An Epidermal Growth Factor Receptor/Gab1 Signaling Pathway Is Required for Activation of Phosphoinositide 3-Kinase by Lysophosphatidic Acid*

Phosphoinositide 3-kinase (PI3K) has been shown to play an essential role in G protein-induced signaling even in non-myeloid cells where few agonists of G protein-coupled receptors are known to activate PI3K. We have identified adherent cell lines where lysophosphatidic acid (LPA) strongly and rapidly activates the accumulation of PI3K lipid products. The process is not modified by expression of a kinase-dead mutant of the Gβγ-responsive PI3K p110γ. In contrast, it is inhibited by genistein or expression of a dominant negative mutant of p85 and potentiated by overexpressing wild-type p110α or -β but not -γ. By using a specific chemical inhibitor of the epidermal growth factor receptor (EGFR) and expression of a dominant negative mutant, we have observed that recruitment of p85/p110 PI3Ks occurs through transactivation of the EGFR by LPA and downstream mobilization of the docking protein Gab1 that associates with p85 upon LPA stimulation. Finally, we show that LPA cannot activate PI3K in cell lines lacking the EGFR/Gab1 pathway, including cells that transactivate the PDGF receptor. Altogether, these results demonstrate that activation of PI3K by LPA is conditioned by the ability of LPA to transactivate an EGFR/Gab1 signaling pathway.

Phosphoinositide 3-kinase (PI3K) has been shown to play an essential role in G protein-induced signaling even in non-myeloid cells where few agonists of G protein-coupled receptors are known to activate PI3K. We have identified adherent cell lines where lysophosphatidic acid (LPA) strongly and rapidly activates the accumulation of PI3K lipid products. The process is not modified by expression of a kinase-dead mutant of the G␤␥responsive PI3K p110␥. In contrast, it is inhibited by genistein or expression of a dominant negative mutant of p85 and potentiated by overexpressing wild-type p110␣ or -␤ but not -␥. By using a specific chemical inhibitor of the epidermal growth factor receptor (EGFR) and expression of a dominant negative mutant, we have observed that recruitment of p85/p110 PI3Ks occurs through transactivation of the EGFR by LPA and downstream mobilization of the docking protein Gab1 that associates with p85 upon LPA stimulation. Finally, we show that LPA cannot activate PI3K in cell lines lacking the EGFR/Gab1 pathway, including cells that transactivate the PDGF receptor. Altogether, these results demonstrate that activation of PI3K by LPA is conditioned by the ability of LPA to transactivate an EGFR/ Gab1 signaling pathway.
One of the major discovery of the 1990s in proliferative signaling has been the emergence of the Ras/mitogen-activated protein kinase (MAPK) 1 cascade as the main pathway used by growth factors for non-hematopoietic cells. However, the early mechanisms of activation of this pathway by agonists of G protein-coupled receptors (GPCR) remain elusive (1). Intensive researches have recently focused on the identification of proteins that could make the link between G proteins and the Ras/MAPK cascade, leading to the identification of various protein tyrosine kinases involved in this process, such as Pyk2 (2), Src (3), receptor tyrosine kinases (RTKs) (4,5), and even Syk in myeloid-derived cells (6). However, the mechanisms of activation of these kinases by GPCR have not been elucidated, except for Pyk2 that is recruited by a G q -Ca 2ϩ -dependent pathway (2). On the other hand, phosphoinositide 3-kinase (PI3K) has been recently shown to play a major role in the early mechanisms of GPCR-mediated activation of the Ras/MAPK pathway (7)(8)(9)(10)(11). However, the fact that PI3K inhibitors interfere with GPCR-induced signaling even in non-hematopoietic cell lines is difficult to interpret since GPCRs agonists are not known to induce the synthesis of PI3K lipid products in these cells. In contrast, neutrophils for example produce large amounts of phosphatidylinositol 3,4-bisphosphate (PI3,4P 2 ) and phosphatidylinositol 3,4,5-trisphosphate (PIP 3 ) upon stimulation with GPCR agonists (12). Similarly, stimulation of cells by RTK agonists such as platelet-derived growth factor (PDGF) generate PI3K lipid products through well known pathways, and PI3K inhibitors interfere with RTK-mediated activation of the Ras/MAPK pathway, although the mechanisms have to be clarified (13). Thus, one important question still remains concerning the regulation of PI3K in GPCR-induced signaling in non-myeloid cells.
Recently, two groups have identified a novel isoform of PI3K, p110␥, which contains a pleckstrin homology domain in its N terminus region. Interestingly, p110␥ can be directly activated by G protein subunits (14,15), due both to a constitutive association with a p101 ␤␥-sensitive protein (15) and to a direct interaction with the ␤␥ complex (16). Although this isoform of PI3K plays a role in the activation of the Ras/MAPK pathway by G proteins (17)(18)(19), it is not clear yet whether this enzyme is involved in any GPCR-induced PIP 3 production in non-hematopoietic cell lines (15,17,18). On the other hand, p110␤ was recently found to play a role in GPCR-induced signaling and mitogenesis (20,21). Therefore, this led us to study the activation of PI3K by GPCR in adherent cell lines. We have measured PI3K lipid products upon stimulation with lysophosphatidic acid (LPA), and we have identified cell lines producing large amounts of PI3,4P 2 and PIP 3 upon treatment with a GPCR agonist. The mechanism does not seem to involve the p110␥ isoform of PI3K but recruits the p85/p110 isoforms through LPA-induced mobilization of the EGF receptor (EGFR) and subsequent engagement of the docking protein Gab1. Finally, we show that LPA cannot activate PI3K in various cell types lacking the EGFR/ Gab1 pathway, thereby demonstrating the pivotal role of this transactivation pathway for PI3K activation by LPA in nonmyeloid-derived cell lines.
Cell Culture and Transfection-Cos, Vero, Rat1, and IMR90 cells were maintained in DMEM supplemented with 10% fetal bovine serum and antibiotics. For B82 L cells, 10% dialyzed newborn calf serum was used. For transfection, cells were incubated 4 h with 10 l of Lipo-fectAMINE (Life Technologies, Inc.) and 2 g of plasmid DNA per ml of Opti-MEM (Life Technologies, Inc.). The transfection mixture was then replaced by DMEM supplemented with 10% serum for 24 h. Before stimulation, cells were serum-starved for 24 h.
Analysis of PI Polyphosphate-Cells grown in 10-cm plates were serum-starved for 24 h upon reaching 80 -90% confluence and then labeled for 5 h with 0.2 mCi of [ 32 P]H 3 PO 4 (Amersham Pharmacia Biotech) per ml in phosphate-free DMEM. Cells were then stimulated for the indicated time and washed once with ice-cold phosphate-buffered saline before addition of 3.75 ml of 2.4 M HCl solution. Then lipid extraction was performed as described previously (27). Briefly, lipids were solubilized by addition of 3 ml of chloroform and 4.5 ml of methanol followed by vortexing. After centrifugation, the lower phase containing the lipids was collected, and the upper phase was washed with 4.5 ml of chloroform. The lower phases were then combined and evaporated under nitrogen, and lipid extracts were solubilized in 250 l of chloroform/methanol (1/1, v/v) and first resolved by thin layer chromatography (TLC) using chloroform/acetone/methanol/acetic acid/water (80/30/26/24/14,v/v). The spots corresponding to PI4,5P 2 /PI3,4P 2 , and PIP 3 were then scraped off, deacylated by 20% methylamine, and analyzed by HPLC on a Whatman Partisphere 5 SAX column. For measurements of PI polyphosphate after transfection of Cos cells with dominant negative mutants, the results represent the mean Ϯ S.E. of three independent experiments. For each experiment, the inhibitory effect of the mutant has been normalized for the percentage of transfected cells. This was determined concurrently using ␤-galactosidase as reporter and following a standard procedure. The efficiency of transfection was routinely around 40%.
Immunoblotting, Immunoprecipitation, and GST Pull-down Experiments-Stimulations were carried out at 37°C in serum-free DMEM containing 20 mM Hepes. Cells were washed once with ice-cold phosphate-buffered saline before lysis. For immunoblotting of crude lysates, cells were scraped off in SDS-PAGE sample buffer and boiled 5 min, then resolved by SDS-PAGE, and analyzed by immunoblotting using an enzyme-linked chemiluminescence system (ECL, Amersham Pharmacia Biotech). Results obtained with anti-phospho-Akt, and Erk antibodies were quantified by densitometry. We have verified that results obtained with the anti-phospho-Erk antibody did not differ from in vitro kinase assays of immunoprecipitated HA-Erk1 (construct kindly provided by Dr. J. Pouyssegur) using myelin basic protein as a substrate. For immunoprecipitations, cells were scraped off in lysis buffer containing 150 mM NaCl, 20 mM Tris-HCl, pH 7.4, 1% Brij (Sigma catalog number P9641), 1 mM Na 3 VO 4 , and 10 g/ml aprotinin and leupeptin. After gentle shaking during 20 min at 4°C and centrifugation (13,000 rpm for 10 min), the supernatants were incubated 1 h with antibodies followed by addition of 10% (w/v) protein A-Sepharose CL4B (Sigma) for 1 h. The immunecomplexes were washed twice with 1 ml of lysis buffer containing 0.1% Brij, 100 M Na 3 VO 4 , and 1 g/ml aprotinin and leupeptin and finally boiled in SDS-PAGE sample buffer. For GST pull-down experiments, cells were processed similarly to the immunoprecipitation protocol, except that cells were incubated 1 h with 2 g of GST fusion protein immobilized on glutathione-Sepharose 4B beads (Amersham Pharmacia Biotech).

LPA Rapidly Activates PI3K Independently of p110␥ but
Recruits p85/p110 Isoforms-To gain insight into the regulation of PI3K by LPA, we have measured the amount of PI3K lipid products in serum-starved cells stimulated with LPA. 1 M LPA induced an important accumulation of PI3,4P 2 and PIP 3 in both Vero and Cos cells, at a level up to 10-fold higher than control, with a maximum after 2 min stimulation (Fig. 1,  A and B). Dose-response assays indicated that the effect was detectable with as low as 0.1 M LPA and increased up to 10 M LPA (Fig. 1C). In contrast, another PI3K product, phosphatidylinositol 3-monophosphate, was found in relative abundance in resting cells and poorly accumulated upon LPA stimulation (counts in resting, 5600 Ϯ 1980; stimulated 2 min, 8850 Ϯ 1280). To elucidate the mechanism of PI3K activation by LPA, we have first studied p110␥ which can be directly activated by G protein subunits. The accumulation of PI3,4P 2 and PIP 3 upon LPA was measured in Cos cells transiently transfected with a kinase-dead mutant of p110␥ (K832R) that inhibits the G␤␥-induced activation of the Ras/MAPK pathway (Fig. 2B). Expression of this mutant somewhat reduced the number of cells retrieved 2 days after transfection, leading to a small decrease of extracted lipids including PI4,5P 2 ( Fig. 2A). However, upon stimulation by LPA, PIP 3 levels were only moderately lowered by expression of kinase-dead p110␥, even after normalizing the results for the percentage of transfection. Thus, the inhibition on PIP 3 production appeared similar to the inhibitory effect on PI4,5P 2 . Moreover, PI3,4P 2 production was not modified in Cos cells expressing the K832R p110␥ mutant. In contrast, the accumulation of PI3,4P 2 and PIP 3 induced by 10 M LPA was nearly abolished in Cos cells treated with genistein ( Fig. 2C), thereby suggesting the implication of p85/ p110 PI3Ks. To evaluate this hypothesis, we have measured PI3K activation by LPA in Cos cells transfected with a dominant negative form of p85␣ lacking the p110-binding site (⌬p85) (23). After normalization of the results for the percent-age of transfection, we have observed that expression of ⌬p85 nearly abolished the accumulation of PI3,4P 2 and PIP 3 induced by 10 M LPA (Fig. 2D). In addition, we have determined whether overexpression of wild-type p110␣, -␤, or -␥ had any potentiating effect on PI3K activation triggered by a submaximal dose of LPA (1 M). Analysis of PI3K stimulation limited for convenience to measurements of PI3,4P 2 showed that p110␣ and p110␤ to a greater extent potentiated PI3K activation by LPA, whereas overexpression of p110␥ had no effect (Fig. 2E). Altogether, these results demonstrate that the activation of PI3K by LPA is mediated by the tyrosine kinase-dependent p85/p110 isoforms of PI3K.
Activation of PI3K by LPA Occurs through Recruitment of the EGF Receptor-To identify the tyrosine kinase(s) involved in this process, we have searched for major tyrosine-phosphorylated proteins in crude lysates from cells treated 2 min with LPA. The major phosphotyrosine signal induced by LPA in Vero or Cos cells was found in the 180-kDa region (Fig. 3A), suggesting an involvement of the two major RTKs expressed in fibroblasts, i.e. the receptors for PDGF and EGF. First, we have determined that the PDGFR-specific inhibitor tyrphostin AG1296 had no effect on LPA-induced tyrosine phosphorylation of the 180-kDa protein, whereas the EGFR inhibitor AG1478 nearly abolished the signal in both Vero and Cos cells (Fig. 3A). In addition, anti-EGFR immunoblotting of cell lysates confirmed that the 180-kDa protein colocalized with EGFR, whereas an antibody against ␤PDGFR gave no signal in Vero cells and a very faint band in Cos cells (Fig. 3B). Finally, following immunoprecipitation, the EGFR was found to be tyrosine-phosphorylated after treatment of Cos cells with LPA (Fig. 3C). Thus, these data demonstrate that phosphorylation of the EGFR is the major early tyrosine phosphorylation event induced by LPA in Cos and Vero cells.
We have then determined whether this event was important for PI3K recruitment by LPA. In both Cos and Vero cells, the accumulation of PI3,4P 2 and PIP 3 induced by LPA was dramatically inhibited by the EGFR inhibitor AG1478 (Fig. 3D). This compound also abolished the activation of Akt by LPA but did not interfere with Akt stimulation by insulin in Vero cells (not shown). In addition, AG1478 did not alter PI3K activity itself in Rat1 cells stimulated with PDGF (counts of PI3,4P 2 , control Ͻ200; PDGF 9255 Ϯ 475; PDGF ϩAG1478 100 nM, 8728 Ϯ 836). Furthermore, Cos cells were transfected with an EGFR mutant truncated at amino acid 688 (EGFRc688) which has a strong ability to dimerize upon activation. 2 This mutant inhibited by over 70% the LPA-induced accumulation of PI3,4P 2 and PIP 3 after normalizing the results for the percentage of transfection (Fig. 3E). Although these results demonstrate that the EGFR plays a crucial role in LPA-induced PI3K activation, EGF is not a typical agonist of PI3K, and PIP 3 levels are insensitive to EGF in various cell types. Nevertheless, we have observed that 10 ng/ml EGF induced an important accumulation of PI3,4P 2 and PIP 3 in Cos cells (Fig. 3F). In addition, the increase in PI3K lipid products occurred earlier upon EGF stimulation (Fig. 3F) than in the presence of LPA (Fig. 1B), which is compatible with a recruitment of PI3K by LPA occurring downstream the EGFR. Finally, expression of ⌬p85 in Cos cells suppressed the synthesis of PI3K lipid products induced by EGF to an extent similar to that upon LPA stimulation (Figs. 3G and 2D). This suggested that both LPA and EGF use a same p85-dependent pathway to stimulate PI3,4P 2 and PIP 3 production.
The EGFR-dependent Activation of PI3K by LPA Mobilizes Gab1-Since the activation of PI3K by EGF seems to differ from one cell type to another, we have studied various pathways possibly involved in the EGFR-dependent activation of PI3K by LPA. Using pull-down experiments with GST-p85 fusion protein, the EGFR and p85 were found to coprecipitate upon stimulation with LPA (Fig. 4A), thereby corroborating that p85 is recruited by LPA through an EGFR-dependent pathway. Although one of the major mechanisms of EGFRmediated recruitment of p85 is the heterodimerization of the EGFR with ErbB3 (28), we did not find any LPA-or EGFinduced association of ErbB3 in p85 immunoprecipitates or GST-p85 pull-downs (not shown). To identify other candidates possibly involved in recruitment of p85, we have looked for tyrosine-phosphorylated proteins in GST-p85 pull-downs and p85 immunoprecipitates from LPA-or EGF-treated cells. The major phosphotyrosine signal appearing upon stimulation was located close to the 115-kDa marker (Fig. 4B). This molecular mass led us to consider the adaptor protein Gab1 as a candidate. By performing both p85 immunoprecipitates and GST-p85 pull-downs, Gab1 was found to associate with p85 following cell stimulation with LPA or EGF (Fig. 4B). To confirm this observation, we have performed anti-Gab1 immunoprecipitates, and p85 was found to coprecipitate with Gab1 upon stimulation with LPA or EGF (Fig. 4C). Finally, the EGFR inhibitor AG1478 was found to abolish the association of Gab1 with p85 (Fig. 4D). Altogether, these results demonstrate that activation of PI3K by LPA occurs through an EGFR/Gab1 pathway.
The EGFR/Gab1 Pathway Is Essential to PI3K Activation by LPA-To determine whether LPA could activate PI3K using other mechanisms, we have first studied IMR90 human fibroblasts where activation of MAPK by LPA is independent of RTK activities (Fig. 5A). Although LPA and PDGF stimulated Erk to similar levels in these cells, LPA produced only a minor accumulation of PI3K lipid products, whereas they readily accumulated upon treatment with PDGF (Fig. 5A), suggesting that specific RTK transactivation is required for activation of PI3K by LPA. To determine if RTKs other than the EGFR could participate in the process, we have studied mouse B82 L fibroblasts that do not express the EGFR and where transactivation of the PDGFR is required for Erk activation by LPA (Fig. 5B) (29). In these cells, LPA produced only a faint accumulation of PI3K lipid products that readily accumulated upon PDGF, 2 K. Tanner, J. Kyte, and G. Gill, manuscript in preparation.

FIG. 4. LPA induces the association of p85 with Gab1 in an EGFR-dependent fashion.
A, quiescent Cos cells were stimulated 2 min with 10 M LPA or 10 ng/ml EGF and then lysed and incubated with 2 g of Sepharose-GST-p85 fusion protein for GST-p85 pull-down (PD) assays. Precipitated proteins were analyzed by imunoblotting (IB) using an anti-EGFR antibody. B, left, cells were stimulated as above and then processed for GST pull-down assays, followed by anti-phosphotyrosine (pY) and Gab1 immunoblotting. Right, anti-p85 immunoprecipitates (IP) from control or stimulated cells were immunoblotted with the indicated antibodies (NRS, normal rabbit serum). C, Gab1 was immunoprecipitated from control (Ctrl) or stimulated cells and then the precipitated proteins were revealed with the indicated antibodies. D, before stimulation, cells were incubated with AG1478 when indicated and then association of Gab1 with p85 was analyzed using GST-p85 pull-down experiments.
whereas both growth factors activated Erk to comparable levels (Fig. 5B).
Finally, in Rat1 cells where transactivation of the EGFR is required for Erk activation by LPA similarly to Vero cells (4,29), we have observed that PI3,4P 2 and PIP 3 were hardly detectable in LPA-treated cells (Fig. 6A). As a control, PDGF induced a massive accumulation of PI3K lipid products, and both LPA and PDGF activated Erk to comparable levels. To gain insight about the missing mechanism in Rat1 cells, we have also measured PI3K activation upon stimulation with EGF. Interestingly, levels of PI3,4P 2 and PIP 3 were not modified by EGF, although Erk activation by EGF was comparable to the PDGF response (Fig. 6A). This suggested that the EGFRdependent pathway of PI3K activation present in Vero and Cos cells was deficient in Rat1 cells. By using GST-p85 pull-downs assays, we have observed that Gab1 did not associate with p85 in Rat1 cells stimulated with LPA or EGF (Fig. 6B). Therefore, we have compared the recruitment of Gab1 in Cos and Rat1 cells. By immunoblotting lysates of Cos cells stimulated with EGF, Gab1 was found to undergo a shift in its apparent molecular weight that is typical of this adaptor (30) (Fig. 6C, top). In contrast, Gab1 migration was hardly modified in Rat1 cells stimulated with EGF. Similarly, following immunoprecipitation, we have found that Gab1 was not tyrosine-phosphorylated, and its migration was unchanged in LPA-or EGFtreated Rat1 cells (Fig. 6C, bottom). DISCUSSION The recent discovery of p110␥ that can be directly activated by G protein subunits (14) led us first to evaluate the role of this enzyme in LPA-mediated activation of PI3K. However, our results demonstrate that the synthesis of PI3K lipid products induced by LPA occurs independently of p110␥, based on the lack of inhibitory effect of a kinase-dead mutant and supported by the non-potentiating effect of the overexpressed wild-type enzyme. This observation was rather surprising in light of recent well documented reports showing that p110␥ is involved in G␤␥-induced signaling, such as activation of the Ras/MAPK pathway (17,31). However, we have investigated the natural signaling resources of cells stimulated by endogenous LPA receptor(s), whereas the studies by Lopez-Ilasaca and co-workers (17,31) were based mainly on transient expression of G␤␥ subunits and cotransfection of effectors. Substantial differences might exist between these two complementary models in terms of recruitment of effectors, such as G protein subunits for example. In addition, although p110␥ is not apparently involved in generation of PI3K lipid products, a recent report demonstrated that the protein kinase activity of p110␥ is essential for MAPK activation (18), in agreement with our observation that a catalytically dead mutant of p110␥ moderately inhibited cell growth without influencing PIP 3 levels. However, it is also important to consider that the expression of p110␥ in fibroblasts is marginal in comparison to blood platelets, 3  type where activation of PI3K by the thrombin GPCR has been reported to engage p110␥ (32,33). Similarly, in neutrophils where p110␥ is readily expressed (15), wortmannin has been shown to inhibit GPCR-induced signaling independently of p85 (34). Therefore, the facts that p110␥ seems preferentially expressed in myeloid-derived cells and the inability of most adherent cells to produce PIP 3 upon GPCR agonists (12) suggest that the role of p110␥ in GPCR-induced signaling could be restricted to hematopoietic cells.
In contrast, we have found that adherent cells stimulated with LPA recruit phosphotyrosine-dependent PI3Ks, in agreement with our recent report showing a major role for p110␤ in LPA-induced mitogenesis of fibroblast cells (21). Interestingly, this isoform can be activated synergistically by G␤␥ subunits and phosphotyrosyl peptides (20), which may account for a more pronounced effect of overexpressing p110␤ than p110␣ on PI3,4P 2 production induced by LPA. In addition, we have shown that engagement of p85/p110 PI3K by LPA occurs mainly through transactivation of an EGFR/Gab1-signaling pathway. Although the GPCR-induced tyrosine phosphorylation of the EGFR has been described as a docking effect for downstream effectors (5), it has been shown using AG1478 that the kinase activity of the EGFR is also required for LPAinduced activation of the Ras/MAPK pathway (4). Here we show that inhibition of EGFR activity blocks PI3K activation by LPA, indicating that transactivation of the EGFR is an essential step for crucial events in LPA-induced signaling, including activation of MAPK and PI3K.
In addition, we did not find a significant activation of PI3K in cells where LPA induces transactivation of the PDGFR, one of the best activators of PI3K when activated by its ligand. These results demonstrate that important differences exist between stimulation of RTKs by their ligands and GPCR-induced transactivation, although both pathways lead to similar levels of activation of Erk1/2. The differences are most likely due to the strength of early signaling events since, for example, tyrosine phosphorylation of RTKs mediated by GPCRs is much weaker than phosphorylation induced by RTK ligands 3 (29,35). Thus, the LPA-induced recruitment of PDGFR might be sufficient to fully activate Erk1/2 in B82 L cells but not the PI3K. The difference of sensitivity between activation of these kinases could be due to the fact that Erk1 and Erk2 are stimulated through an amplification cascade, whereas PI3K is activated directly by the receptor.
Various mechanisms have been proposed regarding the activation of PI3K by EGFR which lacks the YXXM p85-binding motif found in the PDGFR sequence for example. One of the best documented mechanisms is the EGF-induced dimerization of EGFR with a related protein, ErbB3, that contains several YXXM motifs (28). However, we did not find any association of ErbB3 with p85 in LPA or EGF-treated cells, in conditions where the EGFR and p85 were found to form a complex. A similar mechanism of RTK heterodimerization involving association of EGFR with ␤PDGFR has been recently described (36). However, this possibility can be excluded in the case of the events described herein, based on the very weak expression of ␤PDGFR in Cos cells that we and others have observed (29,37), as well as their very weak responsiveness to ␤PDGF in terms of activation of PI3K and MAPK (not shown) and the undetectable expression of ␤PDGFR in Vero cells. Recently, a novel docking protein associated to Grb2, Gab1, has been shown to mediate PI3K activation by various RTKs (30), through an interaction YXXM/SH2 domains of p85 (38). Our data demonstrate that this adaptor is also involved in the EGFR-dependent activation of PI3K by LPA, since it is the major tyrosinephosphorylated protein associated with p85 in LPA-treated cells. In addition, we have observed that Gab1 was indispensable to the EGFR-dependent activation of PI3K by LPA. Indeed, in Rat1 cells, neither LPA which transactivates EGFR nor EGF itself are able to recruit p85, due to a poor recruitment of Gab1 by EGFR in this cell type. The reason for this observation remains undetermined but could be due to a fewer number of EGFR molecules in Rat1 than in Cos cells 3 that is sufficient to activate the Ras/MAPK pathway but not to significantly phosphorylate Gab1. Alternatively, the EGFR-dependent phosphorylation of Gab1 might require an intermediate protein tyrosine kinase poorly expressed in Rat1. Nevertheless, these data obtained in Rat1 cells further confirm the pivotal role of the EGFR/Gab1 pathway for activation of PI3K by LPA in non-myeloid-derived cell lines.
Although the mechanism of GPCR-induced transactivation of RTKs remains completely unknown, one possibility could be the secretion of EGF induced by LPA, in light of the dependence of transactivation processes on calcium and protein kinase C (39,40) that are crucial factors for secretion. Although EGF has not been found in conditioned medium of cells treated with GPCR agonists (40,41), one cannot exclude that secreted EGF would remain cell-associated and work in an autocrine fashion, as described for the fibroblast growth factor (42). However, our study shows that activation of PI3K can be measured as early as 30 s after adding LPA, which seems hardly compatible with a process involving the whole cellular machinery for secretion. In contrast, a scavenger of reactive oxygen species partly inhibited PI3K activation by LPA, due to an inhibitory effect on tyrosine phosphorylation of the EGF, 3 as described recently in HeLa cells (43). This suggests that in unstimulated cells, protein tyrosine phosphatases involved in maintaining RTK in an inactivated state could play a role in the mechanism of transactivation.