Roles of Gab1 and SHP2 in Paxillin Tyrosine Dephosphorylation and Src Activation in Response to Epidermal Growth Factor*

Epidermal growth factor (EGF) induces paxillin tyrosine dephosphorylation and Src activation, but the signaling pathways that mediate these responses were largely undefined. We found that Gab1, a docking protein for the SHP2 protein-tyrosine phosphatase in EGF-stimulated cells, was associated with paxillin. SHP2 dephosphorylated paxillin and caused dissociation of Csk, a negative regulator of Src, from paxillin but had no effect on paxillin-Src association. A lower level of Src Tyr-530 phosphorylation was detected in paxillin-associated Src in EGF-stimulated cells. Expression of an SHP2 binding defective mutant of Gab1 (Gab1FF) or a catalytically inactive mutant of SHP2 (SHP2DN) prevented paxillin tyrosine dephosphorylation and Src activation induced by EGF. Importantly, Gab1FF blocked paxillin-SHP2 complex formation, Src Tyr-530 dephosphorylation, Erk activation, and cell migration induced by EGF. Inhibition of Src tyrosine kinase activity abrogated EGF-stimulated Erk activation and cell migration. Together, these results reveal that Gab1 recruits SHP2 to dephosphorylate paxillin, leading to dissociation of Csk from the paxillin-Src complex and Src activation and that Src is an SHP2 effector involved in EGF-stimulated Erk activation and cell migration.

Src tyrosine kinase is basally inactive due to auto-inhibition by intramolecular interactions between its Src-homology-2 (SH2) domain and a phosphotyrosine (Tyr-527 in chicken, Tyr-530 in human) in the C-terminal region and/or between its SH3 domain and a Pro-rich sequence in the linker region (10). Mutational studies have shown that disrupting SH2 domain phosphotyrosine (Tyr(P))-527 interaction is sufficient to activate Src tyrosine kinase activity (11)(12)(13). In fact a mutated human c-Src that loses interaction between the SH2 domain and Tyr(P)-530 has been identified as a human oncogene (14). Phosphorylation of Src Tyr-530 is catalyzed by C-terminal Src kinase (Csk). Csk is a cytosolic protein-tyrosine kinase that is regulated by binding to its docking proteins for co-localization with Src to exert its inhibitory effect. Paxillin and Cbp/PAG are two known Csk-docking proteins (15)(16)(17)(18).
Paxillin is an adapter protein located primarily at sites of cell adhesions to the extracellular matrix in the proximity of plasma membrane (19). Besides its four LD motifs and five LIM domains that interact with actin-associated proteins, integrins, and other proteins, paxillin contains a proline-rich SH3 domain binding sequence and at least four tyrosine phosphorylation sites that could bind SH2 domains (16). Tyrosine-phosphorylated paxillin has been identified as a major binding protein of Csk (15). Interestingly, Src also binds to paxillin through its SH3 domain (16,20). This raises the possibility that paxillin tyrosine phosphorylation may negatively regulate Src activity by bringing Src and Csk to close proximity (16). In addition to Src and Csk, members of focal adhesion kinase (Fak) also interact with paxillin. Interaction between focal adhesion kinase and paxillin appears to involve both LD and phosphotyrosine-based motifs (19).
Several studies have shown that EGF, heregulin, and insulin-like growth factor 1 induce tyrosine dephosphorylation of paxillin in human carcinoma A431, MDA-MB-468, MCF-7, and DU145 cells (21)(22)(23)(24). Insulin-induced paxillin tyrosine dephosphorylation has also been observed in NIH3T3-derived A14 cells that overexpress human insulin receptor (25). The growth factor-induced paxillin dephosphorylation is inhibited by protein-tyrosine phosphatase (PTPase) inhibitors. In MCF-7 and A14 cells, it has been found that expression of a PTPaseinactive SHP2 mutant can block paxillin tyrosine dephosphorylation induced by heregulin, insulin-like growth factor 1, and insulin (23)(24)(25). These data suggest that SHP2 or an SHP2regulated PTPase is responsible for tyrosine dephosphorylation of paxillin in growth factor-stimulated cells. However, it was unclear how the activation signal from a growth factor receptor, such as the EGF receptor, is relayed to SHP2 for paxillin tyrosine dephosphorylation and whether paxillin tyrosine dephosphorylation leads to Src activation.
SHP2 is a cytoplasmic PTPase with narrow substrate specificity (26,27). SHP2 contains two SH2 domains at its Nterminal region and is regulated by binding to its SH2 domain docking proteins, which activate the SHP2 PTPase and tether the activated enzyme to appropriate cellular location (27)(28)(29). An SHP2 docking protein in growth factor-and cytokine-stimulated cells is Gab1 (30 -35). Gab1 is a pleckstrin-homology domain-containing docking protein that becomes tyrosinephosphorylated in cells activated by growth factors and cytokines (36). Gab1-SHP2 association is necessary for EGF-induced Erk mitogen-activated protein kinase activation (37,38). Interestingly, we recently found that expression of a Gab1-SHP2 fusion protein could activate Src and Erk (28). However, it was unclear whether Gab1-SHP2 interaction is involved in EGF-induced paxillin dephosphorylation or Src activation. We provide evidence here that paxillin is a substrate of the SHP2 PTPase and that Gab1 is responsible for recruiting SHP2 to paxillin for tyrosine dephosphorylation and for activation of Src tyrosine kinase in EGF-stimulated cells. Furthermore, our data show that Src kinase activity is involved in EGF-stimulated Erk activation and cell migration.

EXPERIMENTAL PROCEDURES
Primary Antibodies-Polyclonal antibodies to paxillin, SHP2, Csk, Src, EGF receptor, total Erk1/2, and FLAG tag were from Santa Cruz Biotechnology. Mouse monoclonal antibodies to paxillin and phosphotyrosine (RC20H) were from BD Biosciences. Polyclonal antibodies to Gab1 were either from Upstate biotechnology (rabbit) or Santa Cruz Biotechnology (goat). The mouse monoclonal antibody (clone 327) to Src was from Oncogene Sciences. An antibody against phosphorylated Src Tyr-530 was from Cell Signaling. The antibody to active Erk1/2 was from Promega.
DNA Constructs-A single-vector doxycycline (Dox)-inducible expression system, pSTAR (39,40), was provided by Dr. Wanjin Hong. The SHP2 binding defective Gab1 (Gab1FF), in which Tyr-627 and Tyr-659 have been mutated to Phe, has been reported (37). Human SHP2 cDNA (41,42) was subcloned into a pcDNA3.1-FLAG vector (34). SHP2DN that contains double mutations of the active site Cys residue (Cys-459) to Ser and the general acid Asp residue (Asp-425) to Ala of SHP2 were made by sequential PCR using primers that confer these mutations. The sequence of SHP2DN was verified by DNA sequence analysis. Coding sequences for FLAG-tagged Gab1FF and SHP2DN, respectively, were subcloned into the pSTAR vector to generate pSTAR-Gab1FF and pSTAR-SHP2DN. pGEX-GST-Gab1CT was made by cloning the coding sequence for the C-terminal region of human Gab1 (amino acids 592-694) into BamHI and EcoRI sites of pGEX2T.
Cell Culture and Establishment of Dox-inducible Cell Lines-MDA-MB-468 human breast carcinoma cells and A431 human epidermoid carcinoma cells were obtained from American Type Culture Collection and were maintained in Dulbecco's modified Eagle's medium (DMEM) containing 10% fetal calf serum (FCS) at 37°C/5% CO 2 .
To establish Dox-inducible cell lines for expression of Gab1FF and SHP2DN, MDA-MB-468 cells were transfected with pSTAR-Gab1FF, pSTAR-SHP2DN, or the pSTAR vector that had been linearized by PvuI digestion. Transfected cells were cultured in tetracycline free DMEM, 10%FCS (Clontech) in the presence of 0.4 mg/ml G418. G418-resistant cell colonies from empty pSTAR vector transfected cells were pooled (MpSTAR) and used as the control. Individual G418-resistant cell colonies/lines from pSTAR-Gab1FF-or pSTAR-SHP2DN-transfected cells (24 -36 lines/each) were screened for Dox-inducible expression of the FLAG-tagged Gab1FF or SHP2DN by immunoblot analysis of cell lysates.
Immunoprecipitation and Immunoblot Analysis-Near-confluent cells were serum-starved in DMEM, 0.1% BSA for 18 h and then stimulated with EGF or mock-treated as indicated in figure legends. Cells were lysed in Buffer A (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 25 mM NaF, 5 mM sodium pyrophosphate, 1 mM Na 3 VO 4 , 2 g/ml aprotinin, 2 g/ml leupeptin, 100 g/ml phenylmethylsulfonyl fluoride, 1 mM dithiothreitol, 20 mM p-nitrophenyl phosphate, 1% Triton X-100). For experiments involving in Src Tyr-530 phosphorylation analysis, 0.5 M Csk tyrosine kinase inhibitor, PD180970 (43), was also included in the lysis buffer. Precleared supernatants were incubated with specific antibodies indicated in figure legends. Immune complexes were collected with protein G-agarose or protein A-agarose. Immunoprecipitated proteins were separated on 8% SDS-polyacrylamine gels and transferred to nitrocellulose membranes. Immunoblot analysis was performed essentially as described previously (44).
Paxillin Tyrosine Dephosphorylation Assay-A GST fusion protein of the SHP2 PTPase domain (45) was expressed in Escherichia coli BL21 and affinity-purified using glutathione-agarose. The PTPase activity was measured by hydrolysis of p-nitrophenyl phosphate in 80-l reaction mixtures containing 50 mM Hepes, pH 7.0, 20 mM NaCl, 1 mM dithiothreitol, 10 mM p-nitrophenyl phosphate, and various amounts of the enzyme at 30°C for 10 min. The reaction was stopped by the addition of 80 l of 1 M NaOH. The absorbance of the mixtures at 405 nm (A 405 ) was determined in 96-well plates. One unit of PTPase was defined as the amount of enzyme that produces 1 A 405 unit in this PTPase assay. For the paxillin dephosphorylation assay, paxillin was immunoprecipitated from confluent MDA-MB-468 cells that had been serumstarved in DMEM, 0.1% BSA for 18 h. The paxillin immunoprecipitates were incubated with different amounts of SHP2-PTPase in 50 mM Hepes, pH 7.0, 20 mM NaCl, and 1 mM dithiothreitol in an 80-l reaction volume at 30°C for 20 min. After the reaction, paxillin immunoprecipitates were washed with 1 ml of Buffer A 3 times and resolved by SDS-polyacrylamide gel electrophoresis. Paxillin tyrosine dephosphorylation was then examined by immunoblotting with an anti-phosphotyrosine antibody (RC20H).
Src Kinase Assay-Cells were lysed in Buffer B (25 mM Tris-HCl, pH 7.5, 5 mM EDTA, 5% glycerol, 1% deoxycholate, 0.1% SDS, 1% Triton X-100, 150 mM NaCl, 2 g/ml leupeptin, 2 g/ml aprotinin, 1 mM phenylmethylsulfonyl fluoride, 0.5 mM Na 3 VO 4 ). Src was immunoprecipitated from cell lysate supernatants using anti-Src monoclonal antibody 327. Src immune complex kinase assay was performed at 30°C for 15 min in a 30-l reaction mixture containing 10 mM Tris-HCl, pH 7.4, 10 mM MgCl 2 , 1 mM dithiothreitol, 5 g of GST-Gab1CT, and 10 M [␥-32 P]ATP (5,000 cpm/pmol). The reaction was terminated by the addition of 4ϫ SDS-gel loading buffer and heat denaturation. After resolving on a 10% SDS-polyacrylamide gel, proteins were transferred to a nitrocellulose membrane and processed for autoradiography. After the decay of radioactivity, the membrane was then used for analysis of Src protein presented on the membrane by immunoblotting.
Cell Migration Assay-Transwell cell migration assay was performed as described (46,47). Transwell cell culture insert polycarbonate membrane was coated overnight with rat tail type I collagen (10 g/ml) at 4°C and then air-dried. MDA-MB-468 cells were cultured with or without 2 g/ml Dox for 48 h and serum-starved for 18 h. Cells were detached from plates by trypsin digestion, and cell number was determined. DMEM, 10% FCS (0.6 ml) was placed into the lower chamber in 24-well plates. Cells (5 ϫ 10 4 ) in 0.2 ml DMEM, 0.1% BSA were placed in the upper chamber in triplicate with or without 5 ng/ml EGF in the presence or absence of PP1 (10 M), PD180970 (0.5 M), or U0126 (10 M). After incubation at 37°C, 5% CO 2 for 8 h, cells on the upper member surface were mechanically removed with a cotton swab. Cells that had migrated to the lower side of membrane were fixed and stained using the HEMA3 reagents (Fisher) and enumerated under a microscope in three randomly chosen 1.2 ϫ 0.8-mm fields.

EGF Stimulates Tyrosine Dephosphorylation of Paxillin-To
verify that EGF induces paxillin tyrosine dephosphorylation as reported by others (21), we immunoprecipitated paxillin from MDA-MB-468 breast carcinoma cells and A431 epidermoid carcinoma cells at various times after EGF stimulation. Paxillin tyrosine phosphorylation was then examined by immunoblotting with an anti-phosphotyrosine antibody. Fig. 1 shows that high levels of paxillin tyrosine phosphorylation were detected in serum-starved MDA-MB-468 and A431 cells. EGF-induced paxillin tyrosine dephosphorylation in MDA-MB-468 cells was detectable within 2.5 min. The maximal dephosphorylation (90%) occurred at 10 min after EGF stimulation (Fig. 1A). A lesser (40%) and more transient, but consistent, decrease in paxillin tyrosine phosphorylation was detected in A431 cells after EGF treatment (Fig. 1B). Control immunoblots (Fig 1, A and B, bottom panels) show that equal amounts of paxillin were present in the samples.
Tyrosine-phosphorylated paxillin is a major binding protein of Csk (15). To determine whether the change in paxillin tyrosine phosphorylation in EGF-stimulated cells affects paxillin-Csk association, we analyzed co-immunoprecipitation of Csk with paxillin from MDA-MB-468 cells before and after EGF treatment. As shown in Fig. 1C, a high level of paxillin-Csk complex was observed in serum-starved cells. EGF stimulation decreased the paxillin-Csk complex (Fig. 1C). These data confirm that EGF induces paxillin tyrosine dephosphorylation in these carcinoma cells and raises the question about how the EGF receptor activation signal is transmitted to paxillin.
Gab1 Binds Paxillin, and SHP2 Is Recruited to Gab1 and Paxillin after EGF Stimulation-During our study of Gab1interacting proteins we identified paxillin as a novel Gab1binding protein in various cell lines (data not shown). As illustrated in Fig. 2A (also in Fig. 4C), the Gab1-paxillin complex was observed in resting MDA-MB-468 cells. EGF stimulation further increased the amount of paxillin detected in Gab1 immunoprecipitates.
A minimal residual level of phosphotyrosine was detected in Gab1 in serum-starved cells. As reported previously (37), EGF markedly stimulated Gab1 tyrosine phosphorylation and the formation of the Gab1-SHP2 complex ( Fig. 2A). This correlates with the recruitment of SHP2 to paxillin after EGF stimulation (Fig.  2B). These results raise the possibility that Gab1 may recruit SHP2 to dephosphorylate paxillin in EGF-stimulated cells.
SHP2 Dephosphorylates Paxillin in Vitro-To evaluate whether SHP2 could be responsible for paxillin tyrosine dephosphorylation in EGF-stimulated cells, we tested whether SHP2 can dephosphorylate paxillin. Paxillin was immunoprecipitated from serum-starved MDA-MB-468 cells and incubated with various amounts of a GST fusion protein of the SHP2 catalytic domain (SHP2-PTPase). Fig. 3A shows that incubation of paxillin with the GST-SHP2 fusion protein resulted in tyrosine dephosphorylation of paxillin in a PTPase concentration-dependent manner. The GST control protein had no apparent effect on paxillin tyrosine dephosphorylation in vitro (Fig. 3A, inset). This result demonstrates that paxillin is a substrate of the SHP2 PTPase.
We next examined whether paxillin tyrosine dephosphorylation by SHP2 in vitro resulted in loss of paxillin-associated Csk and Src. Paxillin immune complex from serum-starved MDA-MB-468 cells was incubated with SHP2-PTPase or with GST as a control. After incubation, the immune complex was washed to remove proteins that had been dissociated from the immune complex. The immune complex was then analyzed by immunoblotting for Csk, c-Src, or paxillin (control). As shown in Fig. 3B, incubation of paxillin with SHP2-PTPase in vitro resulted in dissociation of Csk from paxillin, suggesting that binding of Csk to paxillin is dependent on tyrosine phosphorylation. In contrast, the SHP2-PTPase treatment did not affect Src binding to paxillin. This result demonstrates that Src can bind paxillin independent of paxillin tyrosine phosphorylation. Thus, paxillin tyrosine dephosphorylation by SHP2 could lead to separation of Csk, the negative regulator of Src tyrosine kinase, from the paxillin-Src complex.
Gab1-SHP2 Interaction Is Necessary for Recruitment of SHP2 to Paxillin and for Tyrosine Dephosphorylation of Paxillin in Response to EGF-Data presented in Fig. 2 suggest but do not prove that formation of the paxillin-SHP2 complex in EGF-stimulated cells is mediated by Gab1. We previously showed that a Gab1 mutant (Gab1FF), in which Tyr-627 and Tyr-659 have been mutated to Phe, is defective in SHP2 binding (37). To determine whether Gab1-SHP2 interaction is necessary for the formation of paxillin-SHP2 complex and tyrosine dephosphorylation of paxillin in EGF-stimulated cells, we used the pSTAR vector (39,40)  (MG12 and MG24) show Dox-inducible expression of Gab1FF (Fig. 4A). MG12 and MG24 cells were cultured in medium with or without Dox to control Gab1FF expression. The empty pSTAR vector-transfected cells (a pool of G418-resistant colonies, MpSTAR) were used a control. These cells were then serum-starved, stimulated with EGF or mock-treated, and analyzed for formation of the paxillin-SHP2 complex, tyrosine dephosphorylation of paxillin, and changes in the paxillin-Csk and paxillin-Src complexes.
Analysis of paxillin immunoprecipitates by immunoblotting with anti-SHP2 antibody showed that EGF induced paxillin-SHP2 complex formation in the MpSTAR control cells regardless if Dox was present (Fig. 4B, top panel). EGF induced paxillin-SHP2 complex formation in MG12 and MG24 cells in the absence of Dox. However, formation of the EGF-induced paxillin-SHP2 complex was blocked when MG12 and MG24 cells were cultured in the presence of Dox (Fig. 4B, top panel). Gab1FF was not detected in MpSTAR cells or in MG12 and MG24 cells cultured in the absence of Dox (Fig. 4C). When MG12 and MG24 cells were cultured with Dox, Gab1FF was expressed in these cells and bound to paxillin (Fig. 4C). Importantly, Gab1FF did not bind SHP2 (Fig. 4C, middle panel).
A high level of paxillin tyrosine phosphorylation was evidenced in serum-starved MG12, MG24, and MpSTAR cells (Fig. 4B, middle panel). Induction of Gab1FF expression by Dox in MG12 and MG24 cells did not affect paxillin tyrosine phosphorylation under serum-starved conditions, in which there was little paxillin-SHP2 complex. When MG12 and MG24 cells were cultured in tetracycline/Dox-free medium, EGF-induced paxillin tyrosine dephosphorylation in these cells similar to that observed in the MpSTAR control cells (Fig. 4B) or in the parental cells (Fig. 1A). However, EGF-induced paxillin dephosphorylation was abrogated when MG12 and MG24 cells were cultured in the presence of Dox (Fig. 4B, middle panel). Consistent with the in vitro dephosphorylation data, EGFinduced paxillin tyrosine dephosphorylation in MpSTAR, MG12, and MG24 cells resulted in loss of Csk from the paxillin immunoprecipitates (Fig. 4B). Expression of Gab1FF blocked EGF-induced dissociation of the paxillin-Csk complex. In contrast, EGF stimulation in these cells had no effect on the paxillin-Src complex (Fig. 4B). Together, these data demonstrate that Gab1-SHP2 interaction is necessary for recruitment of SHP2 to paxillin and for paxillin tyrosine dephosphorylation in response to EGF. Furthermore, paxillin tyrosine dephosphorylation in response to EGF stimulation results in dissociation of Csk, but not Src, from paxillin.
Requirement of SHP2 PTPase Activity for Paxillin Tyrosine Dephosphorylation in EGF-stimulated Cells-To determine whether the SHP2 PTPase activity is necessary for EGF-induced paxillin tyrosine dephosphorylation, we prepared two MDA-MB-468-derived cell lines (MS3 and MS65) that show Dox-inducible expression of a PTPase-inactive SHP2 mutant (SHP2DN) (Fig. 5A). MS3 and MS65 cells were cultured either with or without Dox to control SHP2DN expression, and EGFinduced paxillin tyrosine dephosphorylation was analyzed. As shown in Fig. 5B, EGF induced paxillin tyrosine dephosphorylation in MpSTAR cells and in MS3 and MS65 cells cultured under Dox-free conditions. Induction of SHP2DN with Dox in MS3 and MS65 cells inhibited EGF-induced tyrosine dephosphorylation of paxillin. Thus, the SHP2 PTPase activity is required for EGF-induced paxillin tyrosine dephosphorylation in MDA-MB-468 cells.
EGF-induced Src Activation Requires SHP2 PTPase Activity and Gab1-SHP2 Interaction-We found previously that expression of a Gab1-SHP2 fusion protein could cause Src activation through an unknown mechanism (28). Our observation above that Gab1-SHP2 interaction mediates EGF-induced paxillin tyrosine dephosphorylation and that paxillin tyrosine dephosphorylation leads to dissociation of Csk from paxillin, whereas Src remains associated with paxillin, suggests that paxillin tyrosine dephosphorylation may regulate Src activation. To test this possibility, we utilized the MDA-MB-468-derived cells (MG12, MG24, MS3, MS65) in which EGF-stimulated paxillin tyrosine dephosphorylation can be regulated by Dox-inducible expression of Gab1FF or SHP2DN. Fig. 6 shows that EGF induced Src activation in the MpSTAR control cells, which was not affected by Dox in the culture medium. EGF activated Src in MG12, MG24, MS3, and MS65 cells when these cells were cultured in Dox-free medium. Induction of Gab1FF or SHP2DN expression by Dox in these cells effectively blocked Src activation in EGF-stimulated cells (Fig. 6). These data support the notion that paxillin tyrosine dephosphorylation is a mechanism for Src activation induced by EGF.
EGF Reduces Phosphorylation of Tyr-530 in Paxillin-associated Src-Csk-catalyzed phosphorylation of Tyr-530 on human Src is known to negatively regulate Src activity. To determine whether EGF reduces Src Tyr-530 phosphorylation, we immunoprecipitated Src or paxillin from MDA-MB-468 cells treated with or without EGF and analyzed these two populations of Src for Tyr-530 phosphorylation by immunoblotting with an antibody specifically to phosphorylated Src Tyr-530. An EGF-induced change in Src Tyr-530 phosphorylation was not detectable in Src immunoprecipitates (Fig. 7B), indicating that EGF either did not induce Tyr-530 dephosphorylation or that the change was too small to be detected in the total Src population. Interestingly, EGF-induced Src Tyr-530 dephosphorylation  was detected in paxillin immunoprecipitates (Fig. 7A). Consistent with previous results, a high level of SHP2 was detected in the paxillin immunoprecipitates from EGF-stimulated cells when the same filter was reprobed with an anti-SHP2 antibody. These results indicate that the paxillin-bound Src represents a major fraction of Src where Tyr-530 dephosphorylation occurs in EGF-stimulated cells.
To determine whether Gab1-SHP2 interaction is required for EGF-induced Src Tyr-530 dephosphorylation, we analyzed Src Tyr-530 dephosphorylation in MpSTAR, MG12, and MG24 cells. These cells were cultured with or without Dox, serumstarved, and treated with EGF. After immunoprecipitation of paxillin, Src Tyr-530 phosphorylation, SHP2, and Src protein in the paxillin immunoprecipitates were analyzed. As shown in Fig. 7C, EGF reduced the level of Src Tyr-530 phosphorylation in MpSTAR, MG12, and MG24 cells in the absence of Dox. This correlated with an increase in paxillin-associated SHP2, whereas there was no change in paxillin-associated Src protein.
Dox-induced expression of Gab1FF in MG12 and MG24 cells blocked recruitment of SHP2 to paxillin and dephosphorylation of Src Tyr-530 in response to EGF in these cells, whereas the control experiment showed no effect of Dox in MpSTAR cells. These results demonstrate that Gab1-SHP2 interaction is necessary for the reduction of Tyr-530 phosphorylation in paxillinassociated Src in EGF-stimulated cells.
Src Kinase Is Involved in EGF-stimulated Erk Activation-It is well documented that Gab1-SHP2 interaction plays a positive role in Erk activation by growth factors, but the signaling pathway linking Gab1 and SHP2 to Erk activation is largely undefined (for review, see Ref. 36). Data presented in Fig. 6 show that Gab1-SHP2 interaction mediates EGF-induced Src activation. Interestingly, it was observed previously that Src kinase is essential for the rapid phase of Erk activation by EGF and neuregulin in T47D and SKBR3 cells (3). These raise the possibility that Src is a Gab1-SHP2 effector that mediates Erk activation by EGF.
To assess if Src kinase is involved in EGF-induced Erk activation in MDA-MB-468 cells, we analyzed the effect of an Src tyrosine kinase inhibitor, PP1, on EGF-induced Erk activation. Fig. 8A shows that PP1 effectively inhibited EGF-induced Erk activation. Similar results were obtained in SKBR3 and U87MG cells (data not shown). Analysis of EGF receptor immunoprecipitates by immunoblotting with an anti-phosphotyrosine antibody indicated that PP1 had little effect on tyrosine phosphorylation of EGF receptor and other proteins co-immunoprecipitated with the EGF receptor (Fig. 8B). Thus, the inhibitory effect of PP1 is unlikely due to nonspecific inhibition of the EGF receptor tyrosine kinase. To further assess the involvement of Src tyrosine kinase in EGF-induced Erk activation, we examined the effects of another Src tyrosine kinase inhibitor, PD180970 (43, 48), on EGF-induced Erk activation and EGF receptor tyrosine phosphorylation. As illustrated in Fig. 8C, EGF-induced Erk activation was inhibited by the Src Equal amounts (30 g) of cell lysate supernatants were analyzed by immunoblotting with antibodies to active Erk1/2 or to total Erk1/2 (A and C). EGF receptor was immunoprecipitated from cell lysate supernatants. A portion of each immunoprecipitate (IP) was analyzed by immunoblotting (IB) with an anti-phosphotyrosine (␣-pY) antibody, whereas the rest was probed with an antibody to EGF receptor (B and D). The filters were either exposed for shorter times to give more a accurate comparison of EGF receptor tyrosine phosphorylation or for longer time to visualize EGF receptor-associated tyrosine-phosphorylated proteins. E, MpSTAR, MG12, and MG24 cells were cultured with or without Dox, serumstarved, and stimulated with EGF as indicated. Equal amounts of cell lysate supernatants (30 g) were analyzed by immunoblotting with antibodies to active Erk1/2 or to total Erk1/2. kinase inhibitor PD180970. PD180970 had no effect on EGFstimulated tyrosine phosphorylation of EGF receptor and EGF receptor-associated proteins (Fig. 8D). These data demonstrate the involvement of Src in EGF-induced Erk activation.
To confirm that Gab1-SHP2 interactions are involved in Erk activation in MDA-MB-468 cells, MpSTAR, MG12, and MG24 cells were grown in the presence or absence of Dox, serumstarved, and stimulated with EGF, and Erk activation was determined. Fig. 8E shows that Dox did not affect EGF-induced Erk activation in MpSTAR cells. As predicted, induction of Gab1FF expression in MG12 and MG24 cells with Dox resulted in inhibition of EGF-induced Erk activation (Fig. 8E).
Both Gab1-SHP2 Interaction and Src Kinase Are Involved in EGF-stimulated Cell Motility-SHP2 has a positive function in growth factor-induced cell migration, and this appears to be related to paxillin tyrosine dephosphorylation (23,49,50). We now show that Gab1 recruits SHP2 to dephosphorylate paxillin and that paxillin tyrosine dephosphorylation leads to Src activation. Therefore, we next determined if Gab1-SHP2 interaction and Src kinase activity are involved in EGF-induced cell motility.
To assess the role of Gab1-SHP2 interaction in cell migration, MpSTAR, MG12, and MG24 cells were cultured with or without Dox to control Gab1FF expression. Cell motility was analyzed using the Transwell cell migration assay. As shown in Fig. 9A, expression of Gab1FF in MG12 and MG24 cells reduced the basal cell migration activity. The effect of Gab1FF on the basal cell migration activity is probably due to 10% FCS included in the lower chamber as chemoattractant, which may stimulate Gab1-SHP2. In the presence of EGF there was an ϳ2-fold increase in cell migration activity. Although Dox had no effect on the EGF-stimulated cell migration activity in Mp-STAR cells, induction of Gab1FF in MG12 and MG24 cells with Dox blocked the EGF-stimulated cell migration activity (Fig. 9A).
We next determined the effects of Src tyrosine kinase inhibitors PP1 and PD180970 on EGF-stimulated MDA-MB-468 cell migration activity. Again, EGF increased MDA-MB-468 cell migration activity about 2-fold (Fig. 9B). The EGF-stimulated cell migration activity was completely inhibited by PP1 and PD180970 (Fig. 9B). Thus, Src tyrosine kinase activity is required for MDA-MB-468 cell migration in response to EGF. Similar to previous reports (46,47), the Mek1 kinase inhibitor U0126 also blocked EGF-stimulated MDA-MB-468 cell migration.

DISCUSSION
Although functional and signaling interactions between EGF receptor and Src tyrosine kinase have been documented, less is known about how EGF induces Src activation. We demonstrate here that EGF induces Gab1-dependent paxillin tyrosine dephosphorylation by SHP2, which dissociates Csk from the paxillin-Src complex and leads to Src Tyr-530 dephosphorylation and Src tyrosine kinase activation (Fig. 10).
Previous evidence suggests that SHP2 is involved in growth factor-induced paxillin tyrosine dephosphorylation (23,24). However, it was unclear if paxillin was dephosphorylated directly by SHP2. We demonstrate here that SHP2 can dephosphorylate paxillin in vitro. Consistent with previous studies in other cell lines, we show here that expression of a PTPaseinactive SHP2 inhibits EGF-induced paxillin tyrosine dephosphorylation in MDA-MB-468 cells. These data indicate that SHP2 is the PTPase responsible for dephosphorylation of paxillin in EGF-stimulated cells.
SHP2 is a cytoplasmic PTPase. Interaction between SHP2 and its docking proteins is absolutely required for the cellular function of SHP2. We and others report that binding of SHP2 to

FIG. 10. A mechanism for EGF-induced Src activation.
In resting cells Csk binds to paxillin through Tyr(P)-dependent interaction, whereas Src and Gab1 are constitutively associated with paxillin. Colocalization of Csk and Src to paxillin allows Csk to effectively phosphorylate Src Tyr-530 and inhibit Src tyrosine kinase activity. When EGF receptor (EGFR) is activated, the EGF receptor tyrosine kinase phosphorylates Gab1 at Tyr-527 and Tyr-659, which bind and activates SHP2. The Gab1-bound SHP2 dephosphorylates paxillin, resulting in release of Csk from the paxillin-Src complex and loss of its effectiveness for phosphorylation of Src Tyr-530. Consequently, the Src Tyr-530 phosphorylation level decreases, and Src tyrosine kinase is activated. pY, phosphotyrosine.
Gab1 mediates Erk mitogen-activated protein kinase activation by growth factors and cytokines through an unknown mechanism (33)(34)(35)(36)(37)(38). Here we show that Gab1 binds to paxillin and recruits SHP2 to dephosphorylate paxillin in response to EGF stimulation. Consistently, a previous study found that the SHP2 SH2 domains do not bind phosphopeptides derived from paxillin (51). Therefore, we have uncovered the signaling pathway linking EGF receptor activation to paxillin tyrosine dephosphorylation.
We previously found that expression of a Gab1-SHP2 fusion protein could activate Src and caused Src-dependent Erk activation in transfected cells (28). This observation suggests that Src could be a downstream effector of SHP2 and that Src could be one of the long sought-after SHP2 signal outputs that mediate Erk activation. However, three issues remained to be addressed. First, how does SHP2 activate Src? Second, does Gab1-SHP2 interaction plays a role in EGF-induced Src activation? Third, does Src activation by SHP2 play a role in Erk activation by EGF?
A potential model for Src activation by SHP2 is that SHP2 may directly dephosphorylate the negative regulatory Tyr(P)-530 residue on Src. In fact, a phosphopeptide containing Tyr(P)-530 was used by us as an SHP2 substrate (37). However, we have been unable to detect tyrosine dephosphorylation of Src protein or activation of Src tyrosine kinase activity by SHP2 PTPase in vitro in many attempts. Therefore, it appears unlikely that SHP2 activates Src by directly dephosphorylating Tyr(P)-530.
Another possible model for SHP2-mediated Src activation is by separating the Src-inhibiting kinase, Csk, from Src. Csk is a cytoplasmic tyrosine kinase that requires co-localization with Src to exert its inhibitory effect. A major Csk binding protein is paxillin (15,16). Interestingly, Src also binds to paxillin. Thus, simultaneous binding of Src and Csk to paxillin could constitute a mechanism for inhibition of Src tyrosine kinase (16). Importantly, we found that tyrosine dephosphorylation of paxillin by SHP2 resulted in dissociation of Csk, but not Src, from paxillin. It was observed previously that the Csk SH2 domain was essential for binding of a GST fusion protein of the Csk SH3-SH2 domains to paxillin, whereas a GST-Csk SH3 domain fusion protein did not bind paxillin in vitro (15,52). Furthermore, a GST-Csk SH3-SH2 domain fusion protein bound to paxillin in chicken embryo cell lysates only when these cells were treated with vanadate, suggesting that the interaction is dependent upon tyrosine phosphorylation (52). In contrast, although both of the Src SH2 and SH3 domains could interact with paxillin in vitro, the Src SH3 domain alone was sufficient to bind strongly to paxillin (20,52). Furthermore, Src was not able to bind synthetic phosphopeptides derived from known paxillin phosphorylation sites in vitro (51). Thus, our data are consistent with these previous observations. Significantly, inhibition of paxillin tyrosine dephosphorylation by expression of Gab1FF and SHP2DN in Dox-inducible cell lines prevents EGF-induced Src activation. These results support the model in which SHP2 activates Src indirectly through dephosphorylation of paxillin (Fig. 10).
Another known Csk-binding protein is Cbp/PAG (17,18). Studies on Cbp/PAG to date were mostly done in lymphocytes or by expression of exogenous Cbp in fibroblasts. Although it is possible that Cbp-Csk interaction may contributes to regulation of Src activity, several observations suggest that paxillin-Csk interaction plays a major role in EGF-induced Src activation in cells used in this study. First, reduced phosphorylation of the Csk substrate site (Tyr-530) on Src was only detectable in paxillin-associated Src but not in the total Src immunoprecipitates. These data suggest that paxillin-associated Src represents a major fraction of EGF-activated Src. Second, Cbp ap-pears to express at low levels in MDA-MB-468 cells and A431 cells. Third, our data show that Gab1 binds paxillin and EGF induces SHP2-paxillin complex formation. In contrast, we have not been able to detect Gab1-Cbp or SHP2-Cbp complex either in serum-starved or in EGF-stimulated cells (data not shown).
Although we showed previously that Gab1-SHP2 interaction could activate Src (28), it was unclear if Gab1-SHP2 interaction plays a role in EGF-induced Src activation pathway. Using the Dox-inducible cell lines for expression of SHP2-binding defective Gab1 and catalytically inactive SHP2, we demonstrated here that both Gab1-SHP2 interaction and SHP2 PTPase activity are required for EGF-induced Src activation in MDA-MB-468 cells. Importantly, we further demonstrate that Gab1-SHP2-dependent EGF-induced Src activation correlates with Tyr-530 dephosphorylation in paxillin-bound Src. These results not only reveal a novel signal transduction pathway linking EGF receptor to Src activation but also provide clear evidence that EGF activates Src activity through regulation of Tyr-530 phosphorylation.
Although EGF-induced Src Tyr-530 dephosphorylation was observed in paxillin immunoprecipitates, such a change was not detectable in Src immunoprecipitates. These data indicate that paxillin-associated Src is the major fraction of Src activated by EGF through Tyr-530 dephosphorylation. Because only a fraction of cellular Src is associated with paxillin, one would predict from these data that the extent of Src activation by EGF measured in Src immunoprecipitates should be much less than that under conditions where most Src in the cells contain unphosphorylated Tyr-530. Indeed, several previous studies show that the Src Tyr-527 mutant has 10-times higher activity than the wild type Src (11)(12)(13). However, EGF-induced Src activation as determined by Src immune complex kinase assay is difficult to detect in most cell lines, although other evidence suggests that Src is involved in EGF signaling in these cells. In those cases where EGF-induced Src kinase activation was observed, the maximal activation measured in Src immunoprecipitates was about 1.5-3-fold (1, 4, 7-9). These observations are consistent with the notion that only a fraction of cellular Src is activated by EGF. Interestingly, after submission of our original manuscript it was reported that paxillin could serve as a scaffold for the Raf-Mek-Erk kinase cascade in mouse epithelial cells (53). Because our data show that Src is involved in regulation of Erk activity, activation of the specific, paxillin-bound Src by EGF may have particular function relevance.
Gab1-SHP2 interaction mediates a branch of signaling pathway in Erk activation by EGF, but the mechanism is largely unclear (36). Several lines of evidence now point to Src as the SHP2 downstream effector that mediates Erk activation by EGF. First, expression of a Gab1-SHP2 fusion protein activated Src and caused Src-dependent Erk activation, indicating Gab1-SHP2 could activate Erk through Src (28). Second, inhibition of Src activity abrogated EGF-induced Erk activation. Third, dominant negative Gab1FF and SHP2DN mutants inhibited both Src and Erk activation by EGF. All of these data are consistent with the notion that Src is a downstream effector of SHP2 that mediates Erk activation by EGF.
EGF stimulates cell motility (21,46). In fact, ErbB overexpression correlates with aggressive tumor behavior and metastasis (54,55). Interestingly, previous studies show that SHP2 plays a positive role in cell migration (49,50). We found here that SHP2 mediates Src activation by EGF and that paxillinassociated Src represents the major fraction of EGF-activated Src. Paxillin is known to interact with integrins and cytoskeleton and serve as a convergent point for control of cell motility (16,19). Inhibition of Src activation by Gab1FF or by Src tyrosine kinase inhibitors blocked cell migration stimulated by EGF. These data demonstrate the involvement of Src in EGFstimulated cell migration activity. Consistent with previous findings (46,47), we show here that Erk activity is also required for EGF-stimulated cell migration activity. Therefore, although Src may mediate EGF-stimulated cell motility through multiple mechanisms, one of these is regulation of Erk activation.