J. Mol. Biol. (1992) 224, 1127-1141

Refined Structure of the Complex betwetn Guanylate Kinase and its Substrate GMP at 2-O A Resolution Thilo Stehle and Georg E. Schulz Institut

fiir Organ&he Chemie und Biochemie der Universittit Albert&. 21, D-7800 Freiburg i. Br., Germany

(Received 14 November

1991; accepted 15 January

1992)

The crystal structure of guanylate kinase from Saccharomyces cerevisiae complexed with its substrate GMP has been refined at aOresolutiqn of 2.0 A. The final crystallographic R-factor is 17.3% in the resolution range 7.0 A to 2.0 A for all reflections of the lOOo/o complete Gata set. The final model has standard geometry with root-mean-square deviations of @016 A in bond lengths and 3.0 ’ in bond angles. It consists of all 186 amino acid residues, the N-terminal acetyl group, the substrate GMP, one sulfate ion and 174 water molecules. Guanylate kinase is structurally related to adenylate kinases and G-proteins with respect to its central P-sheet with connecting helices and the giant anion hole that binds nucleoside triphosphates. These nucleotides are ATP and GTP for the kinases and GTP for the G-proteins. The chain segment binding the substrate GMP of guanylate kinase differs grossly from the respective part of the adenylate kinases; it has no counterpart in the G-proteins. The binding mode of GMP is described in detail. Probably, the observed structure represents one of several structurally quite different intermediate states of the catalytic cycle. Keywords: X-ray structure;

guanylate kinase; structure induced-fit

1. Introduction

role in the recovery of (d)GMP and thus in the biosynthesis of GTP and dGTP. Its reaction and its biological function closely resemble those of the adenylate kinases, which phosphorylate (d)AMP. In contrast to the adenylate kinases, which have been studied extensively over the last 20 years resulting, for instance, in several X-ray structures (Egner et al., 1987; Dreusicke et al., 1988; Miiller & Schulz, 1988; Diederichs & Schulz, 1991), much less is known about the guanylate kinases. The only elucidated sequence and structure is that of the yeast enzyme, which is reported here. The structure of the main part of GK (residues 1 to 32 and 81 to 186) agrees well with that of the adenylate kinases (Stehle & Schulz, 1990). Like the adenylate kinases, F,-ATPase, myosin and other nucleoside triphosphate binding proteins, GK contains a glycine-rich loop near the N terminus (residues 8 to 15) that forms a giant anion hole (Dreusicke & Schulz, 1986) accepting the CI and flphosphate of ATP in the adenylate kinases (Miiller & Schulz, 1992). It is most likely that ATP binds in the same manner to GK. The main part of GK is also structurally homologous to the GTP-binding proteins (also known as G-proteins) EF-Tu and H-ras-p21, the structures of which are known

Guanylate kinase (GKT, ATP : GMP-phosphotransferase, EC 2.7.4.8) is a small monomeric protein that catalyzes the phosphorylation of GMP (and dGMP) at the expense of ATP according to Mg*+ ATP + (d)GMP c ADP + Mg*+ + (d)GDP (Miech $ Parks, 1965). Mg*+ is necessary for catalysis. The molecular mass of the polypeptide is 20,548 as derived from the known sequence (Rerger et al., 1989). The enzyme plays an essential tAbbreviations used: CK, guanylate kinase from crrpvisiae; AK 1, cytosolic adenylate kinase from pig muscle; AKeco, adenylate kinase from Escherichia coli; AKyst, adenylate kinase from

fhxAarom.ycr,s

Raccharomycrs cerevisiae; AK3,

adenylate

refinement; GMP-binding;

kinase from

beef heart mitochondrial matrix; AMPbd, AMPbinding domain of the adrnylate kinases; Ap,A, I”,P5-bis(5’-adenosyl-)pentaphosphate; B-factor, c~rystallographic~ isotropic temperature factor; EF-Tu, elongation factor Tu from lkherichia coli; GMPbd, (:MP-binding domain of guanylate kinase; INSERT, domain that distinguishes the large variants from the short variants of the adenylate kinases; MAW, domain comprising the main part of the structures of nucleoside monophosphate kinases; g, standard deviation; r.m.s., root-mean-syuare. 1127 0022-2836/92/O&H

127-15

$03.00/O

0

1992 Academic

Press Limited

1128

T. Stehle

and

(Jurnak, 1985; LaCour et al., 1985; DeVos et al., 1988; Pai et aE., 1990; Tong et al., 1991). GTP binds to these G-proteins at a position corresponding to the ATP site in the adenylate kinases and presumably in GK. The G-proteins have no binding site equivalent to the GMP site of GK or the AMP site of the adenylate kinases. While there is a particular chain fold motif for binding the nucleoside triphosphate, nothing similar exists for the monophosphates. The chain fold motif of the GMP-binding domain of GK is quite different from the AMP-binding motif of the adenylate kinases, and it has nothing in common with the motifs of ribonucleases (Arni et al., 1988; Sevcik et al., 1991; Nonaka et al., 1991) or glutaminyl-tRNA synthetase (Rould et al., 1991) that bind guanine nucleotides. Still, there do exist local similarities in guanine binding that are discussed here. After reporting a partially refined model of GK (Stehle & Schulz, 1990), we now describe the improvement of this motel by refining the structure at 2.0 A resolution (1 A = @l nm) to convergence, and we interpret the result.

2. Materials and Methods (a) Crystal

data

GK was isolated from Saccharomyces cerevisiae and purified as described by Berger et al. (1989). The yield of

G. E. 9chulz

the isolation procedure was about 20%, resulting in 4 mg of enzyme from 1 kg of yeast. The purity was checked by isoelectric focusing. Crystals were grown at 20°C frorn ammonium sulfate with the hanging drop method (Stehle & Schulz, 1990). The lo-y1 drops contained 7 mg GK/ml, 40 mM-potassium phosphate (pH %5), 1 mM-GMP, 1 “/o (w/v) polyethyleneglycol-1500 and 1.0 M-ammonium sulfate as precipitant. As the reservoir, we used 500 ~1 of 40 mw-potassium phosphate (pH 5.5) with 2.0 M-ammonium sulfate. Since no crystal storage buffer could be found, the crystals were mounted directly from the mother liquor. The enzyme crystallizes in space group P4,2,2 with 1 enzyme : substrate complex per asymmetric unit. The unit cell dimensions are a = b = 50% A

and c = 1553 A; the crystals are of bipyramidal shape with sizes up to 1000 pm x 500 grn x 300 pm. The solvent content of the crystals is 49% (V, = 2.4 A3/I>a). The crystals are mechanically labile. For the analysis, crystals had to be selected according to their X-ray diffraction because they grew mixed with a second, related crystal form with space group C222,.

pattern,

(b) Ilejnement Native data collection on a 4-circle diffractometer and an initial refinement of the multiple isomorphous replacement model of GK have been described (Stehle & Schulz, 1990). vative data were derived from 6 different crystals to 2.0 A resolution. The radiation damage was 4% on average. The internal R-factor between symmetry-related reflections was RF,in, = Fi.O’yo/,. The data set contained 14,130 unique reflections, and it was 100% complete. The resulting preliminary sbructure had reasonably good

Table 1 Simulated annealing rejinement protocol using XPLOR REPELb WA” (Mcal/mol)

Round 1 to 3 4 to 7 to 9 10 11 12 to 15 16 17 18 19 20

2 6 8

14

82 a2 82 82 82 82 82 82 113 113 117 117 150 150

300 Kd (PSI

(cycles) 10 10 10 10

MIN-2’

MIN-1’

80 50 40 40

025 030 @20 -

40

0.20

025

BREF’

MIN-2

(cycles) 40 50 40 40 100 150 100 100 100 50 130 150 100 200

10 10 10 10 10

10 10 10 10 10

10

10

15

50

10 10 10

30 30 30

8WA, weight relating the effective energy term accounting for the diffraction data, E,(XRAY), to the empirical potential energy (1 cal = 4184 J). The respective weight WP for the phase information was set to zero in all rounds. bREPEL, conjugate gradient minimization with soft repulsive potential, using the limit AF = 0.05 hi. If any atom movement exceeds AF, the first derivatives of the effective energy E,(XRAY) are recalculated. ‘MIN-1, conjugate gradient minimization with CHARMM non-bonded potential using AF = @05 A. d300 K, molecular dynamics at 300 K, timestep = 1 fs, AF = 0.2 A. In round 20, the slow-cool protocol was used, starting at a temperature of 300 K and ending at 100 K. “MIN - 2, conjugate gradient minimization using AF = O+lO5 A. ‘BREF, Individual B-factor refinement with standard deviations between B-factors of bonded atoms and B-factors of atoms connected by an angle restrained to 1.5 A2 and 20 A*, respectively. In round 20, the standard deviations of side-chain B-factors of bonded atoms and atoms connected by an angle were merely restrained to 35 A2 and 50 A’, respectively.

Structure

of Guunylate

Kinuse

1129

: GMP

Table 2 Course of the simulated Round” R-factor (O&f Number of water molecules r.m.s. bond length deviation (8) r.m.s. bond angle deviation (deg.) r.m.s. improper angle deviation (deg.) r.m.s. dihedral angle deviation (deg.)

annealing 1 25.6 46

0018 37

1.7

251

rejinement 2

with XPLOR

3

242 46 WO16 3.5

7

21.6 93

19% 151

0.016

1.6

245

13

20

18.7 184

17.3 174

3.1

0.015 3.1

0015 30

1.4

1.7

1.3

249

23.9

0.016 3.0

1.4

23.6

23.7

All refinement rounds were carried out in the resolution range 7.0 to 2.0 A using the 100% complete native data set of 13,646 reflections described by Stehle & Schulz (1990). “After round I, the sulfate ion was incorporated into t,he model; after round 2, the N-terminal acetyl group was included.

stereochemistry; the R-factor was 2S.9o/o for all data in the resolution range 7.0 to 2-O A. The refinement has now been continued until convergence using the simulated annealing method (program XPLOR; Briinger et (II., 1987) according to the protocol given in Table 1. In total, 20 refinement rornds were carried out in the resolution range 7.0 to 20 A using all data without any cutoffs. After each round, the model was inspected by (2F, - FJexpia, and (F, - F,)expia, maps. Mispositioned residues and incorrect conformations were changed manually using the program FRODO (*Jones, 1978) running on a PS330 interactive graphics display (Evans & Sutherland, U.S.A.). After round 1 we detected that the segment 83 to 90 of the early model (Stehle & Schulz, 1990) was out of register and shifted it by 1 residue. Further model building focused exclusively on the correction of side-chain errors. misplaced carbonyl atoms and the modeling of solvent molecules. Water molecules were included manually. They were inspect,ed visually for agreement with the geometric requirements for hydrogen bond formation, and were given initial B-factors of 25 AZ. As a second criterion, water molecules were accepted only if the corresponding density was at. least 4 0 in the (F,-FJexpia, map. After each refinement round, we deleted water molecules with densities below 1 a in the (2F,- F,)expia, map. After round 1, the (F, - F,)expia, map showed a very high positive density peak located in the center of the glycine-rich loop. Since the corresponding glycine-rich loops of AK1 (Dreusicke & Schulz, 1986) and AK3 (Diederichs & Schulz, 1991) bind a sulfate ion, and since the GK crystals have been grown from ammonium sulfate, a sulfate ion was built into this density. All atoms of the sulfate were given initial B-factors of 25 8’. The x terminus of OK is known to be acetylated (Berger et al., 1989); this acetyl group (here sequence position 0) was incorporated after round 2. the R-factor When reached 17.3% and the map showed no more interpretable (F, - FJexpia, features, refinement was halted, yielding a GK model consisting of 1449 non-hydrogen atoms of the polypeptide, 24 atoms of GMP, 5 atoms of the sulfate ion and 174 water molecules. The final f.m.s. bond length and bond angle deviations are 0016 A and 3.0”, respectively. The course of the refinement is shown in Table 2. A comparison with the early model presented byOStehle & Schulz (1990) shows r.m.s. deviations of 1.0 A for the (Y-atoms and of 1.8 -4 for all atoms. Much of this deviation, however, is due to the 1 residue shift of chain segment 83 to 90. Without this segment, the r.m.s. deviation for the Cm-atoms is only &7 A.

3. Results and Discussion (a) Quulity

of the model

The accuracy of the atomic positio;s as derived from a Luzzati (1952) plot is about @2 A (Fig. 1). At the 1 (T level of the final (2F,- F,)expicr, map, all atoms of GK are clearly met by density with the exception of parts of the side-chains of Phe73, Thr139, Glu140, Glu141 and Lysl86 at the Cterminus. These residues are located at the protein surface and are presumably very mobile (see Fig. 6). The final model of GK is shown in Figure 2. (b) Chain conformation Although

the main-chain

dihedral

angles were not

restrained during refinement, all but one of the 173 non-glycine residues of GK are found in favorable regions of the Ramachandran plot in Figure 3(a). The single exception is Glu185 at (78”, -31”) which is close to the C-terminus and is not well defined by its density. The torsion angles of three non-glycine residues (Asn65, Ser74 and Asnl14) fall in the lefthanded a-helical region around (60”, 40”). In Figure 3(b) the x1 and x2 torsion angles for the side-chains of leucine and isoleucine residues are

35 r-----

A

30

25 E

/'

/ 0

1 A'

11 04

I 0.2 2 5," e/x

Figure 1. Luzzati

I, 0.3

! 04

1

B ,A

I 05

cH-'1

(1952) plot for the refined model of GK. Depicted is the R-factor for all non-centric reflections. The thin lines emerging from the origin show the upper limits for the theoretical coordinate errors: A, 023 A; B. 020 A; C, o-17 A.

1130

7’. 8tehle and C. E. Schulz

3

56

56

(b)

Figure 2. Stereo views of the refined model of the GK : GMP complex. (a) The C” backbone (including the N-acetyl group) together with GMP and the sulfate ion. Some residue numbers are given. (b) The full model, including all protein atoms (thin lines), GMP and the sulfate ion (thick lines), and 174 water molecules (x). With respect to (a), the model has been rotated 40” around a vertical axis to emphasize the cleft at which GMP is bound.

shown as a scatter plot. These residues are concentrated within the regions for staggered conformations, corroborating the quality of the final model; only the x1 angle of Leul13 falls within +20” of an eclipsed position. The observed frequencies of the nine staggered conformations are in good agreement with the statistics of Janin et al. (1978) and the observations of Karplus & Schulz (1987). The assignment of secondary structural elements strongly depends on the hydrogen bond criteria applied. Baker & Hubbard (1984) have shown that peptide hydrogen bonds are best defined by a hydrogen . . . acceptor distance of less than 2.5 A and an N-H 0 angle of more than 120”. Using these criteria, we have assigned u-helices, P-strands, 3,,,-helices and turn conformations manually by inspecting hydrogen bonds and torsion angles of the final model. The results are given in Figure 4. GK contains one central, five-stranded parallel /Isheet (/39,/38,~1,/.?7,~2), which is surrounded by six a-helices (al ,a3,a4,~5,a6,a7) and resembles the respective part of the adenylate kinases. In addition, there is a four-stranded mixed /?-sheet (fi5$6,~3,~4), consisting of three antiparallel and

one parallel p-strand (84). The j-sheets are sketched in Figure 5. The average (4,tj)-angles in the b-sheets and u-helices follow the usual patterns. The total amount of cc-helical plus P-sheet structure is 73%.

(c) Mobility

of the chain

The average isotropic temperature factor (Bfactor) of the refined GK model is 22.8 A2 for all non-hydrogen atoms including water. The average B-factor of the main-chain atoms is as low as 18.7 A2, which is slightly higher than the value of 17 A” that was deduced from a Wilson (1949) plot (data not shown). In Figure 6, the main-chain B-factors of GK are given as a function of residue position, revealing rather high B-factors at the N and C termini and in seven other regions. Of particular interest are the peaks at positions 12,42 and 74, which are at the active center. Tn GK, the glycine-rich loop (residues 8 to 15), which forms the giant anion hole and binds ATP in the adenylate kinases, is rather mobile. This is in contrast to the structures of AK1 (Dreusicke et aZ., 1988) and AK3 (Diederichs & Schulz, 1991), where it is rigid. The

Structure of Guanylate Kinase : GMY 180

.

:

I.

: .

120

.

..‘.. .

* ..

< ._

. . ..: : z :;’ . ..

::

t

.

60 $ 3

0

: ... . ‘. :&.. ,

-60

- 120

-120

-180

-60

120

0 60 I+ (deg.1

180

(0) 360 a

300

8

1131

The mobility peaks at residues 23,90, 1I1 and the termini are all in chain segments that are close together at the bottom of the molecule as it is depicted in Figure 2. This region is also soft in AK1 and AK3. The peak at residue 140 concerns the chain loop of GK that corresponds to domain INSERT of the adenylate kinases (Fig. 4). Domain INSERT undergoes large movements during a catalytic cycle (Schulz et aZ., 1990). The corresponding segment has the highest mobility in AKl, which is the adenylate kinase most similar to GK; it has also high mobility in AK3. In GK : GMP the average B-factor of the bound substrate GMP is 16.2 A’ and thus lower than the main-chain average. A closer look at the B-factors of GMP reveals a mobility gradient extending along the molecule: The average B-factors of the guanine ring, the ribose moiety and the phosphate group are 11.1 A’, 204 A” and 20.7 A’, respectively. This gradient corresponds to the accessible surface areas of GMP; guanine is deeply buried, whereas ribose and the phosphate group point towards the solvent. A similar binding situation has been described for the complex AK3 : AMP (Diederichs & Schulz, 1991).

240 a

? & 3 x”

180

(d) Solvent structure

0

120 Ll

*

8

60

0

a

60

120

180

240

300

360

X, (deg.) (bl

Figure 3. Chain conformation in GK. (a) A scatter plot of the main-chain dihedral angles of all 173 non-glycine residues of GK. (b) A scatter plot of the side-chain dihedral angles of the 12 leucine (A) and 11 isoleucine residues (0) of GK.

peak at residue 42 concerns the loop region 38 to 50 connecting the parallel b-strands 83 and p4. This segment binds the phosphate group of GMP (see Fig. 12(a)). The peak at residue 74 is between strands 85 and fi6 in a solvent-exposed chain segment of the GMP-binding domain. Presumably, the segments around residues 42 and 74 of the GMPbinding domain (see below) move over long distances during catalysis, as deduced from such motions of the corresponding AMP-binding domain of the adenylate kinases (Schulz et al., 1990). Also in AK1 with no bound substrate, the mobility of the respective AMP-binding domain is above average. In the complex AK3 : AMP, the corresponding segment has the highest mobility.

The GK model contains 174 water molecules, all of which have B-factors below 70 A” and densities above 1 0 in the final (2F,-F,)expicr, map. On finishing the refinement, the water molecules have been renumbered according to their final map density, starting at number 189 after all 186 residues, GMP and the sulfate ion. Thus, Wat189 corresponds to the highest and Wat362 to t’he lowest observed density. The solvent distribution as a function of density is given in Figure 7. Wat189 has density at 5.0 6, and the average is about 2.0 0. Furthermore, the average B-factor of the assigned water molecules is plotted as a function of density in this Figure. The overall average is 39 A’. As expected, there is a negative correlation between Bfactor and density. Following the suggestions of Blevins & Tulinsky (1985) and Karplus & Schulz (1987), we consider 15 of these water molecules as an integral part of the enzyme because they have densities above 3 0 in the final (2F,- F,)expicc, map and B-factors below 25 A”. Among these, six are buried in the interior of the protein with solvent accessibility zero. Most integral water molecules are located in the vicinity of GMP and at the inner side of the long helix ~6. Among the assigned Owater molecules, 158 are within a distance of 3.7 A from the nearest protein atom and thus belong to the inner hydration shell. The remaining 16 molecules are located in a second shell around the protein in the range 3.7 to 5.6 A from the nearest protein atom. They are fixed only by solvent-solvent interactions and have an average B-factor of 50 A”, indicating a rather high level of mobility.

50 60 70 80 ,. I I I AC-SRPIVIS~PSGTGKSTLLKKLFAEYPDSFGFSVSSTTRTPRAGEVNGKDYNFVSVDEFKSMIKNNEFIEWAQFSGNYYGSTVASVKQVSKSGXTCI EEEEE HHHHHHHHHHHHHTTEEEEE EEEE TT TTTT EEEHHHHHHHHHHH EEEEETTEEEEEHHHHHHHHHHH cl1 a2 a3 P4 P5 $6 Pl

I32

(--GDLLRAEVSSGSARGKMLSEIMEKGQLVPLETVLD----) I I I I 40 50 60 70

MLRDANVAKVDTSKGFL I 80 9"

100 110 120 130 150 160 2; I n --t ---l ---LII--.--' LqIDMQGVKSVKAI---PELNA-RFIIFIAPPSV~D~~EG~------TET~S~N~AA-ELAYAETGA---HDKVIDFIFAE~ EEEHHHHHHHHH G GGGG EEEEEE HHHHHHHHHHHH HHHHHHHHHHHHHH HHHHHHT a4 as P -IPGY-PREVKQGEEFERKImm-GPETMT&&LK&ETSGRVDDNEETIK 110

120

EEE 07

P3

AC-MEEKLKK~KIIFW~GP~SGKGTQCEqIVQKY--GYTHLST lI I 10 20 30

100

90 I

-130

140

170

EEEEE P9

180

HHHHHHHHHHHHHHH (17

~ ___ LETYYKATEPVIAFYEKRGIVRKVNAEGSVDDVFSQVCTfiLDTL&

4’

150

160

I

170

I

T

180

190

I

Figure 4. Secondary structure of GK. The 1st line gives the amino acid sequence (Berger et al.. 1989). The 2nd line gives the secondary structure assignment using the rules of Karplus & Schulz (1987): H, a-helix; E, b-strand; G, 3,,-helix; and T, isolated turn with (i&+3) hydrogen bond. The 3rd line gives the amino acid sequence of AKl, which has been structuraily aligned to GK (see the text). The 132 structurally corresponding residues of GK and AK1 (distances below 7 A) are marked by overlines. Conserved residues in AK1 and GK are doubly underlined; the sequence identity is 13%. The domain limits of GK are indicated as filled boxes (m). They are derived from the alignment with AK1 (see the text). (e) Crystal

packing

Frogram IXSP (Kabsch & Sander, 1983) is 10,183 A’. The solvent-accessible surface buried on crystallization is the sum of all contact areas given in Table 3; it amounts to 25% of the total surface, which is in the usually observed range.

In the crystal, every GK molecule (reference molecule I) has seven neighboring molecules forming four different types of crystal contacts: III, T-III, I-IV and T-V (Fig. 8). These interactions are detailed in Table 3. Crystal contacts I-II and IIII are rather extensive, with several hydrogen bonds involved; contacts I-IV and I-V are not very well defined by polar interactions. Contact I-II is by far the strongest of all; it is formed across a crystallographic dyad relating three roughly parallel helices and one loop to their counterparts in the neighboring molecule. The total solvent-accessible surface of a GK molecule as calculated with

168

structure as derived from adenylate kinases

The sequence of GK showed that there exists a distant structural relationship with the adenylate kinases because some chain segments could be aligned (Berger et al., 1989). This relationship was confirmed by the three-dimensional structure of GK

I INqo

167

(f) Domain

1 121

2.97 O-N 76

73 N-O 2.99 2.96

,N 0-1;84

O-rGO

0'

78

71

./Y.-N 2.96

N-O 2'77

116

2.99 N-O

I

Io4N5i 36 lplo5i

70

~",'-"l_o~N.~~~ogjN..~o'~ 3

O-N 2'98

O-N

:N

I 37

77

72

95

10 0’2

__H -94

94

3.14 O-N

N 29

Nd.89 -0 2i I

Figure 5. The 2 b-sheets of GK. The N...O distances of the hydrogen bonds are given in A.

Structure qf Guanylate Kinase : GMP

1133 70 60 50

b

4o

i t 2 cb

30 20 IO 0

20

40

60.. 80

100 120 140 160 180

Residue

Figure 6. The main-chain

temperature factors) of GK, averaged for each residue.

factors

(B-

(Stehle & Schulz, 1990). After refining this structure at high resolution, we have established the best alignment of GK with AK1 (Dreusicke et al., 1988),

which is the most closely related adenylate kinase. For this purpose, we superimposed the C” backbones of GK and AK1 visually such that a good fit was achieved at the central parallel P-sheet and at the giant anion hole. Starting from this well-fitting basic structure, we related all OC” atoms to each

other that

were less than

superposition

of the resulting

7 A apart.

any cut-off.

The best

132 C” atom pairs was

then calculated with program OVERLAY 197t) without

(Kabsch,

The r.m.s. deviation

wts

2.6 A; none of the resulting distances exceeded 7 A. In order to improve the superposition we first introduced

a cut-off

of 4 A (114 pairs,

0 0

number

r.m.s.

deviation

I

2

3

4

5

b

Electron density (a)

Figure 7. The distribution of solvent molecules as a fun&ion of electron densit,y in the final (2F,-F,)expisr, map. The average B-factors of the histogram sections are given as dots connected by a line. Water molecules with densities below 1 c of this map or with B-factors larger than 70 A’ have been deleted.

1.9 A) and then a cut-off of 3 A, resulting in 103 C” atom pairs with an r.m.s. deviation of 1.7 A. The superposition matrices of the three computations with OVERLAY differ by a rotation of about 1” from each other: the translational differences are below 0.4 .&. The superposition of GK with AK1 as based on the 103 C” atom pairs is illustrated in Figure 9. The equivalenced set of GK residues is given in the Figure legend. Using this superposition we assigned the insertions and deletions visually and introduced them into the alignment of Figure 4. The structural

Figure 8. 0ystal packing of the complex GK : GMP, illustrated with c” backbone models. The tetragonal unit cell is outlined. Indicated are the contacts between reference molecule I (thick line) and molecules IV, V, III and II (thin lines) arranged clockwise starting at the upper left. The given view is along a diagonal 2-fold axis that relates molecules I and II to each other. In relation to Fig. 2, molecule I is viewed from the bottom. The contacts across the dyad are mainly between helices ~2. ~14,cr6 and ct6, tl4, ct2 of molecules I and II, respectively (Table 3).

7’. Stehle and G’. E. AWmlz

1134

Table 3 Crystal Molecular packing interactionsa

area (A2)b

Number of contacting atom pairs’

1020

220

Buried SUrfaCe

I-II

contacts Involved residuesd

55 59 62 63 65 70 -~~~->-~-~72,7& 21. 101, &, 102, 105,109, 109, 145, 145, 148, 149, 151, 152, 155, 156, 158, 159

I-III,

I-VI

340

96

-> 56 --> 57 -3 60 -> 61 -> 63 -> 64 -) 66 123, 131, 168-172

I-IV,

I-VII

270

72

I-V, I-VII1

160

32

-1->-_> 130 131 133137 157-160, 165, 167 2 >->-,_>->-I 22 23 26 27 48 114, 163, 185, 186

“The neighboring molecules II through VIII in fractional coordinates: ,,I(,

vl:(%

i -8>

-i

+(a)

:I:> +(-~EJ-~

y;

i

Polar interactions Atom 2 Atom 1 LysSJ-NZ Lys63-NZ Trp’iO-0 Gln72-OEl Gly75-0 Tyr77-OH Glnl02-OEl Glu57-OE2 SerSO-OG Lys63-0 Asn64-ODl Asn64-ND2 4rgl3.5.NE

: GlulWOEl : Glul58-OE2 : LyslOb-NZ : Serl48-OG : Ser148-OG : Alal52-N : Glnl02-NE2 : Aspl72-N : Leu17 1-N : Argl31-NH1 : Argl31-NH1 : Aspl69-0 : Glul58-OEI

a -ii)

‘(EJ

+(-8’1,)

330’ 338’ 3.2 1 3.19 3.22 3.35 3.43 3.04 2.97 342 2.96 3.49 (3.52)’

Ala23-0 : Asnll4-ND2

are related to the reference molecule I by the following

8)

Donor-awe tor distance (1 )

I,:(-,

8 -!)

+ (EJ

vlll(-i

8 j)

+ (-ig

rotations

‘:(-i

2.90 and translations

y i)

given

+ (iz,)

bCalculated using DASP (Kabsch & Sander, 1983). “Pairs of polypeptide non-hydrogen atoms across the interface with distances less than 45 A. dResidues contributed by molecule I to contacts I-11, I-III, I-IV and I-V are underlined, while the residues of the neighboring molecules are not. As contact I-II is formed across a crystallographic dgad, the underlined residues of molecule I contact the others of molcule II and vice verse ‘Saltbridges.

alignment results in 26 identical amino acid residues, which is 26/208 = 13% of the total length (Fig. 4). In spite of this low homology value, the alignments of the four chain segments described by Berger et al. (1989) were essentially correct.

Furthermore, the structural alignment allows us to transfer the domain assignments of the adenylate kinases (Schulz et al., 1990) from AK1 to GK. This transfer is clear for domains MAIN and INSERT. INSERT contains only 11 residues in AKl, a small

Figure 9. Superposition of the C” backbones of GK (thick lines) with AK1 of Dreusicke et al. (1988) as oalculated with program OVERLAY (Kabsch, 1978) using a cut-off of 3 A. The equivalenced 103 C” atoms of GK are 1 to 24, 29 to 32,81 to 91,93 to 99,105, 109, 113 to 122, 124 to 132, 139 to 149, 151, 152 and 163 to 185. The corresponding residues of AK1 can be taken from Fig. 4.

Structure of Guanylate Kinase : GMP

1135

Figure 10. Best superposition of GK (thick line) with complex 1 of AK3 : AMP (thin line) of Diederichs &, Schulz (1991) as based on the following equivalenced set of 89 C” atoms: 3 to 25, 28 to 32, 34, 82 to 90, 93 to 99, 116 to 122, 124 to135,142to147,163to167,169to182ofGK;and7to29,30to34,36,74to82,83to89,108to126,166to171,191to 195, 197 to 210 of AK3. The r.m.s. deviation is 1.6 A. variant adenylate kinase. With five residues, INSERT is even smaller in GK. In the large variant adenylate kinases like AK3, domain INSERT consists of 38 residues. The bordering residues of the AMP-binding domain (AMPbd) of AK1 are 38 and 67. Residue 38 of AK1 corresponds to residue 33 of GK. Since residue 67 of AK1 has no counterpart in GK, we define the second border of GMPbd as the end of the P-sheet at residue 80 (Fig. 5). The superposition of GK with AK1 is best in domain MAIN, where larger distances occur only at helix a4 and at the C-terminal end of the long helix a6 (Fig. 9). The structures are particularly well conserved at the central parallel B-sheet and at the N-terminal ends of helices cl1 and a7. The segments equivalent to domain INSERT form connections of different lengths between helices a5 and ~6. In contrast, the domains GMPbd and AMPbd differ grossly: GMPbd consists of a mixed P-sheet with a short helix a2, while AMPbd is completely a-helical. Obviously, the specific binding of the more polar guanine base requires a structure quite different from that binding specifically adenine. (g) The GMP-binding

site

Let us first relate this site to the adenylate kinases. The two substrate-binding sites of the adenylate kinases are known from the closely similar structures of AKeco and AKyst ligated with the two-substrate mimicking inhibitor Ap,A and the structure of the complex AK3 : AMP. As demonstrated by the superposition of AK3 : AMP onto GK : GMP (Fig.lO), which is based on the structural correspondence in MAIN, the GMP-site of GK corresponds to the AMP-site of AK3, and domains GMPbd and AMPbd are approximately at the same location.

In spite of the same general location, however, there are appreciable differences between the bound nucleoside monophosphates: The chain folds of the GMP-binding and AMP-binding domains are totally different. Although the phosphate groups superimpose well, the locations of riboses and bases deviate considerably from each other (Fig. 10). The rihose conformation in GMP is 3’-endo, whereas it is 2’-endo in AMP. The ribose of GMP interacts only with solvent,

whereas there is a good hydrogen

bond

between the 2’-hydroxyl group of AMP and a backbone carbonyl oxygen atom. It is remarkable that guanine is bound by two carboxyl groups, while no such group binds to adenine. A more detailed description of the polar interactions of GMP is given in Figure 11. Stereo views of the GMP site are shown in Figure 12. There are five hydrogen bonds between guanine and side-chain atoms and one to a main-chain atom. The phosphate group forms five hydrogen bonds to two arginines and two tyrosine residues. Among them, Arg38 is well fixed by a saltbridge to Glu44. This saltbridge should be strong, because it has only a small solvent-accessible surface area (Table 4). Still, the level of mobility of the phosphate group is relatively high (B-factor = 2@7 A2) which is consistent with the high level of mobility of loop 38 to 50 (maximum B-factor = 3@9 A’, see Fig. 6). A conspicuous residue is Phe73 shown in Figure 12(b), which is located above the GMP-binding site. Its side-chain is fully solvent-exposed, the phenyl ring is very mobile with an average B-factor of 37 A’. Given the similarity between GK and the adenylate kinases,

it is most likely

that

the GMP-binding

domain undergoes a conformational the enzyme binds ATP in addition expect that these changes are small ring of GMP, which forms contacts

change when to GMP. We at the purine to MAIN (see

T. Stehle and G. E. Schulz

1136

OG---

136'

NE '\

-. y'2.61 =OiS 160' 2'84 ,148' \ 'OPSN-b5-SR-C5R ,' ,

,"2.76 0" 168'

,' ,

/

,' ,‘i.

11

171'

\

C,R----C2R I

\

‘\

150'

\,\2.60

N

,' 120' Oh2

'\

I

L/"lRAl.

OF5S \ \,2.58

--00

2.gg---o1o---j-.&,

Asp100

/ I

22 149’

Ni2

Figure ll.- Schema of all p?lar interactions between GMP and GK. The numbers represent the donor-acceptor distances in A, and the donor-H acceptor angles in degrees. The nucleotide atoms are named as in ApsA, but the letters A and B referring to the adenosine sites A and B of Ap,A are omitted, i.e. atoms Nl and OlR correspond to Ap,A NlB and OlRB.

atoms

below). They are probably larger at the ribose and the phosphate group. Presumably, the chain segments with the greatest mobilities (R-factor peaks at positions 42 and 74, see Fig. 6) will move over the longest distances as observed in the adenylate kinases (Schulz et aE., 1999). C” atom displacements in the range of 5 to 10 A should be expected. Such movements would bury the solvent-exposed Phe73. Although it has been convenient to define a GMPbinding domain, there exist also interactions between GMP and residues of MAIN. The purine ring of GMP is, for instance, held by non-polar interactions with Tle99 of MAIN and Tyr78 of GMPbd. Asp100 of MAIN contributes hydrogen bonds. Most conspicuously, the neighboring Asp98 is a crucial residue in catalysis. Tt contributes to the binding of the magnesium ion in the adenylate kinases (Egner et al., 1987). In order to derive general features of guanine binding, we have compared the hydrogen-bonding pattern in GK with those reported for H-ras-p21,

Table 4 Saltbridges

in guanylate N.

Atom 1

Atom 2

kinase

0 distance (4

Accessible surface area& (AZ)

Lysl4.NZ Lys20-NZ Arg38-NH1 Arg38-NH2

Asp98-OD2 Asp172-ODl Glu44-OE2 Glu44-OEl

285 302 2.84 2.86

73 143

AsplBS-ODl

Lys173-NZ

315

176

“Sum over both whole residues as determined DSSP (Kabsch & Sander, 1983).

ribonuclease T, (which represents a group of 5 homologous enzymes (Sevcik et al., 1990; Nonaka et al., 1991), and glutaminyl-tRNA-synthetase. As shown in Figure 13, the binding patterns in GK, Hras-p21 and ribonuclease T, show some similarities, whereas the pattern of glutaminyl-tRNA-synthetase is quite different. For the first three proteins there is a carboxyl group forming two hydrogen bonds to Nl and Nll of guanine. No such carboxyl group is found in glutaminyl-tRNA-synthetase. Tn GK, the observed hydrogen-bonding pattern including Ser80-OG (Fig. 11) indicates that the respective carboxyl group of Glu69 is protonated. Protonation of Glu69 is also favored by the adjacent carboxyl group of Asp100 binding to Nll, and by the low pH of 5.5 in the crystals. No second carboxyl group is found for H-ras-p21 or ribonuclease T,. Hydrogen bonding to 010 is much less uniform. The binding by two serine side-chains observed in GK is geometrically less stringent than the binding to main-chain atoms in H-ras-p21 and ribonuclease T,. Moreover, there is no similarity between the binding partners of N7 of guanine. In GK, the carboxyl group of Glu69 clearly distinguishes between guanine and adenine. Together with the carboxyl of AsplOO, however, it also renders the guanine binding site rather polar. This polarity difference between adenine and guanine binding sites may have caused the drastic structural difference between the GMP-binding domain of GK and the corresponding AMP-binding domains of the adenylate kinases.

61

with

program

(h) The bound sulfate ion The sulfate ion is bound in the giant anion hole formed by the glycine-rich loop of residues 8 to 15

Structure of Guanylate Kim-s-e : GMP

1137

Figure 12. Binding mode of GMP in GK. (a) Stereo view of the GMP binding pocket. Depicted are GMP and the sidechains of all residues forming hydrogen bonds (broken lines) to the nucleotide. The C” chain fold of residues 33 to 81 (essentially domain GMPbd) is given as a thin line. Asp100 is shown its a full residue. With respect to Fig. 2(a), the molecule has been rotated by 90” around a vertical axis in the plane of the paper. (b) Stereo view of the full model, centered at the Nl atom of GMP. Depicted are residues 33 to 35, 69 to 73, 78 to 80, 97 to 100 and 102 to 103 as well as 4 water molecules (x) In (a) and (b) the chain cuts are indicated by dots.

with the sequence Gly-Pro-Ser-Gly-Thr-Gly-LysSer. Its interactions with the enzyme are illustrated in Figure 14 and detailed in Table 5. This hole is present also in the adenylate kinases; in the crystal structures of AK I (Dreusicke et al., 1988) and AK3 (Diederichs & Schulz, 1991), it accommodates a sulfate ion as well. The superpositions of GK with AK1 (Fig. 9) and AK3 in Figure 10 show that the loops and the sulfate ions are at corresponding positions. A closer inspection reveals, however, that the orientation of the sulfate ion of GK is about 30” different from that of AKI, which is tightly bound to the enzyme. The orientation of the sulfate ion bound to AK3 (Fig. lo), is essentially the same as in AK1 but it is not well established, as it has a B-factor as high as 80 8’.

In the refinement of the crystal structure of GK, the sulfate ion finished with an average B-factor as high as 57 AZ, which, apart from the C terminus, exceeds any part of the main-chain (Fig. 6). Tts density is ellipsoidal, suggesting that the sulfate ion is bound in more than one position and/or orientation. The resulting placement is just the best solution of the refinement program XPLOR. The model shows strong interactions between 01 and the backbone (Table 5). It should be noted that all amide nitrogen atoms of residues 11 to 15 form hydrogen bonds to the sulfate ion. This situation had also been observed in AK1 and had given rise to the name “giant anion hole” (Dreusicke & Schulz, 1986). A partial compensation of the two negative charges of the sulfate ion is provided by Arg135,

T. Ntehle and G. E. Lkhulz

113X GK

POS.

"-a

p21

RNase

AsplIP-OD1 wat2s2

Asn98-0

Glu69-OEl AsplOO-OD2

Aspllg-OD2

Glu46-OE2

ser34-OG SerSO-OG

Ala146-N

D

wat192

Asnlld-ND2

E

AsplOO-N

Glu69-OE2

A B C

Tl

Table 5

GlnRS

Sulfate

Gln399-0

Glu46-OEl

Gln399-0

bin&y

in guanylate I).

lbfi

A”

Ab

Sulfate atom

GKatom

(h

01

Glyll-x Thrl2 N Glyl3-N Lysll-N Glyll-N Arg135-NH1

303 SIB 3.09 306 3.18 356’

I40 I 33 138 161 141 157

Serl5-OQ SerlS-N

2.77 2.89

144 164

(d+%.)

Arg402-NE

A-31144-N Tyr45-N

Arg402-NH1

Asn43-N

Arg402-NE

Arg402-NH1 Ribosyl-05'

02 03 04

“Donor acceptor distance. bDonor-fi acceptor angle. ‘Does not qualify as a hydrogen important interaction (see the text).

0A Figure 13. Interactions between guanine and protein in GK, H-ras p21 (Pai et aE., 1999), ribonuclease T, (RNase T,; Arni et al., 1988) and glutaminyl-tRNA-synthetase (ClnRS; Rould et aE., 1991). Ribonuclease T, represents a group of 5 homologous enzymes (Sevcik et al., 1999; Nonaka et al., 1991). which forms a weak contact. The density (see above) allows for an alternative sulfate position that interacts more strongly with Arg135. When inspecting the conformation of the conserved Lysl4 of the glycine-rich loop, we find that it does not contact the sulfate ion but forms a saltbridge to the essential Asp98 in the cleft as described in Table 4. A similar situation is found in

A

kinuse

but

is possibly

an

AK3, where the corresponding Lysl8 and Asp88 form a saltbridge as well. In AKl, however, the Eamino group of the equivalent Lys21 forms hydrogen bonds to the sulfate ion and to the carbonyl groups of the first two residues of the loop. Taken together, the two structures with the bound nucleoside monophosphates (GK : GMP and AK3 : AMP) have a loosely bound sulfate ion with no contact to the conserved lysine residue, while the structure without substrate (AKl) has a tightly bound sulfate contacting this lysine residue. It is conceivable that this difference reflects movements during catalysis. (i) The ATP-binding

site

Soaking experiments using ATP, ADP and AMP in order to locate the ATP-binding site of GK have been done under various conditions. None of them was successful. In several cases the crystals cracked. In hindsight, one should not have expected to find

l&NH2

Figure 14. Stereo view of the interactions Pro9-SerlO-Glyl1-Thr12-Gly13-Lys14-Serl5,

bond

between the sulfate and the guanidinium

A

135-N112

ion and GK, showing group of Arg135.

the sulfate

ion, residues

Gly8-

Structure of Guanylate Kinase : GMP

1139

Figure 15. Stereo view of the hypothetical model of ATP (thick lines) as bound to GK (thin lines). The binding site is derived from superimposing GK with AKeco : Ap,A (Miiller & Schulz, 1992). It also corresponds to the GTP sites of H-ras-p21 and EF-Tu (see the text). Also shown are the side-chains of the conserved residues Arg38, Arg41, Asp98 Arg131, Arg135, Arg146 and Asnl68. The view corresponds to that of Fig. 2(b).

(bl

Figure 16. Chain fold superpositions of GK with G-proteins. The equivalences comprise 4 strands of the central parallel p-sheet @1$7,/?8,~9 of GK) plus the glycine-rich loop forming a giant anion hole and the following helix al. The connection topologies of this part of the sheet are identical. (a) Best superposition of GK (thin line) with the complex EF-Tu : GDP (thick line) of Jurnak (1985) as baaed on the following set of 40 equivalenced c” atoms: 2 to 21, 94 to 98, 116 to 124, 164 to 169 of GK; and 12 to 31, 76 to 80, 106 to 108, 131 to 136 of EF-Tu. For clarity, only the EF-Tu residues used for the superposition are shown together with GDP and the magnesium ion (e). (b) Best superposition of GK (thin line) with the complex H-ru+p21 : GDP (thick line) of Tong et al. (1991) as based on the following set of 45 equivalenced C” atoms: 1 to 23,93 to 98, 115 to 123, 163 to 169 of GK; and 3 to 25, 52 to 56,58, 76 to 84, 111 to 117 of H-ras-~21. For H-ras-~21, only the residues used for the superposition are shown (except residue 58) together with GDP and the magnesium ion (m).

this site by soaking, because it is partly ~ovrrt~tl ~JJ cargstal conta,ct l--IV (see Table 3). In spite of lacking direct evidence, the ATT’ site of’ GK can k~ederived indirectly from the adenylate kinases. which are clearly homologous in domain MATN. A superposition of the complex GK : G&IT’ OII AK3 : AMP is given in Figure 10. Tn two different structures of admylatr kinases cornplexed with Al),=2 (t(:gner it al.. 1987: Miiller B Schulz. 1988) the inhibitor ApsA binds in t’he same (*onformation and position t,o the enzyme. Site adenosincbI< of the symmet’rical molecule Ap5A is the AMP site (1)iederichs 8 Srhulz, 1991) and c*orresponds to t hcs (:MP site of GK. This leaves site adenosinr-A as the ATP site. as illustrated in Figure 13. The assignment Of adenosinr-A to XTP is further confirmed 1~) t,he observation of the magnexiurn ion at its expected position between the p and y-phosphate groups in AKyst : Ap,A (Egner et al.. 1987). Additional evidence for thrb assignment of the ATT’ site caomes from the structural homology between GK and the adenylatr kinases on the one hand and the (:TP-binding proteins on the oth(lr. As shown in Figure 16. the bound GI)P in t’hr two (;proteins H-ms pdl and IiS-Tu corresponds to the ATT’ site in the adenylate kinases. The assumption of structural homology between these two groups of fjroteins is based on the chain fold similarity shown in I$urr 16. and vvw more strongly on the c~onsensus sequence at the giant anion hole. and on t’he similarity of nuclrotide binding to t’his hole. In racah caase. thr giant anion hole accommodates the /C phosphate grou~r of (:l)T’ and of Ap,A (when c*ounted from adenosinr-A). An interesting cGncidencae between the deducthd ATT’-binding site of GK and the GTP-binding sites should not go without mentioning. In all bhrec cases. t,herr is an asparagine at the end of the fourth equivalenced /?-strand (Fig.16) that binds to the ?r’7 atom of the purine as well as to the backbone at the giant, anion hole. The asparaginr residues of H-ras-p21 and EF-Tu belong to the fingerprint of G-proteins (Dever et sequence -Asn-Lys-X-Aspal., 1987). No asparagine is found at the equivalent positions of adenylate kinases. Tn GK, the equivalent Asn168-ND2 is bound to Thrl2-0 and is poised to bind to N7 of ATP. This arrangement resembles H-ras-~21, where Asnl16-ND2 binds to N7 and the equivalent, Vall4-0. The detailed binding situation of EF-Tu will be known when the structure is refined at high resolution.

c~otrclutlt~that the>r~lucidatc~d strucat urt’ of (:K cliffers from t hc structure allowi~lg IJhosphoryl t rarlsfbr. I’rolJa~Jl~. QY olJservr2 an intrrmr~diat~~ confi)rmatioll that is assumed aftrr binding the suljstratv (:Ml’. On ljinding ATT’ as thr second substrate. thv (:M!II’ Ijinding domain is (~spec*tc~tlto move‘ toward domain XIATN closing the cleft (St&t’Fig. 2(b)). bc~ause suc*h a mot,ion was oljsrrvrd in t hth adenylatcb kirjasvs on binding ATl’ in a,ddition to an already I~ountl AMI’ (Schulz rl crl., 1!1!,o). The similarities of thus catalytics mcxchanism of (;K and adrnvlate kinases are XlS0 dot~umc~ntrd IJF il nurnlxv, o’f. consrrvc~d residuchs. All of’ th(‘sp killas(>s caontain t hc glycinca-rich loof) t.hat ac.c,omrrJoclateh the ~-~Jh~Js~~l~at~~ groufj (of the triphosfjhatcl) in a giant anion Iiolc. I’r~~sumalJly. this hole (2*n Imffer electronic- c~trangrs IJ~ tlistriljutinp them ovf’r thfb main-chain. The thrvr glycincb residues of this loop in (:K have main-chain dihedral angles forbidden for non-glycine rrsidues. !Stailr this 1001)is Asp98 of (:K which is IJrvsvnt in all aclt~nylatc kinasc>s al~tl in the two (:-\~rotrins. Xsl~98 holds thcb requirr~d mapnesium ioil vicl water molt~rles. IJut it is likely to fulfill tilrt hf.1, important func*tions. l~f~aiisc~ it (7~11 form illI intermediate saltbridge with thca c~onsc~rvetl Lysl4 in Ihr. enzynits c~omplrxcs with nric~lf~osid~~ mono1jhoslJhiltc.s ((:K : (:MT’ and AK3 : AMP. set’ abovrl). As with the adenylatr kinases. thr catalytic. cleft residues that of (AK contains a number of arginine are likely to fix the phosphate groups during phosphoryf transfer. Among them Argl31. Arpl35 and Arg146 are conserved in all adenylatr kinases. Arg38 and Arg41 hold the phosphate group of (:Ml’ (Fig. 12(a)). Since they belong t.o domain (:MT’btl. no sequencr-equivalent residues in the adenylat~~ kinases can lye assigned (Fig:. 4). However. there are two functionally equivalent argininr residues. Ary44 and Arg97 of AK I, t’hat are c.onserved through all adenylatr kinases. They are known to hold the phosphate group of AMT’ in AK3 : A&IT-’ (l>iederi& dz Schulz. 1991). This work was supported by the Deutsche Forschungsgemeinschaft and by the Graduiertenkolleg Polymerwissenschaften. The co-ordinates of EF-Tu and H -ran-p2 1 have been obtained from the Brookhaven Protein Data Bank. The refined co-ordinates of GK and the structure factors are deposited with the Brookhaven Protein Dat,a Bank under the accession number 1GKY.

References ,4rni,

(j) Catalysis

The bound ATP deduced from structural homology (Fig. 15) superimposes the observed bound sulfate ion of GK with its /?-phosphate group. The yphosphorus atom of the modelled ATP is still at a distance of 6 A from the nearest oxygen atom of the a-phosphate group of the bound GMP: this is far too great for a nucleophilic attack. Moreover, the yphosphate group is so much exposed to the solvent that hydrolysis cannot be avoided. We therefore

R., Heinemann. U.. Tokuoka, R. & Saenger. W. (1988). Three-dimensional structure Of the ribonuclease T, : 2’-GMP complex at I.9 A resolution. .I. Bid. Chem. 263, 15358-15368. Baker, E. N. & Hubbard, R. E. (1984). Hydrogen bonding in globular proteins. Progr. Riophys. Mol. Biol. 44. 97-179. Berger. A., Schiltz, E. & Schulz, G. E. (1989). Guanylate kinase from Saccharomyces cerevisiae: Isolation and characterization, crystallization and preliminary Xray analysis, amino acid sequence and comparison 184. with adenylate kinases. Eur. J. Biochem.

433-443.

Structurr sf Cuanylate Kinase : GM I' Klevitts, R’. A. h Tulinsky. A. (19%). Comparison of the independent solvent, structures of dimeric a-chymotrypsin wit.h themselves and with y-ehymotrypsin. J. Rio/. (%mt. 260. 886.5~8872. Wriinger. A. T.. Kuripan. .I. & Karplus. M. (1987). (‘rystallopraphi~ R’-factor refinement by molerular dynatnic,s. Scirncu. 235. 458-460. I)rvrr. T. 15.. ($lynias. M. ,I. 8r Merrirk, LV. (‘. (1987). (iTI’-binding domain: Three consensus sequence rlrtnrnts with distinct spacing. /‘ror. ,Z’n/. .-icad. Sri.. l’.A’..3. 84. 1X14-ISIX. I~rVos. .\. $1.. Tong I,.. Milhurn. 31. Y.. Matias, J’.. .Ianc.arik. .J.. Sopucahi. S.. Eishimura. S.. Miura. K.. Ohtsuka. E. S: Kim. S.-H. (1988). Three-dimensional strttct ttre of an oncopenr protein. Catalytic domain of tttt tnan C.-H-TUSp:! I Scimcr. 239. W-893. I)reustc~kt~. I). & S(~huJz. (;. E. (1986). The gl,vrinr-rirh IooJ) of’ adrttylate kitrasr forms a giant anion hole. FEISS /AtYr*. 208. 301~ 304. Dreusic~k~~. I).. Karplu~. I’. X. & Schulz. G. E:. (1988). Krfittcd struc*turr of por(Gne cytosolic, adenylatr kittasr at 2.1 .I resolut’iott. .J. Mol. Bid. 199. 3359-371. I~iederic~hs. K. & St.hutz. (:. E. (1991). Thr refined &rueture of thr c~m~tlrx between adenylatr kinase from brrf heart nritochondrial matrix and its substrate AMJ’at I.% .I rrsolutiott. .J. Mol. Bid. 217. X-FM. Egnrr. C-.. Totttasselli. A. (:. 8 Schulz. (;. E. (1987). Structure of the complrx of yeast adenylate kinase with the inhittitor I”.rs-di(adenosine-,5’-) pentaphos]‘h;ttcL at 2-l; .i resoh,&on. J. MO/. Hiot. 195. 649-6X Janin. .I.. \Vodak. S.. Levitt. M. & Maigret. K. (1978). C”onformat,ion of amino acid side-chains in proteins. .J. .Ilo/. Kid. 125. 357-~386. ,Jonrs. ‘I’. A. ( I!PiX). A graphic’s tnodel building and refinemacromolecules. ?J. Appl. ,s~v*tf’lll to1 tnt‘tlt (‘ry.stalloqr. 1 1. PtiX- :“i? ,Jurnak. F. (19%). Structure of the GDP domain of El