The X-ray crystal structures of Yersinia tyrosine phosphatase with bound tungstate and nitrate. Mechanistic implications.

X-ray crystal structures of the Yersinia tyrosine phosphatase (PTPase) in complex with tungstate and nitrate have been solved to 2.4-Å resolution. Tetrahedral tungstate, WO2−4, is a competitive inhibitor of the enzyme and is isosteric with the substrate and product of the catalyzed reaction. Planar nitrate, NO−3, is isosteric with the PO3 moiety of a phosphotransfer transition state. The crystal structures of the Yersinia PTPase with and without ligands, together with biochemical data, permit modeling of key steps along the reaction pathway. These energy-minimized models are consistent with a general acid-catalyzed, in-line displacement of the phosphate moiety to Cys403 on the enzyme, followed by attack by a nucleophilic water molecule to release orthophosphate. This nucleophilic water molecule is identified in the crystal structure of the nitrate complex. The active site structure of the PTPase is compared to alkaline phosphatase, which employs a similar phosphomonoester hydrolysis mechanism. Both enzymes must stabilize charges at the nucleophile, the PO3 moiety of the transition state, and the leaving group. Both an associative (bond formation preceding bond cleavage) and a dissociative (bond cleavage preceding bond formation) mechanism were modeled, but a dissociative-like mechanism is favored for steric and chemical reasons. Since nearly all of the 47 invariant or highly conserved residues of the PTPase domain are clustered at the active site, we suggest that the mechanism postulated for the Yersinia enzyme is applicable to all the PTPases.

Phosphorylation of tyrosine residues of intracellular or membrane proteins is a fundamental cellular signal for regulating cell growth, differentiation, and proliferation (1). The levels of phosphotyrosine in the cell are governed by the competing actions of protein tyrosine kinases and protein tyrosine phosphatases (PTPases, 1 EC 3.1.3.48).
A variety of receptor-like and nonreceptor-like PTPases is found in all eukaryotic cells. Presumably, this molecular diversity provides specific PTPases for different signal transduction pathways that can occur simultaneously in the cell. Although few PTPases have been assigned specific roles in signaling pathways, much is known about PTPase biochemistry, largely from studies of a homologous enzyme identified in Yersinia (2). The Yersinia PTPase, encoded on a virulence plasmid, is required for pathogenicity in this bacterium (3).
The PTPase reaction proceeds in two steps. First, the phosphate is transferred from tyrosine (R) to a functional group on the enzyme (I) and then the phosphoenzyme intermediate is hydrolyzed (II) (4), in a manner similar to the well studied alkaline phosphatase (5) (Reaction 1).
Mutagenesis and biochemical studies have identified several invariant residues in the PTPase signature sequence, (I/V)H-CXAGXXR(S/T)G, important for phosphotyrosine hydrolysis. The phosphate moiety is transferred to the thiol of Cys 403 (4), 2 the pK a of which is lowered 2.7 pH units by His 402 (6) and 0.6 pH units by Thr 410 (7). Arg 409 is critical for ligand binding and catalysis (8). Of invariant residues outside of the signature sequence, Asp 356 was identified as a putative general acid and Glu 290 was identified as a putative general base on the basis of mutagenesis and pH rate profile experiments (9).
The PTPases, unlike alkaline phosphatase, are very substrate-specific, selecting phosphotyrosines over the other phosphorylated residues (1). In addition, the K m and k cat are greatly improved for the PTPases when using a phosphotyrosine-containing peptide as opposed to phosphotyrosine by itself (10).
Our laboratory recently reported the x-ray crystal structure of the unliganded Yersinia PTPase to 2.5-Å resolution and the complex with tungstate, WO 4 2Ϫ , to 2.6-Å resolution (11). We have also solved the structure of a Cys 403 3 Ser mutant complexed with sulfate, SO 4 2Ϫ (12). The PTPase is an ␣ϩ␤ protein with an eight-stranded, mixed ␤-sheet. The central feature of the structure is a strand-loop-helix motif where the loop contains the PTPase signature sequence, including the Cys 403 nucleophile. Oxyanions bind within this loop, which we have termed the phosphate binding or P-loop. Three of the anion oxygens are coordinated by main chain amides of the P-loop while the fourth, apical, anion oxygen (O-4) points away from the P-loop. Asp 356 , the putative general acid, is located on a movable loop, termed the WpD-loop (for the conserved Trp-Pro-Asp sequence). Upon binding of an oxyanion, the carboxyl group of Asp 356 shifts 8 Å toward the active site and this apical oxygen (12).
To better understand the roles of the conserved residues and the specificity of this enzyme, we have refined the tungstate complex structure to a higher resolution, 2.4 Å. Tungstate, a tight competitive inhibitor (K D ϭ 61 M at pH 7.0; Ref. 8), is tetrahedral, like the phosphate moieties of the substrate (phosphotyrosine) and product (orthophosphate). One of the oxygens of the orthophosphate product is derived from a water molecule. Thus, in any crystal structure of the PTPase in complex with a tetrahedral oxyanion, one of the oxygens of the ligand will be analogous to the water-derived oxygen of the orthophosphate product. In the hopes of observing the nucleophilic water molecule prior to attack, crystals were grown in the presence of nitrate, NO 3 Ϫ , a trigonal planar oxyanion. Nitrate is also isosteric with the planar PO 3 moiety expected in a phosphotransfer transition state.
Energy-minimized models of reaction intermediates that are consistent with the known biochemistry were derived from these crystal structures. These models highlight residues that are likely to be of key importance in the hydrolysis. The models also place certain limits on the nature of the reaction mechanism.
Diffraction intensities were collected on a SDMS multiwire area detector system mounted on a Rigaku RU-200 rotating anode generator (50 kV, 100 mA), integrated (14), and scaled (15). The PTPase-WO 4 amplitudes (and those from a mercury derivative) were used to phase data from isomorphous crystals of the Yersinia PTPase Cys 403 3 Ser mutant containing bound sulfate, PTPase-SO 4 (11). After much of the Cys 403 3 Ser structure had been traced, a tetrahedral tungstate ion was fit into a difference electron density map and the model (residues 186 -468) was further refined with X-PLOR (16) against the PTPase-WO 4 amplitude data. The first 23 residues of the polypeptide chain are not visible and are presumed disordered. Force field interactions between the tungstate and Cys 403 were turned off to allow close approach of these two groups if necessary. An overall anisotropic B-factor was applied. One residue (Cys 418 ) has been modeled with two alternate conformations. The PTPase-WO 4 structure currently has a crystallographic R-factor of 17.9% for all non-zero reflections between 7 and 2.4 Å (Table I). One hundred forty-three crystallographically observed water molecules and a sulfate molecule have been added by inspection of 2F obs Ϫ F calc and F obs Ϫ F calc electron density maps.
PTPase-NO 3 was phased on the protein atoms from the PTPase-WO 4 structure. The final structure, including one nitrate and 105 water molecules, was refined to a final R-factor of 18.1% for all non-zero data from 7 to 2.4 Å.
Atomic Coordinates for PTPase-WO 4 and PTPase-NO 3 have been submitted to the Brookhaven Protein Data Bank (17) (entry codes 1YTW and 1YTN, respectively). The coordinates for the unliganded enzyme and for PTPase-SO 4 are from Protein Data Bank entries 1YPT (11) and 1YTS (12), respectively.
Modeling of Reaction Intermediates-All charge, bond-length, and angle parameters for the ligands were derived from reported small molecule structures (18 -24) or published quantum mechanical calculations (25), or were extrapolated from existing amino acid parameters (26) (Table II).
An atomic model for the substrate complex was constructed starting with the PTPase-WO 4 protein atom coordinates. The phosphohexapeptide Asp-Ala-Asp-Glu-Tyr(P)-Leu (10) was modeled initially in an extended conformation. The phosphate moiety, assumed deprotonated based on the pH profile for k cat /K m (27), was modeled on top of the tungstate coordinates, and the tyrosine was positioned within a clear active site cavity. Grooves in the protein surface were graphically highlighted by an algorithm that subtracts the molecular volume of the enzyme from the convex hull of the enzyme. The substrate peptide was then manually docked into grooves near the active site, taking into account possible electrostatic interactions. A 25-Å sphere of water (about 1000 water molecules) centered around the phosphate moiety was added. The model, including explicit polar hydrogens and a unified atom model for non-polar hydrogens, was subjected to simulated annealing from 1000 K to 300 K and then energy minimized using X-PLOR (16). Only the ligand atoms, water molecules within 15 Å of the active site, Arg 409 , Cys 403 , Asp 356 , and all explicit hydrogens were allowed to move, although all atoms were included in the force field.
Models for the phosphotransfer reactions and the phosphocysteine intermediate were generated similarly, constrained by the nucleophilephosphorus distances given in Table III. Throughout, the active site nucleophile (Cys 403 ) was assumed to be unprotonated (6). The putative general acid, Asp 356 , was modeled as protonated for the second transition state, but at the first transition state the proton was placed on the leaving group.

Anion Binding
Nitrate binds within the P-loop (Fig. 1) and effects the closed conformation of the Asp 356 -containing WpD-loop, as do tung- REACTION 1 13 19 a Measured for all non-hydrogen atoms against X-PLOR target values (26). RMSD, root mean square deviation.
b Percent of residues that can be assigned to one of the rotamers identified by Ponder and Richards within 3 standard deviations (49).
c Number of non-glycine residues that fall outside the allowed regions of a Ramachandranplot.
d Mean B factor, averaging over all non-hydrogen protein atoms.
state (11) and sulfate (12), supporting the hypothesis that oxyanion interactions with Arg 409 precipitate the closing of the WpD-loop (12). The orientation of the oxyanion is the same in each oxyanion-bound crystal structure and results in up to 11 hydrogen bonds between the P-loop and the anion oxygens, along with hydrogen bonds to crystallographically observed active site water molecules (Table IV). Three anion oxygens are coordinated by five of the seven amide nitrogens from the P-loop and also by the side chain of Arg 409 . Arg 409 is in turn coordinated in a bidentate salt-bridge with Glu 290 , the putative general base (9). The apical anion oxygen, O-4, in the PTPase-SO 4 and PTPase-WO 4 complexes points away from the nucleophilic cysteine and the P-loop, and, in PTPase-WO 4 , is coordinated by an invariant glutamine residue, Gln 446 . The anion in PTPase-SO 4 is 0.6 Å deeper in the active site than in PTPase-WO 4 , primarily because the Ser 403 hydroxyl in PTPase-SO 4 is smaller than the Cys 403 thiolate in PTPase-

Coordination of the Nucleophile
Although five of the seven amide hydrogens are directed toward the bound anion, the two remaining P-loop amides point to the nucleophilic Cys 403 (Table V). The amide nitrogens of Gly 406 and Thr 410 and the conserved side chain hydroxyl of Thr 410 make hydrogen bonds with the nucleophile. The amide nitrogen of Arg 404 is within hydrogen bonding distance of both the bound anion and the nucleophile.

Active Site Water Molecules
Several water molecules have been identified in the active site (Table VI). WAT1 is observed in all three anion complexes and has an average crystallographic temperature factor of 16 Å 2 . WAT1 is buried under the ligand and is well coordinated Ϫ0.85 Ϫ0.85

TABLE III Model restraint distances
These numbers reflect a van der Waals radius of 1.4 Å for oxygen, 1.9 Å for phosphorus, and 1.85 Å for sulfur. A bond length of 1.64 Å for an RO-P bond in a dianionic phenylphosphate, and 2.13 Å for a thiophosphate S-P bond were obtained from small molecule structures (23).
Step with four tetrahedrally positioned hydrogen bonds. In PTPase-NO 3 , Asp 356 approaches within 3.1 Å of WAT1. A second, poorly ordered water molecule, WAT2, sits above the oxyanion in the active site. In PTPase-SO 4 , WAT2 is 3.0 Å from Gln 446 ; however, in the other two structures WAT2 makes good hydrogen bonds only with other water molecules.
A unique active site water molecule is present in PTPase-NO 3 (Fig. 1). Nitrate lacks the apical oxygen present in tungstate and sulfate. Instead, in PTPase-NO 3 WAT3 sits 4.2 Å directly above the nitrate, perfectly in line with the sulfur of the nucleophile, Cys 403 (O-N-S angle of 168°, O-S distance of 7.3 Å). WAT3 is well ordered (B ϭ 17 Å 2 ) and is coordinated principally by the side chains of Gln 446 and the Gln 357 . Although Gln 446 is invariant, Gln 357 is found only in the Yersinia PTPase sequence.

Second Anion Binding Site
A second anion binding site, far from the active site, contains a sulfate in the PTPase-WO 4 and PTPase-SO 4 complex structures, coordinated by the side chains of Arg 278 and Ser 388 and the main chain nitrogen of Ser 389 (none of these residues is   highly conserved). PTPase-NO 3 , grown without sulfate, has a water molecule at this alternate oxyanion binding site.

Model of the Michaelis Complex
A phosphotyrosine-containing hexapeptide manually docked into the active site pocket and nearby grooves was subjected to energy minimization, as described above (Fig. 2). The position of the phosphotyrosine side chain is well defined by residues lining the active site. Specifically, the phosphotyrosine ring packs against the hydrophobic face of Phe 229 (Fig. 3A). Ile 232 and Ile 443 also contribute to a hydrophobic binding site, while Gln 446 and Gln 357 define the other side of the phosphotyrosine pocket. This modeled peptide-Yersinia PTPase complex is consistent with the recently reported crystal structure of substrate bound to the human PTPase, PTP1B (Ref. 28, and see "Discussion"). WAT1 is the only water molecule that can fit in the active site binding pocket with the substrate present. Asp 356 , modeled as protonated, moved to within hydrogen bonding distance (2.7 Å) of the ester oxygen during energy minimization.
A phosphoserine-containing peptide modeled in this pocket renders the phosphate group 7 Å away from the nucleophile. Bringing the phosphoserine into contact with the nucleophile requires severe distortion of the geometry of the phosphoserine and steric clashes with the side chain of Phe 229 and the other residues lining this pocket.

Model of the First Phosphotransfer
The geometry of the PTPase active site dictates an in-line displacement mechanism, in which the phosphorous passes through a trigonal bipyramidal state (Fig. 3B). Two models were generated of the midpoint of the transfer of the PO 3 moiety from oxygen of the phosphotyrosine to the Cys 403 sulfur of the PTPase, one reflecting an associative reaction and one for a dissociative reaction. In an associative reaction, the bond from the incoming nucleophile is formed prior to bond breakage to the leaving group. In a dissociative reaction, the bond to the leaving group is broken before the bond to the nucleophile is made. The different pathways are distinguished by the bond orders to the phosphorus, which are reflected in the bond lengths and the location of the three negative charges (Ref. 29 and Table VII).
Of special interest is the interaction of the protein and substrate at three key places: the attacking nucleophile, the PO 3 group, and the leaving group: Stabilization of the Thiolate-Cys 403 must be ionized to participate in the phosphotransfer reaction, and its apparent pK a is 4.7 (6). The crystal structure suggests that the pK a of Cys 403 is lowered by hydrogen bonds to specific P-loop nitrogens and Thr 410 and by electrostatic interactions with His 402 (Table V) (11). His 402 is on the opposite face of a ␤-strand from Cys 403 . Although the N␦1 of His 402 is 6.5 Å from the S␥ of Cys 403 , the histidine is buried deep within the protein, so that its electrostatic effect is not mitigated by bulk solvent. The other nearby charged residues that might modulate the pK a of Cys 403 are Arg 409 (4.3 Å to the N⑀), Asp 356 (6.9 Å to the O␦) and Glu 290 (7.3 Å to the O⑀). Arg 409 and Glu 290 are both slightly solventaccessible, while Asp 356 is well exposed to bulk solvent. (Table IV). Arg 409 makes bidentate hydrogen bonds to the PO 3 moiety.

Stabilization of the PO 3 Moiety-The PO 3 moiety in the models maintains the hydrogen bonds exhibited by the three lower oxygens (O-1, O-2, and O-3) of the oxyanions
Stabilization of the Leaving Group-During the molecular modeling, Asp 356 moved to within 3.0 Å of the leaving group, in a position where it could act as a general acid, donating a hydrogen to the leaving group.

Model of the Second Phosphotransfer
The first phosphotransfer generates a phosphocysteine intermediate (Fig. 3C), which is then hydrolyzed (Fig. 3D) to form free orthophosphate. In modeling the configuration of the active site just before hydrolysis of the phosphocysteine, one water molecule was restrained to be at a van der Waals contact distance from the phosphorus atom. Although Asp 356 donated its proton in Step I, Asp 356 was modeled as protonated at this step since it is on the surface of the enzyme, presumably in equilibrium with bulk solvent.
After energy minimization, the restrained nucleophilic water molecule, WAT Nuc , is coordinated by the phosphocysteine oxygens in addition to the side chains of Gln 446 and Asp 356 (Fig.   FIG. 2. Energy-minimized atomic model of a substrate hexapeptide, Asp-Ala-Asp-Glu-Tyr(P)-Leu, bound to the Yersinia PTPase. The molecular surface of the protein atoms from the PTPase-WO 4 crystal structure is shown in cyan. This surface has been sliced (dark blue surface) to reveal a deep active site pocket into which a model of a substrate hexapeptide was manually docked and subsequently energy minimized. The hexapeptide carbon, nitrogen, oxygen, and phosphorus atoms are shown in yellow, blue, red, and yellow, respectively. The peptide substrate runs from N terminus on the left of the figure to C terminus on the right. 3C). It also forms hydrogen bonds with water molecules in the positions identified in the crystal structures as WAT1 and WAT2. WAT Nuc is in position to take part in a second in-line displacement reaction and is just 0.6 Å from the apical oxygen (O-4) of tungstate in PTPase-WO 4 and 2.2 Å from the position seen for WAT3 in the PTPase-NO 3 structure. We postulate that WAT3 is analogous to WAT Nuc , but WAT3 cannot move closer to the nitrate without colliding with the nitrogen, which is not available for hydrogen bonding (Fig. 1).
For the transfer of the phosphate moiety from the Cys 403 to WAT Nuc , both an associative and a dissociative pathway were modeled (Table VII), although the dissociative-like mechanism is depicted in Fig. 3D. As in Step I, charges on the thiol and PO 3 group are balanced by specific hydrogen bonds and ionic interactions (Tables IV and V).

Analysis of Modeled Substrate Binding-The hexapeptide
Asp-Ala-Asp-Glu-Tyr(P)-Leu is an effective substrate for the Yersinia PTPase, with a k cat of 1381 s Ϫ1 and a K m of 100 M (10). Although generated independently, the modeled peptide-PTPase coordinates used here are similar to those recently FIG. 3. Energy-minimized atomic models of the PTPase reaction. Although an entire hexapeptide was used in all energy calculations, only the phosphotyrosine residue is shown here for clarity. The phosphate moiety is shown in magenta, while the rest of the tyrosine is in brown. The tyrosine ring is surrounded by conserved residues. A, the non-covalent Michaelis complex. Asp 356 is positioned to donate a hydrogen to the ester oxygen. B, the midpoint of a dissociative-type transfer reaction from phosphotyrosine to Cys 403 . Hydrogen bonds from the PO 3 phosphate moiety, the Asp 356 and WAT1 are shown in magenta, red, and cyan, respectively. The thin gold bonds indicate the very low bond order bonds to the PO 3 moiety. C, the phosphocysteine intermediate. Potential hydrogen bonds from the nucleophilic water molecule to the phosphate moiety are shown as black lines, although only one of these hydrogen bonds can be formed at any given time. The modeled nucleophilic water molecule is close to the position for WAT3 in PTPase-NO 3 . D, the midpoint of a dissociative-type transfer reaction from the phosphocysteine to water. A hydrogen bond from one PO 3 oxygen to the main chain of Arg 404 has been omitted for clarity. Asp 356 , Gln 446 , or the phosphate group itself could serve as a general base, receiving a proton from the WAT Nuc .
reported for the crystal structure of the same hexapeptide bound to the human PTPase, PTP1B (28). That crystal structure, like the model here, reveals a deep active site pocket formed in part by the closing of the WpD-loop, which traps an active site water molecule (WAT1). The WpD-loop also appears to trap the phosphate group, although modeling suggests that the dephosphorylated peptide product of step I can be released without opening the WpD-loop.
In both the model presented here and the crystal structure of the peptide with PTP1B, the tyrosine ring of the substrate is positioned near the conserved phenyl ring of Phe 229 (Tyr 46 in PTP1B), with acidic residues of the substrate coordinated by Arg 230 (Arg 47 in PTP1B). Val 49 of PTP1B and Ile 232 of this model make hydrophobic contacts with the leucine side chain of the substrate peptide. Asp 181 of PTP1B, like Asp 356 here, approaches close to hydrogen bond distance to the ester oxygen of the scissile bond, and Arg 221 of PTP1B, like Arg 409 here, forms a bidentate hydrogen bond with two of the phosphate oxygens. Differences between this model and the PTP1B crystal structure are due to either limitations in the modeling or the extensive sequence differences between the Yersinia PTPase and PTP1B (only 15% sequence identity for the residues in the Yersinia crystal structure). Crystallographic studies of this hexapeptide bound to the Yersinia PTPase confirm the general interactions outlined above. 3 Associative Versus Dissociative-Although only an associative and a dissociative mechanism were modeled, these two schemes can be viewed as two extremes on a continuum of possible pathways. Between the associative and dissociative pathways would be a "half-bond" mechanism in which the degree of bond breakage to the leaving group is exactly balanced by bond formation to the incoming nucleophile. In this scenario, the PO 3 group maintains its charge of Ϫ2 and a bond order of 4/3 to the equatorial oxygens (29).
In solution, displacements of phosphomonoesters proceed by a dissociative-like mechanism, wherein the transition state approaches a metaphosphate configuration (30, 31). Free metaphosphate (PO 3 Ϫ ) has only been observed in vacuum and is certainly not a discrete intermediate (32). However, in the transition state, the bond order to the leaving group and to the incoming nucleophile are very low.
An associative-type reaction had long been favored for enzyme-catalyzed displacement reactions involving phosphate monoesters (33), because it is apparent that an enzyme could catalyze such a reaction by stabilizing the increased negative charge of the PO 3 group in the transition state. However, since an associative pathway for phosphate monoester hydrolysis is not observed in solution, it would require greater transition state stabilization for an enzyme to select this pathway over a dissociative one. Indeed, the most direct evidence to date suggests that enzyme-catalyzed phosphomonoester hydrolysis is more dissociative-like (29,34,35). Alkaline phosphatase and hexokinase both show an inverse 18 O-secondary isotope effect, indicating an increase in bond order to the equatorial oxygens, consistent with a dissociative-like mechanism (see Table VII).
A phosphate monoesterase can catalyze a dissociative reaction by balancing three negative charges: at the leaving group, the PO 3 group, and the attacking nucleophile (36,37). In the analogous alkaline phosphatase catalyzed reaction, these charges are coordinated with Zn 2ϩ ions for the serine nucleophile and the leaving group and an arginine for the PO 3 moiety (Fig. 4) (38). As shown in the model, the Yersinia PTPase could use specific charged and uncharged groups on the enzyme to stabilize these three negative charges at places structurally analogous to the arginine and Zn 2ϩ in alkaline phosphatase.
The use of an arginine to ligate the phosphate oxygens would seem to favor an associative pathway, since the associative pathway requires a transient increase in charge on the phosphate oxygens. However, in the case of the Yersinia PTPase at least, the charge on Arg 409 appears to be insufficient for catalytic activity. Changing Arg 409 to alanine reduces k cat by 10,000, and increases K m over an order of magnitude (8). The positive charge alone is sufficient for binding, since the K m of an Arg 409 3 Lys mutant is restored to its wild type value. However, the k cat of the Arg 409 3 Lys mutant is nearly identical to that for the Arg 409 3 Ala mutant. Thus, the geometry of the guanidinium group must be essential for turnover.  Single-bond bond lengths for P-S and P-O were derived from small molecule crystal structures (see "Experimental Procedures"). Bond lengths and bond orders are related by Pauling's rule: R(n) ϭ R(1) Ϫ0.3 ln(n), where R(n) is bond length for a bond order of n and R(1) is the bond length for a single bond (50). The dissociative pathway values assume that the oxygen-sulfur distance remains constant, and the midpoint was defined as a movement of the phosphorus halfway between van der Waals contact distance and a covalent bond. The midpoint of the "half-bond" mechanism assumes a bond order of 0.5 from the phosphorus to the oxygen and sulfur, while the associative pathway assumes a full single bond between the phosphorus and each of its ligands. A requirement for a planar guanidinium group is consistent with a dissociative-like transition state. Since the O-P bonds in metaphosphate have more double-bond character (5/3 each) than in orthophosphate (4/3 each), the planar geometry of the arginine guanidinium group is ideal for donating hydrogen bonds to the planar lone-pairs of the metaphosphate oxygens. In other words, by donating hydrogen bonds in the plane of the PO 3 group, Arg 409 could stabilize sp 2 hybridization of the PO 3 oxygens, which would enhance the double bond character of the O-P bonds, thus lowering the energy for a metaphosphate-like transition state.
In a dissociative mechanism transition state, the donor oxygen and acceptor sulfur can remain 5.4 Å apart, as in the Michaelis complex. In the associative mechanism, however, these atoms must approach within 3.7 Å of each other, requiring substantial movement of the entire phosphotyrosine side chain and the substrate peptide. This compression pulls the ester oxygen slightly away from Asp 356 , the general acid, and pulls the nucleophile slightly away from some of its stabilizing hydrogen bonds (Table V). The PO 3 group is in the same position at the midpoint of both pathways. This position is very close to that occupied by the NO 3 Ϫ in the PTPase-NO 3 complex crystal structure (less than 0.5 Å difference in center of mass).
Both an associative-like and a dissociative-like mechanism can be proposed on the basis of the modeled coordinates. The dissociative-like mechanism is favored, since it offers better transition state stabilization and is consistent with knowledge of phosphomonoester hydrolysis in solution and in alkaline phosphatase and hexokinase (29). Recent secondary isotope effect experiments on the Yersinia PTPase support this model (40). A large isotope effect observed for the scissile oxygen indicates a transition state in which the O-P bond is largely broken, while other isotope effects suggest a small degree of involvement of the attacking nucleophile, similar to the reaction in solution.
WAT Nuc -Breakdown of the phosphocysteine intermediate (Fig. 3C) requires attack by a water molecule, identified as WAT Nuc in the modeling. WAT Nuc in this model has a clear pathway to bulk solvent, consistent with the observation that the PTPases can catalyze the exchange of 18 O-labeled orthophosphate (27,39). For exchange to occur, the orthophosphate must attach to the protein and lose a labeled oxygen, which diffuses into bulk solvent to be replaced by the unlabeled oxygen of a water molecule.
A general base is required to accept a hydrogen from WAT Nuc . Glu 290 had been proposed to act as the general base, since mutation of Glu 290 to glutamine eliminates the acidic limb of the pH dependence of k cat (9). In the crystal structure, however, Glu 290 is 7.1 Å from the apical oxygen of the tungstate, well coordinated by hydrogen bonds to Arg 409 and Ser 287 . In addition, there is no room for an intervening water molecule to possibly act as a proton shuttle. From the structure, it appears the pH dependence on Glu 290 stems solely from its coordination of Arg 409 , which it can only do effectively while unprotonated.
The proton from WAT Nuc could be passed to Asp 356 , and, in the model for step II, Asp 356 is hydrogen bonded to WAT Nuc (Fig. 3C). If Asp 356 is a general base in the forward reaction of step II then Asp 356 must act as a general acid for the reverse of step II, which can be studied by the enzyme-catalyzed exchange of 18 O from labeled orthophosphate. The rate of 18 O exchange is several orders of magnitude slower than the hydrolysis of phosphotyrosine, indicating that the slow step in 18 O exchange is the formation of the phosphocysteine intermediate, i.e. the reverse of step II (27,39). However, the rate of 18 O exchange shows only a 2-fold decrease in the pH range studied, from 5.5 to 7.0 (27). In this pH range, an aspartic acid acting as a general acid (for the protonation of the hydroxyl leaving group) would be expected to show at least an order of magnitude decrease in rate, unless it has a pK a of at least 7.0. The apparent pK a of Asp 356 is about 5.1 from the pH dependence of k cat and k cat /K m (9,27).
If Asp 356 is not the general base for step II, the other ligands of WAT Nuc identified by the model must be considered as potential general bases. Gln 446 is hydrogen bonded to WAT Nuc , but the pK a values of glutamine residues make it an unlikely proton acceptor. Alternatively, the phosphate itself could accept the proton, either directly or via WAT1. The pK a2 of free phosphocysteine is about 5.0 (41). The use of the phosphate moiety of a thiophosphate ester to catalyze its own hydrolysis has been proposed for the non-enzymatic hydrolysis of S-nbutylphosphorothioate (42).
WAT1-Based on our modeling studies and our crystal structure, it is unlikely that WAT1 is WAT Nuc . Rather, WAT1 plays a structural role by bridging together a phosphate oxygen (O-1), the tyrosine leaving group oxygen (in step I) or WAT Nuc (in step II), the conserved Gln 450 , and an amide on the general acid-containing WpD-loop. WAT1 may be functionally more flexible than a side chain moiety, since it can adapt to multiple hydrogen bonding configurations necessary between steps I and II. Further insight into this water will be gained from a detailed analysis of a high resolution structure of a PTPasevanadate complex. 4 Applicability to Other PTPases-Sequence alignment of the PTPase domain from proteins from a wide range of organisms has identified 21 invariant and 29 highly conserved residues (43) (of which only 26 are found in the Yersinia sequence; Ref. 44). Nearly all of the 47 invariant or highly conserved residues are clustered around the active site. An ellipsoidal volume (45) defined by these residues has dimensions of 35 ϫ 30 ϫ 24 Å 3 (compared to 58 ϫ 44 ϫ 38 Å 3 for an ellipsoid defined from all the residues) and contains nearly all of the invariant or highly conserved residues while excluding nearly all the remaining residues. Thus, the roles of most of these conserved residues are probably predominantly structural, supporting the catalytically essential residues and maintaining the proper dielectric environment for the electrostatic interactions involved in catalysis.