Identification and structure–function analyses of an allosteric inhibitor of the tyrosine phosphatase PTPN22

Protein-tyrosine phosphatase nonreceptor type 22 (PTPN22) is a lymphoid-specific tyrosine phosphatase (LYP), and mutations in the PTPN22 gene are highly correlated with a spectrum of autoimmune diseases. However, compounds and mechanisms that specifically inhibit LYP enzymes to address therapeutic needs to manage these diseases remain to be discovered. Here, we conducted a similarity search of a commercial database for PTPN22 inhibitors and identified several LYP inhibitor scaffolds, which helped identify one highly active inhibitor, NC1. Using noncompetitive inhibition curve and phosphatase assays, we determined NC1's inhibition mode toward PTPN22 and its selectivity toward a panel of phosphatases. We found that NC1 is a noncompetitive LYP inhibitor and observed that it exhibits selectivity against other protein phosphatases and effectively inhibits LYP activity in lymphoid T cells and modulates T-cell receptor signaling. Results from site-directed mutagenesis, fragment-centric topographic mapping, and molecular dynamics simulation experiments suggested that NC1, unlike other known LYP inhibitors, concurrently binds to a “WPD” pocket and a second pocket surrounded by an LYP-specific insert, which contributes to its selectivity against other phosphatases. Moreover, using a newly developed method to incorporate the unnatural amino acid 2-fluorine-tyrosine and 19F NMR spectroscopy, we provide direct evidence that NC1 allosterically regulates LYP activity by restricting WPD-loop movement. In conclusion, our approach has identified a new allosteric binding site in LYP useful for selective LYP inhibitor development; we propose that the 19F NMR probe developed here may also be useful for characterizing allosteric inhibitors of other tyrosine phosphatases.


Protein-tyrosine phosphatase nonreceptor type 22 (PTPN22)
is a lymphoid-specific tyrosine phosphatase (LYP), and mutations in the PTPN22 gene are highly correlated with a spectrum of autoimmune diseases. However, compounds and mechanisms that specifically inhibit LYP enzymes to address therapeutic needs to manage these diseases remain to be discovered. Here, we conducted a similarity search of a commercial database for PTPN22 inhibitors and identified several LYP inhibitor scaffolds, which helped identify one highly active inhibitor, NC1. Using noncompetitive inhibition curve and phosphatase assays, we determined NC1's inhibition mode toward PTPN22 and its selectivity toward a panel of phosphatases. We found that NC1 is a noncompetitive LYP inhibitor and observed that it exhibits selectivity against other protein phosphatases and effectively inhibits LYP activity in lymphoid T cells and modulates T-cell receptor signaling. Results from site-directed mutagenesis, fragment-centric topographic mapping, and molecular dynamics simulation experiments suggested that NC1, unlike other known LYP inhibitors, concurrently binds to a "WPD" pocket and a second pocket surrounded by an LYP-specific insert, which contributes to its selectivity against other phosphatases. Moreover, using a newly developed method to incorporate the unnatural amino acid 2-fluorine-tyrosine and 19 F NMR spectroscopy, we provide direct evidence that NC1 allosterically regulates LYP activity by restricting WPD-loop movement. In conclusion, our approach has identified a new allosteric binding site in LYP useful for selective LYP inhibitor development; we propose that the 19 F NMR probe developed here may also be useful for characterizing allosteric inhibitors of other tyrosine phosphatases.
Selective inhibition of protein-tyrosine phosphatases (PTPs) 4 has the potential to be developed as a new therapeutic strategy for the treatment of many human diseases, including cancer, inflammation, diabetes, Alzheimer's disease, and autoimmune diseases (1)(2)(3)(4)(5). However, few selective PTP inhibitors have been developed because of their highly conserved active site, which includes a 9-Å-deep and 6-Å-wide phosphotyrosine-binding pocket surrounded by conserved catalytic residues. In the past, selective inhibitors of several classical PTPs, including PTP1B and LYP, were created by targeting both the phosphotyrosinebinding site and an adjacent site (6 -12). Alternatively, noncompetitive inhibitors have been recently identified for the treatment of several important disease-related PTPs, including PTP1B (2,13) and SHP2 (1,14), as a way to achieve selectivity and circumvent the conserved nature of the PTP phosphotyrosine catalytic site. These inhibitors regulate phosphatase activity through allosteric regulation and exhibit great potential for therapeutic development against cancer or diabetes.
A member of the PTP family, PTPN22, also called LYP, is exclusively expressed in hematopoietic immune cells. Developing specific inhibitors toward LYP has raised tremendous interest in the autoimmunity therapeutics community, and as a gain-of-function mutant of LYP, the R620W mutant (encoded by the C1858T SNP), has been associated with many autoimmune diseases, including type 1 diabetes (15, 16), systemic lupus erythematosus (17,18), myasthenia gravis (19 -21), rheumatoid arthritis (22, 23), and Graves' disease (24). Conversely, impairing the phosphatase activity of LYP was found to reduce the risk of several autoimmune diseases, including systemic lupus erythematosus (25), ulcerative colitis (26), and rheumatoid arthritis (27). With remarkable efforts, several potent inhibitors that selectively target LYP without inhibiting other phosphatase members have been recently identified by our group and by others (3,9,11,12,28,29). However, most of these compounds inhibit LYP via a competitive mode, and an alternative allosteric inhibition mechanism for LYP that can fulfill therapeutic demands has not yet been discovered.
Here, we describe the identification of an allosteric LYP inhibitor (NC1) that was obtained via structural modifications of our previously reported competitive LYP inhibitor (i.e. A15 analogues). Importantly, NC1 displayed a noncompetitive mode of LYP inhibition, showed selectivity in a panel of other phosphatases, and inhibited LYP activity in T cells. Further mechanistic study revealed that NC1 concurrently bound to a "WPD" pocket adjacent to the classic phosphotyrosine-binding site and to a unique LYP-specific insert that accounted for its selectivity. Moreover, we used our newly developed unnatural amino acid F2Y incorporation technology and 19 F NMR spectroscopy to provide direct biophysical evidence for the allosteric mechanism underlying the noncompetitive inhibition of LYP by NC1, in which the compound restricts the closure of the catalytic WPD-loop.

Identification of NC1 as a noncompetitive LYP inhibitor with selectivity against a panel of phosphatases
Our recent efforts using target-ligand interaction-based virtual screening identified a series of competitive LYP inhibitors (28). To explore the diverse chemotypes underlying LYP inhibition, we performed hit-based similarity search of commercial database based on our previously published compound A15 (28) and identified a new scaffold (2-iminothiazolidin-4-one) for LYP inhibition (Fig. 1A). Subsequently, 10 compounds were purchased from the SPECS database and examined by NMR and MS (Figs. S1 and S2). We then assessed their abilities to inhibit the LYP-catalyzed hydrolysis of p-nitrophenyl phosphate (pNPP) ( Table S1). The most active compound, NC1 (Fig.  1B), showed LYP inhibitory activity (K i ϭ 4.3 M) that was comparable with the original compound A15 (K i ϭ 2.87 M). Interestingly, analysis of the inhibition kinetics of NC1 unambiguously indicated a noncompetitive inhibition mode toward LYP (Fig. 1C), which was different from the competitive inhibition mode of the original compound A15. The dialysis analysis and reversible binding assays confirmed that the NC1 is a noncovalent revisable LYP inhibitor (Figs. S3 and S4). Consistently with the enzymology analysis, where the interactions between compound A15 and the LYP-active site are stable during the 20-ns MD simulation, the compound NC1 could not form stable interactions with LYP and dissociate from the active site (Fig. S5). Taken together, both biochemical and computational results suggested the noncompetitive binding mode of compound NC1 to LYP.
Further enzyme inhibition tests indicated that NC1 displayed at least 1.9-fold higher selectivity against a panel of other protein phosphatases, including STEP, PTPN18, Glepp, VHR, and the Ser/Thr phosphatases PPM1A and PP1, among others (Table 1). Unlike its inhibition of LYP, NC1 exhibited a competitive inhibition mode toward other tested phosphatases (Fig. S6).

NC1 enhances TCR signaling in lymphoid T cells
We next examined the ability and specificity of NC1 to inhibit LYP activity in cellular contexts. Downstream of TCR activation in lymphoid T cells, LYP negatively regulates the phosphorylation levels of ERK at the pThr-202/pTyr-204 site and of LCK at the pTyr-394 site. As shown in Fig. 2, TCR activation in Jurkat T cells significantly increased the phosphorylation levels of ERK and LCK, which were substantially augmented by the application of 20 M NC1. Importantly, knockdown of LYP by siRNA increased both the phosphoryla-

Allosteric inhibition of PTPN22 by locking two enzyme sites
tion of ERK and LCK to a similar extent solely for administration of NC1 (Fig. 2, A-C, and Fig. S7). Moreover, without endogenous LYP expression in T cells, the NC1 shows no effect on CD3-induced LCK and ERK phosphorylation in T cells (Fig.  2, A-C, and Fig. S7). Taken together, these results suggest that compound NC1 effectively and specifically inhibited LYP-mediated TCR signaling in T cells.

Identification of a novel interaction mode of NC1 with LYP by mutagenesis and simulation analyses
To dissect the molecular mechanism underlying the inhibition of LYP by NC1, a panel of LYP mutants with mutations located on the LYP catalytic surface ( Fig. 3A and Fig. S8) was selected according to our previously published crystal structures of LYP (12,30). Six out of nine mutations were found to increase the K i values of NC1 toward LYP by more than 1.5-fold (Fig. 3B). Despite the conserved nature of the PTP catalytic pocket, the combination of four of the mutated residues (His-196, Asp-197, Phe-28, and Thr-36) is unique among all PTPs, as indicated by sequence alignment (Fig. 3C). This unique pattern contributes to the selectivity of NC1 for LYP over other phosphatases. In particular, the LYP-specific insert is a unique PTPN22 sequence that is not shared by other phosphatases, as revealed by our previous crystallographic studies (12). More interestingly, two key residues, His-196 and Asp-197, are located in the WPD-loop, and one residue, Thr-36, is located in the LYP-specific insert 35 STKYKADK 42 . The WPD-loop harbors the essential catalytic residue Asp-195, which moves more than 6 Å after substrate binding (12,31,32) to coordinate the stabilization of the leaving group after the phospho-ester bond is broken during catalysis. Notably, the LYP-specific insert is on the other side of the WPD-loop relative to the substrate phosphotyrosine-binding pocket. Therefore, the concurrent interaction of NC1 with both the LYP-specific insert and the WPDloop suggests a unique binding mode of NC1 with LYP that is dissimilar to traditional inhibitor binding to the substratebinding pocket (10,33).
To predict the allosteric binding mode of NC1 to LYP, we analyzed all possible binding pockets around the four key residues using AlphaSpace, a fragment-centric topographic mapping program. To deal with the protein flexibility of LYP, eight available crystal structures of LYP were used in our pocket analysis (3,9,12,30,34). Three different WPD-loop conformations (closed, atypical-open, and open) were found in the eight crystal structures, and the LYP-specific insert can exist in either ␣-helix or loop conformations (Fig. S9). The docking results suggested that NC1 is indeed able to concurrently target the "WPD pocket" and the LYP-specific insert and to interact with the four key residues in the crystal structure 3H2X, which possesses an "atypical-open" WPD-loop conformation ( Fig. S9D) (34,35).
Eight representative docked poses of NC1 that bind the predicted allosteric pockets in opposite directions were selected and subjected to molecular dynamics (MD) simulations to evaluate their binding stabilities ( Fig. S10A). One docked pose of NC1 remained in the predicted allosteric pockets during the 50-ns MD simulation; this pose is recognized as the allosteric binding mode of NC1 ( Fig. 4A and Fig. S10C), whereas others could not bind tightly to LYP, and NC1 moved out of the initial pocket (Fig. S10, B and C). Individual residue contribution to the binding of compound NC1 with LYP was further calculated by the MM/GBSA binding free energy decomposition analysis (Fig. 4B). Residues Arg-266 and Thr-36 were found to contribute substantially to the binding energy of NC1 to LYP. Consistently, the T36E/R266A double mutation showed an 8.2-fold decrease in its binding ability to NC1 (Fig. 3B). According to the well-recognized general acid-base catalysis mechanism, residue Asp-195, which is located in WPD-loop, works as the general acid. Therefore, the conformation of the WPD-loop plays key roles in determination of the activity of LYP. To further examine the allosteric inhibition mechanism of NC1, we measured the distance between Asp-195 and the small artificial substrate pNPP by MD simulation of the LYP-pNPP system with or without NC1 bound. As shown in Fig. 6, the MD analysis suggested that the binding of NC1 in the allosteric pocket, which is located between the substrate-binding pocket and the WPD-loop (Fig. 5C), may block the closure of the WPD-loop and thereby lock LYP in an inactive conformation.
To further understand why NC1 displayed a different inhibition mode for LYP compared with other protein phosphatases (Table 1), we performed a pocket analysis using the crystal structures of PTP1B, STEP, PTPN18, and Glepp, which all have crystal structures with an "open" WPD-loop conformation (35,36). Interestingly, we detected a similar "WPD pocket" in the PTP crystal structures that possess "open" conformations for their WPD-loops (Fig. 6). However, the "secondary pockets" were detected in PTP1B, PTPN18, and Glepp, and they were less connected to their "WPD pockets" than that of LYP. Thus, the binding of NC1 to both the "WPD pocket" and "secondary pocket" provides a potential structural basis for its different inhibition mechanism with LYP compared with its mechanisms with other protein phosphatases.

F NMR spectroscopy reveals a noncompetitive mechanism underlying LYP inhibition by NC1
We next examined the dynamic conformational changes of the WPD-loop using our recently developed unnatural amino acid F2Y incorporation together with 19 F NMR technology (37,38). The unnatural amino acid incorporation causes the fewest structural perturbations and maintains better protein structural integrity than traditional chemical labeling, whereas 19 F NMR is an excellent tool for examining the Table 1 Selectivity of NC1 against a panel of protein phosphatases All measurements were made by using pNPP as a substrate at pH 7.0, 25°C, and ionic strength of 0.15 M. For all statistical analyses, data from at least three independent experiments were quantified and presented as the mean Ϯ S.D.

Allosteric inhibition of PTPN22 by locking two enzyme sites
conformational rearrangement of proteins with higher molecular weights (39,40). Residue Leu-281 was selected as the F2Y incorporation site to generate a 19 F NMR probe to detect WPD-loop dynamics, as this residue is buried by the WPD-loop in the absence of substrate (2P6X) (35) but is substantially exposed after substrate binding (2QCJ) (Fig. 7A) (12). We then mutated residue 281 to an amber stop codon and co-transfected the LYP mutant plasmid with the pEVOL-F2YRS plasmid, which encodes specific Methanocaldococcus jannaschii tyrosyl amber suppressor tRNA/tyrosyl-tRNA synthase mutants, into Escherichia coli strain BL21 and cultured it in medium containing F2Y (Fig. 7B). After purification, we obtained ϳ95% pure L281F2Y-LYP, and MS analysis unambiguously identified the incorporation of F2Y at position 281 ( Fig. 7C and Fig. S11). The L281F2Y incorporation did not perturb the overall LYP structure, as it showed similar activity toward a phosphopeptide substrate (Fig. S12). We then used 19 F NMR to monitor WPD-loop movement in response to the binding of the phosphate mimic Na 3 VO 4 with or without compound NC1. A 0.56 ppm upfield shift was detected after incubation of Na 3 VO 4 with the LYP-L281F2Y probe (Fig. 7D). In contrast, the Na 3 VO 4 -induced upfield shift was reduced to 2/3 of its original value (Ϯ0.39 ppm) following preincubation with NC1, indicating suppressed movement of the WPD-loop after NC1 incubation (Fig. 7D). The NMR results provided direct biophysical evidence that NC1 noncompetitively inhibits LYP by restricting the movement of the WPD-loop, in agreement with the data obtained from the mutagenesis analysis and the MD simulations.

Discussion
The development of selective inhibitors of specific PTPs has been hampered by the fact that their conserved active site is shared by most PTP family members. In the past, potent and selective PTP inhibitors have been developed by targeting nonconserved second-layer residues close to the active site (33) or by simultaneously binding to both the active site and a second pocket in the vicinity (9,10,12). Alternatively, inhibitor selectivity can be achieved via allosteric regulation by targeting a pocket outside of the catalytic center. Such allosteric inhibitors have been identified for PTP1B (2,13) and CD45 (41) and more recently for SHP2 (1,14), and these compounds serve as promising new therapeutics to treat cancer and diabetes. Whereas PTP1B and SHP2 are important drug targets for cancer and diabetes treatment, modulation of LYP activity has the potential to treat autoimmune diseases. Several LYP inhibitors with and of ERK (C) in T cells preincubated with NC1. Statistical comparisons between two groups were performed with Student's t tests. *, p Ͻ 0.05 when the anti-CD3 antibody-treated cells were compared with the untreated cells. Statistical comparisons among the anti-CD3-treated groups were performed with two-way ANOVA analysis. Difference between NC1 groups and control (con) groups was significant (p Ͻ 0.001). Difference between siRNA-treated groups and siRNA-untreated groups was significant (p Ͻ 0.001); the interaction between these two factors was significant (p Ͻ 0.005). For all statistical analyses, data from at least three independent experiments were quantified and presented as the mean Ϯ S.D. (error bars).

Allosteric inhibition of PTPN22 by locking two enzyme sites
both high potency and selectivity have been developed (3,9,11,12). Although one of these known LYP inhibitors has a mixed inhibition mode (3), an allosteric inhibitor for LYP was still lacking, and the mechanism of allosteric regulation of LYP by a small compound had not been revealed. Here, we identified NC1 as a noncompetitive inhibitor of LYP using enzymology. Moreover, the results of site-directed mutagenesis, fragmentcentric topographic mapping, and MD simulations suggested

Allosteric inhibition of PTPN22 by locking two enzyme sites
that NC1 concurrently binds to a "WPD pocket" in WPD-loop and a "secondary pocket" in LYP-specific insert outside the active site of LYP. Two residues (His-196 and Asp-197) in the WPD pocket and two residues (Phe-28 and Thr-36) in the secondary pocket shared low sequence identity compared with other phosphatases, thus contributing to the selectivity of NC1 toward LYP.
Because the fragment-centric topographic mapping and allatom MD simulations suggested that NC1 binds to LYP with the catalytically important WPD-loop assuming an open conformation, and because the efficient catalysis of substrate by PTPs requires the closed form of the WPD-loop (31, 32, 36), we reasoned that NC1 allosterically regulates WPD-loop movement and thus inhibits enzyme activity. Traditionally, structural information can be acquired by co-crystallizing an inhibitor with a phosphatase, which captures a static image of how an inhibitor interacts with a phosphatase (1, 13), but cannot provide further dynamic information. NMR spectroscopy can be used to characterize small phosphatases, such as VHR and PRL, but it is not easily applied to classic tyrosine phosphatases because of their large size. Here, using our newly developed unnatural amino acid F2Y incorporation technology (38), we were able to monitor WPD-loop dynamics with high resolution using 19 F NMR spectroscopy. The binding of LYP to the prod-

Figure 6. Comparison of the potential allosteric pockets in LYP with atypical-open WPD-loop (A, PDB code 3H2X), VHR with closed WPD-loop (B, PDB code 1J4X), PTP1B with open WPD-loop (C, PDB code 2HNP), STEP with open WPD-loop (D, PDB code 2CJT), PTPN18 with open WPD-loop (E, PDB code 2OC3), Glepp with open WPD-loop (F, PDB code 2GJT).
The proteins are presented in transparent white surface with WPD-loop shown as red loop and substrate pNPP shown as yellow sticks. The predicted binding pose of NC1 was derived from representative MD simulation snapshot in Fig. 5A and shown as green sticks. Fragment-centric topographic mapping was performed using AlphaSpace. Good pockets (pocket score Ͼ 100) are presented with green spheres and auxiliary pockets (30 Ͻ pocket score Ͻ 100) are presented with blue spheres. Potential allosteric inhibitor binding pockets, which possess a series connected small pockets, are marked with yellow circles.

Allosteric inhibition of PTPN22 by locking two enzyme sites
uct mimic vanadate caused a significant upfield shift, which was restricted by ϳ1/3 after incubation with compound NC1. Therefore, the F2Y incorporation method together with 19 F NMR spectroscopy provided direct evidence that modulation of WPD-loop movement serves as an underlying mechanism for the noncompetitive inhibition of LYP by NC1. Interestingly, both previously identified allosteric inhibitors of PTP1B also limited the movement of the WPD-loop. Therefore, limiting WPD-loop movement may be a common strategy to develop allosteric inhibitors for PTPs, and our newly developed unnatural amino acid F2Y incorporation method, together with 19 F NMR spectroscopy, may be used to characterize the inhibitory mechanisms of other classic PTPs.

Protein expression and purification
The expression of native proteins, including the catalytic domain of LYP (residues 1-294) with an N-terminal His tag and other His-tagged proteins, was described previously (46). Briefly, BL21 (DE3) cells were transformed with the expression plasmids and cultured in LB medium with shaking at 37°C. The culture temperature was adjusted to 18°C when the cultures reached an A 600 of 0.6, and expression was induced for 12 h with 0.3 mM IPTG at an A 600 of 0.8. For expression of C-terminal His-tagged LYP F2Y-incorporated protein, pEVOL-F2YRS was co-transformed with LYP L281TAG mutation into BL21 (DE3). The expression was induced with 0.3 mM IPTG and 0.02% L-arabinose at an A 600 of 1.0 in the presence of 0.5 mM F2Y. The cells were then harvested by centrifugation and resuspended in lysis buffer (20 mM Tris, pH 8.0, 300 mM NaCl). After centrifugation, the supernatant was incubated with Ni-NTA resin with end-to-end mixing for 1 h at 4°C. The beads were collected and washed with 20 ml of wash buffer (20 mM Tris, 300 mM NaCl, and 5 mM imidazole) and eluted with an imidazole gradient (20 mM Tris, pH 8.0, 300 mM NaCl, and 20 -200 mM imidazole). The protein was further purified through CM Sefin-ose85 with elution by a salt gradient. The low-salt solution contained 20 mM MES, pH 6.0, 100 mM NaCl, 1 mM EDTA, and Figure 7. 19 F NMR spectroscopy reveals suppression of WPD-loop conformational changes by NC1. A, crystal structures of LYP showing conformational changes in the WPD-loop and its adjacent residues with or without substrate. Left, residue Leu-281 is "buried" by the WPD-loop (PDB code 2P6X). Right, residue Leu-281 is "exposed" after substrate binding (PDB code 2QCJ). B, schematic flowchart of the incorporation of F2Y into LYP at position 281. C, purity of the protein was determined by electrophoresis (left panel). The purified protein was subjected to trypsin digestion and analyzed by MS/MS, which indicated the presence of the y12Ϯ F2Y-VYNAVLELFKR fragment (M r 1550) and the y13Ϯ E-F2Y-VYNAVLELFKR fragment (M r 1679). These results confirmed that F2Y was specifically incorporated into LYP at position 281. m/z, mass/charge ratio. D, upfield shift was observed in the 19 F NMR spectrum of the LYP L281F2Y 19 F NMR probe in response to Na 3 VO 4 binding (upper panel). The 19 F NMR spectrum of the LYP-L281F2Y probe in response to Na 3 VO 4 binding after preincubation with compound NC1 (lower panel).

Allosteric inhibition of PTPN22 by locking two enzyme sites
2 mM DTT. The high-salt solution contained 20 mM MES, pH 6.0, 1 M NaCl, 1 mM EDTA, and 2 mM DTT. After purification using CM Sefinose, the protein was further concentrated and stored at Ϫ80°C.

k cat and K m measurements
Initial rate measurements for the enzyme-catalyzed hydrolysis of pNPP were conducted as described previously (36). All assays were carried out at 25°C in 50 mM 3,3-dimethylglutarate, pH 7.0, buffer, containing 2 mM DTT and 1 mM EDTA, with an ionic strength of 0.15 M adjusted by addition of NaCl. For the pNPP reaction, assay mixtures of 100 l in total volume were set up in a 96-well polystyrene plate from Thermo Fisher Scientific. A substrate concentration range from 0.2 to 5 K m was used to determine the k cat and K m values. Reactions were started by the addition of an appropriate amount of enzymes. The reaction mixtures were quenched with 100 l of 1 M sodium hydroxide, and the absorbance at 405 nm was read using a plate reader. All Michaelis-Menten parameters reported are based on nonlinear curve fits of the raw data. The steady-state kinetic parameters were determined from a direct fit of the data to the Michaelis-Menten equation using GraphPad Prism 6.0 as shown in Equation 1.

IC 50 measurements
Kinetics assays for LYP-catalyzed pNPP hydrolysis in the presence of a small-molecular inhibitor were measured as described previously (28,46). The effect of each inhibitor on the LYP-catalyzed pNPP hydrolysis was determined at 25°C in reaction buffer (50 mM 3,3-dimethylglutarate buffer with the ionic strength of 0.15 M adjusted by NaCl). The K m values of LYP toward pNPP hydrolysis (4 mM for pNPP) were used to determine the IC 50 . The reaction was detected by monitoring the absorbance of pNP at 405 nm. The IC 50 values were obtained by fitting the data to Equation 2 using GraphPad Prism 6.0.

K i measurements
The phosphatase-catalyzed hydrolyses of pNPP in the presence of inhibitors were assayed at 25°C. The reaction was initiated by addition of pNPP (ranging from 0.2 to 5 K m ) to a reaction mixture containing different phosphatases and various fixed concentrations of inhibitors and stopped by addition of 1 M NaOH. All K i values were evaluated based on nonlinear curve fits of the raw data using GraphPad Prism 6.0. Inhibition patterns were evaluated by fitting the data to the Michaelis-Menten equations (or Lineweaver-Burk equation) for competitive inhibition (Equations 3 and 4) and noncompetitive inhibition (Equations 5 and 6), using linear regression and the program GraphPad Prism 6.0 as follows.

Cell culture, RNAi, and Western blot analysis
Cell culture and RNAi were performed as described previously (28). Jurkat T cells were preincubated with 20 M (final concentration) inhibitor (NC1) or DMSO for 45 min and then stimulated with 5 g/ml anti-CD3 antibody (OKT3) or medium for 5 min. The stimulation was terminated by transferring cells to ice and then lysed in lysis buffer (50 mM Tris, pH 7.5, 150 mM NaCl, 10 mM NaF, 2 mM EDTA, 10% glycerol, 1% Nonidet P-40, 0.25% sodium deoxycholate, 1 mM Na 3 VO 4 , 1 mM phenylmethylsulfonyl fluoride, 0.3 M aprotinin, 130 M bestatin, 1 M leupeptin, and 1 M pepstatin) for 15 min. The lysates were then centrifuged at 12,000 rpm for 15 min. The supernatants were collected, and the protein concentrations were measured by the BCA protein quantitation kit (Beyotime). Equal amounts of cell lysates were denatured in 2ϫ SDS loading buffer and boiled for 10 min. Protein samples were then subjected to Western blotting with specific anti-Src/pTyr-416 or anti-ERK pThr-202/pTyr-204 antibodies or GAPDH antibodies.

Preparation of ligand and protein structures
The initial structures of compound A15 and NC1 were constructed using SYBYL-x 1.1 (Tripos, Inc.), and the Tripos force field was employed in structure minimization. Eight crystal structures of LYP were retrieved from the Protein Data Bank (2P6X, 2QCT, 2QCJ, 3BRH, 3H2X, 3OLR, 3OMH, and 4J51) and prepared using the protein preparation workflow in SYBYL-x 1.1 (Tripos, Inc.). The mutated residues were changed back using Discovery Studio 2.5. The protonation states of specific residues were determined at constant pH 7 using the PDB code 2PQR server (47). Alignment of eight crystal structures was performed using PyMOL (The PyMOL Molecular Graphics System, Version 1.7. 4 Schrödinger, LLC).

Pocket analysis
Pocket analysis of LYP crystal structures was performed using AlphaSpace (48,49), which utilizes a geometric model based on Voronoi tessellation. Concave interaction space across the protein surface was identified and represented as a set of ␣-atom/␣-space pairs, which are then clustered into discrete fragment-centric pockets. Details for the calculation of pocket score are described in previous study (48,49).

Molecular docking
All docking studies were carried out using the standard setting of Autodock Vina (50). Compound NC1 was docked into eight LYP crystal structures in the existence of substrate (pNPP) to generate the noncompetitive binding model. The initial binding conformation of pNPP was judiciously determined based on the crystal structure of LYP complexed with a phosphotyrosine peptide (PDB code 3OLR). The competitive binding models of compound A15 and NC1 were predicted by docking each ligand separately to the active site of LYP (PDB code 3H2X) in the absence of pNPP. A grid box with 30 Å units Allosteric inhibition of PTPN22 by locking two enzyme sites in x, y, and z directions was used to cover the protein surface around five key interacting residues from mutational analysis. AutoDock Vina reports a series of lowest energy conformations, and eight representative models were selected for NC1 to fully explored the potential noncompetitive inhibitor binding mode. In addition, the competitive binding models for A15 and NC1 were selected according the docking scores. A total of 10 LYP-inhibitor complexes obtained from molecular docking were subjected to molecular dynamics simulations.

Molecular dynamics simulation
Molecular dynamics simulations were carried out using Amber14 package with Amber14SB force field (51). The LYP crystal structure (PDB code 3BRH) that possesses a closed conformation of WPD-loop was used for LYP_pNPP system without NC1 bound. Eight representative docked poses of compound NC1 with LYP crystal structure (PDB code 3HX2) were used for the LYP_pNPP system with pNPP bound. Partial atomic charges for pNPP, A15, and NC1 were obtained from HF/6 -31G(d) calculations using Gaussian 09 package (51)(52)(53).The RESP module in the Amber package was employed to fit the charges to each atomic center (54,55). Each system was neutralized with Na ϩ counterions and solvated with explicit TIP3P water in a rectangular periodic box with 10.0 Å buffer. After a series of minimizations and equilibrations, standard molecular dynamics simulations were performed with periodic boundary conditions. Nonbonded interactions were treated using the Particle Mesh Ewald method (56, 57) with 12.0 Å cutoff. The SHAKE algorithm (58) was utilized to constrain all bonds involving hydrogen atoms. The coordinates were stored every 2 ps, and the simulation time step was 2 fs. Berendsen thermostat method (59) was used to control the system temperature at 300 K. All other parameters were default values. MD trajectories were analyzed using cpptraj module in AmberTools 15. Protein-ligand interaction energies were calculated using the MM/GBSA method (60,61). MM-GBSA calculations were performed by MMPBSA.py module of Amber14. All figures and movies are produced using PyMOL (The PyMOL Molecular Graphics System, Version 1.7.4 Schrödinger, LLC), Chimera (42), and Microsoft Excel.

Trypsin digestion and MS/MS analysis
The LYP-F2Y protein was subjected to electrophoresis, and the protein band was cut into small plugs and washed twice in 200 ml of distilled water for 10 min. The gel bands were dehydrated in 100% acetonitrile for 10 min and dried in a Speedvac (Labconco) for 15 min. Disulfide bonds were reduced by adding 10 ml of 100 mM dithiothreitol (DTT) and subsequently alkylated by 40 mM iodoacetamide, 25 mM NH 4 HCO 3 for 45 min at room temperature in the dark. The sample was then mixed with trypsin by a ratio of 100:1 in Tris buffer and digested at 37°C for 12 h. Digestion was stopped by adding formic acid to a 1% final concentration. Digested samples were purified, desalted, and re-dissolved in 30 ml of 50% CH 3 CN, 0.1% CF 3 COOH buffer before MS/MS analysis.
LC-MS/MS analysis was performed using a Thermo Finnigan LTQ linear ion trap mass spectrometer in line with a Thermo Finnigan Surveyor MS Pump Plus HPLC system. The peptides generated by trypsin digestion were loaded onto a trap column (300SB-C18, 5 ϫ 0.3 mm, 5-m particle) (Agilent Technologies, Santa Clara, CA), which was connected through a zero dead volume union to the self-packed analytical column (C18, 100 m inner diameter ϫ 100 mm, 3-m particle) (SunChrom, Germany). The peptides were then eluted over a gradient (0 -45% B in 55 min, 45-100% B in 10 min, where B ϭ 80% acetonitrile, 0.1% formic acid) at a flow rate of 500 nl min Ϫ1 and introduced online into the linear ion trap mass spectrometer (Thermo Fisher Scientific, San Jose, CA) using nanoelectrospray ionization. MS data were analyzed by Bioworks 3.2 software.

NMR experiments
To detect Na 3 VO 4 -induced LYP WPD-loop conformational changes, 100 M LYP F2Y proteins were mixed with or without a 10-fold molar ratio of Na 3 VO 4 and incubated in binding buffer (20 mM Tris-HCl, pH 8.0, 150 mM NaCl, 10% D 2 O) with end-to-end rotation at room temperature for 30 min. The protein samples were then subjected to 19 F NMR experiments.
To detect the effect of NC1 on Na 3 VO 4 -induced LYP WPDloop conformational changes, 100 M LYP F2Y proteins were preincubated with a 5-fold molar ratio of NC1 in binding buffer with end-to-end rotation at room temperature for 30 min and then mixed with or without a 10-fold molar ratio of Na 3 VO 4 and incubated in binding buffer with end-to-end rotation at room temperature for 30 min. The protein samples were then subjected to 19 F NMR experiments.
All NMR data were collected using an Agilent OD2 600 spectrometer fitted with a 5-mm broad band probe. The 19 F 90°p ulse lengths were 9.9 s, and the spectra were typically obtained using 15,000 scans and a recovery delay of 1 s. Data were processed using 10-Hz Lorentzian line broadening and were referenced to the internal TFA standard (Ϫ76.5 ppm). All of the spectra were recorded at 25°C.

Statistics
The data were analyzed using GraphPad Prism 6. All experiments were performed at least in triplicate, and the data were expressed as mean Ϯ S.D. Statistical comparisons between two groups were performed with Student's t tests. Statistical comparisons between two factors were performed with two-way ANOVA analysis.

Data availability
The authors declare that data supporting the findings of this study are available in the article as well as its supporting information and from the authors on reasonable request.