Identification of Substrates of Human Protein-tyrosine Phosphatase PTPN22*

Stimulation of mature T cells activates a downstream signaling cascade involving temporally and spatially regulated phosphorylation and dephosphorylation events mediated by protein-tyrosine kinases and phosphatases, respectively. PTPN22 (Lyp), a non-receptor protein-tyrosine phosphatase, is expressed exclusively in cells of hematopoietic origin, notably in T cells where it represses signaling through the T cell receptor. We used substrate trapping coupled with mass spectrometry-based peptide identification in an unbiased approach to identify physiological substrates of PTPN22. Several potential substrates were identified in lysates from pervanadate-stimulated Jurkat cells using PTPN22-D195A/C227S, an optimized substrate trap mutant of PTPN22. These included three novel PTPN22 substrates (Vav, CD3ϵ, and valosin containing protein) and two known substrates of PEP, the mouse homolog of PTPN22 (Lck and Zap70). T cell antigen receptor (TCR) ζ was also identified as a potential substrate in Jurkat lysates by direct immunoblotting. In vitro experiments with purified recombinant proteins demonstrated that PTPN22-D195A/C227S interacted directly with activated Lck, Zap70, and TCRζ, confirming the initial substrate trap results. Native PTPN22 dephosphorylated Lck and Zap70 at their activating tyrosine residues Tyr-394 and Tyr-493, respectively, but not at the regulatory tyrosines Tyr-505 (Lck) or Tyr-319 (Zap70). Native PTPN22 also dephosphorylated TCRζ in vitro and in cells, and its substrate trap variant co-immunoprecipitated with TCRζ when both were coexpressed in 293T cells, establishing TCRζ as a direct substrate of PTPN22.

T cell antigen receptor (TCR) 2 signaling is important for the proliferation and differentiation of T cells. Upon binding of the peptide antigen presented by major histocompatibility complex, T cell antigen receptors initiate a cascade of signaling events mediated by proteintyrosine kinases and adaptor proteins (1)(2)(3). TCR engagement brings the Src family kinase Lck into proximity with cytoplasmic domains of the invariant TCR subunits TCR, CD3⑀, CD3␥, and CD3␦, each of which contains one or more immunoreceptor tyrosine-based activation motifs (ITAMs) that are phosphorylated at tyrosine residues within the conserved signature motif (YXXLX 6 -8 YXXL) by Lck (4). Phosphorylation of ITAMs creates high affinity binding sites for the tandem SH2 domains of the Syk family protein-tyrosine kinase Zap70 (1,2), which is subsequently recruited to the TCR complex and activated by Lck (1). Zap70, in turn, phosphorylates LAT (5,6) and SLP76 (7), key adaptor proteins in the TCR signaling pathway that relay the signal to downstream effectors and eventually lead to the activation of T cells. Phosphorylated Zap70 can also serve as a docking site for several components of the TCR signaling pathway including Lck itself (8), Vav (9), and Cbl (10,11).
Lck has two major phosphorylation sites at tyrosine 394 and tyrosine 505 (1,12,13). Tyrosine 394, which is within the activation loop of the kinase domain, is autophosphorylated upon T cell stimulation and phosphorylation at this site is required for maximal Lck kinase activity (1). Tyrosine 505, located at the carboxyl terminus of Lck, is phosphorylated by Csk (14,15) to create an internal binding site to which the SH2 domain of Lck binds, thus inhibiting Lck activity (16). Zap70 is phosphorylated at multiple sites, some of which activate (tyrosines 319 and 493) and others that suppress (tyrosines 292 and 492) TCR signaling (10,17). Phosphorylation at tyrosine 493 by Lck augments the kinase activity of Zap70 (17,18), whereas phosphorylation at tyrosine 319 appears to generate a docking site for Lck but does not alter intrinsic Zap70 kinase activity (8).
T cell signaling is negatively regulated by the activity of protein-tyrosine phosphatases such as PTPH1, SHP1, and PEP that dephosphorylate components of the TCR signaling pathway (19,20). Overexpression of PEP, the mouse homolog of PTPN22, decreases TCR activation (21,22); conversely, genetic ablation of PEP results in an increase in TCR signaling, notably in the effector/memory T cell population (23). PEP interacts with Csk, a negative regulator of Src family kinases, to repress TCR signaling in a synergistic manner (24). The physiological importance of PTPN22 in proper immune system regulation is further demonstrated by recent disease association studies linking a functional R620W protein variant, encoded by a C1858T single nucleotide polymorphism, to a significantly increased risk of autoimmune diseases including type 1 diabetes (25), rheumatoid arthritis (26), and systemic lupus erythematosis (27). The effects of the R620W variant on PTPN22 function have not been entirely defined but one consequence is to disrupt the interaction between Csk and PTPN22 (25,26).
Despite the key role of PTPN22 in T cell signaling, there has to date been no systematic effort to identify its substrates in T cells. Reported substrates for PEP/PTPN22 include Zap70 and Src family kinases such as Fyn and Lck (22)(23)(24). It is likely that other substrates exist, e.g. Cskbinding phosphoproteins. In this study, we developed an unbiased approach to identify substrates of PTPN22 using substrate trapping methods combined with protein identification by mass spectrometry. Using this approach, we identified both known and potentially novel substrates of PTPN22, including Lck, Zap70, VCP, Vav, CD3⑀, and TCR. We also provide biochemical evidence for direct interactions between PTPN22 and activated Lck and Zap70, and characterize the specificity of tyrosine dephosphorylation in these two substrates.

EXPERIMENTAL PROCEDURES
Antibody Reagents, Cell Lines, and Cell Culture-Human Jurkat and 293T cell lines were purchased from ATCC and cultured in RPMI 1640, 10% fetal bovine serum or Dulbecco's modified Eagle's medium, 10% fetal bovine serum, respectively. Sf9 insect cells were obtained from Invitrogen and cultured in ExCel420. HA.7 and HA.11 antibodies to the hemagglutinin affinity tag were obtained from Sigma and Covance, respectively; anti-(His) 5 antibody was purchased from Qiagen. An antibody to phosphotyrosine 416 of Src that cross-reacts with phosphotyrosine 394 of Lck was purchased from Cell Signaling, as were antibodies to LCK phosphotyrosine 505, Zap70, Vav, Zap70 phosphotyrosine 319, and Zap70 phosphotyrosine 493. Anti-TCR (6B10.2), anti-phospho-TCR, and anti-CD3⑀ antibodies were from Santa Cruz Biotechnology. 4G10 anti-phosphotyrosine antibody was from Upstate Biotechnology.
Expression Plasmids and Transient Transfections-A cDNA encoding the catalytic domain (cd) of human PTPN22 (amino acids 1 to 300) was cloned from a human leukocyte cDNA library (Clontech) into the Escherichia coli expression vector pET104.1, with a His 6 tag at the COOH terminus to facilitate purification. The vector provides a biotin conjugation site at the NH 2 terminus of the protein, enabling in vivo biotinylation of PTPN22cd by endogenous E. coli biotin ligases. Substrate trap mutants of PTPN22cd, including D195A, C227S, D195A/ C227S, and D195A/Q274A were constructed using the GeneTailor mutagenesis system (Invitrogen). The full-length PTPN22 cDNA was cloned into the BamHI/NotI sites of pIRES-GFP (Qbiogene), with an HA tag added at the NH 2 terminus to obtain the mammalian expression construct, pCMV5-HAPTPN22. For expression in 293T cells, the fulllength Lck, TCR, and Zap70 cDNAs were cloned from the human leukocyte cDNA library into pCDNA3.1 (Invitrogen). A Y505F mutant of Lck was constructed by site-directed mutagenesis as described above for the substrate trap mutants. For expression of TCR in E. coli, a cDNA encoding the entire cytoplasmic domain (amino acids 52-164) of human TCR (cyto-TCR) was cloned into the BamHI/XhoI sites of pET21a, incorporating a His 6 tag at the protein COOH terminus. For baculovirus expression in Sf9 cells, Lck and Zap70 were cloned into pFastbac1, with a His 6 tag at the NH 2 or COOH terminus, respectively. The cloning, expression, and purification of Lck⌬224 (residues 225-509) was as described (28) except tyrosine 505 was mutated to phenylalanine. Transfection of DNA into 293T cells was performed with Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions.
Expression and Purification of Recombinant Proteins-Sf9 cells were infected with baculovirus containing Lck or Zap70 at a multiplicity of infection of 2. Cells were harvested 48 h after infection and lysed with lysis buffer A (20 mM Tris⅐Cl, pH 7.6, 200 mM NaCl, 20 mM imidazole, 1% Triton X-100, 1ϫ Complete protease inhibitor mixture (Roche), 1 mM TCEP). The cell lysate was loaded on a nickel-nitrilotriacetic acid column (Qiagen), washed extensively with wash buffer A (20 mM Tris⅐Cl, pH 7.6, 300 mM NaCl, 20 mM imidazole, 0.25 mM TCEP), and the proteins were eluted with elution buffer A (20 mM Tris⅐Cl, pH 7.6, 300 mM NaCl, 250 mM imidazole, 0.25 mM TCEP). The recovered proteins were then loaded on a Superdex 200 (Amersham Biosciences) column pre-equilibrated with gel filtration buffer (20 mM Tris⅐Cl, pH 7.6, 150 mM NaCl, 1 mM DTT). The purity of purified Lck and Zap70 was greater than 95% as assessed by SDS-PAGE (data not shown). Purified Lck and Zap70 were used immediately after preparation because a substantial loss of activity was observed in samples that were subjected to freeze-thaw cycles. The cytoplasmic domain of human TCR was expressed in E. coli BL21(DE3) cells. Cells were induced with 1 mM isopropyl 1-thio-␤-D-galactopyranoside at 37°C for 3 h. Cyto-TCR was purified to Ͼ90% homogeneity by affinity and gel filtration chromatography exactly as described for Lck and Zap70. The wild type and substrate trap mutants of PTPN22cd were expressed in E. coli BL21(DE3) cells. Cells were grown to an A 600 ϭ 0.7 and induced with 1 mM isopropyl 1-thio-␤-D-galactopyranoside at 20°C for 16 h. The harvested cell pellet was lysed in lysis buffer B (20 mM Tris⅐Cl, pH 7.6, 1 M NaCl, 20 mM imidazole, 1% Triton X-100, 1ϫ Complete protease inhibitor mixture, 1 mM TCEP). Cell lysates were loaded on a nickel-nitrilotriacetic acid column and proteins were eluted as described above for Lck and Zap70. The pooled protein fractions were dialyzed against MES buffer (20 mM MES, pH 6.0, 50 mM NaCl, 1 mM EDTA, 2 mM DTT) at 4°C overnight, and loaded on a Source S-15 cation exchange column pre-equilibrated with MES buffer. PTPN22cd was eluted with a 50°-250 mM linear NaCl gradient over 25 column volumes. The pooled fractions containing PTPN22cd were concentrated and loaded on a Superdex 200 column pre-equilibrated with Hepes buffer (25 mM Hepes, pH 7.4, 150 mM NaCl, 1 mM EDTA, 1 mM DTT). After gel filtration, the purity of PTPN22cd was greater than 95%. All of the substrate-trapping mutants of PTPN22cd were purified in a similar manner to Ͼ95% homogeneity.
Substrate Trapping-Substrate trapping was performed as described (29), with minor modifications. Briefly, 1 ϫ 10 9 Jurkat cells were treated with 100 M pervanadate for 30 min and collected by centrifugation. The cell pellet was lysed with 10 ml of lysis buffer C (25 mM Hepes, pH 7.4, 150 mM NaCl, 1% Triton X-100, 1ϫ Complete protease inhibitor mixture, 1 mM EDTA), treated with 5 mM iodoacetic acid on ice for 5 min, neutralized by addition of 10 mM DTT, and centrifuged to remove debris. The purified biotinylated PTPN22cd (WT and substrate trap mutants) were coupled to streptavidin beads following the supplier's protocol (Pierce). 100 mg of Jurkat lysate was incubated with 1 mg of protein equivalent of PTPN22cd-coated beads at 4°C for 1 h. The beads were pelleted and washed 3 times for 5 min with lysis buffer C supplemented with 1 mM DTT. Bound proteins were eluted with 1ϫ SDS sample buffer at room temperature for 1 h. The eluted proteins were then boiled, aliquots were analyzed by a 4°-20% gradient of SDS-PAGE, and stained with SilverSnap (Pierce), a reagent that is compatible with subsequent mass spectrometric peptide identification. The vanadate competition experiment was performed as described (30).
Mass Spectrometry Identification-Protein bands excised from a silverstained gel were digested with trypsin (Promega) using an in-gel digestion robot (Investigator ProGest, Genomic Solutions, Ann Arbor, MI). The eluted tryptic peptides were concentrated to 2 l using a SpeedVac and reconstituted to 10 l in 1% formic acid (Fluka). 5 l of the tryptic peptides were analyzed by light chromatography/mass spectrometry using an Ultimate HPLC system (Dionex/LC Packings, Sunnyvale, CA) with a reversed phase C18 column (Dionex) and injected into a mass spectrometer (Qstar Pulsar; Applied Biosystems, Foster City, CA). Protein identification was done by a data base search using the Mascot search engine from Matrix Sciences (London, United Kingdom).
In Vitro Phosphorylation and Dephosphorylation Assays-Purified Lck was autoactivated in kinase buffer (20 mM Tris⅐Cl, pH 7.6, 10 mM MgCl 2 , 1 mM MnCl 2 , 150 mM NaCl) containing 100 M ATP at 37°C for 10 min. Purified Zap70 and cyto-TCR were phosphorylated in vitro by Lck (20:1 Zap70:Lck molar ratio; 60:1 TCR:Lck molar ratio) at 37°C for 10 min. After Lck treatment, the reaction mixtures were dialyzed against Hepes buffer at 4°C overnight. For dephosphorylation studies, 1 g of activated Lck, Zap70, or cyto-TCR was incubated at 37°C for 30 min in Hepes buffer with 10, 20, 40, and 80 ng of wild type PTPN22cd or 80 ng of the C227S active site mutant. To assay the activity of PTPN22cd, 1 g of recombinant protein was incubated with 200 l of 10 mM p-nitrophenyl phosphate in HAC buffer (50 mM acetic acid, pH 5.0, 2 mM DTT) at room temperature for 5 min. The reaction was stopped with the addition of 50 l of 1 N NaOH. The optical density at 405 nm was read in a 96-well plate reader (SpectraMax; Molecular Devices). Wild type PTPN22 is active against p-nitrophenyl phosphate, but the C227S mutant has no detectable activity with this substrate (data not shown).
In Vitro Binding Assays-200 ng of non-activated or activated Lck, Zap70, or TCR were incubated with 5 g of purified PTPN22cd (wild type and substrate trap mutants) coupled to 10 l of streptavidin beads in Hepes buffer at 4°C for 1 h. The bound proteins were washed with the same buffer 3 times, and the beads were boiled in 1ϫ SDS sample buffer for 3 min. 50 ng of input proteins (25% of total input) and 100% of eluted proteins were analyzed by SDS-PAGE and proteins captured by PTPN22 were quantitated by densitometry (ImageQuant).

RESULTS
Identification of Potential Substrates of PTPN22-PTPN22 contains two distinct functional domains, an NH 2 -terminal catalytic domain and a COOH-terminal proline-rich domain. The COOH-terminal domain participates in additional protein-protein interactions that likely contribute to protein localization or scaffolding functions but do not necessarily provide substrates for PTPN22 phosphatase activity. To reduce the complexity of protein-protein interactions and facilitate the identification of potential direct substrates of PTPN22, we produced and evaluated substrate trap variants of the PTPN22 catalytic domain (PTPN22cd). Mutations at key residues important for catalytic function have been employed as substrate traps for several tyrosine phosphatases, either singly or in combination (29 -32). We tested D195A and C227S (equivalent to D181A and C215S in PTP1B, respectively; Ref. 29) as single mutants as well as the double mutants D195A/C227S and D195A/Q274A for their ability to trap tyrosine-phosphorylated proteins from Jurkat cell lysates. The D195A/C227S mutant of PTPN22cd proved superior to the other substrate trap mutants (data not shown) and was then used to pull down potential substrates of PTPN22. Whole cell lysates from Jurkat cells treated with the phosphatase inhibitor pervanadate to stabilize tyrosine-phosphorylated proteins were subjected to affinity purification with the wild type and substrate-trapping DACS mutant of PTPN22cd (Fig. 1). In the absence of pervanadate treatment, neither the wild type protein nor the DACS mutant pulled down significant amounts of proteins (Fig. 1A, lanes 1 and 2). Following pervanadate treatment, several proteins including p270, p100, p90, p70, p56, p40, and p23 were trapped by the DACS mutant but not by the wild type protein (Fig. 1A, lanes 3 and 4). Bands corresponding to the proteins pulled down by the DACS protein were excised from the gel, digested in situ, and peptide fragments were identified by mass spectrometry. The p270 and p40 proteins were identified as spectrin and actin, respectively. Because these two abundant proteins are routinely identified in other mass spectrometry-based peptide identification experiments, we believe they represent nonspecific interactions of no biological significance. The p100 protein was identified as valosin containing protein (VCP), a member of the AAA (ATPases associated with multiple cellular activities) family of ATPases known to be phosphorylated on tyrosine residues following T cell receptor activation (33). The p90 protein was identified as Vav, a guanine nucleotide exchange factor also implicated in T cell signaling (34). The p70, p56, and p23 proteins were identified as Zap70, Lck, and CD3⑀, respectively, all proteins known to participate in T cell signaling. For each of these proteins, at least two (Vav, Lck), and as many as 71 (VCP) distinct peptides were identified by mass spectrometry (see examples in Fig. 1, B-D). The substrate trap protein identifications made by mass spectrometry were confirmed in repeat pull down experiments followed by immunoblotting with antibodies against the corresponding proteins. In agreement with the mass spectrometry results, the DACS mutant of PTPN22cd pulled down Zap70, Lck, Vav, and CD3⑀ from pervanadate-treated Jurkat cell lysates, whereas wild type PTPN22cd did not (Fig. 1E). Although immunoblotting was not performed for VCP, the abundance of peptides identified from this protein and the fact that it was not routinely observed in other mass spectrometry-based peptide identification experiments (data not shown) is consistent with VCP also being a potential substrate for PTPN22. Because CD3⑀ is a component of the TCR that is phosphorylated at an ITAM, we asked whether TCR, another ITAM-containing TCR component, might also be a potential substrate for PTPN22. Indeed, the DACS mutant of PTPN22cd specifically pulled down TCR from pervandate-treated Jurkat cell lysates (Fig. 1E, lanes 9 and 10).
PTPN22 Interacts Directly with Activated Lck and Zap70-Proteins identified from cell lysates by substrate trapping could represent either direct substrates of PTPN22 or non-substrate proteins trapped indirectly as part of a multiprotein complex. To confirm direct PTPN22cdsubstrate interactions, we repeated the in vitro substrate trap experiments with purified recombinant proteins. Native or DACS mutant PTPN22cd were coupled to streptavidin beads and used to trap purified LCK or ZAP70. Lck purified from insect Sf9 cells is phosphorylated at tyrosine 505 but not at tyrosine 394. Following autoactivation in the presence of ATP and Mg 2ϩ , tyrosine 394 became phosphorylated, whereas tyrosine 505 was not significantly affected ( Fig. 2A). When non-activated Lck was incubated with PTPN22cd, neither the wild type nor DACS mutant of PTPN22cd pulled down Lck (Fig. 2B, lanes 1-3). A similar result was obtained when the blot was probed with anti-phosphotyrosine 505 antibody (Fig. 2B, lanes 4 -6), indicating that phosphorylation at tyrosine 505 alone did not promote interaction with PTPN22. In contrast, activated Lck (phosphorylated at tyrosine 394) did interact with PTPN22cd, as ϳ30% of the activated Lck was pulled down by the substrate trap mutant (Fig. 2B, lanes 7-9). These results suggest that PTPN22cd interacts specifically with phosphotyrosine 394 of LCK. Because activated Lck is phosphorylated at tyrosines 394 and 505, it is formally possible that phosphorylation at both sites is required for PTPN22cd binding even though phosphorylation at tyrosine 505 alone is insufficient. To address this possibility, a truncated form of the Lck protein, Lck⌬224, containing the kinase domain and a Tyr to Phe change at position 505 was generated and tested in vitro for its ability to be trapped by the DACS mutant of PTPN22cd. Lck⌬224 purified from Sf9 cells contains no phosphate moieties as determined by mass spectrometry and Western blotting with anti-phosphotyrosine 4G10 antibody (data not shown). Following autoactivation, it was singly phospho-  1 and 2), Lck (lanes 3 and 4), CD3⑀ (lanes 5 and 6), Vav (lanes 7 and 8), and phospho-TCR (lanes 9 and 10). APRIL 21, 2006 • VOLUME 281 • NUMBER 16 rylated at tyrosine 394 as determined by mass spectrometry (data not shown) and Western blot with anti-phosphotyrosine 394 antibody (Fig.  2C, lane 4). Non-phosphorylated Lck⌬224 did not interact with either wild type or the DACS mutant of PTPN22cd (Fig. 2C, lanes 1-3), whereas ϳ30% of phosphorylated Lck⌬224 was trapped by the DACS mutant (Fig. 2C, lanes 4 -6). This result was similar to that obtained with activated full-length Lck, confirming that phosphorylation at tyrosine 394 alone is sufficient for PTPN22cd binding.
Site-specific Dephosphorylation of Lck and Zap70 by PTPN22-Because the DACS mutant of PTPN22 interacted directly with Lck through phosphotyrosine 394 but not phosphotyrosine 505, we examined the dephosphorylation of these sites by PTPN22cd. Activated Lck was incubated with increasing amounts of the wild type or inactive C227S mutant of PTPN22cd and phosphorylation of Lck was monitored with antibodies specific to phosphotyrosines 394 and 505 of Lck.
Wild type PTPN22cd clearly dephosphorylated phosphotyrosine 394 of Lck, but did not dephosphorylate phosphotyrosine 505 of Lck (Fig. 3A). As a control, the inactive C227S mutant of PTPN22cd did not dephosphorylate either site (Fig. 3A). Therefore PTPN22cd specifically dephosphorylated phosphotyrosine 394 but not 505 of Lck in vitro, a result consistent with the tyrosine site specificity observed in the substrate trap experiments.
Next we asked if full-length PTPN22 exhibited the same specificity against Lck phosphorylation sites in a cellular context. The Lck cDNA was co-transfected with either the wild type PTPN22 or DACS mutant of PTPN22 into human 293T cells, which do not express endogenous Lck or PTPN22. In cells co-transfected with mutant PTPN22 and Lck, the phosphorylation level of Lck tyrosine 394 was slightly higher than cells transfected with Lck alone (data not shown), confirming that the DACS mutant of PTPN22 was inactive. In cells transfected with the wild type and DACS mutant of PTPN22, the amounts of Lck in cell lysates were comparable (Fig. 3B, lanes 5 and 6); the DACS mutant was expressed at a slightly higher level than wild type PTPN22 (Fig. 3B, lanes  7 and 8). Phosphorylation of Lck tyrosine 394 was reduced in cells cotransfected with wild type PTPN22 compared with the DACS mutant (Fig. 3B, lanes 1 and 2). In contrast, the phosphorylation level of tyrosine 505 was similar in cells transfected with either the WT or DACS mutant of PTPN22 (Fig. 3B, lanes 3 and 4). Therefore, PTPN22 dephosphoryl-  1 and 2) or pY394  (lanes 3 and 4). B, proteins pulled down by recombinant wild type (WT) or DACS mutant PTPN22cd incubated with activated or non-activated recombinant Lck were analyzed by immunoblotting with antibodies to Lck (lanes 1-3), Lck pY505 (lanes 4 -6), and Lck pY394 (lanes 7-9). The INPUT lane represented 25% of the total input proteins. C, immunoblot analysis of proteins pulled down by recombinant WT or DACS mutant PTPN22cd incubated with non-phosphorylated Lck⌬224 (lanes 1-3) or Lck⌬224 phosphorylated only at Tyr-394 (lanes 4 -6). D, immunoblotting of Lck-activated Zap70 with antibodies to pY319 (lanes 1 and 2) or pY493 (lanes 3 and 4). E, immunoblot analysis of proteins pulled down by recombinant WT or DACS mutant PTPN22cd incubated with non-phosphorylated or phosphorylated Zap70 and probed with antibodies to Zap70 (lanes 1-3) or pY493 (lanes  4-6). pY, phosphotyrosine.
ated Lck in transfected cells with a specificity comparable with that observed with purified proteins in vitro.
In similar experiments we tested the ability of PTPN22 to dephosphorylate Zap70 in vitro and in transfected cells. Interestingly, PTPN22cd dephosphorylated phosphotyrosine 493 (Fig. 3C, lanes 1-5) but not phosphotyrosine 319 (Fig. 3C, lanes 6 -10). The specificity of PTPN22 against Zap70 phosphorylation sites observed in vitro was maintained in intact cells as wild type, full-length PTPN22 reduced the phosphorylation of tyrosine 493 of Zap70 to undetectable levels, whereas phosphorylation of tyrosine 319 was unaffected (Fig. 3D, lanes  1-4). In these experiments, a construct expressing LCK(Y505F) was cotransfected along with PTPN22 and Zap70 to provide a source of Lck to activate Zap70 in trans. Cells cotransfected with the wild type or DACS mutant of PTPN22 expressed comparable amounts of Zap70 as well as both forms of PTPN22 (Fig. 3D, lanes 5-8).
Specific Interactions between TCR and Substrate-trapping Mutants of PTPN22-As one of the four novel PTPN22 substrates identified by substrate trap in Jurkat cells, we investigated the interaction between TCR and PTPN22 in more detail. First, we tested the ability of various substrate trap mutants of PTPN22cd including D195A, C227S, D195A/ Q274A, and DACS to pull down TCR from Jurkat cell lysates. All four substrate-trapping mutants of PTPN22cd pulled down phosphorylated TCR, with the DACS mutant being the most effective as expected, followed by the C227S mutant (Fig. 4A). In the presence of 2 mM sodium vanadate, neither the wild type PTPN22cd nor the D195A mutant interacted with TCR (Fig. 4B, lanes 1 and 2) whereas in the absence of vanadate, only the D195A mutant could pull down TCR (Fig. 4B, lanes  3 and 4). These results demonstrate that the interaction between TCR and the D195A mutant of PTPN22 was dependent on the active site of PTPN22. To confirm that the interaction between PTPN22cd and TCR was direct, recombinant cytoplasmic domain TCR (cyto-TCR) containing all three ITAMs was phosphorylated in vitro with purified  11-15). B, immunoblot analysis of Lck phosphorylation in 293T cells cotransfected with expression plasmids encoding Lck and full-length wild type or DACS mutant PTPN22. Cell lysates were probed with antibodies to pY394 (lanes 1 and 2), pY505 (lanes 3 and 4), Lck (lanes 5 and 6), and the HA epitope tag (lanes 7 and 8). C, immunoblot analysis of recombinant, activated Zap70 incubated with increasing amounts of the WT or inactive CS mutant of PTPN22cd and probed with antibodies to pY493 (lanes 1-5), pY319 (lanes 6 -10), and Zap70 (lanes [11][12][13][14][15]. D, immunoblot analysis of Zap70 phosphorylation in 293T cells cotransfected with expression plasmids encoding Zap70, constitutively active Lck-Y505F, and fulllength wild type or DACS mutant PTPN22. Cell lysates were probed with antibodies to pY493 (lanes 1 and 2), pY319 (lanes 3 and 4), Zap70 (lanes 5 and 6), and HA (lanes 7 and 8). pY, phosphotyrosine.  1-6) or TCR (lanes 7-12). E, dephosphorylation of cyto-TCR as a function of incubation time with WT or CS PTPN22cd. The blot was probed with antibodies to phosphotyrosine (lanes 1-7) or TCR (lanes 8 -14).
Lck and tested as a potential substrate for PTPN22cd. Non-phosphorylated cyto-TCR was not pulled down by either the wild type or DACS mutant of PTPN22cd, (Fig. 4C, lanes 2 and 3). In contrast, ϳ54% of the phosphorylated cyto-TCR was captured by the DACS mutant, whereas none was pulled down by wild type PTPN22cd (Fig. 4C, lanes 4 -6), confirming that the DACS mutant of PTPN22cd interacted directly with phosphorylated TCR. Furthermore, purified wild type PTPN22cd dephosphorylates Lck-treated cyto-TCR in vitro, whereas a catalytically inactive mutant, C227S, does not (Fig. 4D). PTPN22cd-catalyzed dephosphorylation of TCR was complete within 30 min, with the majority of the phosphate groups being removed within the first 5 min (Fig. 4E).
To further confirm the interaction between PTPN22 and TCR in a cellular context, expression plasmids encoding full-length PTPN22 (wild type or the DACS mutant) and TCR were cotransfected into 293T cells. A plasmid encoding Lck mutated at Tyr-505 to enhance activity, Lck-Y505F, was also cotransfected to provide activated Lck in trans to phosphorylate TCR. The proteins co-immunoprecipitated with HA-tagged PTPN22 were analyzed by immunoblotting with anti-TCR and anti-HA antibodies. The anti-HA antibody immunoprecipitated comparable amounts of wild type and DACS mutant of HA-PTPN22 (Fig. 5A, lanes 3 and 4). However, only the DACS mutant of PTPN22 co-immunoprecipitated significant amounts of phosphorylated TCR (seen as multiple bands on the immunoblot; Fig. 5A, lanes 1  and 2), indicating that the substrate-trapping mutant of full-length PTPN22 interacts with TCR in transfected cells. To confirm that TCR is a substrate of PTPN22 in transfected cells, expression plasmids encoding Lck-Y505F, TCR, and full-length PTPN22 (wild type or DACS mutant) were cotransfected into 293T cells and phosphorylation of TCR was analyzed with anti-TCR and anti-phosphotyrosine (4G10) antibodies. Similar levels of unphosphorylated TCR were observed in cells transfected with either the wild type or DACS mutant (lower band, Fig. 5B, lanes 1 and 2) but the more slowly migrating species were reduced in cells transfected with wild type PTPN22, consistent with highly phosphorylated forms of TCR being converted to less phosphorylated species. Re-probing the blot with the 4G10 anti-phosphotyrosine antibody showed that phosphorylation of TCR was also reduced in cells transfected with wild type PTPN22 compared with the DACS mutant as expected (Fig. 5B, lanes 3 and 4).

DISCUSSION
Activation of T cells via antigen stimulation of the T cell receptor triggers the formation of an immune synapse involving sequential phosphorylation of TCR subunits, effector kinases such as Lck and Zap70, and various scaffolding proteins that together form the proximal signaling complex in T cells (1)(2)(3). This process is tightly regulated, both spatially and temporally, to control the timing and extent of immune responses following T cell activation. Phosphatases play a key regulatory role by modulating the phosphorylation, and thus, activation state of the proximal signaling complex in T cells (19,20). PTPN22, also known as Lyp, is a non-receptor tyrosine phosphatase recently shown to negatively regulate T cell signaling (20 -22). Although several substrates of PEP, the mouse homolog of PTPN22, have been reported, similar studies have not been described for PTPN22 itself. We used an unbiased approach combining substrate trap methods with mass spectrometry to: (a) confirm Lck and Zap70 as substrates for human PTPN22 and (b) to identify Vav, CD3⑀, and VCP as novel potential substrates. In addition, we identified TCR as a novel substrate for PTPN22, in vitro and in transfected cells. Substrate trapping has been used to identify potential substrates in various cell types for a number of phosphatases, e.g. PTP1B (29), PTPH1 (36), and SHP2 (37) among others. However, the majority of these studies identified only one or two substrates, and generally relied on a "candidate protein" approach using molecular weight determinations followed by Western blotting with specific antibodies to confirm the identification. Our approach enabled direct identification of multiple potential substrates by mass spectrometry, requiring no a priori knowledge of candidate proteins and, in principle, no antibody detection reagents. Substrate trap mutants vary in their intrinsic efficiencies and the optimal mutant needs to be established for each phosphatase; we found the D195A/C227S double mutant of PTPN22cd to be the most effective variant for substrate trapping studies. Although we used the catalytic domain of PTPN22 in our substrate trap experiments to minimize non-substrate-based protein-protein interactions, it is possible that some proteins pulled down from cell lysates were components of a multiprotein complex and that the PTPN22 interaction was indirect. We established that PTPN22cd interacts directly with Lck, Zap70, and TCR in vitro and in transfected cells, indicating that the pull down results reflected true substrate capture. We also showed that PTPN22 can remove phosphate moieties from each of these three proteins in vitro and in cells. Thus, these three proteins appear to be direct substrates of PTPN22. Interestingly, PTPN22 dephosphorylated Lck in a site-specific manner, removing phosphates from the autoactivation site at tyrosine 394 but not from the inhibitory site at tyrosine 505. SHP1, another protein-tyrosine phosphatase that moderately inhibits TCR signaling, exhibits similar specificity toward Lck (38), suggesting that multiple PTPs regulate the phosphorylation state of Lck at tyrosine 394. Similarly, dephosphorylation of Zap70 by PTPN22 occurred at tyrosine 493 but not tyrosine 319. Dephosphorylation at these sites in  11; lanes 3 and 4). B, immunoblot analysis of TCR phosphorylation in 293T cells cotransfected with TCR, PTPN22, and Lck-Y505F expression constructs. The blots were probed with antibodies to TCR (lanes 1 and 2) or phosphotyrosine (lanes 3 and 4).
Lck and Zap70 would be expected to down-regulate their kinase activities and is consistent with the negative role of PTPN22 in T cell signaling. It is likely that other tyrosine phosphatases target tyrosine 319 in Zap70 to further suppress T cell signaling, e.g. SHP1 (39,40). Conversely, tyrosine 505 of Lck is dephosphorylated by the receptortyrosine phosphatase CD45, a positive regulator of TCR signaling (20). Studies with PEP, the mouse ortholog of PTPN22, have suggested roles for Csk and/or c-Cbl in targeting PTPN22 to the immune synapse where the substrates Lck and Zap70 reside following T cell activation. The SH3 domain of Csk interacts with the proline-rich domain of PEP; PEP also associates with c-Cbl, the SH2 domain that interacts with phosphotyrosine 292 of Zap70. An alternative model has PEP being recruited to these protein-tyrosine kinases through intrinsic affinity for the activated kinases (20). Our data with PTPN22 are consistent with the latter model but the two hypotheses are not mutually exclusive.
Overexpression of PTPN22 in human T cells (41) or PEP in mouse T cells (24) leads, directly or indirectly, to dephosphorylation of TCR. In the latter study (24), a D195A substrate-trapping mutant of mouse PEP failed to interact with TAC-, a chimeric protein containing the extracellular domain of the ␣ chain of the interleukin 2 receptor (TAC) fused to the cytoplasmic domain of TCR, in transfected COS-1 cells. Our data identify TCR as a substrate of human PTPN22 and suggest that the dephosphorylation observed in these earlier studies may be direct. A possible explanation for the apparent discrepancy in substrate trapping results may be the different TCR constructs used. We used native TCR in our co-immunoprecipitation and affinity purification experiments; the presence of the TAC domain in the Tac-fusion protein could impair the interaction between PEP and the TCR domain. Another possibility is that the DACS mutant, a more potent substratetrapping variant of PTPN22, allowed us to detect weaker interactions. Substrate overlap among phosphatases is likely in T cells because PTPH1, a cytoskeletal protein-tyrosine phosphatase that suppresses TCR signaling, also dephosphorylates TCR in vitro and in COS-1 cells (42).
In addition to Lck, Zap70, and TCR, we also identified CD3⑀, Vav, and VCP as other potential substrates of PTPN22. Dephosphorylation of the ITAM domains of TCR and, presumably, CD3⑀ by PTPN22 would be expected to impair recruitment of Zap70 and possibly other proteins to the TCR complex, again acting to suppress T cell signaling. Other ITAM-containing proteins such as CD3␥ and CD3␦ may also be substrates for PTPN22. Although we focused our analysis largely on components of the proximal TCR signaling complex, the repertoire of PTPN22 substrates is likely to be considerably broader in scope. For example, Vav, a guanine nucleotide exchange factor for the small GTPases Rac and cdc42 that acts downstream of the TCR, was identified as a potential substrate in our trapping experiments. Vav, activated by Fyn via tyrosine phosphorylation (43), has been implicated in the cytoskeletal remodeling at the immune synapse that occurs as an integral part of T cell activation (44). VCP, a member of the AAA family of ATPases, is another novel potential substrate of PTPN22 that has reported roles as a molecular chaperone in multiple cellular processes including cell division (45) and dissociation of endoplasmic reticulum proteins into the cytosol (46). VCP was also identified by trapping studies as a substrate for the phosphatase PTPH1 (36). RNA interference directed against VCP shows antiproliferative and pro-apoptotic effects in HeLa cells (47). Although VCP is known to be tyrosine phosphorylated following T cell activation, its function in T cells is undefined as are the effects of phosphorylation. However, its reported function in cell cycle control raises the interesting possibility that VCP, and by extension, PTPN22, plays a role in regulating T cell survival and/or proliferation.