A Function for Phosphoinositide 3-Kinase (cid:1) Lipid Products in Coupling (cid:1)(cid:2) to Ras Activation in Response to Lysophosphatidic Acid*

Although G (cid:1)(cid:2) is thought to mediate mitogen-acti-vated protein kinase (MAPK) activation in response to G protein-coupled receptor stimulation, the mechanisms involved in this pathway have not been clearly defined. Phosphoinositide 3-kinase (PI3K) has been proposed as an early intermediate in this process, but its role has remained elusive. We have observed that dominant negative mutants of p110 (cid:1) , but not of p110 (cid:2) , inhibited MAPK stimulation in response to lysophosphatidic acid (LPA). The role of p110 (cid:1) was located upstream from Ras. To determine which of the lipid or protein kinase activi- ties of p110 (cid:1) were important for Ras activation, we produced a mutant p110 (cid:1) lacking the lipid but not the protein kinase activity. This protein displayed a dominant negative activity similar to a kinase-dead mutant, indicating 50 m M Tris, pH 7.5, 10 m M MgCl , 1 m M dithiothreitol. The reaction was performed in 25 (cid:6) l of kinase buffer containing 10 (cid:6) g of histone 2B (Roche Molecular Biochemicals), (cid:6) M ATP, and 3 (cid:6) Ci of (cid:2) - 32 P]ATP. reaction was incubated during 30 min at 25 °C, stopped by of and Phosphorylation phorImager then subjected followed by lipid or protein kinase perform kinase were then twice in the following lipid kinase m M pH m M 5 m M , m M dithiothreitol, m M were then resuspended (cid:6) l of lipid kinase buffer supple- mented with 150 (cid:6) M phosphatidylserine, 75 (cid:6) M phosphatidylinositol, 20 (cid:6) M ATP and 5 (cid:6) Ci of [ (cid:2) - 32 P]ATP. The reaction was performed 37 °C for 30 min, then stopped by 100 (cid:6) l of HCl (1.5 N ), followed by lipid extraction (40). Lipids were then separated by thin layer chroma- tography and using a the protein kinase assay, immunoprecipitation pellets M

Lysophosphatidic acid (LPA) 1 is an intercellular lipid mediator potentially involved in tissue regeneration, brain develop-ment, tumorigenesis and atherosclerosis, although its precise physiopathological role in vivo remains to be explored (1)(2)(3)(4)(5). LPA is produced by activated cells, notably platelets, and promotes the proliferation or survival of a large number of cell types, similarly to a canonical growth factor. It is now well accepted that the biological activity of LPA is mediated by at least three different G protein-coupled receptors (GPCR), namely LPA 1 , LPA 2 , and LPA 3 (Edg 2, 4 and 7, respectively) which belong to a recently described family of receptors to bioactive lysophospholipids. These receptors can activate in concert the G q , G i and G 12/13 subfamilies of G proteins, even though the coupling specificity between each receptor and the different G proteins remains to be clearly established (6 -8).
A number of studies have started to delineate the intracellular signaling pathways that mediate the biological activity of LPA. This lipid was first found to activate the so-called Ras/ mitogen-activated protein kinases (MAPK) pathway, which controls cell proliferation, differentiation and survival in response to numerous extracellular stimuli (see Refs. 1, 9 for recent reviews). Nevertheless, the early mechanisms involved in MAPK activation in response to LPA are still incompletely defined. LPA can potentially activate this pathway through G q -dependent Ca 2ϩ mobilization leading to stimulation of protein kinase C (PKC) that in turn can activate Raf or MEK (MAP kinase/ERK kinase) (10). However, pertussis toxin that inactivates G i was found to inhibit MAPK stimulation induced by LPA or other G i -coupled receptor agonists (9). In addition, the role of G i was located upstream from Ras, and the involvement of ␤␥ subunits has been well documented (11,12). In contrast, the molecular intermediates between ␤␥ and Ras have not been clearly identified (for recent reviews, see (9,13)). Since Ras is classically activated downstream of tyrosine kinases, various candidates, including Src, Pyk2 and growth factor receptors were shown to mediate MAPK activation in response to GPCR stimulation (14 -17). However, there is little evidence for a participation of ␤␥ in the activation of these tyrosine kinases, and, in the case of Pyk2 or EGF receptor (EGFR), their recruitment has been shown to occur through G q -dependent pathways (14,18,19). Nevertheless, it has been proposed that ␤␥ could participate in Src activation, leading to EGFR phosphorylation and MAPK stimulation (20). G protein-coupled receptor kinases and ␤-arrestins might be intermediates between ␤␥ and Src (21), or even actors of the MAPK pathway (22), but these models have been essentially described in cells overexpressing adrenergic receptors and their relevance to LPA signaling remains to be established.
Besides tyrosine kinases, phosphoinositide 3-kinase (PI3K) has been shown to be an early intermediate of MAPK activation in response to LPA (9,23). PI3K phosphorylates the 3Ј position of the inositol ring of phosphoinositides to produce lipids that are now considered as crucial spatio-temporal organizers of various signaling pathways. From a molecular point of view, three classes of PI3Ks have been defined, and mitogen signaling involves essentially class I enzymes that are subdivided into two subclasses (A and B) (24,25). Class I A PI3Ks include the catalytic subunits p110␣, ␤, and ␦ associated with a p85 regulatory subunit and activated by phosphotyrosine-containing motifs encountered on receptors or adapter proteins. In the case of p110␤, G protein ␤␥ subunits were shown to dramatically increase its activation by phosphotyrosine motifs, even though the molecular determinants of this synergism have to be further defined (26,27). Class I B is represented by the catalytic subunit p110␥ that has the unique feature to be directly activated by G protein ␤␥ subunits. p110␥ can be associated with a p101 regulatory subunit that has no homology with any known protein but contributes to the sensitivity of p110␥ to G␤␥ (28,29). From a mechanistic point of view, processes by which PI3Ks activate signaling pathways have been recently unraveled. Their lipid products interact with a number of signaling proteins, resulting in their membrane targeting and/or modulation of their enzyme activity. For example, phosphatidylinositol 3,4,5-trisphosphate (PtdIns(3,4,5) P 3 ) binds to a conserved protein motif called the pleckstrin homology (PH) domain, thereby inducing the activation of the serine/threonine kinase Akt/protein kinase B (PKB) and its upstream activators, the phosphoinositide-dependent kinases (30,31).
The hypothesis that PI3K might participate in the MAPK pathway has emerged when blockade of PI3K was found to interfere with MAPK activation in response to LPA (9,23). Further studies have suggested that p110␥ was a strong candidate to mediate this effect (32)(33)(34), and to date it is commonly accepted that p110␥ can link GPCR stimulation to MAPK activation (13). However, the mechanism of this connection has been incompletely characterized. Although it has been initially reported that p110␥ linked G␤␥ to MAPK activation through an Shc-Grb2-Sos-Ras pathway (32), the putative tyrosine kinase activated downstream of p110␥ remains to be identified. In addition, in a particular CHO cell line displaying a deficient activation of Ras, Takeda et al. (33) have observed that LPA can activate MAPK at the level of MEK through a pathway involving p110␥ upstream of PKC. PKC can be stimulated directly by PtdIns(3,4,5)P 3 or by the phosphoinositide-dependent kinases (35,36). However this model has been challenged by the demonstration that the protein kinase activity of p110␥, but not its lipid kinase activity, was important for MAPK activation (34). Indeed, PI3K catalytic subunits have an intrinsic protein kinase activity that was thought to simply regulate their lipid kinase activity (37,38), but, at least in the case of p110␥, the protein kinase activity was shown to be critical for MAPK activation by promoting MEK phosphorylation (34). However, the protein substrate of p110␥ remains to be identified. Further studies are certainly required to complete these different models and define their physiological relevance, but the fact that p110␥ seems essentially expressed in blood cells suggests that its role between G␤␥ and MAPK might be that of a tissue-specific function.
We and others have previously reported (39 -41) that LPA can activate p110␤ in different non-hematopoietic cell lines. The mechanism is thought to implicate cooperation between ␤␥ and a transactivation pathway involving the EGFR and the adapter protein Gab1 that provides the consensus phosphotyrosine motifs. Activation of p110␤ is essential for cell cycle progression of NIH-3T3 cells stimulated with LPA (39), which suggested that this PI3K was an important actor of the signaling pathways involved in the mitogenic activity of LPA. We have thus searched for a role of p110␤ in the MAPK pathway in a non-transformed cell line stimulated through endogenous LPA receptors. By using a strategy based on the transfection of dominant negative mutants, p110␤ was found to contribute to MAPK activation upstream from Ras. In addition, the lipid kinase activity of p110␤ was found to be essential, pointing out a key element of Ras activation that involves sequentially PtdIns(3,4,5)P 3 , the PH domain-containing adapter Gab1, and the tyrosine phosphatase SHP2. Moreover, we present data indicating that this pathway provides a novel connection between ␤␥ and Ras.

EXPERIMENTAL PROCEDURES
Materials-LPA, pertussis toxin, and wortmannin were from Sigma Chemical Co. LY294002 and AG1478 were from BIOMOL. Ro-31-8220 was from Calbiochem. Monoclonal anti-phosphoErk and anti-GST antibodies were from Sigma. Polyclonal antibodies against ERK2, Grb2, SHP2, and EGFR and monoclonal anti-myc were from Santa Cruz Biotechnology. Polyclonal antibody against Gab1 was from Upstate Biotechnology Inc. Polyclonal antibody against phosphoAkt/PKB (Ser-473) was from Cell Signaling. 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 Invitrogen.
Expression Plasmids-A construct encoding C-terminal Myc/Histagged ERK1 was obtained by subcloning ERK1 (kindly provided by Dr. E. vanObberghen, Nice, France) into pcDNA3.1-MycHis (Invitrogen). The pRK5 plasmid encoding wild type HA-tagged p110␤ was kindly provided by Drs. P. Hu and J. Schlessinger (New York University) (44). The kinase-inactive K805R mutant of p110␤ was obtained by site-directed mutagenesis (QuikChange, Stratagene) with the following mutagenic primer: 5Ј-GTTGGAGTGATTTTTAGAAATGGTGATGATT-TACG-3Ј (the changed nucleotide is underlined). p110␤ "protein kinase only" mutant was obtained by deleting amino acids 946 to 955 using the following mutagenic primer: 5Ј-ATTCTTGGAAATTTCGTGCCTTTTA-TTCTT-3Ј. The mutations were verified by sequencing, and the integrity of the encoded proteins was controlled by immunoblotting and kinase assays following expression in COS cells. A construct encoding full-length p110␤ in antisense orientation was obtained by excising p110␤ from the pRK5 construct using EcoRI and HindIII. The insert was then subcloned through the same sites into pHM6 (Roche Molecular Biochemicals, Meylan, France), and the construction was verified by restriction mapping. The plasmids encoding p110␥, ␤ARK, PTEN, and Gab1 mutants have been previously described (32,(45)(46)(47). The p85 and SHP2 constructs were kindly provided by Drs. W. Ogawa (University of Kobe, Japan) and N. Rivard (Sherbrooke University, Canada), respectively.
Cell Culture, Transfection, and Stimulations-Vero cells (ATCC CCL 81) were maintained in Dulbecco's modified Eagle 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 medium containing 2 g of total DNA and 6 l of each LipofectAMINE and Plus reagents (Invitrogen). Cells were incubated overnight in serum-free medium before stimulation with LPA (10 M, 5 min). When indicated, cells were incubated with 100 nM wortmannin or 25 M LY294002 15 min before stimulation.
Cell Lysis, Immunoprecipitation, and Immunoblotting-Cells were scrapped off in lysis buffer containing 20 mM Tris, pH 7.4, 150 mM NaCl, 1 mM EDTA, 10% glycerol, 1% Nonidet P-40, 10 g/ml of each aprotinin and leupeptin, 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 incubated with protein A-or p110␤ Links LPA and ␤␥ to Ras protein G-Sepharose (Sigma) for 1 h, then collected by centrifugation and washed with lysis buffer containing 0.1% Nonidet P-40, 1 g/ml of each aprotinin and leupeptin, 0.1 mM orthovanadate. Pellets were then processed for in vitro kinase assays or resuspended in electrophoresis sample buffer and analyzed by immunoblotting. Blots were developed using chemiluminescence (Amersham Biosciences, Inc.) and semi-quantified using a Bio-Rad gel analysis device and the software IMAGE (National Institutes of Health). For immunoblotting analysis of cell lysates, cells were directly scrapped off in electrophoresis sample buffer, then boiled and processed for immunoblotting.
Measurements of ERK Phosphorylation in Transfected Cells-Cells were transfected with 1 g of each plasmid 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) during 2 h at 4°C then washed three times with lysis buffer supplemented with 5 mM imidazole. The pellets were then resuspended in electrophoresis sample buffer and processed for anti-phosphoERK and anti-His immunoblotting.
In Vitro Akt/PKB Kinase Assay-Vero cells were cotransfected with 0.5 g of DNA encoding HA-tagged Akt/PKB and 1.5 g of plasmid encoding the indicated dominant negative protein. In the case of ⌬p85, only 1 g was used to avoid nonspecific effects, supplemented with 0.5 g of empty vector. After stimulation, cells were scrapped off in lysis buffer then subjected to anti-HA immunoprecipitation. Immunoprecipitates were washed twice with lysis buffer, then twice with kinase buffer containing 50 mM Tris, pH 7.5, 10 mM MgCl 2 , 1 mM dithiothreitol. The reaction was performed in 25 l of kinase buffer containing 10 g of histone 2B (Roche Molecular Biochemicals), 50 M ATP, and 3 Ci of [␥-32 P]ATP. The reaction was incubated during 30 min at 25°C, then stopped by addition of electrophoresis sample buffer and analyzed by SDS-PAGE. Phosphorylation of histones was quantified using a Phos-phorImager and the software ImageQuant (Molecular Dynamics).
In Vitro Lipid and Protein Kinase Assays-COS cells were transfected with the constructs encoding HA-tagged p110␤ mutants. Cells were then subjected to anti-HA immunoprecipitation, followed by lipid or protein kinase assays. To perform the lipid kinase assay, pellets were then washed twice in the following lipid kinase buffer: 50 mM Tris, pH 7.4, 100 mM NaCl, 5 mM MgCl 2 , 1.5 mM dithiothreitol, 0.5 mM EDTA. Pellets were then resuspended in 100 l of lipid kinase buffer supplemented with 150 M phosphatidylserine, 75 M phosphatidylinositol, 20 M ATP and 5 Ci of [␥-32 P]ATP. The reaction was performed at 37°C for 30 min, then stopped by adding 100 l of HCl (1.5 N), followed by lipid extraction (40). Lipids were then separated by thin layer chromatography and revealed using a PhosphorImager. To perform the protein kinase assay, immunoprecipitation pellets were washed twice in the following protein kinase buffer: 50 mM Tris, pH 7.4, 100 mM NaCl, 1 mM MnCl 2 . Pellets were then resuspended in 50 l of protein kinase buffer supplemented with 40 M ATP and 3 Ci of [␥-32 P]ATP. The reaction was incubated at 37°C for 30 min, then stopped by adding electrophoresis sample buffer and analyzed by SDS-PAGE. Autophosphorylated p110␤ was revealed using a PhosphorImager.
Activated Ras Affinity Precipitation Assay-The assay was performed essentially as described previously (43). The recombinant Rasbinding domain (RBD) of Raf1 (kindly provided by Dr. F. R. McKenzie, Nice, France) was expressed as a GST fusion protein in Escherichia coli and extracted using glutathione-Sepharose beads. To measure Ras activation in stimulated cells, 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% deoxycholate, 1% Nonidet P-40, 0.1% SDS, 1 mM phenylmethylsulfonyl fluoride, 1 g/ml of each 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 then boiled, and proteins were resolved by SDS-PAGE. Immunoblotting was performed with anti-pan Ras antibodies. To study Ras activation in transfected cells, the cells were cotransfected with 0.5 g of plasmid encoding HA-tagged wild type Ras (kindly provided by Dr. B. M. Burgering, Utrecht, The Netherlands) and 1.5 g of the indicated effector, unless otherwise indicated. 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 essentially as described (48). Cells were scrapped off in lysis buffer containing 20 mM Tris, pH 7.4, 150 mM NaCl, 1 mM EDTA, 10 mM EGTA, 2 mM MgCl 2 , 10 g/ml of each aprotinin and leupeptin, 1 mM orthovanadate, then Dounce-homogenized. The homogenate was centrifuged at 100,000 ϫ g for 1 h. The pellet was washed once in lysis buffer then dissolved in lysis buffer supplemented with 1% Triton X-100. The in-soluble material was spun out, and the supernatant was taken as the solubilized membrane fraction.
RT-PCR Analysis-This study was performed essentially as described (42). Total RNA were extracted from cultured cells by the TRIzol method (Invitrogen) followed by purification on NucleoSpin RNA II columns (Macherey-Nagel, Germany). The presence of contaminating genomic DNA was monitored by PCR amplification of ␤-actin. Pure RNA preparations were then reverse-transcribed with SuperScript II (Invitrogen) using oligo-dT or random hexamer primers as indicated.

MAPK Activation by LPA Requires p110␤
Rather Than p110␥-To examine the role of PI3K in MAPK activation, we have used a non-transformed monkey kidney cell line (Vero) that displays a strong increase of phosphorylated ERK during LPA stimulation (Fig. 1A). The involvement of PI3K was first approached using the conventional inhibitors wortmannin and LY294002. As shown in Fig. 1A, each drug inhibited by approximately 60% ERK phosphorylation in response to LPA. In contrast, each compound abolished Akt/PKB phosphorylation, indicating that they fully blocked PI3K activation in our experimental conditions. These results suggested that MAPK activation in Vero cells stimulated with LPA was partially dependent on PI3K. Because the function of PI3K in the MAPK pathway is poorly understood, we have studied the involvement of different PI3K isoforms using a dominant negative strategy. Cells were thus transiently cotransfected with distinct PI3K mutants and His-tagged ERK1. Following cell stimulation, ERK1-His was extracted and its phosphorylation was analyzed by immunoblotting. We have first examined the role of p110␥, because this enzyme was shown to participate in MAPK stimulation mediated by ␤␥ subunits (32). This was achieved by cotransfecting ERK1-His with the dominant negative p110␥-K832R mutant. Fig. 1B shows that expression of this mutant did not reduce ERK1-His phosphorylation induced by LPA. As a control, we have verified that p110␥-K832R acted as a dominant negative molecule by blocking ERK activation induced by ␤␥ overexpression (Fig. 1C), in agreement with previous reports (32,40). This suggested that p110␥ did not significantly participate in MAPK activation in response to LPA. In contrast, ERK1-His phosphorylation was strongly inhibited in cells transfected with a form of p85 lacking the p110 binding site (⌬p85) (Fig. 1B). This protein is a widely used dominant negative mutant for class I A PI3K, which suggested the involvement of p110␤ because other class I A enzymes are not activated by GPCRs. The role of p110␤ was further studied using a kinase-inactive mutant of the catalytic subunit (p110␤-K805R). This protein significantly reduced ERK1-His phosphorylation induced by LPA, although the effect was less pronounced than that of ⌬p85. To understand this difference, we have compared the capacity of the two mutants to interfere with PI3K activation in response to LPA. For convenience, the activation of Akt/PKB was used as readout of PI3K stimulation, by cotransfecting HA-tagged Akt/PKB with ⌬p85 or p110␤-K805R, followed by an in vitro kinase assay. As shown in Fig. 1D, whereas ⌬p85 abolished the stimulation of HA-Akt/ p110␤ Links LPA and ␤␥ to Ras PKB induced by LPA, p110␤-K805R produced only a partial inhibition, consistent with the respective capacity of the two mutants to interfere with MAPK activation. In addition, p110␥-K805R did not significantly interfere with HA-Akt/PKB stimulation induced by LPA, further suggesting that in Vero cells p110␥ does not significantly participate in LPA signaling. Altogether, these data strongly suggested that p110␤ was involved in MAPK activation in response to LPA.
Role of p110␤ Upstream from Ras-To study p110␤ function in the MAPK pathway, we have first determined at which level of the cascade p110␤ was involved. Because Ras is located in a central position, we have determined whether p110␤ was necessary for Ras activation. This was achieved by using the precipitation assay for activated Ras that takes advantage of a GST fusion protein containing the Ras-binding domain of Raf (RBD) to extract selectively activated GTP-bound Ras (43). The amount of activated Ras in the pull-down assays was determined by anti-Ras immunoblotting. As shown in Fig. 2A, LPA strongly stimulated endogenous Ras in Vero cells. Pretreatment of the cells with wortmannin or LY294002 significantly reduced this activation, suggesting an involvement of p110␤ upstream of Ras. To test this hypothesis, the activation of HA-tagged wild type Ras was studied in cotransfection experiments with PI3K mutants. As shown in Fig. 2B, the ⌬p85 and p110␤-K805R mutants significantly inhibited HA-Ras activation induced by LPA, with a relative efficiency corresponding to their ability to block Akt/PKB activation (Fig. 1D). In addition, dominant negative p110␥ mutant did not prevent Ras activation induced by LPA (Fig. 2B), further excluding a role for this PI3K in this pathway. Finally, we have used an antisense . The graphs represent immunoblot quantification by densitometry using the software IMAGE (National Institutes of Health), unless otherwise indicated. B, cells were cotransfected with constructs encoding ERK1-His and one of the following proteins: empty vector (V); GST-tagged kinase-dead p110␥-K832R (␥KR); HA-tagged dominant negative p85 (⌬p85); HA-tagged kinase-dead p110␤-K805R (␤KR). After stimulation and lysis, cells were subjected to ERK1-His extraction, followed by anti-phosphoERK (upper panel) and anti-His (lower panel) immunoblotting (IB). C, the dominant negative activity of the p110␥ mutant was verified on ERK1-His phosphorylation induced by ␤␥ overexpression, as previously described (32). D, cells were cotransfected with HA-tagged Akt/PKB and the indicated constructs, following the conditions described under "Experimental Procedures." After cell stimulation, HA-Akt/PKB was immunoprecipitated, then subjected to in vitro kinase assay using histone 2B (H2B) and [␥-32 P]ATP as substrates. Radiolabeled histones were revealed using a PhosphorImager (upper panel) and quantified using the software ImageQuant (graph). The amount of immunoprecipitated HA-Akt/PKB was controlled by anti-HA immunoblotting (lower panel). E, PI3K mutant expression was verified in lysates from transfected cells using the appropriate antibody. Mean Ϯ S.E. are from at least three independent experiments. *, different from stimulated control or empty vector; ns, not significant; p Ͻ 0.05, paired t test.
p110␤ Links LPA and ␤␥ to Ras strategy to confirm the involvement of p110␤ in Ras activation. We have constructed a plasmid encoding full-length p110␤ in antisense orientation. Transfection of this plasmid in Vero cells produced a partial reduction of endogenous p110␤, which might be due to the low percentage of transfected cells (Fig. 2C). In these conditions, we have tested whether this construct prevented Ras activation in cotransfection experiments to circumvent the low percentage of transfection. As shown in Fig. 2C, transfection of antisense p110␤ led to a partial but significant inhibition of Ras activation. Taken together, these data indicated that p110␤ contributed to Ras stimulation in response to LPA.
Role of the Lipid and Protein Kinase Activities of p110␤-Because each of the lipid or protein kinase activities of PI3K can be involved in the MAPK pathway, we have determined for p110␤ the respective participation of each activity in Ras activation. We have thus prepared a protein kinase only (PKO) mutant by deleting the region that putatively interacts with the phosphoinositide polar head, according to the sequence alignment between p110␤ and p110␥ (34). The effect of this deletion on p110␤ was verified using immune complex kinase assays following expression in COS cells. The lipid kinase assay was performed using phosphatidylinositol as a substrate, and autophosphorylation was used as a protein kinase assay. As controls, the assays were also performed with wild type (wt) and kinase-inactive p110␤ (K805R). Fig. 3A shows that, as expected, the PKO protein has lost the ability to phosphorylate phosphatidylinositol. In contrast, its protein kinase activity was conserved and was even higher than that of wild type p110␤ (Fig. 3B), which supported the notion that the two catalytic activities of PI3K are antagonistic (37). Therefore, the above deletion did convert p110␤ into a protein kinase only mutant. To determine whether this mutant could exhibit dominant negative effects, we have tested its ability to interfere with HA-Akt/PKB activation in cotransfection experiments. As shown in Fig. 3C, the PKO mutant inhibited HA-Akt/PKB stimulation as efficiently as the kinase dead K805R mutant, demonstrating that the loss of p110␤ lipid kinase activity had turned the PKO protein into a dominant negative mutant for D3-phosphoinositide-dependent processes. We have thus tested this mutant in cotransfection experiments with HA-Ras to determine whether p110␤ lipid kinase activity was dispensable for Ras activation. However, Fig. 3D shows that the PKO protein inhibited HA-Ras stimulation as efficiently as the kinase-dead mutant, which suggested that p110␤ lipid kinase activity was essential to Ras activation. To strengthen this result, Ras activation was measured in cells overexpressing PTEN, a phosphatase that degrades PI3K lipid products (46). As shown in Fig. 3E, overexpression of wild type PTEN significantly inhibited the stimulation of HA-Ras induced by LPA. Because PTEN has also a protein phosphatase activity that might interfere with Ras activation, we have used as a control FIG. 2. Involvement of p110␤ in LPA-induced Ras activation. A, control (ctrl) or LPA-treated cells, preincubated or not with wortmannin (ϩW) or LY294002 (ϩLY), were subjected to the precipitation assay of activated Ras using a GST fusion protein containing the Rasbinding domain of Raf (GST-RBD). 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). B, cells were cotransfected with 0.5 g of plasmid encoding HA-tagged Ras and one of the construct encoding the following PI3K mutants: empty vector (V, 1.5 g) dominant negative p85 (⌬p85, 1 g ϩ 0.5 g of empty vector); catalytically inactive p110␤ (␤K805R, 1.5 g); catalytically inactive p110␥ (␥K832R, 1.5 g). Following stimulation, cells were processed as in A. C, HA-tagged Ras activation was measured as in A in cells transfected with empty vector (V) or a construct encoding full-length p110␤ in antisense orientation (AS p110␤). Lower panel: cell lysates were immunoblotted with anti-p110␤ antibody to verify inhibition of expression. The graphs represent quantifications (mean Ϯ S.E.) from three independent experiments. *, less than stimulated control; ns, not significant; p Ͻ 0.05, paired t test.
p110␤ Links LPA and ␤␥ to Ras the "protein phosphatase only" mutant of PTEN (G129E) that does not interfere with PI3K signaling (46). Unlike wild type PTEN, this mutant did not modify HA-Ras activation induced by LPA, which indicated that the lipid phosphatase activity of PTEN was essentially responsible for Ras inhibition. Taken together, these results showed that the lipid products of p110␤ were critical for Ras stimulation in response to LPA.
Gab1 and SHP2 Link p110␤ to Ras-To define p110␤ function upstream from Ras, we have studied the proteins that associate with Grb2 in a PI3K-dependent manner. This was achieved by immunoprecipitating Grb2 from LPA-treated cells, preincubated or not with wortmannin. Immunoprecipitates were then analyzed by antiphosphotyrosine immunoblotting. As shown in Fig. 4A, LPA increased the phosphorylation of six different proteins, five of which were insensitive to wortmannin: The 180-kDa protein comigrated with the epidermal growth factor receptor (EGFR), the 66-, 52-, and 46-kDa proteins comigrated with Shc, and we failed to identify the 95-kDa protein (data not shown). In contrast, wortmannin inhibited the phosphorylation of the 115-kDa protein (Fig. 4A). This protein comigrated with Gab1 upon reblotting the samples with anti-Gab1 antibody (data not shown). Although Gab1 is an essential activator of p110␤ by providing p85-binding motifs (40), this result, and the fact that Gab1 contains a PH domain (49), suggested that Gab1 might also be the downstream effector of p110␤ lipid products. To assess this hypothesis, we have first studied the effect of PI3K inhibitors on Gab1 phosphorylation. This was achieved by immunoprecipitating Gab1 from control or LPA-treated cells, followed by antiphosphotyrosine immunoblotting. As shown in Fig. 4B, Gab1 phosphorylation in response to LPA was strongly reduced when cells were incubated with PI3K inhibitors. In addition, Gab1 recruitment was transiently transfected with constructs encoding the following HA-tagged p110␤ mutants: empty vector (V); wild type (␤wt); kinase dead K805R (␤K805R); protein kinase only (␤PKO). After 24-h expression, the three HA-p110␤ variants were immunoprecipitated and assayed for their lipid and protein kinase activities. A, lipid kinase assay: immunoprecipitates were incubated with phosphatidylinositol and [␥-32 P] ATP. Lipids were then extracted and separated using TLC. Radiolabeled lipids were revealed with a PhosphorImager. B, protein kinase assay: immunoprecipitates were incubated with [␥-32 P]ATP. Autophosphorylated p110␤ was resolved by SDS-PAGE and revealed using a PhosphorImager. C, cells were cotransfected with HA-tagged Akt/PKB and the indicated p110␤ constructs. Following cell stimulation, HA-Akt/PKB was immunoprecipitated and subjected to in vitro kinase assay. Histone phosphorylation (H2B) was revealed with a PhosphorImager (upper panel), and HA-Akt/PKB amounts were controlled by immunoblotting (IB HA, lower panel). D, cells were cotransfected with HA-tagged Ras and the indicated constructs. Following stimulation, cells were processed for the GST-RBD assay. The amount of activated HA-Ras associated with the beads was determined by anti-HA immunoblotting and quantified using IMAGE. E, cells were cotransfected with HA-tagged Ras and the indicated construct: empty vector (V); PTEN wild type (wt); PTEN protein phosphatase only mutant (G129E). HA-Ras activation in response to LPA was determined as in D. Data shown represent the mean Ϯ S.E. from three independent experiments. *, less than stimulated control; ns, not significant; p Ͻ 0.05, paired t test. D and E, lower panels: anti-HA immunoblotting of cell lysates to determine the expression level of transfected proteins.
p110␤ Links LPA and ␤␥ to Ras examined by studying its association with membrane fractions. Fig. 4C shows that Gab1 was strongly enriched in membrane fractions prepared from LPA-treated cells. This redistribution was almost abolished when cells were preincubated with wort-mannin or LY294002. These data showed that PI3K was essential for Gab1 recruitment in LPA signaling, suggesting that Gab1 was the downstream effector of p110␤ lipid products. To further establish the respective role of p110␤ and p110␥ in this pathway, we have studied the effect of their dominant negative mutants on Gab1-myc phosphorylation. As shown in Fig. 4D, the p110␤, but not p110␥, mutant inhibited LPA-induced Gab1 phosphorylation, further suggesting that Gab1 was the downstream target of p110␤ lipid products.
Gab1 is a multifunctional adapter protein that contains binding sites for numerous signaling proteins, including Grb2, p85, and the tyrosine phosphatase SHP2 that can be involved in Ras stimulation (47,50). This suggested that SHP2 might be a downstream effector of p110␤ and Gab1 required for Ras activation in response to LPA. To get further insight on the role of Gab1 and its partners in this pathway, we have studied the activation of both Ras and Akt/PKB in cells expressing Gab1 mutated on its SHP2 binding site (Gab1-Y627F) or on its three p85 binding sites (Gab1-YF3). These mutants were previously characterized using coimmunoprecipitation experiment as unable to associate with SHP2 or p85, respectively (47). As expected, Gab1-YF3 that prevents PI3K activation, significantly reduced HA-Ras stimulation induced by LPA (Fig. 5A). Interestingly, Gab1-Y627F strongly inhibited HA-Ras activation, suggesting that the association between Gab1 and SHP2 was important for Ras stimulation. To determine whether this interaction was required upstream or downstream from PI3K we have studied the ability of Gab1-Y627F to interfere with PI3K activation using Akt/PKB as readout. As shown in Fig. 5B, Gab1-Y627F slightly increased Akt/PKB stimulation induced by LPA compared with Gab1-wt, whereas, as a control, Gab1-YF3 nearly abolished this activation. These results suggested that the association between Gab1 and SHP2 was important for activation of Ras, but not of PI3K, implying that SHP2 was a downstream effector of PI3K and Gab1 in this pathway. To assess this hypothesis, we have studied the role of PI3K in the recruitment of SHP2 in Ras activation complex. This was achieved by examining its association with Grb2 in coimmunoprecipitation experiments. As shown in Fig. 5C, LPA induced the association of SHP2 with Grb2, and this interaction was strongly inhibited by wortmannin. This showed that SHP2 was recruited downstream of PI3K, and supported the idea that the phosphatase was an intermediate between p110␤ and Ras. To further explore this pathway, we have determined whether SHP2 was required for Ras activation in response to LPA. This was achieved by studying HA-Ras activation in cotransfection experiments with a catalytically inactive mutant of SHP2 (C/ S), or wild type SHP2 as control. Fig. 5D shows that SHP2-C/S, but not wt-SHP2, inhibited HA-Ras stimulation, which indicated that SHP2 catalytic activity was necessary for Ras activation induced by LPA. Taken together, these data showed that Gab1 and SHP2 mediated the participation of p110␤ to Ras stimulation in response to LPA.
These results also suggested that the pathway described herein might provide a novel link between ␤␥ and Ras, because ␤␥ is thought to participate in p110␤ activation. On the other hand, this pathway could be simply activated downstream of G q , because it requires EGFR transactivation. To gain insight on the G protein specificity of this pathway, we have first studied Ras activation in cells treated with pertussis toxin or transfected with the C-terminal domain of the ␤-adrenergic receptor kinase (␤ARK-CT) that acts as a ␤␥-scavenger molecule (51). As shown in Fig. 6 (A and B), pertussis toxin and ␤ARK-CT strongly inhibited Ras activation in response to LPA, which indicated that G i -derived ␤␥ played an essential role in this pathway. Although p110␤ might be the downstream effec- FIG. 4. Role of PI3K in Gab1 recruitment. A, anti-Grb2 immunoprecipitates were obtained from control (ctrl) or LPA-treated cells, preincubated or not with wortmannin (ϩW), as indicated. Samples were analyzed by immunoblotting with antiphosphotyrosine (IB P-Tyr) and anti-Grb2 (IB Grb2) antibodies. As a control, the immunoprecipitation was performed without adding Grb2 antibody (ϪAb). B, before stimulation with LPA, cells were incubated with PI3K inhibitors when indicated (ϩW, ϩLY). Cells were then subjected to Gab1 immunoprecipitation, followed by anti-phosphotyrosine (IP P-Tyr) and anti-Gab1 (IB Gab1) immunoblotting. C, membrane fractions from cells treated as in A were prepared by ultracentrifugation. Membrane fractions were then analyzed by anti-Gab1 immunoblotting (upper panel), followed by anti-EGFR immunoblotting to control gel loading (lower panel). D, cells were cotransfected with Gab1-myc and the indicated constructs: empty vector (V); kinase-dead p110␤ (␤KR); kinase-dead p110␥ (␥KR). After stimulation, cells were subjected to Gab1-myc immunoprecipitation followed by anti-phosphotyrosine (upper panel) and anti-myc (lower panel) immunoblotting. The results displayed are representative of at least two independent experiments performed in duplicate.
p110␤ Links LPA and ␤␥ to Ras tor of ␤␥, it has also been proposed that ␤␥ could participate in EGFR recruitment (20). To distinguish between these two possibilities, we have compared the effect of pertussis toxin and ␤ARK-CT on LPA-induced EGFR transactivation and p110␤ activation. As shown in Fig. 6 (C and D), pertussis toxin and ␤ARK-CT significantly inhibited Akt/PKB activation that was used as marker of p110␤ activation. In contrast, pertussis toxin did not modify EGFR phosphorylation induced by LPA (Fig.  6E). We have also studied EGFR transactivation in cells transfected with ␤ARK-CT. To circumvent the low efficiency of transfection, we have attempted to cotransfect Vero cells with a His-tagged EGFR and analyze its phosphorylation in response to LPA, but this produced a constitutive activation of this receptor (data not shown). Vero cells were thus transfected with a construct encoding myc-tagged Grb2, and its association with EGFR and Shc was used as a readout of EGFR transactivation. As shown in Fig. 6F, LPA induced the coimmunoprecipitation of Grb2-myc with the EGFR and Shc, and these associations were not modified in cells expressing ␤ARK-CT. Altogether, these data indicated that G i and ␤␥ were involved in p110␤ activation but not in EGFR transactivation, suggesting that p110␤ and its downstream effectors provide a novel connection between ␤␥ and Ras.
As shown in Fig. 1, the PI3K-dependent pathway described herein seems to account for about half of ERK activation induced by LPA. As a preliminary study of the other half of this activation, we have used Ro-31-8220, an inhibitor of all PKC family members, alone or in combination with wortmannin or pertussis toxin. As shown in Fig. 6G, Ro-31-8220 partially reduced ERK phosphorylation induced by LPA. In addition, when cells were treated with both Ro-31-8220 and wortmannin, ERK phosphorylation was nearly abolished, suggesting that PKC and PI3K acted on distinct pathways leading to MAPK activation. In agreement with this, Ro-31-8220 displayed also an additive effect with pertussis toxin that was shown to block the PI3K pathway at the level of G i (see Fig.  6C). Moreover, Ro-31-8220 did not interfere with LPA-induced activation of both Akt/PKB and Ras (Fig. 6, A and C), further suggesting that PKC and the G i /p110␤/Ras pathway provide two distinct routes leading to ERK activation in response to LPA.
The ability of LPA to stimulate signaling pathways is thought to be mediated through at least three different GPCR, namely LPA 1 (Edg2), LPA 2 (Edg4), and LPA 3 (Edg7). As a first approach to define their role in the pathway described herein, we have studied their expression using RT-PCR analysis. As FIG. 5. Role of Gab1 and SHP2 in Ras and Akt activation. A, cells were cotransfected with HA-tagged Ras and the indicated construct: empty vector (V); wild type Gab1-myc (wt); Gab1-myc mutated on its SHP2 binding site (Y627F); Gab1-myc mutated on its three p85 binding sites (YF3). Following stimulation, cells were lysed and processed for GST-RBD assay. The amount of precipitated activated HA-Ras was determined by anti-HA immunoblotting (upper panel). Lysates were also directly subjected to anti-HA (middle panel) and anti-myc immunoblotting (lower panel) to control expression of HA-Ras and Gab1-myc constructs. The graph represents a quantification (mean Ϯ S.E.) from three independent experiments. *, less than empty vector; ns, not significant; p Ͻ 0.05, paired t test. B, cells were cotransfected with HA-tagged Akt/PKB and the indicated Gab1 constructs, then processed for HA-Akt/PKB immunoprecipitation followed by in vitro kinase assay. Phosphorylation of histones (H2B) was quantified with a PhosphorImager (upper panel ϩ graph, mean Ϯ S.E. from three experiments) and HA-Akt/PKB was revealed by immunoblotting (lower panel). C, anti-Grb2 immunoprecipitates were performed from cells pretreated or not with wortmannin and stimulated with LPA as indicated. Immunoprecipitates were then analyzed by anti-SHP2 (upper panel) and anti-Grb2 (lower panel) immunoblotting. D, cells were cotransfected with HAtagged Ras and empty vector (V), catalytically inactive SHP2 (C/S) or wild type SHP2 (wt) as indicated. Cells were then stimulated and subjected to the GST-RBD pull-down assay as in A.
p110␤ Links LPA and ␤␥ to Ras shown in Fig. 7, both LPA 1 and LPA 3 could be amplified from Vero cell cDNA. However, LPA 1 was produced more readily, suggesting that it was expressed to a higher level. Because LPA 1 and LPA 3 preferentially activate G i and G q , respectively (6 -8), these results suggested that LPA 1 might be involved in the G i /p110␤/Ras pathway whereas LPA 3 could be responsible for the PKC-dependent pathway leading to MAPK activation. DISCUSSION Although PI3Ks are important effectors of LPA signaling, their role in the MAPK pathway has remained somewhat elusive. p110␥ appeared to be the best candidate to participate in this pathway, but the mechanisms connecting p110␥ to MAPK activation have been incompletely characterized. We report here a novel mechanism involving p110␤, a widely expressed p110␤ Links LPA and ␤␥ to Ras PI3K, upstream of a complex signaling pathway leading to Ras activation in response to LPA.
The involvement of p110␤ in this pathway is based on the dominant negative effects of three different PI3K mutants. First, ⌬p85, the standard inhibitor of class I A PI3Ks, inhibited MAPK activation induced by LPA, which strongly suggested the involvement of p110␤, because other class I A enzymes are not known to be activated by GPCR stimulation. In agreement with this, a kinase-inactive mutant of p110␤ catalytic subunit was found to interfere with the stimulation of both Ras and MAPK, although less efficiently than ⌬p85. This difference seems to be due to the differential ability of p85 and p110 mutants to interfere with PI3K stimulation, as shown by studying Akt/PKB activation. An explanation for this latter observation could be that p110 mutants must displace a pre-existing complex between endogenous p85 and p110 to produce dominant negative effects, whereas ⌬p85 can act on its own. Moreover, through a different approach using antisense cDNA of p110␤, we have also observed that this protein was important for Ras activation in response to LPA. Finally, the fact that the PKO mutant also produced a dominant negative effect on Ras activation further demonstrated the involvement of p110␤.
The PKO mutant was produced to distinguish which of the protein kinase or lipid kinase activities of p110␤ were involved in this pathway, considering that each activity can potentially participate in MAPK activation. However this mutant displayed a dominant negative activity on Ras activation, suggesting that the lipid kinase activity of p110␤ was essentially involved. In addition, overexpression of the phosphatase PTEN that degrades PI3K lipid products was found to specifically interfere with Ras activation, further supporting a major role for p110␤ lipid products in this pathway. Therefore, the functional link between PI3K protein kinase activity and the MAPK pathway remains thus far restricted to p110␥, although its protein substrate remains to be identified.
Moreover, it seems that the notion that p110␥ is a major intermediate between GPCR stimulation and the MAPK pathway must be somewhat moderated. Although it has been reported that dominant negative p110␥ interfered with MAPK activation in COS or CHO cells stimulated with LPA, we failed to reproduce this result in Vero cells. However, the study in COS cells was performed with a very low dose of LPA (40 nM), suggesting that the role of p110␥ is limited to weak stimulations of the MAPK pathway (34). A common theme may also be evoked to explain the results in CHO cells, because the involvement of p110␥ was essentially demonstrated in a particular cell line where Ras activation had been compromised (33). This suggests that p110␥ provides in these cell lines a limited contribution to MAPK activation that has been masked under our experimental conditions. Indeed, we have studied the response of normal cells stimulated with a mitogenic dose of LPA (10 M) that engages at least two pathways, i.e. PKC and Ras, leading to MAPK activation. In addition, although p110␥ is abundant in neutrophils or platelets, its level of expression is much lower in Vero and other adherent cell lines (COS, Rat1, and NIH-3T3) (data not shown). We therefore assume that the capacity of p110␥ to contribute significantly to the MAPK pathway, and other cellular responses, might be restricted to hematopoietic cells. In agreement with this, the phenotypes described to date in p110␥ knockout mice are associated with blood cell functions (52)(53)(54)(55), taking into account that the role of p110␥ in colorectal cancer has been recently challenged (56). Nevertheless, additional studies are necessary to determine whether p110␥ is physiologically important for MAPK activation in hematopoietic cells.
Several biochemical mechanisms have suggested a role for PI3K in the MAPK pathway at the level of Raf or MEK, because these enzymes are potential targets of kinases activated through PI3K-dependent processes (10,36,(57)(58)(59). However, we have observed that PI3K can play an important role upstream from Ras in cells stimulated with LPA. Indeed, the lipid products of p110␤ were found to facilitate the recruitment of the docking protein Gab1 that participates in Ras activation. Gab1 was thought to provide the phosphotyrosine-binding motifs required for p110␤ activation, but we have observed that it is also a downstream target of p110␤. We thus propose the following model connecting LPA to Ras where Gab1 is involved at two different steps (Fig. 8). Following LPA stimulation, Gab1 is recruited to phosphorylated EGFR through Grb2 or direct binding to the EGFR (49,60). This induces the phosphorylation of Gab1 on p85 and SHP2 binding sites, leading to p110␤ activation in synergy with ␤␥ subunits released during stimulation of LPA receptors. Activation of p110␤ produces PtdIns(3,4,5)P 3 that will recruit additional Gab1 molecules in the vicinity of the EGFR through binding to its PH domain. We assume that this mechanism provides a way to amplify the recruitment of p85 and SHP2, because the physical association of Gab1 with the EGFR is limited by the low number of EGFR molecules that are phosphorylated in response to LPA (data not shown). Downstream of Gab1, we have observed that SHP2 could provide a link with Ras, although the nature of this connection remains obscure. It has been proposed that SHP2 can function as an adapter protein in platelet-derived growth factor signaling, because it can bind to both the receptor and Grb2, and therefore contributes to the recruitment of Grb2-Sos (61). However, we have observed that the catalytic activity of SHP2 was important for Ras activation in response to LPA, in agreement with a recent report showing that this phosphatase was necessary for MAPK activation in HEK293 cells stimulated through LPA 1 (62). This suggests that SHP2 activates by dephosphorylation a protein involved in Ras activation, or down-regulates an inhibitor of Ras. One possible candidate FIG. 7. RT-PCR analysis of LPA receptors in Vero cells. RNA from Vero cells was extracted, purified, and reverse-transcribed using random hexamers (hex) or oligo-dT (dT) primers. cDNAs were then amplified by PCR using specific primers for LPA 1 (A), LPA 2 (B), or LPA 3 (C) and ␤-actin as control. Amplification of ␤-actin and LPA 1 was performed with 2.5 units of Taq DNA polymerase, whereas LPA 2 and LPA 3 were amplified with 5 units. cDNAs from the human T-cell leukemia cell line Jurkat (LPA 1 and LPA 2 ) and human fibroblast cell line IMR-90 (LPA 3 ) were used as positive controls.
p110␤ Links LPA and ␤␥ to Ras could be the GTPase-activating proteins that promote Ras deactivation. Interestingly, in U937 cells, PI3K was found to play a permissive role in basal activation of Ras through inhibition of GTPase-activating proteins (63). However, the mechanism of this regulation is not known, and additional studies are required to define whether Ras GTPase-activating proteins are a downstream target of SHP2.
One important aspect of the pathway described herein is that it provides a substantial connection between ␤␥ and the MAPK pathway. Although it is well established that ␤␥ is important for Ras activation, the effectors of this pathway have not been clearly identified (9,13). Interesting models have suggested that ␤␥ might be involved in EGFR recruitment through ␤-arrestin and Src (20, 21), but we have observed that ␤␥ was required at the level of p110␤, but not EGFR transactivation in Vero cells stimulated with LPA. Considering that EGFR stimulation per se can stimulate PI3K, one may wonder why ␤␥ and p110␤ are particularly required in this pathway. We assume that the direct recruitment of class I A PI3Ks to transactivated EGFR produces only a marginal PI3K stimulation corresponding to the strength of this transactivation. Therefore, the "boosting" effect of the interaction between ␤␥ and p110␤ (26,27) is probably important to initiate a significant activation of PI3K that will be subsequently amplified through Gab1.
In conclusion, p110␤ appears to be a pivotal link between ␤␥ and Ras, due to its unique ability to integrate two separate signals arising from GPCR stimulation, i.e. the release of ␤␥ and EGFR transactivation. We presume that the mechanism described here will apply to other cellular systems and other GPCR agonists, considering the relative ubiquity of p110␤ expression and EGFR transactivation.