Effective Dephosphorylation of Src Substrates by SHP-1*

The protein-tyrosine phosphatase SHP-1 is a negative regulator of multiple signal transduction pathways. We observed that SHP-1 effectively antagonized Src-dependent phosphorylations in HEK293 cells. This oc-curred by dephosphorylation of Src substrates, because Src activity was unaffected in the presence of SHP-1. elevated to phosphorylation in , and SHP-1 variants with mutated phosphorylation sites in the C terminus, SHP-1 Y538F, SHP-1 Y538F,Y566F less active toward Src-generated phosphoproteins different proteins interaction of Src-generated phosphoproteins with the SHP-1 catalytic domain the SH2

The SH2 1 domain PTP SHP-1 regulates multiple signal transduction events by dephosphorylation (1)(2)(3). These comprise signaling of cytokine receptors such as the erythropoietin receptor (4), and the interleukin-3 receptor (5), and of receptor tyrosine kinases such as c-Kit (6 -8), the colony-stimulating factor-1 receptor (9,10), and the epithelial kinase Ros (11). SHP-1 modulates also the function of immunoreceptors (12), and cytoplasmic tyrosine kinases such as Lck (13,14). In these and many other cases, SHP-1 regulates signaling in a negative manner. In other pathways, SHP-1 may also exert a positive function. Thus, a role of SHP-1 for differentiation of glia cells (15), and for Ras-dependent activation of mitogen-activated protein kinase (16) have been reported. Also, SHP-1 has the capacity to activate Src kinase by dephosphorylation of the inhibitory phosphotyrosine in the Src C terminus and may thus stimulate Src-dependent phosphorylations in certain cell types (17). SHP-1 can translocate into the nucleus, however, its nuclear substrates are still elusive (18 -21). Among the recently identified substrates of SHP-1 is p120 ctn (22), a cytoplasmic protein that possesses armadillo-like repeats, participates in cell-cell adhesion complexes, and has the capacity to translocate to the nucleus. p120 ctn is also a prominent substrate for the cytoplasmic tyrosine kinase Src (23). The function of p120 ctn tyrosine phosphorylation and hence, the role of PTPmediated dephosphorylation, is not yet known. In addition to SHP-1, the transmembrane PTPs PTP and Dep-1 were reported to dephosphorylate p120 ctn (24 -26).
The substrate selectivity of PTPs in intact cells depends in a combinatorial way on targeting domains, and on the selectivity of the PTP catalytic domain (27). In the case of SHP-1 and the structurally related PTP SHP-2, this concept has been supported by domain swapping experiments (28,29). Both, selectivity of the SH2 domains and the catalytic domains contribute to selective signaling effects in different cell models.
The activity of SHP-1 is regulated by several mechanisms. A firmly established regulation mechanism is activation of SHP-1 through occupation of the SH2 domains by tyrosine-phosphorylated ligands (30). This mechanism has been inferred by multiple lines of evidence. Very recently, determination of the crystal structure of SHP-1 has directly shown that, exactly as previously observed for SHP-2 (31), SHP-1 exists in a closed, autoinhibited conformation (32). In this fold, the N-terminal SH2 domain occludes the catalytic center. Binding of a phosphorylated ligand to the N-terminal SH2 domain stabilizes a conformation of this domain, which can no longer bind into the active site and thereby the enzyme is released from autoinhibition. The binding selectivity for the SHP-1 N-terminal SH2 domain has been characterized in much detail and a binding consensus hXpYXXh (x, any amino acid; h, hydrophobic; pY, phosphotyrosine) has been defined (33,34). In addition to SH2 domain ligand binding, other mechanisms have been proposed to modulate SHP-1 activity. Notably, several modifications of the C-terminal tail domain of SHP-1 lead to activity changes in vitro. These include truncation (30,35), phospholipid binding (36), and tyrosine phosphorylation (37,38). A high affinity binding site for phosphatidic acid has been assigned to the last 41 amino acids of SHP-1 (36). Major phosphorylation sites have been mapped at tyrosines 536 and 554. Different tyrosine kinases were shown to phosphorylate one or both of these sites, including the insulin receptor tyrosine kinase and Lck (37,38). Uchida et al. (38) could show that phosphorylation at tyrosine 536 enhances SHP-1 activity in vitro. It is currently not clear whether lipid binding or tyrosine phosphorylation play a role for SHP-1 activity regulation in vivo. However, the existence of an SHP-1 variant with an alternate C terminus, SHP-1L (39), suggests that this domain of the molecule may fulfill an important function. Potentially, it could also be involved in targeting of SHP-1. For example, a nuclear localization signal has been mapped to the C terminus (20) and the phosphorylation sites may have the capacity to recruit SHP-1 substrates (40).
Compared with the SH2 domain specificity, the substrate selectivity of the SHP-1 catalytic domain is less well characterized. Comparison of different phosphopeptide substrates (35), determination of the catalytic domain structure in complex with a substrate phosphopeptide (41), and a phosphopeptide library approach (42) have provided some insights. According to these data, negatively charged amino acids N-terminal to the phosphotyrosine and possibly hydrophobic residues in the ϩ3 position may be critical for substrate binding to the SHP-1 catalytic domain. Here we describe novel phosphopeptide substrates for SHP-1 that are superior to previously described ones. They all correspond to efficient Src phosphorylation sequences. Based on enzymatic data and modeling studies we propose that Src and SHP-1 may have complementary substrate selectivity. In intact cells, Src substrates are efficiently dephosphorylated by SHP-1. Efficient recognition of Src substrates by the SHP-1 catalytic domain and the activation of SHP-1 by Src-dependent phosphorylation at the C-terminal phosphotyrosines play a role in this process. Src substrates only poorly interact with SHP-1 SH2 domains and activation of SHP-1 by C-terminal phosphorylation may be particularly important under these conditions.

MATERIALS AND METHODS
cDNA Constructs, Recombinant Proteins, and Antibodies-A human c-src cDNA was kindly provided by Dr. D. Fujita, University of Calgary, and subcloned into the eukaryotic expression vector pRK5RS. GST-Src was obtained from a pRK5RS c-src/EcoRI fragment, cloned into EcoRIdigested pBluescript KS (pBKKS, Stratagene) in such orientation that the EcoRV site of pBKKS was in front of the ATG of c-src. This construct was treated with EcoRV and NotI and the obtained c-src DNA was subcloned into SmaI-and NotI-digested pGEX-5X-2 (Amersham Biosciences). The constructs for eukaryotic expression of SHP-1 wild type and SHP-1 C455S and for bacterial expression of GST-SHP-1 wild type and SHP-1 domains have been described earlier (11,36). Throughout, the 597-amino acid epithelial SHP-1 variant (GenBank TM NM_080548) was used and numbering of mutations refers to this sequence. The SHP-1 Y538F, SHP-1 Y566F, and SHP-1 Y538F,Y566F mutations were obtained by PCR mutagenesis using pairs of oligonucleotides: 5Ј-GCC-AGGAGTCGGAGTTCGGGAACATCACC-3Ј/5Ј-GGTGATGTTCCCGA-ACTCCGACTCCTGGC-3Ј (for Y538F) and 5Ј-CAAGGAGGATGTGTT-TGAGAACCTGCAC-3Ј/5Ј-GTGCAGGTTCTCAAACACATCCTCCTT-G-3Ј (for Y566F). Two PCR fragments for each mutation were obtained by using one of these oligonucleotides and either the corresponding upstream or downstream oligonucleotide (5Ј-GACCATCCAGATGG-TGCGGG-3Ј/5Ј-CTAGTCTAGAGGACAGCACCGCTCACTTCCT-3Ј) as primers. PCR fragments were then fused by another PCR using the upstream/downstream pair of oligonucleotides as primers. The corresponding wild type DNA in GST-SHP-1 was replaced by the mutagenized fragments using Eco47III and XbaI (for Y538F or Y566F single mutation) or BglII and StuI (for Y538F,Y566F double mutation) sites. pcDNA3 constructs of these mutants were obtained by subcloning them from the GST constructs using EcoRI and XbaI sites flanking the SHP-1 gene. All constructs were verified by DNA sequencing. Expression and purification of GST fusion proteins and GST-free recombinant SHP-1 was done as described earlier (36,43).
Dephosphorylation and GST Pull-down Assays-HEK293 cells were transfected with the different expression constructs as indicated in the figure legends using a calcium phosphate method and cell extracts were analyzed by immunoblotting as described earlier (28). For pull-down assays using different fusion proteins of SHP-1 domains, HEK293 cells were transfected with 10 g of Src expression construct per 10-cm dish. The cells were lysed with 1 ml of lysis buffer (20 mM HEPES, pH 7.4, 150 mM NaCl, 2 mM EDTA, 2 mM EGTA, 20 M zinc acetate, 50 mM NaF, 10 mM NaPP, 1% Triton X-100, 1 mM phenylmethylsulfonyl fluoride, 1 g/ml pepstatin A, 1 g/ml aprotinin, 10 g/ml leupeptin, and 1 mM sodium orthovanadate) and were precleared by incubation with glutathione-Sepharose for 1 h at 4°C. 20 g of the different GST fusion proteins were allowed to bind to 10 l of glutathione-Sepharose (Amersham Biosciences) and were then incubated with the precleared HEK293 cell lysate for 1 h at 4°C. The beads were washed three times with lysis buffer and proteins were eluted by boiling with SDS-PAGE sample buffer.
Enzyme Assays-To assay Src activity, cell extracts were subjected to immunoprecipitation with agarose-linked anti-v-Src monoclonal antibody 327 (Calbiochem-Novabiochem, Bad Soden, Germany). Activity assays were done with acid-treated enolase as a substrate in the presence of [␥-32 P]ATP as described by Bagrodia et al. (44), except that the ATP concentration was increased to 10 M and no ␤-mercaptoethanol was included into the kinase reaction. Alternatively, Src optimal substrate peptide AEEEIYGEFEAKKK (45) was used as a substrate. Aliquots of the immunoprecipitates were incubated with the peptide at 30 M and [␥-32 P]ATP (10 M, 10 Ci per sample) in a buffer containing 10 mM MnCl 2 , 50 mM HEPES, pH 7.5, and 0.2 mM sodium orthovanadate in a final volume of 100 l at 30°C. At different time points 20-l aliquots were removed and mixed with 20 l of 40 mM EDTA. The samples were briefly centrifuged and 10-l aliquots of the supernatant were spotted on phosphocellulose paper (Whatman P81) and allowed to dry in air. The paper was washed 3 times with 75 mM phosphoric acid, dried, and exposed to the screen of a GS250 Molecular Imager (Bio-Rad). To test activation of SHP-1 by Src-mediated phosphorylation in vitro, 1 g of recombinant SHP-1 (or the GST fusion proteins as indicated in the figure legends) was incubated under agitation with the ϳ1/100 molar amount of GST-Src in 30 l of a solution containing 1 mM ATP, 50 mM HEPES, pH 7.4, and 10 mM MgCl 2 at 30°C for 30 min. Controls were done without GST-Src or without ATP. Then, pNPP and DTT were added to final concentrations of 10 and 100 mM, respectively (final volume 50 l), and the incubation was continued for another 30 min. The reaction was stopped by addition of 100 l of 1 N NaOH and pNPP hydrolysis was measured at 405 nm. Enzymatic activity of SHP-1 toward different phosphopeptides was assayed using malachite green for quantification of released phosphate with tyrosine phosphatase assay kit 1 from Upstate Biotechnology (Lace Placid, NY). Briefly, the phosphopeptides were incubated at a concentration of 200 M with 4.4 ng/ml of the GST fusion protein of the SHP-1 catalytic domain (GST-SHP-1CAT) in a reaction buffer containing 10 mM Tris-HCl, pH 7.4, and 5 mM DTT, at room temperature for 5 min. The reaction was stopped by addition of malachite green reagent and the amount of released free phosphate was determined by measuring the absorption at 620 nm. To obtain kinetic constants, a continuous fluorimetric assay for proteintyrosine phosphatase activity according to Zhang et al. (46) was used. Different concentrations of phosphopeptide were incubated in reaction buffer (50 mM 3,3-dimethylglutaric acid, pH 6.6, 150 mM NaCl, 1 mM EDTA, 5 mM DTT) at 30°C for 5 min. The reaction was started by addition of catalytic amounts (0.8 -8 nM) of GST-SHP-1CAT. Dephosphorylation of phosphotyrosine was determined by following the increase of fluorescence at 305 nm after excitation with 282 nm with a "FluoroMax-2 fluorescence spectrometer" (ISA JOBIN YVON-SPEX, Instruments S.A., Inc.). Data fitting according to Zhang et al. (46) resulted in large variations of the constants obtained for different substrate concentrations. Therefore, a conventional analysis was performed using the program Grace-5.1.7. (Weizmann Institute of Science, Rehovot, Israel). Initial velocities were determined by linear regression of points in the first 10 s of reaction. Means of data from at least triplicate determinations were subjected to fitting according to the Michaelis-Menten model. Kinetic values obtained with the single data sets were in a range of Ϯ15%.
Molecular Modeling-Structural models for the complexes of the SHP-1 catalytic domain with various phosphopeptides were generated using the molecular modeling package InsightII (Accelrys Inc., San Diego, CA). The model structures are based on the 2.5-Å crystal structure of the catalytic domain of human SHP-1 (synonym: PTP-1C) complexed with the SHPS-1/SIRP␣-derived phosphopeptide substrate SIRP␣pY469 (Ref. 41, Protein Data Bank code 1fpr). Note that numbering refers to the hematopoietic 595-amino acid variant of SHP-1 (GenBank TM NM_002831). The peptides were initially modeled using the builder and biopolymer modules of InsightII. Docking into the active binding site was performed by exchanging the corresponding residues of the peptide substrate in the experimental complex structure. The side chains of the substrate in the crystal structure (EDTLT-pYADLD) were replaced by those of the phosphorylated optimal Src substrate or the p120 ctn pY296 peptide (Table I). To analyze the contribution of individual residues of the optimal Src substrate, single residues were replaced by the corresponding ones in p120 ctn pY296 and the SIRP␣pY469 peptide. In addition, residues of SHP-1 that are expected to play a role for substrate specificity (47) were replaced by the corresponding ones in PTP1B (Barford et al. (48), Protein Data Bank code 2hnp). The modeled complex conformations were used as starting structures for a 600 step-conjugated gradient energy minimization with the protein backbone harmonically restrained to the x-ray coordinates (force constant: 100.0 kcal/mol/Å2) and free peptides using the cvff-force field as implemented in the discover module of InsightII. An implicit solvent model with a distant dependent dielectric constant of 4.0 was applied. The ranking corresponds to the differences of interaction energies between peptide and catalytic domain relative to the optimal Src substrate (in kcal/mol). It should be noted that the calculated interaction energy is basically a measure of the steric and electrostatic peptideprotein complementarity.

SHP-1 Antagonizes Src-dependent Phosphorylation-We
have previously reported that SHP-1 can negatively regulate autophosphorylation and signaling of the epidermal growth factor receptor (28,43,49). Src kinase is a known effector and modulator of EGF receptor signaling. In an attempt to explore the possible role of Src kinase for regulation of EGF receptor signaling by SHP-1, we observed that Src-induced tyrosine phosphorylation in HEK293 cells is very efficiently antagonized by SHP-1 coexpression. In contrast, dephosphorylation of the EGF receptor is detectable, but clearly less efficient (Fig.  1A). Abrogation of Src-dependent phosphorylations depends on the catalytic activity of SHP-1, because expression of the SHP-1 C455S mutant protein had no effect on Src-induced tyrosine phosphorylations (Fig. 1A). We considered the following possible mechanisms for the surprisingly efficient Src-SHP-1 antagonism. First, Src could be inactivated by SHP-1 coexpression, second, Src could activate SHP-1, and third, Src could generate very efficient interaction partners for SHP-1. The first possibility seemed not very likely because in fact activation of Src by SHP-1 had been reported (17). To evaluate Src activity in the absence and presence of SHP-1, we first tested the phosphorylation state of c-Src tyrosines 418 and 529, which are involved in regulation of Src activity, by immunoblotting. No obvious differences were detectable (Fig. 1B). We then assayed the activity of immunoprecipitated Src that was expressed alone or coexpressed with SHP-1 wt or SHP-1 CS in HEK293 cells. With two different exogenous substrates we could not observe any significant activity difference of Src in the absence or presence of either SHP-1 variant (Fig. 1C). Thus, inactivation of Src by SHP-1 is probably not the basis for the efficient antagonism and the down-regulation of Src-dependent phosphorylations by SHP-1 is likely to occur at the level of dephosphorylation of Src substrates.
Src Activates SHP-1-We next explored the possibility of SHP-1 activation by Src. This should occur by phosphorylation of SHP-1 at tyrosine residues. Candidate phosphotyrosines are those in the C-terminal tail of SHP-1. Phosphorylation of SHP-1 at tyrosines 536 and 564 has been reported earlier (37,38). If tyrosine phosphorylations of SHP-1 in the C-terminal tail would play a role for efficient dephosphorylation of Src substrates, respective YF mutants of SHP-1 should have reduced activity against Src substrates in coexpression assays. This is indeed the case. As shown in Fig. 2, the SHP-1 Y538F mutant (note: numbering differs from that in the above quoted papers by ϩ2 because we used the epithelial SHP-1 variant for our experiments) had a reduced capacity to suppress Src-induced phosphorylations. Mutation of another reported phosphorylation site, Tyr 566 , may also have some effect on SHP-1 activity but this was clearly minor compared with the Tyr 538 mutation. The SHP-1 Y538F,Y566F variant had a reduced activity similar to the Y538F mutant (Fig. 2). To further establish the possibility of a positive regulation of SHP-1 activity by Src, we performed in vitro phosphorylation of SHP-1 and subsequent PTP assays. A possible complication in these experiments is the capacity of SHP-1 to autodephosphorylate. To obtain phosphorylation of SHP-1 at high stoichiometry we found it necessary to inactivate SHP-1. To do this, we took advantage of the reversible inactivation of PTPs by oxidation in the absence of reducing agents (50). We found that incubation of recombinant SHP-1 at room temperature in the absence of DTT for 15 min was sufficient to nearly completely inactivate the enzyme. The activity could then be nearly fully restored by treatment with DTT (not shown). Recombinant SHP-1 was purified, inactivated, and subjected to phosphorylation by recombinant GST-Src. Thereafter, SHP-1 activity was restored with DTT and assayed against pNPP as a substrate. Under these conditions, we found an ϳ5-fold activation of free SHP-1 subsequent to phosphorylation by Src (Fig. 3). Also, a GST-SHP-1 fusion protein can be activated by Src-mediated phosphorylation (Fig. 3). In this case, the achieved Src-mediated activation is lower as with free SHP-1, presumably because GST-SHP-1 already has constitutively elevated activity. We used this setting to also analyze SHP-1 mutants. Consistent with an important role of the C-terminal tyrosines for Src-dependent activation of SHP-1, the SHP-1 Y538F,Y566F mutant could no longer be activated (Fig. 3). In summary, Src can activate SHP-1 by phosphorylation, and phosphorylation of tyrosine 538 is required for optimal activity of SHP-1 toward Src substrates in intact cells.
Src Substrates Are Good SHP-1 Substrates-We next explored the possibility that Src might generate very efficient interaction partners for SHP-1 by tyrosine phosphorylation of cellular proteins. Tyrosine-phosphorylated proteins can efficiently interact with SHP-1 in at least two ways: by binding to the SH2 domains, leading to recruitment and activation of SHP-1, or by presenting efficient substrates. The latter possibility would be associated with efficient binding to the SHP-1 catalytic domain. To explore these possibilities, we subjected lysates of Src overexpressing HEK293 cells to pull-downs with different domains of SHP-1. As shown in Fig. 4, the SHP-1 SH2 domains were very inefficient in binding Src-phosphorylated proteins. In contrast, a GST fusion protein of the catalytically inactive SHP-1 C455S catalytic domain pulled down several tyrosine-phosphorylated binding partners. A pull-down with the GST fusion protein of catalytically inactive full-length SHP-1 C455S contained abundant amounts of different Src substrates, whereas the corresponding SHP-1 wild type fusion protein interacted with a similar pattern of phosphoproteins, but much less efficiently. Taken together, these experiments suggest that the lysates of Src overexpressing cells contain multiple efficient interaction partners for SHP-1. Among single domains, binding to the catalytic domain appears more efficient than binding to the SH2 domains. An important role of the catalytic domain in recognition of Src substrates is also suggested by the strong difference in the efficiency of interaction of wild type and catalytically inactive SHP-1. Catalytic activity of SHP-1 strongly destabilizes the complexes, also suggesting that complex formation is dependent on enzyme-substrate interactions. We, therefore, considered the possibility that Src generates efficient SHP-1 substrates by virtue of the sequence selectivity for Src-mediated phosphorylation. We tested this possibility by comparing different phosphopeptides, which are good Src substrates in their unphosphorylated form for their dephosphorylation efficiency by the SHP-1 catalytic domain. Three peptides designated optimal Src substrate pep-FIG. 1. SHP-1 abrogates Src-dependent phosphorylations but does not inhibit Src activity. HEK293 cells were cotransfected with expression constructs for epidermal growth factor receptor (EGFR), or Src kinase, and active SHP-1 (wt), or the catalytically inactive SHP-1 C455S variant (CS), as indicated. A, cell lysates were subjected to immunoblotting analysis with anti-phosphotyrosine (PY), anti-Src, and anti-SHP-1 antibodies. B, the phosphorylation state of Src Tyr(P) 418 and Tyr(P) 529 in corresponding cell lysates was analyzed by immunoblotting with phosphorylation site selective antibodies. The identity of the Src band and the amounts of Src in the cell lysates were evaluated by subsequent stripping and reprobing the blots with anti-Src antibodies (lower panels). C, Src was immunoprecipitated from corresponding lysate aliquots and the activity was assayed by phosphorylation of acid-denatured enolase in the presence of [␥-32 P]ATP, and subsequent autoradiography (upper panel). Alternatively, immunoprecipitates were subjected to an activity assay using the peptide AEEEIYGEFEAKKK as substrate (lower panel). Values in this experiment are means of triplicate determinations and are corrected for the amounts of Src detected in a corresponding immunoblot. tide, YEEI peptide, and Cdc2 peptide were reported as very good Src substrates with V max /K m values of 7.76, 6.24, and 0.82, respectively (51). Interestingly, the corresponding phosphopeptides were also very good or good SHP-1 substrates with apparently the same ranking of efficiency: optimal Src substrate Ͼ YEEI peptide Ͼ Cdc2 peptide (Fig. 5, A and B, Table  I). We also tested a peptide from the sequence of p120 ctn , a component of cell-cell adhesion complexes. P120 ctn is a known substrate for both, Src and SHP-1 in intact cells (22,23). The Src phosphorylation sites of p120 ctn have recently been mapped (23) and among the different phosphorylation sites the sequence around tyrosine 296 has the highest similarity with the consensus for Src phosphorylation (45). The corresponding phosphopeptide p120 ctn pY296 was also a very good substrate for SHP-1 (Fig. 5A, Table I). These findings suggest that Src and SHP-1 display a similar selectivity with respect to the primary sequence environment of a tyrosine or phosphoty-rosine that is required for efficient phosphorylation or dephosphorylation, respectively.
Molecular Modeling of Substrate Peptide Binding-To better understand the structural basis for the described substrate selectivity of SHP-1, molecular modeling was performed taking advantage of the published three-dimensional structure of the SHP-1 catalytic domain in complex with a substrate peptide EDTLT(p)YADLD (SIRP␣pY469) (Ref. 41, Protein Data Bank code 1fpr). For these studies we focused on the two phosphopeptides with the best substrate properties, the optimal Src substrate and the p120 ctn pY296 peptide. They were docked into the catalytic domain in the same orientation as described for the co-crystallized SHP-1 substrate phosphopeptide SIRP␣pY469 (Fig. 6A). Comparative interaction energies were then calculated by minimizing the energy using an implicit solvent model. The calculated interaction energy ranking is a qualitative measure of the steric and electrostatic fit between peptide and SHP-1 and is reported as the relative rank with respect to the optimal Src substrate (Tables I and II). Interestingly, according to the calculations, both the optimal Src substrate and the p120 ctn pY296 sequences interact more favor- able with SHP-1 than the previously co-crystallized peptide SIRP␣pY469 (Table I). The calculated rank order optimal Src substrate peptide Ͼ p120 ctn pY296 peptide Ͼ SIRP␣pY469 peptide is consistent with the results obtained in dephosphorylation experiments.
To estimate the contribution of individual substrate amino acid residues for peptide-SHP1 interaction, single residues of the optimal Src substrate peptide were replaced by the corresponding ones in p120 ctn pY296 and the SIRP␣pY469 peptide (Table II). According to these docking experiments, the major contributions to efficient interaction are likely to arise from negatively charged side chains of the optimal Src substrate in positions Ϫ4, Ϫ3, and Ϫ2, which are all engaged in ionic interactions with a positively charged cluster at the SHP-1 surface formed by side chains of Arg 277 , Lys 358 , Arg 360 , and Lys 362 . Similarly, negatively charged residues in positions Ϫ4 and Ϫ3 of the p120 ctn pY296 peptide interact with this cluster. However, lack of a negative residue in position Ϫ2 and of a corresponding interaction with Arg 277 and Lys 362 may explain the somewhat lower affinity of the p120 ctn pY296 peptide. The favorable interaction energy calculated for an aspartate at the solvent exposed position Ϫ1 may be because of the hydrophilic character of aspartate compared with a hydrophobic isoleucine.
Replacement of residues C-terminal to the Tyr(P) residue from the ones in the original optimal Src substrate sequence to those in the p120 ctn pY296 and in particular those in the SIRP␣pY469 peptide is predicted to be unfavorable in most of the cases (Table II). Only at position ϩ1 is an alanine residue a better fit into a hydrophobic pocket formed by SHP-1 Ile 281 , Ile 459 , and Tyr 278 than a glycine residue. The subpocket ϩ3 is formed by Lys 259 , Ser 498 , Ile 281 , and Leu 282 . Here a phenylalanine as in the optimal Src substrate goes deep into the hydrophobic pocket. The calculations indicate that a methionine residue as in the p120 ctn pY296 peptide is less favorable than a phenylalanine in this position. A leucine residue as in the SIRP␣pY469 peptide is least favorable. The glutamate at position ϩ4 of the peptide is in contact with the positively charged surface residues Arg 494 and Lys 259 , which could be the reason for the energy increase of 9.4 kcal/mol, when replacing the glutamate by a serine residue.
According to the sequence alignment of human PTP domains, residues in several sequence motifs are highly variable and were proposed to contribute to substrate specificity (47). For SHP-1, these motifs comprise sequences containing amino acid residues Gln 254 , Lys 279 , Asn 280 , Asn 361 , His 422 , Ser 498 , and Arg 360 (Fig. 6B). Analyzing the SHP-1 catalytic domain structure in complex with phosphopeptide SIRP␣pY469 reveals that only residues Asn 280 , Ser 498 , and Arg 360 are involved in substrate binding, whereas Gln 254 , Lys 279 , Asn 361 , and His 422 have no contact with the associated substrate. Arg 360 , Asn 280 , and Ser 498 belong to the peptide substrate binding sites P Ϫ 4, P ϩ 1, and P ϩ 3, respectively. To explore the possible contribution of these catalytic domain amino acids to SHP-1 substrate selectivity, they were replaced separately by the corresponding residues of PTP1B and the interaction with the optimal Src substrate was again analyzed. Replacement of Arg 360 by a corresponding serine residue of PTP1B clearly decreased the calculated interaction energy (not shown) suggesting that this residue may be important for SHP-1 substrate selectivity. DISCUSSION The efficiency of substrate dephosphorylation of PTPs in intact cells is determined in a combinatorial way by different mechanisms. These include targeting domains that determine localization in subcellular compartments and mediate PTP substrate recruitment through protein-protein interactions. Furthermore, catalytic domains of different PTPs have different selectivity for phosphotyrosines in a certain sequence context. Finally, post-translational modifications of PTPs such as by proteolysis or phosphorylation can modify access to substrates, intrinsic activity, and possibly also substrate selectivity.
We show here that substrates that are efficiently phosphorylated by Src kinase are in turn efficient substrates for SHP-1. This was detectable for Src-phosphorylated phosphoproteins in intact cells as well as for phosphopeptides that correspond to excellent Src substrates. The phosphopeptide AEEEIpY-GEFEA (optimal Src substrate) that exactly matches the consensus sequence for optimal Src phosphorylation is to our knowledge the best currently known substrate for SHP-1 with a k cat /K m of 5.63 ϫ 10 6 M Ϫ1 s Ϫ1 . Molecular modeling studies revealed that both amino acids N-and C-terminal of the phosphotyrosine contribute to the efficient interaction. At the Nterminal side, the negatively charged side chains in positions Ϫ4, Ϫ3, and Ϫ2 are engaged in ionic interactions with a positively charged cluster at the SHP-1 surface. This observation is in agreement with a recently published study reporting screen- FIG. 5. Phosphopeptides that correspond to good Src substrates are good SHP-1 substrates. A, synthetic phosphopeptides of the following sequences RRLIEDAEpYAARG (Kit reference), AEEE-IpYGEFEA (optimal Src substrate), EEEPQpYEEEIPIYL (YEEI peptide), IGEGTpYGVVYK (Cdc2 peptide), and DDLDpYGMMSD (p120 ctn pY296 peptide), as indicated, were incubated at a concentration of 200 M with 4.4 ng/l of a GST fusion protein of the SHP-1 catalytic domain at room temperature for 5 min. The released phosphate was determined by a malachite green assay. B, example of a dephosphorylation reaction of AEEEIpYGEFEA (optimal Src substrate), monitored by tyrosine fluorescence at 305 nm (excitation at 282 nm). Reaction over time at 30°C is presented with different phosphopeptide concentrations indicated in the graph. The fluorescence is in arbitrary units.
ing of a phosphopeptide library for optimal SHP-1 substrates (42). According to these data, SHP-1 strongly prefers an acidic residue in the Ϫ2 position. In the Ϫ1 position, acidic residues are also preferred but other residues are likewise tolerated. The modeling also suggests a favorable role of an acidic residue in Ϫ1. The crystal structure of the SHP-1 catalytic domain in  The sequence of this peptide continues PIYL. Kinetic constants were determined by a fluorimetric assay (Fig. 5B, "Materials and Methods"). Because of its relatively high K m , no reliable data for the Cdc2 peptide could be obtained with the fluorimetric PTP assay. The rank given in kcal/mol corresponds to the difference of calculated peptide-PTP interaction energy compared to that of the optimal Src substrate bound to the SHP-1 domain (0 kcal/mol).
c Kinetic data taken from Ref. 41.
FIG. 6. Active binding site residues of the human SHP-1 catalytic domain in complex with the optimal Src substrate. A, catalytic domain residues are shown in a space filling model, whereas the bound peptide is shown in a stick model; the coloring is according to their hydrophobicity from blue (hydrophilic) to red (hydrophobic). The docking configuration corresponds to the position of local minimum potential energy, starting from the crystallographic backbone orientation of the SIRP␣pY469 peptide bound to the SHP-1 catalytic domain (Ref. 41, Protein Data Bank code 1fpr). The structures of the catalytic domain in the three modeled complexes (SHP-1 with SIRP␣pY469, p120 ctn pY296, and optimal Src substrate) were almost identical, with a root mean square deviation of ϳ0.1 Å. They were also similar to the crystallographic catalytic domain structure with an root mean square deviation of 0.74 Å. B, a number of PTP catalytic domain residues were previously proposed to contribute to substrate selectivity (47); these comprise residues Arg 24 , Arg 47 , Asp 48 , Lys 116 , Gly 117 , Ser 118 , Leu 119 , Phe 182 , Met 258 , and Gly 259 of PTP1B corresponding to the residues Gln 254 , Lys 279 , Asn 280 , Lys 358 , Gly 359 , Arg 360 , Asn 361 , His 422 , Ser 498 , and Gly 499 , respectively, of SHP-1 (numbering for the 595-amino acid hematopoietic SHP-1 variant). Of these residues, the positions of those that differ in SHP-1 from PTP1B are shown in the complexes with the optimal Src substrate.

Role of individual amino acid residues in substrate peptides for interaction with the SHP-1 catalytic domain
Calculated changes of relative binding energy (rank in kcal/mol), when replacing single residues of the Src optimal peptide by the corresponding residues in the p120 ctn pY296 peptide or the SIRP␣pY469 peptide, as indicated.
Position relative to Tyr(p) (0) Amino acid exchange Src optimal peptide 3 p120 ctn pY296 Rank (kcal/mol) Src optimal peptide: 0 Amino acid exchange Src optimal peptide 3 pY496 Rank (kcal/mol) Src optimal peptide: 0 complex with substrate peptides derived from the sequence of the physiological substrate SHPS-1/SIRP␣ also highlights the importance of acidic amino acids N-terminal of the phosphotyrosine for efficient interaction (41). Notably, according to the structures, acidic residues in the Ϫ4 position can form a salt bridge with Arg 360 in the SHP-1 catalytic domain. In the Ϫ2 position, the peptides in the co-crystals contain hydrophobic residues. According to our modeling studies and in agreement with the library data such hydrophobic residues while obviously compatible with substrate recognition are less favorable for binding to the SHP-1 catalytic domain. In this point some caution is, however, required because the modeling did not take into account conformational changes of the catalytic domain structure. The crystal structures revealed a considerable degree of conformational flexibility of the domain, particularly N-terminal of the phosphotyrosine pocket (41). Still, an unfavorable role of a hydrophobic residue in Ϫ2 is also in agreement with the less efficient dephosphorylation of the p120 ctn pY296 and the SIRP␣pY469 peptides compared with the optimal Src substrate peptide (Table I). Hitherto, rather little is known about the contribution of residues C-terminal of phosphotyrosine for SHP-1 substrate recognition. Comparing a series of phosphopeptide substrates derived from sequences of different receptor tyrosine kinases, Dechert et al. (35) suggested a preference of SHP-1 for C-terminal acidic residues. The modeling study clearly supports a role of C-terminal residues for substrate recognition, however, the quantitative contribution of these interactions may be lower than of those at the N-terminal side. Importantly, all tested alterations from the consensus for efficient Src phosphorylation, except replacement of glycine in ϩ1 by alanine, led to a decrease in SHP-1 interaction efficiency. The major contributions came from residues in ϩ3 and ϩ4. A phenylalanine side chain in ϩ3 is effectively accommodated by the ϩ3 subpocket of the SHP-1 catalytic domain. On the other hand, phenylalanine in ϩ3 is particularly strongly preferred for Src-mediated substrate phosphorylation. The excellent phosphopeptide substrates we have found match only partially the consensus sequence for SHP-1 proposed by Yang et al. (41): (D/E)X(L/I/V)X n pYXX(L/I/V) (n ϭ 1 or 2). Taken together with previously published data and our results, a consensus (D/ E)(D/E)(D/E/L)XpYXX(F/h)(D/E) (X, any amino acid; h, hydrophobic; pY, phosphotyrosine) may be more reasonable. Do our findings indicate a possible substrate selectivity of SHP-1 in comparison with other PTPs? A preference for peptide substrates with acidic residues N-terminal of the phosphotyrosine has been described for a number of PTPs analyzed so far, including PTP1B (52-54), HPTP␤ (55,56), and TC-PTP (57). PTP1B exhibits a remarkable plasticity, allowing efficient interaction with substrates having quite variable residues in Ϫ1, including acidic and aromatic residues (54,58). Phosphopeptides with a positive residue in Ϫ4 were very poorly hydrolyzed by the transmembrane PTP Dep-1 (59). Thus, the SHP-1 catalytic domain selectivity for N-terminal residues seems to be similar to that of multiple other PTPs. There may, however, be subtle differences. Interestingly, exchanging in the model SHP-1 Arg 360 , which is part of the Ϫ4 subpocket, to serine, the corresponding residue in PTP1B, led to a drop in the efficiency of interaction with both good SHP-1 substrates, optimal Src peptide and p120 ctn pY296 (not shown). With respect to residues at the C-terminal side, several studies reported that they may be of lesser importance for PTP substrate recognition. For example, an Ala scan of a phosphopeptide derived from the EGF receptor sequence around Tyr(P) 992 revealed only minor effects of C-terminal substitutions for substrate efficiency with PTP1B (53). Also, truncation of sequence from ϩ4 on in a lysozyme-derived phosphopeptide only weakly affects dephos-phorylation efficiency by PTP1B (52). For the receptor-like PTP RPTP␣, even the complete truncation of the C-terminal sequence in an EGF receptor-derived phosphopeptide led only to a 2-fold drop in activity (60). On the other hand, for several PTPs basic residues C-terminal of the phosphotyrosine were reported to impair the performance of substrate interaction, including a basic residue in ϩ2 for HPTP␤ (56) and in ϩ3 for Dep-1 (59). For PTP1B, a preference for a second phosphotyrosine in ϩ1 in a sequence context DEpY(0)pY(ϩ1)R/K has been described recently (61). This preference forms the basis for efficient dephosphorylation of kinase activation loops harboring tandem phosphotyrosines, such as in the insulin receptor (61) and JAK2 (62,63). Although a stringent comparison remains to be performed, it is possible that our finding of a preference of SHP-1 for a phenylalanine in ϩ3 and an acidic residue in ϩ4 defines a substrate selectivity of SHP-1 compared with other PTPs. Substrate selectivity regulation by a residue in this part of the PTP catalytic domain has also been described by Peters et al. (64). Glycine 259 in PTP1B that corresponds to glycine 499 in SHP-1 allows a wider substrate recognition opposed to a glutamine in this position in RPTP␣.
Apart from substrate selectivity, another reason for the efficient counteraction of Src phosphorylations by SHP-1 is the Src-mediated tyrosine phosphorylation of SHP-1 at the C terminus. In vitro experiments indicated that C-terminal tyrosine-phosphorylated SHP-1 has an elevated catalytic activity with pNPP as a substrate. This observation is in agreement with earlier reports. Activation of SHP-1 by C-terminal phosphorylation has been described by Uchida et al. (38) in 1993. Lorenz et al. (37) demonstrated that the Src family kinase Lck can phosphorylate SHP-1 at tyrosine 536 to high stoichiometry, however, effects on SHP-1 activity were not analyzed in this study. Very recently, Zhang et al. (65) reported the elevated activity of semisynthetic SHP-1 proteins containing phosphotyrosine mimetics at the Tyr 536 and Tyr 564 sites (numbering for hematopoietic version). The semisynthetic modification circumvented the problem of SHP-1 autodephosphorylation. Activation was mainly attributed to modification of Tyr 536 , which led to 4 -8-fold stimulation of activity. As a mechanism of activation the authors propose an intramolecular interaction with the N-terminal SH2 domain. It is likely that the lower activity of a SHP-1 Y538F mutant toward Src substrates in intact cells observed by us is because of a reduced intrinsic SHP-1 activity. Interestingly, Src overexpression did not generate efficient ligands for the SHP-1 SH2 domains. It is tempting to speculate that C-terminal phosphorylation of SHP-1 plays a role for regulation of SHP-1 activity particularly in such a scenario, whereas C-terminal phosphorylation may be less important in the presence of efficient SH2 domain ligands. In addition, tyrosine phosphorylation of SHP-1 may also take part in substrate recruitment and should be impaired in the case of the C-terminal Y538F mutant. For RPTP␣, substrate recruitment via a C-terminal tyrosine phosphorylation site has been shown (66). The binding of a 30/32-kDa phosphoprotein in hematopoietic cells to the SHP-1 C-terminal tail has been reported earlier (40). Also, the phosphatidylinositol 3-kinase subunit p85 can bind to the SHP-1 C terminus and is in turn dephosphorylated (67).
In our experiments, the activity of Src was unaffected by coexpressed SHP-1. Somani et al. (17) reported earlier that SHP-1 has the capacity to activate Src by C-terminal dephosphorylation. This conclusion was based on the analysis of Src in thymocytes and other cells of wild type and SHP-1-deficient mice as well as in vitro experiments (17). Appropriate cellular backgrounds and expression levels may be critical to observe SHP-1-dependent Src activation. On the other hand, the activity of the Src family kinase Lck is inhibited by SHP-1 that has the capacity to directly dephosphorylate the Lck activating phosphotyrosine 394 (14). In addition, SHP-1 inhibits Lck signaling by dephosphorylation of phosphatidylinositol 3-kinase p85 (67). Thus, SHP-1 directly inactivates Lck, whereas it does not inactivate Src. The systems have, however, in common that both Src and Lck kinase have the capacity to activate SHP-1 by phosphorylation. In turn, kinase-dependent signaling is effectively suppressed by SHP-1.
What may be the biological relevance of the described Src-SHP-1 interaction? There are several examples of kinases that activate antagonizing phosphatases, for example, Erk2 activates the dual specificity phosphatase MKP3 by direct binding (68). Such regulatory loops may play an important role to determine intensity and duration of signaling events (69). A prediction from our findings would be that Src-dependent signals should be very transient in the presence of SHP-1. Under conditions where SHP-1 can even activate Src (17) they may also be of higher intensity. There is at least one known common substrate of Src and SHP-1, p120 ctn , which was discovered as a prominent Src substrate. The Src phosphorylation sites on p120 ctn have recently been mapped (23). SHP-1 can bind to p120 ctn and dephosphorylate it. We show here that one phosphopeptide from the p120 ctn sequence, p120 ctn pY296, is an excellent SHP-1 substrate. P120 ctn -SHP-1 interaction may take place at the cytoplasmic face of the plasma membrane or in the cell nucleus, into which both molecules can translocate in epithelial cells. The association of SHP-1 to p120 ctn could not be assigned to a single p120 ctn phosphotyrosine residue, rather the p120 ctn tyrosine phosphates had an additive effect on SHP-1 binding (23). These findings may reflect a predominant role of the SHP-1 catalytic domain in recognition of multiple phosphorylation sites on p120 ctn . Our observations further support the concept of a combinatorial role of different domains in PTPs for the selective substrate interaction.