack, A. & Levitt, M. (1978). Acta Crystdogr. sect. A, 34, 931-935. Johansson, C.-J. & Pettersson, G. (1977). Biochim. Biophys. Acta, 484, 208-215, 11, 268-272. Jones, T. A. (1978). J. Appl. Crystallogr. Kabsch, W. (1978). Acta Crystallogr. sect. A, 34; 827-828. Luzzati, V. (1952). Crystallography, 5, 8022810. Marquart, M., Deisenhafer, J., Huber, R. & Palm, W. (1980). J. Mol. Biol. 141, 3699391. Remington, S., Wiegand, G. & Huber, R. (1982). J. Mol. Biol. 158, 111-152. Rubenstein, P. & Dryer, R. (1980). J. Biol. Chem. 255, 7858-7862. Schwager, P., Bartels, K. & Jones, T. A. (1975). J. Appl. Crystallogr. 8, 275-280. Sim, G. A. (1959). Acta Crystallogr. 12, 813-815. Steigemann, W. (1974). Doctoral thesis, Technical University, Munich. Sussman, J. L., Holbrook, S. R., Church, 6. M. & Kim, S. (1977). Acta Crystallogr. sect. A, 33, 800-804. Weidman, S. W. & Drysdale, G. R. (1979). Biochemistry, 18, 3822-3827.

Edited by J. C. Kendrew