Crystal structure of common type acylphosphatase from bovine testis

Research Article

69

Crystal structure of common type acylphosphatase from bovine testis Marjolein MGM Thunnissen1, Niccolo’ Taddei2, Gianfranco Liguri2, Giampietro Ramponi2 and Pär Nordlund1* Background: Acylphosphatase (ACP) is a low molecular weight phosphomonohydrolase catalyzing with high specificity the hydrolysis of the carboxyl-phosphate bond present in acylphosphates. The enzyme is thought to regulate metabolic processes in which acylphosphates are involved, such as glycolysis and the production of ribonucleotides. Furthermore the enzyme is capable of hydrolyzing the phospho-aspartyl intermediate formed during the action of membrane pumps such as (Ca2+ + Mg2+) ATPase. Although the tertiary structure of a muscle ACP has been determined by NMR spectroscopy, little is known about the catalytic mechanism of ACP and further structures might provide an increased understanding. Results: The structure of ‘common type’ ACP from bovine testis has been determined by X-ray crystallography to a resolution of 1.8 Å. The structure has been refined to an R factor of 17.0 % using all data between 15 and 1.8 Å. The binding of a sulphate and a chloride ion in the active centre allows a detailed description of this site. The overall protein folds of common type and muscle ACP are similar but their loops have very different conformations. These differences, in part, are probably caused by the binding of the ions in the active site of the common type form. The phosphate-binding loop of ACP shows some remarkable similarities to that of low molecular weight protein tyrosine phosphatase.

Addresses: 1Department of Molecular Biology, University of Stockholm, S-106 91 Stockholm, Sweden and 2Department of Biochemical Sciences, University of Florence, Viale Morgagni 50, 50134 Florence, Italy. *Corresponding author. E-mail: [email protected] Key words: acylphosphatase, phosphoric monoester hydrolases, substrate-assisted catalysis, X-ray crystallography Received: 27 August 1996 Revisions requested: 23 September 1996 Revisions received: 25 October 1996 Accepted: 4 November 1996 Electronic identifier: 0969-2126-005-00069 Structure 15 January 1997, 5:69–79 © Current Biology Ltd ISSN 0969-2126

Conclusions: The active site of ACP has been located, enabling a reaction mechanism to be suggested in which the phosphate moiety bound to Arg23 acts as a base, abstracting a proton from a nucleophilic water molecule liganded to Asn41. The transition-state intermediate is stabilized by the phosphate-binding loop. We suggest the catalysis to be substrate assisted, which probably explains why this enzyme can only hydrolyze acylphosphates.

Introduction Acylphosphatase (ACP; E.C. 3.6.1.7) is one of the smallest enzymes known, with a molecular weight of 11 kDa (98 residues). It catalyzes the hydrolysis of acylphosphates, such as 1,3-bisphosphoglycerate, acetylphosphate, carbamoylphosphate, succinyl phosphate, and b-aspartylphosphate membrane pump intermediates [1]. ACP differs from alkaline phosphatases, acid phosphatases or protein phosphatases by its inability to hydrolyze other phosphate esters or anhydrides. Although there is some debate about the exact biological function of ACP, it is increasingly apparent that it could play a regulatory role in all metabolic processes in which acylphosphates are involved, most notably glycolysis and pyrimidine biosynthesis [1]. In these processes, acylphosphatase is suggested to be involved in feedback mechanisms or control of intracellular levels of carbamoylphosphate or 1,3-bisphosphoglycerate. Furthermore it is thought to be involved in the hydrolysis of the phosphorylated aspartyl intermediates formed during the

action of membrane pumps such as the (Ca2+ + Mg2+) ATPase and the (Na+/K+)ATPase of the neuronal and erythrocyte cell membranes [2,3]. ACP from erythrocytes has a very high affinity for the intermediate formed during the action of the erythrocyte membrane sodium pump and is thought to hydrolyze these intermediates [4]. ACP could thereby affect the coupling between ATP hydrolysis and cation transport, modifying the normal ordered reaction sequence and short-circuiting the cation transport system. Since ACP indirectly promotes polyphosphoinositide dephosphorylation, probably through regulation of intracellular Ca2+ levels, it has been suggested that ACP has a role in inositol lipid metabolism in platelets [5]. Acylphosphatase is an ubiquitous enzyme that is found in many tissues, mainly of vertebrate species, as a cytosolic protein. Cytosolic ACP exists as two isoenzymes sharing over 50 % sequence homology, which suggests that both are derived from a common ancestor by gene duplication [6–8]. The two isoenzymes are present at different ratios in various tissues and have been called muscular and

70

Structure 1997, Vol 5 No 1

erythrocyte (or common type) acylphosphatase. The amino acid sequence of the two isoenzymes in different animal classes is highly conserved [9]. Both enzymes exhibit similar kinetic properties, for example a pH optimum of 5.3 to 5.5 [6], although the common type enzyme has a somewhat higher specific activity. The three-dimensional structure of the muscular isoenzyme from horse muscle has been determined by NMR spectroscopy [10,11]. It has a globular a/b fold, consisting of a b sheet with five antiparallel strands and two a helices packed parallel on the same side of the sheet, forming an a/b sandwich protein. The secondary structure elements can be considered as two intercalating bab units, which gives rise to a 4-1-3-2 b strand topology. This arrangement has also been observed in ferrodoxin [12], the activation domain of pro-carboxypeptidase [13,14], the RNA-binding domain of the small nuclear ribonucleoprotein (RNP) A and C [15], the C-terminal domain of ribosomal protein L7/L12 [16] and L30 [17]. The histidine containing phospho-carrier protein (HPr) displays a similar fold but has a different topology [18,19]. Since ACP, HPr and RNP are all involved in phosphate binding and share a similar fold despite having no significant sequence homology, a common phosphate-binding motif has been suggested [20]. This motif is located at the N terminus of the first helix where, in all three proteins, important basic residues involved in the binding of phosphate are located. Although many sequences of ACP from different organisms are available and an NMR structure for horse muscle acylphosphatase is known, there is still surprisingly little known about the reaction mechanism. The proposed mechanism for the muscle enzyme suggests that the substrate is held primarily by the binding of its phosphate moiety to a positively charged group with pKa 12.0 and less so to a second residue with a pKa of 7.9 [21]. A water molecule bound locally, probably to a hydrophilic sidechain, is likely to be the nucleophilic agent during the first step of hydrolysis. The second (rate-limiting) step is electron withdrawal by the substrate acyl group. Chemical modification studies and site-directed mutagenesis showed that the conserved Arg23 at the N-terminal end of the first helix is the principal residue involved in the binding of the phosphate group of the substrate [22]. However, there has been no clear identification of the second positively-charged residue in this mechanism or of the residue that activates the nucleophilic water molecule. From NMR studies using a competitive inhibitor and [Cr(CN)6]3– as a relaxation probe it became clear that the active site was in the neighbourhood of the C terminus in a rather shallow cavity [10]. This has been partly confirmed by recent mutational studies [23,24]. Site-directed mutagenesis experiments on Asn41 have shown that this residue is essential for catalysis [25]. Since only Kcat is affected by the mutations and not Km or Ki, it is thought that Asn41 is probably directly

involved in the reaction mechanism by polarizing and orientating the water molecule. Although the NMR structure showed the three-dimensional fold of acylphosphatase, the composition of the active site and hence the structural basis for the catalytic action of acylphosphatase are still unknown. Here we describe determination of the structure of common-type ACP from bovine testis by X-ray diffraction. A comparison with the NMR structure is made. The presence of a sulphate ion and a chloride, both competitive inhibitors of the enzyme, in the active centre allowed the site to be located and described in detail.

Results and discussion Overall structure

The structure of acylphosphatase has been solved to a resolution of 1.8 Å with an R factor of 17.0 % for all data between 15 and 1.8 Å. The structure shows a five-stranded mixed b sheet with two a helices running parallel to the b sheet. The b sheet is slightly curved with a right-handed twist. The a helices protect one side of the b sheet, whereas the other side is exposed to the solvent. Hydrophobic residues from the a helices and b sheet intercalate to form the core of the enzyme. The overall structure reveals a very compact protein with the dimensions 35 × 35 × 45 Å. The loops between secondary structure elements are relatively short at the bottom side of the molecule, whereas at the top they are slightly longer (Fig. 1). At the top of the molecule, loop 14–21, connecting the first strand in b sheet to the first a helix, forms an extended stretch with two type IV b turns (residues 14–17 and residues 17–20 ) (Fig. 2). This uncommon conformation is due to the unusual f/ψ combinations of Gly15, Gly19 and Phe21. This loop is stabilized by interactions of Val17, which occupies a hydrophobic pocket formed by residues Ile13, Phe22, Val47 and Ile75. Further stabilization results from hydrogen bonds between the amide group of Phe14 and carbonyl oxygen of Arg77, the Nε2 Gln18 and carbonyl oxygens of Thr46 and Thr42, and the carbonyl oxygen of Phe21 and the amide group of Tyr25. This loop is involved in some crystal packing interactions; the sidechain of Phe21 points towards an interface between two crystallographically related molecules, where it is stabilized by hydrophobic interactions with His60, Glu63, Trp64 and Lys68 of the neighbouring molecule. The loop 42–45 forms a tight type I b turn connecting strands 2 and 3 of the b sheet. Loop 68–74 has an extended conformation and runs along loop 14–21, resulting in hydrogen-bonding interactions between the two. Loop 68–74 makes a type I b turn at residues from 70 to 73. At the bottom of the molecule, one short loop (residues 33–35) connects helix I to strand 2, whereas no loop is

Research Article Common type acylphosphatase structure Thunnissen et al.

71

Figure 1 The tertiary structure of acylphosphatase. (a) Front view, a helices are in red, b strands in blue. (b) Side view, orientation is approximately 90° to Figure 1a. (c) Stereo diagram of a Ca trace of ACP. Every tenth residue is labelled.

found between strand 3 and the second helix. A long loop (residues 86–93) running across the whole of the bottom of the molecule connects strand 4 to the C-terminal strand 5. This loop is stabilized by several hydrogen bonds, both from sidechain atoms and backbone atoms.

The b bulge observed in the NMR structure at position 88 is missing and the b sheet is regular in the X-ray structure. The beginnings and ends of the secondary structure elements also differ slightly between the X-ray and NMR structures.

Comparison to the NMR structure of muscle type acylphosphatase.

A comparison between Ca positions in the averaged and regularized NMR structure and the X-ray structure shows a root mean square (rms) difference of about 3.1 Å for all 98 Ca positions. When only the 76 core Ca positions are compared, the rms difference decreases to 1.3 Å (Table 1 for comparison of individual NMR structures). The atoms omitted during the comparison are located in the loops connecting secondary structure elements. The conformation of these loops is very different in the averaged NMR and X-ray structures. In particular loops 14–21, 42–45, 51–55 and 68–76 are mobile in the NMR structure. The X-ray structure does not show any disordered regions except for the N-terminal residues; the density of the residues in the

The NMR structure of muscle form acylphosphatase is known [11]. This enzyme has 50% homology with the common type acylphosphatase. The NMR data consist of an ensemble of five structures. To facilitate the comparison of the X-ray structure and the NMR ensemble, an averaged NMR structure was calculated (see Materials and methods). Comparison of the averaged NMR and X-ray structure of the two acylphosphatases shows that the secondary structure elements are preserved and well conserved in both structures and that only some of the details differ.

72

Structure 1997, Vol 5 No 1

Figure 2 Stereo diagram of the structure of loop 13–24. For clarity, the backbone of the protein is drawn as a ribbon and the sidechains are represented by a ball-and-stick model. Atoms are shown in standard colours.

loops is well defined and the conformation of the residues is unambiguous. The biggest difference between the two structures is found in loop 14–21. The position of this loop in the X-ray structure is up to 6 Å from the mean Ca positions in the averaged NMR structure, and the loop in the X-ray structure has a well defined conformation that has not been observed in any of the five structures in the NMR ensemble (see Fig. 3). It is possible that the cause of these differences is the binding of a sulphate and a chloride ion (see below). The exact conformation of the loops might be induced by this binding, especially that of the chloride ion, because it makes many interactions with backbone atoms. A sequence comparison between muscle and common type ACPs shows that the residues of the loop are quite well preserved and not likely to be the cause of the conformational change. Since this loop is also involved in some crystal contacts, these could also be a cause of the difference in conformation. However the conformation of this loop is very similar to the phosphate-binding loop found in low molecular weight protein tyrosine phosphatase (lmPTP) [26] and it seems very likely that crystal contacts are not a cause of the difference. Another possibility is that the loop structures are underdetermined in the NMR structure. These differences cause the X-ray structure to be more compact than the NMR structure. This is also observed in

other proteins for which NMR and X-ray structures have been compared [27,28]. Crystal contacts

A range of crystal contacts is made in the lattice. Some of these are hydrophobic interactions such as those made by Phe21, His60, Glu63, Trp64 and Lys68. These interactions are further extended by the interaction of Lys68 with the sulphate ion bound in the active site of a symmetry-related molecule. These contacts are extensive enough to give the impression that the enzyme could act as a dimer, but no biochemical evidence has been found to corroborate such an assumption. In muscle ACP, residue 68 is not conserved as a lysine. Residue 67 is a lysine, but mutations of this residue have no effect on activity (N Taddei, unpublished results). The position of Phe21 in common type ACP is curious, since it is located on the surface and would be fully exposed in a monomeric form of the protein. The fact that Phe21 is exposed suggests that it has some functional importance and that this area might be part of a protein–protein interaction surface and/or the binding site for substrate molecules. The corresponding residue is a cysteine in muscle acylphosphatase. The position of this residue is unknown since it is part of a mobile loop in the muscle acylphosphatase structure.

Table 1 Structures of the NMR ensemble of muscle type acylphosphatase compared with X-ray structure of common type acylphosphatase in root mean square deviations. ACP All residues (Å) Core residues (Å)

NMR1

NMR2

NMR3

NMR4

NMR5

3.377 1.94 (76)

3.446 1.471 (75)

3.911 2.049 (78)

3.494 1.481 (76)

3.698 1.320 (74)

Number of compared residues in the core is shown in parentheses.

Research Article Common type acylphosphatase structure Thunnissen et al.

73

Figure 3 The X-ray structure of common type acylphosphatase superimposed on the NMR ensemble of the muscle form acylphosphatase. (a) The complete structure; the X-ray structure is labelled every tenth residue. (b) Close up of the superimposition of loop 12–26; the X-ray structure is labelled at every second residue. The NMR structures are drawn as thin white coils and the X-ray structure as a thick black coil.

Sulphate-binding sites and the active site

Together the three loops at the top of the enzyme create a shallow groove on the surface into which a sulphate ion is bound to the N-terminal end of the first a helix. The sulphate molecule makes extensive contacts with the guanidinium group of Arg23 and makes additional hydrogen bonds to the backbone amide groups of Phe21, Arg23 and Lys24 (Fig. 4). Furthermore, it makes a hydrogen bond to Lys68 from a symmetry-related molecule. Several water molecules participate in hydrogen bonds with the sulphate molecule. During the refinement it was noticed that

a water molecule near the sulphate ion had developed a very low B factor; this was interpreted to mean that it could actually be a chloride ion, a competitive inhibitor for acylphosphatase [21]. Chloride was present in the crystallization set-up as the buffer was titrated to the correct pH using HCl, and could also be present as a contaminant of the PEG4000 used. In addition, the observed interactions supported the assumption that this peak might be a chloride ion, and it was refined as such. This ion makes many interactions to the backbone amide groups of residues 19–21; it is also in contact with the sulphate ion and a

Figure 4

The active site of acylphosphatase. (a) Stereo diagram of the structure of the active site; atoms in are shown in standard colours. (b) Schematic overview of the sulphate-binding site.

74

Structure 1997, Vol 5 No 1

Figure 5 Stereo diagram of the superposition of the phosphate-binding loops in protein phosphotyrosine phosphatase (PTP, white) and acylphosphatase (dark grey). Labels for residues in ACP are on top, those for PTP at the bottom.

water molecule (Fig. 4). The short distance between the sulphate and the chloride ions suggests that the sulphate ion is protonated at pH 3.5 which allows the presumed interaction between these ions. The Arg23 sidechain is at 3.8 Å from the sidechain of Asn41. Asn41 makes a further contact to a water molecule, but the latter is not in contact with the sulphate ion. No other polar groups are found in the neighbourhood of the sulphate ion. A comparison of the phosphate-binding loops of ACP with these of lmPTP [26], which has a cradle-like loop structure common to all protein tyrosine phosphatases, shows that they are remarkably similar (rms difference for 37 equivalent atoms is 1.33 Å) (Fig. 5). In both enzymes the main interactions involved in the binding of the phosphate

moiety are made through an arginine residue, which makes a bidentate interaction through its guanidinium head group, and through backbone nitrogens of residues in the phosphate-binding cradle. In both enzymes a sulphate ion is found in the active site; however, the sulphate ion in lmPTP is bound deeper into the pocket and makes more extensive contacts with backbone atoms. In the superposition it can be observed that one of the oxygens of the sulphate ion in lmPTP is replaced by the chloride ion in ACP (Fig. 5). In lmPTP, the nucleophilic agent is a cysteine at position 12, at the edge of the phosphate-binding loop. No such residue can be found in acylphosphatase because the phosphate-binding loops of both enzymes are very different

Figure 6

Sulphate binding in acylphosphatase. (a) The two sulphate-binding sites in acylphosphatase. The active site sulphate ion is numbered Sul-1, the other sulphate ion Sul-2. The protein is depicted as a coil, the sidechains of Phe22, Arg23, Asn41, His60, Trp64 and Lys68 are shown as balland-stick models. (b) The crystallographic dimer. The interactions with

the symmetry-related molecule in the region of the sulphate-binding site. The symmetry-related molecule is labelled by inverted commas. (c) Close-up view of the sulphate-binding site and dimeric interactions. Labels for residues from a symmetry-related molecule are in inverted commas. Atoms are shown in standard colours.

Research Article Common type acylphosphatase structure Thunnissen et al.

towards the edge of the loops. The entrance to the phosphate-binding cradle is also relatively different in the two enzymes. The second catalytic residue in lmPTP, Asp129, is found near Arg18 in close proximity to the sulphate ion. This residue functions as proton donor for the tyrosine leaving group in the lmPTP reaction. In ACP, Asn41 can be found at an equivalent position. This conserved residue has been shown to be important for activity [25]; however, in the present structure Asn41 does not make any direct or indirect interactions to the sulphate ion, nor is it directly hydrogen bonded to Arg23. A second sulphate molecule is located on a depression on the surface of the protein bound to Lys32, His60 and Lys68. Glu29 also makes contact with this sulphate ion. The density shows that this site is not fully occupied. No residues involved in binding this sulphate ion have been showed to be essential for the activity of the enzyme or conserved within the acylphosphatase family (N Taddei, personal communication). Due to crystal packing, Lys68 makes contact to both sulphate ions (Fig. 6). Implications for the catalytic mechanism of acylphosphatases

In the present structure a well defined active-site structure is formed in the presence of anions such as sulphate Figure 7 Schematic picture of the proposed reaction mechanism.

75

and chloride; this site appears to be suitable for catalyzing phosphoryl transfer reactions. However, since chloride is a competitive inhibitor of the enzyme, it is very likely that the sulphate ion has been pushed somewhat out of the true phosphate-binding site by the chloride ion and that the new position of the sulphate ion is stabilized by interaction with Lys68 from a crystallographically-related molecule (Fig. 6c). Hence crystal contacts also affect the binding of the sulphate ion. It is difficult to estimate the extent of distortion in the conformation of loop 14–21 caused by crystal contacts; however, the similarity with the phosphate-binding loop of lmPTP suggests that such distortions are limited. Since Lys68 is not conserved within the acylphosphatase family and mutations on a positionally equivalent lysine (Lys67) did not affect activity (N Taddei, unpublished results), it is unlikely that the protein works as a dimer with this lysine residue in the active site. The functionally relevant phosphate-binding site is instead probably deeper down in the active-site pocket. Modelling studies show that it is possible to position the sulphate ion further down in the pocket, and that the phosphate group, in this position, would make many more interactions with the enzyme. It was found that the sulphate ion is best positioned when one of its oxygen atoms is put in the position of the present chloride. With a phosphate ion bound at this position, the water molecule in the current structure, bound to Asn41 and the backbone

76

Structure 1997, Vol 5 No 1

oxygen of Val17, would be well positioned to act as nucleophile in the reaction. Another possibility is that the phosphate moiety of the substrate binds even deeper down in the pocket and is occupying the position of this water molecule; however, this is much less likely since from the modelling it can be seen that the phosphate would make direct contact to the carbonyl oxygen of Val17, which is stereochemically unfavourable. As in protein tyrosine phosphatases, the key feature of the ACP active site is the phosphate-binding cradle. This structure appears to be particularly well suited for binding the phosphate moiety as well as stabilizing highly negatively charged reaction intermediates. Although the relationship of the phosphate-binding cradle to other catalytic elements is somewhat different in the active sites of ACP and lmPTP, they most probably have very similar roles in the reaction mechanisms of the two enzymes. We expect that the phosphate-binding loop and the position of Arg23 are very similar in both isoforms of ACP and that these elements have a similar function in phosphate binding and intermediate stabilization. This is confirmed by the lowering of the pKa of free benzoyl-phosphate (pKa = 4.8) upon binding to the enzyme (pKa = 3.7) [21]. Mutants of all polar/charged residues near the active site and mutants with inserted or deleted residues at the C terminus have been made (N Taddei, unpublished results and [23,24]). None of these mutants has any major catalytic effect except mutations of Arg23 and Asn41 [22,25]. We conclude that, of the residues with pKa’s of >12 and 7.9 implicated in the reaction mechanism, Arg23 is responsible for the pKa of >12. Since there are no available candidates for the residue with a pKa of 7.9 we argue that this residue is not directly involved in the reaction mechanism. Another likely candidate for a catalytic residue in our structure is Asn41. On the basis of the modelled sulphatebinding site, the water molecule interacting with Asn41 is well positioned to function as the attacking nucleophile. Furthermore, a Asn41→Asp mutant only shows some residual activity (1 % of wild type activity), indicating that the neutral charge of the asparagine may be important for the catalytic process (N Taddei, personal communication). There is a possibility that this residue could have a double role in the activation of the water nucleophile, both by orienting the water molecule and by serving as a base, accepting a proton from the water molecule. However, it is very uncommon for an asparagine to have the role of a general base. It is clear from the structure that any mutation of Asn41 would strongly affect the positioning of this water molecule, which might explain why all mutations made at this site result in no or little catalytic activity [25]. Therefore the role of Asn41 is most likely to position the nucleophilic water molecule.

The strongest hypothesis for the reaction mechanism is that the phosphate ion itself accepts the proton of the water nucleophile in the reaction cycle. Such a mechanism has been suggested to be the case for the GTPase catalyzed hydrolysis of GTP [29], which is similar in the sense that the GTPases also act in the absence of metals or protein-derived nucleophiles. In this mechanism Asn41, together with the phosphate-binding loop, can stabilize the transition intermediate (see Fig. 7 for the proposed reaction mechanism). The second step in the reaction mechanism will be a dissociation of the leaving acyl group and the phosphate group. Therefore the catalytic ability of the acylphosphatases might not involve any protein catalyzed general acid/base chemistry but would instead be mediated mainly through the phosphate-binding cradle and the correct positioning of the water nucleophile. This may explain their inability to dephosphorylate phosphate esters or anhydrides in which the leaving group is less favourable.

Biological implications Phosphatases play an important role in the regulation of many important physiological processes from signal transduction and glycogen metabolism to fatty acid synthesis. In recent years, structural information has become available for a wide range of these enzymes. These studies provide suggestions for the mechanisms of phosphatases as well as the structural basis for their substrate specificity. Acylphosphatase (ACP) is a small phosphomonohydrolase thought to participate in the regulation of glycolytic pathways and the synthesis of pyrimidines. Furthermore, it has a high specificity for the phosphoaspartyl-intermediate of Na+/K+ membrane-pumps and it is thought to affect the coupling between ATP hydrolysis and cation transport. Here we describe the X-ray crystal structure of common type ACP. The overall structure of the protein is similar to the structure of muscle form ACP, but the conformation of loop regions is markedly different. This is most probably caused by the binding of a sulphate and a chloride ion. The binding of these ions in the active site allows a detailed description of the active site. The phosphate-binding loop of ACP is remarkably similar to that in protein phosphotyrosine phosphatases. Based on the active site, a mechanism is proposed in which a water molecule bound to Asn41 acts as a nucleophile, while the phosphate group is bound to Arg23. The phosphate group acts as a base in the first step, subtracting a proton from the water molecule. The protein molecule itself has no nucleophile, which may explain ACP’s inability to dephosphorylate phosphate esters or anhydrides whose leaving groups are less favourable; this is probably the cause of the substrate specificity of acylphosphatase. This mechanism can serve as a model

Research Article Common type acylphosphatase structure Thunnissen et al.

for other acyl-phosphatases, such as the aspartyl-phosphatases involved in bacterial sporulation signal transduction [30].

Materials and methods Crystallization and native data collection The crystallization has been reported previously [31]. The protein was purified as previously described [9]. Crystals were grown by the hanging drop method at 4°C from 25 mM acetate buffer (pH 3.5) 30 % PEG4000 and 0.2 M ammonium sulphate. The space group of these crystals is C2 with a = 63.3 Å, b = 36.2 Å, c = 45.3 Å and b = 104.1° and one acylphosphatase molecule in the asymmetric unit. A complete native data set to a resolution of 1.8 Å and three heavy atom derivative data sets have been collected on a MAR-Research image plate using CuKa from a Siemens rotating anode equipped with a graphite monochromator. Data collection, reduction and scaling were carried out using the programs MARXDS and MARSCALE [32].

Molecular replacement studies Molecular replacement (MR) studies were started using structures from the NMR ensemble as starting search models. All MR studies were performed using AMoRe [33]. Several different approaches were tried, from using one structure out of the ensemble to summations of structures or using an average model based on all five structures from the NMR ensemble, or only a subset of the latter. The NMR structures were used as polyalanine or polyserine models. In an alternative approach, non-conserved residues were changed into alanines. The B factors of the models were either all set to 25 Å2 or increased B factors for the mobile loops were used. Combinations of all the above possibilities were tried.

77

decreased to 41 %, Rfree increased to 60 %. 10 % of the data, randomly chosen, was used for calculations of Rfree [34]. X-PLOR was run using different protocols and resolution limits, but none of these resulted in convergence of refinement. At that stage it was decided to solve the structure by multiple isomorphous replacement (MIR) methods. After the structure was solved by MIR it appeared that the rotation function solution was correct, but that the translation function had misaligned the b sheet by one strand.

Multiple isomorphous replacement In total, three isomorphous derivative data sets have been collected (Table 3). The compounds were potassium osmate (K2OsO4.2H2O), p-chloro-mercury-benzoyl sulphonic acid (PCMBS) and uranyl-acetate (UO2(CH3COO)2). All derivatives were prepared by dissolving the heavy-atom derivative in native mother liquor. Scaling and merging of the derivative data was performed using software from the software package BIOMOL (protein crystallography group, university of Groningen, the Netherlands) and the CCP4 suite of programs [35]. The K2OsO4 difference Patterson maps had low noise and a principal heavy atom site could easily be identified. A second site was later identified by inspection of a difference Fourier map. Difference Fourier maps were also used for the identification of heavy atom positions for the other two derivatives. Anomalous differences for PCMBS and UO2(CH3COO)2 were included in the phasing and were used to establish the correct hand of the structure. Refinement of the heavy atom parameters was carried out using MLPHARE [36]. Individual B factors for the different heavy atom sites were not refined but kept fixed at 20 Å2. The figure of merit was 0.54 for 2475 reflections between 30.0 Å and 2.8 Å resolution (Figure of merit = |F(hkl)best| / |F(hkl)| with F(hkl)best = ΣaP(a)Fhkl(a) / ΣaP(a)).

Density modification and refinement Using three structures out of the ensemble of five, superimposed as polyserine chains, a reasonable solution could be obtained from AMoRe (Table 2). Inspection of the packing of the molecules in the crystal did not show any major violations. However the subsequent refinement of this solution with X-PLOR did not converge and while the R factor Table 2 Results from AMoRe using NMR1, NMR4 and NMR5 together as polyserine model.

Rotation Translation FitFun

Proper solution peak*

Next highest*

13.8, -, (5) 0.22, 57 %, (1) 0.54, 48.4 %, (1)

0.18, -, (1) 0.16, 58.8 %, (2) 0.42, 52.1 %, (2)

Figures refer to correlation, R factor, peak number in (in parentheses) in that order. Where *R factor = {Σhkl ( || Fobs | – | Fcalc || ) / Σhkl ( | Fobs | )} × 100 %.

The program DM [37] was used to calculate a map which was modified by solvent flattening, histogram matching and skeletonizing. Phases from back transformation of this map were used to further refine the heavy-atom parameters [38], which resulted in an average change of ~10 % in occupancy and B factors and allowed the identification of a fifth uranyl position using difference Fourier maps. This process was carried out twice, at which point the refinement had converged and the mean figure of merit between 30.0 and 2.8 Å had improved from 0.54 to 0.59 (see Table 4 for more statistics for the heavy-atom derivatives). The skeletonized map calculated at 2.8 Å using the optimized heavyatom parameters showed the five stranded b sheet very clearly and one of the structures from the NMR ensemble was roughly positioned in the density. This model was used as a guide in the tracing of the common type ACP molecule. Residues 6 to 90 and the sidechains for about 80 residues could be built into the map using O [39]. The incomplete model was refined against the modified phases using X-PLOR [40] and a new map was calculated using SIGMAA weighting [41] to combine

Table 3 Data collection statistics.

Native KOs (10 mM 20h) UO2(CH3COO)2 (12 mM 1200 h) Hg(CH3COO)2 (10 mM 504 h)

a (Å)

b (Å)

c (Å)

b (Å)

Rmerge* (%)

Comp. (%)

Res. (Å)

Refl.

Uniq. Refl.

Rderiv† (%)

63.2 62.5

36.3 36.2

45.2 45.1

104.1 104.1

5.9 6.2

98.8 73

1.8 2.1

41285 5487

8949 3840

32.6

64.5

36.5

45.2

103.8

8.4

91

2.6

10753

2926

43.1

62.8

36.3

45.0

104.1

8.3

97.8

2.8

9016

2541

25.1

*Rmerge = {Σhkl Σrefl. | I (hkl,j) – I (hkl) | / Σhkl Σrefl. | I (hkl,j) |} × 100 %. †Rderiv = {Σhkl || Fderiv (hkl) | – | Fnative (hkl) / Σhkl | Fnative | } × 100%.

78

Structure 1997, Vol 5 No 1

Table 4 Heavy atom refinement parameters. Phasing power* acentric

Rcullis† acentric

Phasing power* centric

Rcullis† centric

Rcullis† anom

1.87 1.58 1.46

0.64 0.73 0.73

1.92 1.21 1.17

0.57 0.77 0.76

0.96 0.95

KOs UO2(CH3COO)2 Hg(CH3COO)2

*Phasing power = [Σn | FH |2 / Σn | E |2]½, with Σn | E |2 = Σ {| FPH | (obs) – | FPH | (calc)}2 †Rcullis = {Σhkl ( || FPH ± FP | – | FH (calc) || / Σhkl ( | FPH ± FP | )} × 100 %. Figure 8 Stereo diagram showing 2Fo–Fc density contoured at 1.1s. Part of the active site surrounding the sulphate molecule is shown. Figures created using a modified version of MOLSCRIPT [49] (modified by R Esnouf, personal communication) and Raster3D [50].

the partial model phases and the MIR phases between 2.8 and 15 Å. The new map allowed full tracing of the chain and all sidechains. The model was subsequently refined using TNT [42,43]. During refinement the resolution was gradually extended to 1.8 Å resolution. SIGMAAweighted 2|Fo|–|Fc|, |Fo|–|Fc| and OMIT maps [44] were used for model inspection and rebuilding using the program O. Water molecules were located in |Fo|–|Fc| maps. Peaks with a density > 2.5s above the mean were identified and used as possible water positions. Only water molecules at proper distances from protein atoms were kept. Those that developed B factors higher than 65 Å2 were removed from the model. In total 75 water molecules were identified. During refinement electron density that has been identified as two sulphate ions, one of them partially occupied, and a chloride ion also appeared. At an R factor of 17.9 % and an Rfree of 23.5 % for data between 15 and 1.8 Å it was decided to incorporate the 10 % ‘free’ data for a final round of refinement. The final R factor for all data between 15 and 1.8 Å is 17.0 %. The final model is of good quality as tested by PROCHECK [45] and there are no outliers in the Ramachandran plot [46] (90.5 % residues in core, 9.5 % of residues in additionally allowed regions). The rms deviations from the ideal values of Engh and Huber [47] are 0.010 Å for bonds, 1.7° for bond angles and 16.9° for dihedrals. The mean thermal B factor for all atoms is 26.4 Å2. The estimated rms positional error in the coordinates, derived from a SIGMAA plot, is 0.14 Å. Except for the first two N-terminal residues and the sidechains for some residues on the surface of the molecule for which the density is weak, the density for the model is good (Fig. 8).

Averaged NMR structure The NMR ensemble of horse muscle acylphosphatase available from the Protein Data Bank [48], code 1APS, contains five structures [11]. All of these structures were used to obtain an average structure using the program Moleman (GJ Kleywegt & TA Jones, unpublished results); this structure was then regularized using geometry refinement by the TNT package. The averaged structure was compared to the individual structures of the NMR ensemble (see Table 1) and the X-ray structure

of the common type acylphosphatase by using the program SUPPOS from the BIOMOL package.

Accession numbers The coordinates have been submitted to the protein databank at Brookhaven (ID code 2acy).

Acknowledgements We thank Dr Robert Esnouf for modifications of MOLSCRIPT and Drs Cristina Cecchi and Alessandro Pieri for protein purification. This work has been supported by a grant from the Swedish Cancer Society for MT and PN and by grants from the Italian MURST (fondi 40 %), CNR (target project structural biology) and the Italian Association for Cancer research (AIRC) for NT, GL and GR.

References 1. Stefani, M. & Ramponi, G. (1995). Acylphosphate phosphohydrolases. Life Chemistry Reports 12, 271–301. 2. Liguri, G., et al., & Ramponi, G. (1994). Expression of human acylphosphatase in Escherichia coli affects intracellular calcium levels. Biochem. Mol. Biol. Int. 34, 109–117. 3. Nassi, P., Nediani, C., Liguri, G., Taddei, N. & Ramponi, G. (1991). Effects of acylphosphatase on the activity of erythrocyte membrane Ca2+ pump. J. Biol. Chem. 266, 10867–10871. 4. Nassi, P., Marchetti, E., Nediani, C., Liguri, G. & Ramponi, G. (1993). Acylphosphatase induced modifications in the functional properties of erythrocyte membrane sodium pump. Biochim. Biophys. Acta 1147, 19–26. 5. Berti, A., et al., & Ramponi, G. (1988). Effect of exogenously added acylphosphatase on inositol lipid metabolism in human platelets. FEBS Lett. 235, 229–232. 6. Liguri, G., et al., & Ramponi, G. (1986). A new acylphosphatase isoenzyme from human erythrocytes: purification, characterisation and primary structure. Biochemistry 25, 8089–8094. 7. Fujita, H., Mizuno, Y. & Shiokawa, H. (1987). Purification and properties of porcine testis acylphosphatase. J. Biochem. 102, 1405–1414. 8. Mizuno, Y., et al., & Shiokawa, H. (1990). Distribution and classification of acylphosphatase isozymes. Arch. Biochem. Biophys. 278, 437–443.

Research Article Common type acylphosphatase structure Thunnissen et al.

9. Pazzagli, L., Cappugi, G., Camici, G., Manao, G. & Ramponi, G. (1993). Bovine testis acylphosphatase: purification and amino acid sequence. J. Protein Chem. 12, 593–601. 10. Saudek, V., Williams, R.J.P. & Ramponi, G. (1989). The structure and properties of horse muscle acylphosphatase in solution. Mobility of antigenic and active regions. FEBS Lett. 242, 225–232. 11. Pastore, A., Saudek, V., Ramponi, G. & Williams, R.J.P. (1992). Threedimensional structure of acylphosphatase. Refinement and structure analysis. J. Mol. Biol. 224, 427–440. 12. Adman, F.T., Sicker, L.C. & Jensen, L.H. (1972). The structure of a bacterial ferrodoxin. J. Biol. Chem. 248, 3987–3996. 13. Coll, M., Guasch, A., Aviles, F.X. & Huber, R. (1991). Threedimensional structure of porcine procarboxy-peptidase B: a structural basis for its inactivity. EMBO J. 10, 1–9. 14. Vendrell, J., Billeter, M., Wider, G., Aviles, F.X. & Wüthrich, K. (1991). The NMR structure of the activation domain isolated from porcine procarboxy-peptidase B. EMBO J. 10, 11–15. 15. Nagai, K., Oubridge, C., Jessen, T.H., Li, J. & Evans, P.R. (1990). Crystal structure of the RNA-binding protein of the U1 small nuclear ribonucleoprotein A. Nature 348, 515–520. 16. Leijonmarck, M. & Liljas, A. (1987). Structure of the C-terminal domain of the ribosomal protein L7/L12 from Escherichia coli at 1.7 Å. J. Mol. Biol. 195, 555–580. 17. Wilson, K.S., Appelt, K., Badger, J., Tanaka, I. & White, S.W. (1986). Crystal structure of a prokaryotic ribosomal protein. Proc. Natl. Acad. Sci. USA 83, 7251–7255. 18. Herzberg, O., et al., & Kapadia, G. (1992). Structure of the histidine containing phosphocarrier protein HPr from Bacillus subtilis at 2.0 Å resolution. Proc. Natl. Acad. Sci. USA 89, 2499–2503. 19. Jia, Z., Vandonselaar, M., Quail, J.W. & Delbaere, L.T.J. (1993). Active centre torsion-angle strain revealed in 1.6 Å resolution structure of histidine containing phosphocarrier protein. Nature 361, 94–97. 20. Swindells, M.B., Orengo, C.A., Jones, D.T., Pearl, L.H. & Thornton, J.M. (1993). Recurrence of a binding motif? Nature 362, 299. 21. Satchell, D.P.N., Spencer, N. & White, G.F. (1972). Kinetic studies with muscle acylphosphatase. Biochim. Biophys. Acta 268, 233–248. 22. Taddei, N., et al., & Ramponi, G. (1994). Arginine-23 is involved in the catalytic site of muscle acylphosphatase. Biochim. Biophys. Acta 1208, 75–80. 23. Taddei, N., et al., & Ramponi, G. (1995). Properties of N terminus truncated and C terminus mutated muscle acylphosphatase. FEBS Lett. 362, 175–179. 24. Taddei, N., et al., & Ramponi, G. (1996). C-terminal region contributes to muscle acylphosphatase three-dimensional structure stabilization. FEBS Lett. 384, 172–176. 25. Taddei, N., et al., & Ramponi, G.(1996). Looking for residues involved in muscle acylphosphatase catalytic mechanism and structural stabilization: the role of Asn41, Thr42 and Thr46. Biochemistry 35, 7077–7083. 26. Su, X.-D., Taddei, N., Stefani, M., Ramponi, G. & Nordlund, P. (1994). The crystal structure of a low molecular weight phosphotyrosine protein phosphatase. Nature 370, 575–578. 27. Engh, R.A., et al., & Oschkinat, H. (1993). Conformational variability of chicken cystatin. Comparison of structures determined by X-ray diffraction and NMR spectroscopy. J. Mol. Biol. 234, 1060–1069. 28. Gallagher, T., Alexander, P., Bryan, P. & Gilliland, G. (1994). Two crystal structures of the B1 immunoglobulin-binding domain of streptococcal protein G and comparison with NMR. Biochemistry 33, 4721–4729. 29. Schweins, T., et al., & Wittinghofer, A. (1995). Substrate-assisted catalysis as a mechanism for GTP hydrolysis of p21ras and other GTPbinding proteins. Nat. Struc. Biol. 2, 36–44. 30. Perego, M., et al., & Hoch, J.A. (1995). Multiple protein aspartate phosphatases provide a mechanism for the integration of diverse signals in the control in B. subtilis sporulation. Cell 79, 1047–1055. 31. Thunnissen, M.M.G.M., et al., & Nordlund, P. (1995). Crystallization and preliminary X-ray analysis of the ‘common type’ acylphosphatase. FEBS Lett. 364, 243–244. 32. Kabsch, W. (1993). Automatic processing of rotation diffraction data from crystals of initially unknown symmetry and cell constants. J. Appl. Cryst. 26, 795–800. 33. Navaza, J. (1994). AMoRe: an automated package for molecular replacement. Acta Cryst. A 50, 795–800. 34. Brünger, A.T. (1992). Free R value: a novel statistical quantity for assessing the accuracy of crystal structures. Nature 355, 472–475. 35. Collaborative Computing Project No. 4 (1994). The CCP4 suite: programs for protein crystallography. Acta Cryst. D 50, 760–763.

79

36. Otwinowski, Z. (1991). Maximum likelihood refinement of heavy atom parameters. Isomorphous replacement anomalous scattering. In Proceedings of the CCP4 study weekend. pp. 80–85, SERC Daresbury Laboratory, Warrington, UK. 37. Cowtan, K.D. & Main, P. (1996). Phase combination and cross validation in iterated density modification calculations. Acta Cryst. D 52, 43—48. 38. Cura, V., Krisnaswamy, S. & Podjarny, A.D. (1992). Heavy atom refinement against solvent flattened phases. Acta Cryst. A 48, 756–764. 39. Jones, T.A., Zou, J.Y., Cowan, S.W. & Kjeldgaard, M. (1991). Improved methods for building protein models in electon density maps and the location of errors in these models. Acta Cryst. A 47, 110–119. 40. Brünger, A.T., Kuriyan, J. & Karplus, M. (1987). Crystallographic R factor refinement by molecular dynamics. Science 235, 458–460. 41. Read, R. (1986). Improved Fourier coefficients for maps using phases from partial structures with errors. Acta Cryst. A 42, 140–149. 42. Tronrud, D.E., ten Eyk, L.F. & Matthews, B.W. (1987). An efficient general purpose least squares refinement program for macromolecular structures. Acta Cryst. A 43, 481–501. 43. Tronrud, D.E. (1992). Conjugate-direction minimisation: an improved method for the refinement of macromolecules. Acta Cryst. A 48, 912–916. 44. Bhat, T.N. (1988). Calculation of an OMITMAP. J. Appl. Cryst. 21, 279–281. 45. Laskowski, R.A., MacArthur, M.W., Moss, D.S. & Thornton, J.M. (1993). PROCHECK: a program to check the stereochemical quality of protein structures. J. Appl. Cryst. 26, 283–291. 46. Ramakrishnan, C. & Ramachandran, G.N. (1965). Stereochemical criteria of polypeptide and protein chain conformations II. Allowed conformations for a pair of peptide units. Biophys. J. 5, 909–933. 47. Engh, R.A. & Huber, R. (1991). Accurate bond and angle parameters for protein structure refinement. Acta Cryst. A 47, 392–400. 48. Bernstein, F.C., et al., & Tasumi, M.(1977). The Protein Databank: a computer-based archival file for macromolecular structures. J. Mol. Biol. 112, 535–542. 49. Kraulis, P.J. (1991). MOLSCRIPT: a program to produce both detailed and schematic plots of proteins. J. Appl. Cryst. 24, 946–950. 50. Merrit, E.A. & Murphy, M.E.P. (1994). Raster3D version 2.0: a program for photorealistic molecular graphics. Acta Cryst. D 50, 869–873.