Insights into the Reaction of Protein-tyrosine Phosphatase 1B

Catalysis by protein-tyrosine phosphatase 1B (PTP1B) occurs through a two-step mechanism involving a phosphocysteine intermediate. We have solved crystal structures for the transition state analogs for both steps. Together with previously reported crystal structures of apo-PTP1B, the Michaelis complex of an inactive mutant, the phosphoenzyme intermediate, and the product complex, a full picture of all catalytic steps can now be depicted. The transition state analog for the first catalytic step comprises a ternary complex between the catalytic cysteine of PTP1B, vanadate, and the peptide DADEYL, a fragment of a physiological substrate. The equatorial vanadate oxygen atoms bind to the P-loop, and the apical positions are occupied by the peptide tyrosine oxygen and by the PTP1B cysteine sulfur atom. The vanadate assumes a trigonal bipyramidal geometry in both transition state analog structures, with very similar apical O–O distances, denoting similar transition states for both phosphoryl transfer steps. Detailed interactions between the flanking peptide and the enzyme are discussed.

The phosphorylation of tyrosine residues by protein-tyrosine kinases and the reverse action by protein-tyrosine phosphatases (PTPs) 4 is a common mechanism for the control of biological pathways (1)(2)(3). Protein-tyrosine phosphatase 1B (PTP1B) is a biomedically important phosphatase with several roles, including negative regulation of insulin signaling by dephosphorylation of the insulin receptor tyrosine kinase (4). Knock-out studies show that loss of PTP1B is associated with an increased insulin sensitivity and suppression of weight gain in mice (5). As a result, this enzyme has been considered a significant target for treatment of type 2 diabetes and obesity (6,7). PTP1B also down-regulates cell growth by dephosphorylating the epidermal growth factor receptor (8). Overexpression has been observed in human breast and ovarian cancer, where it is believed to suppress potential tumors by antagonizing signaling of oncogenic factors (9,10). Other PTPs are known virulence factors for a number of human diseases (11)(12)(13).
Reactions catalyzed by PTPs take place by a ping-pong mechanism ( Fig. 1) (2). In the first step, a nucleophilic cysteine thiolate attacks the phosphate ester moiety of the substrate, resulting in formation of a phosphoenzyme intermediate with release of the peptidyl tyrosine. The second step occurs via attack of water on the phosphoenzyme intermediate and yields the final products inorganic phosphate and the regenerated enzyme. The central binding site for the substrate is the P-loop, a region at the bottom of a pocket that includes the nucleophilic cysteine and backbone amide groups oriented in a horseshoe fashion. The amide protons in the P-loop, together with a conserved arginine residue, hydrogen-bond to the phosphoryl group of the substrate and orient it for nucleophilic attack, providing transition state (TS) stabilization. Substrate binding is followed by conformational changes that culminate with closure of the active site pocket by a conserved and flexible loop of sequence WPD. The aspartic acid in this loop functions as a general acid in the first step and as a general base catalyst in the second step.
The design of PTP inhibitors is a very active area of research. A common strategy is focused on the synthesis of competitive inhibitors that block the active site (6,11,14). Binding affinities in these cases range from millimolar to micromolar, and specificity is often compromised by the high similarity between PTP active sites. In order to achieve more potency and selectivity, more extensive interactions are achieved by using doubly targeted inhibitors (9). For example, in PTP1B, targeting of a second Tyr(P) binding site in the vicinity of the active site has been used. In such cases, inhibition constants can be improved to the nanomolar range (15), and selectivity is gained because the second Tyr(P) site in PTP1B is not conserved among all PTPs.
One important reason for the limited success of single site inhibition in PTPs is the fact that substrate binding is far surpassed by the affinity of the TS. Like other PTPs, PTP1B catalyzes the hydrolysis of phosphate ester dianions, which have very slow rates of uncatalyzed hydrolysis (e.g. the half-life for the uncatalyzed attack of water on the dianion of alkyl phosphate esters is ϳ1.1 ϫ 10 12 years (k ϭ 2 ϫ 10 Ϫ20 s Ϫ1 at 25°C)) (16). Catalytic efficiency is achieved by selective affinity for the TS relative to the substrate. According to the TS binding paradigm, TS affinity is equivalent to the ratio between the specificity constant (k cat /K m ) for the catalyzed reaction and its uncatalyzed rate constant (17). Based on kinetic constants for catalyzed (18) and uncatalyzed (16)  when the substrate is the peptide acetyl-DADEpYL-NH 2 .
Besides the known closure of the WPD-loop, the existence of other catalysis-associated conformational changes that might affect protein-substrate interactions is uncertain. In order to better understand structural changes associated with catalysis, we have crystallized transition state analogs for the first and second catalytic steps. Kinetic isotope effects (KIEs) for the reaction of p-nitrophenyl phosphate by PTP1B were measured to verify that its transition state structure is similar to those of other PTPs. The first step TS analog reported here is a ternary complex that involves a central metavanadate with two apical ligands, the nucleophilic cysteine thiolate and the phenol leaving group, the latter as the tyrosyl oxygen atom in the peptide sequence DADEYL. This peptide is a fragment of the epidermal growth factor receptor, which, in phosphorylated form, exhibits strong affinity for PTP1B (18). A previous crystallographic study shows a Michaelis complex between a catalytically inactive C215S mutant of PTP1B and the phosphopeptide DADEpYL (19). The second step TS analog reported here is a complex between PTP1B and orthovanadate. Although a related structure has been previously described (18), the coordinates are not publicly available, and it was therefore necessary to determine a new structure approximating the second transition state. Comparisons of these newly determined and previously reported structures highlight significant protein conformational changes arising during catalysis.
These TS analogs are based on vanadate ions, which are tetrahedral in solution but in active sites of PTPs form stable pentavalent structures exhibiting trigonal bipyramid geometries (18, 20 -23). Such structures are widely accepted as sharing close structural characteristics with the metaphosphate-like transition state in phosphate transfer (24). In comparison with other analogs (25)(26)(27), such as WO 4 2Ϫ , MgF 3 Ϫ , and NO 3 Ϫ , complexes based on vanadate have the advantage of displaying the same overall charge as phosphate and can easily derivatize and interact with ligands at their axial positions.
The new TS analog structures, together with previously solved structures of the apoenzyme (28), phosphoenzyme intermediate (29), and product complex (30), now provide a depiction of all of the catalytic steps involved in the mechanism of PTP1B. This full array of structures permits a step-by-step description of the structural changes involved in binding and catalysis, particularly those related to the following protein segments: 1) the P-loop (residues 215-222), which includes the nucleophilic cysteine (Cys 215 ) and an important serine (Ser 222 ) that stabilizes its thiolate form, and the arginine

EXPERIMENTAL PROCEDURES
The expression and purification of wild type PTP1B were slightly modified from previously reported methods (32) (see supplemental material). Protein concentrations were monitored by UV using an A 1 mg/ml 280 nm ϭ 1.24. The peptide DADEYL (Ͼ98% pure) was a custom synthesis by Anaspec. Labeled and natural abundance pNPP (dicyclohexylammonium salt) were synthesized according to previous methods (33). Sodium orthovanadate (Na 3 VO 4 ) was purchased from Fisher.
Kinetic Isotope Effect Determinations-KIEs were measured using the internal competition method and thus are isotope effects on V/K (34). In the commonly used notation, a leading superscript of the heavier isotope is used to indicate the isotope effect on the following kinetic quantity (e.g. 15 (V/K) denotes 14 (V/K)/ 15 (V/K), the nitrogen-15 KIE on V/K).
Natural abundance pNPP was used for measurements of 15 (V/K). The 18 O KIEs 18 (V/K) bridge and 18 (V/K) nonbridge were measured by the remote label method, using the nitrogen atom in p-nitrophenol as a reporter for isotopic fractionation in the labeled oxygen positions (35). The isotopic isomers used are shown in the supplemental material. Isotope effect determinations were made at 25°C in 50 mM bis-tris buffer, pH 5.5, containing 1 mM dithiothreitol. The pNPP concentration was 18 mM, and the reaction was started by the addition of wild type PTP1B to a final concentration of 0.21 M. Within 15-20 min, the reactions reached 40 -60% completion and were stopped by titrating with HCl to pH 2-3. Protocols for isolation of p-nitrophenol and isotopic analysis and calculation of the isotope effects were the same as previously described (36) and are described in the supplemental material.
Crystallization-Crystals were grown using sitting drop vapor diffusion at 4°C by mixing 2 l of protein solution, 0.5 l of 30% (w/v) sucrose, and 3 l of precipitant solution (0.1 M Hepes, pH 7.5, 0.2 M magnesium acetate, and 15-17% polyethylene glycol 8000). Single crystals appeared after 3 days. The  MAY 21, 2010 • VOLUME 285 • NUMBER 21 protein solution was prepared as follows. 0.36 l of 100 mM of Na 3 VO 4 and 10 l of 50 mM DADEYL peptide (at pH 8.5-9.0) were mixed and allowed to react for 1-1.5 h; 50 l of native PTP1B (12 mg/ml in 10 mM Tris, pH 7.5, 25 mM NaCl, 0.2 mM EDTA, and 3 mM dithiothreitol) was then added to the peptide/ vanadate mixture and immediately placed in the crystallization tray. The final protein solution contained a ratio of 1:3:30 between PTP1B, Na 3 VO 4 , and the peptide DADEYL. Crystallization at higher concentrations of Na 3 VO 4 resulted in different crystal morphology; a 30:3 vanadate/peptide ratio gave crystals that proved to be a binary complex between PTP1B and VO 4 3Ϫ . Cryoprotection was performed by transferring crystals stepwise into stabilization solution with increasing glycerol amounts to a final concentration of 15% and the respective initial concentrations of ligands present in the protein solution, and then flash-cooled in liquid nitrogen.

The Mechanism of Catalysis by Protein-tyrosine Phosphatase 1B
Data Collection, Structure Determination, and Refinement-Diffraction data were collected using a home source generator and detector (Rigaku RU-200/Raxis IVϩϩ). Data were indexed and processed using d*TREK in the program Crystal Clear (37). Molecular replacement was performed using Phaser (38) from the CCP4 program suite (39,40). The search model was the apo wild type PTP1B structure (Protein Data Bank (PDB) entry 2CM2) with the active site water molecules removed. Refinement was performed using the program Phenix (41). The TLS groups were generated by the TLS Motion Determination server (42). Coot (43) and MolProbity (44) were used for model building and validation. The crystallographic data and statistics of structure refinement are given in Table 1. The observed differences in B-factors between TSA1 (transition state analog 1) and TSA2 (TSA1 Ͼ TSA2) are due to tighter crystal packing along the c axis of the unit cell in TSA2. The bound peptide in TSA1 partially disrupts packing along the c axis, resulting in increased spacing and conformational flexibility between adjacent molecules. Structure factors and coordinates for the PTP1B-VO 3 -DADEYL and PTP1B-VO 4 structures have been deposited in the PDB under accession numbers 3I7Z and 3I80, respectively. All figures depicting crystal structures were prepared using PyMOL (45).

RESULTS
Kinetic Isotope Effects-Reactions of wild type PTP1B with pNPP were carried out at the optimum pH of 5.5 and 25°C, where K m ϭ 0.58 Ϯ 0.01 mM and k cat ϭ 24.4 Ϯ 0.4 s Ϫ1 . The isotope effect data are shown in Table 2. As the enzymatic substrate of PTPs is the dianionic form of pNPP, the 18 (V/ K) nonbridge isotope effect was corrected considering the isotope fractionation for protonation of the nonbridge oxygen atoms because pNPP is present as a mixture of monoanion and dianion forms at this pH. Data for the reaction catalyzed by the other PTPs YopH from Yersinia and rat PTP1 are also shown (36) for comparison purposes.
Crystallography-The two transition state analogs were crystallized under similar conditions, except for the concentrations of peptide and vanadate, which were optimized to give the structures of the transition state analog for the first step (TSA1) and the second step (TSA2). With a ratio of peptide/vanadate/ enzyme of 3:30:1, hexagonal prism crystals were obtained. The composite omit map of the active site revealed electronic density for a single vanadate, exhibiting a trigonal bipyramidal structure conforming to TSA2, and the map showed no evidence for a bound peptide. When the peptide/vanadate ratio was 30:3, crystals grew with hexagonal bipyramidal morphology. Composite omit maps for the active site reveal clear electronic density for the vanadyl-peptide complex and a trigonal bipyramid structure that is consistent with TSA1. Supplemental Fig. S2 displays composite omit maps for residues in the active site of TSA1 and TSA2. Fig. 2 shows key active site interactions in the TS analog structures. The apical positions of the vanadate moieties of both TSA1 and TSA2 are occupied by the sulfur of Cys 215 and by an oxygen atom at distances of 2.5 and 2.1 Å, respectively, from the central vanadium atom. The equatorial oxygen atoms hydrogen-bond to backbone amide groups of the P-loop and and is calculated using all data; R free is the R-factor based on 5% of the data excluded from refinement. c Ramachandran statistics were calculated using the MolProbity server (44).

The Mechanism of Catalysis by Protein-tyrosine Phosphatase 1B
the guanidinium group of Arg 221 , which is rotated in relation to the apo structure by about 82°around the -3 dihedral angle (see Fig. 3). Substrate and competitive inhibitors cause the same conformational change in Arg 221 (19,28,46). In this conformation, the side chain of Arg 221 hydrogen-bonds with the backbone carbonyl of Trp 179 and assists the WPD-loop in a shift from the open, catalytically inactive to the closed, active conformation. This brings the WPD-loop residue Asp 181 into proximity to protonate the leaving group oxygen (47). The essential role of this conserved arginine in facilitating WPD-loop closure has been shown in the closely related PTP YopH, where the R221K mutation disables general acid catalysis, presumably by affecting WPD-loop movement (33). Gln 262 in TSA1 occupies a similar position as in the Michaelis complex and is rotated 132°counterclockwise in relation to TSA2 (Fig. 2b). This residue is believed to position a nucleophilic water molecule in the second catalytic step (29). In TSA1, the amide proton of the side chain of Gln 262 hydrogen bonds to the carbonyl oxygen of Gly 259 (not depicted). The same interaction is observed in the apo and Michaelis complex. In TSA2, the side chain of Gln 262 rotates into the active site, and the amide carbonyl oxygen hydrogen-bonds to the apical oxygen in the trigonal bipyramidal vanadate moiety.
In both TSA1 and TSA2 structures, Lys 120 adopts a similar conformation, hydrogen-bonding to Asp 181 and Glu 115 at donor-acceptor distances of 2.6 and 2.7 Å, respectively. In the Michaelis complex, distances are 2.8 and 3.0 Å, respectively. In contrast, in the phosphoenzyme, the Lys 120 side chain does not hydrogen-bond to other protein residues. This residue has been suggested to be part of dynamic events during substrate and inhibitor binding (31,48). The K120A mutation has little effect on K m and k cat for pNPP hydrolysis and K i values for vanadate (49). How-   (29); e, second transition state complex, between native PTP1B and orthovanadate (new structure, PDB entry 3I80); f, PTP1B-tungstate product analog complex (PDB entry 2HNQ) (30). These structures are further summarized in supplemental Table S1. MAY 21, 2010 • VOLUME 285 • NUMBER 21 ever, the use of a nonbiological substrate may preclude a proper definition of the role of this residue. In fact, protein alignments reveal that most PTPs retain a Lys or Arg at this position, suggesting an important role for this residue (50). The crystal structure of a complex between PTP1B and insulin receptor tyrosine kinase reveals that Lys 120 is involved in a salt bridge interaction with insulin receptor tyrosine kinase (51).

The Mechanism of Catalysis by Protein-tyrosine Phosphatase 1B
Extensive interactions occur between the peptide of TSA1 and a surface groove adjacent to the active site (supplemental Fig. S4). These interactions include hydrogen bonds between the backbone amide proton of Arg 47 in PTP1B and the backbone carbonyl oxygen of Asp Ϫ2 in the peptide and between the side chain carboxyl of Asp 48 in PTP1B and the backbone amide protons of Tyr(P) and Leu ϩ1 in the peptide. 5 Similar interactions are observed in the Michaelis complex along the core of the peptide, although positions of the peptide backbone and N-terminal residue side chains are slightly different. The change in position of the peptide backbone is accompanied by movement of the aryl ring of the substrate, which is about 0.45 Å farther away from the active site in TSA1 than in the Michaelis complex, as measured by the distance between the aryl ring and C␣ of Arg 221 . Differences in N-terminal regions of the peptides may arise from crystal packing. Hydrogen bonds along this region are observed with adjacent protein molecules; in TSA1, the side chain of Asp Ϫ4 hydrogen-bonds to the side chain of Thr 168 in an adjacent molecule and to Arg 47 in the same molecule. In the Michaelis complex structure, the respective aspartic acid hydrogen bonds to the backbone carbonyl of Gln 166 in an adjacent molecule. There are no other significant changes in the enzyme or peptide between Michaelis complex and TSA1 structures that can be attributed to crystal packing. Fig. 3 combines the present TS structures with published ones to depict the active site arrangement at each step of the pathway. Supplemental Table S1 summarizes the structures used for each step.

Kinetic Isotope Effects and Transition State Structure
The KIEs for PTP1B are similar to those reported previously for other PTPs (Table 2). Because they were measured using the competitive method, the isotope effects are on the kinetic quantity V/K, which includes events up to and including the first irreversible step. These thus reflect the first catalytic step, phosphoryl transfer from substrate to enzyme.
Kinetic isotope effects reflect differences in bonding to the labeled atom between the ground state and transition state (52). Consequently, they are useful in probing transition state structures and reaction mechanisms of reactions. A primary kinetic isotope effect at an atom undergoing bond cleavage will be normal (Ͼ1), due to the preference of the heavier isotope for the lower energy (more tightly bonded) position. Secondary kinetic isotope effects are normal when the labeled atom becomes more loosely bonded in the transition state or inverse (Ͻ1) when bonding becomes tighter. KIEs have been reported for a number of enzymatic and uncatalyzed phosphate ester reactions (35, 53-57), which provides a background for the interpretation of the KIE data measured in this work.
The three KIEs give different information regarding transition structure and charge. The magnitude of 15  The experimental and most theoretical results indicate that the Michaelis complex consists of the dianion form of the substrate, with the nucleophilic cysteine in the thiolate form and the conserved Asp residue on the WPD-loop in the protonated form, followed by a loose metaphosphate-like phosphoryl transfer in which the leaving group is fully neutralized (2, 36, 59 -62), although some theoretical studies have suggested that PTPs utilize the monoanionic form of the substrate (63,64).

Crystal Structures
An optimal transition state analog would consist of a peptide sequence corresponding to a natural substrate incorporating a tyrosine, with metavanadate (VO 3 ) poised between the nucleophilic cysteine sulfur atom and the tyrosine hydroxyl group. In solution, orthovanadate (VO 4 3Ϫ ) forms vanadate esters with hydroxylic species, and these are found to be substrates for a range of enzymes, including dehydrogenases, isomerases, and aldolases (65). The formation constants of vanadate esters in aqueous solution are on the order of 0.1-0.2 M Ϫ1 for aliphatic esters and about 3 times larger for aromatic esters (65). One precedent for vanadate ester formation and enzymatic recognition is the reaction between D-glucose and vanadate. D-Glucose contains one primary and several secondary hydroxyl groups that can form vanadate esters. Despite the many species that form in solution, the small amount of glucose-6-vanadate that forms is selectively recognized by glucose-6-phosphate dehydrogenase (66,67). In the present work, the tyrosyl vanadate ester of the peptide DADEYL was formed by preincubation of 5 Amino acid position numbers marked as positive and negative (Asp Ϫ2 , Leu ϩ1 , etc.) represent positions on the peptide relative to Tyr(P).
orthovanadate and peptide and was subsequently bound by PTP1B and co-crystallized. Vanadate anions form stable tetrahedral structures in solution but exhibit clear preference for trigonal bipyramidal structures in the presence of polydentate ligands (65), including the active site of enzymes, such as PTPs (20,22). In these enzymes, vanadate exhibits K i values ranging from 0.4 to 5 M (23, 68,69) and structural characteristics resembling the trigonal bipyramidal transition state in phosphoryl transfer reactions. This geometry is formed by nucleophilic attack of the cysteine (Cys 215 in PTP1B) on the vanadium atom, with the nucleophilic sulfur and leaving group oxygen occupying apical positions. The equatorial oxygen atoms of this structure extensively interact with the P-loop (Fig. 2). The binding mode of vanadate in the active site of PTPs has been considered a fair structural transition state analog, showing clear advantages in relation to other anions (24). For example, tetrahedral anions that do not undergo similar stabilization in the active site of PTPs, exhibit weaker inhibition. The K i values for phosphate and arsenate range from 10 to 30 mM (62,70), and K i values for tungstate range from 5 to 60 M (70, 71). It has been noted that in some instances, PTP complexes with vanadate resemble a more associative structure than the dissociative TS determined experimentally (68).

Binding of Substrate and Water Molecules in the Active Site
Substrate binding to PTPs is accompanied by the expulsion of several water molecules from the active site. Three water molecules are commonly found in crystal structures of apo-PTPs at positions occupied by the phosphoryl nonbridge oxygens when substrate is bound (31). Displacement of these waters may supply the entropic driving force assisting the substrate binding. Other water molecules have important roles in catalysis by PTP1B (29). The water molecule W1 in Fig. 4 is found in all x-ray structures in the catalytic pathway of PTP1B and is also observed in other members of the PTP family (e.g. Yersinia PTP (72) and PTP␤ (22)). In TSA1, the coordination shell for W1 includes the side chains of Asp 181 and Gln 266 , the backbone amide proton of Phe 182 in the WPD-loop, and axial and equatorial oxygens of the trigonal bipyramid structure. In TSA2, the coordination shell of W1 also includes the side chain of Gln 262 . In view of the short distance between the side chain of general acid catalyst Asp 181 and the oxygen leaving group in the substrate, it is difficult to say that the water molecule W1 is involved in any proton transfer during catalysis. However, this water molecule may be an important structural component of the active site through hydrogen bonding with key side chains. Additionally, this water may be an energy sink by partially solvating the phosphate moiety in the active site. It is important to point out that Gln 262 and Gln 266 , both in the coordination shell of water W1, are highly conserved residues in PTPs.
Each crystal structure depicted in Fig. 3 represents a discrete form of PTP1B in the ping-pong mechanism. Next, we provide a step-by-step description of structural changes associated with binding and catalysis.

Changes Accompanying Conversion of the Apoenzyme (PDB Entry 2CM2) to the Michaelis Complex (MC) (PDB Entry 1PTU)
P-loop-Substrate binding to the P-loop occurs with side chain rotation of Arg 221 , which is rotated by Ϫ71°around -3. New hydrogen bonds are formed with the phosphate moiety of the substrate, and with the carbonyl oxygen of Trp 179 . This interaction stabilizes the WPD-loop in the active, closed conformation.
WPD-loop-The closed position of this loop brings the general acid Asp 181 into position for catalysis (19,46). Stabilization of the closed conformation is further achieved by interaction of the Phe 182 side chain with the phosphotyrosine ring of the substrate. Phe 182 and the highly conserved Tyr 46 flank the phenyl ring of the substrate.
Q-loop-The Gln 262 residue occupies the same position in both structures. Steric constraints in the MC prevent the Gln 262 side chain from interactions with the active site, precluding a catalytic function in the first catalytic step. The side chain is in sufficient proximity to participate in van der Waals interactions with the Tyr(P) side chain of the substrate.
Lys-loop-Lys 120 is rotated Ϫ100°around the -4 angle. In the apo structure, the ammonium moiety is embedded in a polar environment interacting with the hydroxyl groups of Ser 118 , Ser 216 , and Tyr 46 . In the MC, the Lys 120 side chain hydrogen-bonds to the Asp 181 and Glu 115 side chains.
Tyr(P) Recognition Loop-The position of Asp 48 is similar for the apo and MC states, but in the MC, the Asp 48 side chain hydrogen-bonds to the peptide main chain amides of Tyr(P) and Leu ϩ1 . Peptide binding has a strong influence in the position of Arg 47 , which moves to make hydrogen bonds with Glu Ϫ1 and the main chain carbonyl of Asp Ϫ4 .

Changes Accompanying Conversion of the MC (PDB Entry 1PTU) to the Transition State Analog of the First Step, TSA1 (PDB Entry 3I7Z)
The TSA1 structure depicts a pentacoordinate species consisting of a planar metavanadate with apical ligands and equatorial oxygen atoms interacting with the P-loop. This structure models the TS of the first catalytic step.
P-loop-The Cys 215 sulfur attacks the vanadyl center in line with leaving group departure. The nonbridge oxygen positions are not appreciably different from their position in the MC, but the vanadium position has moved relative to that of phosphorus. The unusually low pK a of 4.6 for the Cys 215 thiol (73,74) results in part from a stabilizing hydrogen bond with Ser 222 . The sulfur-oxygen distance establishing this interaction is not significantly different in the apo (3.2 Å), MC (3.0 Å) (this is shorter because PDB entry 1PTU is a C215S mutant), and TSA1 (3.3 Å) structures. The constancy of this distance indicates that the low pK a of the nucleophile is permanently built into the enzymatic structure.
WPD-loop-The Asp 181 carboxyl group moves closer to the Tyr(P) leaving group oxygen (2.6 Å) than in the MC structure (3.5 Å). This distance is 3.0 Å in a related MC structure (PDB entry 1G1G) with the peptide sequence ETDYpYRKGGKLL bound to an inactive C215A mutant (46). In the first catalytic step, the phosphotyrosine is attacked by the nucleophilic Cys 215 accompanied by proton transfer from Asp 181 to the leaving group oxygen. KIE data confirm that the efficiency of leaving group protonation is such that the leaving group is fully neutralized in the TS.
Comparing TSA1 with MC structure (PDB entry 1PTU), the phenyl ring of Phe 182 is rotated Ϫ23°around the -1 dihedral; in MC (PDB entry 1G1G), this residue is rotated 15°. The Ϫ23°r otation of Phe 182 permits extensive interaction between the Phe 182 and the vanadyl tyrosine aromatic rings. However, in the MC 1G1G, the side chain of Arg ϩ1 is very close to the Phe 182 ring, and because of van der Waals or steric clashes, the Phe 182 assumes a different position compared with MC 1PTU. This variability in the details of peptide-enzyme interactions fits with the substrate flexibility of PTP1B.
Q-loop-The Gln 262 position does not change appreciably from its position in the MC and remains excluded from prox-imity with the active site by steric clashes with the vanadyl Tyr. This verifies the proposal of Pannifer et al. (29) that Gln 262 has no catalytic function in the first step.
Lys-loop-The Lys 120 residue rotates around the -4 dihedral by Ϫ20°. In both states, this residue hydrogen-bonds to Asp 181 and Glu 115 . However, the interaction in the TSA1 seems to be more favorable than in the MC; donor-acceptor distances involving Lys 120 are 2.6 Å with Asp 181 and 2.7 Å with Glu 115 but 0.2-0.3 Å longer in the MC (PDB entry 1PTU). In the other Michaelis complex (PDB entry 1G1G), Lys 120 is found in a similar position as PDB entry 1PTU and does not exhibit hydrogen bonds as short as those observed in TSA1.
Tyr(P) Recognition Loop-As in other PTP1B structures with bound peptides (PDB entries 1EEO (75), 1G1G, and 1G1H (46)), in TSA1, Asp 48 forms hydrogen bonds to the main chain amides of vanadyl-Tyr(P) and Leu ϩ1 residues. The Asp 48 residue is conserved among all PTPs and may be important in determining the overall substrate orientation, in positioning the peptide substrate for an optimal orientation for the attack of Cys 215 and proton transfer from Asp 181 (46). Our TSA1 structure indicates that this interaction is maintained in the transition state. The C␣ position of Asp 48 moves by 0.5 Å from the MC to the TSA1 structures in order to maintain proper structural requirements for the hydrogen bond between Asp 48 and the peptide.
The position of Arg 47 differs from the MC 1PTU by about 3.5 Å in the distal part of the side chain. This is probably related to the bond lengthening of the scissile bond observed from the ground to the transition state (Fig. 5). In the MC, weak hydrogen bonds are present between Arg 47 and Glu Ϫ1 and the main chain carbonyl of Asp Ϫ4 of the peptide. There are no apparent hydrogen bonds involving the Arg 47 in TSA1, although it is embedded in a position with negative charges from Glu Ϫ1 , Asp Ϫ2 , and Asp Ϫ4 .
The combined changes in the positions of Arg 47 and Asp 181 from the ground to the transition state reflect the needs of catalysis. In the transition state, the protein structure at the WPDloop becomes tighter to facilitate proton transfer but loosens around Arg 47 in order to permit leaving group departure.

Changes Accompanying Conversion of TSA1 (PDB Entry 3I7Z) to the Phosphoenzyme (PDB Entry 1A5Y)
P-loop-Cys 215 is phosphorylated. No other significant changes are observed in this region.
WPD-loop-This loop slides horizontally, and Asp 181 is displaced 1.2 Å relative to its position in TSA1.
Q-loop-A comparison is not possible in this case because PDB entry 1A5Y is a Q262A mutant.
Lys-loop-In the phosphoenzyme, the Lys 120 side chain is not involved in hydrogen bonds with other protein residues. In TSA1, this residue is involved in hydrogen bonds with Asp 181 and Glu 115 . These interactions may be important for positioning Asp 181 during catalysis because this residue in the phosphoenzyme is found 1.2 Å away from its catalytically competent position observed in TSA1.
Tyr(P) Recognition Loop-Relative to TSA1, Asp 48 and Arg 47 move away from their positions involved in hydrogen bonds with the peptide. This loop backbone is similar to its position in other structures without bound peptide.

Changes Accompanying Conversion of the Phosphoenzyme (PDB Entry 1A5Y) to the Transition State Analog of the Second
Step, TSA2 (PDB Entry 3I80) P-loop-The side chains and backbone positions of TSA1, the phosphoenzyme, and TSA2 are very similar. We note that the hydrogen bonding distances between the Cys 215 sulfur atom and the oxygen of Ser 222 (3.3 and 3.2 Å respectively) remain unchanged from those of the apo, MC, and TSA1 structures and are essentially constant through the entire catalytic cycle. In the transition state of the second step, this interaction serves to facilitate the leaving group propensity of the sulfur atom.
WPD-loop-This loop is found in a similar position in TSA1 and TSA2. In the phosphoenzyme structure, the Asp 181 residue is slightly displaced 1.2 Å away from its position in TSA1 and TSA2. In the second catalytic step, this residue is a general base for the deprotonation of a nucleophilic water molecule.
Q-loop-Comparing TSA2 with TSA1, the Gln 262 residue is rotated by Ϫ120°around the -2 angle. This brings the amide moiety into the active site, where it positions the nucleophilic water molecule for phosphoenzyme hydrolysis. This structure explains the observation that the Q262A mutation hinders phosphoenzyme hydrolysis but not the first step of catalysis (29).
Lys-loop-In the phosphoenzyme intermediate, the Lys 120 side chain is not involved in hydrogen bonds with other protein residues. In TSA2, as in TSA1, Lys 120 is involved in hydrogen bonds with Asp 181 and Glu 115 .
As revealed by the TSA1 and TSA2 structures, the Lys 120 may interact with key residues in the transition state. This contrasts with molecular dynamics simulations of the ground state, which observed concurrent movements between the WPDloop and Lys-loop during WPD-loop closure (31). Together, these observations imply that the movement of the Lys-loop probably lags behind the WPD-loop closure in the early stages of each catalytic step.
Tyr(P) Recognition Loop-In TSA2, the side chains of Arg 47 and Asp 48 hydrogen bond with each other. The backbone of this region is in a similar position in the phosphoenzyme and TSA2.
The distances between the two apical oxygen atoms are identical within the uncertainty limits of the structure in the two transition state analogs (Fig. 5). This implies a similar transition state for the two phosphoryl transfer reactions.

Changes Accompanying Conversion of TSA2 (PDB Entry 3I80) to the Product Complex (PDB Entry 2HNP)
P-loop and WPD-loop-Both loops return to an open conformation, as observed in the apo enzyme.
Q-loop-Relative to TSA2, in the product complex, Gln 262 rotates back to the position observed in the apo form.
Lys-loop-In contrast to TSA1 and TSA2, in the intermediate and the product complexes, Lys 120 is positioned away from Asp 181 and Glu 115 and does not hydrogen-bond to protein residues.
Tyr(P) Recognition Loop-The interactions between residues Arg 47 and Asp 48 in TSA2 are lost in the product complex.
Complementing these specific step-by-step observations, we also notice that several precedents from the modeling of enzymatic features in small molecules indicate that decreased entropy and proximity have a profound effect in nucleophilic attack and general acid catalysis. In both cases, these effects have shown catalytic contributions of over 10 6 -fold (76 -78), which, combined, may partially explain the catalytic proficiency of enzymes. Upon formation of the Michaelis complex in PTPs, WPD-loop closure and the resulting proximity of the Asp 181 side chain to the leaving group oxygen cause the pK a of the leaving group tyrosine to drop below its acidity in water. It is interesting to observe that lengthening of the scissile P-O bond causes the peptide chain to slightly shift in the active site of PTP1B in TSA1 relative to the MC structure. This change probably occurs upon transformation of the Michaelis complex to the transition state because the apical bond in the trigonal bipyramidal transition state is longer than the respective bond in the ground state. However, the phosphotyrosine ring (and consequently the leaving group oxygen) and nucleophile (Cys 215 ) do not appreciably change their position (Fig. 5). The major part of the reaction coordinate, implied by these structures, is movement of the phosphorus atom between the leaving group and the nucleophile.
The proximity of the general acid in TSA1, brought into the active site by the conformational change associated with WPD-loop closure, implies that the electronic effects of proton transfer to the ester oxygen, rather than only structural consequences, are important for the very small leaving group dependence in PTPs. Site-directed mutagenesis and Brönsted relationships for YopH, another member of the PTP family, show that abolishment of general acid catalysis by Asp is accompanied by a change of the Brönsted ␤ lg from about Ϫ0.1 in the wild type to Ϫ1 in mutants (70). In the transition state, the general acid catalyst Asp 181 may drive the proton transfer causing the P-O bond lengthening. The charge in the phosphate moiety changes from Ϫ2 in the substrate to approximately Ϫ1 in the transition state. Negatively charged nucleophiles are unlikely to react with phosphate monoesters in the dianion form. However, this has been observed in cases where, as in the PTP1B reaction, proton transfer is present or the phosphate moiety is neutral or monoanionic (78).

CONCLUSIONS
The close similarity of the apical O-O distances in TSA1 and TSA2 support the notion that the transition states of the two catalytic steps are very similar. KIE and linear free energy relationship data have suggested that the first TS is loose for the related PTPs, YopH and Stp (36,62,79), as is the second TS in Stp1 (80). Because PTPs exhibit very similar active sites, it is likely that the same is true in the transition states of reactions catalyzed by other PTP family members.
Several general conclusions are revealed by the step-by-step analysis. The leaving group propensity in the first catalytic step is modulated by Asp 181 through an interaction that is controlled by motions of a protein loop. In contrast, the pK a of the Cys 251 thiol (the nucleophile in the first step, the leaving group in the second) is controlled by a hydrogen bond in a fixed structural element that is unchanged throughout the catalytic cycle. The unusually low pK a makes catalytic sense in the second step, in which Cys 215 is a leaving group. The first step would logically be assisted by a weakened hydrogen bond that would raise this pK a , but the evidence from the structures is that the low pK a is fixed. Lys 120 hydrogen-bonds to key residues involved in catalysis, but its main function may remain in structural and protein-protein interactions. This may be an important factor for development of PTP inhibitors. The transition states of the two catalytic steps are indicated to be essentially identical. A structural rationale for the role of Gln 262 in the second catalytic step, but not the first, is now evident.