The three-dimensional structure of shikimate kinase

J. Mol. Biol. (1998) 278, 983±997 The Three-dimensional Structure of Shikimate Kinase Tino Krell2, John R. Coggins2 and Adrian J. Lapthorn1* 1 Department of Chemistry and Division of Biochemistry and Molecular Biology, Institute of Biological and Life Sciences University of Glasgow Glasgow G12 8QQ, Scotland UK 2 The three-dimensional structure of shikimate kinase from Erwinia chrysanthemi has been determined by multiple isomorphous replacement. Ê model and a 2.6 A Ê Two models are presented: a high resolution 1.9 A model which contains bound Mg-ADP. The enzyme is an a/b protein consisting of a central sheet of ®ve parallel b-strands ¯anked by a-helices with overall topology similar to adenylate kinase. Evidence is presented that shikimate kinase undergoes major conformational changes on ligand binding. It resembles adenylate kinase in having a P-loop containing core structure and two ¯exible domains which undergo induced ®t movement on substrate binding. The binding of Mg2‡ in the active site of shikimate kinase involves direct interaction with two protein side-chains which is different from the situation found in adenylate kinase. Shikimate kinase has a readily identi®able Walker A-motif and a recognisable but modi®ed Walker B-motif. Comparison of shikimate kinase to adenylate kinase has led to the identi®cation of an adenine-binding motif (I/VDAXQ/NXP). Difference Fourier calculations have revealed the shikimate binding site which corresponds to the location of the AMP-binding site in adenylate kinase. A model for shikimate-binding is presented. # 1998 Academic Press Limited *Corresponding author Keywords: shikimate kinase; X-ray analysis; phosphoryl transfer; shikimate pathway; drug design Introduction The shikimate pathway is the seven-step biosynthetic route which generates chorismic acid, the major branch point in the synthesis of aromatic compounds, from phosphoenolpyruvate and erythrose 4-phosphate. The pathway is essential for plants and micro-organisms, but is absent from animals making the enzymes important targets for the development of potentially non-toxic antimicrobial agents (Davies et al., 1994) and herbicides (Coggins, 1989). Knowledge of the three-dimensional structures of the enzymes will undoubtedly aid the design of useful inhibitors. Abbreviations used: Ap5A, P1,P5-bis(50 -adenosyl)pentaphosphate; AK, adenylate kinase; B-factor, crystallographic temperature factor; DTT, dithiothreitol; GK, guanylate kinase; MIRAS, multiple isomorphous replacement with anomalous scattering; NCS, noncrystallographic symmetry; NMP, nucleoside monophosphate; P-loop, phosphate-binding loop; rms, root mean square; SIRAS, single isomorphous replacement with anomalous scattering; SK, shikimate kinase; UK, uridylate kinase; 3D, three-dimensional; s, standard deviation; CD, circular dichroism; gN-ATP, adenylyl-imidodiphosphate. 0022±2836/98/200983±15 $25.00/0/mb981755 Shikimate kinase (SK; EC 2.7.1.71), the ®fth enzyme of the pathway, catalyses the speci®c phosphorylation of the 3-hydroxyl group of shikimic acid using ATP as a co-substrate. In Escherichia coli, this reaction is catalysed by two different isoforms, a type I enzyme (SKI) and a type II enzyme (SKII), which share 30% sequence identity (Whipp & Pittard, 1995). The occurrence of two isoforms is unusual for an enzyme in the middle of a biosynthetic pathway and it has been suggested that shikimate may be at a branch point of two different pathways (Weiss & Edwards, 1980). The major difference between the isoenzymes is their Km value for shikimate, 200 mM for the SKII and 20 mM for the SKI enzyme (De Feyter & Pittard, 1986). It seems likely that SKII plays the major role in the biosynthesis of chorismate, while the role for SKI is not clear (De Feyter et al., 1986). Since mutations in SKI are associated with sensitivity to the antibiotic mecillinam (Vinella et al., 1996) it has recently been suggested that SKI may have an alternative biological function and that it phosphorylates shikimate only fortuitously (De Feyter & Pittard, 1986). The presence of a phosphate-binding loop (P-loop) is expected in both enzymes, since each has a Walker A-motif (GXXXXGKT/S; # 1998 Academic Press Limited 984 The Structure of Shikimate Kinase Walker et al., 1982). P-loop containing proteins form an extremely diverse class which includes proteins such as p21-Ras (Pai et al., 1990), F1ATPase (Abraham's et al., 1994) and adenylate kinase (Dreusicke et al., 1988). It was observed that all P-loop containing proteins have a central core domain consisting of four parallel strands with two pairs of ¯anking helices (Milner-White et al., 1991). The only exceptions to this generalisation are F1-ATPase and RecA (Story & Steitz, 1992). The SK from Erwinia chrysanthemi has been overexpressed in E. coli to up to 40% of total cellular protein using the T7 expression system (Studier & Moffat, 1986) (unpublished data). The puri®ed enzyme has been shown to be a monomer in the presence of DTT (Krell et al., 1997). Electrospray mass spectrometry revealed the presence of two species in the enzyme preparation, one (60% of total SK) with Mr ˆ 18,955 con®rming the mass derived from the DNA sequence (Minton et al., 1989), and the other (40%) with Mr ˆ 18,824 corresponding to enzyme with the N-terminal methionine cleaved off (Krell et al., 1997). This SK is a type II enzyme (Minton et al., 1989), sharing 53% amino acid sequence identity with E. coli SKII. Despite a very low degree of sequence similarity (19% identity) E. chrysanthemi SKII has been predicted to be structurally similar to porcine AK (Matsuo & Nishikawa, 1994). Unlike AK, where mechanistic and structural studies are complicated by the similarity of the two nucleotide substrates, SK has two structurally very different substrates. This coupled with the fact that it is among the smallest kinases so far reported make it a good model-enzyme to study the catalysis of phosphoryl transfer. Results and Discussion As all the crystals of SK were grown in the presence of 5 mM shikimate, 5 mM ADP, 10 mM MgCl2, it was expected that the enzyme would crystallise as a dead-end ternary complex: SKADP/Mg2‡-shikimate. This was not the case. The structure was deduced from two separate batches of crystals: one grown at pH 7.1 which gave a Ê native dataset and the other grown at pH 6.9 2.6 A Ê native dataset (Table 1). From which gave a 1.9 A these two native datasets two models were obtained, each containing two enzyme molecules in the asymmetric unit (A and B) which are related Ê by a non-crystallographic 2-fold axis. In the 2.6 A model 1, molecule A contains only shikimate Ê whilst molecule B contains only ADP. In the 1.9 A model 2, shikimate is bound to molecule A whilst molecule B is unliganded (Table 2). The presence of ADP in only one of the two crystals and the differential ligand-binding is hard to account for as in both cases the active sites are open to the sol- Table 1. Data collection, heavy atom derivatisation and MIR phasing Data collection SRS Daresbury beam line Ê) Resolution (A Total number of reflections Completeness (%) Average multiplicity Rmerge (%)a Riso (%)b Wilson B Ê) MIR phasing (5.8± 2.6 A Soak (mM/min) No. of sites Occupancy (%) Site 1 Site 2 Site 3 Temperature factor Site 1 Site 2 Site 3 Phasing powerc Rcullisd Fractional co-ordinates of heavy atoms (x, y, z) Site 1 Site 2 Site 3 a Native 1 Native 2 Hg(CN)2 (NH4)2PtCl4 9.5 28± 2.6 114,899 98.8 5.8 10.7 9.6 24±1.9 338,195 99.4 6.4 6.8 52.3 48.7 9.5 24±3.0 40,959 73.8 4.6 10.8 25.2 56.8 7.2 25±2.75 136,492 99.7 8.5 9.1 28.0 58.0 1/1 2 4.5/5 3 63.2 50.6 100 76.4 54.2 40.5 40.5 41.7 40.3 6.8 0.7 0.9 1.0 0.8 Hg(CN)2 (NH4)2PtCl4 0.739, 0.097, 0.687 0.839, 0.275, 0.404 0.932, 0.738, 0.107 0.229, 0.571, 0.101 0.775, 0.3110, 0.120 Rmerge ˆ h(j|Ij,h ÿ hIhi|/Ij,h), where h ˆ set of Miller indices and j ˆ set of observations of re¯ection h. Riso ˆ h||FDer|ÿ|FNat||/h|FNat|, where |FDer| ˆ observed derivative structure factor amplitude and |FNat| ˆ observed native structure factor amplitude. p c Root mean square (rms) fh/residual ˆ (f 2h/(FDer ÿ FPH)2), where hh ˆ calculated heavy atom structure factor, FPH ˆ calculated derivative structure factor. d RCullis ˆ ||fh|ÿ(|FDer|ÿ|FNat|)/|||FDer|ÿ|FNat||. b 985 The Structure of Shikimate Kinase Table 2. Final model statistics Data Ê) Resolution (A Rwork (%) Rfree (%) No. of unique reflections Model No. of No. of No. of No. of No. of No. of enzyme molecules disulphide bonds amino acids protein atomsf heteroatoms water molecules Stereochemistry Overall G-factore Ramachandran quality, % in Most favoured regions Allowed regions rms deviation Ê) Bond length (A Bond angle (degrees) Thermal parameter Mean B-factor Main-chain Side-chain Solvent Model 1 Model 2 20±2.6 18.6 26.3 19,785 20±1.9 17.4 22.1 52,843 2 1 320a 2495 31d 154 ÿ0.17 93.8 6.2 0.014 2.6 49.87 52.09 43.51 2 1 317b 2485 3c 475 0.07 94.4 5.6 0.013 1.7 47.33 50.25 60.93 a The total length of SK is 173 amino acids, molecule A contains residues 1 to 112 and 128 to 173; molecule B contains residues 1 to 112 and 123 to 172. b Molecule A contains residues 1 to 112 and 128 to 173; molecule B contains residues 1 to 112 and 126 to 172. c 3 Mg2‡. d 4 Mg2‡ and one molecule of ADP. e Calculated using the program PROCHECK (Laskowski et al., 1994). f Calculated including atoms with an occupancy of 0.5, there are 13 amino acids side-chains with a dual conformations in model 1 and 15 in model 2. vent. The possibility that this was an artifact of the re®nement protocol was ruled out as the molecules were carefully averaged and omit maps calculated at a number of stages during re®nement. Another unexpected observation is that molecules A and B are covalently linked in the crystal by a disulphide bridge between the Cys162 from each molecule. SK functions as a monomer (Millar et al., 1986) and it was shown by dynamic light scattering to be a monomer immediately before crystallisation (Krell et al., 1997). To investigate this disulphide formation, an SK solution without DTT was left at 20 C for several weeks. The electrospray mass spectrum of the resulting enzyme showed two species, one corresponding to the monomeric mass, the other to two times the monomeric mass minus 2 Da, clearly indicating the formation of a disulphide bond. Since SK has been crystallised in the absence of DTT, we presume that the disulphide bond formed during crystal growth because of the proximity of the free sulphydryl groups in the crystal. In addition to the active site magnesium ion (see below) three more magnesium ions were found in the structure of SK. However, their position on either the enzyme surface or at the interface between molecules A and B (Table 3) does not suggest any functional or structural role. Their presence may simply be a consequence of the high concentration of MgCl2 (10 mM) used for crystallisation. The enzyme fold SK is an a/b protein consisting of a central ®ve stranded parallel b-sheet with the strand order 23145, ¯anked on either side by a-helices (a1 and a8 on one side, a4, a5 and a7 on the other; Figure 1). The walker A-motif (Walker et al., 1982) is located between the ®rst b-strand (b1) and the ®rst a-helix (a1) forming a characteristic phosphate-binding loop (Saraste et al., 1990). The core of the SK structure forms a classical mononucleotide-binding fold (reviewed by Schulz, 1992) found in a number of structurally diverse proteins such as myosin (Smith & Rayment, 1995), elongation factor EF-Tu (Berchthold et al., 1993), p21-Ras (Pai et al., 1990) and AK (Dreusicke et al., 1988). To compare the structural similarity of SK with other P-loop containing proteins, representative structures from the Brookhaven databank (Bernstein et al., 1977) were superimposed onto SK. The number of Ca atoms used in the alignment and the rms difference in their positions are listed in Table 4. The nucleoside monophosphate (NMP) kinases, particularly yeast AK, appear to have more Ca atoms in similar positions to SK than other nucleotide-binding proteins as exempli®ed by p21-Ras (Pai et al., 1990) and Gia1 (Coleman et al., 1994). A core region of 44 Ca atoms, corresponding to SK strands b1, b3, b4, b5, in the main are in similar positions in all of the aligned structures. In addition to this common core region, SK shares structural similarities with the NMP kinases and others with p21-Ras and Gia1. The Ca atoms of b2 and a6 in the SK structure (Figure 1) align to equivalent Ca atoms in the NMP kinases, but this similarity does not extend to p21-Ras and Gia1. The loop involving Ca atoms 77 to 80 in SK is very similar to that found in p21-Ras and Gia1, which is part of the Walker B-motif (see below) in these proteins. The precise ordering of the strands 23145 in the parallel b-sheet classi®es SK as belonging to the same structural family as the NMP kinases for which structures are known for AK (Dreusicke et al., 1988; Schlauderer & Schulz, 1996), guanylate kinase (Stehle & Schulz, 1990), uridylate kinase (MuÈller-Dieckmann & Schulz, 1994) and thymidine kinase (Wild et al., 1995). This family is not restricted to enzymes catalysing phosphate transfer to NMPs but also includes enzymes transferring phosphate to hydroxyl groups like the kinase domain of rat testis 6-phosphofructo2-kinase/fructose-2,6-bisphosphatase (Hasemann et al., 1996) and now SK. A characteristic feature of the NMP kinases is that they undergo large conformational changes 986 The Structure of Shikimate Kinase Table 3. Crystal contacts between shikimate kinase molecules A and B Molecular packing interaction Buried surface Ê 2) area (A Residues involveda Source atom Polar interactionsb Ê) Target atom Distance (A Molecules A± A1 (2-fold crystallographic symmetry) 559 29± 32, 35, 36, 38± 43, 73 PheA30-N PheA30-O SerA41-O GlyA42-O GlnA38-NE2 ArgA73-NH2 3.14 3.31 3.44 Molecules B ±B1 (2-fold crystallographic symmetry) 519 47, 50±55, 58 134, 135, 137±139 ValB49-O AlaB50-O AlaB50-O ArgB139-NH2 ArgB139-NE ArgB139-NH2 2.92 2.68 3.00 Molecules A± B (NCS 2-fold symmetry) 400 20± 21, 24± 25, 156, 158, 162, 165, 172 GluA210OE1 GluA21-OE2 GluA21-OE2 CysA162 ArgB24-NE ArgB24-NE ArgB24-NH2 CysB162 3.06 (2.79) 3.41 (3.50) 2.94 (3.01) 2.01c Molecules A± B2 (NCS 2-fold symmetry) 870 1±3, 85, 89, 92± 96 140 ±142, 144, 145 147 ±149, 168 ±169 GluA3-N GluA3-OE2 ThrA96-OG1 AlaA93-O AlaA141-O AspA145-OD2 ArgA92-O GlyA95-O AlaA93.O GlnB89-OE1 ArgB92-NE HisB148-NE2 HisB94-NE2 ArgB169-NH2 ArgB169-NH2 Mg2‡903 Mg2‡903 Mg2‡903 3.29 2.82 2.94 2.90 3.15 3.05 2.15 2.43 2.58 (3.15) (2.78) (2.75) (3.18) (2.98) (2.51) (2.22) (2.47) Molecules A and B form the asymmetric unit. The transformation from A to B is [ÿ0.461, ÿ0.884, 0.077, ÿ0.887, 0.456, ÿ0.066, 0.023, ÿ0.099, ÿ0.994] as rotation matrix and [0.735, 0.466, 0.309] as the translation vector. Molecules A1, B1 and B2 are related to the respective reference complex at [X, Y, Z] by [X, Y, ÿZ], [X, Y, 1 ÿ Z] and [1/2 ÿ X, 1/2 ‡ Y, ÿ1/4 ‡ Z], respectively. For molecules related by crystallographic symmetry only one of the interactions is shown, for molecules related by NCS the distances in brackets are for the corresponding interaction. a Ê between any atom of shown residues to atoms of neighbouring residues was used as a criterion. A distance below 4.5 A b Ê and D-H


A angles above 90 . Hydrogen bonds are de®ned by donor


acceptor distances below 3.6 A c Disulphide bond. during catalysis. There are two ¯exible regions of the structures that are responsible for movement: one is the NMP-binding site which is formed by a series of helices between strands 1 and 2 of the parallel b-sheet and the other is the so-called lid domain, a region of varied size and structure following the fourth b-strand of the sheet (MuÈller et al., 1996; Gerstein et al., 1993). We have shown that SK undergoes conformational changes on substrate and co-factor binding. The CD spectra of unliganded SK and enzyme in the presence of either 2 mM shikimate or 2 mM of the ATP analogue (gN-ATP (Figure 2) are signi®cantly different, indicating ligand-induced changes in the enzyme secondary structure. It is dif®cult to correlate the CD data precisely with the structural data. Analysis of the CD spectrum of the unliganded enzyme using the CONTIN procedure (Provencher & GloÈckner, 1981) over the range of 195 to 240 nm gave 29(1)% a-helix and 27(2)% b-sheet. This differs from the secondary structure content as calculated from the 3D-structure (50% a-helix and 13% b-sheet). This discrepancy would be signi®cantly reduced if, like adenylate kinase (Schlauderer & Schulz, 1996), the disordered lid domain of SK contains some b-sheet. Using the CONTIN procedure the changes in the CD spectrum on binding shikimate and gN-ATP correspond approximately to a 10% decrease in a-helix content and a 10% increase in b-sheet content. The changes observed in the 3D-structures are concerted movements of secondary structure elements but do not involve conversion of a-helix into b- sheet. Thus the changes in the CD spectra are indicative of conformational change but do not give a precise measure of secondary structure content. The ADP and shikimate-binding sites are indicated schematically in Figure 1B. The electron density for ADP was suf®cient in native dataset 1 for the inclusion of the molecule in the structure (Figure 1A), but the electron density for shikimate was too poor to accurately locate this substrate. There is only weak incomplete electron density for residues 113 to 122 (model 1), which indicates that the polypeptide chain is disordered (Figure 1), and prevents the chain being traced in this region. These disordered residues form part of the ¯exible lid domain which, by analogy to the NMP kinases, stretches from residues 112 to 126. The g-phosphate group of ATP is presumably necessary for complete lid closure in SK, as has been shown for the NMP kinases (Vonrhein et al., 1995). The lid domains of NMP kinases are quite variable in size, e.g. AKs occur in two forms: the small cytosolic variants have about 195 residues and a lid domain consisting of 11 residues, whereas the larger mitochondrial variants have about 225 residues with lid domains comprising 38 residues (Gerstein et al., 1993). The lid domain in SK contains 15 residues, which makes SK more similar to the cytosolic variants of AK. The structural similarity of SK to AK was predicted by Matsuo & Nishikawa (1994) using a novel sequence-structure compatibility method. Considering the low sequence similarity of both The Structure of Shikimate Kinase 987 Figure 1. The structure of shikimate kinase. A, Stereo view of shikimate kinase complexed with ADP (model 1). b-Strands are shown as arrows and the N and C termini are labelled as N and C, respectively. There is a break in the polypeptide chain between residues 112 to 123 of the lid domain (112 to 126) for which there was no clear electron density. B, Topology diagram of shikimate kinase. a-Helices are represented as cylinders and labelled a1 to a8, there is one 310 helix labelled 310. b-Strands (labelled as b1 to b5) are shown as arrows. The approximate position of bound ADP and shikimate are indicated. The broken line represents the missing residues 113 to 122 of the lid domain (112 to 126). C, Sequence of Erwinia chrysanthemi shikimate kinase. Secondary structure elements are underlined and labelled as in B. The shaded region indicates the position of the A-motif (Walker et al., 1982) and the broken line shows the missing residues of the lid domain. Part A of this Figure and subsequent Figures were prepared using MOLSCRIPT (Kraulis, 1991) 988 The Structure of Shikimate Kinase Table 4. Structural alignment of various P-loop containing proteins with shikimate kinase Protein Adenylate kinase (Abele & Schulz, 1995±2aky) Uridylate kinase (MuÈller-Dickmann & Schulz, 1995) Guanylate kinase (Stehle & Schulz, 1992) Gia1 (Coleman et al., 1994) p21-Ras (Pai et al., 1990) Similarities in the Ca positions Ê) Number of atoms rms deviation (A 97 92 85 63 62 1.85 1.76 2.21 1.82 1.75 Protein structures were obtained from the Protein Data Bank (Bernstein et al., 1977) and then superimposed separately with shikimate kinase (model 2) using the lsq-commands of the program O (Jones et al., 1991). An initial superimposition was achieved by aligning the P-loop (using lsq_explicit). This alignment was extended over the entire polypeptide chain using the command lsq_improve. The number of Ca atoms matched and their rms deviations are listed. enzymes (only 19% identity), the accuracy of the prediction is impressive. We have compared the secondary structure prediction of Matsuo & Nishikawa (1994) and a prediction using the program PHD (Rost & Sander, 1993) with the crystal structure (Table 5). PHD correctly predicted the secondary structure of 79% of the residues in the structure (compared with an average secondary prediction accuracy of 72% over a wide range of proteins), whilst the prediction by Matsuo & Nishikawa was better, with an accuracy of 82%. PHD did not predict all the secondary structure elements of SK correctly (strand b2 was predicted as helix and strand b5 as a loop) while Matsuo & Nishikawa correctly predicted all the secondary structure elements, with errors only in the exact boundaries of these elements. A structure-based sequence alignment between SK and yeast AK (Abele & Schulz, 1995; Figure 3) shows extensive Figure 2. Superimposed circular dichroism spectra of unliganded and liganded shikimate kinase from Erwinia chrysanthemi. ( ÐÐ ) Unliganded enzyme; (- - - - - -) enzyme in the presence of 2 mM adenylyl imidodiphosphate (gN-ATP); (





) enzyme in the presence of 2 mM shikimate. Spectra were recorded in a Jasco J-600 spectropolarimeter using cylindrical quartz cells of path length 0.02 cm. Protein solutions of 0.5 mg/ml were in 10 mM Tris-HCl (pH 7.5). For the spectra of the liganded enzyme an aliquot of a 20 mM ligand solution in 10 mM Tris-HCl (pH 7.5) (pH readjusted using conc. KOH) was added to the protein solution and the resulting spectra were corrected for dilution. The CD spectra were also corrected for the contribution of the ligands; this correction was less than 5%. similarities between both enzymes in the arrangement of secondary structure elements. Ligand-binding ADP/Mg 2‡-binding SK and many other nucleotide-binding enzymes contain a short conserved stretch of sequence GXXXXGKT/S (the Walker A-motif, Walker et al., 1982). This motif forms the P-loop, a giant anion hole which accommodates the b-phosphate of the ADP by donating hydrogen bonds from several backbone amides (reviewed by Smith & Rayment, 1996). The binding of ADP by SK (Figure 1A) is analogous to that seen in other proteins containing the Walker A-motif, the protein-nucleotide interactions are summarised in Table 6. The missing residues of the lid domain (Figure 1), such as the conserved Arg120, are likely to form more interactions with the nucleotide. In close association with the P-loop there is a binding site for the Mg cation essential for enzyme activity. This cation-binding in many Ploop proteins such as myosin (Smith & Rayment, 1995), elongation factor EF-Tu (Berchthold et al., 1993), p21-Ras (Pai et al., 1990) and the heterotrimeric G-proteins (Coleman et al., 1994) involves hexa-coordination of the Mg2‡ by two oxygen atoms (from the b and g-phosphates of the bound nucleotide), two water molecules and two protein ligands. One protein ligand is invariantly the hydroxyl side-chain of the Thr/Ser at position 8 of the A-motif, the second is often a Thr from another region of the protein, but without a clear consensus pattern within P-loop proteins. In NMP kinases, the Mg2‡ binding is different, since position 8 of the A-motif is occupied by a Gly instead of Thr or Ser. It has been dif®cult to elucidate the protein residues which co-ordinate the Mg2‡ as the cation was not found in a number of crystal structures of AK (Dreusicke et al., 1988; Diederichs & Schulz, 1990; MuÈller & Schulz, 1992; Berry et al., 1994). Recently, a high resolution structure of yeast AK complexed with the ``two in one'' substrate adenosine-(phosphate)5-adenosine (Ap5A; Abele & Schulz, 1995) has revealed the Mg2‡ position and shown that the protein does not interact directly with the cation, but instead it is co-ordinated by 989 The Structure of Shikimate Kinase Table 5. Comparison of the accuracy of the secondary structure predictions for shikimate kinase from Erwinia chrysanthemi from the sequence-structure compatibility method (Matsuo & Nishikawa, 1994) and PHD (Rost & Sander, 1993) No. of amino acids for which secondary structure was predicted correctly; (% accuracy) Shikimate kinase (model 1) ±No. of amino acids in secondary structure elements In In In In b-sheet: a-helix: loops: total: Sequence-structure compatibility method 22 92 49 163 the oxygen atoms of the b and g-phosphates of Ap5A and three water molecules. Unlike the NMP kinases, SKs have a Thr at position 8 of the Walker A-motif. The binding pattern of the active site Mg2‡ in SK is shown in Figure 4. The four ligands seen in model 1 are the oxygen of the b-phosphate of ADP, a water and two direct interactions with the protein involving residues Thr16 and Asp32, which are conserved in all SKs. It should be noted that the distance from the Mg2‡ to its ligands is greater than seen in most P-loop proteins. The missing ligands that make up the hexa-coordination of the Mg2‡ are likely to be either missing residues from the ¯exible lid domain or water molecules. Hence, although SK is structurally very similar to the NMP kinases its mode of 21 79 33 133 (95.5%) (85.5%) (67.4%) (81.6) PHD 14 81 34 129 (66.6%) (88.0%) (69.3%) (79.1%) magnesium-binding appears to be rather different and more closely resembles that of the other Ploop proteins. In addition to the Walker A-motif, the majority of purine-nucleotide binding proteins contain a second conserved sequence, called the B-motif (Walker et al., 1982). This motif, Z-Z-Asp-X-X-Gly (where Z is hydrophobic and X is any residue), is usually located on the C-terminal segment of the second strand (b3) of the central b-sheet (Smith & Rayment, 1996; Figure 3). The conformation of this motif is very similar in p21-Ras (residues 55 to 60; Pai et al., 1990), EF-TU (residues 79 to 84; Berchthold et al., 1993) and the G-proteins (residues 198 to 203; Coleman et al., 1994). The motif forms a loop around the g-phosphate of the nucleotide, Figure 3. A structural alignment of shikimate kinase with yeast adenylate kinase (Abele & Schulz, 1995, 2aky) comparing the secondary structure elements. The Walker A and B-motifs are shaded. Conserved residues present in the adenine binding pocket are boxed. The missing residues in the lid domain of shikimate kinase are bracketed. 990 The Structure of Shikimate Kinase Table 6. The binding of ADP ADP O1B O2B O3B O2A O3A N6 Contacting atom Distance/angle Ê /degree)a (A Lys15-N Gly14-N Thr16-N Thr16-OG1 Wat14 Gly12-N Thr17-N Gly14-N Gln155-O Wat27 3.17/126.6 3.08/122.4 3.21/156.4 3.27/105.4 3.43 3.28/175.9 3.11/158.1 3.21/126.9 2.53/150.4 2.92 a Hydrogen bonds are de®ned by donor


acceptor distances Ê and D-H


A angles above 90 . below 3.6 A with the amide nitrogen of the conserved glycine residue hydrogen-bonding to one of the g-phosphate oxygen atoms. In p21-Ras, the conserved Asp (Asp57) has a dual function: it forms a hydrogen bond with one of the two water molecules liganded to the active site Mg2‡ and a hydrogen bond with the Ser following the P-loop lysine (Pai et al., 1990). The B-motif present in the NMP kinases does not have a conserved glycine residue. The loop following the Asp in NMP kinases has a completely different conformation from that in proteins with a full Walker B-motif. The loop turns away from the nucleotide and so prevents the formation of hydrogen bonds between the g-phosphate and the peptide amide nitrogen atoms. The binding of the gphosphate in AK is accomplished instead by arginine residues (Abele & Schulz, 1995). The conserved Asp in this motif is involved in binding two of the three water molecules co-ordinating the active site Mg2‡ (Abele & Schulz, 1995). SK also has a modi®ed B-motif, in this case retaining the Gly (Gly79) but not the Asp which is replaced by Ala76. The two functions performed by the B-motif Asp57 in p21-Ras (hydrogen bonds to a Mg2‡-bound water and to the hydroxyl group of residue 8 of the A-motif) are accomplished by two conserved aspartate residues in SK (Asp32 and Asp34) located on the C-terminal segment of the right-hand strand in the b-sheet (b2 in Figure 1). Asp32 forms a hydrogen bond to Thr16 (residue 8 of the A-motif), whilst Asp34 hydrogen bonds to a water co-ordinated to the active site Mg2‡ (Figure 4). The conserved Gly79 of SK (Figure 1 and 6) is in an almost identical position to the conserved Gly found in proteins with the full B-motif and its amide nitrogen is correctly positioned to hydrogen-bond to the g-phosphate of a bound ATP. No sequence motif has been reported for adenine binding; instead, Moodie et al. (1996) have suggested a ``fuzzy-recognition template''. While comparing SK to the other NMP kinases we were surprised to ®nd that the binding sites for adenine in SK, yeast AK (Abele & Schulz, 1995) and bovine AK, isoenzyme II (Schlauderer & Schulz, 1996), were very similar. Figure 5A shows the adeninebinding pockets of SK and yeast AK (Abele & Schulz, 1995, 2aky) after superimposing the adenine moieties of ADP (SK) and Ap5A (AK). In both cases the adenine is sandwiched between an Arg and a Pro. The Pro forms part of a loop, connecting the ®fth b-strand (b5) with the C-terminal helix (Figures 1 and 3), which wraps around the bound adenine. The Arg110 (Figure 3) is part of a segment just upstream of the lid domain and its side-chain is parallel to the adenine ring. Adenine can, in principle, establish ®ve hydrogen bonds; two from the donor N6 and one each from acceptors N1, N3 and N7. However, in both structures only the two N6 hydrogen bonds are formed, to the backbonecarbonyl atom of a Gln and to a water molecule. The latter is co-ordinated to backbone-carbonyl atoms of an Ala and to the P-loop residue located two positions upstream of the conserved lysine of the Walker A-motif (Figure 5A). The N6 of adenine, which binds the protein, appears to be responsible for the selectivity of both enzymes for adenine nucleotides over guanine nucleotides. In both structures the P-loop is further linked with the adenine-binding pocket by a hydrogen bond between the NH1 atom of the Arg110 (Figure 5A) and the backbone carbonyl atom of the residue three positions upstream of the A-motif lysine. It appears that this similarity in adenine-binding is not coincidental as there is a conserved sequence motif (Figure 5B) common to the bacterial SKIIs, AK type II isoenzymes (which are located in the intermembrane space of mitochondria) and to AK type I isoenzymes from yeast. This adenine-binding motif has the consensus sequence Val/Ile-AspAla-X-Gln/Asn-X-Pro (X is any residue). Consider- Figure 4. Stereo view of the binding site of the active site Mg2‡ in shikimate kinase (model 1) liganded with ADP. Residues involved in binding are labelled Ê ). and bond distances are shown (A 991 The Structure of Shikimate Kinase Figure 5. The adenosine binding motif. A, Stereo view of the adenosine binding pocket of shikimate kinase (model 1) and adenylate kinase from S. cerevisiae (Adbele & Schulz, 1995; 2aky) after superimposition of adenine moieties of bound ADP (shikimate kinase) and Ap5A (adenylate kinase). Shown are amino acids and a water molecule which are in the vicinity of the adenine; the ®rst number after the residue name refers to shikimate kinase, the second number to adenylate kinase; black line, shikimate kinase; grey line, adenylate kinase; (- - - - - -) hydrogen bonds. B, Sequence alignment of shikimate kinases (isoenzymes II) and adenylate kinases (isoenzymes II and yeast isoenzyme I) in the region of the adenine binding pocket (loop between the ®fth b-strand and the C-terminal helix (see Figure 1). Conserved residues are shaded. The numbering above the alignment corresponds to the E. chrysanthemi shikimate kinase, the numbering below to S. cerevisiae adenylate kinase. ing that only backbone interactions are responsible for adenine recognition it is remarkable to ®nd such a conserved sequence motif. Previous attempts to ®nd sequence motifs associated with the adenine-nucleotide binding have been unsuccessful. This may simply re¯ect the enormous functional variety of the proteins surveyed, which have included, for example, NAD(P)-binding proteins, AMP-binding proteins and FAD-binding proteins (Moodie et al., 1996). Our motif is restricted to a relatively narrow group of ATP-binding proteins. Shikimate-binding A peak of more than 5s in the ®nal Fo ÿ Fc difference Fourier map indicates the position of bound shikimate in the electron density of molecule A in both models of SK. However, the electron density was not clear enough to include shikimate in the molecular structure. The ambiguous density for shikimate is probably due to a combination of relatively weak substrate-binding (Km [shikimate] ˆ 300 mM) and the high ionic strength (2.16 M NaCl) used for crystallisation. From the location of the difference density the shikimate appears to bind in a position analogous to the nucleotide monophosphate in NMP kinases. The shikimate-binding domain, which follows strand b2, consists of the helices a2 and a3 and the N-terminal region of helix a4 (see Figure 1). This corresponds to the AMP-binding site of AK (Abele & Schulz, 1995). Several conserved, charged amino acids and three glycine residues (Figure 6) are grouped around the peak in the difference map, implicating them in shikimate-binding. If we consider the structure of shikimic acid, it is a six-membered ring with a carboxylate group at position 1 and hydroxyl groups at positions 3, 4 and 5. It seems likely that conserved residues Arg58 and Arg139 (Figure 6) are involved in binding the carboxylate group. In addition the backbone NH-groups of Gly78 and Gly80 may also contribute to carboxylate-binding in a similar manner to that proposed for Gly395 and Ala397 in carboxylate-binding via a water molecule in 3-phosphoglycerate kinase (Harlos et al., 1992). The two other conserved charged residues are in positions to co-ordinate the hydroxyl groups of shikimate: Asp34, which is indirectly involved in co-ordinating the active site Mg2‡ (Figure 4), is suitably positioned to bind to the hydroxyl groups at C3 and/or C4, while Glu61 is suitably positioned to bind the C5 hydroxyl group (Figure 6). Induced fit movements Kinases need to protect their active sites from the omnipresent water to avoid ATP hydrolysis (Jencks, 1975). This is achieved by induced-®t 992 The Structure of Shikimate Kinase Figure 6. The shikimate binding. Shikimate kinase was co-crystallised with ADP and shikimate. The electron density for ADP allowed its positioning in the molecular structure as shown (Figure 1). There was substantial electron density for shikimate, but an unambiguous positioning of the molecule was not possible. The charged residues grouped around the density for shikimate, which are most likely involved in its binding are shown. All labelled residues are conserved in a sequence alignment of both isoenzymes of shikimate kinase and the shikimate kinase domains of the AROM pentafunctional enzyme complexes. For clarity only residues 5 to 86 and 128 to 144 of the protein are shown. (Koshland, 1958) movements of enzymes as observed, for example, in hexokinase (Bennet & Steitz, 1980) and AK (Schulz et al., 1990). In AK such movements have been described for both the domains involved in substrate-binding. The NMPbinding domain of AK was shown to undergo a rigid-body rotation of 39 , but, more spectacularly, Ê and undergoes a the lid domain moves 30 A 90 hinge bending rotation (Schulz et al., 1990; Gerstein et al., 1993) on ATP-binding. Spectroscopic studies have established that SK undergoes major conformational changes on binding either shikimate or ATP analogues (Figure 2). The single tryptophan residue (54) is positioned close to the shikimate-binding site and serves as a reporter group to allow the effect of ligand-binding to be directly monitored. In the absence of ligands, the ¯uorescence emission maximum of 346 nm and the Stern-Volmer constant of 4.8 Mÿ1 (Idziak et al., 1997) indicate that the Trp side-chain is highly exposed to solvent (Eftink & Ghiron, 1976). Addition of shikimate causes a decrease in ¯uorescence (with a 3 nm blue shift in the emission maximum) and a remarkable decrease in the SternVolmer constant to 1.8 Mÿ1. These changes are consistent with the loop containing this Trp sidechain becoming more deeply buried within the Figure 7. B-factors for side-chain atoms (A) and mainchain atoms (B) (averaged per residue) of shikimate kinase, model 2; (ÐÐ ) molecule A, with shikimate bound; (- - - - - -) molecule B, no shikimate bound. protein following ligand-binding (Idziak et al., 1997). In the crystal structure of SK the averaged B-factors per residue show clear evidence of the ¯exibility of the molecule (Figure 7). The temperature factors for both molecules show two regions of high mobility; this is most prominent in molecule B. The ®rst region corresponds to the shikimatebinding site (residues 32 to 58) and the other corresponds to the lid domain and its ¯anking regions (residues 100 to 140). Similar B-factor pro®les have been reported for the NMP-binding domain and the lid domain of AKs (Abele & Schulz, 1995; MuÈller et al., 1996). The models presented contain two enzyme molecules in the asymmetric unit: one complexed with shikimate and free of bound ADP (molecule A) and the other either uncomplexed or complexed solely with ADP (molecule B). This situation is not surprising for an enzyme which, like AK (Noda, 1973), can bind either substrate independently. The region 32 to 50 (which forms part of the shikimate-binding domain) has elevated B-factors in the shikimate-free molecule B in comparison to molecule A, where shikimate is bound. In addition, crystal contacts between symmetryrelated A molecules are mainly formed by residues 29 to 43 (Table 3), while this region in molecule B is exposed to the solvent and not involved in crystal contacts. The differences in the temperature factors therefore cannot be simply attributed to shikimate-binding, as crystal contacts undoubtedly play a role. The residues on both sides of the lid domain show elevated B-factors, indicating the ¯exibility of this domain. Comparing the B-factor pro®les of molecules A and B, there is no observable change in chain ¯exibility in any region removed from the binding sites, which could act as an energetic counterweight on substrate-binding as proposed for AK by MuÈller et al. (1996). A comparison of the crystal structures of the unliganded enzyme and the enzyme liganded to the two-in-one substrate adenosine-(phosphate)4-shikimate (Ap4S) might allow such changes to be observed. So far we The Structure of Shikimate Kinase 993 Figure 8. Flexibility of shikimate kinase. A, Comparison of molecule A (with bound shikimate) and molecule B (no bound shikimate) from model 2. Molecules were superimposed using the program LSQKAB from the CCP4 program suite (CCP4, 1994). The regions of structure which were essentially identical are represented in a schematic way, with the exception of residues 101 to 128, which have been omitted for clarity. The main structural differences are located in two domains (32 to 60 and 134 to 142) where main-chain atoms are shown of molecule A ( ÐÐ ) and molecule B (- - - - - -). The conserved residues ARg139, Phe57 and Val45, which may contribute to the orientation of bound shikimate, are labelled. B, The difference in Ca positions between molecule A and B; ( ÐÐ) model 2; (- - - - - -) model 1. have failed to obtain suitable crystals for such an analysis. In model 1, where ADP is bound in molecule B (Table 2), there is clearer density for the lid domain, which allowed the inclusion of residues 123 to 126 for this molecule but not for molecule A. However, molecule B still shows no electron density for the rest of the lid domain (Figure 1). Judging from the weak density obtained for the gap in molecule B, the lid in its closed state forms a short loop folded over the bound nucleotide. The presence of ADP in molecule B appears to have somewhat improved the order of the lid domain, whereas it appears to be fully disordered in the ADP-free molecule. The domain closure in molecule B is incomplete with ADP bound, suggesting that the g-phosphate of ATP plays a crucial role in the completion of the domain movement. A comparison of the differences in Ca positions after superimposing molecule A (with shikimate) 994 on to molecule B (without shikimate) (Figure 8B) shows that the major differences are con®ned to the shikimate-binding domain (32 to 60) and to a region following the lid domain (134 to 142). These two regions appear to be involved in shikimate recognition (Figure 6). In molecule A residues 32 to 60, which form the shikimate-binding domain, are Ê movement towards involved in a concerted 1.5 A the observed density for shikimate (Figure 8A). Furthermore, crystal contacts between pairs of symmetry-related A molecules, involving residues 29 to 43, stabilise the closed conformation of the shikimate-binding domain (Table 3) in contrast to molecule B. These induced-®t movements are relatively modest in comparison to those found in AKs (Vonrhein et al., 1995). We have proposed that Arg139 is involved in recognition of the shikimate-carboxylate group (Figure 6). There seems to be a signi®cant difference in the conformation of Arg139 between molecules A and B. In the shikimate-free molecule B, the arginine forms several hydrogen bonds as part of the crystal contacts to a symmetry-related B molecule (Table 3). As a result, Arg139 and neighbouring residues have been displaced towards the symmetry-related B molecule. This movement of Ê prethe guanidium group of Arg139 by nearly 2 A vents any possibility of direct hydrogen-bonding to shikimate. It may be that this involvement of Arg139 in crystal contacts results in the dissociation of bound shikimate from the active site during crystallisation and thus the shikimate-binding domain (residues 32 to 60) is free to move back into an unliganded state. These crystal contacts may account for the differences in ligand binding between molecules A and B and are consistent with the ability of the enzyme to bind each ligand independently. Materials and Methods Crystallisation and heavy atom derivatisation SK crystals were prepared by sitting-drop vapour-diffusion using 12 ml drops containing a 50:50 (v/v) mix of protein (16 mg/ml in 20 mM Tris-HCl (pH 7.6), 5 mM shikimate, 5 mm ADP, 10 mM MgCl2) and of reservoir solution (either 2.16 M NaCl, 100 mM Hepes (pH 6.9) (native 2 and derivatives) or pH 7.1 (native 1; see Table 1). Crystals appeared within 10 to 12 days and continued to grow as tetragonal bipyramids to a maximum size of 0.7 mm  0.2 mm  0.2 mm, as reported previously (Krell et al., 1997). Heavy atom derivatives were prepared by soaking crystals in native mother liquor containing various concentrations of metal salts as listed in Table 1. Data collection and processing All data were collected as 1 deg. oscillation frames at stations 7.2, 9.5 and 9.6 at the Daresbury SRS (Table 1). Crystals were loop-mounted in a cryoprotectant containing 17.5% (v/v) glycerol and ¯ash-cooled to 100 K using an Oxford Cryosystems cryostream. Data were processed with DENZO and scaled with SCALEPACK The Structure of Shikimate Kinase (Otwinowski, 1993). The crystals were shown to belong to the primitive tetragonal crystal system, with unit-cell Ê and c ˆ 92.8 A Ê . Analysis dimensions of a ˆ b ˆ 108.5 A of the systematic absences in the data revealed absences at h ˆ 2 and l ˆ 4n along the (h00) and (00l) axes, respectively, which are consistent with the space group of P41212 or its enantiomorph. Two molecules per asymmetric unit are predicted from the packing density of Ê 3 Daÿ1, which corresponds to a solvent conVm ˆ 3.6 A tent of 66% (Matthews, 1968). MIR phasing, model building and refinement Ê native and a 3.0 A Ê mercury derivative dataset A 2.6 A were collected at station 9.5 at Daresbury SRS. The wavelength for the collection of the mercury derivative Ê to optimise the anomolous dataset was tuned to 0.83 A contribution at the mercury absorption edge. The soak time for the mercury derivative was extremely short as the crystals shattered after only three minutes incubation in 1 mM Hg(CN)2. The derivative and native data were scaled using SCALEIT from the CCP4 suite of programs (1994) and SHELX-90 (Sheldrick, 1991) was used to locate the heavy atom positions using the Patterson method (Table 1). The positional parameters (x, y, z), temperature factors and relative real and anomalous occupancies of the heavy atoms sites were re®ned in the space groups P41212 and P43212 using MLPHARE (Otwinowski, 1991). From the anomalous occupancy the space group was determined as P41212. Single isomorphous replacement with anomalous scattering (SIRAS) phases were used as starting phases for density modi®cation procedures to re®ne and extend the phases to Ê resolution using the program DM (Cowtan, 1994). 2.8 A The electron density fell short of being interpretable, but was of suf®cient quality to recognise regions of secondary structure and con®rm that there were two molecules in the asymmetric unit. It was not possible to de®ne a clear molecular boundary and there was no indication of 2-fold non-crystallographic symmetry (NCS) in the selfrotation function. Therefore, electron density corresponding roughly to one molecule was used as a search model for molecular replacement with the program AmoRe (Navaza, 1994). The solution for two molecules was used to obtain an initial matrix for performing 2-fold (NCS) averaging. A platinum derivative dataset was collected Ê resolution on station 7.2 at Daresbury SRS at to 2.8 A Ê wavelength. Using the mercury SIRAS phases, 1.49 A three sites were located in difference Fourier maps (Table 1). Heavy atom parameters were re®ned using MLPHARE and used to generate MIRAS phases, which were improved by solvent ¯attening, histogram matching and 2-fold NCS averaging using DM. The ®nal density-modi®ed electron density map, Ê was of excellent quality and easily interphased to 2.8 A pretable. One molecule was traced into the electron density using the program O (Jones et al., 1991); the Ca atoms were positioned using skeletonized electron density generated by the program BONES (Kleywegt & Jones, 1994). The high quality of the experimentally determined phases allowed the positioning of over 80% of the side-chains in their electron density. The initial model was subjected to rounds of simulated annealing, positional and temperature factor re®nement, with strict NCS using the program X-PLOR (BruÈnger et al., 1987). After rebuilding into the averaged electron density, where the majority of the missing side-chains were included, further re®nement of the structure with strict NCS did not improve the free R-factor as would have 995 The Structure of Shikimate Kinase Figure 9. The ®nal 2Fo ÿ Fc electron density map including all data Ê contoured at 2.4 from 30.0 to 1.9 A s above the mean of the entire map. The region shown corresponds to amino acid residues 149 to 169 of model 2. This Figure was prepared using the program Sector (Evans, 1993). been expected. After examination of 2Fo ÿ Fc difference Fourier maps calculated with phases derived from the averaged structure, it became apparent that there were signi®cant differences in conformation between the two molecules in the asymmetric unit. Positional and B-factor re®nement was continued using X-PLOR with NCS restraints over the majority of the protein molecules (speci®cally excluding residues 35 to 55) and using all Ê ) with a bulk solvent correction. data (28 to 2.6 A A further four rounds of rebuilding and re®nement resulted in a model with Rwork ˆ 25.5% and Ê) Rfree ˆ 29.5%. At this stage a higher resolution (1.9 A native dataset was collected on station 9.6 at Daresbury SRS, and the re®nement was continued using these data (native 2, Table 1). The maximum likelihood re®nement program REFMAC (Murshudov et al., 1996) was used to re®ne the structure using the bulk solvent correction (imported from X-PLOR) and an overall anisotropic temperature factor scaling. The program ARP (Lamzin & Wilson, 1993) was used to include and reject water molecules during the re®nement. After 12 cycles of re®nement and rebuilding the ®nal model contained two protein molecules, three magnesium ions and 475 water molecules. Although the enzyme was co-crystallised in the presence of 5 mM ADP (product) there was no density corresponding to ADP and the electron density for shikimate was ambiguous and so this ligand was not included in the model. A representative section of the ®nal 2Fo ÿ Fc map contoured at 2.4 s illustrates the quality of the electron density (Figure 9). Ê model was used to phase the initial The ®nal 1.9 A Ê resolution native dataset (Table 1). In the Fo ÿ Fc 2.6 A difference Fourier map there was clear density for one ADP molecule. The ADP was modelled into the structure and further rounds of re®nement with REFMAC and ARP resulted in a model which contains two protein molecules, four magnesium ions, one molecule of ADP and 154 water molecules. The model resulting from the Ê dataset is referred to as model 1 and the model 2.6 A Ê data as model 2. The ®nal statistics based on the 1.9 A for both models are shown in Table 2. Acknowledgements We thank Professor N. C. Price and Drs E. J. MilnerWhite and S. M. Kelly for helpful discussions and the recording of the CD spectra. Many thanks to members of the molecular enzymology and protein crystallography groups in Glasgow with special thanks to Dr Paul Emsley, Dr Steve Prince, Dr John Maclean and Mr Andrew Elwell. Financial support from the BBSRC is acknowledged. The re®ned co-ordinates of both models of SK and the structure factors are deposited with the Brookhaven Protein Data Bank under the accession numbers 1shk (model 2) and 2shk (model 1). References Abele, U. & Schulz, G. E. (1995). High-resolution structure of adenylate kinase from yeast ligated with inhibitor Ap5A, showing the pathway of phosphoryl transfer. Protein Sci. 4, 1262±1271. Abrahams, J. P., Leslie, A. G. W., Lutter, R. & Walker, Ê resolution of F1J. E. (1994). Structure at 2.8 A ATPase from bovine heart mitochondria. Nature, 370, 621± 628. Bennet, W. S. & Steitz, T. A. (1980). Structure of a complex between yeast hexokinase A and glucose. II. Detailed comparisons of conformation and active site con®guration with the native hexokinase B monomer and dimer. J. Mol. Biol. 140, 211± 230. Berchthold, H., Reshetnikovca, L., Reiser, C. O. A., Schrimer, N. K., Sprinzl, M. & Hilgenfeld, R. (1993). Crystal structure of active elongation factor Tu reveals major domain rearrangements. Nature, 365, 126± 132. Bernstein, F. C., Koetzle, T. F., Williams, G. J. B., Meyer, E. F., Jr, Brice, M. D., Rogers, J. R., Kennard, O., Shimanouchi, T. & Tasuma, M. (1977). The protein data bank: a computer-based archival ®le for macromolecular structures. J. Mol. Biol. 112, 535± 542. Berry, M. B., Meador, B., Bilderback, T., Liang, P., Glaser, M. & Phillips, G. N., Jr (1994). The closed conformation of a highly ¯exible protein: The structure of E. coli adenylate kinase with bound AMP an dAMPPNP. Proteins: Struct. Funct. Genet. 19, 183± 198. BruÈnger, A. T., Kuriyan, J. & Karplus, M. (1987). Crystallographic R-factor re®nement by molecular dynamics. Science, 235, 458± 460. CC4, (1994). The CCP4 suite: programs for protein crystallography. Acta Crystallog. sect. D, 50, 760± 763. Coggins, J. R. (1989). The shikimate pathway as a target for herbicides. In Herbicides and Plant Metabolism 996 (Dodge, A., ed.), pp. 97 ± 112, Cambridge University Press. Coleman, D. E., Berghuis, A. M., Lee, E., Linder, M. E., Gilman, A. G. & Sprang, S. R. (1994). Structures of active conformations of G1a1 and the mechanism of GTP hydrolysis. Science, 265, 1405± 1412. Cowtan, K. (1994). DM, An automated procedure for phase improvement by density modi®cation. Joint CCP4 SF-EACBM Newsletter Protein Crystallog. 31, 24 ± 28. Davies, G. M., Barrett-Bee, K. J., Jude, D. A., Lehan, M., Nichols, W. W., Pinder, P. E., Thain, J. L., Watkins, W. J. & Wilson, R. G. (1994). (6S)-6-Fluoroshikimic acid, an antimicrobial agent acting on the shikimate pathway. Antimicrob. Agents Chemother. 38, 403± 406. De Feyter, R. C. & Pittard, J. (1986). Puri®cation and properties of shikimate kinase II from Escherichia coli K-12. J. Bacteriol. 165, 331± 333. De Feyter, R. C., Davidson, B. E. & Pittard, J. (1986). Nucleotide sequences of the transcription unit containing the aroL and aroM genes from Escherichia coli K-12. J. Bacteriol. 165, 233± 239. Diederichs, K. & Schulz, G. E. (1990). Three-dimensional structure of the complex between mitochondrial matrix adenylate kinase and its substrate AMP. Biochemistry, 29, 8138± 8144. Dreusicke, D., Karplus, A. & Schulz, G. E. (1988). Re®ned structure of porcine cytosolic adenylate Ê resolution. J. Mol. Biol. 199, 359± 371. kinase at 2.1 A Eftink, M. R. & Ghiron, C. A. (1976). Exposure of tryptophanyl residues in proteins. Quantitative determination by ¯uorescence quenching studies. Biochemistry, 15, 672± 680. Evans, S. V. (1993). Sector: Hardware lighted threedimensional solid model representations of macromolecules. J. Mol. Graphics, 11, 134±138. Gerstein, M., Schulz, G. E. & Chothia, C. (1993). Domain closure in adenylate kinase: Joints on either side of two helices close like neighbouring ®ngers. J. Mol. Biol. 229, 494± 501. Harlos, K., Vas, M. & Blake, C. F. (1992). Crystal structure of the binary complex of pig muscle phosphoglycerate kinase and its substrate 3-phospho-Dglycerate. Proteins: Struct. Funct. Genet. 12, 133±144. Hasemann, C. A., Istvan, E. S., Uyeda, K. & Deisenhofer, J. (1996). The crystal structure of the bifunctional enzyme 6-phosphofructo-2-kinase/fructose-2,6-bisphosphatase reveals distinct domain homologies. Structure, 4, 1017± 1029. Idziak, C., Price, N. C., Kelly, S. M., Krell, T., Boam, D. J., Lapthorn, A. J. & Coggins, J. R. (1997). The interaction of shikimate kinase from Erwinia chrysanthemi with substrates. Biochem. Soc. Trans. 25, S627. Jencks, W. P. (1975). Binding energy, speci®city, and enzyme catalysis: the circe effect. Advan. Enzymol. 43, 219± 410. Jones, T. A., Zou, J.-Y., Cowan, S. W. & Kjelgaard, M. (1991). Improved methods for building protein models in electron density maps and the location of errors in these models. Acta Crystallog. sect. A, 47, 110± 119. Kleywegt, G. J. & Jones, T. A. (1994). Halloween. Masks and bones. In From First Map to Final Model Processing (Bailey, S., Hubbard, R. & Waller, D., eds), pp. 59 ± 66, SERC Daresbury Laboratory, Warrington, UK. The Structure of Shikimate Kinase Koshland, D. E., Jr (1958). Application of a theory of enzyme speci®city to protein synthesis. Proc. Natl Acad. Sci. USA, 44, 98 ± 104. Kraulis, P. J. (1991). MOLSCRIPT-A program to produce both detailed and schematic plots of protein structures. J. Appl. Crystallog, 24, 946± 950. Krell, T., Coyle, J. E., Horsburgh, M. J., Coggins, J. R. & Lapthorn, A. J. (1997). Crystallisation and preliminary X-ray crystallographic analysis of shikimate kinase from Erwinia chrysanthemi. Acta Crystallog. sect. D, 53, 612±614. Lamzin, V. S. & Wilson, K. S. (1993). Automated re®nement of proteins. Acta Crystallog. sect. D, 49, 129± 147. Laskowski, R. A., MacArthur, M. W. & Thornton, J. M. (1994). Evaluation of protein coordinate sets. In From First Map to Final Model (Bailey, S., Hubbard, R. & Waller, D. A., eds), pp. 149±159, SERC Daresbury Laboratory, Warrington, UK. Matthews, B. W. (1968). Solvent content of proteins. J. Mol. Biol. 33, 491± 497. Matsuo, Y. & Nishikawa, K. (1994). Protein structural similarities predicted by a sequence-structure compatibility method. Protein Sci. 3, 2055± 2063. Millar, G., Lewendon, A., Hunter, M. G. & Coggins, J. R. (1986). The cloning and expression of the aroL gene from Escherichia coli K-12. Biochem. J. 237, 427± 437. Milner-White, J. E., Coggins, J. R. & Anton, I. A. (1991). Evidence for an ancestral core structure in nucleotide-binding proteins with the type A-motif. J. Mol. Biol. 221, 751± 754. Minton, N. P., Whitehead, P. J., Atkinson, T. & Gilbert, H. J. (1989). Nucleotide sequence of an Erwinia chrysanthemi gene encoding shikimate kinase. Nucl. Acids Res. 17, 1769. Moodie, S. L., Mitchell, J. B. O. & Thornton, J. M. (1996). Protein recognition of adenylate: An example of a fuzzy recognition template. J. Mol. Biol. 263, 486± 500. MuÈller, C. W. & Schulz, G. E. (1992). Structure of the complex between adenylate kinase from E. coli and Ê resolution. the inhibitor Ap5A re®ned at 1.9 A J. Mol. Biol. 224, 159± 177. MuÈller, C. W., Schlauderer, G. J., Reinstein, J. & Schulz, G. E. (1996). Adenylate kinase motions during catalysis: an energetic counterweight balancing substrate binding. Structure, 4, 147± 156. MuÈller-Dieckmann, H.-J. & Schulz, G. E. (1994). The structure of uridylate kinase with its substrates, showing the transition state geometry. J. Mol. Biol. 236, 361± 367. Murshudov, G. N., Dodson, E. J. & Vagin, A. A. (1996). Application of maximum likelihood methods for macromolecular re®nement. In Macromolecular Re®nement (Dodson, E., Moore, M. & Bailey, S., eds), pp. 93 ± 104, SERC Daresbury Laboratory, Warrington, UK. Navaza, J. (1994). AmoRe ±An automated package for molecular replacement. Acta Crystallog. sect. A, 50, 157± 163. Noda, L. (1973). Adenylate kinase. In The Enzymes (Boyer, P. D., ed.) vol. 8, pp. 279± 305, Academic Press, New York. Otwinowski, Z. (1991). Maximum likelihood re®nement of heavy atom parameters. In Isomorphous Replacement and Anomalous Scattering (Wolf, W., Evans, P. R. & Leslie, A. G. W., eds), pp. 80 ± 85, SERC Daresbury Laboratory, Warrington, UK. The Structure of Shikimate Kinase Otwinowski, Z. (1993). Oscillation data reduction program. In Data Collection and Processing (Sawyer, L., Isaacs, N. & Bailey, S., eds), pp. 56 ± 62, Daresbury Laboratory, Warrington, UK. Pai, E. F., Krengel, U., Petski, G. A., Goody, R. S., Kabsch, W. & Wittinghofer, A. (1990). Re®ned crystal structure of the triphosphate conformation of Ê resolution: implications for the H-ras p21 at 1.35 A mechanism of GTP hydrolysis. EMBO J. 9, 2351± 2359. Provencher, S. W. & GloÈckner, J. (1981). Estimation of globular protein secondary structure from circular dichroism. Biochemistry, 20, 33 ± 37. Rost, B. & Sander, C. (1993). Prediction of protein secondary structure at better than 70% accuracy. J. Mol. Biol. 232, 584± 599. Saraste, M., Sibbald, P. R. & Wittinghofer, A. (1990). The P-loop. A common motif in ATP and GTP binding proteins. Trends Biochem. Sci. 15, 43 ± 434. Schlauderer, G. J. & Schulz, G. E. (1996). The structure of bovine mitochondrial adenylate kinase: Comparison with isoenzymes in other compartments. Protein Sci, 5, 434±441. Schulz, G. E. (1992). Binding of nucleotides by proteins. Curr. Opin. Struct. Biol. 2, 61 ± 67. Schulz, G. E., MuÈller, C. W. & Diederichs, K. (1990). Induced-®t movements in adenylate kinases. J. Mol. Biol. 213, 627± 630. Sheldrick, G. M. (1991). Heavy atom location using SHELX-90. In Isomorphous Replacement and Anomalous Scattering (Wold, W., Evans, P. R. & Leslie, A. G. W., eds), pp. 80 ± 86, SERC Daresbury Laboratory, Warrington, UK. Smith, C. A. & Reyment, I. (1995). X-ray structure of the magnesium(II)-pyrophosphate complex of the truncated head of Dictyostellium discoideum myosin to Ê resolution. Biochemistry, 34, 8973± 8981. 2.7 A 997 Smith, C. A. & Rayment, I. (1996). Active site comparisons highlight structural similarities between myosin and other P-loop proteins. Biophys. J. 70, 1590± 1602. Stehle, T. & Schulz, G. E. (1990). Three-dimensional structure of the complex of guanylate kinase from yeast with its substrate GMP. J. Mol. Biol. 211, 249± 254. Story, R. M. & Steitz, T. A. (1992). Structure of the recA protein-ADP complex. Nature, 355, 374± 376. Studier, F. W. & Moffatt, B. A. (1986). Use of bacteriophage T7 RNA polymerase to direct selective highlevel expression of cloned genes. J. Mol. Biol. 189, 113± 130. Vinella, D., Gagny, B., Joseleau-Petit, D., D'Ardi, R. & Cashel, M. (1996). Mecillinam resistance in Escherichia coli is conferred by loss of a second activity of the aroK protein. J. Bacteriol. 178, 3818± 3828. Vonrhein, C., Schlauderer, G. J. & Schulz, G. E. (1995). Movie of the structural changes during a catalytic cycle of nucleoside monophosphate kinases. Structure, 3, 483± 490. Walker, J. E., Saraste, M., Runswick, M. J. & Gay, N. J. (1982). Distantly related sequences in the a- and bsubunits of ATP synthase, myosin, kinases and other ATP-requiring enzymes and a common nucleotide binding fold. EMBO J. 1, 945± 951. Weiss, U. & Edwards, J. M. (1980). The Biosynthesis of Aromatic Compounds, John Wiley and Sons, New York. Whipp, M. J. & Pittard, A. J. (1995). A reassessment of the relationship between aroK- and aroL-encoded shikimate kinase enzymes of Escherichia coli. J. Bacterial. 177, 1627± 1629. Wild, K., Bohner, T., Aubry, A., Folkers, G. & Schulz, G. E. (1995). The three-dimensional structure of thymidine kinase from Herpes simplex virus type 1. FEBS Letters, 368, 289± 292. Edited by K. Nagai (Received 18 September 1997; received in revised form 17 February 1998; accepted 4 March 1998)