Residue 259 is a key determinant of substrate specificity of protein-tyrosine phosphatases 1B and alpha.

The aim of this study was to define the structural elements that determine the differences in substrate recognition capacity of two protein-tyrosine phosphatases (PTPs), PTP1B and PTPalpha, both suggested to be negative regulators of insulin signaling. Since the Ac-DADE(pY)L-NH(2) peptide is well recognized by PTP1B, but less efficiently by PTPalpha, it was chosen as a tool for these analyses. Calpha regiovariation analyses and primary sequence alignments indicate that residues 47, 48, 258, and 259 (PTP1B numbering) define a selectivity-determining region. By analyzing a set of DADE(pY)L analogs with a series of PTP mutants in which these four residues were exchanged between PTP1B and PTPalpha, either in combination or alone, we here demonstrate that the key selectivity-determining residue is 259. In PTPalpha, this residue is a glutamine causing steric hindrance and in PTP1B a glycine allowing broad substrate recognition. Significantly, replacing Gln(259) with a glycine almost turns PTPalpha into a PTP1B-like enzyme. By using a novel set of PTP inhibitors and x-ray crystallography, we further provide evidence that Gln(259) in PTPalpha plays a dual role leading to restricted substrate recognition (directly via steric hindrance) and reduced catalytic activity (indirectly via Gln(262)). Both effects may indicate that PTPalpha regulates highly selective signal transduction processes.

Protein-tyrosine phosphatases (PTPs) 1 and protein-tyrosine kinases control the phosphotyrosine (Tyr(P)) levels of cellular proteins and are thus key regulators of signal transduction (reviewed in Refs. 1 and 2). PTPs are a diverse family of intracellular and receptor-type enzymes that are characterized by one or two structurally conserved catalytic domain(s) of about 250 amino acid residues. The catalytic domain in intracellular PTPs is often associated with proximal or distal sequence homology containing regulatory elements such as SH2 and FERM domains (3) directing protein-protein interactions, subcellular localization, or enzyme stability. Most receptortype PTPs contain two catalytic domains in their intracellular region and, in addition, have a single transmembrane region and an extracellular domain. The diversity of the extracellular domains suggests that these PTPs are regulated by specific extracellular ligands.
The intricate regulation of signal transduction processes into temporarily and spatially ordered pathways requires recruitment of regulatory molecules, including PTPs, into macromolecular assemblies (reviewed in Ref. 4). Thus, subcellular localization is considered to play a significant role in defining the substrates that are dephosphorylated by specific PTPs (5). However, there is increasing evidence that substrate specificity is also conveyed directly by the catalytic domains of PTPs. Thus, several studies with synthetic tyrosine-phosphorylated substrates (6 -9) or peptides containing nonhydrolyzable phosphotyrosine analogs (10) have shown significant preferences for specific residues adjacent to the phosphorylated tyrosine or Tyr(P) mimetic. One of the most convincing demonstrations of the functional importance of direct substrate recognition by PTPs came from recent studies of the highly homologous PTPs, SHP-1 and SHP-2. By using chimeric constructs of these two PTPs, Böhmer and co-workers (11) demonstrated that the differential interaction of SHP-1 and SHP-2 with the epidermal growth factor receptor is due to the specificity of the catalytic domains rather than the SH2 domains. In a detailed study of the influence of SHP-1 and SHP-2 on Xenopus oocyte maturation, O'Reilly and Neel (12) showed that the most important substrate recognition elements reside in the catalytic domain. Furthermore, in a functional cell-based rescue assay, we found PTP␣ and PTP⑀ to be the most efficient negative regulators of insulin signaling, whereas other receptor-type PTPs had little or no influence (13,14). Finally, studies with so-called substrate-trapping mutants show selective substrate recognition that seems to be mediated by the catalytic domains (15,16).
PTPs are generally considered potentially important therapeutic targets due to their pivotal roles as control elements in signal transduction pathways (17). As an example, several candidate PTPs have been proposed as negative regulators of the insulin receptor signaling pathway, including PTP1B and PTP␣ (reviewed in Ref. 18). Thus, selective inhibitors of these enzymes could potentially be useful in the treatment of diabetes. Recently, analysis of PTP1B knock-out mice (19) and treatment of diabetic ob/ob mice with selective PTP1B inhibitors (20) provided support for the view that PTP1B is a major regulator of insulin signaling. By using a structure-based design, we have recently been able to make selective, low molecular weight non-phosphorus PTP1B inhibitors. A basic nitrogen in the inhibitor forms a salt bridge with the selectivitydetermining residue Asp 48 in PTP1B. In other PTPs with an asparagine in the equivalent position, this basic nitrogen causes repulsion. The net effect is a remarkable selectivity for PTP1B (21). Furthermore, replacing the basic nitrogen in the inhibitor structure with an oxygen atom increased the potency of the inhibitor for all PTPs with an asparagine in position 48.
In the present study, we decided to get further insight into the elements that determine the substrate specificity of different PTPs and, hence, by inference also information that can be used in structure-based design of selective inhibitors. In particular, we wanted to study the difference in substrate recognition of PTP1B and PTP␣. Tyrosine-phosphorylated synthetic peptides have been used extensively to analyze the substrate specificity of PTPs. The peptide Ac-DADE(pY)L-NH 2 , which is derived from the epidermal growth factor receptor, has been particularly useful. This peptide is well recognized by PTP1B but less efficiently by PTP␣. Based on C␣ regiovariation analyses and primary sequence alignments, we have previously identified regions in close proximity to the catalytic cleft that are likely to confer specificity onto PTPs. 2 The combination of residues 47, 48, 258, and 259 (PTP1B numbering) might be a specificity-determining region. PTP1B and PTP␣ mutants, in which these four residues are exchanged either in combination or alone, were used to analyze the substrate specificity against Ac-DADE(pY)L-NH 2 peptide and a set of analogs. Residue 259, which is a glutamine in PTP␣ and a glycine in PTP1B, was found to be a major determinant of substrate recognition capacity and hydrolysis.

EXPERIMENTAL PROCEDURES
Materials-Phosphopeptides were purchased from Neosystem (Strasbourg, France). The purity of the peptides was greater than 97%. Water was purified in a Millipore purification system (18 megohm-cm; Millipore Inc.). Glutathione-Sepharose, Sephadex G-25, Q-Sepharose Fast Flow, Mono Q, and Superdex 200 were from Amersham Pharmacia Biotech.
Phosphate Reagent-One volume of 10% (w/v) ammonium molybdate was mixed with 3 volumes of 0.2% (w/v) malachite green in 4 N HCl.
After stirring for 30 min at room temperature, the solution was filtered through a 0.22-m Millipore filter (Millex-GV). The solution was stored at 4°C in a foil-wrapped container. Before usage, the solution was stirred for 30 min and filtered through a 0.22-m Millipore filter.
Buffers-All enzyme assays with synthetic peptide substrates were performed at a constant ionic strength of 100 mM using a three-components buffer system consisting of 50 mM Tris, 50 mM Bis-Tris, and 100 mM acetate (22,23). The buffer contained 5 mM 2-mercaptoethanol and 0.1% (v/v) Tween 20 (24,25). Concentrations are given as final concentrations in the 75-l assay.
Cloning, Expression, and Purification of Recombinant Proteins-Cloning, expression, and purification of the catalytic domains of PTP1B (26) and PTP␣ domain 1 (27) were done as described previously (28). The following PTP mutants were made by overlap extension polymerase chain reaction using appropriate restriction sites for cloning purposes (29): PTP1B to PTP␣, (i) R47V, D48N, C258M, and G259Q; (ii) R47V and D48N; (iii) C258M and G259Q; and (iv) G259Q; PTP␣ to PTP1B (PTP1B numbering): (v) V47R, N48D, M258C, and Q259G; (vi) M258C and Q259G; and (vii) Q259G. All constructs were inserted into pGEX expression vectors (Amersham Pharmacia Biotech). In addition, for x-ray protein crystallography the cDNA encoding the first 321 amino acids of PTP1B and the PTP1B (R47V,D48N,C258M,G259Q) mutant were inserted in the pET11a expression vector. All coding sequences were confirmed by DNA sequencing. Expression and purification of the glutathione S-transferase fusion proteins and the proteins for x-ray protein crystallography were done as described previously (28).
Determination of Kinetic Constants-The phosphatase activity was determined using the malachite green microtiter plate assay (30). The catalytic reaction was performed in half-area 96-well plates (Costar) with a final volume of 75 l. Both the reaction solution containing appropriate amounts of peptide and the diluted enzyme solution were temperature-equilibrated for at least 15 min at 30°C. The reaction was started by addition of PTP and terminated after 20 min by addition of 25 l of malachite green solution. The plates were then incubated for 60 min at room temperature with continuous shaking before measuring the absorbance in an enzyme-linked immunosorbent assay plate reader at 620 nm. A standard curve of KH 2 PO 4 used for calculating the release of inorganic phosphate in the samples was determined parallel to and on the same plate as the samples. Kinetic parameters were determined by a nonlinear least squares fit of the initial rates at various substrate concentrations to the Michaelis-Menten equation. In some of the kinetic experiments, the K m values of particular peptides were larger than the maximum substrate concentration used. When K m is much larger than the substrate concentration used, the Michaelis-Menten equation reduces to a first-order rate equation, where the catalytic efficiency can be obtained by linear least squares fitting of the rate of production versus substrate concentration. The reported standard deviations are calculated from at least four independent experiments using the nonbiased method.
Determination of Inhibitor Constants, K i -The enzyme reactions were carried out using standard conditions essentially as described by Burke et al. (31). The assay conditions were as follows. Diluted inhibitors (6 different concentrations, 2-fold dilution) were added to the reaction mixtures containing different concentrations of the substrate, p-nitrophenyl phosphate (p-NPP) (usual range, 0.16 -10 mM final assay concentration). The buffer used (total volume, 100 l) consisted of 100 mM sodium acetate, 50 mM sodium chloride, 0.1% (w/v) human serum albumin, 5 mM glutathione, and 1 mM EDTA, pH 5.5. The reaction was started by addition of the enzyme and carried out in microtiter plates at 25°C for 20 min. The reactions were stopped by addition of 1 N NaOH. The enzyme activity was determined by measuring the absorbance at 405 nm with appropriate corrections for absorbance of the compounds and p-NPP. The data were analyzed using nonlinear regression hyperbolic fit to classical Michaelis-Menten enzyme kinetic models. Inhibition is expressed as K i values in M.
X-ray Data Collection-Data collection was performed at 100 K. The following cryo conditions were used: to the sitting drop 3 l of 50% glycerol (containing 0.5 mmol of inhibitor) were added. The crystal was removed from the drop after 20 min and transferred to 50% glycerol (containing 0.5 mmol of inhibitor) for 5-10 s and flash-frozen. Data were collected using a mar345 image plate detector at the 711 beam line at the MAX-lab synchrotron facilities at Lund University (Sweden). A 1°o scillation was used for 60 images. A full data set to 2.1-Å resolution was obtained. The space group was determined to be P3121. Data processing was performed using Denzo, Scalepack, and the CCP4 program suite (32,33). For further details see Table I. Structure Refinements-As P3121 contains a polar axis and, thus, possesses more than one indexing possibility, a molecular replacement solution using Amore (33,34) was found prior to the refinements. A high resolution PTP1B structure was used as a starting model (28), where ligand and water molecules were omitted from the structure. Furthermore, alanine substitutions in the mutated positions were performed. All refinements were performed with Xplor. version 3.851 (Molecular Simulations Inc.). Interchanging cycles of model building using X-build (Molecular Simulations Inc.) and refinement were performed. The alanines for the mutated residues were substituted with the correct side chains and build into the 2F o Ϫ F c maps. The 2F o Ϫ F c maps were inspected by the use of X-ligand (Molecular Simulations Inc.) at a 1.3 level for densities that could correspond to the structure of the inhibitor. A well suited inhibitor electron density was identified in the active site pocket, see Fig. 6 below. No other densities were identified to fit the inhibitor. Water molecules were inserted using the X-solvate program (Molecular Simulations Inc.). For further details see Table I.

RESULTS
The goal of this study was to identify structural elements that are critically involved in determining the substrate specificity of PTPs. We have chosen to describe at the molecular level the difference in substrate recognition by two widely expressed PTPs, PTP1B and PTP␣. Both enzymes have been implicated as negative regulators of the insulin receptor tyrosine kinase, and hence they are potential therapeutic targets in type 2 diabetes. An understanding of the difference in substrate recognition should assist in designing selective PTP inhibitors. Whereas PTP1B seems to recognize a broad variety of synthetic tyrosine-phosphorylated substrates, PTP␣ shows limited recognition capability.
The Ac-DADE(pY)L-NH 2 Peptide Can Be Used to Discriminate between PTP1B and PTP␣-The synthetic peptide Ac-DADE(pY)L-NH 2 and analogs thereof have previously been used extensively to study substrate recognition by PTPs (36). In the present context it is of significance that this peptide is recognized well by PTP1B but not by PTP␣ (6,7,37) and that the binding mode of this peptide in PTP1B has been elucidated by x-ray crystallography (38). Therefore, the Ac-DADE(pY)L-NH 2 peptide offers a unique possibility for identifying key structural elements that determine the differences in substrate recognition by PTP1B and PTP␣. In accordance with previous studies, we found that PTP1B dephosphorylates this peptide about 30 times more efficiently than PTP␣ (Table II).
Residues 47, 48, 258, and 259 Define an Important Substrate Recognition Site-By using primary sequence alignments and C␣ regiovariation-score analyses, we have previously identified residues 47, 48, 258, and 259 (PTP1B numbering) as a potential selectivity determining area of PTPs. 2 To get initial insight into the significance of this region for substrate recognition, we first made a PTP1B mutant in which these four amino acid residues were substituted for the corresponding residues in PTP␣ (i.e. Arg-Asp-Met-Gly to Val-Asn-Cys-Gln). Table II shows that introduction of the putative selectivity-determining residues from PTP␣ into PTP1B caused a remarkable decrease in activity against the Ac-DADE(pY)L-NH 2 peptide. Importantly, xray crystallography of this mutant shows that the overall fold-ing and structure is similar to that of the wild type enzyme (see below). Thus, the observed differences in K m and k cat values are due to the introduction of the four "PTP␣ residues" into PTP1B thereby providing significant support for the view that these four residues are critically involved in the substrate recognition capacity and hydrolysis.
Residues 258 and 259 Are the Major Determinants for Selectivity-To dissect further the role of the individual amino acid residues in the putative selectivity-determining region, we next introduced either residues 47-48 or 258 -259 from PTP␣ into PTP1B. Table III shows that the replacement of residues 47-48 resulted in a 1.5-fold reduction in catalytic efficiency (k cat /K m ), whereas the introduction of residues 258 -259 reduced the overall catalytic efficiency about 4-fold. In comparison to the wild type enzyme, the lower catalytic efficiency for PTP1B (R47V,D48N) is due to the 1.4-fold reduction in the turnover number (Table III), whereas K m is apparently not affected by these mutations. In contrast, introduction of a cysteine (residue 258) and a glutamine (residue 259) resulted in an almost 2-fold increase in K m value. Thus, when comparing the PTP1B (R47V,D48N) and PTP1B (M258C,G259Q) mutants, it seems that the region comprising residues 258 -259 is the most important determinant for the catalytic efficiencies of these enzymes toward the DADE(pY)L peptide. Hence, we decided to focus our attention on residues 258 and 259.
Residues 258 -259 -The published x-ray structure of PTP1B co-crystallized with the DADE(pY)L-NH 2 peptide shows that the leucine residue (pYϩ1) is located on a hydrophobic region of the protein surface forming van der Waals contacts with the side chains of Val 49 , Ile 219 , and Gln 262 . In addition, a watermediated hydrogen bond was observed between the pYϩ1 amide group and the side chain of Gln 262 . Since these residues are positioned adjacent to Gly 259 and Met 258 , we reasoned that in particular introduction of the corresponding residues from PTP␣ into PTP1B could influence/impair the interaction of the pYϩ1 residue with PTP1B. Indeed, Cys 258 and Gln 259 seemed to be responsible for the limited catalytic activity against the Ac-DADE(pY)L-NH 2 peptide of the PTP1B (M258C,G259Q) mutant and possibly PTP␣ (Table III). It should be noted that previous studies with murine PTP␣ have shown that the k cat /K m value for DADE(pY)-NH 2 is about 3-fold higher than for DADE(pY)L-NH 2 , thus indicating a negative influence of the pYϩ1 residue (36). Therefore, to study in closer detail the influence of the pYϩ1 position on substrate recognition by PTP1B, we compared a series of Ac-DADE(pY)L-NH 2 analogs as shown in Fig.  2. Although some differences are observed, the wild type enzyme in general seemed to be influenced very little by the actual pYϩ1 residue, probably indicating that the contribution of this residue to the overall binding affinity is mainly due to the above water-mediated hydrogen bond to Gln 262 . By introduction of residues 258 -259 from PTP␣ into PTP1B, a significant increase in K m and a corresponding decrease in k cat /K m values were observed for all peptides. It is particularly noteworthy that the effects of the pYϩ1 residues seem to be independent of charge, suggesting that structural constraints are important in determining substrate specificity.
In contrast to PTP1B, the peptide side chain of the pYϩ1 residue had significant influence on substrate recognition by PTP␣ (Fig. 3). Thus, it was not possible to measure K m with an isoleucine or a threonine residue in the pYϩ1 position. It is of significance that the highest catalytic efficiency for PTP␣ was observed for the peptide without a residue in the pYϩ1 position, pointing to steric hindrance as a cause of the reduced substrate recognition in PTP␣. When Met 258 and Gly 259 were replacing Cys 258 and Gln 259 , these differences are less pronounced, i.e. with little influence of the nature of pYϩ1 side chain as in wild type PTP1B. Gln 259 Causes Steric Hindrance in PTP␣-The above results indicated to us that the architecture of the binding region 258 -259 causes steric hindrance and hence contributes to the general mechanism by which PTP␣ discriminates between dif-ferent substrates. We speculated that Gln 259 , due to its structural position and bulkiness, would be the major determinant for substrate recognition. For the substitution of Met 258 to Cys 258 , we would expect a less significant effect, since this side chain in the x-ray structure of PTP1B complexed with DADE(pY)L-NH 2 does not interact with the pYϩ1 side chain (38). Thus, in order to define unambiguously the role of residue 259, we made single mutants of PTP1B and PTP␣ in which residue 259 was exchanged. As seen in Fig. 4, introduction of a glutamine in position 259 in PTP1B gave rise to increases in K m values and a concomitant decrease in k cat /K m values similar to that of the double mutant shown in Fig. 2. Consistently, by replacing Gln 259 in PTP␣ with a glycine resulted in an enzyme with almost similar K m values as the double mutant PTP␣ (C258M,Q259G) for this series of synthetic peptide substrates (Figs. 3 and 4). Noticeably, the catalytic efficiency is almost comparable to that of PTP1B, irrespective of the residue in the pYϩ1 position in the substrate (Fig. 4    influence on the binding of peptides with a residue in the pYϩ1 position but also on the recognition of Ac-DADE(pY)-NH 2 (Figs. [2][3][4]. This indicated that Gln 259 indirectly would interfere with the binding of other parts of the peptide than the pYϩ1 residue, perhaps even with the tyrosine phosphate group itself. We hypothesized that the bulky side chain of Gln 259 could affect the positioning and the rotational flexibility of the side chains of structurally neighboring residues. By comparing the published apo structures of PTP1B (39) and PTP␣ (40), it became apparent that Gln 262 in PTP␣ was pointing into the active site cleft and thus seemed to impair the access of Tyr(P) substrates to the PTP signature motif (Fig. 5A). In contrast, Gln 262 in PTP1B points away from the active site thereby promoting substrate binding (Fig. 5B). In other words, it seems conceivable that the side chain of Gln 259 forces the Gln 262 side chain into a conformation that impairs substrate binding and/or hydrolysis. Protein X-ray Crystallography-To gain further insight at the structural level of the importance of Gln 259 /Gln 262 on inhibitor/substrate binding, we next attempted to co-crystallize PTP␣ with low molecular weight inhibitors. However, although we were able to obtain crystals that diffracted well, in all cases there was no inhibitor bound to the active site. Instead, as reported by Bilwes and co-workers (40), PTP␣ crystallized as a dimer with the so-called wedge inserted into the active site and probably therefore preventing binding of the inhibitors. Assuming that the PTP1B (R47V,D48N,M258C,G259Q) mutant could be used as a model for PTP␣ in the surrounding of Gln 259 , we then co-crystallized this mutant with compound 1, which inhibits PTP␣ with a K i value of ϳ80 M (see below). The mutated residues and the inhibitor are clearly defined in the electron density maps shown in Fig. 6. As shown in Fig. 6B, there are no direct interactions between the inhibitor and Gln 259 /Gln 262 . Most importantly, no changes were observed in the secondary or tertiary structure of the mutated enzyme validating the general concept of using mutants for enzyme kinetics analysis. After binding of compound 1, Gln 262 is found in a conformation pointing away from the active site pocket similar to the side chain conformation seen for apoPTP1B. This conformation is stabilized by a hydrogen bond to the backbone of Gln 259 and van der Waals interactions between these two residues (Fig. 6). We assume that similar structural constraints are at play in PTP␣. Therefore, it is likely that this arrangement in turn will increase the energy penalties for binding of substrates, which have bulky pYϩ1 residues. In addition, when the Gln 259 and Gln 262 conformations first are formed, this may impair the capability of the latter to position a nucleophilic water molecule correctly (see below). The side chain of Gln 259 is fairly flexible in the PTP1B mutant structure (average B-factor 41 Å 2 ), whereas Gln 262 is stable (average B-factor 20 Å 2 ). The average B-factor for the whole PTP1B mutant structure is 20 Å 2 .
Gln 262 in PTP␣-As indicated above, energy penalties could be expected for ligand binding in PTP␣ due to the positioning of Gln 259 /Gln 262 . In accordance with this notion, replacement of Gln 259 in PTP␣ with a glycine residue significantly decreases the K m values for all peptides analyzed (Fig. 4). However, since Gln 259 seems to have both negative direct (steric hindrance) and indirect (via Gln 262 ) effects on the binding and/or hydrolysis of peptide substrates, it is difficult to assess the contribution from Gln 262 . The above protein x-ray crystallographic analysis of compound 1 complexed with the PTP1B (R47V,D48N,M258C,G259Q) mutant (Fig. 6) shows that the minimal unit of this inhibitor (2-(oxalyl-amino)-thiophene-3carboxylic acid) is too small to be directly influenced by Gln 259 . We therefore reasoned that low molecular weight inhibitors representing the minimal unit could be used to semi-quantify the influence of Gln 262 on the accessibility of the active site pocket. Consequently, compounds 2-4 were tested against PTP1B, PTP␣, and the single mutants PTP1B (G259Q) and PTP␣ (Q259G) . It appears from Table V that introduction of Gln 259 into PTP1B caused a significant reduction in the affinity for all three minimal unit inhibitors, most conceivably due to a direct interference of Gln 262 with ligand binding. When Gln 259 in PTP␣ was replaced by a glycine, a substantial increase in affinity was observed for all inhibitors. A schematic representation of the proposed positioning of residues 259 and 262 in the apo structures of PTP1B and PTP␣ and during substrate binding and hydrolysis is shown in Fig. 7.

DISCUSSION
Intensive studies of the insulin signaling pathway have pointed to several PTPs as key regulators, including PTP1B, PTP␣, and PTP-LAR (18). It is generally believed that selective inhibitors of PTPs that negatively regulate insulin signaling could be useful in the treatment of diabetes. Using structurebased design we have recently developed a novel, low molecular weight, non-phosphorus, and highly selective inhibitor of PTP1B (21). This was achieved by introducing a basic nitrogen into a general PTP inhibitor leading to salt bridge formation with Asp 48 in PTP1B. In other PTPs, containing an asparagine in the equivalent position, the basic nitrogen causes repulsion. The net result is a marked selectivity for PTP1B. The present study was undertaken to expand further our knowledge regarding the structural requirements for substrate and inhibitor recognition by PTPs. In particular, our focus has been to delineate the structural basis for differences in substrate recognition by PTP1B and PTP␣.
The synthetic peptide DADE(pY)L (derived from the epidermal growth factor receptor) and analogs thereof have previously been used extensively to study substrate recognition of PTPs (6,7,41,42). Furthermore, structural information is provided by x-ray crystallography of the PTP1B-DADE(pY)L-NH 2 complex that defines the binding mode (38). Importantly, the DADE(pY)L peptide is recognized well by PTP1B but less efficiently by PTP␣ (36,37). Thus, this peptide seems an ideal tool to probe and identify the structural requirements involved in defining substrate specificity. By using a series of PTP mutants in which residues 47, 48, 258, and 259 were exchanged between PTP1B and PTP␣ in combination with a series of Ac-DADE(pY)L-NH 2 analogs, we here demonstrate that the key selectivity-determining residue is residue 259. In PTP␣ this residue is a glutamine, and in PTP1B it is a glycine. By replacing Gln 259 with a glycine in PTP␣, this enzyme exhibits the same broad substrate recognition capacity and catalytic activity as PTP1B. Conversely, introduction of glutamine in position 259 almost turns PTP1B into a PTP␣-like enzyme. Thus, glutamine in position 259 causes steric hindrance and limits the substrate recognition capability, whereas a glycine allows broad substrate recognition. In comparison, in the evolution of phosphatases steric hindrance is utilized to provide selectivity for phosphotyrosine. As noted by Barford and coworkers (39), the catalytic cleft is about 9 Å deep, which only allows phosphotyrosine, but not phosphoserine, to reach the active cysteine and bind to the PTP signature motif.
As indicated above, the Ac-DADE(pY)L-NH 2 peptide and analogs thereof have previously been widely used to study the substrate specificity of PTPs (6,7,10,42). However, these studies in most cases focused on the residues positioned Nterminally to the Tyr(P) residue. Consequently, little is known about the importance of residues located C-terminally to Tyr(P). In a detailed Ala scan of the longer Ac-DADE(pY)LIPQQG-NH 2 peptide, only a minor influence of changing the pYϩ1 from a leucine to an alanine was observed when testing against PTP1 and Yersinia PTP (7). In accordance with this, we observe little influence of the pY ϩ1 residue on PTP1B. In contrast, this residue dramatically influences the binding affinity to PTP␣. It is tempting to speculate that this is due to a steric clash with Gln 259 , since replacement of this bulky residue with a glycine causes a significant increase in affinity. By using low molecular weight PTP inhibitors and the substrate p-NPP, we have further provided evidence that Gln 259 indirectly may influence the position of other residue(s) in the active site of PTP␣. Thus, the K m value of p-NPP is about 5-fold higher in the PTP1B (G259Q) mutant than in the wild type enzyme. Similarly, the affinity of the low molecular weight inhibitors is significantly lower in this PTP1B mutant (Table V). Due to their small size, it is unlikely that their binding could be directly influenced by Gln 259 . We hypothesize that a glutamine in position 259 in addition to its direct influence on substrate binding also changes the position or limits the rotational freedom of other residues. In particular, the conserved Gln 262 found in all active PTPs seems an attractive candidate.
X-ray crystallographic studies indirectly provide support for the notion that Gln 259 influences the positioning of Gln 262 in PTP␣. Thus, structure analysis of apoPTP␣ shows that, in contrast to the apoPTP1B structure, the side chain of Gln 262 is pointing into the catalytic site and thereby impairing substrate recognition. In PTP1B complexed with Ac-DADE(pY)L-NH 2 , the side chain amide group of Gln 262 forms a hydrogen bond to the backbone carbonyl of Gly 259 (38). Thereby, a steric clash with the phenyl side chain of the tyrosine-phosphorylated peptide is avoided, and the active site is sufficiently open to easily accommodate tyrosine phosphate.
When PTP1B is complexed with peptides (38), Tyr(P) (38), and low molecular weight inhibitors (28), the Gln 262 side chain shows a similar conformation as that observed in the apo structure (39). In contrast, the side chain of Gln 262 has to move in PTP␣ to provide sufficient space for a substrate to bind. It is evident from the apoPTP␣ structure that the side chain of Gln 262 has limited rotational freedom due to the bulkiness of Gln 259 , and simultaneous conformational changes of Gln 259 and Gln 262 seem to be required for substrate binding. To gain further insight into the putative interaction between these two residues during ligand binding, we undertook co-crystallization of PTP1B (R47V,D48N,M258C,G259Q) with a low molecular weight inhibitor (compound 1). No changes were observed in the secondary or tertiary structure of the mutated enzyme. Therefore, it seems conceivable that the PTP1B (R47V,D48N,M258C,G259Q) mutant will mimic PTP␣ in relation to the influence of these four residues on substrate and inhibitor binding. The x-ray structure of PTP1B (R47V,D48N,M258C,G259Q) co-crystallized with compound 1 reveals that the side chain of Gln 262 simultaneously forms a hydrogen bond to the backbone of Gln 259 and van der Waals contact to the side chain, thereby stabilizing the side chains of both residues. Assuming that these conformational changes also take place in PTP␣ upon substrate or inhibitor binding, we speculate that the interaction between the glutamines further decreases their rotational freedom leading to an even more efficient blockage of this area of the enzyme (Fig.  7).
During hydrolysis of the cysteinyl-phosphate intermediate, Gln 262 in PTP1B swings into the catalytic site to position or activate a nucleophilic water molecule (43). Mutation of Gln 262 in PTP1B to an alanine reduces k cat by 100-fold and K m 10-fold (44). This indicates that the positioning of Gln 262 plays a major role for efficient hydrolysis. Since it is conceivable that Gln 262 in PTP␣ forms a hydrogen bond and van der Waals contacts to Gln 259 , this would implicate that Gln 262 moves less freely in PTP␣ than in PTP1B. We here demonstrate that replacement of Gly 259 with a glutamine in PTP1B causes a significant decrease in the k cat value, irrespective of the residue in the pYϩ1 position in the substrate (Fig. 4). Similarly the catalytic efficiency against p-NPP is decreased in this mutant (Table V).
Although not formally proven, we hypothesize that the low catalytic efficiency observed for wild type PTP␣ could in part be due to ligand-induced stabilization of Gln 259 and Gln 262 . In addition, Gln 259 may also impair correct positioning of Gln 262 relative to the above nucleophilic water molecule (Fig. 7).
The steric hindrance caused by residues in position 259 may potentially be useful in inhibitor design. Thus, inhibitors containing substituents addressing this area of the enzyme may be accommodated in PTP1B with a glycine in this position, but not by PTPs with bulky residues. In this context, it is of interest that Zhang and co-workers (45) utilizing a second aryl phosphate-binding site in PTP1B have recently made a series of phosphonate-based non-peptide inhibitors that showed a remarkable selectivity for PTP1B (46). The most important residues in this second binding site are Arg 24 and Arg 254 which were found to coordinate phosphate (45). Furthermore, additional points of interactions were found to include Met 258 and Val 49 . In other words, to bind simultaneously to the PTP loop and the second aryl phosphate-binding site, the inhibitors needed to "pass over" Gly 259 and Met 258 (46). Although no structural information was provided, we hypothesize that the high selectivity for PTP1B obtained with the above phosphonate-based inhibitors may, at least in part, be due to steric hindrance imposed by more bulky residues in position 259 in other PTPs.
In summary, we have shown that the residue in position 259 (PTP1B numbering) is a key determinant in substrate recognition capacity and hydrolysis by PTPs. PTPs with a bulky residue such as a glutamine show restricted substrate recognition, whereas PTPs with a glycine in this position have broad specificity. In PTP␣, Gln 259 further influences the positioning of Gln 262 . As a consequence, the ligand-induced stabilization of both residues leads to an increased efficiency of steric hindrance and a decreased catalytic efficiency. Thus, we propose that Gln 259 in PTP␣ plays a dual role leading to restricted substrate recognition and reduced catalytic rate. Both effects could indicate that PTP␣ is involved in the control of highly selective signal transduction processes.