Phosphotyrosines 627 and 659 of Gab1 Constitute a Bisphosphoryl Tyrosine-based Activation Motif (BTAM) Conferring Binding and Activation of SHP2*

A major Grb2-associated binder-1 (Gab1) binding partner in epidermal growth factor (EGF)-stimulated cells is protein-tyrosine phosphatase (PTPase) SHP2, which contains tandem SH2 domains. The SHP2 PTPase activity is required for activation of the extracellular signal-regulated kinase (ERK) subfamily of mitogen-activated protein (MAP) kinase by EGF. To investigate the mechanism by which Gab1 and SHP2 mediate ERK activation, we characterized the Gab1-SHP2 interaction. We found that both Tyr-627 and Tyr-659 of Gab1 were required for SHP2 binding to Gab1 and for ERK2 activation by EGF. Far Western blot analysis suggested that the tandem SH2 domains of SHP2 bind to Gab1 in a specific orientation, in which the N-SH2 domain binds to phosphotyrosine (Tyr(P))-627 and the C-SH2 domain binds to Tyr(P)-659. When assayed with peptide substrates, SHP2 PTPase was activated by a bisphosphopeptide containing both Tyr(P)-627 and Tyr(P)-659, but not by monophosphopeptides containing Tyr(P)-627 or Tyr(P)-659 or a mixture of these monophosphopeptides. These results suggest that Tyr(P)-627 and Tyr(P)-659 of Gab1 constitute a bisphosphoryl tyrosine-based activation motif (BTAM) that binds and activates SHP2. Remarkably, while a constitutively active SHP2 (SHP2ΔN) could not rescue the defect of a SHP2-binding defective Gab1 (Gab1FF) in ERK2 activation, expression of a Gab1FF-SHP2ΔN chimera resulted in constitutive activation of ERK2 in transfected cells. Thus, physical association of activated SHP2 with Gab1 is necessary and sufficient to mediate the ERK mitogen-activated protein kinase activation. Phosphopeptides derived from Gab1 were dephosphorylated by active SHP2 in vitro. Consistently, substrate-trapping experiments with a SHP2 catalytic inactive mutant suggested that Gab1 was a SHP2 PTPase substrate in the cells. Therefore, Gab1 not only is a SHP2 activator but also is a target of its PTPase.

SHP2 is a protein-tyrosine phosphatase (PTPase) 1 with two Src homology-2 (SH2) domains (1,2). These two SH2 domains, termed N-SH2 and C-SH2 domains, are arranged in tandem at the amino (N)-terminal portion of the protein. SHP2 has a low basal PTPase activity that can be activated by deletion of N-SH2 or both SH2 domains or by specific phosphopeptides that bind to the SH2 domains. The crystal structure of SHP2 shows that the N-SH2 domain is inserted into the catalytic cleft of the phosphatase domain in the absence of a ligand for the N-SH2 domain, thus maintaining the phosphatase in an autoinhibitory state (3). Binding of a specific tyrosine-based activation motif (TAM) to the N-SH2 domain is predicted to induce an allosteric change that disrupts the inhibitory interaction between the N-SH2 and the catalytic domains, leading to phosphatase activation (3). Although there is no direct surface contact between the C-SH2 and the catalytic domains, the C-SH2 domain connects the N-SH2 domain to the PTPase domain and contributes to the selectivity and high affinity binding of the tandem SH2 domains to bisphosphoryl TAMs (4 -6).
It has been demonstrated that the SHP2 PTPase activity is required for activation of the extracellular signal-regulated protein kinase (ERK) subfamily of mitogen-activated protein (MAP) kinase by epidermal growth factor (EGF) (7,8). However, the mechanism by which SHP2 exerts its positive role in ERK MAP kinase activation by EGF is not yet known. The autoinhibitory configuration of the unbound SHP2 necessitates a binding partner for activation of the SHP2 PTPase. Although insulin receptor substrate-1 (IRS-1) and SHP substrate-1 (SHPS-1, also called BIT, SIRP␣) can bind and activate SHP2 (5,6), IRS-1 is not tyrosine phosphorylated in response to EGF (9). Moreover, the interaction between IRS-1 and SHP2 is not required for ERK activation by insulin (10). In fact, interaction between IRS-1 and SHP2 appears to have negative effects on insulin signaling (10). Similarly, it has been reported that SHPS-1 and its interaction with SHP2 play a negative role in the EGF-induced ERK activation pathway (11).
We previously found that Gab1 with a single mutation at Tyr-627 lost SHP2 binding activity and inhibited EGF-stimulated ERK2 (p42 MAPK ) activation (15). However, the mechanism by which the Gab1-SHP2 interaction mediates ERK activation has not been elucidated. In the present study, we provide evidence that two tyrosine residues (Tyr-627 and Tyr-659) in the carboxyl (C)-terminal region of Gab1 are both required for SHP2 binding to Gab1 and for EGF-stimulated ERK activation. These two tyrosine residues constitute a bisphosphoryl tyrosine-based activation motif (BTAM) that mediate binding and activation of SHP2 by Gab1. Co-expression of a constitutive active SHP2 (SHP2⌬N) did not rescue the defect of a SHP2 binding-defective Gab1 (Gab1FF), whereas expression of a Gab1FF-SHP2⌬N fusion protein resulted in constitutive activation of ERK2 in COS-7 cells. These data indicate that the activated SHP2 has to be physically associated with Gab1 to mediate EGF-stimulated ERK2 activation and identify Gab1 as the SHP2 activator for the ERK MAP kinase pathway in EGFstimulated cells. Furthermore, our in vitro dephosphorylation experiments and substrate-trapping study in cultured cells suggest that Gab1 is a substrate of the SHP2 PTPase.
Preparation of the Gab1 Mutants-Expression vectors for FLAGtagged Gab1 (pGab1) and pGab1Y627F have been described (15). Expression plasmids for Gab1 with a Tyr-659 to Phe mutation (Gab1Y659F) and Gab1 with double mutations of both Tyr-627 and Tyr-659 to Phe (Gab1FF) were prepared with the GeneEditor in vitro site-directed mutagenesis system (Promega) using pGab1 and pGab1Y627F, respectively, as templates. The mutagenic primer was 5Ј-pGATGAGAGAGTGGAATTTGTTGTTGTTGAC (the changed nucleotides are underlined). Mutations were confirmed by DNA sequencing.
To construct a plasmid for the expression of a Gab1FF-SHP2⌬N chimeric protein, a cDNA fragment encoding for amino acids 108 -597 of SHP2 (23) was amplified by polymerase chain reaction with Pfu DNA polymerase. The polymerase chain reaction primers used were: 5Ј-TGTCTAGAGGAGGCGGTACCTCTGAAAGGTGGTTCC (sense, un-derline represents DNA sequence encoding for a three-residue Gly linker), and 5Ј-CATCTAGAACTCCTCTGCTGCTGCATGAG (antisense). The polymerase chain reaction fragment was inserted into a XbaI site between the last codon of Gab1FF and the FLAG tag coding sequence in pGab1FF.
Cell Culture and Transfection-COS-7 cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum. Transfection was performed on cells grown in 60-mm plates using LipofectAMINE (Life Technologies) according to the supplier's instructions. Briefly, cells were incubated with DNA (2 g total)-Lipo-fectAMINE (12 l) complexes in serum-free medium for 5-7 h. The DNA complexes were then removed from the medium and the cells were incubated with complete growth medium. Twenty-four hours after transfection, the cells were starved in Dulbecco's modified Eagle's medium with 0.1% fetal calf serum for 20 h and then used for experiments.
Far Western Blotting-For Far Western blot analysis of GST fusion proteins binding to the wild type Gab1, Gab1Y627F, Gab1Y659F, and Gab1FF, immunoprecipitates of these Gab1 proteins were resolved on 8.5% SDS-polyacrylamide gels and transferred to Immobilon-P filters (Milllipore). The filters were blocked with 5% ovalbumin in buffer B (10 mM Tris-HCl, 100 mM NaCl, 0.05% Tween 20) for 1.5 h, and then incubated with GST, GST-N-SH2, GST-C-SH2 proteins (2 g/ml) in the same buffer for 1.5 h. The filters were washed 6 times (5 min each), and then incubated sequentially with an anti-GST antibody and a horseradish peroxidase-conjugated secondary antibody, and developed with the SuperSignal chemiluminescent substrate-horseradish peroxidase substrate system (Pierce). Quantification was achieved by scanning the films with an optical scanner (Hewlett-Packard ScanJet 6100C/T with transparency adapter) and then analyzing band intensities of the images with the ImageQuant program (Molecular Dynamics).
ERK Kinase Assay-Cells were washed in cold phosphate-buffered saline and lysed in buffer A containing 1 mM dithiothreitol and 20 mM p-nitrophenyl phosphate. HA-ERK2 was immunoprecipitated with a monoclonal antibody against the HA-tag. One-half of each lysate was used to determine ERK2 activity, while the other half was analyzed for ERK2 protein by immunoblotting. For the kinase assay, the immune complexes were incubated for 5 min at 30°C in 40 l of kinase buffer (20 mM Hepes, pH 7.5, 10 mM MgCl 2 , 1 mM dithiothreitol, 10 mM p-nitrophenyl phosphate, 40 M ATP, 0.375 mg/ml myelin basic protein (MBP)) containing 10 Ci of [␥-32 P]ATP (3000 Ci/mmol). The reaction was terminated by the addition of SDS gel loading buffer and heat denaturation at 95°C for 5 min. The reaction mixtures were resolved on 11% SDS-polyacrylamide gels. The gels were processed by autoradiography. For quantification, the gels were analyzed with a PhosphorImager. The relative ERK2 activity (fold) of each sample was then calculated by comparing the signal intensity of each MBP band with that of the control (serum-starved, empty vector, or Gab1-transfected cells), which had been set as 1. A polyclonal antibody against ERK2 was used for immunoblotting analysis of ERK2 protein in the immunoprecipitates.
Determination of Peptide Concentrations-The concentrations of non-phosphorylated peptides were determined by measuring A 280 and using molar extinction coefficients calculated by the DNASTAR program. To determine the concentrations of phosphopeptides, each phosphopeptide was completely hydrolyzed in 30 l of perchloric acid (25%, v/v) and sulfuric acid (25%, v/v) at 190°C for 1 h in a screw-capped, 13 ϫ 100-mm Pyrex test tube (Corning) (24,25). After acid hydrolysis, 95 l of Hepes (100 mM, pH 7.2) was added and 25 l of this mixture was used to determine the amount of free phosphate by the malachite green assay (see below). A standard curve of free phosphate concentrations was prepared using KH 2 PO 4 that had been treated exactly the same way as the phosphopeptides (25).
PTPase Assays-SHP2 PTPase activity was determined by measuring the phosphate released from phosphopeptides using the malachite green assay (25). The PTPase reaction was carried out at 30°C for 5 min in a 25-l reaction mixture containing 50 mM Hepes, pH 7.2, 0.2% bovine serum albumin, 1 mM EDTA, 1 mM dithiothreitol, and the indicated amounts of phosphopeptides and PTPase. The reaction conditions were set up so that Ͻ20% of the substrate was dephosphorylated by the end of the reaction. The reaction was stopped by addition of 50 l of malachite green/Tween 20 solution prepared as described (25). After further incubation for 30 min at room temperature, the absorbance of the mixture at 620 nm (A 620 ) was determined in a half-area, 96-well plate (Costar) with a microplate reader. A standard curve of free phosphate concentrations was prepared using KH 2 PO 4 in conditions identical to the PTPase assay. Although the standard curve was linear between 50 and 4000 pmol of phosphate, a more accurate standard curve was prepared each time in the 25-1000 pmol range.

RESULTS
Both Tyrosine 627 and Tyrosine 659 of Gab1 Are Required for SHP2 Binding to Gab1 and for EGF-stimulated ERK2 Activation-We previously showed that Gab1 with a Tyr-627 to Phe mutation (Gab1Y627F) lost SHP2 binding activity in EGFstimulated cells, indicating that Tyr-627 of Gab1 is involved in Gab1-SHP2 interaction (15). Inspection of the Gab1 amino acid sequence indicates that Tyr-659 of Gab1 (Y 659 VVV), a residue known to be phosphorylated by the EGF receptor (26), is also located in a potential binding site for the SH2 domain of SHP2. To test whether this tyrosine residue may coordinate with Tyr-627 to interact with the tandem SH2 domains of SHP2, we studied the effects of single and double mutations of Tyr-659 and Tyr-627 on SHP2 binding to Gab1. Mutated or wild type Gab1 was expressed in COS-7 cells by transfection. Gab1-SHP2 binding was then examined in serum-starved cells treated with or without EGF by immunoblot analysis after immunoprecipitation of Gab1 proteins with an anti-FLAG antibody (M2).
Little SHP2 was associated with Gab1 in serum-starved cells (Fig. 1A) in which Gab1 was not tyrosine-phosphorylated (Fig.  1B). Upon EGF stimulation, tyrosine residues in Gab1 became phosphorylated (Fig. 1B) and SHP2 was bound to the wild type Gab1 (Fig. 1A). Tyr to Phe mutation of Gab1 at either Tyr-627 (Gab1Y627F) or Tyr-659 (Gab1Y659F) or both (Gab1FF) resulted in loss of the EGF-dependent SHP2 binding activity (Fig. 1A). A control immunoblot showed that equal amounts of Gab1 proteins were present in Gab1 immunoprecipitates (Fig.  1C). This observation indicates that both Tyr-627 and Tyr-659 are required for SHP2 binding to Gab1 in response to EGF.
Immunoblotting analysis with an anti-phosphotyrosine antibody (RC20) indicated that Gab1Y627F, Gab1Y659F, and Gab1FF had lower levels of phosphotyrosine content than that of the wild type Gab1 in EGF-stimulated cells (Fig. 1B). This observation suggests that Tyr-627 and Tyr-659 were phosphorylated in the wild type Gab1 in EGF-stimulated cells, in agreement with a previous report (26). Tyrosine phosphorylation of Gab1FF, although reduced, was still detectable in EGF-stimulated cells, consistent with the observation that tyrosine residues besides Tyr-627 and Tyr-659 were phosphorylated upon EGF stimulation (15,26).
To determine whether Tyr-659 is required for EGF-stimulated ERK2 activation, Gab1Y659F, Gab1FF, and control constructs (Fig. 2) were co-expressed with HA-tagged ERK2 in COS-7 cells, and EGF-stimulated ERK2 activation was determined. As shown in Fig. 2, EGF treatment resulted in approximately a 12-fold activation of ERK2. Expression of the wild type Gab1 had little effect on EGF-stimulated ERK2 activation, while expression of Gab1Y627F had a dominant inhibitory effect on EGF-stimulated ERK2 activation as reported previously (15). Similar to Gab1Y627F, expression of Gab1Y659F or Gab1FF blocked EGF-stimulated ERK2 activation. Thus, both Tyr-627 and Tyr-659 of Gab1 are necessary for Gab1 to mediate the EGF-stimulated ERK2 activation.
N-SH2 Domain of SHP2 Interacts Selectively with Tyr(P)-627 of Gab1-To further characterize the interaction between Gab1 and SHP2, we performed Far Western blot analysis using GST fusion proteins containing either the N-SH2 or C-SH2 domain of SHP2. FLAG-tagged Gab1, Gab1Y627F, Gab1Y659F, and Gab1FF were immunoprecipitated from EGFstimulated COS-7 cells expressing Gab1, Gab1Y627F, Gab1Y659F, or Gab1FF. The immunoprecipitates were resolved on SDS-polyacrylamide gels, transferred to Immobilon-P filters, and probed with GST, GST-N-SH2, or GST-C-SH2, followed by an anti-GST antibody (Fig. 3, A-D). Alternatively, the filters were probed with an anti-FLAG antibody (Fig. 3E) or an anti-phosphotyrosine antibody (Fig. 3F). The binding of GST protein to Gab1 was minimal (Fig. 3B). In contrast, GST-N-SH2 and GST-C-SH2 bound to the wild type Gab1 in an EGF-dependent manner but not to Gab1FF (Fig. 3,  A, C, and D). GST-N-SH2 bound to Gab1Y659F with a slightly reduced intensity but not to Gab1Y627F, indicating that the N-SH2 domain of SHP2 selectively binds to the Tyr(P)-627 site in Gab1 (Fig. 3, A and C). On the other hand, GST-C-SH2 preferentially bound to Gab1Y627F over Gab1Y659F (Fig. 3, A  and D). These results support our prediction based on data presented in Fig. 1 that both SH2 domains of SHP2 could interact with Gab1 in EGF-stimulated cells. These data also suggest that SHP2 interacts with Gab1 in a specific orientation in which the N-SH2 domain binds the Tyr(P)-627 site while the C-SH2 domain binds the Tyr(P)-659 site.
Activation of SHP2 PTPase by a Gab1 Bisphosphopeptide Containing Tyr(P)-627 and Tyr(P)-659 -To determine whether phosphopeptides containing Tyr-627 and/or Tyr-659 can activate the PTPase activity of SHP2, we assayed the SHP2 PTPase activity using two phosphopeptides as substrates (Fig.  4). The phosphopeptide SrcPY was used initially as a SHP2 substrate in our experiments. SrcPY contains the autoinhibitory Tyr(P)-527 site of c-Src. Previous reports have identified this peptide as a good substrate of constitutively active SHP2 that does not affect SHP2 activity (6,27). As shown in Fig. 4A  (open bars), under the assay conditions, the SrcPY (100 M) was dephosphorylated at a slow rate of 17 pmol/min by SHP2 (80 nM). Addition of monophosphopeptides (10 M) containing either Tyr(P)-627 (PY627) or Tyr(P)-659 (PY659) resulted in a slight increase in the phosphate release rate. This small change in phosphate release rate was due to dephosphorylation of PY627 or PY659 by SHP2, because the slightly higher rate was approximately the sum of the phosphate release rates of SrcPY and PY627 or PY659 when they were incubated separately with SHP2. Therefore, there was only an additive increase in the phosphate release rates, indicating that PY627 and PY659 did not activate the SHP2 PTPase. Similar results were obtained with higher concentrations of PY627 and PY659 that we have tested (up to 100 M PY627 or PY659, data not shown).
Similarly, a mixture of PY627 and PY659 could not activate SHP2 at all concentrations that we have tested (Fig. 4A, and data not shown). We then tested the effect of a bisphosphopeptide (PY627PY659) on SHP2 activity. PY627PY659 is a 43amino acid residue bisphosphoryl peptide containing Tyr(P)-627 and Tyr(P)-659 of Gab1 and the naturally occurring amino acid residues between and around Tyr(P)-627 and Tyr(P)-659. As shown in Fig. 4A, addition of PY627PY659 (10 M) to the reaction mixture resulted in 3-fold activation of SHP2. Activation of SHP2 by PY627PY659 was concentration-dependent (Fig. 4B). The activating activity of PY627PY659 requires phosphorylation of Tyr-627 and Tyr-659, because a non-phosphorylated peptide (Y627Y659) with identical amino acid residues did not have any apparent effect on SHP2 (Fig. 4, A and B, and data not shown). In fact, the non-phosphorylated Y627Y659 peptide was included as a negative control with SrcPY in our experiments presented in Fig. 4, A and B. Together, these data indicate that PY627PY659 can activate SHP2 PTPase and that this effect of PY627PY659 depends on both phosphorylation of Tyr-627 and Tyr-659 and the physical constraint of these two phosphotyrosine residues in a single peptide.
To confirm that PY627PY659 can activate SHP2, we performed the SHP2 PTPase activity assay using a Gab1-derived phosphopeptide, PY589, as substrate. PY589 contains Tyr(P)-589, which is one of the phosphatidylinositol 3-kinase-binding sites in Gab1 (15,18) and a likely target of SHP2 in the cells. As shown in Fig. 4A (black bars), we found that SHP2 was not activated by monophosphopeptides PY627, PY659, a mixture of PY627 and PY659, or the non-phosphorylated peptide Y627Y659. Again, addition of the bisphosphopeptide PY627PY659 to the reaction mixture resulted in activation of SHP2 PTPase (Fig. 4A).
To exclude the possibility that the apparent activation of SHP2 by PY627PY659 was due to kinetic difference of the phosphatase toward SrcPY, PY589, PY627, PY659, and FIG. 3. Far Western blot assay for binding of SH2 domains of SHP2 to Gab1. The wild type Gab1, Gab1Y627, Gab1Y659F, and Gab1FF were expressed in COS-7 cells. The cells were stimulated with EGF or left untreated. Gab1, Gab1Y627, Gab1Y659F, and Gab1FF were immunoprecipitated with an anti-FLAG antibody, resolved on 8.5% SDS-polyacrylamide gels, and transferred to Immobilon-P filters. The filters were incubated with GST (B), GST fusion protein of the Nterminal SH2 domain of SHP2 (GST-N-SH2, C), GST fusion protein of the C-terminal SH2 domain of SHP2 (GST-C-SH2, D), followed by an anti-GST antibody. Alternatively, the filters were probed with an anti-FLAG antibody (␣FLAG, E) or an anti-phosphotyrosine antibody (␣PTyr, F). A, relative band intensities from four experiments. PY627PY659, we measured the dephosphorylation of these phosphopeptides by a constitutively active SHP2. An N-SH2 domain deletion mutant of SHP2 (SHP2⌬N) was used for this purpose. Similar results were obtained in reactions performed at 50 and 100 M peptide concentrations (Fig. 4C). As shown in Fig. 4C, the active SHP2 dephosphorylated PY627PY659 faster than SrcPY but at a slower rate than PY589. Because PY627PY659 increased the reaction rate of SHP2 when either SrcPY or PY589 was used as substrate, the higher rate could not be attributed to a potential preferential hydrolysis of PY627PY659. Second, the active SHP2 dephosphorylated PY627 faster than PY627PY659, yet PY627 did not increase the SHP2 PTPase activity as PY627PY659 did. This result again rules out the possibility that activation of the SHP2 PTPase activity by PY627PY659 observed in Fig. 4, A and B, was due to preferential utilization of substrate.
Association of Activated SHP2 with Gab1 Is Required for EGF-induced ERK2 Activation-Our in vitro study suggests (Fig. 4) that a consequence of Gab1-SHP2 interaction is activation of SHP2 PTPase. Because SHP2 PTPase activity is required for ERK2 activation in response to EGF stimulation (7,8), we asked if activation of SHP2 is the only function of Gab1-SHP2 interaction. To test this possibility, we determined whether expression of a constitutively active SHP2 could rescue the inhibitory effect of Gab1FF, which neither binds SHP2 nor is predicted to activate its PTPase activity. As shown in Fig. 5A, the N-SH2 domain deletion mutant of SHP2 (SHP2⌬N) had a constitutively activated PTPase activity, which could not be activated further by the Gab1 bisphosphoryl peptide PY627PY659.
The wild type SHP2 or SHP2⌬N was co-expressed with the wild type Gab1 or Gab1FF in COS-7 cells and EGF-stimulated ERK2 activation in the transfected cells was analyzed (Fig.  5B). Expression of SHP2 or SHP2⌬N with the wild type Gab1 had little effect on EGF-stimulated ERK2 activation (compare Fig. 2 and Fig. 5C). As shown previously in Fig. 2, Gab1FF has a dominant inhibitory effect on EGF-stimulated ERK2 activation. Fig. 5, C and D (lane 8), shows that overexpression of SHP2 could not rescue the inhibitory effect of Gab1FF on ERK2 activation. Similarly, expression of the constitutively active SHP2⌬N did not rescue the inhibition of Gab1FF on ERK2 activation (Fig. 5, C and D, lane 9). This result suggests that the activated SHP2 must associate with Gab1 in order to mediate ERK2 activation.
Expression of a Chimeric Protein Consisting of SHP2 Binding-defective Gab1 and Active SHP2 PTPase Results in Constitutive Activation of ERK2-To further test the possibility that association of activated SHP2 PTPase with Gab1 is necessary for ERK2 activation, we constructed a plasmid for expression of a chimeric protein in which SHP2⌬N is fused to the C terminus of Gab1FF. Fig. 5B shows that the Gab1FF-SHP2⌬N chimera was expressed in COS-7 cells. Interestingly, cells transfected with Gab1FF-SHP2⌬N contained constitutively elevated ERK2 activity (Fig. 5, C and D, and Fig. 6, A and B). Figs. 5E and 6C show that similar amounts of HA-ERK2 were present in the immune complexes used for the ERK2 activity assay. The ERK2 activity in transfected cells was dependent on the amount of Gab1FF-SHP2⌬N cDNA used to transfect cells (Fig.   FIG. 4. SHP2 PTPase activity assay 6). At a ratio of 10:1 for Gab1FF-SHP⌬N cDNA:HA-ERK2 cDNA, the constitutive ERK2 activity in COS-7 cells transfected with Gab1FF-SHP2⌬N was ϳ6-fold above the basal level. This level of constitutive ERK2 activity was similar to that detected in COS-7 cells transfected with expression vectors for the G ␤1 /G ␥2 subunits of heterotrimeric G proteins (Fig.  6). G ␤1/ G ␥2 is one of the best combinations of G ␤ and G ␥ subunits known to activate ERK2 constitutively in COS-7 cells (28,29).
Catalytic Cysteine Mutant of SHP2 (SHP2CS) Protects Tyrosine Residues in Gab1 from Dephosphorylation-Our in vitro study indicated that phosphopeptides containing Tyr(P)-627 and Tyr(P)-659 are SHP2 substrates (Fig. 4). To evaluate whether Tyr(P)-627 and Tyr(P)-659 of Gab1 are dephosphorylated by SHP2 in the cells, we co-expressed FLAG-tagged Gab1 with the wild type or a catalytically inactive SHP2 (SHP2CS, catalytic Cys to Ser mutation) in COS-7 cells. Following EGF stimulation for various times, tyrosine phosphorylation of Gab1 and the amount of SHP2 retained in the Gab1 complex were compared in cells transfected with SHP2 and SHP2CS.
It has been shown that the catalytic Cys to Ser mutants of some PTPases can form a stable complex with their substrates and protect these substrates from dephosphorylation (30 -32). Fig. 7, A and D, shows that little tyrosine phosphorylation of Gab1 was detectable in serum-starved cells transfected with the wild type SHP2. EGF stimulation resulted in a rapid increase in Gab1 tyrosine phosphorylation that peaked at 1 min. The Gab1 tyrosine phosphorylation gradually decreased to about half of the maximal at 30 min after EGF stimulation in these cells. This corresponded to the change in the amount of SHP2 in the Gab1 immune complex (Fig. 7B), suggesting that Tyr(P)-627 and/or Tyr(P)-659 of Gab1 were dephosphorylated in these cells.
Cells transfected with SHP2CS had an elevated level of Gab1 tyrosine phosphorylation and Gab1-SHP2 complex in the serum-starved condition (Fig. 7, A and D). After EGF stimulation, there was a continuous rise in Gab1 tyrosine phosphorylation. Importantly, this corresponded to a continuous increase in the amount of SHP2 associated with Gab1 (Fig. 7B). Therefore, SHP2CS was able to trap Gab1 in a tyrosine-phosphorylated state in EGF-stimulated cells, suggesting that Gab1 is a substrate of SHP2 in the cells. DISCUSSION Previous experiments have shown that Tyr-627 is required for SHP2 binding to Gab1 in cells stimulated with EGF or insulin (15,18). The present study provides evidence that Tyr-659, like Tyr-627, is required for SHP2 binding to Gab1 in EGF-stimulated cells and for ERK2 activation by EGF. Because SHP2 contains two SH2 domains that are arranged in tandem in the N-terminal portion of SHP2, the requirement of both Tyr-627 and Tyr-659 for Gab1-SHP2 association suggests that both phosphotyrosines may simultaneously interact with the tandem SH2 domains of SHP2. If this is the case, Tyr(P)-627 and Tyr(P)-659 should be able to bind to different SH2 domains in SHP2. Our Far Western blot binding assay showed that Tyr(P)-627 selectively interacted with the N-SH2 domain  1, 3, 5, 8, and 10 so that equal amounts of HA-tagged HA-ERK2 could be immunoprecipitated from each cell lysate. B, 9 g of each cell lysate supernatant was analyzed by immunoblotting with an antibody against the C-terminal region of SHP2. SHP2 indicates the endogenous SHP2. Note that only a fraction of cells in each sample was predicted to be transfected with the plasmids and express the exogenous proteins. Therefore, there were higher levels of exogenous proteins than endogenous SHP2 in transfected cells than what appears in Panel B. HA-ERK2 was immunoprecipitated from cell lysate supernatants. One-half of each immunoprecipitate was used to determine ERK2 kinase activity using MBP as substrate (C and D). The other half of each immunoprecipitate was analyzed by immunoblotting with an anti-ERK antibody (E). C represents the means Ϯ S.D. of ERK2 activity from three independent experiments. D, a representative autoradiograph. Chimera, Gab1FF-SHP2⌬N. of SHP2, whereas Tyr(P)-659 preferentially bound to the C-SH2 domain of SHP2. This result suggests that it is possible for the tandem SH2 domains to bind to Tyr(P)-627 and Tyr(P)-659 simultaneously in a specific orientation, in which Tyr(P)-627 binds to the N-SH2 domain and Tyr(P)-659 binds to the C-SH2 domain. In other experiments, we assayed SHP2 PTPase activity using p-nitrophenyl phosphate as substrate. We found that peptide PY627, but not peptide PY659, was sufficient to activate SHP2 PTPase in this assay (data not shown). Because SHP2 activation requires binding of a TAM to its N-SH2 domain, this observation implies that PY627 has a higher affinity than PY659 for the N-SH2 domain of SHP2. Therefore, this observation provides another line of evidence to support the notion that Tyr(P)-627 binds to the N-SH2 domain while Tyr(P)-659 binds to the C-SH2. Interestingly, the same orientation specific binding of the tandem SH2 domains of SHP2 to the BTAM of IRS-1 has also been suggested (33).
Crystal structure analysis indicates that, in the absence of N-SH2 domain binding peptide, the N-SH2 domain of SHP2 interacts with its catalytic domain and thus the enzyme remains in an inactive state (3,4). Binding of phosphopeptide to the N-SH2 domain induces a conformational change that activates the PTPase (3,5). We found that a bisphosphopeptide (PY627PY659) derived from Gab1 that contains both the Tyr(P)-627 and Tyr(P)-659 TAMs could activate SHP2 PTPase. In previous studies with IRS-1 BTAM, artificial aminohexanoic acid spacers were used to link the two TAMs that are separated by a long amino acid sequence (3,5). PY627PY659 used in our study represents the longest BTAM with a natural amino acid sequence that has been tested for SHP2 activation.
In a previous study using artificial IRS-1 BTAM, SHP2 activation was detected at 0.5 M of optimally spaced Tyr(P)-1172 and Tyr(P)-1222 TAMs, and a maximal activation of about 7-fold was observed (4). In our study, a net increase in phosphate release was detectable at the lowest concentration of PY627PY659 (0.25 M) that we have tested. PY627PY659 at 1 and 10 M activates SHP2 about 1-and 3-fold, respectively. In order to keep a low concentration ratio of the activating peptide (PY627PY659) to the designated substrate (SrcPY or PY589), we did not attempt to achieve maximal SHP2 activation. Furthermore, different PTPase assays used in these two studies may contribute to the slight difference in the results.
Our observation that PY627PY659 activates SHP2 PTPase suggests that SHP2 binding to Gab1 would result in SHP2 activation, which is necessary for EGF-stimulated ERK2 activation. Importantly, monophosphopeptides PY627 and PY659 FIG. 6. Constitutive activation of ERK2 in COS-7 cells expressing Gab1FF-SHP2⌬N chimera. COS-7 cells in 60-mm plates were transfected with 0.2 g of HA-ERK2 DNA, the indicated amounts of Gab1FF-SHP2⌬N DNA (and pcDNA3.1 to keep total DNA at 2 g for each transfection), or the expression plasmids (0.9 g/each) for G ␤1 (pCDM8-G ␤1 ) and G ␥2 (pCDM8-G ␥2 ). Cells were serum-starved for 20 h and then either left untreated or treated with EGF as indicated. HA-ERK2 activity was determined by immune complex kinase assay as described in legend for Fig. 5. A, average Ϯ ranges of ERK2 activity from two experiments. B, a representative autoradiograph showing MBP phosphorylation by HA-ERK2. C, a representative immunoblot filter showing HA-ERK2 protein in the immunoprecipitates. Chimera, Gab1FF-SHP2⌬N; ␤ 1 ␥ 2 , pCDM8-G ␤1 ϩ pCDM8-G ␥2 . could not activate SHP2 when SHP2 was assayed using peptide substrates. A mixture of PY627 and PY659 also could not activate SHP2 when peptide substrates were used to assay SHP2. These observations reinforce our finding that both Tyr-627 and Tyr-659 of Gab1 are required for EGF-stimulated ERK2 activation and indicate that the phosphorylated Tyr-627 and Tyr-659 of Gab1 constitute a BTAM for the binding and activation of SHP2. Furthermore, these results suggest that the physical constraint between the two TAMs in the Gab1 BTAM is necessary for SHP2 activation when natural phosphopeptides, rather than p-nitrophenyl phosphate, were its substrates.
We found that expression of a constitutively active SHP2 was insufficient to overcome the inhibition by Gab1FF of EGFstimulated ERK2 activation. Interestingly, fibroblasts isolated from mouse embryos with a targeted deletion mutation of the N-SH2 domain of SHP2 were defective in ERK activation (34), suggesting that the activated SHP2 needs to associate with one or more cellular proteins through either its N-SH2 or through both SH2 domains to mediate ERK activation. Importantly, we found that expression of a fusion protein consisting of the constitutively active SHP2⌬N-linked to the C terminus of the SHP2 binding defective Gab1FF resulted in a constitutive activation of ERK2 in COS-7 cells. These results indicate that the activated SHP2 must associate with Gab1 in order to mediate ERK activation by EGF and that constitutive association of activated SHP2 with Gab1 will result in a constitutively elevated ERK2 kinase activity in the cells.
The reason that the activated SHP2 has to associate with Gab1 in order to mediate ERK activation is unclear at present, but several possibilities may exist. For example, (a) SHP2 may need to dephosphorylate a specific negative phosphorylation site on Gab1; (b) SHP2 may need to dephosphorylate a Gab1associated phosphoprotein as suggested (17); (c) SHP2 may require Gab1 to translocate it to a specific cellular compartment to dephosphorylate a target that is not associated with Gab1.
Phosphopeptides derived from Tyr(P)-589, Tyr(P)-627, and Tyr(P)-659 of Gab1 were dephosphorylated by SHP2 in vitro. Our substrate trapping study showed that SHP2CS protected Gab1 from dephosphorylation with a concomitant increase in the amount of Gab1⅐SHP2CS complex, indicating that SHP2CS was able to trap Gab1 Tyr(P) residues. These experiments demonstrate that Gab1 not only is a SHP2 activator in EGFstimulated cells but also is a target of the activated SHP2 PTPase in the cells.