Thermodynamic Study of Ligand Binding to Protein-tyrosine Phosphatase 1B and Its Substrate-trapping Mutants*

The binding of several phosphonodifluoromethyl phenylalanine (F2Pmp)-containing peptides to protein-tyrosine phosphatase 1B (PTP1B) and its substrate-trapping mutants (C215S and D181A) has been studied using isothermal titration calorimetry. The binding of a high affinity ligand, Ac-Asp-Ala-Asp-Glu-F2Pmp-Leu-NH2, to PTP1B (K d = 0.24 μm) is favored by both enthalpic and entropic contributions. Disruption of ionic interactions between the side chain of Arg-47 and the N-terminal acidic residues reduces the binding affinity primarily through the reduction of the TΔS term. The role of Arg-47 may be to maximize surface contact between PTP1B and the peptide, which contributes to high affinity binding. The active site Cys-215 → Ser mutant PTP1B binds ligands with the same affinity as the wild-type enzyme. However, unlike wild-type PTP1B, peptide binding to C215S is predominately driven by enthalpy change, which likely results from the elimination of the electrostatic repulsion between the thiolate anion and the phosphonate group. The increased enthalpic contribution is offset by reduction in the binding entropy, which may be the result of increased entropy of the unbound protein caused by this mutation. The general acid-deficient mutant D181A binds the peptide 5-fold tighter than the C215S mutant, consistent with the observation that the Asp to Ala mutant is a better “substrate-trapping” reagent than C215S. The increased binding affinity for D181A as compared with the wild-type PTP1B results primarily from an increase in the ΔH of binding in the mutant, which may be related to decreased electrostatic repulsion between the phosphate moiety and PTP1B. These results have important implications for the design of high affinity PTP1B inhibitors.

moval of the phosphoryl moiety from phosphotyrosine (Tyr(P)) and play important roles in signal transduction pathways that regulate cell proliferation, differentiation, migration, metabolism, and cell death (1)(2)(3). The PTPase superfamily is presently comprised of approximately 100 enzymes, which includes tyrosine-specific, dual specificity, and low molecular weight phosphatases (4). PTP1B is one member of this enzyme family, which in common with all PTPases has the active site signature motif C(X 5 )R(S/T) (5). Extensive biochemical and structural studies have led to an understanding of the mechanism by which PTPases catalyze phosphate monoester hydrolysis (4,6). PTPases share a common catalytic mechanism, utilizing a Cys nucleophile (Cys-215 in PTP1B) in the formation of a thiophosphoryl covalent enzyme intermediate (Fig. 1), which undergoes hydrolysis in a second step (7,8). The invariant Arg residue (Arg-221 in PTP1B) functions in substrate binding and in transition state stabilization (5,9). The initial phosphoryl transfer step is assisted by the conserved Asp (Asp-181 in PTP1B), which protonates the leaving group (10 -12), thereby acting as a general acid catalyst (Fig. 1).
Research into PTPases has relied heavily on the use of "substrate-trapping" mutants (13)(14)(15)(16)(17)(18)(19)(20)(21). Two types of substrate-trapping mutants have been used to isolate PTPase substrates. In the first, the active site Cys residue is replaced by a Ser, whereas in the second, the general acid Asp residue is substituted by an Ala. Although the Cys to Ser mutant has no measurable phosphatase activity (22) and the catalytic activity of the Asp to Ala mutant is reduced 10 5 -fold toward a protein substrate (21), both mutants retain the ability to bind substrates, enabling their use as "affinity reagents" for the isolation of in vivo PTPase substrates. Interestingly, in several instances, the Asp to Ala mutant has been found to be a better substrate-trapping reagent than the Cys to Ser mutant (17,21). Substrate-trapping mutants have also been employed extensively in structural studies of PTPases. In fact, most of the structural information regarding PTPase-substrate interactions has been obtained not from wild-type PTPases but from substrate complexes with the active site Cys to Ser substratetrapping mutants (23)(24)(25)(26). Although it is often assumed that substrate-trapping mutants retain the structural and binding properties of the wild-type PTPases, significant differences have been found in the few studies that have addressed this issue (22,27,28). Given the importance of PTPases in signal transduction, and the importance of substrate-trapping mutants in the studies of PTPases, a clear understanding of the molecular basis of the ability of these mutants to bind substrates is highly desirable. PTP1B has been implicated as a negative regulator of insulin signaling (29 -32). In addition, mice lacking functional PTP1B exhibit increased sensitivity toward insulin and are resistant to obesity (33,34). These results, taken together, suggest that specific inhibition of PTP1B may be therapeutically beneficial for the treatment of Type II diabetes and obesity. Thus, there is considerable interest in understanding the molecular basis of PTP1B-ligand interactions. An improved understanding of the interactions between PTP1B and its ligands (substrates and/or inhibitors) should facilitate the development of potent and selective PTP1B inhibitors.
We describe here a detailed thermodynamic study of ligand binding to PTP1B and its substrate-trapping mutants using isothermal titration calorimetry. The major advantage of isothermal titration calorimetry is that the association constant as well as the energetics of binding can be measured simultaneously. Differences in ligand binding affinity can be understood in terms of the relative contributions of the enthalpy change (⌬H) and the entropy change (⌬S) for binding of each ligand, thereby providing a thermodynamic basis for the ability of PTP1B and several mutants to bind ligands. Significant differences between the thermodynamic consequences of the interactions between various forms of PTP1B and a specific phosphonopeptide are reported herein. In addition, the energetics of binding of several ligands to a number of different PTP1B mutants is described, which permits an improved understanding of the observed differences and similarities in ligand binding affinity in terms of the relative contributions of the ⌬H and the ⌬S of binding for each ligand.

EXPERIMENTAL PROCEDURES
Materials-p-Nitrophenyl phosphate (pNPP) was purchased from Fluka Co. Other chemicals were from Fisher Co. Solutions were prepared using deionized and distilled water. The nonhydrolyzable Tyr(P) mimetic phosphonodifluoromethyl phenylalanine (F 2 Pmp)-containing peptides, Ac-Glu-F 2 Pmp-Leu-NH 2 , Ac-Asp-Ala-Asp-Glu-F 2 Pmp-Leu-NH 2 , and Ac-Ser-Ser-Val-Leu-F 2 Pmp-Thr-NH 2 , were synthesized using a modification of previously reported solid-phase technique (35), in which F 2 Pmp reagent (36) lacking phosphonate protection (37) was used. Final high pressure liquid chromatography-purified products provided NMR and mass spectral data consistent with their assigned structures.
Protein Expression and Purification-The catalytic domain of PTP1B (residues 1-321) was used in this study. Site-directed mutagenesis for C215S, D181A, D181N, and R47E was described previously (11,24,38,39). The recombinant wild-type and mutant PTP1Bs were expressed in Escherichia coli and purified to homogeneity as described (24,40). Protein concentration was determined from absorbance measurement at 280 nm using an absorbance coefficient of 1.24 for 1 mg/ml PTP1B.
Inhibition Study-The PTPase activity was assayed at 25°C in a reaction mixture (0.2 ml) containing an appropriate concentration of pNPP as a substrate. The reaction was initiated by addition of the enzyme and quenched after 2-3 min by addition of 1 ml of 1 N NaOH. The nonenzymatic hydrolysis of the substrate was corrected by measuring the control without the addition of the enzyme. The amount of p-nitrophenol produced was determined from the absorbance at 405 nm using a molar extinction coefficient of 18,000 M Ϫ1 cm Ϫ1 . The inhibition constants for the F 2 Pmp-containing peptides were determined for PTP1B in the following manner. At various fixed concentration of inhibitors, the initial rate at various pNPP concentrations was measured as described (40). All inhibition experiments were performed in 50 mM 3,3-dimethylglutarate buffer (pH 7.0). The ionic strength of each solution was adjusted to 0.15 M by addition of NaCl. Inhibition was competitive with respect to the substrate, and data were fit to Equation 1 using KINETASYST (IntelliKinetics, State College, PA) to obtain the inhibition constant (K i ): where V max is the maximal velocity, K m , the Michaelis constant, S, the substrate concentration, I, the inhibitor concentration, and K i , the inhibition constant. Isothermal Titration Calorimetry-All isothermal titration calorimetry experiments were performed using an MCS isothermal titration calorimetry system from Microcal Inc. (Northampton, MA). Experiments at pH 7.0 were conducted at 25°C, in 50 mM 3,3-dimethylglutarate buffer, containing 1 mM dithiothreitol. The ionic strength of the buffer was adjusted to 0.15 M by addition of NaCl. Protein concentration in the calorimeter cell was 27-85 M, whereas the ligand in the syringe was 0.37-1 mM. The PTP1B samples used in the isothermal titration calorimetry experiments were dialyzed completely against buffer. High concentration stock solutions were prepared for ligands with distilled water and adjusted to pH 7.0. Stock was diluted at least 26-fold with 50 mM 3,3-dimethylglutarate buffer before titration. Protein dilution during titration was determined by titration of buffer into the protein solution. The heat of protein dilution was found to be negligible. The heat of ligand dilution was corrected by subtracting the average heat of injection after saturation. The binding data were analyzed using Origin software (41). Binding constants, K, and enthalpy changes, ⌬H, were used to calculate free energy change, ⌬G, and entropy change, ⌬S, according to Equation 2: where R is the gas constant, and T is the absolute temperature. Calculation of Accessible Surface Areas-Accessible surface area (ASA) calculations were carried out using the program Delphi-Insight II (Biosym Technologies, Inc.). A 1.4-Å probe radius was employed. The structures used in this study have the following accession numbers in the Protein Data Bank at Rutgers University: PTP1B (2HNP); PTP1B/ C215S complexed with Asp-Ala-Asp-Gly-Tyr(P)-Leu-NH 2 (1PTU); PTP1B/C215S complexed with Ac-Asp-Glu-Tyr(P)-Leu (1PTT). To calculate the change in accessible surface area (⌬ASA) upon peptide binding, the surface areas of ligand-free PTP1B (ASA PTP1B ), free peptide (ASA peptide ), and PTP1B-peptide complex (ASA PTP1B-peptide ) were determined for both the hexapeptide and the tetrapeptide. The change in accessible surface area was then calculated from ⌬ASA ϭ ASA PTP1B ϩ ASA peptide Ϫ ASA PTP1B-peptide (42). ASA peptide was calculated as the sum of the accessible surface areas of the individual residues in the extended Ala-Xaa-Ala geometry. ASA PTP1B and ASA PTP1B-peptide were calculated from the corresponding crystal structures using coordinates for residues 5-282 in PTP1B to keep the ligand-free and ligand-bound form constant.

RESULTS AND DISCUSSION
Isothermal titration calorimetry allows a simultaneous determination of the binding constant (K), stoichiometry, as well as the enthalpy change (⌬H) associated with the binding of a ligand to a macromolecule (41). From these parameters, the Gibbs free energy of binding (⌬G) and the entropy change (⌬S) of binding can also be derived from the expression ⌬G ϭ ϪRTlnK ϭ ⌬H Ϫ T⌬S. Evaluation of these thermodynamic parameters yields important insight into the nature of binding reaction that is not evident from the measurement of binding constants alone (43,44). For example, from changes in the magnitude and the sign of the ⌬H and ⌬S and in conjunction with structural data provided by NMR and x-ray crystallography, one can obtain information about structural alterations that accompany ligand binding in terms of 1) changes in solvation state of the ligand or protein; 2) interactions between ligand and macromolecule, such as hydrogen bonding, dipoledipole, electrostatic, van der Waals forces, and hydrophobic interaction; and 3) changes in conformation/dynamics induced by ligand binding or by mutation. In the following, we describe a thermodynamic analysis of binding reactions between PTP1B and several site-directed mutants with phosphonodifluoromethyl phenylalanine (F 2 Pmp)-containing peptides using isothermal titration calorimetry.
The phosphopeptide Ac-Asp-Ala-Asp-Glu-Tyr(P)-Leu-NH 2 (modeled after the epidermal growth factor receptor autophosphorylation site Tyr-992, residues 988 -993) is an excellent substrate for PTP1B with a k cat /K m value (1.88 ϫ 10 7 M Ϫ1 s Ϫ1 ) 2100-fold higher than that of Tyr(P) (45). This indicates that amino acid residues flanking Tyr(P) contribute to efficient PTP1B binding and catalysis. The crystal structure of the catalytically inactive PTP1B/C215S complexed with Asp-Ala-Asp-Glu-Tyr(P)-Leu-NH 2 provides a structural snapshot of the interaction of PTP1B/C215S with this peptide 2 (23). However, to fully appreciate the interactions between a protein binding site and its ligands, a detailed thermodynamic description is a highly desirable complement to the structural data. F 2 Pmp-containing Peptides Are Excellent Nonhydrolyzable Substrate Analogs for PTPases-Because of inherent hydrolytic activity, it has not been possible to study binding interactions between a PTPase and its substrate directly. In fact, most structural studies have employed catalytically inactive, active site Cys to Ser mutant PTPases to visualize enzyme-substrate complexes (23)(24)(25)(26). To compare binding interactions between wild-type PTP1B and its various mutants (either active or inactive), we must use nonhydrolyzable phosphopeptide analogs. The most commonly used phosphorus-based Tyr(P) analogs are phosphonomethyl phenylalanine (Pmp) (45) and F 2 Pmp (46) (Fig. 2). Peptides bearing F 2 Pmp are over 1000 times more potent PTPase inhibitors than the analogous peptides containing Pmp (46,47). This has been attributed to a direct interaction between the fluorine atoms and PTPase ac-tive site residues (47). Specific interactions between one of the fluorine atoms in difluoronaphthylmethyl phosphonic acid and PTP1B have been observed (48). Biochemical and structural studies indicate that the phenolic oxygen in the substrate receives a proton from the general acid Asp-181 (10,11,23) (Fig.  1). The enhanced affinity of F 2 Pmp over Pmp likely arises from the ability of the fluorine atoms in F 2 Pmp to interact with the PTPase active site residues in a fashion analogous to that involving the phenolic oxygen and side chains in the active site of PTPases. Thus, F 2 Pmp-containing peptides are excellent nonhydrolyzable substrate analogs for PTPases.
Binding of Ac-Asp-Ala-Asp-Glu-F 2 Pmp-Leu-NH 2 to PTP1B-The F 2 Pmp-containing peptide Ac-Asp-Ala-Asp-Glu-F 2 Pmp-Leu-NH 2 is a nonhydrolyzable, competitive inhibitor of PTP1B with a K i value of 0.25 M at pH 7.0 and 25°C (Table I). This value is similar to a previous measurement of 0.18 M at pH 7.0 and 30°C (47). Unless stated otherwise, all experiments were performed at pH 7.0 and 25°C in this study. The dissociation constant (K d ) and thermodynamic parameters for the binding of Ac-Asp-Ala-Asp-Glu-F 2 Pmp-Leu-NH 2 to PTP1B were determined by isothermal titration calorimetry and are summarized in Table I. A typical titration curve and binding isotherm is shown in Fig. 3. Binding of Ac-Asp-Ala-Asp-Glu-F 2 Pmp-Leu-NH 2 was exothermic at 25°C (⌬H ϭ Ϫ3.9 kcal/mol). From curve fitting of such binding isotherms, the stoichiometry for the binding of the peptide to PTP1B was determined to be 1:1. The dissociation constant K d was 0.24 M, which is very close to the K i values determined from inhibition studies.
The ⌬G for binding of Ac-Asp-Ala-Asp-Glu-F 2 Pmp-Leu-NH 2 to PTP1B is Ϫ9.0 kcal/mol at pH 7.0 and 25°C. The association process between PTP1B and the F 2 Pmp-containing peptide is both enthalpically (⌬H ϭ Ϫ3.9 kcal/mol) and entropically (T⌬S ϭ 5.1 kcal/mol) favored. The relative magnitude of ⌬H is usually associated with ligand-protein interactions such as hydrogen bonding, ionic and positive van der Waals interactions, whereas the magnitude of the T⌬S term is associated with solvent reorganization and other entropic contributions to binding (49). In the crystal structure of PTP1B/C215S with Asp-Ala-Asp-Glu-Tyr(P)-Leu-NH 2 , the oxygen atoms of the phosphoryl moiety form an extensive array of hydrogen bonds with the main-chain nitrogens of the PTP loop (residues 215-221) and the guanidinium side-chain of Arg-221 ( Fig. 4 and Ref. 23). In addition to the electrostatic interactions between the phosphoryl moiety and the PTP loop in the active site, binding of phosphopeptides also include favorable ⌬H terms due to hydrogen bonding, ionic, and van der Waals interactions between amino acid residues of the peptide and PTP1B. The Tyr(P) phenyl ring is effectively buried within an active site cavity, formed by the nonpolar side chains of Ala-217 and Ile-219 of the PTP loop, Phe-182 of the WPD loop (residues 179 -189), and Tyr-46, Val-49, and Gln-262 (Fig. 4). Specific polar interactions between the enzyme and the peptide backbone stabilize the binding interface. Most notably, Asp-48 forms two hydrogen bonds to the main chain nitrogens of the Tyr(P) and the ϩ1 residues (Fig. 4). The main chain nitrogen of Arg-47 forms a hydrogen bond with the main chain carbonyl of the Glu(Ϫ2) in the peptide. In addition, the guanidinium group of Arg-47 forms salt bridges with the carboxylate groups at the Ϫ2 and Ϫ1 positions of the peptide substrate and a long hydrogen bond with the main-chain carbonyl at the Ϫ4 position (Fig. 4).
Hydrophobic interactions as well as other specific interactions between residues of the hexapeptide and PTP1B and the binding stabilized WPD loop closure are responsible for the burial of an extensive amount of total surface area (869 Å 2 , calculated using Delphi-Insight II, see "Experimental Procedures"). This value is similar to the calculated total buried surface area (902 Å 2 ) upon docking of Asp-Ala-Asp-Glu-F 2 Pmp-Leu, whose structure was determined by NMR-transferred nuclear Overhauser effects, to the active site of PTP1B (50). Thus, the large favorable T⌬S term (5.1 kcal/mol) may include the release from the interacting surface of water molecules to the bulk solvent, which provides the major driving force for the association of PTP1B with the peptide.
Binding of Ac-Glu-F 2 Pmp-Leu-NH 2 to PTP1B-The affinity of the tripeptide Ac-Glu-F 2 Pmp-Leu-NH 2 for PTP1B is only 6-fold less than Ac-Asp-Ala-Asp-Glu-F 2 Pmp-Leu-NH 2 (Table I). A comparison of the crystal structures of PTP1B/C215S with Ac-Asp-Glu-Tyr(P)-Leu-NH 2 and with Asp-Ala-Asp-Glu-Tyr(P)-Leu-NH 2 (23) reveals that the two hydrogen bonds between the Asp-48 carboxylate side chain and the main-chain nitrogens of the Tyr(P) and the ϩ1 residue, as well as interactions with the Tyr(P) moiety, are conserved in the two structures. One noticeable difference is the amount of buried surface area upon complex formation with PTP1B for the hexapeptide (869 Å 2 ) and the tetrapeptide (669 Å 2 ). Presumably, the tripeptide Ac-Glu-F 2 Pmp-Leu-NH 2 interacts with PTP1B in a similar mode as does the tetrapeptide and buries less surface area than 669 Å 2 . Thus, the greater than 200 Å 2 decrease in the buried surface area for the tripeptide may be responsible for the reduced T⌬S term for the binding of Ac-Glu-F 2 Pmp-Leu-NH 2 to PTP1B (Table I). Another noticeable difference between the two complexes is that the distance between the main-chain nitrogen of Arg-47 and the main-chain carbonyl of the Ϫ2 residue is shortened from 3.0 Å in the hexapeptide to 2.6 Å in the tetrapeptide. The large enthalpic gain for the binding of Ac-Glu-F 2 Pmp-Leu-NH 2 to PTP1B may be the result of a shorter and therefore stronger hydrogen bond (51) formed between the main-chain nitrogen of Arg-47 and the acetyl carbonyl group in the tripeptide. However, given the resolution of the structures (2.6 Å for the hexapeptide complex and 2.9 Å for the tetrapeptide complex), the observed difference in distance between the main-chain nitrogen of Arg-47 and the main-chain carbonyl of the Ϫ2 residue may be at the threshold of significance. Alternatively, it is also possible that the increased enthalpic contribution to binding for the tripeptide may be the result of elimination of unfavorable enthalpic interactions in the PTP1B/hexapeptide complex. However, this more favorable ⌬H is more than offset by the smaller T⌬S, which may result from a smaller surface area buried during complexation between the tripeptide and PTP1B.
Binding of Ac-Ser-Ser-Val-Leu-F 2 Pmp-Thr-NH 2 to PTP1B-We  also measured the binding of Ac-Ser-Ser-Val-Leu-F 2 Pmp-Thr-NH 2 to PTP1B. This sequence originated from the platelet-derived growth factor receptor residues 1005-1010. The rationale for choosing this peptide was to study the binding of a "nonspecific" peptide to PTP1B, which lacks acidic residues at the N terminus. However, it became clear during the course of this work that the Ϫ1 binding site in PTP1B can also accommodate a Leu residue (52). Furthermore, crystallographic structural analysis of peptide substrate-PTP1B/C215S complexes shows that the two hydrogen bonds between the Asp-48 carboxylate side chain and the main-chain nitrogens of the Tyr(P) and the ϩ1 residue, and a third between the main-chain nitrogen of Arg-47 and the main-chain carbonyl of the Ϫ2 residue, occur in all PTP1B/peptide interactions (23,26). In addition, the Tyr(P) moiety is engaged in similar interactions with PTP1B in all complexes. Thus, the binding affinity of Ac-Ser-Ser-Val-Leu-F 2 Pmp-Thr-NH 2 for PTP1B is only 7.5-fold lower than that of Ac-Asp-Ala-Asp-Glu-F 2 Pmp-Leu-NH 2 (Table I). This lower affinity apparently results from a reduction of the T⌬S term (Table  I). Because Ac-Ser-Ser-Val-Leu-F 2 Pmp-Thr-NH 2 lacks acidic residues at its N terminus, it is unable to engage in ionic interactions with the guanidinium group of Arg-47. Consequently, the distal N-terminal residues may not be able to make surface contact with PTP1B as extensively as does the acidic peptide. The reduced T⌬S term may be the result of decreased binding interface (less solvent displacement) between Ac-Ser-Ser-Val-Leu-F 2 Pmp-Thr-NH 2 and PTP1B.
Temperature Dependence of Ac-Asp-Ala-Asp-Glu-F 2 Pmp-Leu-NH 2 Binding to PTP1B-The temperature dependence of the binding of Ac-Asp-Ala-Asp-Glu-F 2 Pmp-Leu-NH 2 to PTP1B was studied over the range of 16 -30°C, with the thermodynamic parameters ⌬H , T⌬S, and ⌬G as a function of temperature being shown in Table II. It is clear that the association becomes increasingly enthalpically driven as the temperature increases. This pattern of temperature dependence is characteristic of hydrophobic association (53). The temperature dependence data can be used to determine the heat capacity change (⌬C p ϭ ␦⌬H /␦T) for the binding reaction. A plot of ⌬H versus temperature was linear in this temperature range (Fig.  5A). The slope of the line yields ⌬C p of Ϫ220 Ϯ 20 cal mol Ϫ1 deg Ϫ1 for the binding reaction between Ac-Asp-Ala-Asp-Glu-F 2 Pmp-Leu-NH 2 and PTP1B. The ⌬C p for the binding reaction is modest and is typical of a rigid body interaction that does not involve significant gross conformational change in the protein (54). This correlates with crystallographic studies of the free and bound states of PTP1B, which show conformational changes limited only to the WPD loop between the two states (23).
It is interesting to note that, although the ⌬H decreases as the temperature increases, the ⌬G is relatively invariant with temperature (Table II). Hence, ⌬H and ⌬S compensate to provide a relatively constant ⌬G. Fig. 5B shows an enthalpy/ entropy compensation plot for the interaction between the peptide and PTP1B. Similar compensation phenomena have been observed for many macromolecule interactions and are attributed to the unique properties of weak intermolecular interac-tions in aqueous solution (55,56).
Binding of F 2 Pmp-containing Peptides to PTP1B/R47E-The structure of PTP1B/C215S in complex with Asp-Ala-Asp-Glu-Tyr(P)-Leu-NH 2 revealed that the guanidinium group of Arg-47 forms both salt bridges with the carboxylate groups at the Ϫ2 and Ϫ1 positions of the peptide substrate and a long hydrogen bond with the main-chain carbonyl at the Ϫ4 position (Fig. 4). To test the importance of Arg-47 in PTP1B substrate recognition, Arg-47 was mutated to a Glu. It has been previously shown that the abrogation of the electrostatic interactions between Arg-47 and the acidic residues in the substrate led to an 8-fold decrease in k cat /K m for the substrate Asp-Ala-Asp-Glu-Tyr(P)-Leu-Ile-Pro-Gln-Gln-Gly. In addition, PTP1B/ R47E displays a greater decrease in k cat /K m than PTP1B/R47A for the same peptide, which is consistent with charge repulsion between the Glu residue at position 47 and the acidic residues in the substrate (39). Interestingly, the affinity for the binding of Ac-Asp-Ala-Asp-Glu-F 2 Pmp-Leu-NH 2 to PTP1B/R47E is 110-fold lower than the wild-type PTP1B (Table III). The major effect of the Arg-47 to Glu mutation on binding is a large decrease in the T⌬S term (Tables I and III). This may result from a reduction in the amount of buried surface area due to electrostatic repulsion between the acidic residues in the pep-  5. Temperature dependence of ligand binding to PTP1B. A, effect of temperature on ⌬H for the binding of Ac-Asp-Ala-Asp-Glu-F 2 Pmp-Leu-NH 2 to wild-type PTP1B. The solid line is a linear regression fit to the data. B, plot of ⌬H versus T⌬S for the binding of Ac-Asp-Ala-Asp-Glu-F 2 Pmp-Leu-NH 2 to the wild-type PTP1B. The solid line is a linear regression fit to the data. tide and Glu-47 in the mutant. To further probe the role of Arg-47 in peptide binding, we also determined the binding affinity of PTP1B/R47E for Ac-Glu-F 2 Pmp-Leu-NH 2 (Table  III), which is 4-to 12-fold lower than that of the tripeptide for the wild-type PTP1B. This clearly shows the importance of the interaction between Arg-47 and the Ϫ1 acidic residue in the peptide. As a control, we also measured the binding of Ac-Ser-Ser-Val-Leu-F 2 Pmp-Thr-NH 2 to PTP1B/R47E (Table III). Because Ac-Ser-Ser-Val-Leu-F 2 Pmp-Thr-NH 2 lacks the acidic residues that can engage in electrostatic interactions with Arg-47, it should bind wild-type PTP1B and the R47E mutant with equal affinity. This is indeed the case found (Table I and Table  III). Thus, electrostatic interactions between Arg-47 and the N-terminal acidic residues are important for high affinity binding between PTP1B and peptides. Because the kinetic parameter k cat /K m is an apparent second order constant that contains elements of both substrate binding and catalysis, it does not provide an accurate description of the intrinsic affinity between enzyme and substrate.
Binding of Ac-Asp-Ala-Asp-Glu-F 2 Pmp-Leu-NH 2 to PTP1B/ C215S-As discussed previously, the active site Cys residue in PTPases acts as a nucleophile to attack the substrate phosphorous atom. Although replacing this Cys residue with a Ser abolishes the PTPase activity, the mutant protein can still bind substrates, as it has been used in substrate-trapping experiments and x-ray crystal structural studies. However, it is not known how tightly the Cys to Ser mutant binds substrates as compared with the wild-type PTPase. The affinity of PTP1B/ C215S for Ac-Asp-Ala-Asp-Glu-F 2 Pmp-Leu-NH 2 and the thermodynamic parameters associated with binding were measured using isothermal titration calorimetry (Table IV). The dissociation constant of PTP1B/C215S for the peptide is 0.19 M, which is similar to that of the wild-type PTP1B (0.24 M). Thus, substitution of Cys-215 in PTP1B by a Ser does not change PTP1B's affinity for peptide substrates, even though the phosphatase activity was completely abolished. The Cys to Ser mutation also does not affect PTPase's ability to bind suramin, an active site-directed, competitive PTPase inhibitor (38). However, although the free energies of binding to the peptide of the wild-type PTP1B and the C215S mutant were found to be similar, significant differences in binding ⌬H and T⌬S were observed (Table IV). Unlike the wild-type PTP1B, for which peptide binding is driven by both favorable enthalpic and entropic terms, the ⌬H and T⌬S for peptide binding to PTP1B/ C215S are Ϫ10.4 and Ϫ1.2 kcal/mol, respectively, indicating that peptide binding is primarily driven by enthalpy with an unfavorable entropic contribution. Although the enthalpic contribution to the binding of peptide to PTP1B/C215S is 6.5 kcal/mole more favorable than the wild-type enzyme, the entropic contribution is disfavored by 6.3 kcal/mol. Thus the changes in enthalpy are not independent of changes in entropy. This is a hallmark of enthalpy/entropy compensation, which means that perturbations that increase the enthalpy can also increase the entropy, with little or no effect on the free energy (55).
Enthalpy/entropy compensation has also been observed for the binding of vanadate by the Yersinia PTPase C403S mutant (28). The dissociation constant of C403S for vanadate is 2.3 M, which is similar to that of the wild-type PTPase (1.3 M). However, the ⌬H and T⌬S for the binding of vanadate to the Yersinia PTPase are Ϫ10.0 and Ϫ2.0 kcal/mol, respectively, whereas those for the C403S mutant are Ϫ14.9 and Ϫ7.2 kcal/ mol, respectively. Thus the enthalpic contribution to the binding of vanadate for C403S is 4.9 kcal/mol more favorable than the wild-type enzyme, whereas the entropic contribution is disfavored by 5.2 kcal/mol. To understand the observed enthalpy/entropy compensation in ligand binding of the Cys to Ser mutant, it is important to point out that the side chain of the active site Cys exists as a thiolate anion at physiological pH (12,57). Thus, the Cys-215 to Ser mutation is not a simple substitution of OH for SH but rather a replacement of the negatively charged thiolate anion with a neutral hydroxyl group. The enhanced enthalpic contribution for the association of PTP1B/C215S with the peptide may be explained by the removal of the repulsive interaction between the thiolate anion and the negatively charged phosphonic acid moiety. The decrease in the entropic contribution to ligand binding by the Cys to Ser mutant may be the result of enthalpy/entropy compensation, i.e. perturbations that decrease the enthalpy (e.g. due to tightening-up of the system) can also decrease the entropy. Alternatively, the reduction in the value of T⌬S may also be the result of increased entropy of the unbound state (free protein) due to this mutation (see below).
The thiolate anion in the Yersinia PTPase is stabilized by hydrogen bonds between the backbone amides of the PTP loop, helix ␣5 (58 -61), and the side chains of His-402 and Thr-410 (57,62). In other words, the thiolate anion in the wild-type protein may be responsible for holding the PTP loop and the surrounding loops together (low entropy state). In the Cys to Ser mutant, structural and/or dynamic features that are important for the stabilization of the thiolate in the wild-type PTPase may be disrupted. The loss of the negative charge in the active site may perturb the electrostatic balance leading to more relaxed dynamics in the active site region (high entropy state). Thus, compared with the wild-type protein, the Cys to Ser mutant may be more flexible and therefore possesses higher entropy. Consistent with this, the Cys to Ser mutant of the Yersinia PTPase is considerable less stable (22) and experiences increased H/D exchange in the active site region relative to the wild-type protein (28). The overall negative T⌬S for the binding reaction between PTP1B/C215S and the peptide may thus be the result of a large reduction in protein entropy that more than offsets any positive entropic contribution derived from the liberation of water molecules into bulk solvent due to complex formation (Table IV).
Binding of Ac-Asp-Ala-Asp-Glu-F 2 Pmp-Leu-NH 2 to PTP1B/ D181A and PTP1B/D181N-The Asp-181 3 Ala mutant PTP1B has been found to trap substrates more efficiently than the Cys-215 3 Ser mutant (21). To provide a quantitative comparison with the Cys to Ser mutant, we have measured the binding of Ac-Asp-Ala-Asp-Glu-F 2 Pmp-Leu-NH 2 by PTP1B/ D181A (Table IV). PTP1B/D181A binds the peptide 5-fold tighter than PTP1B/C215S. Similarly, both PTP1B/D181A and Yersinia PTP/D356A displayed higher affinity (5-fold) than the wild-type enzymes toward suramin (38). These are consistent with the observation that the Asp to Ala mutant PTPase is a better "substrate-trapping" reagent than the active site Cys to Ser mutant for the identification of physiological PTPase substrates in vivo.
The affinity of PTP1B/D181A for the peptide is also 6-fold higher than that of the wild-type PTP1B (Table IV). An examination of the thermodynamic parameters in Table IV shows that the increased binding affinity for the general acid-deficient mutant as compared with the wild-type enzyme results primarily from the increase in the ⌬H of binding in the mutant. In the crystal structures of PTP1B/C215S substrate complexes (23), the carboxyl group of Asp-181 makes a hydrogen bond with the phenolic oxygen of Tyr(P) and another one with a phosphoryl oxygen through a water molecule (Fig. 4). Presumably, the F 2 Pmp moiety interacts with PTP1B in a similar fashion as Tyr(P) does. The mutation of Asp-181 to an uncharged Ala will result in a decrease in negative charge at this position. As suggested previously, this decrease in negative charge would reduce the electrostatic repulsion between the phosphate moiety and PTP1B and increase binding affinity (21). Our thermodynamic analyses are fully consistent with this suggestion. Furthermore, because Asn lacks the negative charge but retains the hydrogen bonding potential, one would predict that the affinity of PTP1B/D181N for the peptide should be in between those of the wild-type and the Asp to Ala mutant. Indeed, the affinity of PTP1B/D181N for the peptide was determined to be 2-fold higher than wild-type and 3-fold lower than PTP1B/D181A (Table IV).
Implications for PTPase Inhibitor Design-It has been shown that potent PTPase inhibitors can be obtained when the F 2 Pmp moiety is incorporated into an appropriate template. However, there is a concern that the dianionic nature of the phosphonate group may compromise its ability to cross cell membranes. On the other hand, there is also worry that elimination of the negative charges from the phosphonate may also decrease its affinity for the enzyme. The results described above show that the active site Cys to Ser mutant PTPases display enhanced enthalpic contribution for the binding of negatively charged ligands (e.g. F 2 Pmp-containing peptides, suramin, and vanadate), possibly because of the removal of the repulsive interaction between the thiolate anion and the negatively charged phosphate mimics. By the same token, a properly functionalized phosphate surrogate with less or no negative charge when attached to an appropriate aromatic framework would be expected to effectively occupy the native PTPase's active site. Indeed, high affinity nonphosphate-containing small molecule PTP1B inhibitors have been reported recently (63).
Summary-In the present paper, we report the results of a detailed study of binding thermodynamics of a number of F 2 Pmp-containing peptides to PTP1B and several mutant forms using isothermal titration microcalorimetry. In addition to the determination of energetics of ligand binding, this study leads to a more enhanced understanding of the molecular basis of PTP1B substrate recognition. The binding of the high affinity ligand Ac-Asp-Ala-Asp-Glu-F 2 Pmp-Leu-NH 2 to PTP1B is favored by both enthalpy and entropic contributions. Significant differences between the thermodynamic consequences of the interactions between the peptide and various forms of PTP1B have been determined. Elimination of the ionic interactions between the side chain of Arg-47 and the N-terminal acidic residues reduces the binding affinity primarily through the reduction of the T⌬S term. The role of Arg-47 is to maximize surface burial upon complex formation, which contributes to high affinity binding. Substitution of the general acid Asp-181 by a neutral residue improves the binding affinity of the mutant for the ligands, possibly due to the reduction of chargecharge repulsion between the carboxylate and the phosphate group. The fact that D181A binds Ac-Asp-Ala-Asp-Glu-F 2 Pmp-Leu-NH 2 5-fold more tightly than the C215S mutant supports the conclusion that it is a superior substrate-trapping reagent. PTP1B/C215S binds ligands with the same affinity as the wildtype enzyme. However, unlike the wild-type PTP1B, peptide binding to C215S is predominately driven by enthalpy change, which likely results from the elimination of the electrostatic repulsion between the thiolate anion and the phosphonate group. The increased enthalpic contribution is offset by reduction in the binding entropy change, which may be the result of increased entropy of the unbound protein due to this mutation. This study also demonstrates how structural and dynamic data can be complemented by thermodynamic measurements. The thermodynamic data support conclusions reached from spectroscopic experiments (22,27) and H/D exchange experiments (28), that the conformation and/or dynamic properties of the Cys to Ser mutant is different from those of the native PTPase.