J. Mol. Biol. (1998) 276, 449±459

The 2.3 AÊ X-ray Crystal Structure of S. cerevisiaePhosphoglycerate Mutase

Daniel J. Rigden1,2*, Dmitriy Alexeev2, Simon E. V. Phillips1

and Linda A. Fothergill-Gilmore3

1Department of Biochemistryand Molecular BiologyUniversity of Leeds, LeedsLS2 9JT, UK2Department of BiochemistryUniversity of EdinburghSwann Building, KingsBuildings, EdinburghEH9 3JZ, Scotland3Department of BiochemistryUniversity of EdinburghHugh Robson BuildingGeorge Square, EdinburghEH8 9XD, Scotland

Abbreviations used: PGAM, phosmutase; 2,3-BPG, 2,3-bisphosphoglybisphosphoglycerate mutase; F2, 6Bbisphosphate phosphatase; RPAP, rphosphatase; NCS, non-crystallogra

0022±2836/98/070449±11 $25.00/0/mb9

The high resolution crystal structure of Saccharomyces cerevisiae phospho-glycerate mutase has been determined. This structure shows importantdifferences from the lower resolution structure deposited in 1982. Thecrystal used to determine the new structure was of a different form, hav-ing spacegroup P21. The model was re®ned to a crystallographic R-factorof 18.9% and a free R-factor of 28.4% using all data between 25 and2.3 AÊ and employing a bulk solvent correction. The enzyme is a tetramerof identical, 246 amino acid subunits, whose structure is revealed to be adimer of dimers, with four independent active sites located well awayfrom the subunit contacts. Each subunit contains two domains, the largerwith a typical nucleotide binding fold, although phosphoglyceratemutase has no physiological requirement to bind nucleotides. The cataly-tic-site histidine residues are no longer in a ``clapping-hands'' confor-mation, but more resemble the conformation seen in the distantly relatedenzymes prostatic acid phosphatase and fructose-2,6-bisphosphatase.However, the catalytic histidine residues in the mutase are found to bemuch closer to each other than in the phosphatase structures, perhapsdue to the absence of bound ligands in the mutase crystal. An intricateweb of H-bonds is found around the catalytic histidine residues, high-lighting residues probably important for maintaining their correct orien-tation and charge. The positions of certain other residues, including somefound near the catalytic site and some lining the catalytic-site cleft, havebeen changed by the correction of registration errors between sequenceand electron density in the original structure. Electron density was appar-ent for a portion of the functionally important C-terminal tail, which wasabsent from the earlier structure, showing it to adopt a mainly helicalconformation.

# 1998 Academic Press Limited

Keywords: phosphoglycerate mutase; crystal structure; phosphohistidine;glycolysis

*Corresponding author

Introduction

Phosphoglycerate mutases constitute a family ofenzymes which catalyse the transfer of phosphogroups among the three carbon atoms of phospho-glycerates (reviewed by Fothergill-Gilmore &Watson, 1989). A striking feature of the mechanismof these enzymes is the formation of a phosphohis-tidine intermediate. There are at least four types of

phoglyceratecerate; BPGAM,Pase, fructose 2,6-at prostatic acidphic symmetry.

71554

phosphoglycerate mutase (PGAM) which are kine-tically and structurally distinct, but which never-theless have many properties in common. Themutase in the glycolysis/gluconeogenesis path-ways (EC 5.4.2.1) plays an essential role by catalys-ing the interconversion of 2-phosphoglycerate and3-phosphoglycerate. One member of the PGAMfamily requires the cofactor 2,3-bisphosphoglyce-rate (2,3-BPG), and is found in vertebrates and inyeasts. A second type (found in Bacillus spp.) isindependent of bisphosphoglycerate, but requiresmanganese. A third type does not require anycofactors, and is found in organisms such as higherplants and invertebrates. The fourth member of thefamily is a closely related enzyme (EC 5.4.2.4/EC

# 1998 Academic Press Limited

450 Phosphoglycerate Mutase Crystal Structure

3.1.3.13) which catalyses the synthesis of 2,3-BPG.It plays a major role in controlling haemoglobinoxygen af®nity, and is of consequent pharmaco-logical interest. This enzyme is known as bis-phosphoglycerate mutase (BPGAM), and can beconsidered to be an isoenzyme of the vertebratePGAM because it shares high sequence identity(about 50%), and also ef®ciently catalyses the gly-colytic mutase reaction. The cofactor-dependentmutases have been studied in most detail, partlybecause of their relative ease of puri®cation. Theseenzymes are active as monomers, dimers or tetra-mers depending upon the organism from whichthey have been isolated. The subunit size is about250 amino acid residues, and numerous aminoacid sequences have been determined (reviewed byFothergill-Gilmore & Michels, 1993). There is noapparent sequence similarity between the cofactor-dependent and independent mutases. Quite unex-pectedly, it was discovered that PGAMs are hom-ologous (albeit with low percentage identity) to thephosphatase portion of the bifunctional enzymeresponsible for catalysing the synthesis and degra-dation of fructose 2,6-bisphosphate (Lively et al.,1988), and also to prostatic acid phosphatase(Schneider et al., 1993).

The tetrameric PGAM from Saccharomyces cerevi-siae was one of the ®rst enzymes for which a crys-tal structure was solved (Campbell et al., 1974),showing that the enzyme possesses the now fam-iliar feature of a central b-sheet surrounded by a-helices. The topology of the polypeptide backboneincludes a nucleotide-binding motif, despite thelack of physiological requirement to bind nucleo-tides. Ligand-soaking experiments revealed the 3-phosphoglycerate binding site (Winn et al., 1981),but frustratingly were not useful for studying thecatalytically competent phospho form of theenzyme because the crystals cracked when cofactorwas added. Two histidine rings (residues 8 and181) were prominent in a parallel ``clappinghands'' conformation at the active site. The guani-dinium group of Arg59 was in a suitable positionto form an electrostatic interaction with the car-boxyl group of the phosphoglycerate ligands. Thecoordinates of this early structure were depositedin the Brookhaven Data Bank (Watson, 1982), andhave been invaluable in developing an understand-ing of the properties of this enzyme. However,there are some important limitations. The structureis of only modest resolution (2.8 AÊ ), has been thesubject of only limited re®nement and relied on aprotein sequence determined by manual methodsthat was subsequently shown by DNA sequencingto suffer from mistakes in four regions (White &Fothergill-Gilmore, 1988). Moreover, an examin-ation of the phi and psi angles adopted by thepolypeptide backbone shows that only 49% of theresidues occur in ``most favoured regions'' of aRamachandran plot (Laskowski et al., 1993). Thevalue expected for a typical protein should be inexcess of 80%.

The catalytic cycle of PGAM proceeds via anenzyme-substitution pathway in which an active-site histidine is phosphorylated (reviewed byFothergill-Gilmore & Watson, 1989). In the case ofPGAM from S. cerevisiae, His8 is phosphorylated(Rose, 1971), and the catalytically competent phos-phoenzyme has been shown by mass analysis tohave a half-life of approximately 35 minutes (Nairnet al., 1995). It has been proposed that the secondhistidine at the catalytic site plays a role as a pro-ton donor/acceptor (Rose, 1980), and the import-ance of this residue has been con®rmed by site-directed mutagenesis (White & Fothergill-Gilmore,1992; White et al., 1993a). The C-terminal 14 resi-dues of PGAM have attracted interest becausetheir removal by limited proteolysis correlates witha substantial decrease in activity (Sasaki et al.,1966; Price et al., 1985). Unfortunately these resi-dues were not visible in the electron density mapof the early structure, presumably because of ¯exi-bility. However, early model building studies indi-cated that they could modulate access to the activesite (Fothergill-Gilmore & Watson, 1989).

Many aspects of the properties of PGAM, andthe related enzymes BPGAM and fructose 2,6-bisphosphate phosphatase (F2,6BPase), have beenstudied by site-directed mutagenesis. As expected,alterations to either of the two catalytic-site histi-dine residues caused dramatic loss of activity in allthree enzymes (White & Fothergill-Gilmore, 1992;Garel et al., 1993; Tauler et al., 1990). Changes toother residues at the catalytic sites of PGAM andBPGAM (Ser11, Glu15, Gly21, Ala60 and Arg87,PGAM numbering) affected substrate binding andspeci®city (Ravel et al., 1996; Walter, R. A.,personal communication). Experiments withF2,6BPase have shown that modi®cations to a glu-tamate residue at the catalytic site (equivalent toGlu86 in PGAM) brought about substantial loss ofactivity (80 to 98%), and it was suggested that thisresidue may act as a base, and/or in¯uence theprotonation state of the essential histidine residues(Lin et al., 1992). The C-terminal tails of PGAM andBPGAM have been subjected to study by aminoacid replacements and partial deletions (Walter &Fothergill-Gilmore, 1995; Garel et al., 1989). Itseems likely that these residues play a role in sub-strate binding and product release, but the effectsare subtle and remain poorly understood.

We report here the determination of a high-resol-ution crystal structure of PGAM from S. cerevisiae,that forms a sound basis for continuing proteinengineering studies, the design of inhibitormolecules as potential lead compounds for drugdevelopment, and the interpretation of NMRspectroscopy studies (UhrõÂnova et al., 1997).

Results and Discussion

Quality of the final model

The ®nal model is of good sterochemical qualityas shown in Table 1. The r.m.s. differences of

Table 1. Re®nement statistics

Number of non-hydrogen protein atoms 8326Number of non-hydrogen solvent atoms 827Resolution range 25-2.3 AÊ 2.38-2.3 AÊ

Number of reflections 43973 4432R (%) 19.3 24.5Rfree (%) 28.5 35.5Mean temperature factor B (AÊ 2)

All atoms 21.7Main-chain 20.8Side-chain 22.4Solvent 22.5

R.m.s. deviation from ideal valuesBond lengths (AÊ ) 0.011Bond angles (deg.) 2.1Improper dihedrals (deg.) 1.5

NCS r.m.s. deviations for non-hydrogenatoms (AÊ )

Restrained subunits 0.04Unrestrained subunits 0.73

Figure 2. B-factor plot of subunit A. Within each residuemain-chain atoms were assigned one group B factor andside-chain atoms another. The continuous line representsmain-chain B-factors and the dotted line side-chainB-factors.

Phosphoglycerate Mutase Crystal Structure 451

bonds and angles from small molecule-derivedideal values are 0.01 AÊ and 2.1�, respectively.Figure 1 shows a Ramachandran plot producedusing the program Procheck (Laskowski et al.,1993) with 89.4% of the residues in the mostfavoured regions of the plot. The remainder fallinto additional allowed areas, with the exceptionof Ala180 from subunits A and D for which thereis clear electron density. Procheck also identi®esno side-chains as having unusual w1/w2 confor-mations.

A plot of main-chain and side-chain B-factors ofsubunit A is shown in Figure 2. The most notablefeature is a peak around position 107 where themain-chain B factors exceed 40 for residues 104 to

Figure 1. Ramachandran plot of the ®nal structureproduced using PROCHECK (Laskowski et al., 1993).Glycine residues are represented by triangles and otherresidues by squares.

110. This region comprises the loop betweenhelices 4 and 5 together with the beginning of helix5, a region distant from the main body of the struc-ture (Figures 3 and 4). The very high B factors seenhere could be a consequence of the breakdown ofthe imposed non-crystallographic symmetry sinceanalysis of crystal contacts shows that residueLys102 of all subunits contacts Glu12 of neighbour-ing tetramers in the crystal lattice. However thedensity in this area is very weak, and when theseresidues were excluded from NCS restraints thefree R-factor failed to improve.

Description of overall structure

The PGAM subunit contains two domains (seeFigure 3). The larger contains a beta-pleated sheetcore surrounded by helices. The beta strands com-prising the sheet, D, C, G, A, I and J run parallelwith the exception of I. The ¯anking helices arenumbered 8, 9 and 10 on one side, 2 and 3 on theother. The similarity of this domain to a nucleotidebinding fold has been noted before (Campbell et al.,1974). Such folds are frequently found to containligand binding sites at the edges of the sheet(BraÈnden, 1980), particularly at the carboxy ends ofthe strands, and PGAM falls in to this category.The catalytic site, as de®ned by mutational studies(discussed below) and crystal soaking (Winn et al.,1981) centres on histidine residues 8 and 181, atthe carboxy ends of strands A and G, respectively.

The smaller domain had appeared in the 1982lower resolution structure (designated 3PGM) to bealmost devoid of secondary structure. The newstructure shows it to contain both beta strands andalpha helices, albeit relatively short ones. Particu-larly interesting is the appearance of helix 5,formed by residues 106 to 113. The last of these,

Figure 3. A ribbon diagram of subunit A, with alphahelices labelled from 1 to 12 and beta strands from A toK. The N terminus and the last residue at the C termi-nus for which electron density was seen are alsolabelled, as is the region 104 to 110 which has highB-factors. The three labelled Arg residues de®ne theapproximate extent of the active site. (Drawn withMOLSCRIPT (Kraulis, 1991), as are Figures 4, 7 and 9).

452 Phosphoglycerate Mutase Crystal Structure

Arg113, and its neighbour Arg114, surround theentrance to the catalytic site cleft.

The C-terminal tail of the protein has also beenimplicated in the reaction mechanism by limitedproteolysis experiments (Sasaki et al., 1966; Priceet al., 1985; D.J.R., Walter, R. A & L.A.F.-G., unpub-lished results) and mutational studies (Walter &Fothergill-Gilmore, 1995). In the 3PGM structurethe C-terminal region of the protein was notde®ned after Ala232 due to poor electron density.Inositol hexaphosphate, a potent PGAM inhibitorshown to interact with the tail (D.J.R., Walter, R. A.& L.A.F.-G., unpublished results), was included inthe mother liquor in an attempt to crystallise a con-formation of PGAM with a structured tail. Notrace of the inhibitor could be seen, however, inthe electron density map but in the new structureextra residues could be built at the C termini ofthe subunits. In the cases of subunits B and D,model building could be carried out up toGly236 but for A and C, only residues up toAla235 and Ala234, respectively could be built,forming an alpha helix from 230 to 234. As can

be seen in Figures 3 and 4, the C-terminal helixmakes no contacts to the rest of the protein con-sistent with its being mobile. The disappearanceof traceable density after Gly236 suggests thatthis residue is responsible for extra disorder in agenerally disordered region, and it would, there-fore, be an interesting target for mutagenesisexperiments.

The PGAM subunits associate to form a tetramerwith 222 symmetry. It was originally suggestedthat one subunit interface within the tetramer ismuch more extensive than the other (Campbellet al., 1974). This was partly due to the omission oftwo residues from the original structure, Asp169and Leu170. The current structure shows the inter-faces bury approximately the same amount of sol-vent-accessible area, 970 and 920 AÊ 2 for the A-Band A-C interfaces, respectively. A view of thePGAM tetramer emphasising regions involved insubunit association is shown in Figure 4. At the A-B interface, the majority of contacts involve helix 3and strand D. The side-chains of Glu135 andArg136 from helix 6 also form a large number ofcontacts. At the A-C interface the contacts aredivided between a number of different regions ofthe protein sequence. The larger contributors areArg83, residues 163-164 from the end of helix 8and the region 138 to 144. Also contributing areresidues 168 to 172 of helix 9 (around the site ofthe two missing residues in 3PGM) and the region218 to 220, a loop between strands I and J. In sum-mary, the two types of interface bury similaramounts of solvent-accessible area and involvesimilar numbers of the various types of atomiccontact.

The quaternary structure of PGAM is of interestbecause this enzyme is unusual in being active aseither a tetramer, dimer or monomer dependingupon the organism in which it occurs. It is prob-able that the A±C interface is present only in thetetrameric form of the enzyme. Inspection of pro-tein sequences in this region shows that all mam-malian dimeric forms of the enzyme have a prolineat residue 168, in contrast to the lysine present inhelix 9 of the tetrameric yeast PGAM. Site-directedmutagenesis of Lys168 to a Pro does indeed desta-bilise the tetramer (by 20 kJ/mol), con®rming thatthe A±C interface is the one not present in thedimeric PGAMs (White et al., 1993b). Presumablythe disruption of helix 9 by a proline is suf®cient todestroy many of the contacts at the A±C interface.The monomeric PGAM from Schizosaccharomycespombe has, in addition to the proline at 168, sub-stantial differences in the A±B interface, includinga deletion of residues 122 to 146, and non-conser-vative amino acid replacements in the region 74 to78 (Nairn et al., 1994).

Differences from 3PGM

The sequence changes made in the new structurecompared to 3PGM are the replacement of Gly46with Lys, Asp52 with Leu, Val170 with Ser, Ser174

Figure 4. A diagram of the PGAMtetramer. The orientation was cho-sen to show the subunit interfacesas clearly as possible. The subunitsA to D are labelled and colouredred, blue, green and black, respect-ively. Only secondary structuralelements involved in contactsbetween subunit A and subunits Band C are drawn as ribbons. Alsoshown and labelled are the loopsbetween helices 6 and 7 andbetween strands I and J. The N ter-mini and the last residues at the Ctermini for which electron densitywas seen are also labelled. Theregions between residues 104 and110 with high B-factors in each sub-unit are shown as thick black tubesand labelled. The catalytic Hisresidues, numbers 8 and 181,are drawn and labelled in eachsubunit.

Phosphoglycerate Mutase Crystal Structure 453

with Val, Thr207 with Ile and Ile208 with Pro. Twoextra residues, Asp169 and Leu170, were inserted.

The r.m.s. difference plot shown in Figure 6highlights some important differences between3PGM and the new structure. The peak betweenresidues 14 and 21 is due to a local register errorbetween sequence and electron density in 3PGM.This has the consequence that in the new structurethe catalytic site is changed in some importantways. For example, the side-chain of Asn14 nowpoints towards the catalytic histidine residueswhereas previously Glu15 occupied this position,and Gly21, a residue implicated in PGAM/BPGAM substrate speci®ty, has moved signi®-cantly. The side chain of Phe19 is now favourablyburied near Phe109, Tyr89 and Leu92 in contrastto its solvent exposure in 3PGM. The peakaround residue 94 arises simply from differenttracing and has no obvious consequences forenzyme function. Another register error is respon-sible for the peak between residues 113 and 117.In the new structure Arg113 and Arg114 makethe active site cleft narrower and deeper than itappeared in 3PGM. The sharp peak around pos-ition 170 is caused by the insertion of the tworesidues missing in 3PGM.

Comparisons to other structures

Similarities of the new PGAM structure to otherstructures of the Protein Data Bank (Bernstein et al.,1977) were analysed using the program DALI(Holm & Sander, 1993). Unfortunately, theF2,6BPase structure had not been deposited andwas therefore not available for comparison. Apartfrom the obsolete PGAM structure 3PGM, thehighest match was found for 1RPA, the structureof rat prostatic acid phosphatase (RPAP) bound totartaric acid (Lindqvist et al., 1993). 159 Ca atomscould be aligned to the PGAM structure with ar.m.s. deviation of 3.6 AÊ . The structural similaritybetween this phosphatase and PGAM has beennoted before (Schneider et al., 1993). Following thiswere binding proteins for leucine/isoleucine,glucose/galactose and ribose, preceding the ®rstmatch against a nucleotide binding protein (1xaa;isopropylmalate dehydrogenase). The programWHATIF (Vriend, 1990) was used to compare thestructure of F2,6BPase (which was kindly madeavailable by the authors; Lee, Y. H., personal com-munication) to that of PGAM. For this enzyme, 170Ca atoms were aligned to the PGAM structurewith a r.m.s. deviation of 2.5 AÊ .

Figure 5. A structure based alignment of yeast PGAM and related proteins with yeast PGAM numbering. The helicesand strands are indicated and labelled above the sequences. Helices 5 and 9 and strand C, which ¯ank insertions, aresplit for clarity. The structural alignments of yeast PGAM with RPAP and F2,6BPase, produced with DALI andWHAT IF, respectively, are separately boxed. Shaded residues are identical in the four sequences. The sequences arefrom the following publications: yeast PGAM (White & Fothergill-Gilmore, 1988); human BPGAM (Joulin et al., 1988);rat PAP (Roiko et al., 1990); rat F2,6BPase (Lively et al., 1988). The Figure was produced using Alscript (Barton, 1993).

Figure 6. A plot showing the r.m.s. difference per resi-due between the new PGAM structure and 3PGM. Thecontinuous line shows the r.m.s. difference betweenmain-chain atoms and the dotted line that between side-chain atoms. Zero values indicate positions where thesequence differs between the two structures.

Phosphoglycerate Mutase Crystal Structure 455

The structure-based sequence alignment ofPGAM with related proteins is shown in Figure 5.Among the conserved positions are the catalyticsite residues Arg7, His8, Gly9, Arg59 and His181(PGAM numbering). Of the others there is no clearstructural reason for the conservation of Ala65,Pro77 Glu107 or Pro206. Interestingly, three con-served leucine residues, numbered 84, 158 and 188are found to lie close to each other in space, withLeu158 forming van der Waals contacts with theother two. This, clearly very stable, packingarrangement lies around 9 AÊ from the catalytic siteand may be a means of ``stapling'' together thethree distant stretches in which the leucine residuesare contained, the loop between strands D and Eand helices 8 and 10. The catalytic His181 liesimmediately before the start of helix 10. It may bethat the interaction of these three leucine residuesis important for the correct orientation of His181within the catalytic site. Their evolutionary conser-vation makes them attractive targets for mutagen-esis.

The catalytic site

The catalytically competent form of the enzymeis phosphorylated on His8, and has so far provedimpossible to crystallise. Phosphorylation of theprotein contained in the crystals used here, and aprevious form (Winn et al., 1981), leads to crystalcracking. This suggests that phosphorylation canlead to signi®cant structural changes but, with thisin mind, the structure reported here can still shedlight on observations made from sequence com-parisons and mutagenesis.

The catalytic site is situated at the bottom of alarge cleft at least 7 AÊ deep with a minimum diam-

eter of 7 AÊ . The diameter would be readilyincreased by small movements of the ¯exible side-chains of arginine and lysine residues. Figure 7shows the site of the catalytic histidine residues atthe bottom of the cleft and other signi®cant resi-dues in the vicinity. The mouth of the cleft isde®ned at the top of the diagram by Lys97 on oneside and arginine residues 113 and 114 on theother. Residues Gly21 and Ser11 on the left-handside of the cleft are those implicated by sequencecomparisons in the de®nition of substrate speci-®city between the PGAMs and the BPGAMs.Arg59 has been suggested as a ligand for thecarboxylate group of the substrate 2,3-BPG(Fothergill-Gilmore & Watson, 1989). Its positionnear the catalytic histidine residues, and the drasticconsequences of its mutation in BPGAM (Ravelet al., 1996), are consistent with this proposal. Theimportance of Glu86 has been shown by itsmutation in PGAM (Walter, R. A. & L.A.F.-G.,unpublished results) and mutagenesis of the corre-sponding Glu327 in F2,6BPase (Lin et al., 1992).Arg7 has not been the subject of mutagenesisstudies but, from its position, is likely to bind the2-phospho group of 2,3-BPG as it is transferred toHis8. Ser55 at the bottom of the diagram is gener-ally a Ser or Thr amongst the PGAMs, theBPGAMs and the related phosphatases. One poss-ible reason for this is readily apparent. The side-chain hydroxyl group of Ser55 hydrogen bondsatom NE2 of the His181 side-chain and contributesto the correct orientation of the imidazole ring.This interaction is conserved in the RPAP structurebetween Thr75 and His257, and in F2,6BPasebetween Ser52 and His141. The catalytic site cleftcontains many basic residues, resulting in a largeoverall positive electrostatic potential (Figure 8).The effect is even larger in the new structurethan the earlier analysis of 3PGM suggested(Warwicker, 1986). Such a concentration of basicresidues may be necessary to attract a highly nega-tively-charged substrate to a site containing acatalytically essential glutamate residue.

The protein hydrogen bonding network in thevicinity of His8 is shown in more detail in Figure 9.His181 and Arg59 are both within hydrogen bond-ing range of His8, suggesting that His8 is unproto-nated. The catalytic site histidine residues arecloser in this structure than they are in the relatedprotein structures. Here they are as little as 3.2 AÊ

apart whereas in the RPAP the minimum distanceis 4.7 AÊ and in the F2,6BPase structure they are4.2 AÊ apart (Lee et al., 1996). This may be related tothe absence of salt or substrate in the PGAM struc-ture compared to the presence of a phosphate inthe F2,6BPase structure, and a tartrate molecule inthe RPAP structure. It may indicate movements ofthe catalytic site on ligand binding. Glu86, a poss-ible candidate to raise the pKa of His181, on thebasis of similarity to F2,6BPase (Lin et al., 1992) issituated some 4.5 AÊ from the ND1 atom of His181,and is therefore not likely to exert enough in¯u-ence in its present conformation. However, nearer

Figure 7. View of the catalytic histidine residues, 8 and 181, at the bottom of the catalytic site cleft. Also included arebasic residues in the vicinity and other residues mentioned in the text.

456 Phosphoglycerate Mutase Crystal Structure

the pH optimum of 7.0, when His181 may becomeprotonated, Glu86 may be induced to move closer.His8 is also hydrogen bonded, by its ND1 atom, tothe backbone carbonyl group of Gly9, a residuetotally conserved amongst the mutases and relatedphophatases. The reason for the conservation ofthis Gly appears not to lie in unusual backboneangles (phi, psi � ÿ 66, ÿ44), but with the absenceof space for a side-chain. Interestingly, Arg59 alsoH-bonds the backbone carbonyl of another Gly,residue 21. In this case the glycine residue does

have unusual backbone torsion angles (phi,psi � 66, ÿ119). As mentioned before, this positionis occupied by serine in the BPGAMs, and since aserine residue would be unlikely to adopt the back-bone conformation seen for Gly21 here, theBPGAM structure may well differ signi®cantly inthis region. This difference could easily be trans-mitted via Arg59 to other catalytic site residues.Perhaps this is at least partly responsible for thedifferent substrate speci®cites of the PGAMs andthe BPGAMs.

Figure 8. The solvent-accessiblesurface with positive and negativeelectrostatic potential coloured blueand red respectively. The units ofthe scale are kT where k is theBoltzmann constant and T is tem-perature. The catalytic site cleft isshown in the centre of the Figure.In comparison to the rest of themolecular surface, the concen-tration of basic residues in the cata-lytic site cleft is obvious. Thecatalytic histidine residues arelocated in the upper left area of theelectrostatically positive region.Arginine residues 113 and 114 formthe lip visible towards the lowerright of the electrostatically positiveregion. The small electrostaticallynegative area at the base ofthe cleft is due to Glu86. TheFigure was produced using GRASP(Nicholls et al., 1991).

Figure 9. The H-bonding network around His8 and the position of Glu86. H-bonds are represented by dotted linesand their lengths in AÊ indicated. Also drawn are residues included in the two-sites hypothesis (Ravel et al., 1996; seethe text). A |2Fo ÿ Fc| map of the area is shown contoured at 1s.

Phosphoglycerate Mutase Crystal Structure 457

Based on analysis of BPGAM mutants, two sep-arate binding sites have been postulated for mono-phosphoglycerate and for 2,3-BPG (Ravel et al.,1996). The suggested residues correspond inPGAM to Ser11, Glu15 and Lys245 (monophospho-glycerate) and Gly21 and Arg87 (2,3-BPG). Thenew structure only partially supports this idea.Gly21 and Arg87 are close, having atoms within6 AÊ . Ser11 and the side-chain of Glu15 are a similardistance apart, but Glu15 is solvent-exposed on adifferent face of the protein, not within the catalyticsite cleft. Lys245 is within the ¯exible C-terminaltail not seen in the electron density map. Ser11,Arg87 and Gly21 are shown in Figure 9.

Materials and Methods

Crystallization

The protein, puri®ed as previously described (White& Fothergill-Gilmore, 1992), was crystallized by the sit-ting drop method. Equal volumes of protein solution, ata concentration of 10 mg/ml, and well solution weremixed. The well solution contained 60 mM Tris-HCl buf-fer (pH 8.65), 120 mM lithium sulphate and 22 to 24%PEG 4000. The protein solution contained 1 mM inositolhexakisphosphate, a potent PGAM inhibitor, but notrace of this was seen in the ®nal structure. Crystals ofan elongated rod shape grew within a week to a typicalsize of 0.4 mm � 0.1 mm � 0.1 mm.

Crystals were ¯ash frozen in liquid nitrogen after briefimmersion in a cryoprotectant solution of mother liquorcontaining 15% (v/v) glycerol. They were stored inliquid nitrogen until the diffraction experiments.

Data collection and processing

Data were collected on MAR 18 cm image plate andCCD detectors (MAR Research, Hamburg, Germany)

using a synchrotron X-ray source (l � 0.95 AÊ , stationX31 at the EMBL outstation, Hamburg, Germany). Dueto lack of time only 60� of diffraction data were collectedwith the image plate and 140� on a CCD detector. There¯ections fully recorded with CCD have increased thequality of the data, although minor errors in the proces-sing software did not allow us to include partial re¯ec-tions. The crystal was maintained at a temperature of100 K throughout data collection. The crystal was foundto be twinned but we have processed the minor twindata and it is estimated to diffract about 25 to 30% asstongly as the major component. The high Rmerge valuesof our data (8.2% overall and 25.3 in the highest resol-ution shell) are the result of the inclusion of data fromtwo different detectors and crystal twinning. The imageswere integrated using DENZO and scaled and mergedusing SCALEPACK (Minor, 1993). Overall there were49,635 unique re¯ections with an average multiplicity of1.96 for the whole dataset, and 1.91 in the highest resol-ution shell. The dataset was 89.9% complete in thehighest resolution shell and 88.7% complete overall.The I/s(I) values were 2.4 in the highest resolutionshell and 5.3 overall. The crystals were found to be ofspacegroup P21 with cell dimensions a � 81.7 AÊ ,b � 84.6 AÊ , c � 88.9 AÊ , b � 111.7�.

Structure determination

The molecular replacement package AMORE (Navaza,1994) was used to determine the position and orientationof the protein in the unit cell. A tetrameric structurederived from the PDB entry 3PGM was used as a searchmodel, and X-ray data from 8 to 4 AÊ were included inthe calculations. Four strong solutions, well separatedfrom others, were obtained from the rotational search.These were found to arise from C2 pseudosymmetry atlow resolution. After translation the best of these sol-utions yielded an R-factor of 58.4%. This improved to46.6% after rigid-body re®nement with X-PLOR 3.1(BruÈ nger, 1992a).

458 Phosphoglycerate Mutase Crystal Structure

Cycles of simulated annealing (BruÈ nger et al., 1987,1990; BruÈ nger, 1992b) using X-PLOR 3.1 and X-PLORonline 3.843 (BruÈ nger, 1997 available at http://xplor.csb.yale.edu/xplorinfo/xploronline.html) and rebuild-ing, using O (Jones et al., 1991) and FRODO (Jones,1978), were then carried out with re¯ections F > 2s(F)included. The improvement of the model was monitoredwith the help of the free R-factor. This was calculatedfrom 10% of randomly selected re¯ections that wereexcluded from the re®nement procedure. Programs ofthe CCP4 suite (CCP4, 1994) were used for manipulatingcoordinates and for map calculations. In regions wherethe electron density was unclear simulated annealingomit maps were calculated. In this way phase biastowards 3PGM was reduced. In the early stages strictnon-crystallographic symmetry (NCS) was imposed onthe four subunits of the tetramer. This was replaced inthe later stages by strong (500 kcal/AÊ 2) harmonic NCSrestraints enforcing ``dimer-of-dimer'' symmetry. Thismove was justi®ed by a 3% drop in both the R-factorand the free R-factor, plus an improvement in stereo-chemistry. Removal of these restraints produced noimprovement in the free R-factor. The AVE suite of mol-ecular averaging programs (Kleywegt & Jones, 1994) wasused to produced phase-averaged maps and the stereo-chemistry of the model was monitored with PROCHECK(Laskowski et al., 1993). The protein sequence derivedfrom DNA sequencing (White & Fothergill-Gilmore,1988) was used. When the R-factor was below 24% watermolecules were introduced into difference map peakswithin range of suitable hydrogen-bonding partners. Afterre®nement water molecules for which there was no cleardensity in the |2Fo ÿ Fc| map, or which had drifted out ofhydrogen bonding range, were deleted. Group B-factorre®nement was carried out, assigning one value each forthe side-chain and main-chain of each protein residue, andone value for each water molecule. B-factors of residues inrestrained subunits were restrained with a target deviationof 1.0 AÊ 2. Individual B-factor re®nement was unjusti®edsince it caused no improvement in the free R-factor.

Three-dimensional structural comparison of PGAMand other proteins was carried out using DALI (Holm &Sander, 1993) and WHAT IF (Vriend, 1990). Secondarystructure assignment was carried out using the YASSPAcommand of O. This assigns secondary structure on thebasis of local similarity to helical and strand templates.Unlike DSSP (Kabsch & Sander, 1983) beta strands arenot required to be hydrogen-bonded to other strands.The de®nitions of helices produced by the two programsare very similar.

The structure has been deposited with the ProteinData Bank under the code 4pgm. The structure factorshave been deposited as r4pgmsf.

Acknowledgements

We thank Rebecca Walter for providing the protein.The invaluable help of Drs Matthew Newman and CarrieWilmot during the re®nement is acknowledged. Wethank the European Community STD Programme for®nancial support, and the European Union for supportof the work at EMBL Hamburg through the HCMPAccess to Large Installation Project, Contract numberCHGE-CT93-0040. This work bene®tted from the use ofthe SEQNET facility. S.E.V.P. is an International ResearchScholar of the Howard Hughes Medical Institute.

References

Barton, G. J. (1993). ALSCRIPT, a tool to format multiplesequence alignments. Protein Eng. 6, 37±40.

Bernstein, F. C., Koetzle, T. F., Williams, G. J. B., Meyer,E. F., Jr, Brice, M. D., Rodgers, J. R., Kenard, O.,Shimanouchi, T. & Tasumi, M. (1977). The ProteinData Bank: a computer-based archival ®le formacromolecular structures. J. Mol. Biol. 112, 535±542.

BraÈnden, C.-I. (1980). Relation between structure andfunction of a/b-proteins. Quart. Rev. Biophys. 13,317±338.

BruÈ nger, A. T. (1992a). X-PLOR (Version 3. 1) Manual,Yale University Press, New Haven and London.

BruÈ nger, A. T. (1992b). The free R-value: a novel statisti-cal quantity for assessing the accuracy of crystalstructures. Nature, 355, 472±474.

BruÈ nger, A. T., Kuriyan, J. & Karplus, M. (1987).Crystallographic R-factor re®nement by moleculardynamics. Science, 235, 458±460.

BruÈ nger, A. T., Krukowski, A. & Erickson, J. (1990).Slow-cooling protocols for crystallographic re®ne-ment by simulated annealing. Acta Crystallog. sect.A, 46, 585±593.

Campbell, J. W., Watson, H. C. & Hodgson, G. I. (1974).Structure of yeast phosphoglycerate mutase. Nature,250, 301±303.

CCP4, (1994). The CCP4 suite: programs for proteincrystallography. Acta Crystallog. sect. D, 50, 760±763.

Fothergill-Gilmore, L. A. & Michels, P. A. M. (1993).Evolution of glycolysis. Prog. Biophys. Mol. Biol. 59,105±236.

Fothergill-Gilmore, L. A. & Watson, H. C. (1989). Thephosphoglycerate mutases. Advan. Enzymol. RelatedAreas Mol. Biol. 62, 227±313.

Garel, M.-C., Joulin, V., Le Boulch, P., Calvin, M.-C.,Prehu, M.-O., Arous, N., Longin, R., Rosa, R., Rosa,J. & Cohen-Solal, M. (1989). Human bisphosphogly-cerate mutase: expression in Escherichia coli and useof site-directed mutagenesis in the evaluation of therole of the carboxyl-terminal region in the enzy-matic mechanism. J. Biol. Chem. 264, 18966±18972.

Garel, M.-C., Lemarchandel, V., Calvin, M.-C., Arous,N., Craescu, C. T., Prehu, M.-O., Rosa, J. & Rosa, R.(1993). Amino acid residues involved in the cataly-tic site of human erythrocyte bisphosphoglyceratemutase. Eur. J. Biochem. 213, 493±500.

Holm, L. & Sander, C. (1993). Protein structure compari-son by alignment of distance matrices. J. Mol. Biol.233, 123±138.

Jones, T. A. (1978). A graphics model building andre®nement system for macromolecules. J. Appl.Crystallog. 11, 268±272.

Jones, T. A., Zou, J.-Y., Cowan, S. W. & Kjeldgaard, M.(1991). Improved methods for building proteinmodels in electron density maps and the location oferrors in these models. Acta Crystallog. sect. A, 47,110±119.

Joulin, V., Garel, M.-C., Le Boulch, P., Valentin, C.,Rosa, R., Rosa, J. & Cohen-Solal, M. (1988).Isolation and characterization of the human 2,3-bisphosphoglycerate mutase gene. J. Biol. Chem. 263,15785±15790.

Kabsch, W. & Sander, C. (1983). Dictionary of proteinsecondary structure: pattern recognition of hydro-gen-bonded and geometrical features. Biopolymers,22, 2577±2637.

Phosphoglycerate Mutase Crystal Structure 459

Kleywegt, G. J. & Jones, A. T. (1994). Halloween . . .masks and bones. In From First Map to Final Model(Bailey, S., Hubbard, R. & Walter, D., eds), pp. 59±66, SERC Daresbury Laboratory, Warrington, UK.

Kraulis, J. (1991). MOLSCRIPT: a program to produceboth detailed and schematic plots of proteinstructures. J. Appl. Crystallog. 24, 946±950.

Laskowski, R., Macarthur, M., Moss, D. & Thornton, J.(1993). Procheck: a program to check stereochemicalquality of protein structures. J. Appl. Crystallog. 26,283±290.

Lee, Y.-H., Ogata, C., P¯ugrath, J. W., Levitt, D. G.,Sarma, R., Banaszak, L. J. & Pilkis, S. J. (1996).Crystal structure of the rat liver fructose-2,6-bispho-sphatase based on selenomethionine multiwave-length anomolous dispersion phases. Biochemistry,35, 6010±6019.

Lin, K., Li, L., Correia, J. J. & Pilkis, S. J. (1992). Glu 327is part of a catalytic triad in rat liver fructose-2,6±bisphosphatase. J. Biol. Chem. 267, 6556±6562.

Lindqvist, Y., Schneider, G. & Vihko, P. (1993). Three-dimensional structure of rat acid phosphatase incomplex with L(�)-tartrate. J. Biol. Chem. 268,20744±20746.

Lively, M. O., El-Maghrabi, M. R., Pilkis, J., D'Angelo,G., Colosia, A. D., Ciavola, J.-A., Fraser, B. A. &Pilkis, S. J. (1988). Complete amino-acid sequenceof rat-liver 6-phosphofructo-2-kinase/fructose-2,6-bisphosphatase. J. Biol. Chem. 263, 839±849.

Minor, W. (1993). ``XDISPLAYF'' Program, PurdueUniversity, USA.

Nairn, J., Price, N. C., Fothergill-Gilmore, L. A., Walker,G. E., Fothergill, J. E. & Dunbar, B. (1994). Theamino acid sequence of the small monomeric phos-phoglycerate mutase from the ®ssion yeast Schizo-saccharomyces pombe. Biochem. J. 297, 603±608.

Nairn, J., Krell, T., Coggins, J. R., Pitt, A. R., Fothergill-Gilmore, L. A., Walter, R. & Price, N. C. (1995). Theuse of mass spectrometry to examine the formationand hydrolysis of the phosphorylated form of phos-phoglycerate mutase. FEBS Letters, 359, 192±194.

Navaza, J. (1994). AMoRe: an automated package formolecular replacement. Acta Crystallog. sect. A, 50,157±163.

Nicholls, A., Sharp, K. A. & Honig, B. (1991). Proteinfolding and association: insights from the interfacialand thermodynamic properties of hydrocarbons.Proteins: Struct. Funct. & Genet. 11, 281±286.

Price, N. C., Duncan, D. & McAlister, J. W. (1985).Inactivation of rabbit muscle phosphoglyceratemutase by limited proteolysis with thermolysin. Bio-chem. J. 229, 167±171.

Ravel, P., Arous, N., Croisille, L., Craescu, C. T., Rosa,J., Rosa, R. & Garel, M. C. (1996). Determination ofamino acid residues implicated in substrates andcofactor binding sites in human bisphosphoglyce-rate mutase. A tool for therapeutic approaches. Br.J. Haem. 93, Abstract 717.

Roiko, K., Jaenne, O. A. & Vihko, P. (1990). Primarystructure of rat secretory acid-phosphatase and

comparison to other acid-phosphatases. Gene, 89,223±229.

Rose, Z. B. (1971). The phosphorylation of yeast phos-phoglycerate mutase. Arch. Biochem. Biophys. 146,359±360.

Rose, Z. B. (1980). The enzymology of 2,3-bisphosphoglycerate. Advan. Enzymol. Relat. AreasMol. Biol. 51, 211±253.

Sasaki, R., Sugimoto, E. & Chiba, H. (1966). Yeast phos-phoglyceric acid mutase-modifying enzyme. Arch.Biochem. Biophys. 115, 53±61.

Schneider, G., Lindqvist, Y. & Vihko, P. (1993). Three-dimensional structure of rat acid phosphatase.EMBO J. 12, 2609±2615.

Tauler, A., Lin, K. & Pilkis, S. J. (1990). Hepatic 6-phos-phofructo-2-kinase/fructose-2,6-bisphophatase: Useof site-directed mutagenesis to evaluate the roles ofHis-258 and His-392 in catalysis. J. Biol. Chem. 265,15617±15622.

UhrõÂnovaÂ, S., UhrõÂn, D., Nairn, J., Price, N. C.,Fothergill-Gilmore, L. A. & Barlow, P. N. (1997).Backbone assignment of double labelled 23.7 kDaphosphoglycerate mutase from Schizosaccharomycespombe. J. Biomol. NMR, 10, 309±310.

Vriend, G. (1990). WHAT IF: a molecular modeling anddrug design program. J. Mol. Graphics, 8, 52±56.

Walter, R. A. & Fothergill-Gilmore, L. A. (1995).Mutating mutases: phosphoglycerate mutase ligandinteractions studied by site-directed mutagenesis. InBiopolymers and Bioproducts: Structure, Function andApplications (Svasti, M. R. J., ed.), pp. 369±375,Samakkhisan Co. Ltd., Bangkok.

Warwicker, J. (1986). Continuum dielectric modelling ofthe protein-solvent system, and calculation of thelong-range electrostatic ®eld of the enzyme phos-phoglycerate mutase. J. Theoret. Biol. 121, 199±210.

Watson, H. C. (1982). Protein Data Bank Entry: Phospho-glycerate Mutase (Yeast), Brookhaven, New York.

White, M. F. & Fothergill-Gilmore, L. A. (1988).Sequence of the gene encoding phosphoglyceratemutase from Saccharomyces cerevisiae. FEBS Letters,229, 383±387.

White, M. F. & Fothergill-Gilmore, L. A. (1992).Development of a mutagenesis, expression andpuri®cation system for yeast phosphoglyceratemutase. Eur. J. Biochem. 207, 709±714.

White, M. F., Fothergill-Gilmore, L. A., Kelly, S. M. &Price, N. C. (1993a). Substitution of His-181 byalanine in yeast phosphoglycerate mutase leads tocofactor-induced dissociation of the tetramericstructure. Biochem. J. 291, 479±483.

White, M. F., Fothergill-Gilmore, L. A., Kelly, S. M. &Price, N. C. (1993b). Dissociation of the tetramericphosphoglycerate mutase from yeast by a mutationin the subunit contact region. Biochem. J. 295, 743±748.

Winn, S. I., Watson, H. C., Harkins, R. N. & Fothergill,L. A. (1981). Structure and activity of phosphoglyce-rate mutase. Phil. Trans. Roy. Soc. London, Ser. B,293, 121±130.

T. Richmond

(Received 13 June 1997; received in revised form 21 November 1997; accepted 24 November 1997)