Protein-tyrosine Phosphatase ϵ Regulates Shc Signaling in a Kinase-specific Manner

Individual protein tyrosine kinases and phosphatases target multiple substrates; this may generate conflicting signals, possibly within a single pathway. Protein-tyrosine phosphatase ϵ (PTPϵ) performs two potentially opposing roles: in Neu-induced mammary tumors, PTPϵ activates Src downstream of Neu, whereas in other systems PTPϵ can indirectly down-regulate MAP kinase signaling. We now show that the latter effect is mediated at least in part via the adaptor protein Shc. PTPϵ binds and dephosphorylates Shc in vivo, reducing the association of Shc with Grb2 and inhibiting downstream ERK activation. PTPϵ binds Shc in a phosphotyrosine-independent manner mediated by the Shc PTB domain and aided by a sequence of 10 N-terminal residues in PTPϵ. Surprisingly, PTPϵ dephosphorylates Shc in a kinase-dependent manner; PTPϵ targets Shc in the presence of Src but not in the presence of Neu. Using a series of point mutants of Shc and Neu, we show that Neu protects Shc from dephosphorylation by binding the PTB domain of Shc, most likely competing against PTPϵ for binding the same domain. In agreement, PTPϵ dephosphorylates Shc in mouse embryo fibroblasts but not in Neu-induced mammary tumor cells. We conclude that in the context of Neu-induced mammary tumor cells, Neu prevents PTPϵ from targeting Shc and from reducing its promitogenic signal while phosphorylating PTPϵ and directing it to activate Src in support of mitogenesis. In so doing, Neu contributes to the coherence of the promitogenic role of PTPϵ in this system.

The human genome contains 90 genes that encode tyrosine kinases (1) and 107 genes that encode tyrosine phosphatases (2). Similar numbers of genes from either category generate products that act on proteins, 85 PTKs 2 and 81 PTPs (2). Tyro-sine kinases and phosphatases are then vastly outnumbered by their potential substrates. These enzymes target multiple substrates each and may play distinct, sometimes opposite, roles in different physiological circumstances. One such example is PTP1B, which down-regulates insulin receptor signaling and several other growth factor receptor pathways (3,4) but up-regulates Ras signaling (5). Taken to an extreme, a single enzyme may activate a signaling pathway by targeting one substrate but simultaneously contribute toward inactivating the same pathway via a second substrate.
Nevertheless, the outcomes of signaling pathways are specific, coherent, and reproducible, implying that molecular mechanisms exist to prevent or to control situations where opposite signals can be generated by the same enzyme. Much is known about regulation of PTPs by mechanisms such as proteolysis (6,7), inhibitory dimerization (8 -13), reversible oxidation (14 -16), and phosphorylation (13,(17)(18)(19). However, it remains unclear how the potential conflicts between the activities of a particular PTP on several of its substrates within a single system are prevented or resolved.
The two major protein forms of PTP⑀ are the receptor type (RPTP⑀) and non-receptor type (cyt-PTP⑀) enzymes (20 -22). Both forms can support or inhibit mitogenic signaling in a context-dependent manner. We have shown that RPTP⑀ supports the transformed phenotype of mouse mammary tumors induced in vivo by an activated Neu transgene. RPTP⑀ activates c-Src; lack of RPTP⑀ in Neu-induced mammary tumor cells reduced c-Src activity, caused the cells to display aberrant morphology, and reduced their proliferation rate in culture and in vivo (23). Activation of c-Src by RPTP⑀ was shown to be dependent upon phosphorylation of the PTP at its C-terminal tyrosine Tyr 695 by Neu (24). A Neu-RPTP⑀-Src signaling pathway exists therefore in mammary tumor cells induced by Neu, in which Neu phosphorylates RPTP⑀, thereby driving the phosphatase to activate c-Src and to contribute to the transformed phenotype of these cells (24). On the other hand, a clear role for RPTP⑀ and cyt-PTP⑀ in down-regulating signaling events was shown in the context of insulin receptor signaling (25)(26)(27), ERK and MAPK signaling (28,29), endothelial cell proliferation (30), and transformation and tumorigenicity of M1 leukemia cells (31). The molecular mechanisms that regulate and balance between these opposing roles remain unclear.
The ShcA gene yields three protein products, p46, p52, and p66. Shc proteins can associate with a variety of activated tyrosine kinase growth factor receptors and can be phosphorylated by them on tyrosines Tyr 239 /Tyr 240 /Tyr 317 (numbered as in p52 Shc). The phosphorylated Shc proteins serve as adaptors, leading to recruitment of the Grb2-SOS complex and resulting in activation of downstream signaling. p66 Shc was found to be a proapoptotic factor activated in response to oxidative stress (32). In this study, we show that PTP⑀ dephosphorylates the three protein products of the ShcA gene following phosphorylation by Src. In so doing, PTP⑀ also inhibits the Shc-Grb2 association and phosphorylation of ERK1 and ERK2 further downstream. In the context of Neu-induced tumor cells, this activity of RPTP⑀ should convey an antimitogenic signal, thus antagonizing the promitogenic contribution of RPTP⑀ through its activation of Src. Surprisingly, RPTP⑀ does not target Shc in cells in which the predominant kinase is Neu, including in Neutransformed mammary tumor cells, due to strong physical interaction between Neu and Shc. We conclude that Neu balances between two opposing activities of RPTP⑀. Neu phosphorylates RPTP⑀ and drives it to activate Src while binding Shc tightly and preventing the phosphatase from dephosphorylating the adaptor molecule. As a result, Neu prevents the potential conflict between these pro-and antimitogenic activities of PTP⑀ in mammary tumors, enabling PTP⑀ to support mitogenic signaling in a self-consistent and coherent manner.
Cell Culture-Human embryonic kidney (HEK293) cells and SYF cells (immortalized mouse fibroblasts lacking Src, Yes, and Fyn (39)) were grown in Dulbecco's modified Eagle's medium (DMEM) (Invitrogen) supplemented with 10% fetal calf serum (Invitrogen), 4 mM glutamine, 50 units/ml penicillin, and 50 g/ml streptomycin. HEK cells were transfected by the calcium phosphate technique as described (33), whereas SYF cells were transfected by the FuGENE 6 transfection reagent (Roche Applied Science). Metabolic labeling of cells was performed at 70% confluence; cells were incubated for 30 min in methioninedeficient DMEM, followed by overnight incubation in similar medium supplemented with [ 35 S]methionine (175 Ci (1000 Ci/mmol) per 5 ml of medium per plate). Spontaneously immortalized embryonic fibroblasts were prepared from PTP⑀deficient or wild-type mice by the 3T3 method. NIH3T3 cells were transformed by expression of v-Ha-Ras. Both types of cells were grown in DMEM supplemented as above, which, for the embryonic fibroblasts, also contained 6 ϫ 10 Ϫ5 M ␤-mercaptoethanol. Mammary tumor cells isolated from spontaneous tumors of mice carrying the V664E Neu transgene and genetically lacking PTP⑀ (23) were grown in DMEM supplemented with 10% fetal calf serum (Hyclone, Logan, UT) and glutamine and antibiotics as above. RPTP⑀ was expressed in these cells by retrovirus-mediated infection as described (23).
Immunoprecipitation and Protein Blot Analysis-Cells or tissues were lysed in buffer A (50 mM Tris-HCl, pH 7.5, 100 mM NaCl, 1% Nonidet P-40), supplemented with protease inhibitors (1 mM N-(␣-aminoethyl) benzene-sulfonyl fluoride, 40 M bestatin, 15 M E64, 20 M leupeptin, 15 M pepstatin; Sigma) and 0.5 mM sodium pervanadate. Pervanadate was replaced with 5 mM sodium iodoacetate when the association between Shc and D245A cyt-PTP⑀ was examined. Protein amounts were determined using the Bio-Rad protein assay system with bovine serum albumin as a standard. Immunoprecipitation was performed using 1-3 mg of total cell lysate and protein A and/or protein G-Sepharose beads (Amersham Biosciences or Santa Cruz Biotechnology), as described (24). Eukaryotic GST pulldown studies were performed by rocking 800 g of cellular proteins with glutathione-agarose beads (Pierce) that had been previously blocked in 1% skim milk, followed by three washes in buffer A. Immunoprecipitated material and crude cell lysates were subject to SDS-PAGE and blotting as described (7).
Purification of Bacterial GST Fusion Proteins-Bacterial GST fusion proteins were grown in Escherichia coli DH5␣ bacteria and purified by binding to glutathione-agarose beads (Pierce) previously blocked in 1% milk. Following three washes in NETE buffer (0.5% Nonidet P-40, 20 mM Tris, pH 8, 100 mM NaCl, 1 mM EDTA), proteins were eluted in 20 mM glutathione, 50 mM Tris, pH 8.0. The quantity and purity of the purified proteins were examined by SDS-PAGE followed by Gel-code staining (Pierce).
In Vitro Dephosphorylation of Shc-Eukaryotic GST-Shc fusion proteins were expressed in HEK293 cells together with Neu or Y527F Src. Cells were lysed in buffer A as above, supplemented with sodium pervanadate, and rocked gently with glutathione-agarose beads (Pierce) for 3 h followed by 10 washes in buffer A. FLAG-tagged cyt-PTP⑀ was expressed in HEK293 cells and was immunoprecipitated using anti-FLAG M2 affinity beads (Sigma) followed by three washes with buffer A, two washes with buffer B (100 mM KCl, 0.5 mM EDTA, 20 mM Hepes, pH 7.6, 0.4% Nonidet P-40, 20% glycerol), and two washes with buffer 54K (150 mM NaCl, 50 mM Tris, pH 7.9, 0.5% Triton X-100), all without pervanadate supplement. FLAGtagged cyt-PTP⑀ was then eluted by incubation of the beads with Elution Buffer (20 mM Hepes, pH 7.3, 200 mM NaCl, 1 mM EDTA, 0.1 mM EGTA, 1 mg/ml FLAG peptide) (Sigma). The integrity, purity, and amount of GST-Shc and PTP⑀ were analyzed by SDS-PAGE followed by silver staining (Bio-Rad) using a bovine serum albumin standard curve. Dephosphorylation reactions were conducted in 150 l of activity buffer (50 mM MES, pH 7, 0.5 mM dithiothreitol, 0.5 mg/ml bovine serum albumin), to which 100 ng of purified, eluted cyt-PTP⑀ and a similar amount of Shc bound to beads were added. Following gentle rocking at 32°C for 7 h, samples were analyzed by SDS-PAGE.

Shc and PTP⑀ Associate in Vivo-Previous studies have
shown that PTP⑀ can down-regulate MAPK signaling but does so indirectly (28), most likely by targeting an upstream molecule(s). In order to examine this issue, we studied the effect of PTP⑀ on the adaptor protein Shc as a possible candidate for mediating these effects. Endogenous Shc proteins formed stable associations with endogenous cyt-PTP⑀ or RPTP⑀ in several tissues and cell types, including in Ras-transformed NIH3T3 cells, Jurkat T-cells, and mouse brain (Fig. 1). p66 Shc and p52 Shc bound PTP⑀; p46 Shc was not visible in these experiments due to interference of the precipitating antibody heavy chain. Similar association was observed in HEK293 cells, which express endogenous p46 and p52 Shc, upon expression of fulllength, WT cyt-PTP⑀ or of various other forms of PTP⑀ ( Fig.  2A). The substrate-trapping mutant D245A cyt-PTP⑀ (34,40) bound cyt-PTP⑀ as strongly as WT Shc; similar binding was observed also when the inactive R283M cyt-PTP⑀ or Y683F cyt-PTP⑀, which lacks the key phosphorylation site analogous to Tyr 695 in RPTP⑀, was used ( Fig. 2A). These results indicate that a stable complex containing both cyt-PTP⑀ and Shc exists and that this interaction does not require PTP⑀ activity or C-terminal phosphorylation of the phosphatase. In contrast, the receptor RPTP⑀ and the shorter p67 form associated with Shc less strongly ( Fig. 2A). The only structural difference between cyt-PTP⑀ and p67 is the presence of a unique sequence of 27 amino acid residues at the N terminus of cyt-PTP⑀ (36). In order to determine whether sequence motifs present in this region affect binding to Shc, we examined the ability of a series of N-terminal mutant cyt-PTP⑀ proteins to bind Shc. cyt-PTP⑀ truncation mutants lacking the first 4, 8, or 11 amino acids tended to bind Shc more strongly than full-length cyt-PTP⑀. Mutants of cyt-PTP⑀ lacking residues 11-27 or lacking the first 22 residues bound Shc similar to p67 (Fig. 2B). We conclude that sequences between positions 12 and 22 of cyt-PTP⑀ are important for binding Shc, whereas residues 1-4 of cyt-PTP⑀ may inhibit this association. No obvious sequence motifs that bind PTB domains are present within this region, possibly suggesting that these residues contribute indirectly to the Shc-cyt-PTP⑀ association, as discussed further below.
The Shc-PTP⑀ interaction was recapitulated also when bacterially produced GST fusion proteins of various domains of Shc were used to pull down cyt-PTP⑀ from lysates of eukaryotic cells. A fusion protein containing the PTB domain of Shc bound cyt-PTP⑀ strongly, whereas a protein containing the Shc SH2 domain did not bind cyt-PTP at all (Fig. 2C). Interestingly, a GST fusion of full-length p52 Shc bound cyt-PTP⑀ less well than a fusion protein containing only the PTB domain. These results indicate that the Shc-PTP⑀ interaction is mediated by the PTB domain of Shc and suggest that other sequences within Shc may antagonize this interaction. Similar binding was observed when the GST-Shc fusion proteins were used to pull down hyperphosphorylated cyt-PTP⑀ from cells pretreated FIGURE 1. Endogenous PTP⑀ interacts with endogenous Shc. A, Ras-transformed NIH3T3 cell lysates, which express endogenous cyt-PTP⑀, were precipitated with PTP⑀ antibodies, and precipitates were blotted against Shc (top) or PTP⑀ (bottom; the arrow marks cyt-PTP⑀). Lys, cell lysate; Ab, precipitating antibody; IP, immunoprecipitation; WB, Western blot. B, similar to A using Jurkat T-cells, documenting co-immunoprecipitation of endogenous cyt-PTP⑀ with Shc in these cells as well. The arrow (bottom) marks cyt-PTP⑀. C, RPTP⑀ binds Shc in brain. p66 and p52 Shc were precipitated from lysates of brains of WT and PTP⑀-deficient (KO) mice. Precipitates were blotted against PTP⑀ (top) and Shc (bottom). In all panels, the asterisks mark the heavy chain of the precipitating antibody, which prevents viewing of p46 Shc; molecular mass markers in this and in subsequent figures are in kDa.
with sodium pervanadate, indicating that the Shc-cyt-PTP⑀ interaction is independent of tyrosine phosphorylation of the phosphatase (Fig. 2C). The Shc PTB domain binds the nonreceptor type phosphatase PTP-PEST by targeting a tyrosineless NPLH sequence motif located at the C terminus of the PTP (41). cyt-PTP⑀ contains a similar sequence, NPSH, at positions 266 -269 within its D1 PTP domain. However, mutating the above four residues to alanines did not affect binding of cyt-PTP⑀ to Shc, indicating that the NPSH motif is not a target of the Shc PTB domain (not shown). Replacement of tyrosines 239, 240, or 317 on Shc, which play important roles in binding Grb2 and in Shc signaling (42), with nonphosphorylatable phenylalanine residues also had no effect on cyt-PTP⑀ binding (not shown). Together with the strong ability of bacterially produced GST-Shc proteins to bind cyt-PTP⑀ (Fig. 2C), this result indicates that tyrosine phosphorylation of Shc is also not required for the interaction with cyt-PTP⑀. In all, we conclude that the Shc-cyt-PTP⑀ interaction is mediated by the PTB domain of Shc and is supported by residues 12-22 of cyt-PTP⑀; the interaction is constitutive and is independent of the tyrosine phosphorylation status of either Shc or cyt-PTP⑀.
cyt-PTP⑀ Reduces Shc Phosphorylation in a Kinase-specific Manner-In order to examine the physiological consequences of the Shc-cyt-PTP⑀ interaction, we expressed the phosphatase in cells and examined its effect on phosphorylation of endogenous Shc in 293 cells. Basal phosphorylation of Shc in this system is low, hence Shc was activated by co-expression of PTKs. When the endogenous p46 and p52 forms of Shc were phosphorylated in the presence of activated (Y527F) Src, cyt-PTP⑀ reduced their phosphorylation levels significantly (Fig.  3A). Catalytically inactive R283M cyt-PTP⑀ did not reduce Shc phosphorylation, indicating that this effect requires PTP⑀ activity. Similar results were obtained also when wild-type Src was used (not shown) and when experiments were repeated in fibroblasts lacking endogenous Src, Fyn, and Yes (Fig.  3B). The latter cells also express p66 Shc, whose phosphorylation was reduced in the presence of cyt-PTP⑀. Interestingly, the presence of cyt-PTP⑀ did not reduce phosphorylation of Shc when the latter was phosphorylated by Neu (Fig. 3, C  and D). The distinction between Neu and Src also affected downstream signaling events; the presence of cyt-PTP⑀ reduced the association of Shc with Grb2 and reduced activation of ERK1 and ERK2 in the presence of Src but not in the presence of Neu (Fig. 3D). Similar effects and a similar distinction between Neu and Src were observed when cyt-PTP⑀ was replaced with RPTP⑀ (Fig. 3, E and F). In separate experiments, cyt-PTP⑀ could not reduce phosphorylation of Shc in the presence of the epidermal growth factor receptor (not shown), suggesting that the "protective" effect of Neu may be shared also by related kinases.
Importantly, PTP⑀ affected Shc phosphorylation in a kinasedependent manner also when endogenous proteins were examined. Phosphorylation of Shc and of ERK was increased in spontaneously immortalized embryonic fibroblasts from mice genetically lacking PTP⑀ (28,34) in comparison with WT cells that express the phosphatase (Fig. 4A). In contrast, phosphorylation of Shc was consistently similar in mammary tumors induced in vivo by activated Neu in mice lacking PTP⑀ (23) and in the same cells reconstituted with RPTP⑀ (Fig. 4B), in line with observations presented in Fig. 3, D and E. Since cyt-PTP⑀ and  1 and 3). The amount of precipitated cyt-PTP⑀ (panels 2 and 4) and input of GST fusion proteins (panel 5) are also shown. Cells were either treated (ϩV) or not (ϪV) with 0.5 mM sodium pervanadate prior to lysis. Note that pervanadate treatment increases binding of GST-Grb2 to cyt-PTP⑀, which is dependent upon PTP⑀ phosphorylation (33). IP, immunoprecipitation; WB, Western blot. FEBRUARY 22, 2008 • VOLUME 283 • NUMBER 8 RPTP⑀ affect Shc phosphorylation and downstream signaling in similar manners (Fig. 3, D and E) (results not shown), we believe that this difference in Shc phosphorylation is due to the presence of Neu as the dominant kinase in these mammary tumor cells (Fig. 4B) versus the absence of a single dominant kinase in immortalized fibroblasts (Fig. 4A). We note that in some experiments phosphorylation of ERK1 and ERK2 was unchanged in PTP⑀-deficient mammary tumor cells (e.g. see Fig. 4B), whereas in other studies of these cells, ERK phosphorylation was increased relative to cells expressing RPTP⑀. This suggests that RPTP⑀ affects phosphorylation of ERK in this system not only through Shc.

PTP⑀ Dephosphorylates Shc in a Context-dependent Manner
Further studies focused on cyt-PTP⑀. A possible explanation for the kinase-dependent ability of cyt-PTP⑀ to dephosphoryl-ate Shc is that Src and Neu phosphorylate Shc at distinct sites, some of which may not be dephosphorylated by cyt-PTP⑀. In order to study this possibility, we examined phosphorylation of Shc at tyrosine residues 239, 240, and 317 using phospho-specific antibodies. Neu and Src phosphorylated Shc at these sites equally well (Fig. 5A); in agreement with previous results, cyt-PTP⑀ reduced phosphorylation at Tyr 239/240 and Tyr 317 in the presence of Src but not in the presence of Neu (Fig. 5A). This result indicates that although both kinases can phosphorylate Shc at these critical sites, cyt-PTP⑀ is nevertheless prevented from dephosphorylating them in the presence of Neu.
In separate experiments, analysis by two-dimensional gel electrophoresis revealed different migration patterns along the pI axis for Shc isolated from cells expressing Src versus Neu (not shown), strongly suggesting that these kinases do modify Shc covalently, directly or indirectly, in manners that are at least partially distinct. Nevertheless, these distinct modification patterns do not account for kinase-dependent dephosphorylation of Shc, since kinase specificity was not detected when purified Shc was dephosphorylated by cyt-PTP⑀ in vitro. In these exper-  iments, Shc and cyt-PTP⑀ were expressed separately in HEK293 cells expressing activated Src or Neu, precipitated, and washed extensively; their integrity, purity, and (in the case of Shc) phosphorylation were verified by SDS-PAGE analyses followed by protein blotting or silver staining. When Shc was incubated with cyt-PTP⑀, in vitro kinase specificity was not observed; phospho-Shc was dephosphorylated by cyt-PTP⑀ irrespective of whether Shc had been phosphorylated previously by Src or Neu (Fig. 5B). This last result indicates that the inability of cyt-PTP⑀ to dephosphorylate Shc in the presence of Neu is not due to covalent modification of Shc itself but rather possibly to interactions of Shc with another molecule.
Neu Binds Shc and Protects It from Dephosphorylation by cyt-PTP⑀-In order to examine molecules that associate with Shc, we expressed this protein in cells that co-expressed either Neu or Src and that had been metabolically labeled with [ 35 S]methionine. Analysis of proteins that co-precipitated with Shc in Neu-expressing cells revealed prominent bands at 180 kDa and at 25 kDa; in agreement with significant previous data published by numerous groups, these bands were subsequently identified by protein blotting as Neu and Grb2, respectively ( Fig. 5C) (data not shown). Neu and Grb2 bound Shc strongly and similarly both in the presence or absence of cyt-PTP⑀. Shc associated with Grb2 also when Neu was replaced with Src, but in this case co-expression of cyt-PTP⑀ reduced this association significantly; no co-precipitation of Shc and Src was observed (Fig. 5C). These results suggest that Neu itself may be the molecule whose association with Shc protects it from dephosphorylation by cyt-PTP⑀. Alternative possibilities, such as sequestration of PTP⑀ by Neu, are less probable, since Neu and Src do not bind PTP⑀ or do so very weakly (24) (data not shown).
A significant part of the interactions between Shc and Neu is mediated by the PTB domain of Shc binding individual tyrosine autophosphorylation sites located at the C-terminal region of Neu. Tyrosines 1226/1227 of Neu (numbering as in Ref. 37) are prominent in this respect (37,43,44), but binding to additional tyrosines has been described (45,46). The role of the Shc SH2 domain in binding Neu is not clear, although it has been reported to assist binding (45,46). We have not observed a significant contribution of the Shc SH2 domain in this respect (see Fig. 7). Shc and Src have been reported to interact under specific conditions (25,47,48); although Src does phosphorylate Shc (Figs. 3-5) (42, 49), we have not observed stable association between both proteins in our system (Fig. 5C) (data not shown). In agreement with the above information, our results suggest that Neu protects Shc from dephosphorylation by physically binding Shc and that Src does not bind Shc as tightly and hence cannot protect it. According to this model, the presence of Src should not prevent Neu from binding Shc and from protecting it from dephosphorylation, leading us to expect that cyt-PTP⑀ would not dephosphorylate Shc in the presence of both kinases. In order to evaluate this possibility, we examined phosphorylation of endogenous Shc in HEK293 cells in the presence of both kinases (Fig. 6). Upon expression of Src, robust FIGURE 5. Mechanistic dissection of the kinase dichotomy. A, PTP⑀ dephosphorylates Shc at sites Tyr 239/240 and Tyr 317 . HEK293 cells expressing cyt-PTP⑀, Y527F Src, and Neu, as indicated, were analyzed for phosphorylation of Shc at Tyr 239/240 and Tyr 317 using phospho-specific antibodies. B, purified PTP⑀ can dephosphorylate Shc in vitro irrespective of the phosphorylating kinase. GST-Shc was purified from HEK293 cells expressing Y527F Src or Neu. Following incubation with cyt-PTP⑀ that had been purified from separate cells, phosphorylation of GST-Shc was analyzed by protein blotting (top). Also shown are amounts of purified GST-Shc and PTP⑀ present (bottom). C, Neu, but not Src, adheres tightly to Shc. HEK293 cells expressing cyt-PTP⑀, FLAGtagged Shc, Neu (N), or activated Src (S), as indicated, were labeled with [ 35 S]methionine. Shc was precipitated via its FLAG tag, and the precipitates were analyzed by SDS-PAGE on a 6 -15% gradient gel. Proteins were transferred to membrane, which was then exposed to film. Proteins noted were identified by probing the membrane with specific antibodies (not shown). IP, immunoprecipitation; WB, Western blot. phosphorylation of Shc was detected; phosphorylation was reduced significantly upon co-expression of cyt-PTP⑀ (Fig. 6,  lanes 3 and 4). However, upon added expression of Neu, Shc phosphorylation was not reduced and remained at levels observed in cells expressing Src alone (Fig. 6, lanes 3, 4, and 6). This level of phosphorylation was similar to that observed in cells expressing Neu, with or without added cyt-PTP⑀ (Fig. 6,  lanes 1 and 7). This result indicates that the "protective" effect of Neu dominates over the "permissiveness" of Src.
An additional prediction of the above model is that Neu that cannot physically bind Shc would be unable to protect it from dephosphorylation. In order to examine this possibility, we turned to catalytically active Neu in which all known autophosphorylation sites had been mutated (Y1024F,Y1144F,Y1201F, Y1226F,Y1227F,Y1253F Neu; NYPD Neu (37)) and which cannot bind Shc. Indeed, NYPD Neu did not protect Shc from dephosphorylation by cyt-PTP⑀ in the presence of Src and did not prevent down-regulation of ERK1 and ERK2 further downstream (Fig. 6, lanes 3-5).
In order to confirm the results obtained using NYPD Neu, we disrupted the Neu-Shc association in separate experiments by mutating Shc. For this purpose, we used S154P Shc, whose mutated PTB domain cannot bind phosphotyrosine but maintains its phosphotyrosine-independent interactions (50). S154P Shc was FLAG-tagged to distinguish it from endogenous Shc. As expected, S154P Shc bound significantly less Neu and was less phosphorylated by the kinase than wild-type Shc or Shc in which the SH2 domain had been inactivated (R401L Shc; Fig.  7A, top two panels). S154P and R401L Shc bound cyt-PTP⑀ similarly to wild-type Shc in the presence of Neu (Fig. 7A, lanes  5-7). In contrast, all three forms of Shc (wild type, S154P, and R401L) were phosphorylated by Src, were dephosphorylated by cyt-PTP⑀ in the presence of Src, and bound cyt-PTP⑀ equally well (Fig. 7B). These results agree with the central role of the Shc PTB domain in mediating the phosphotyrosine-based interactions between Shc and Neu and with the phosphotyrosine-independent nature of the association between cyt-PTP⑀ and Shc. These results also demonstrate that in this system, the Shc SH2 domain does not play a significant role in mediating associations of Shc with Neu, Src, or cyt-PTP⑀.
In order to examine whether S154P Shc, which cannot bind Neu, is protected from dephosphorylation by cyt-PTP⑀, we expressed wild type or S154P Shc in HEK293 cells together with Neu, Src, and cyt-PTP⑀ (Fig. 7C). Both WT and S154P Shc were phosphorylated by Src (Fig. 7C, lanes 3 and 4) and were dephosphorylated by cyt-PTP⑀ (Fig. 7C, lanes 5 and 6). Additional expression of Neu protected wild-type Shc from dephosphorylation, in agreement with previous results in this study but did not protect S154P Shc from dephosphorylation (Fig. 7C, lanes 7  and 8). In these studies, S154P Shc retained 20.1 Ϯ 8.6% of the phosphorylation of WT Shc (mean Ϯ S.E., n ϭ 3, p ϭ 0.011 by Student's t test). These results and those obtained with the NYPD Neu mutant lead us to conclude that disruption of the association between Shc and Neu prevents Neu from protecting Shc from dephosphorylation by PTP⑀. Finally, the model by which Shc is protected from dephosphorylation by physical binding to Neu suggests that virtually all phosphorylated Shc found in cells expressing Neu should be associated with and protected by the kinase (and possibly also with other molecules that protect it). In order to verify this, we expressed Neu with or without added cyt-PTP⑀ in cells, exhaustively precipitated over 90% of cellular Neu, and examined the distribution of Shc and phospho-Shc between the precipitated material and remaining supernatant. In agreement with the model, all detectable Tyr(P) 239/240 Shc was found in the precipitated fraction, whereas significant amounts of nonphosphorylated Shc were found in both the precipitated and nonprecipitated fractions. Similar results were obtained in the presence or absence of cyt-PTP⑀ (Fig. 8). We conclude that although not all Shc molecules associate with Neu, phospho-Shc molecules are found only in association with the Neu signaling complex. In all, these results are consistent with tight association between Neu and Shc preventing cyt-PTP⑀ from dephosphorylating Shc and down-regulating signaling events further downstream.

DISCUSSION
Results presented here indicate that cyt-PTP⑀ and RPTP⑀ can down-regulate phosphorylation of Shc at its major phosphorylation sites Tyr 239/240 and Tyr 317 , thus reducing association of Shc with Grb2 and activation of ERK1 and ERK2 further downstream. The ability of purified cyt-PTP⑀ to dephosphorylate purified phospho-Shc in vitro together with presence of both molecules in a stable complex suggests that Shc is a substrate of PTP⑀. PTP⑀ down-regulates activity of ERK1 and ERK2 and other mitogenactivated protein kinases in an indirect manner (28); results presented here suggest therefore that Shc is one of the upstream molecules that mediate the effect of PTP⑀ on ERK.
Shc and cyt-PTP⑀ interact constitutively via the PTB domain of Shc and in a manner supported by the presence of residues 12-22 in cyt-PTP⑀. Of note, most cyt-PTP⑀ molecules are cytosolic, whereas ϳ10% are nuclear and another 20% are membrane-associated. The absence of the first 27 residues of cyt-PTP⑀ abolishes membrane localization (36), whereas residues 1-10 are required for nuclear localization (35). Since Shc is often phosphorylated by or associated with membrane-associated molecules, residues 12-22 of cyt-PTP⑀ may contribute indirectly to the PTP⑀-Shc interaction by promoting membranal localization of the phosphatase. We note, however, that residues 12-22 are present also in RPTP⑀, which is an integral membrane protein and which binds Shc effectively but apparently more weakly. RPTP⑀ contains also unique membranespanning and extracellular domains that are absent from cyt-PTP⑀ and that may mask the contribution of residues 12-22 to binding Shc in RPTP⑀. Our results also suggest that residues 1-4 of cyt-PTP⑀, which are unique to this form of PTP⑀, decrease association with Shc. Further studies are required to fully resolve this issue.
An unexpected finding in this study was the inability of PTP⑀ to dephosphorylate Shc in the presence of Neu, despite being able to do so in the presence of Src or when using purified Shc and PTP⑀. The "protective" effect of Neu in cells is dominant, since the presence of Neu protects Shc from dephosphorylation even in the presence of Src. Further studies strongly suggested that the molecular basis for protection by Neu is its constitutive interaction with Shc, which is mediated by the Shc PTB domain-binding phosphotyrosine residues located in the cyto- FIGURE 7. Disrupting the Shc-Neu association by mutating Shc prevents Neu from protecting Shc from dephosphorylation; an intact PTB domain is required for Shc to bind and be phosphorylated by Neu but not by Src. A, FLAG-tagged WT Shc; S154P Shc (PTB), which carries an inactivating mutation in the PTB domain; R401L Shc (SH2), which carries an inactivating mutation in the SH2 domain; Neu; and cyt-PTP⑀ were expressed in HEK293 cells as indicated. Following precipitation of Shc via its FLAG tag, phosphorylation of Shc and its binding to PTP⑀ and to Neu were examined by blotting. B, similar to A, examining the ability of activated (Y527F) Src to phosphorylate the various Shc mutants and to affect the Shc-PTP⑀ association. C, FLAG-tagged WT or S154P Shc (PTB) was expressed in HEK293 cells together with Src, Neu, and cyt-PTP⑀ as indicated.
Phosphorylation of precipitated Shc and ERK are shown, as are expression of PTP⑀, Src, and Neu. A-C are representative of 3-4 repeats each. IP, immunoprecipitation; WB, Western blot. solic domain of Neu. Indeed, eliminating the phosphotyrosinebinding ability of the Shc PTB domain or mutating the residues in Neu to which this domain can bind disrupted the Shc-Neu association and allowed PTP⑀ to dephosphorylate Shc in the presence of catalytically active Neu. Since Shc binds Neu and PTP⑀ via the same PTB domain, it is likely that the same Shc molecule cannot bind both Neu and PTP⑀ simultaneously. According to this model, Shc molecules that bind Neu are phosphorylated by the kinase and are protected from dephosphorylation by PTP⑀ by the tight association with Neu and by their inability to bind PTP⑀ (Fig. 9, A and B).
Previous studies of Neu-induced mammary tumor cells have established that RPTP⑀ supports the transformed phenotype of these cells by activating Src (23). In particular, Neu phosphorylates RPTP⑀ at its C-terminal tyrosine Tyr 695 ; this phosphorylation event is critical for Src activation, since the nonphosphorylatable Y695F RPTP⑀ cannot activate Src (24). Among other results, the present study identifies Shc as a second molecular target of RPTP⑀ activity. Shc is an important supporter of growth factor-mediated mitogenesis (e.g. see Refs. [51][52][53], and its phosphorylation at residues Tyr 239/240 and Tyr 317 is important in this respect (38,42,54,55). Dephospho-rylation of Shc by PTP⑀ in the mammary tumor system would then contradict the mitogenic signal initiated by Neu and conveyed in part by RPTP⑀-mediated dephosphorylation of Src. However, this potential for conflicting signals is avoided by the activities of Neu, the predominant kinase in the system. On one hand, Neu phosphorylates RPTP⑀ and channels its activity toward Src (24), whereas on the other hand, Neu binds Shc tightly and prevents it from being dephosphorylated by RPTP⑀. Furthermore, the "protective" effect of Neu toward Shc is dominant, ensuring that RPTP⑀ is not driven to dephosphorylate Shc as Src activity increases (Fig. 9C). In all, the present study describes a molecular mechanism by which PTP activity is modulated by directing it toward one substrate while avoiding another. In so doing, the potential for the PTP to initiate signals that are potentially conflicting is reduced, and its contribution to cell signaling is made more coherent.