The switch between two conformations of adenylate kinase

The switch between two conformations of adenylate kinase

J. Mol. Biol. (1988) 203, 1021-1028 The Switch Between Two Conformations of Adenylate Kinase Dirk Dreusicke and Georg E. Schulz Institut fiir Organi...

3MB Sizes 0 Downloads 51 Views

J. Mol. Biol.

(1988) 203, 1021-1028

The Switch Between Two Conformations of Adenylate Kinase Dirk Dreusicke and Georg E. Schulz Institut fiir Organische Chemie und Biochemie der Universittit Albertstr. 21, 7800-Freiburg i.Br., F.R.G. (Received 12 October 1987) Crystalline adenylate kinase from porcine muscle cytosol can assume two interconvertible structures. Here, we report the refined structure of crystal form B at 3.3 A resolution and compare it with crystal form A. Crystal forms A and B can be interconverted by protonation and deprotonation of His36, which is located deep in the active center cleft. The changes concern the molecular packing as well as the polypeptide chain conformation. On conversion from crystal form A to B, the N-terminal a-helix unwinds, the active center cleft opens to some extent and the nucleotide-binding glycine-rich loop 15-22 at the active center is detached from the bulk protein. This loop has counterparts in various important mononucleotide-binding proteins and is known to bind a phosphoryl group in adenylate kinase and in the oncogenic ras proteins. It is most likely involved in the phosphoryl transfer and the concomitant conformational changes. It is suggested that the two observed conformations are relevant for enzyme action in solution: they represent two of a series of three known snapshots depicting the enzyme during the substrate binding process.

1. Introduction

to the resolution limit of the crystals, which is 3.3 A (1 A=O.l nm). The results are compared with the high-resolution structure of crystal form A (Dreusicke et al., 1988), showing the conformational differences.

The suspicion that protein conformations in crystals may differ radically from those in solution is still alive (Harrison et al., 1975; Brown et al., 1986; Furey et al., 1986). On crystallization, there do occur limited perturbations (Blevins & Tulinsky, 2. Materials and Methods 1985). Crystal forces can change the chain mobilities at the packing contacts (Sheriff et al., (a) Data collection 1985) and/or they can select and freeze one or Crystals of porcine cytosohc adenylate kinase were several among a number of different conformations grown at pH 6.9 in crystal form A (Schulz et al., 1973). (Huber, 1979). Multiple conformations of the same For the conversion to crystal form B, and for storage and protein are of great interest, since they may reveal data collection in form B, the crystals were transferred to intermediates of mechanical movements of this a solution containing 3 M-ammonium sulfate in protein. 0.15 M-Tris-maleate at pH 58. Crystal form B belongs to Here, we are concerned with adenylate kinase space group P3,21 with unit cell axes a= 6 =48.5 A and (ATP : AMP phosphotransferase, myokinase, EC c= 139 A. Space group and unit cell dimensions agree 2.7.4.3), a small ubiquitous kinase (Noda, 1973) with crystal form A except for a shortening of the c-axis by2A. that undergoes large conformational changes during A data set of the native protein, consisting of 3257 catalysis (Kalbitzer et al., 1982; Egner et al., 1987), reflections with h, k, 120 in the resolution range co to as has been postulated for all kinases (Jencks, 3.3 A, was collected from 5 crystals at about 4°C 1975). The reported species is porcine cytosolic following a procedure described earlier (Schulz et al., adenylate kinase, which exists in two crystal forms, 1973). While the diffraction limit of crystal form B is at A and B. These crystal forms can be interconverted 3.3 A, the X-ray data of crystal form A could be collected by pH changes and by soaking with a number of to 2.1 A resolution. This difference is also reflected in effecters (Pai et al., 1977). Following an earlier Wilson plots, which show overall temperature factors of analysis at low resolution (Sachsenheimer & Schulz, about 50 A2 and 25 A2 for crystal forms B and A. 1977), the analysis of form B has now been carried respectively. 1021

0022-2836/88/201021-08

$03.00/O

0

1988

Academic press Limited

1022

D. Dreusicke

and G. E. Schulz

Table 1 Course of structural Number of cycles completed Resolution range (A) Number of reflection# R-factor (o/c) Bond length (A)’ Bond angle (degree)e Planarity (trigonal) (A)’ Planarity (non-trigonal) (A)’ Bad contacts (A)’ Temperature factor variation

(L%*)~

rejinement

0” 19-3.3 3ooo 58.3 0.026 3.3 0.020 0.028 0.77 0

21sb 193.3 3000 30.1 0.023 57 04x29 0.047 0.26 6.4

400’ lcL3.3 3000 26.8 0.027 4.0 0.024 0.022 0.08 5.2

“Refined molecule of crystal form A (Dreusicke et al., 1988) after displacing it by (+ 1.0, -2.6, + 2.0) A, given in orthogonal co-ordinates, to its position in form B. The shift vector was derived from a comparison of the main heavy-atom sites of the m.i.r. analyses of forms A and B (Schulz er al., 1974; Sachsenheimer & Schulz, 1977). A uniform temperature factor of 33 A* was used. ‘This intermediate model contains 5 C’ atoms with the wrong chirality. ’ Final refined model. All chiralities are correct, 1 water molecule and 1 sulfate ion are included. The final shift vector derived from the r.m.s. overlay of the 135 c” atoms with residual distances of less than 2 A is ( + 1.0, -2.7, + 1.6) A in orthogonal co-ordinates. JJNumber of independent reflections used that is 94% of all independent reflections of the resolution range. The original data set contained 3257 reflections with h,k,Z20 in the resolution range co to 3.3 A. The refinement program of Tronrud et al. (1987) takes only reflections with F>O into account. The data quality is described by the internal R-factor, R,,, = 8%. This R-factor is defined as Rin, = 2 x C(F,,c -F,J/Z(FhLc + F&, where the sums are taken” over all h > k. ’ r.m.s. deviations from standard geometries. f r.m.s. deviation from the minimum distance between 2 non-bonded atoms as calculated for all bad contacts ( = collisions). gr.m.s. difference between temperature factors of neighboring bonded atoms.

(b) Structure re&ement The crystal structure was refined with the program package of Tronrud et al. (1987). In this restrained procedure, all reflections with F>O are given unit weights, while the weights on the geometric restraints can be set according to the progress of the refinement. As initial co-ordinates, we took the refined molecular model of crystal form A (Dreusicke et al., 1988) and placed it on its approximate position in form B, using the shift vector derived from the main heavy-atom sites of the m.i.r.t analyses of forms A and B (Schulz et al., 1974; Sachsenheimer & Schulz, 1977). All solvent molecules were removed from the model. Some data of the refinement are given in Table 1. The apphed mesh srze for the fast Fourier transforms was 1.0 A (50 x 50 x 144 grid points) and the cycle time was 130min on a Vaxll-730. With the 1519 atoms of the model, the number of refined co-ordinates exceeded the number of reflections so that restraining to standard geometries was essential. Table 1 shows the refinement at 3 stages. The initial structure gave an R-factor as high as 58.3O/o. Its r.m.s. deviation of the bad contacts of 0.77 A is much higher than in crystal form A (Dreusicke et al., 1988), indicating that the differences between crystal forms A and B are more substantial than a mere molecular shift. During the refinement we intervened 9 times by manual model fitting to the electron density maps. Mostly, we built into (2F,,,-Fc,,,) maps after checking the status with (Elobs-F,,,J difference Fourier maps. The most prominent conformational changes at the N terminus and at the glycine-rich loop were followed with omit-maps (Bhat & Cohen, 1984). t Abbreviations used: m.i.r., multiple replacement; r.m.s., root-mean-square.

isomorphous

After cycle 218, the R-factor had dropped appreciably, but the geometry was not satisfactory (Table 1). Although the r.m.s. deviation of the bad contacts had decreased to one-third of the starting value, it was still much higher than the r.m.s. value of 0.1 A suggested by Tronrud et al. (1987). Moreover, the bond angle deviations were high and there were 5 c” atoms with wrong chiralities. After 400 cycles the refinement was stopped, because no further improvement could be expected. The final R-factor is 26.8% and the deviations from standard geometry have acceptable values. In spite of the limited resolution, 2 solvent molecules have been included because they were obvious. We refrained from the possibility of combining up to one-third of the resulting phases with previously derived m.i.r. phases (Sachsenheimer & Schulz, 1977) because these m.i.r. phases suffered strongly from non-isomorphism and they were dominated by a single heavy-atom site close to the glycine-rich loop 15-22. Under such circumstances one has to expect that the m.i.r. phases introduce electron density distortions around the heavyatom site and thus affect loop 15-22 and its environment, which is the most interesting part of the structure.

3. Results and Discussion (a) Accuracy

of the model

analysis was clearly hampered b the high overall temperature factor of about 50 A 9, limiting the available data. A plot according to Luzzati (1952) gives an upper estimate of 0.6 A for the co-ordinate error. Therefore, all structural interpretations have to proceed with The

reported

structure

caution. The quality of the electron density map can be visualized in Figure 1. This map shows that,

although

backbone dihedral

angles can be derived,

Conformational

Change

1023

of Adenylate Kinase

Figure 1. Stereo view of the glycine-rich loop in crystal form B. Depicted are residues 15 through 23 with the sequence Gly-Gly-Pro-Gly-Ser-Gly-Lys-Gly-Thr. Arg132 is given in addition because it contacts the bound sulfate ion in crystal form A. The electron density is from the final (2F,,,- Fcalc) map, cut at 16% of the maximum.

their accuracy remains limited. A particular difficulty arose at the extended N-terminal chain, which could be safely placed only after omit-maps were applied. (b) Crystal

packing

In order to establish the relative molecular positions and orientations in crystal forms A and B, we superimposed the reference molecules by applying the algorithm of Kabsch (1978) to the 135 C” atoms with residual differences of less than 2~0 A. The resulting center of mass shift is (AZ, Ay, Az)=(+l.O, -2.7, + 1.6) A in orthogonal coordinates. The rotational differences around the orthogonal z, y and x-axes are 1*3”, 1.3” and l”, respectively. These angles are below the error limit, because the largest shifts caused by these rotations are appreciably smaller than the Luzzati error estimate. The differences between the molecular packings in crystal forms B and A are shown in Figure 2. A detailed survey of all molecular contacts is compiled in Table 2. The switch from crystal form A t,o B augments the number of contacted neighbor molecules by seven. The largest contact is across a dyad and involves the “open mouths” of the active center regions; the molecules “bite each other”. Compared to form A, this contact has enlarged in form B, where they “bite each other ferociously”. In contrast, the smaller contact II across a second dyad has decreased as a result of the unwound N-terminal a-helix. All new contacts in form B as compared to form A are rather small. Adding all contact areas of the reference molecule in form B gives 4700 AZ, which is 33 o/o larger than

the equivalent value in form A. However, the surface area of the reference molecule in form B is 13,900 8’ and thus 25% larger than in form A, so that the fraction of surface buried on crystallization is rather similar, 34% in form B and 32 y. in form A. The increase of surface in form H correlates with its more open structure, the most obvious indication of which is the unwinding of the N-terminal a-helix. (c) Conformational

changes

In order to establish the conformational changes undergone by adenylate kinase, and the molecular shift in the unit cell, the reference molecules of crystal forms A and B have been overlaid as described above. The residual differences are depicted in Figure 3 and quantified in Figure 4. Besides the unwinding of the N-terminal a-helix and the movement of the glycine-rich loop 15-22 with the sequence Gly-Gly-Pro-Gly-Ser-Gly-LysGly, there are several movements over distances up to 8 A. At the corners around residues 54, 63 and 135 there are changes opening the active center cleft in form B. The changes at the corners around residues 83 and 107 are close to the N terminus and may therefore be coupled to the unwinding of the N-terminal helix. It should be noted that the unwinding of this helix is well supported by the final 2Fobs- Fcalc map. Moreover, the molecular shift from packing A to B would cause an intact Nterminal helix (molecule conformation A) to collide seriously at contact II, because Met1 would touch Asp78’. In contrast to the low-resolution results (Sachsenheimer & Schulz, 1977), the movements of

1024

D. Dreusicke and G. E. Schulz

Figure 2. Molecular packing as illustrated with C” backbone models. The vertical and horizontal straight lines indicate the 3-fold screw axes 3, and the dyads in the crystal with space group P3r21, respectively. The reference molecule is given with thick lines, the given neighbor molecules are numbered in accordance with Table 2. Top; the arrangement of crystal form B. Bottom; the arrangement of crystal form A given for reference purposes.

helix 23-30 near to the main heavy-atom sites of the mi.r. analysis, of helix 120-133 and of the long, interrupted helix 145165 are rather small. Part of this discrepancy may be related to the inaccuracy of the molecular shift vector derived at low resolution. The r.m.s. difference between all C” atoms of forms A and B (Kabsch, 1978) is as high as 3.5 A, mostly due to the N-terminal differences. Disregarding the N terminus by omitting the C” atoms of residues 1 through 5 from the calculation, the r.m.s. difference still remains at 2.3 A, which reflects the substantial conformational changes. (d) Structural changes in the active center The switch from crystal form A to B causes an opening of the active center cleft, which can be visualized in Figure 3. The edges of the cleft, the “lips of the open mouth”, move farther apart, and

the glycine-rich

loop 15-22 is detached from the If we overlay the structure of the complex between yeast adenylate kinase and the substrate analog ApsA (Egner et al., 1987) with the porcine enzyme in forms A and B, we can arrange the structures according to cleft size. bulk

protein.

Form B binds

no ligand

and has the widest

Form A binds one sulfate ion tightly rich

loop

15-22

forming

a

giant

cleft.

at the glycineanion

hole

(Dreusicke & Schulz, 1986), and another sulfate less

tightly

at

guanidinium

groups

of

arginines

(Dreusicke et al., 1988), and it has a narrower cleft. The complex with a substrate analog, finally, is accompanied by a complete closure of the cleft (Egner et al., 1987). So, we seem to have a series of snapshots showing how this kinase closes up on binding substrates. The A to B switch on pH change is most likely triggered by the protonation of His36, because the pK of this histidine (6.3; McDonald et al., 1975; Noda et al., 1975) corresponds with the center of the switching hysteresis at pH 6.4 (Schulz et al., 1973). Moreover, His36 sits deep in the active center cleft, and switching can also be accomplished by binding a number of compounds close to His36 (Sachsenheimer & Schulz, 1977). A more detailed view of the active center changes is given in Figure 5. In form A, His36-NE2 is involved H-bond to Asp93-ODl (N.. 0 ~sta~c,w-e?a~ A N-H . 0 angle = 130”) which ‘in turn is strongly bound to Cys25 (S . . . 0 distance = 2.7 A). This situation changes in form B where His36-NE2 forms a more linear H-bond to Asp93-OD2 (or ODl), which in turn is strongly bound to Ser38-OG. The H-bond to Cys25 is broken. This rearrangement seems to disrupt the H-bonding pattern around sulfate ion-l, held by loop 15-22 in form A, so profoundly that the ion dissociates and the loop moves out toward neighbor molecule 1 in form B. Possibly, the compensation of

Conformation&

Change of Adenylate Kinuse

1025

Table 2 Molecular contacts in crystal form B

Packing interaction I I1 III III IV IV v v VI VI VII VII VIII IX IX Total

Contacting residues’

Buried surface areab (A*)

Area difference to crystal form A’ (AZ)

Relative shift of moleculesd (4

a’-aif t+b, A3 d-c, ef5 f-e, g-h, h-g, i-j, j-60 k-l,, l-k,, m-ml3 n-14 *n15

1730 300 870 870 135 135 120 120 80 80 80 80 40 30 30

+350 -360 + 270 + 270 + 35 + 35 + 80 + 80 + 80 + 80 + 80 + 80 f40 + 30 + 30

3.3 5.4 3.3 3.3 0 0 5.4 5.4 5.0 5.0 5.0 5.0 6.2 5.0 5.0

4700

+ 1180

a Contacting residue lists are represented by a,b,c o, whereas neighbor molecule numbers are given as indices 1,2, 15. The threshold contact distance is 5 A. The lists are: a: 17, 18, 38 to 44, 53 to 72, 75, 93, 94, 97, 104, 135, 136, 138, 139, 144, 149. b : 53, 78, 79, 81 to 84, 107 to 109. c: 117, 119 to 122, 150 to 155, 158, 162, 165, 166, 168, 170 to 172, 174 to 176, 181, 185, 189, 192. d : 23, 26, 27, 29, 30, 34 to 36, 41, 44, 48, 49, 128, 131 to 135, 179. e: 185, 188, 189, 191. f:50, 52, 54, 55, 59. g : 0( =acetyl), 1 to 4. h: 100, 103, 163. i: 147, 148, 151, 152, 155. j : 0( = acetyl), 1. k : 123, 127. I : 85, 86. m : 140, 142. n : 140. 0 : 108. ‘The buried surface area has been derived from the solvent accessible surface areas of the residue lists a through o alone, and in the respective contact arrangement using the program of Kabsch & Sander (1983). ‘Packing interactions I through V are also present in crystal form A, whereas interactions VI through IX occur only in crystal form B. There are substantial differences in the contacting residue lists (see Dreusicke et al., 1988) and contact areas. The total accessible surface areas of the reference molecules in crystal forms B and A are 13,SOOAZ and 11,100 A*, respectively. d These are the lengths of the difference vectors between the shifts of the neighbor molecules and the shift of the reference molecule. ‘The reference molecule is located in the z,y,z ranges (-0.88.. f0.34, -0.54.. f0.31, -0.11.. +0,34). f The neighbor molecules 1,2. .15 are related to the reference molecule by the rotations and/or translations in fractional co-ordinates:

1026

D. Dreusicke and G. E. Schulz

Figure 3. Stereo view of the c” backbone models of the molecules in crystal form B (thick line) and crystal form A (thin line). The overlay is based on the 135 C” atoms with residual distances of less than 2 8. Segments 42 to 66 and 126 to 149, which border the large active center cleft located at the upper right-hand side, form a more open cleft in form B as compared to form A; the “lips of the open mouth” move farther apart in form B. The magnitudes of the shifts are

given in Fig. 4.

the negative charge at the carboxyl by protonated His36 removes the charge compensation for the rather buried E-amino group of Lys21, so that Lys21 is expelled, turning the glycine-rich loop 1522 inside out. In form B, loop 15-22 is detached from the bulk protein and makes a contact to neighbor molecule 1 by forming an H-bond between Prol7-0 and Gly64’-N. Comparing the energy of a single H-bond in form B with the energy of the complete contact between loop and bulk protein in form A, it seems unlikely that the loop is pulled out by the neighbor. Rather, the loop seems to record a possible conformational change in an isolated molecule. The notion that both crystal forms represent reasonably

0

20

40

60

80

stable conformations is further corroborated by the observations that (1) the (de)protonation of His36 as the initializing event occurs far away from all crystal contacts, (2) the closely homologous enzymes from man (von Zabern et al., 1976) and rabbit (R. H. Schirmer t G. E. Schulz, unpublished results) crystallize only in form B, and (3) the porcine enzyme crystallizes best in form A, which can then be converted to form B. For a more detailed analysis, however, energy calculations are necessary. The conformation of loop 15-22 in form B is clear enough (Fig. 1) to allow a discussion of backbone dihedral angles. Apart from positions 16 and 19, where Gly and Ser exchange with Ala, all these

100 Residue

120

140

160

180

200

number

Figure 4. Residual differences of the main chain (thick line) and side-chain (thin line) atoms of the molecules in crystal form B and A. The diagram is based on an r.m.s. overlay of 135 c” atoms with residual distances of less than 2 A. Averaging is done by residue.

Conformational

1027

Change of Adenylate Kinase

Figure 5. Stereo view of the active center region in crystal form B (thick lines) and crystal form A (thin lines). Sulfate ion 1 of form A is shown at the center of the picture between residues 17 and 21. Chain ends are marked by dots at the N and 0 atoms.

residues are strictly conserved in the adenylate kinase family (Schulz et al., 1986). While the protein structure in crystal form A showed the reason for conservation of glycine in positions 18, 20 and 22 (see Table 5 of Dreusicke et al., 1988), there was no indication why position 15 cannot tolerate a sidechain. Now in form B, Gly15 assumes a conformation forbidden for residues with side-chains. Moreover, peptide 15-16 is rotated by 180” (Fig. 5) as compared to form A, demonstrating that Gly15 functions as a swivel in the loop movement. Note that Gly15 is the first residue after the B-sheet in form A (Dreusicke et al.. 1988). It should be kept in mind that Gly15 and Lys21, which show the most conspicuous changes (Fig. 5), have been used in all pattern searches for mononucleotide binding sites (Walker et al., 1982; Robson, 1984; Mijller & Amons, 1985; Doolittle et al., 1986; Higgins et al., 1986; Sigal et al., 1986; Bos et al., 1987). They are also conserved in the p21 proteins coded by ras-type genes, where the corresponding loop is known to bind a phosphoryl group of GDP (S. H. Kim, personal communication). In view of the substantial mechanical movements of adenylate kinases it seemed worthwhile to analyze Pro96 in the active center. Since the isomerization of prolines is known to influence the chain-folding process (Schmid & Baldwin, 1978), it is conceivable

that

cis to tram

isomerization

of

Pro96 participates in substrate binding mechanics. In form B, the respective electron density was clear enough to assign a cis peptide at Pro96. A cis peptide has also been found in form A. In the third picture of the snapshot series, the complex with a substrate analog, the respective peptide isomer is

not yet established (Egner et al., 1987). An isomerization during catalysis is further contradicted by the observation that the exchange of Pro96 to Ser in the adenylate kinase of Escherichia coli did not abolish enzyme activity, but rendered the enzyme temperature sensitive (Gilles et al., 1986). In conclusion, we report here on the most open conformation of an adenylate kinase, which appears to be poised for accepting a nucleotide, and which closes up on the incoming substrate. The very well conserved sequence of the glycine-rich loop 15-22 appears to be intimately connected with these movements. This loop binds a phosphoryl group in the oncogenic ras proteins, where it is most likely involved in the conformational changes on GTP hydrolysis. In adenylate kinase, the observed complex mechanics seem to be necessary for burying the transferred phosphoryl group deep inside the protein.

We thank Dr R. H. Schirmer for providing us with crystals of the enzyme and Dr P. A. Karplus for discussions.

References Bhat, T. N. & Cohen, G. H. (1984). J. Appl. CrystaZZogr. 17, 246248. Blevins, R. A. & Tulinsky, A. (1985). J. Biol. Chem. 260, 42644275. Boa, J. L., Fearon, E. R., Hamilton, S. R., Verlaan-de Vries, M., van Boom, J. H., van der Eb, A. J. & Vogelstein, B. (1987). Nature (London). 327, 293297.

1028

D. Dreusicke

and G. E. Schulz

Brown, W., Wagner, G., Wiirgotter, E., V&k, M., Klgi, J. H. R. & Wiithrich, K. (1986). J. Mol. Biol. 187, 125129. Doolittle, R. F., Johnson, M. S., Husain, I., Van Houten, B., Thomas, D. C. & Sancar, A. (1986). Nature (London), 323, 451452. Dreusicke, D. & Schulz, G. E. (1986). FEBS Letters, 208, 301-304. Dreusicke, D., Karplus, P. A. & Schulz, G. E. (1988). J. Mol. Biol. 199, 359372. Egner, U., Tomasselli, A. G. & Schulz, G. E. (1987). J. Mol. Biol. 195, 649658. Furey, W. F., Robbins, A. H., Clancy, L. L., Winge, D. R., Wang, B. C. & Stout, C. D. (1986). Science, 231, 704710. Gilles, A. M., Saint-Girons, I., Monnot, M., Fermandjian, S., Michelson, S. t Barzu, 0. (1986). Proc. Nat. Acad. Sci., U.S.A. 83, 57995802. Harrison, L. W., Auld, D. S. & Vallee, B. L. (1975). Proc. Nat. Acad. Sci., U.S.A. 72, 43564360. Higgins, C. F., Hiles, I. D., Salmond, G. P., Gill, D. R., Downie, J. A., Evans, I. J., Holland, I. B., Gray, L., Buckel, S. D., Bell, A. W. & Hermodson, M. A. (1986). Nature (London), 323, 448450. Huber, R. (1979). Trends B&hem. Sci. 4, 271-276. Jencks, W. P. (1975). Advan. Enzymol. 43, 219410. sect. A, 34, 827-828. Kabsch, W. (1978). Aeta Crystallogr. Kabsch, W. & Sander, C. (1983). Biopolymers, 22, 25777 2637. Kalbitzer, H. R., Marquetant, R., Rosch, P. & Schirmer, R. H. (1982). Eur. J. Biochem. 126, 531-536. Luzzati, V. (1952). Acta Crystallogr. 5, 802-810. McDonald, G. G., Cohn, M. & Noda, L. (1975). J. BioE. Chem. 250, 6947-6954. Edited

Miiller, W. & Amons, R. (1985). FEBS Letters, 186, l-7. Noda, L. (1973). The Enzymes (Boyer, P. D., ed.), vol. 8, pp. 279-305, Academic Press, New York. Noda, L., Schulz, G. E. & von Zabern, I. (1975). Eur. J. B&hem. 51, 229-235. Pai, E. F., Sachsenheimer, W., Schirmer, R. H. & Schulz, G. E. (1977). J. Mol. Biol. 114, 3745. Robson, R. L. (1984). FEBS Letters, 173, 394-398. Sachsenheimer, W. & Schulz, G. E. (1977). J. Mol. Biol. 114, 23-36. Schmid, F. X. & Baldwin, R. L. (1978). Proc. Nat. Acad. Sci., U.S.A. 75, 47644768. Schulz, G. E., Biedermann, K., Kabsch, W. & Schirmer, R. H. (1973). J. Mol. Biol. 80, 857-864. Schulz, G. E., Elzinga, M., Marx, F. & Schirmer, R. H. (1974). Nature (London), 250, 120-123. Schulz, G. E., Schiltz, E., Tomasselli, A. G., Frank, R., Brune, M., Wittinghofer, A. & Schirmer, R. H. (1986). Eur. J. B&hem. 161, 127-132. Sheriff, S., Hendrickson, W. A., Stenkamp, R. E., Sieker, L. C. & Jensen, L. H. (1985). Proc. Nat. Acod. Sci., U.S.A. 82, 1104-l 107. Sigal, I. S., Gibbs, J. B., D’Alonzo, J. S. & Scolnick, E. M. (1986). Proc. Nat. Acad. Sci., U.S.A. 83, 47254729. Tronrud, D. E., TenEyck, L. F. C Matthews, B. W. (1987). Acta Crystal&p. sect. A, 43, 489501. von Zabern, I., Wittmann-Liebold, B., Untucht-Grau, R., Schirmer, R. H. & Pai, E. F. (1976). Eur. J. Biochem. 68, 281-290. Walker, J. E., Saraste, M., Runswick, M. J. & Gay, N. J. (1982). EMBO J., 1, 945-951.

by R. Huber