A Critical Role for Phosphoinositide 3-Kinase Upstream of Gab1 and SHP2 in the Activation of Ras and Mitogen-activated Protein Kinases by Epidermal Growth Factor*

Although the mechanisms involved in the activation of mitogen-activated protein kinases (MAPK) by receptor tyrosine kinases do not display an obvious role for phosphoinositide 3-kinases (PI3Ks), we have observed in the nontransformed cell line Vero stimulated with epidermal growth factor (EGF) that wortmannin and LY294002 nearly abolished MAPK activation. The effect was observed under strong stimulation and was independent of EGF concentration. In addition, three mutants of class Ia PI3Ks were found to inhibit MAPK activation to an extent similar to their effect on Akt/protein kinase B activation. To determine the importance of PI3K lipid kinase activity in MAPK activation, we have used the phosphatase PTEN and the pleckstrin homology domain of Tec kinase. Overexpression of these proteins, but not control mutants, was found to inhibit MAPK activation, suggesting that the lipid products of class Ia PI3K are necessary for MAPK signaling. We next investigated the location of PI3K in the MAPK cascade. Pharmacological inhibitors and dominant negative forms of PI3K were found to block the activation of Ras induced by EGF. Upstream from Ras, although association of Grb2 with its conventional effectors was independent of PI3K, we have observed that the recruitment of the tyrosine phosphatase SHP2 required PI3K. Because SHP2 was also essential for Ras activation, this suggested the existence of a PI3K/SHP2 pathway leading to the activation of Ras. In addition, we have observed that the docking protein Gab1, which is involved in PI3K activation during EGF stimulation, is also implicated in this pathway downstream of PI3K. Indeed, the association of Gab1 with SHP2 was blocked by PI3K inhibitors, and expression of Gab1 mutant deficient for binding to SHP2 was found to inhibit Ras stimulation without interfering with PI3K activation. These results show that, in addition to Shc and Grb2, a PI3K-dependent pathway involving Gab1 and SHP2 is essential for Ras activation under EGF stimulation.

The mitogen-activated protein kinases (MAPK) 1 extracellu-lar signal-regulated kinases (ERK) 1 and 2 transduce proliferative signals to the nucleus (1). The mechanisms leading to their activation by ligands of receptor tyrosine kinases appear well understood and the GTPase Ras plays a central role (2). For example, epidermal growth factor (EGF) activates its receptor tyrosine kinase, which autophosphorylates, creating binding sites for SH2-domain containing proteins, including the adapter proteins Shc and Grb2. In addition to its SH2 domain, Grb2 binds through its SH3 domains to the guanine nucleotide exchange factor Sos. Thus, the binding of Grb2 to phosphorylated EGF receptor (EGFR) results in the recruitment of Sos to the plasma membrane and has been proposed as a model for activation of membrane-bound Ras (3). In addition, EGF-induced activation of Ras may be transduced via Shc, which binds to activated EGFR and becomes phosphorylated, creating an additional binding site for Grb2 (4). Once Ras has been activated by Sos, GTP-bound Ras stimulates the serine/ threonine kinase Raf. Activated Raf stimulates the downstream kinase MAPK/ERK kinase (MEK), which in turn phosphorylates ERK. In addition, activation of ERK under EGF stimulation can be mediated by Ras-independent pathways, through protein kinase C (PKC) and calcium-mediated mechanisms (5).
The phosphoinositide 3-kinases (PI3Ks) also transduce proliferative signals. PI3Ks phosphorylate phosphoinositides at the 3Ј-position of the inositol ring, and their major lipid product is phosphatidylinositol 3,4,5-trisphosphate (PIP 3 ), which is produced during cell stimulation by various mitogens (6,7). Three classes of PI3Ks have been defined, and class I enzymes are involved in mitogen signaling. The members of the subclass Ia include the catalytic subunits p110␣, p110␤, and p110␦ associated with a regulatory subunit p85 and activated through protein tyrosine kinases, and subclass Ib is represented by p110␥, which is activated by heterotrimeric G proteins. On a functional point of view, subclass Ia PI3Ks are required for growth factor-induced mitogenesis (8,9), and it was recently reported that embryos of p110␣ knockout mice die at early age due to a profound proliferative defect (10). The mechanisms by which PI3Ks activate signaling pathways have been recently unraveled. PIP 3 has binding affinity for a conserved peptidic sequence called the pleckstrin homology (PH) domain, thereby inducing the localization of PH-domain-containing proteins to membrane-associated signaling complexes (6). Several targets for PI3K lipid products have been proposed, including the proto-oncogene product Akt/protein kinase B (PKB) and its upstream activators, the phosphoinositide-dependent kinases. These kinases activate various enzymes that are important for cell growth, including the p70-S6 kinase and the glycogen synthase kinase 3 (6,7). In addition, the catalytic subunits of PI3Ks possess an intrinsic protein kinase activity, which is involved in the down-regulation of their lipid kinase activity (11,12). Interestingly, at least in the case of p110␥, this protein kinase activity has been reported to participate to MAPK signaling, whereas its lipid kinase activity appeared not to be necessary (13).
Although the model for growth factor-induced MAPK activation described above does not show an obvious role for PI3K, many reports have documented inhibition of MAPK activation by pharmacological inhibitors of PI3K (14 -16). However, recent data have suggested that the role of PI3K depends on signal strength and is in fact limited to weak activations. In Swiss 3T3 cells, PI3K was found to be required during stimulation induced by low, but not high, doses of platelet-derived growth factor (17). In insulin-treated Chinese hamster ovary cells, the requirement for PI3K was reported to depend on the number of insulin receptors expressed on the cell surface (17). Similarly, in COS cells, pharmacological inhibitors of PI3K were found to inhibit MAPK activation induced only by low doses of EGF (18). On a molecular point of view, two major mechanisms have been proposed to illustrate a possible involvement of PI3K in MAPK activation. One involves the ability of PIP 3 to activate members of the PKC family, directly or via the phosphoinositide-dependent kinases (19,20). Activated PKC can then stimulate Raf (21). The second involves the ability of p21-activated kinase, a downstream target of PI3K via Rac, to promote stimulation of Raf and MEK (22,23).
In the nontransformed cell line Vero, we have observed that various PI3K inhibitors block MAPK activation induced by EGF, independently of signal strength. This led us to show that a PI3K-dependent pathway involving Gab1 and SHP2 is required, in addition to Shc and Grb2, for the activation of Ras by EGF.

EXPERIMENTAL PROCEDURES
Materials-Human recombinant EGF was from Calbiochem. Polyclonal antibodies against Grb2, SHP2, and EGFR and monoclonal anti-Myc were from Santa Cruz Biotechnology. Polyclonal antibodies against Gab1, Sos, Shc, p85, and monoclonal anti-phosphotyrosine (4G10) were from Upstate Biotechnology Inc. Anti-phospho-ERK antibody was from Promega. Monoclonal anti-pan Ras was from Oncogene Research. Monoclonal anti-His tag antibody was from Invitrogen, anti-HA tag was from Roche Molecular Biochemicals, and anti-T7 tag was from Novagen. Cell culture reagents were from Life Technologies, Inc.
Cell Culture, Transfection, and Stimulations-Vero cells (a monkey kidney cell line, ATCC CCL 81) were maintained in Dulbecco's modified Eagle's medium supplemented with 7.5% fetal bovine serum and antibiotics. For transfection experiments, cells in 60-mm plates were incubated 3 h with 2 ml of Dulbecco's modified Eagle's medium containing 2 g of total DNA, 6 l of LipofectAMINE, and 6 of Plus reagent (Life Technologies, Inc.). Before stimulation, cells were blocked overnight by serum starvation. Unless otherwise indicated, cells were stimulated for 5 min with 10 ng/ml EGF. Before stimulation, cells were incubated for 15 min with 100 nM wortmannin (Sigma Chemical Co.) or 25 M LY294002 (BioMol) where indicated.
Cell Lysis, Immunoprecipitation, and Immunoblotting-Cells were scrapped off in lysis buffer containing 20 mM Tris, pH 7.4, 150 mM NaCl, 10 mM EDTA, 10% glycerol, 1% Nonidet P-40, 10 g/ml of each aprotinin and leupeptin, and 1 mM orthovanadate. After shaking for 15 min at 4°C, soluble material was incubated with the appropriate antibody for 2 h at 4°C. The antigen-antibody complexes were collected with protein A-or protein G-Sepharose (Sigma) for 1 h and washed three times with lysis buffer. Blots were developed using chemiluminescence (Amersham Pharmacia Biotech).
Preparation of Expression Plasmids-A plasmid encoding wild type HA-tagged p110␤ was kindly provided by Drs. P. Hu and J. Schlessinger (New York University) (24). The kinase-inactive K805R mutant of p110␤ was obtained by replacing with arginine the lysine 805 present in the putative ATP binding site using site-directed mutagenesis (QuikChange, Stratagene) with the following mutagenic primer: 5Ј-GTTGGAGTGATTTTTAGAAATGGTGATGATTTACG-3Ј (the changed nucleotide is underlined). The cDNA of Tec was cloned using RT-PCR (Superscript II, Life Technologies) from poly(A)ϩ RNA purified from the megakaryoblastic cell line Dami and inserted into pET21 (Novagen) to introduce a N-terminal T7 epitope tag. The cDNA of T7-Tec was then subcloned into pCI-neo (Promega) for mammalian expression. A construct encoding the N-terminal PH domain was obtained by deleting the fragment between the two StuI internal sites which encompasses the SH3, SH2 and kinase domains. The R29C mutant was obtained using the following mutagenic primer: 5Ј-CGCCCTTAAACTACAAAGAGT-GCCTTTTTGTACTTACAAAGTCC-3Ј. The cDNA of Gab1 was kindly provided by Dr. A. Ullrich (Max Planck Institute, Germany). To generate the mutant of Gab1 defective for p85 binding (YF3), tyrosines 447, 472, and 589 were replaced with phenylalanines using the following primers, respectively: 5Ј-CTGGATGAAAATTTCGTCCCAATGAATC-3Ј; 5Ј-CAGGAAGCAAATTTTGTGCCAATGACTC-3Ј; 5Ј-CAGTGAA-GAGAATTTTGTTCCCATGAACC-3Ј. The SHP2 binding site of Gab1 was mutated by replacing tyrosine 627 with phenylalanine using the following primer: 5Ј-GGAGACAAACAGGTGGAATTCTTAGATCTCG-ACTTAGA-3Ј. Gab1 deleted of the PH domain was generated by polymerase chain reaction amplification of the cDNA sequence encoding amino acids 104 -694. In this mutant, translation is initiated on the second methionine at position 104. All Gab1 constructs have a Cterminal Myc tag. In addition, HA-tagged Gab1 was obtained by subcloning the entire coding sequence of wild type Gab1 into pcDNA-HA using polymerase chain reaction amplification. All the mutations were verified by sequencing. A construct encoding His/Myc-tagged ERK1 was obtained by subcloning ERK1 (kindly provided by Dr. E. vanObberghen, Nice, France) into pcDNA3.1-MycHis (Invitrogen). The constructs encoding HA-tagged PTEN and its G129E mutant have been already described (25). The vectors encoding ⌬p85 and p110␣-K802R mutants were kindly provided by Drs. W. Ogawa (University of Kobe, Japan) and M. Wymann (University of Fribourg, Switzerland), respectively. The plasmids encoding SHP2 and its catalytically inactive mutant (C/S) have been kindly provided by Dr. N. Rivard (Sherbrooke University, Canada).
In Vitro Kinase Assays-To measure activation of Akt/PKB, cells were cotransfected with 0.5 and 1.5 g of DNA encoding HA-tagged Akt/PKB and the indicated effector, respectively. After stimulation, cells were scrapped off in lysis buffer, then subjected to HA immunoprecipitation using 12CA5 antibody (Roche Diagnostics). Immunoprecipitates were washed with lysis buffer, then with kinase buffer containing 50 mM Tris, pH 7.5, 10 mM MgCl 2 , 1 mM dithiothreitol and incubated with Histone 2B (Roche Diagnostics) and [␥-32 P]ATP as described previously (26). To measure MAPK activation, cells were transfected with 1 g of each DNA encoding ERK1-His and the indicated effector protein. After stimulation, cells were harvested in lysis buffer supplemented with 300 mM NaCl. Soluble material was incubated with 30 l of ProBond Resin (Invitrogen), then washed with lysis buffer supplemented with 5 mM imidazole, followed by washes with kinase buffer. Phosphorylation of myelin basic protein (MBP) was performed as described (27). Reactions were stopped by addition of Laemmli sample buffer and analyzed by SDS-PAGE. Phosphorylation of MBP and histones was quantified using a PhosphorImager.
Activated Ras Affinity Precipitation Assay-The assay was performed essentially as described previously (28). The recombinant Rasbinding domain (RBD) of Raf1 (kindly provided by Dr. F. R. McKenzie, Nice, France) was expressed as GST fusion protein in Escherichia coli and extracted using glutathione-Sepharose beads. To measure Ras activation, Vero cells were scrapped off in 1 ml of lysis buffer containing 50 mM Tris, pH 8.0, 150 mM NaCl, 10 mM MgCl 2 , 0.5% sodium desoxycholate, 1% Nonidet P-40, 0.1% SDS, 1 mM phenylmethylsulfonyl fluoride, and 1 g/ml each of aprotinin and leupeptin. Cleared lysates were incubated at 4°C for 30 min with 30 g of GST-RBD bound to glutathione-Sepharose beads. Beads were washed three times in lysis buffer and then boiled, and proteins were resolved by SDS-PAGE. Immunoblotting was performed with anti-pan Ras antibodies. To study the activation of Ras in transfected cells, cells were cotransfected with 1 g of each plasmid encoding HA-tagged wild type Ras (kindly provided by Dr. B. M. Burgering, Utrecht, The Netherlands) and the indicated effector. The GST-RBD pull-down assay was performed as above, except that immunoblots were revealed with anti-HA antibody.
Membrane Fractions-Membrane fractions were prepared as described (29). Briefly, cells were scrapped off in hypotonic lysis buffer then Dounce-homogenized. The homogenate was centrifuged at 100,000 ϫ g for 1 h. The pellet was dissolved in 1% Triton X-100 lysis buffer, and the insoluble material was spun out. This was taken as the solubilized membrane fraction.

RESULTS
PI3K Is Required for EGF-induced MAPK Activation, Independently of Signal Strength-By using anti-phospho-ERK immunoblotting, we have observed in Vero cells that the phosphorylation of ERK2 induced by 10 ng/ml EGF is abolished when cells are preincubated with wortmannin or LY2940002 (Fig. 1A). Because the role of PI3K in ERK activation is thought to be limited to weak stimulations, we have measured the activation of transfected ERK1-His using an in vitro kinase assay. As shown in Fig. 1B, treatment with 10 ng/ml EGF induced a Ͼ10-fold increase in ERK1-His activity, in agreement with the strong phosphorylation of endogenous ERK2 observed using anti-phospho-ERK immunoblotting. Preincubation of the cells with PI3K inhibitors reduced by over 80% the activation of purified ERK1-His. In addition, the requirement for PI3K appeared independent of EGF concentration, because wortmannin treatment also inhibited the phosphorylation of endogenous ERK2 induced by 30 or 50 ng/ml EGF (Fig. 1C), To confirm these data obtained with pharmacological inhibitors, we have studied the ability of three PI3K mutants to inhibit ERK1-His activation in cotransfection experiments. In parallel, we have determined their efficiency to inhibit PI3K signaling by measuring their ability to interfere with the activation of Akt/PKB. This was achieved in cotransfection experiments with HA-Akt/PKB, followed by an in vitro kinase assay. First, cells were transfected with a form of p85␣ lacking the p110 binding site (⌬p85), which is a widely used dominant negative mutant for class Ia PI3Ks (30). As shown in Fig. 2A, expression of ⌬p85 almost completely inhibited the activation of ERK1-His induced by EGF. Because ⌬p85 contains SH2 and SH3 domains that might interfere with MAPK activation, we have tested catalytically inactive mutants of p110␣ (␣K802R) and p110␤ (␤K805R) subunits. Expression of these mutants was found to produce a partial inhibition of ERK1-His activation, the p110␤ construct being more efficient than p110␣. Lysates were also subjected to anti-ERK2 immunoblotting (lower panel).

FIG. 2. Inhibition of ERK1-His and HA-Akt/PKB activation by expression of PI3K mutants.
A, cells were cotransfected with ERK1-His and one of the indicated constructs: empty vector (V); dominant negative p85 (⌬p85); catalytically inactive p110␣ (␣K802R); catalytically inactive p110␤ (␤K805R). After cell stimulation and lysis, the activation of purified ERK1-His was determined by in vitro kinase assay. Right panels, expression of PI3K mutants was verified in lysates from cells transfected with the indicated construct. B, cells were cotransfected with HA-tagged Akt/PKB and the same constructs as in A. Following cell stimulation, HA-Akt/PKB was immunoprecipitated and incubated with histones (H2B) and [␥-32 P]ATP. Phosphorylation of histones was revealed using a PhosphorImager. The bottom graph represents the mean Ϯ S.E. of three independent experiments.
However, these mutants also inhibited partially the activation of HA-Akt/PKB, and the p110␤ construct was more efficient (Fig. 2B). Thus, the ability of the PI3K mutants to inhibit MAPK activation was correlated to their capacity to interfere with PI3K signaling. Taken together, these results indicate that, in Vero cells stimulated with EGF, PI3K plays an essential role in MAPK activation.
PIP3 Is Essential for MAPK Activation-One important question regarding the involvement of PI3Ks in MAPK activation is the respective role of their lipid kinase and protein kinase activities, because PIP 3 has been shown not to be necessary for p110␥-mediated activation of MAPK (13). To define the importance of PI3K lipid products in EGF signaling, we have taken advantage of PTEN, a protein phosphatase that is also capable of dephosphorylating PI3Ks lipid products (25). As shown in Fig. 3, overexpression of PTEN inhibits ERK1-His activation induced by EGF. To determine whether the PTEN effect is due to its protein or lipid phosphatase activity, we have used as a control the "protein phosphatase only" mutant of PTEN (G129E), which does not interfere with PI3K signaling (25,31). Transfection of PTEN-G129E did not significantly inhibit ERK1-His activation, whereas this mutant was somewhat more expressed than wild type PTEN (Fig. 3). To confirm the importance of PI3K lipid products, we have used the PH domain of Tec as a competitor for binding to PIP 3 . Tec belongs to a family of tyrosine kinases containing a PIP 3 -sensitive PH domain (32,33). As shown in Fig. 3, expression of the Tec PH domain significantly inhibited ERK1-His activation induced by EGF. As a control, we have used the Tec PH domain containing the R29C mutation, which decreases the affinity of Tec kinases for PIP 3 (32,33). This mutant did not significantly inhibit ERK1-His activation. Altogether, these results suggested that PIP 3 produced during EGF stimulation is necessary for MAPK activation.
PI3K Is Required Upstream of Ras and Sos-We next investigated the location of PI3K function in the MAPK pathway. Considering that PI3K-dependent mechanisms have been proposed at the level of Raf and MEK (19 -23), and that MAPK activation by EGF can occur via Ras-independent pathways (5), we have first examined the role of Ras in Vero cells. Expression of dominant negative RasN17 was found to abolish ERK1-His activation induced by EGF (data not shown), indicating that MAPK activation is strictly dependent on Ras in Vero cells. Because MAPK activation also requires PI3K, we have determined whether PI3K was involved in the activation of Ras. This was achieved using a precipitation assay for activated Ras. Following cell stimulation and lysis, endogenous activated Ras was extracted using a GST fusion protein containing the Ras-binding domain of Raf (RBD), which interacts specifically with GTP-bound Ras (28). The amount of activated Ras in the GST-RBD pull-down assays was determined by anti-Ras immunoblotting. As shown in Fig. 4A, treatment of Vero cells with 10 ng/ml EGF induced the coprecipitation of Ras with GST-RBD, and preincubation of the cells with wortmannin or LY294002 strongly inhibited this association, suggesting that PI3K is involved in Ras activation. To confirm this hypothesis, the activation of HA-tagged Ras was studied in cells cotransfected with dominant negative PI3Ks. As shown in Fig. 4B, expression of ⌬p85 or p110␤-K805R inhibited the amount of HA-Ras associated with the GST-RBD protein. Taken together, these results indicate that PI3K plays a critical role in the activation of Ras induced by EGF.
Upstream from Ras, membrane translocation of Sos is thought to be the limiting event for Ras activation. To determine whether PI3K was involved in Sos redistribution, we have prepared membrane fractions from EGF-treated cells preincubated or not with PI3K inhibitors. As shown in Fig. 4C, the membrane enrichment of Sos induced by EGF was nearly abolished by PI3K inhibitors. In addition, overexpression of Sos was found to partially overcome the need for PI3K in ERK activation without increasing the basal activation of ERK1-His (Fig. 4D). This suggested that PI3K was involved at the level or upstream of Sos. However, expression of constitutively activated PI3K (p110␣-CAAX) was not sufficient to activate MAPK in unstimulated cells (data not shown). This indicated that PIP 3 cannot directly induce the redistribution of Sos, suggesting the existence of a PIP 3 -dependent mechanism upstream from Sos.
SHP2 and Ras Are Activated Downstream of PI3K and Gab1-To identify a PI3K-dependent event upstream from Sos, we have analyzed by immunoblotting the proteins coimmunoprecipitated with Grb2. As expected, Fig. 5A shows that the EGFR and Shc readily precipitated with Grb2 upon EGF stimulation and this was not modified by PI3K inhibitors. In addition, PI3K inhibitors did not influence the constitutive association of Grb2 with Sos (data not shown). In contrast, the coimmunoprecipitation of the tyrosine phosphatase SHP2 with Grb2 was found to depend on PI3K (Fig. 5A). In agreement with this, the recruitment of SHP2 in anti-phosphotyrosine immunoprecipitates upon EGF stimulation was also reduced by PI3K inhibitors (Fig. 5B), suggesting that SHP2 is a downstream effector of PI3K in EGF signaling. To determine whether SHP2 was important for Ras activation in Vero cells, we have transfected a catalytically inactive mutant of SHP2 (SHP 2 -C/S) and analyzed the activation of HA-tagged Ras in cotransfection experiments. As shown in Fig. 5C, expression of SHP2-C/S completely inhibited the precipitation of HA-Ras with GST-RBD, whereas wild type SHP2 had no effect. This indicated that the catalytic activity of SHP2 is involved in the activation of Ras and suggested the existence of a PI3K/SHP2 pathway leading to Ras stimulation because SHP2 is recruited downstream of PI3K. We next attempted to identify a substrate of SHP2 involved in the activation of Ras using the SHP2-C/S mutant in a "substrate trapping" experiment. Cells transfected with SHP2-C/S, or wild type SHP2 as a control, were stimulated with EGF, subjected to SHP2 immunoprecipitation, then anti-phosphotyrosine immunoblotting. As shown in Fig. 5D, both wild type SHP2 and SHP2-C/S coimmunoprecipitated with three phosphoproteins of apparent molecular masses of around 180, 115, and 100 kDa, whereas only the SHP2-C/S mutant associated with a protein of approximately 135 kDa. Upon reblotting, the 180-kDa protein comigrated with the EGFR and the 115-kDa hyperphosphorylated protein comigrated with the docking protein Gab1 (data not shown). We attempted to identify the ϳ135-kDa protein using antibodies directed against Ras effectors, including Ras-GTPase-activating protein and Sos2, but we failed to label this protein (data not shown).
These results also suggested that the docking protein Gab1 could mediate the PI3K-dependent recruitment of SHP2 and the subsequent activation of Ras. Gab1 sequence displays three binding sites for p85 and is thought to mediate the activation of PI3K during EGF stimulation (34). In addition, Gab1 contains one binding site for SHP2 and a PIP 3 -specific PH domain (35,36). However, the role of this PH domain in EGF signaling is not clear (37), and it is not known if the interaction of Gab1 with SHP2 requires PI3K. As shown in Fig. 6 (A and B), EGF induced the coimmunoprecipitation of Gab1 with SHP2, and this interaction was reduced by PI3K inhibitors. In addition, PI3K inhibitors nearly abolished the tyrosine phosphorylation of Gab1 induced by EGF (Fig. 6B). These data strongly suggested that Gab1 mediates the PI3K-dependent recruitment of SHP2. However, it is not known if Gab1 is important for Ras activation. To examine this question, we have produced Gab1 mutants deficient for binding to p85 or SHP2 and tested their effect on the activation of HA-Ras in cotransfection experiments. As shown in Fig. 6C, expression of Gab1-YF3 lacking the three p85 binding sites strongly inhibited the binding of HA-Ras to GST-RBD, and transfection of Gab1-Y627F deficient for SHP2 binding also blocked this association. This indicated that interaction of Gab1 with p85 and SHP2 is required for Ras activation. To determine whether Gab1 mutants blocked the pathway upstream or downstream of PI3K, we have studied their effect on the activation of HA-Akt/PKB. As shown in Fig.  6D, this activation was not impaired in cells transfected with Gab1-Y627F in comparison with cells expressing wild type Gab1. In contrast, activation of HA-Akt/PKB was strongly inhibited when Gab1-YF3 was expressed. In addition, Gab1-Y627F did not bind less p85 than wild type Gab1 in coimmunoprecipitation experiments (Fig. 6E). This demonstrated that Gab1-YF3 blocked the activation of Ras upstream of PI3K, whereas Gab1-Y627F interfered with Ras activation without preventing PI3K activation. Altogether, these results show that, in addition to Shc and Grb2, a PI3K-dependent pathway involving Gab1 and SHP2 participates to the activation of Ras under EGF stimulation.
These data also suggested that the major function of PIP 3 in the activation of Ras is to promote the recruitment of Gab1 in EGF signaling, leading to the recruitment of SHP2. However, Gab1 can associate to the activated EGFR through Grb2 (34) and directly through the Met-binding domain (MBD), which constitutes a novel phosphotyrosine-binding motif (36,38). We have thus further examined the role of PIP 3 in Gab1 recruitment by preparing a mutant deleted of the PH domain (Gab1-⌬PH). As expected, this mutant has lost the ability to associate with membrane fractions in a PI3K-dependent manner (Fig.  7A). We next studied its involvement in EGF signaling. As shown in Fig. 7B, the phosphorylation of Gab1-⌬PH and its coimmunoprecipitation with SHP2 were reduced in comparison to Gab1-wt, and these events were insensitive to wortmannin. This indicates that the PH domain is important for the recruitment of Gab1 in EGF signaling. However, it is not clear if the interaction between the PH domain and PIP 3 simply stabilizes the association of Gab1 with the EGFR, or whether PIP 3 promotes the recruitment of additional Gab1 molecules in the vicinity of the receptor.
As a first approach to answer this question, we have studied the ability of Gab1-⌬PH to interfere with EGF signaling. This was achieved by measuring MAPK activation in cotransfection experiments. Fig. 7C shows that Gab1-⌬PH did not significantly modify Erk1-His activation. As a control, transfection of FIG. 4. Role of PI3K in the activation of Ras. A, Cells treated with wortmannin (W) or LY294002 (LY) when indicated were stimulated with EGF then lysed. Cleared lysates were incubated with glutathione-Sepharose beads bound to a GST fusion protein containing the Rasbinding domain of Raf (GST-RBD). The beads were then washed and proteins resolved by SDS-PAGE. The amount of activated endogenous Ras associated to the GST-RBD beads was determined by anti-Ras immunoblotting (upper panel). Cell lysates were also directly subjected to anti-Ras immunoblotting to verify that equal amounts of Ras were present in each sample (lower panel). As a control, the pull-down assay was also performed from cells transfected with constitutively activated RasV12 (right lane). B, cells were cotransfected with HA-tagged Ras and the indicated constructs: empty vector (V); dominant negative p85 (⌬p85); catalytically inactive p110␤ (␤K805R). Following stimulation, cells were lysed and incubated with GST-RBD beads. The amount of HA-Ras associated with the beads was determined by anti-HA immunoblotting. The immunoblots shown are representative of at least two independent experiments. C, membrane fractions from control or EGFtreated cells were prepared by ultracentrifugation and analyzed by anti-Sos immunoblotting (upper panel), followed by anti-EGFR immunoblotting to verify that an equal amount of fractions were loaded on the gel (lower panel). D, cells were cotransfected with ERK1-His and a construct encoding wild type Sos or empty vector (V) as indicated. Following treatment or not with wortmannin (W) and stimulation with EGF as indicated, measurements of ERK1-His activation were performed. *, different than empty vector; p Ͻ 0.05, paired t test. Right panel, cell lysates were subjected to Sos immunoblotting to verify expression of transfected Sos. the Gab1-Y627F mutant that inhibited the activation of Ras (Fig. 6C) was found to reduce by more than 70% MAPK activation (Fig. 7C). Because Gab1-⌬PH is less phosphorylated than Gab1-wt (Fig. 7B), the fact that Gab1-⌬PH did not inhibit EGF signaling suggested that this mutant, albeit overexpressed, could not prevent the recruitment of endogenous wild type (wt) Gab1. To test this hypothesis, we have determined whether overexpression of Gab1-⌬PH interferes or not with the phosphorylation of Gab1-wt induced by EGF. This was achieved by studying the phosphorylation of HA-tagged Gab1-wt cotransfected with Gab1-⌬PH in a 1:20 ratio. As shown in Fig. 7D, overexpression of Gab1-⌬PH did not modify the phosphorylation of HA-Gab1-wt. As a control, overexpression of other Gab1 constructs (Gab1-Myc-wt or Gab1-Myc-Y627F) strongly inhibited the phosphorylation of HA-Gab1-wt. These results indicate that Gab1-⌬PH cannot prevent the phosphorylation of Gab1-wt by the EGFR, suggesting that PIP 3 alone is sufficient to recruit Gab1 in the vicinity of the receptor. DISCUSSION Although the mechanisms involved in growth factor-induced activation of MAPK do not display an obvious role for PI3K, pharmacological inhibitors of PI3K were found to strongly interfere with MAPK activation in Vero cells stimulated with EGF. In agreement with this, expression of mutants of class Ia PI3K inhibited the activation of MAPK to an extent similar to their effect on the activation of Akt/PKB, a major effector of PI3K. Moreover, the requirement for PI3K was observed under strong activation and independently of EGF concentration, which indicated that this enzyme can play an important function in the mechanisms leading to MAPK activation.
We have first studied the role of PI3K lipid products in this pathway, because it has been reported that only the protein kinase activity of p110␥ is necessary for MAPK activation (13). We have shown that overexpression of two proteins interfering with PIP 3 impaired MAPK activation, suggesting that, in contrast to p110␥, the lipid kinase activity of class Ia PI3Ks is essential for MAPK signaling. PTEN has already been shown to interfere with growth factor-induced MAPK activation, but it has been proposed that this effect was due to its ability to dephosphorylate proteins potentially involved in MAPK signaling, including the Fak kinase and Shc (39,40). We have thus used as a control the G129E mutant of PTEN, which retains the catalytic activity but has lost its ability to interact with phosphoinositides and, consequently, does not interfere with PI3K signaling (25,31). In our model, this mutant was not active on MAPK signaling, suggesting that the lipid phosphatase activity of PTEN is primarily responsible for MAPK inhibition. In agreement with this, overexpression of the PH domain of Tec also inhibited MAPK, whereas mutation of an amino acid residue involved in PIP 3 binding produced an inactive protein.
Based on biochemical approaches, multiple reports have suggested an involvement of PI3K in MAPK activation downstream of Ras, considering that the activation of Raf or MEK FIG. 5. SHP2 is recruited downstream of PI3K and is essential for Ras activation. A, anti-Grb2 immunoprecipitation was performed from cells treated with PI3K inhibitors when indicated (W, wortmannin; LY, LY294002) and stimulated, or not (Ctrl), with EGF. Immunoprecipitates were analyzed by immunoblotting with the indicated antibodies. As a control, the immunoprecipitation was performed without adding Grb2 antibody (ϪAb). B, anti-phosphotyrosine immunoprecipitations were performed from cells treated with PI3K inhibitors when indicated and stimulated with EGF. Immunoprecipitates were then analyzed by anti-SHP2 (upper panel) and anti-EGFR (lower panel) immunoblotting. C, cells were cotransfected with HAtagged Ras and the indicated constructs: empty vector (V); catalytically inactive SHP2 (C/S); wild type (wt) SHP2. Cells were then stimulated and incubated with GST-RBD beads to measure the activation of HA-Ras. The amount of HA-Ras associated with the beads was determined by anti-HA immunoblotting (upper panel). As a control, one pull-down assay was performed from a lysate of stimulated cells incubated with GST alone (lane ϪRBD). Lower panel, lysates were subjected to anti-HA immunoblotting to verify that equal amounts of HA-Ras were present in each sample. D, cells were transfected with wild type (wt) or C/S SHP2 as indicated, before stimulation or not (ctrl) with EGF. The cells were then subjected to anti-SHP2 immunoprecipitation, followed by anti-phosphotyrosine (upper panel) and anti-SHP2 (lower panel) immunoblotting. The immunoblots shown in this figure are representative of at least two independent experiments performed in duplicate.
can be mediated by targets of PI3K signaling (19 -23). In contrast, we have observed using inhibitors and dominant negative mutants that PI3K has a function upstream of Ras during EGF stimulation. PI3K was also found to be necessary for the redistribution of Sos, but expression of constitutively activated PI3K was not sufficient to activate MAPK. This suggested that PIP 3 cannot directly induce the membrane translocation of Sos, although Sos contains a PH domain that has some affinity for this lipid (33). Nevertheless, we cannot exclude that PIP 3 participates in Sos redistribution by stabilizing its interaction with the plasma membrane. In addition, although PI3K does not seem to contribute to the formation of the complex between Grb2 and the EGFR, Shc, or Sos, we have observed that the recruitment of SHP2 in this complex was dependent on PI3K. Because the activation of Ras in response to EGF was also strongly dependent on SHP2, this suggested the existence of a PI3K/SHP2 pathway that is important, in addition to Shc and Grb2, for the activation of Ras.
The docking protein Gab1 was found to participate in the process at two different levels (Fig. 8). First, we have verified that Gab1 is effectively involved in PI3K activation during EGF stimulation, because expression of Gab1 deficient for p85 binding sites abolished the activation of Akt/PKB. Second, Gab1 mediates the recruitment of SHP2 downstream of PI3K and the subsequent activation of Ras. Indeed, the phosphorylation of Gab1 and its association with SHP2 induced by EGF were blocked by PI3K inhibitors, and disruption of its SHP2 binding site suppressed the stimulation of Ras without interfering with PI3K activation. In addition, we have observed that the PH domain is important for efficient phosphorylation of Gab1 and binding to SHP2. Therefore, these data support the notion that PIP 3 is essential for the recruitment of Gab1 in EGF signaling, as suggested by the fact that Gab1 contains a PH domain that binds specifically PIP 3 (35,36). Yet, one may wonder what precise role PIP 3 plays, because Gab1 can associate to the phosphorylated EGFR through the MBD and indirectly through Grb2 (34,36). PIP 3 could simply stabilize the association of Gab1 with the EGFR, leading to an enhanced phosphorylation of Gab1. However, the results obtained with Gab1-⌬PH suggest that PIP 3 alone could promote the recruitment of Gab1 molecules in the vicinity of the EGFR. Indeed, the observations that Gab1-⌬PH is phosphorylated upon EGF stimulation, albeit less strongly than Gab1-wt, and that this phosphorylation is independent of PI3K suggest that Gab1-⌬PH can compete with Gab1-wt for binding to the receptor through Grb2-or MBD-mediated interactions. Nevertheless, overexpression of Gab1-⌬PH does not have a dominant negative effect on EGF signaling, i.e. phosphorylation of Gab1-wt and MAPK activation. This indicates that Gab1-wt is effectively recruited to the receptor in cells overexpressing Gab1-⌬PH, through a mechanism for which Gab1-⌬PH cannot compete. This is most likely the interaction between the PH domain and PIP 3 . In any case, this question must be further investigated, by studying, for example, the recruitment of Gab1 mutants deleted of the MBD and the Grb2-binding regions.
An important question is the physiological significance of FIG. 6. Role of Gab1 in the recruitment of SHP2 and Ras activation. Vero cells were preincubated with PI3K inhibitors when indicated before stimulation with EGF. Cell lysates were then subjected to SHP2 (A) or Gab1 (B) immunoprecipitation.
Immunoprecipitates were analyzed by anti-Gab1, anti-SHP2, and antiphosphotyrosine immunoblotting as indicated. C, cells were cotransfected with HA-tagged Ras and the indicated construct: empty vector (V); wild type (wt) Gab1-Myc; Gab1-Myc mutated on its three p85 binding sites (YF3); Gab1-Myc mutated on its SHP2 binding site (Y627F). Following stimulation, cells were lysed and incubated with GST-RBD beads. The amount of HA-Ras precipitated with the beads was determined by anti-HA immunoblotting. Lysates were also directly subjected to anti-HA (middle panel) and anti-Myc immunoblotting (lower panel) to verify the expression of HA-Ras and Gab1-Myc constructs. D, cells were cotransfected with HA-Akt and the indicated Gab1 constructs. Following cell stimulation, activation of HA-Akt was determined using in vitro kinase assay. E, cells expressing the indicated Gab1-Myc constructs were treated or not with EGF and subjected to anti-Myc immunoprecipitation followed by immunoblotting analysis using the indicated antibodies. this PI3K-dependent pathway in regard to the canonical Shc/ Grb2/Sos pathway. Further studies are necessary to determine whether the PI3K-dependent pathway is cell-type specific or if it is more generally involved in MAPK activation. A recent study in Cos cells has shown that the role of PI3K in MAPK activation is limited to weak stimulations induced by low doses of EGF (18), suggesting that the Shc/Grb2/Sos pathway is prominent in this cell type. In contrast, in cells derived from SHP2 or Gab1 knockout mice, MAPK are activated by EGF at much lower levels than in control cells (41,42). It is not known if MAPK activation is also altered in p110␣ knockout mice that have profound proliferative defects (10), but these findings strongly support the notion that activation of Ras by EGF does not depend entirely on the Shc/Grb2 pathway in normal cells.
The next step in the understanding of this PI3K-dependent pathway will be brought by the identification of the SHP2 substrates involved in the activation of Ras. It has been proposed that SHP2 can function as an adaptor in PDGF signal-ing, because it can bind to both the receptor and the SH2 domains of Grb2 and therefore contributes to the recruitment of Grb2-Sos (43). However, we have found that expression of catalytically inactive SHP2 abolished Ras activation in EGFtreated Vero cells, which is consistent with the fact that catalytically inactive SHP-2 was shown to inhibit MAPK activation induced by EGF in other cell types (37,44). This suggests that SHP2 activates by dephosphorylation a protein promoting the translocation of Sos or down-regulates an inhibitor of Sos redistribution. A candidate could be the 135-kDa protein that is associated with catalytically inactive SHP-2 in our substrate trapping experiment. Another candidate could be an unidentified 90-kDa protein associated to Gab1 and for which phosphorylation is up-regulated in SHP2 knockout mice (41). Further efforts are required to identify these proteins. FIG. 7. Role of the PH domain in the recruitment of Gab1. A and B, cells were transfected with Gab1-Myc containing (wt) or not (⌬PH) the PH domain before preincubation with wortmannin (ϩW) and stimulation with EGF when indicated. Following cell lysis, membrane fractions were prepared by ultracentrifugation and analyzed by anti-Myc immunoblotting (A, upper panel), followed by anti-EGFR immunoblotting to control gel loading (A, lower panel). B, lysates were subjected to anti-Myc immunoprecipitation, followed by immunoblotting with the indicated antibodies. C, cells were cotransfected with ERK1-His and the following constructs: empty vector (V); wild type Gab1-Myc (wt); Gab1-Myc deleted of the PH domain (⌬PH); Gab1-Myc mutated on its SHP2 binding site (Y627F). After stimulation with EGF, ERK1-His was extracted, and its activation was measured using an in vitro kinase assay with MBP. Data shown represent the mean Ϯ S.E. from three independent experiments. D, cells were cotransfected with 0.1 g of DNA encoding HA-tagged wild type Gab1 and 2 g of the indicated Gab1-Myc constructs. After stimulation, cells were processed for anti-HA immunoprecipitation followed by immunoblotting with the indicated antibodies. Lysates were also directly subjected to anti-Myc immunoblotting to verify the expression level of the Gab1-Myc constructs (bottom panel) .   FIG. 8. Model outlining the roles of PI3K, Gab1, and SHP2 in the activation of Ras during EGF stimulation. The activated receptor autophosphorylates on tyrosine residues involved in binding to Grb2 and Gab1 MBD. This leads to the association of Gab1 with the receptor and its phosphorylation on SHP2 and p85 binding sites, which triggers the activation of PI3K and the production of PIP 3 . This lipid recruits additional Gab1 molecules in the vicinity of the EGFR through binding to Gab1 PH domain, leading to an increased recruitment of p85 and SHP2. Downstream of SHP2, the relocation of Sos to membranebound Ras is facilitated by dephosphorylation of an unidentified SHP2 substrate.