Conformation-assisted Inhibition of Protein-tyrosine Phosphatase-1B Elicits Inhibitor Selectivity over T-cell Protein-tyrosine Phosphatase*

PTP-1B represents an attractive target for the treatment of type 2 diabetes and obesity. Given the role that protein phosphatases play in the regulation of many biologically relevant processes, inhibitors against PTP-1B must be not only potent, but also selective. It has been extremely difficult to synthesize inhibitors that are selective over the highly homologous TCPTP. We have successfully exploited the conservative Leu119 to Val substitution between the two enzymes to synthesize a PTP-1B inhibitor that is an order of magnitude more selective over TCPTP. Structural analyses of PTP-1B/inhibitor complexes show a conformation-assisted inhibition mechanism as the basis for selectivity. Such an inhibitory mechanism may be applicable to other homologous enzymes.

PTP-1B represents an attractive target for the treatment of type 2 diabetes and obesity. Given the role that protein phosphatases play in the regulation of many biologically relevant processes, inhibitors against PTP-1B must be not only potent, but also selective. It has been extremely difficult to synthesize inhibitors that are selective over the highly homologous TCPTP. We have successfully exploited the conservative Leu 119 to Val substitution between the two enzymes to synthesize a PTP-1B inhibitor that is an order of magnitude more selective over TCPTP. Structural analyses of PTP-1B/inhibitor complexes show a conformation-assisted inhibition mechanism as the basis for selectivity. Such an inhibitory mechanism may be applicable to other homologous enzymes.
The quest for agents that can intervene in type 2 diabetes continues to be a major research focus and challenge in many laboratories. As impaired insulin action is an underlying mechanism in type 2 diabetes, the insulin signaling pathway has naturally been the focus of research in attempts to identify suitable therapeutic target(s) for drug intervention against the disease. Insulin signaling begins with the activation of the insulin receptor (IR) via tyrosine phosphorylation and culminates in the uptake of glucose into cells by the glucose transporter, Glut4 (1). The activated IR must then be deactivated and returned to a basal state, a process that is believed to involve protein-tyrosine phosphatase-1B (PTP-1B). 4 PTP-1B has been shown to directly interact with the activated insulin receptor (2)(3)(4)(5)(6). Disruption of the gene that codes for PTP-1B in mice results in sensitivity to insulin and also increased resistance to diet-induced obesity (7,8). This phosphatase is therefore a very attractive therapeutic target for the treatment not only of type 2 diabetes but also of obesity. Predictably, there has been an intense research effort by many groups to identify potent and selective inhibitors of PTP-1B (for recent reviews, see Refs. 9 and 10). PTP-1B belongs to the protein-tyrosine phosphatase (PTP) superfamily of enzymes, which includes ϳ100 members involved in signal transduction and regulation of cellular processes such as growth, differentiation, and proliferation. Inhibitors against the enzyme must thus be not only potent but also selective. Sequence alignment of amino acids residues within the PTPs catalytic domain shows that, in general, all phosphatases have less than 40% identity to PTP-1B: a remarkable exception is the T-cell protein-tyrosine phosphatase (TCPTP), which has a 72% identity (11)(12)(13)(14). Inhibitors that target PTP-1B and incorporate previously identified selectivity determinants (15)(16)(17) are generally selective (Ͼ30-fold) over all PTPs tested but are equipotent on TCPTP. Although it is not evident that co-inhibition of TCPTP (with PTP-1B) will result in serious adverse effects, mice lacking the TCPTP gene die within 3-5 weeks after birth from defects in hematopoiesis and immune function (18). Hence, it is highly desirable to design PTP-1B inhibitors that are selective over TCPTP as potential therapeutic agents particularly as type 2 diabetes is a chronic disease. The only demonstrable structural determinant(s) of inhibitor selectivity between PTP-1B and TCPTP reported to date is located in the enzyme secondary binding site (16), and inhibitors targeting that area have been shown to have some discriminatory capability between the two enzymes (19 -23). Very recently, Wiesmann et al. (24) reported the discovery of a novel allosteric site that could be used for PTP-1B inhibition. Although the discovery of this site appears to be promising for the design of inhibitors with enhanced pharmacological properties, the issue of selectivity versus TCPTP remains partially unaddressed. The two enzymes share over 95% of sequence similarity in this area, and are structurally identical, although TCPTP lacks the phenylalanine at position 280 that appears to play a role in inhibitor binding. In order to further address the selectivity problem, we report here the identification of a previously unknown structural determinant between PTP-1B and TCPTP, and the biochemical and mutagenesis data that confirmed it. This new site may be exploited, alone or in conjunction with other selectivity determinants, to increase potency and elicit selectivity.
For TCPTP, the mutations were introduced into pFlag2-hTCPTP-(1-296) that was used as template DNA. The oligonucleotides used for introduction of the mutation that codes for the V121L substitution were as follows: 5Ј-GTGGAGAAAGAATCGCTGAAATGTGCACAGTAC-3Ј.
The correct construction of the mutant derivatives was verified by DNA sequencing. DNA was subsequently isolated from the positive clones and introduced into Escherichia coli BL21 cells for protein expression.
Protein expression was assessed by growing E. coli BL21 cells to an optical density of 0.7 at 600 nm in Luria Bertani broth supplemented with 100 g/ml ampicillin (LB-Amp) 2 . The culture was then induced by addition of isopropyl-1-thio-␤-D-galactopyranoside to a final concentration of 1 mM. Cells were harvested 3-h postinduction and the lysate analyzed by electrophoresis on a 10 -20% (w/v) polyacrylamide gradient gel containing sodium dodecyl sulfate (SDS-PAGE). The separated bacterially expressed proteins were transferred onto nitrocellulose and probed with an anti-FLAG monoclonal antibody by chemiluminescence. The Super Signal TM West Pico kit was used for the detection of proteins by following the suggested protocol.
Activity Assays-The FLAG-tagged catalytic domains of WT and mutant PTP-1B and TCPTP proteins were expressed in E. coli and purified as reported previously (15,25). The activity of all enzymes was assayed with fluorescein diphosphate (FDP) as substrate as previously reported (23,26). Kinetic constants were obtained by fitting the observed rates of reactions to the Michaelis-Menten equation with the aid of the non-linear curve fitting software program, Grafit 4.0.10 (Erithacus Software Inc.). Similarly, inhibitor concentrations that give 50% inhibition of enzyme activity (IC 50 ) were derived by fitting rates obtained in the presence and absence of inhibitor to a four-parameter equation. Under the conditions of the assay, all progress curves remained linear.
Synthesis of Inhibitors-The synthesis of compounds was carried out as described previously (22).
Crystal Structure Determination-Crystals of the isolated catalytic domains of WT PTP-1B and L119V mutant in complex with compound 1 were obtained as previously reported (23). The data sets were collected from a single crystal using synchrotron radiation at the beamline 17ID in the facilities of the Industrial Macromolecular Crystallographic Association Collaborative Access Team (IMCA-CAT) at the Advanced Photon Source (Argonne National Laboratory, Argonne, IL). Crystals were dipped for about 10 -30 s in cryoprotectant (15% MPD in mother liquor) before flash-freezing them in the liquid nitrogen stream. The data were integrated and scaled using X-gen (27). The WT PTP-1B:compound 1 structure was solved by Molecular Replacement using the software AMORE (28) and the coordinates from 1PTY (16), without ligand and solvent molecules, as search model. Refinement of the model was carried out by alternating cycles of manual rebuilding of the model in O (29) and computer-based refinement using CNX (Accelrys, San Diego, CA, Ref. 30). For all data sets, 5% of the data were flagged for R-free (31); bulk solvent correction was applied throughout the entire refinement, and the refinement was performed using the cross-validated maximum likelihood approach (32). NCS was applied at the initial stages of the refinement, and released toward the end. Typically, two cycles of torsion angle dynamics (33), positional and temperature factor refinement were run in each cycle. Omit maps were used to rebuild regions of the protein that were poorly defined in the initial maps. Table 1 summarizes the refinement statistics. For the structure of the mutant protein, initial A-weighted (34) 2 F o Ϫ F c and F o Ϫ F c difference electron density maps were calculated in CNX using phases derived from the WT PTP-1B:compound 1 structure, with ligand and solvent molecules removed, to check for the presence of the bound inhibitor. Refinement was then carried out essentially as described for the WT enzyme, but for the use of Refmac5 (35)(36)(37)(38) as the refinement program. Table 1 summarizes the statistics for data collection and structure refinement.
The coordinates and structure factors for the structures described here have been deposited with the Protein Data Bank, accession codes 2FJM and 2FJN, respectively.

RESULTS
Lead Compound Identification-Our initial screening campaign of thousands of compounds did not yield any inhibitors that were selective for PTP-1B over TCPTP. Similarly, efforts to identify differences between TCPTP and PTP-1B by enzyme kinetic analysis did not reveal any discernible differences between the two enzymes that could be exploited for inhibitor design to achieve selectivity (23). Furthermore, a structural comparison of the three-dimensional crystal structures of both enzymes did not reveal obvious determinants of selectivity within  the active site region. The structural comparison showed a high homology of 92% identity between the enzymes among the first shell residues (within 5 Å) of the inhibitor binding locus. Thus, we decided to re-assess our PTP-1B inhibitor data base to identify any compounds that showed a hint of selectivity between the two enzymes. Following a detailed and an in-depth re-evaluation of the data base, it became apparent that the presence of an ester group in a few inhibitors as exemplified by compound 1 in Fig. 1 resulted in a very modest selectivity over TCPTP. Re-evaluation of compound 1 showed a reproducible 2-fold selectivity between the enzymes. Subsequent structure-activity relationship studies revealed that the non-esterified derivative of the compound was not selective between the enzymes (compound 2, Fig. 1).
Identification of Selectivity Determinants-Using the crystal structure of PTP-1B in complex with other inhibitors as our guide (23), we hypothesized that the ester group would be likely positioned near the loop spanning residues 110 -121 hereafter referred to as loop 110 -121. We speculated that this disordered loop may serve as a potential source of selectivity, albeit at least ϳ6.4 Å away from the nearest chemical moiety examined. To identify the residue(s) responsible for the selectivity we aligned the amino acid sequences of PTP-1B and TCPTP to determine substitutions in this loop region. Four substitutions were found at positions 113, 114, 117, and 119 (PTP-1B numbering adopted). Valine 113 in PTP-1B is substituted with isoleucine in TCPTP; methionine 114 is replaced with a valine; glycine 117 is changed to glutamic acid, and leucine 119 is a valine. The amino acid changes found in TCPTP were then introduced into PTP-1B by site-directed mutagene-     sis. The recombinant proteins were purified to homogeneity and characterized kinetically. Because residues 113 and 114 are nearest neighbors and could exert a concerted effect the doubly substituted protein, V113I/M114V, was also constructed. The introduction of the amino acid substitutions did not appreciably alter the kinetic parameters of PTP-1B ( Table 2). The K m for FDP remained unchanged for all the mutant proteins, whereas the k cat was within 2-fold of the value obtained for the wild-type protein. To ensure the structural integrity of the loop, we also examined the effect of the amino acid substitutions on the inhibitory potency of a nonselective compound, BzN-EJJ-amide, whose binding orientation on PTP-1B has been reported and does not involve loop 110 -121 (39). No significant differences were observed between the loop mutants and wild-type PTP-1B with respect to inhibitory potency of BzN-EJJ-amide (Table 3). Thus, we conclude that the amino acid changes in loop 110 -121 do not significantly alter the overall conformation of the proteins. Next, we examined the effect of the amino acid substitutions on the potency of the modestly selective compound 1. As shown in Table 4, the mutant proteins maintained affinity for compound 1 with the exception of the L119V-substituted protein.
The leucine to valine amino acid substitution at position 119 resulted in a 2-fold loss in potency of compound 1, equivalent to the selectivity that was observed over TCPTP.

Crystal Structures of PTP-1B WT and L119V in Complex with
Compound 1-To further understand the involvement of Leu 119 in the esterdependent selectivity and provide a structural basis for its effects, we determined the crystal structures of the isolated catalytic domains of WT-PTP-1B and of the L119V mutant in complex with compound 1. Table 1 summarizes the statistics for the refined complexes. In both cases the protein crystallized in space group P2 1 2 1 2 1 , with two molecules per AU. In the WT structure, the inhibitor binds with the phosphonate in the primary binding site, or P-loop (colored in red in Fig. 2A), where the catalytic cysteine (Cys 215 ) is located. The benzotriazole moiety is located under the YRD loop (yellow in Fig. 2A), where it makes one hydrogen-bonding interaction with the main chain nitrogen of Asp 48 . The cinnamyl moiety binds toward the secondary binding site (as defined by Puius et al., Ref. 16; blue in Fig. 2A), and the benzyl ester is located in the proximity of the loop spanning residues 110 -121 (colored according to atom type in Fig. 2A). This loop, in most of the structures reported to date, was always held by interactions with symmetry-related molecules in an "open" and well ordered conformation. In the structure of BzN-EJJ-amide, the loop is not involved in interactions with other molecules, and appears to be very flexible (39). In the structures of several other inhibitors similar to compound 1 but with just a phenyl ring in proximity of the loop (23), it appears again to be disordered and/or to assume a different conformation from that observed in the 1PTY structure. However, none of these inhibitors show selectivity between PTP-1B and TCPTP as exemplified by compound 2 (see Fig. 1). In the compound 1 structure, the loop assumes two somewhat different conformations in the two molecules (Fig. 2B), but it appears in both cases to be in a "close" conformation, with the side chain of Leu 119 within van der Waals distance from the benzyl ester. The overall L119V mutant structure is extremely similar to the WT structure, and the inhibitor binds in the same fashion. The only difference is that in the mutant structure the loop 110 -121 is highly disordered, and density for the residues 117-119 was not available (Fig. 3B). This further suggests that the presence of Leu 119 is necessary for ordering of the loop.
The V121L Mutant in TCPTP-Crystal structures suggested a structural mechanism for the role of Leu 119 as a selectivity determinant between PTP-1B and TCPTP. To establish whether or not that residue alone was sufficient to elicit the ester-dependent selectivity between the two enzymes, we engineered the reciprocal Val to Leu substitution into TCPTP. As shown in Table 5 (left panel), substitution of Val at position  5 Leu to Val substitutions between PTP-1B and TC-PTP influence inhibitor selectivity 121 (equivalent to position 119 in PTP-1B) elicited selectivity between the derivative and the wild-type proteins. The modest 2-fold selectivity was retained between the two derivatives as was observed for PTP-1B indicating that the residue at position 119 (PTP-1B numbering) is necessary and sufficient to confer selectivity. Also shown in Table 5 (right  panel) is compound 3, a PTP-1B inhibitor that takes advantage of the loop differences and is an order of magnitude more selective over TCPTP. Absence of the oxadiazole group from this inhibitor abolishes selectivity between the two phosphatases (data not shown). In this case, the oxadiazole moiety also serves as a metabolically more stable replacement for the methyl ester of compound 1. As expected, substitution of Leu 119 to Val abolishes the selectivity observed with PTP-1B over TCPTP whereas the reciprocal V121L replacement engenders a similar effect in TCPTP. Thus, the single amino acid change from Leu to Val interconverts PTP-1B and TCPTP with respect to the ester and oxadiazole-dependent selectivity.

DISCUSSION
We have previously reported on structural determinants on PTP-1B that mediate selectivity over other PTPs with the exception of TCPTP (15). By exploiting those determinants it was possible to synthesize inhibitors that were Ͼ30-fold selective over other PTPs and three orders of magnitude more selective over cdc25A, a dual-specificity phosphatase. Here, we have reported the identification of a selectivity determinant between PTP-1B and the structurally highly homologous TCPTP by exploiting a previously unknown characteristic of PTP-1B inhibitors, the ability to induce an ordered conformation out of the otherwise unstructured loop 110 -121.
Our data suggest that the size of the inhibitor substituent near the loop influences the position of the loop itself. The side chain of Leu 119 clearly interacts with the bound inhibitor, providing some degree of binding energy (compound 2, the compound 1 analog without the ester has an IC 50 of 109 nM, versus the 39 nM of compound 1, Fig. 1). The favorable interaction may be explained by the fact that, although the entropic factor is unfavorable during loop stabilization, there is a gain in energy obtained by removing the mostly hydrophobic loop from the solvent and locking it into a more hydrophobic environment. Substitution of Leu with Val (as in the TCPTP sequence), probably has the effect of reducing the extent of interactions between the loop and the inhibitor, with consequent loss of binding affinity, and concomitant increase in the inhibitor's IC 50 . Comparison of the two structures (WT and L119V mutant) clearly indicate that a smaller side chain in position 119 is not able to exploit the same function as the Leu thus suggesting that Leu 119 is indeed the selectivity determinant (at least in this class of compounds) between PTP-1B and TCPTP. To confirm this, we engineered the reciprocal Val to Leu substitution into TCPTP. The prediction was that the reciprocal substitution should induce a similar 2-fold selectivity between wild-type TCPTP and its (V121L) derivative if that residue alone suffices to elicit selectivity and it did (Table 5).
In this report, we also demonstrated the ability to exploit the structural determinant to design a more selective PTP-1B inhibitor over TCPTP. Because of the obvious potential of this discovery, we have also looked closely at this region in all PTPs in the data base (data not shown). Of all the tyrosine-specific PTPs, PTP-1B is the only enzyme that carries Leu at position 119. Hence, it seems possible that even if other PTPs fold similarly to PTP1B, this region may still be exploited for selectivity. We hope our work will provide the framework for the design and synthesis of even more potent and selective PTP-1B inhibitors. As the structured loop 110 -121 is positioned fairly close to the active site of PTP-1B, targeting Leu 119 should facilitate the design and synthesis of selective inhibitors that are less bulky compared with inhibitors designed to target the secondary phosphotyrosine binding site denoted by Phe 52 (PTP-1B numbering). Such inhibitors will have a better chance of achieving oral bioavailability, another important issue associated with the design and synthesis of PTP-1B inhibitors as therapeutic agents. The identification of Leu 119 as a determinant of inhibitor selectivity was not possible from a strict structural comparison of the crystal structures of PTP1B and TCPTP alone (11). We propose "conformation-assisted inhibition" to describe such instances where the targeted structural feature is not readily identifiable from the crystal structure as it may not be properly induced and/or spatially aligned appropriately upon substrate (or analog) binding to influence inhibitor design. Our study is a testament to the remarkable power of combining protein engineering, medicinal chemistry and x-ray crystallography to structure-function relationships. We anticipate that conformation-assisted inhibition may be applicable to other structurally highly homologous proteins under similar circumstances.