INTRODUCTION

Focal adhesion kinase (FAK)¹ is a cytoplasmic tyrosine kinase involved in integrin-mediated signal transduction pathways (1, 2, 3, 4). In adherent cells, FAK colocalizes with integrins in focal contacts. FAK activation and tyrosine phosphorylation have been shown in a variety of cell types to be dependent on integrins binding to their extracellular ligands (2). Furthermore, FAK-deficient mouse embryos generated by FAK gene knockout exhibit a general deficiency in mesoderm which is very similar to what is seen in fibronectin-deficient mice (5, 6, 7). These results complement the observations in cell culture systems which suggest a unique role of FAK in signaling pathways initiated by integrin binding to fibronectin.

Recent studies have suggested that complexes of FAK with other cellular proteins may play important roles in signal transduction by integrins (1). In fibroblasts, integrin engagement promotes the association of FAK with both c-Src and the adaptor protein Grb2 (8) and results in activation of mitogen-activated protein kinase (8, 9, 10). Tyrosine residue 397 has been identified as the major site of FAK autophosphorylation (11, 12) and the binding site for the SH2 domain of Src family kinases (13, 14). Tyr-925 of FAK, which is phosphorylated by Src in vitro, has been identified as the binding site for the SH2 domain of Grb2 (8). Based on these results, it has been proposed that the interaction of FAK with Src and Grb2 can link integrin-initiated signals to the Ras/mitogen-activated protein kinase pathway (1, 8).

Studies from both our laboratory (15) and that of Guinebault et al. (16) have shown that FAK can also bind to phosphatidylinositol (PI) 3-kinase upon ligand engagement of integrins. PI 3-kinase phosphorylates at the D-3 position of the inositol ring of phosphatidylinositides to produce PI(3)P, PI(3,4)P₂, and PI(3,4,5)P₃, which are potential second messengers that can activate protein kinase C (17) or , , and  (18), and the related c-Akt protein kinase (19, 20). PI 3-kinase is a heterodimer composed of a 110-kDa catalytic p110 subunit and an 85-kDa regulatory p85 subunit. The latter contains an SH3 domain and two SH2 domains (21, 22). Binding of PI 3-kinase to FAK upon cell adhesion is mediated by the p85 subunit (15, 16). Guinebault et al. (16) have demonstrated that a GST fusion protein containing the p85 SH3 domain can bind to FAK as well as a synthetic peptide derived from a proline-rich region of FAK, indicating a role of the SH3 domain of p85 in FAK association with PI 3-kinase. However, several lines of evidence suggest that the p85 SH2 domains may also be responsible for the association. First, cell adhesion stimulates FAK association with PI 3-kinase in vivo (15) as well as its binding to the NH₂-terminal SH2 domain of p85 in vitro (8), concomitant with FAK activation and autophosphorylation. Second, autophosphorylation of recombinant FAK in vitro increases its binding to PI 3-kinase (15). Finally, FAK contains three tyrosines (Tyr-180, Tyr-652, and Tyr-950) in the YXXM motif for optimal binding to p85 SH2 domains (23), although it is not clear whether any of these sites is phosphorylated in addition to Tyr-397.

In this paper, we have investigated further the nature of the interaction between FAK and PI 3-kinase. Our results suggest that the major autophosphorylation site of FAK (Tyr-397) is the binding site for the SH2 domains of p85 and is responsible for the in vivo association of FAK with PI 3-kinase, even though it is not in the typical YXXM motif for binding p85 SH2 domains. We have also shown that Tyr-397 is responsible for the increased FAK:PI 3-kinase association upon platelet-derived growth factor (PDGF) stimulation (24), implying that no additional site of FAK was involved in its binding to PI 3-kinase after PDGF stimulation. Lastly, we present evidence that interaction of PI 3-kinase with Tyr-397 of FAK stimulates its activity. Together, these results suggest that FAK activation and autophosphorylation at Tyr-397 may lead to its association with PI 3-kinase through the SH2 domains of p85, which can subsequently activate PI 3-kinase during cell adhesion.

EXPERIMENTAL PROCEDURES

Protein A-Sepharose 4B, glutathione-agarose beads, rabbit anti-mouse IgG antibody, and PDGF were purchased from Sigma. G418 and LipofectAMINE were purchased from Life Technologies, Inc. The phosphopeptide pY397 surrounding Tyr-397 of FAK (SETDDpYAEIIDE) was purchased from Chiron Mimotopes (Victoria, Australia). The control peptide corresponding to the COOH-terminal 15 residues of FAK (IDQARLKMISQSRPH) was synthesized by the Biotechnology Program of Cornell University. The expression vector pKH3 was a generous gift of Dr. I. Macara (University of Vermont) and described previously (25). Monoclonal antibody (12CA5) against an epitope of the hemagglutinin (HA) protein of the influenza virus (YPYDVPDYA, HA epitope) was described previously (26). Anti-p85 serum was prepared in rabbits using GST-p85 as antigen.

pGEX-p85 was kindly provided by Dr. L. C. Cantley (Harvard University). pGEX-Src·SH2 was a generous gift of Drs. Bibbins and Varmus (27). pGEX-p85·NSH2, pGEX-p85·CSH2, and pGEX-Grb2·SH2 were kindly provided by Dr. T. Pawson (Mt. Sinai Hospital, Toronto) and described previously (28, 29). GST fusion proteins were produced and purified as described previously (14).

Preparation of insect (Sf21) cell lysates containing HA epitope-tagged FAK (wild type) and its mutants kd (kinase-defective; Lys-454 mutated to Arg), N (deletion of amino acids 50-376), C1 (deletion of amino acids 721-857), C2 (deletion of amino acids 853-963), and C3 (deletion of amino acids 965-1012) was as described previously (26). The Tyr-397  Phe mutation was introduced to HA epitope-tagged FAK using site-directed mutagenesis by overlap extension using the polymerase chain reaction, as described previously (30). The desired mutation was confirmed by dideoxy DNA sequencing (U. S. Biochemical Corp.). It was then cloned into expression vector pBlueBac2 (Invitrogen) and used to prepare Sf21 lysates containing HA epitope-tagged recombinant protein Y397F, as described previously (26).

For both transient expression in human 293 cells and stable expression in NIH 3T3 cells, FAK and some of its mutants were cloned into an expression vector pKH3 (25). cDNAs encoding FAK and its kd mutant were as described previously (15). The Y652F mutant (converting Tyr-652 to Phe) was generated by the methods as described above for the Y397F mutant. cDNAs encoding FAK and its mutants kd, Y397F, and Y652F were inserted into pKH3 at the BamHI cloning site, which generated in-frame fusions of a sequence encoding three HA epitopes (YPYDVPDYA) to the 5 end of the FAK coding sequences. The resulting expression plasmids were designated as pKH3-FAK, pKH3-kd, pKH3-Y397F, and pKH3-Y652F, respectively.

Human 293 cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum (Life Technologies, Inc.). One day after plating 5 × 10⁵ cells on 60-mm dishes, the cells were transfected with 2 µg of expression plasmids (pKH3-FAK, pKH3-Y397F, or pKH3-Y652F) or pKH3 (control) using LipofectAMINE. Two days after transfection, cell lysates were prepared and used for in vitro binding assays as described below.

Cell lysates were prepared in 1% Nonidet P-40 lysis buffer (20 mM Tris, pH 8.0, 137 mM NaCl, 1% Nonidet P-40, 10% glycerol, 1 mM Na₃VO₄, 1 mM phenylmethylsulfonyl fluoride, 0.2 trypsin inhibitory units/ml aprotinin, and 20 µg/ml leupeptin) as described previously (31). GST fusion proteins (5 µg) were immobilized on glutathione-agarose beads and then incubated with Sf21 cell lysates (40 µg) containing recombinant FAK proteins or lysates (100 µg) from transfected 293 cells in 1% Nonidet P-40 lysis buffer for 90 min at 4 °C. The complexes were washed four times with 1% Nonidet P-40 lysis buffer, resolved by SDS-polyacrylamide gel electrophoresis, and analyzed by Western blotting with 12CA5 (1:1,000 dilution) using the Amersham Enhanced Chemiluminescence system, as described previously (26). For peptide competition assays, immobilized GST fusion proteins were preincubated with various concentrations of oligopeptides in 1% Nonidet P-40 lysis buffer for 15 min at 4 °C and then incubated with 12 µg of Sf21 cell lysates containing recombinant FAK in the presence of the oligopeptides. The bound proteins were analyzed as described above.

NIH 3T3 cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% calf serum (Life Technologies, Inc.). Cells were cotransfected with 2 µg of pSV2neo (32) and 18 µg of the expression plasmids or pKH3 using the calcium phosphate precipitation method (33). Two days after transfection, cells were split 1:15 and plated in medium containing 0.5 mg/ml G418. The medium was replaced every 3-4 days. G418-resistant clones were selected after approximately 14 days and analyzed for the expression of exogenous FAK by Western blotting using 12CA5 as described above.

To detect the in vivo association of FAK with PI 3-kinase, cell lysates were prepared in 1% Nonidet P-40 lysis buffer from NIH 3T3 cell clones stably expressing HA epitope-tagged FAK or its mutants, as described previously (15). The lysates (1 mg) were immunoprecipitated by incubation with 5 µl of 12CA5 for 2 h at 4 °C. Immune complexes were collected on protein A-Sepharose beads that had been precoated with rabbit anti-mouse IgG antibodies. After washing four times with 1% Nonidet P-40 lysis buffer the immune complexes were divided into two parts. One portion was analyzed for the expression of the HA epitope-tagged FAK or mutants by Western blotting with 12CA5. The other aliquot was used in PI 3-kinase activity assays as described previously (15). In some experiments, cells were serum starved for 18 h and then treated with 25 µg/ml PDGF for 10 min at 37 °C before lysis.

To detect the effect of oligopeptides on the PI 3-kinase activity, PI 3-kinase was immunoprecipitated from lysates of serum-starved NIH 3T3 cells by rabbit anti-p85 serum and then assayed for its activity in the presence or absence of 80 µM oligopeptides, as described previously (34, 35). The region containing the labeled PI phosphate was excised and quantified by scintillation counting.

RESULTS

To determine the potential p85 binding sites on FAK, we first performed a series of in vitro binding assays using recombinant FAK and its mutants produced in insect Sf21 cells via baculovirus expression systems. The wild type FAK and mutants kd, N, C1, C2, and C3 were as described previously (26). The mutant Y397F was obtained by using similar methods, as described under ``Experimental Procedures.'' All recombinant FAK proteins were tagged with an epitope derived from the influenza virus HA sequence (YPYDVPDYA, HA epitope), which allows detection of the recombinant proteins by the monoclonal antibody 12CA5 (26; see Fig. 1A).

In vitro binding assays were performed using cell lysates prepared from Sf21 cells infected with various recombinant viruses encoding HA epitope-tagged FAK or its mutants. An aliquot of each sample was analyzed by Western blotting with 12CA5 to show that similar amounts of recombinant proteins were used in the binding assays (Fig. 1A). The remaining samples were incubated with immobilized GST fusion proteins containing p85 (GST-p85) or its NH₂- or COOH-terminal SH2 domains (GST-p85·NSH2 or GST-p85·CSH2). After washing, the bound proteins were resolved by SDS-polyacrylamide gel electrophoresis and detected by Western blotting with 12CA5. As shown in Fig. 1, wild type FAK and all four FAK deletion mutants (N, C1, C2, and C3) bound GST-p85 (panel B) as well as GST-p85·NSH2 and GST-p85·CSH2 (panels C and D, respectively). In contrast, the FAK mutant kd or Y397F mutant did not bind GST fusion proteins containing p85 or its two individual SH2 domains (panels B-D). The Y397F mutant bound the SH2 domain of Grb2 and had tyrosine kinase activity comparable to that of wild type FAK (data not shown), rendering it unlikely that the point mutation of Tyr-397 to Phe caused an overall conformational change leading to the inability of Y397F mutant to interact with p85. None of the recombinant FAK proteins bound to GST alone (data not shown). Taken together, these results suggested that FAK could associate with p85 by binding to either of its NH₂- or COOH-terminal SH2 domain. They also showed that kinase activity and Tyr-397 were required for FAK binding to p85 domains, suggesting that autophosphorylated Tyr-397 bound to these SH2 domains.

FAK contains three tyrosines (Tyr-180, Tyr-652, and Tyr-950) in the YXXM motif described for optimal binding to p85 SH2 domains (23), although it is not clear whether any of these sites is phosphorylated in addition to Tyr-397. Results shown in Fig. 1 indicated that neither Tyr-180 nor Tyr-950 was responsible for FAK binding to p85 because FAK mutants N (lacking Tyr-180) and C2 (lacking Tyr-950) bound to GST-p85 as effectively as wild type FAK. To test whether Y652 mediates FAK binding to p85, a FAK mutant Y652F (Tyr-652 mutated to Phe) was generated, transiently expressed in 293 cells, and used for in vitro binding assays shown in Fig. 2. Panel A shows that similar expression levels of the HA epitope-tagged wild type FAK, Y397F, and Y652F mutants were obtained from 293 cells. Panel B shows that wild type FAK bound GST-p85 but Y397F mutant did not, as expected. The Y652F mutant bound to GST-p85 as efficiently as the wild type FAK, suggesting that Tyr-652 was not involved in FAK binding to p85. Therefore, none of the three tyrosine residues of FAK in YXXM motif was responsible for FAK binding to p85 in vitro.

To examine the specificity of the interaction of p85 with phosphorylated Tyr-397 of FAK, a synthetic phosphopeptide pY397 (12-mer containing phosphorylated Tyr-397 and its flanking sequences) was tested for its ability to inhibit the binding of recombinant FAK to GST-p85. As shown in Fig. 3A, the phosphopeptide pY397 inhibited FAK binding to GST-p85 or GST-p85·NSH2. It had no effect on FAK binding to GST-Grb2·SH2, which is mediated by Tyr-925 (8). These results confirmed that the phosphorylated Tyr-397 of FAK was responsible for its interaction with p85 via binding to the SH2 domains of p85.

Phosphorylation of Tyr-397 in FAK has been shown to be responsible for its association with Src through its SH2 domain (11, 36). Indeed, the phosphopeptide pY397 also inhibited FAK binding to GST fusion protein containing the Src SH2 domain (GST-Src·SH2; Fig. 3A). Fig. 3B shows that pY397 inhibited the binding of recombinant FAK to GST-p85 or GST-Src·SH2 in a dose-dependent manner. The IC₅₀ to inhibit FAK:GST-p85 and FAK:GST-Src·SH2 associations were approximately 10 and 3 µM, respectively. A control peptide corresponding to the carboxyl 15 residues of FAK did not inhibit FAK binding to either GST-p85 or GST-Src·SH2 even at 200 µM.

From our in vitro data to this point, it appeared that phosphorylated Tyr-397 in FAK was responsible for FAK binding to PI 3-kinase through the SH2 domains of p85. To test if this was also the case in vivo, we established NIH 3T3 cell lines, which stably expressed HA epitope-tagged wild type FAK and three mutants (kd, Y397F, and Y652F), as described under ``Experimental Procedures.'' Association of PI 3-kinase with the transfected FAK and its mutants was assessed by immunoprecipitation with 12CA5 followed by PI 3-kinase activity assays, as described previously (15). Fig. 4A shows that similar amounts of HA epitope-tagged FAK and its mutants were present in all immunoprecipitates with the exception of cells transfected with pKH3 alone. Fig. 4B shows that PI 3-kinase was associated with wild type FAK and the Y652F mutant but not Y397F or kd mutants. Fig. 4C shows that PDGF increased the quantity of PI 3-kinase activity associated with wild type FAK and the Y652F mutant, consistent with our previous observation that PDGF could specifically stimulate the interaction of FAK with PI 3-kinase in NIH 3T3 cells (24). However, no PI 3-kinase activity was detected in association with either the kd or Y397F mutant, even after PDGF stimulation. Therefore, phosphorylation of Tyr-397 in FAK was also required for its association with PI 3-kinase in vivo in response to both cell adhesion and PDGF stimulation. Furthermore, these results indicated that other regions of FAK besides Tyr-397 were not involved in the increased FAK:PI 3-kinase association in response to PDGF.

To examine the potential functional consequences of FAK association with PI 3-kinase, we tested if the binding of the phosphopeptide pY397 to p85 could affect PI 3-kinase activity. Cellular PI 3-kinase was immunoprecipitated from NIH 3T3 cells using polyclonal anti-p85 and assayed for its activity in the presence or absence of pY397 or the control peptides. Fig. 5 shows that PI 3-kinase activity was increased 2-3-fold in the presence of pY397 but was not affected by the control peptide. These results suggested that association of FAK with PI 3-kinase might result in the activation of PI 3-kinase.

DISCUSSION

In mammalian cells, PI 3-kinase has been implicated in signal transduction pathways triggered by a variety of cell surface receptors (37, 38, 39, 40). For example, PI 3-kinase interacts with activated growth factor receptor tyrosine kinases upon ligand binding (41, 42, 43). This association has been demonstrated to be mediated by binding of autophosphorylated tyrosine residues in the receptor tyrosine kinase to the SH2 domains of p85 subunit of PI 3-kinase (44, 45, 46, 47). Recent studies have also shown that PI 3-kinase associates with activated FAK in response to integrin binding to its ligands (15, 16). However, less is known about the nature of FAK:PI 3-kinase interactions or the binding sites on FAK. In this paper we demonstrated that FAK kinase activity and the major FAK autophosphorylation site Tyr-397 were required for FAK association with PI 3-kinase. A mutation converting Tyr-397 to Phe abolished recombinant FAK binding to p85 or its SH2 domains in vitro. Consistent with this, a phosphopeptide corresponding to Tyr-397 and its flanking sequences (pY397) inhibited the binding of FAK to p85. Furthermore, the Tyr-397  Phe mutation also prevented FAK association with PI 3-kinase in vivo in response to either cell adhesion or PDGF stimulation. Finally, we observed that the binding of the phosphopeptide pY397 to p85 resulted in the activation of PI 3-kinase by 2-3-fold (Fig. 5). Together, these results suggested that FAK activation and autophosphorylation at Tyr-397 might lead to its association with PI 3-kinase through the SH2 domains of p85, which could subsequently activate PI 3-kinase during cell adhesion.

Guinebault et al. (16) have observed that the SH3 domain of p85 could bind to a proline-rich sequence in FAK (amino acids 875-880) using in vitro binding experiments, suggesting that the SH3 domain of p85 may play a role in FAK:PI 3-kinase association. Similarly, we also observed that the GST fusion protein containing the p85 SH3 domain was able to bind to recombinant FAK in vitro (data not shown). However, this proline-rich sequence appeared not to be necessary for FAK binding to p85 itself because the C₂ mutant (lacking amino acids 875-880) bound to GST-p85 as effectively as the wild type FAK (Fig. 1). Indeed, our results here strongly suggested that the primary FAK binding sites in PI 3-kinase were the SH2 domains of p85. Nevertheless, our results did not rule out the possibility that the p85 SH3 domain binding to the proline-rich sequences of FAK might also participate in FAK:PI 3-kinase association in response to certain stimuli in different cell types.

Our identification of Tyr-397 in FAK as its primary binding site for p85 SH2 domains was surprising because Tyr-397 was not in the YXXM motif for optimal binding to p85 SH2 domains reported previously (23). Although there are three tyrosine residues in FAK within the YXXM motif, our results here clearly ruled out their involvement in FAK binding to p85 (Figs. 1 and 2). Furthermore, there are several precedents where tyrosine residues not in the YXXM motif have been reported to bind to the SH2 domains of p85. These include YVXV in the hepatocyte growth factor receptor (48), YVNA in the vascular endothelial growth factor receptor (49), and YVAC in the erythropoietin receptor (50). One possible explanation for these exceptions is provided by recent results showing that the residues amino-terminal to the phosphotyrosine may also contribute to specific binding of phosphotyrosine to SH2 domains (51). Using a degenerate library of peptides in which residues both NH₂- and COOH-terminal to the phosphotyrosine are varied, it has been found that the optimal motif for binding the p85 NH₂-terminal SH2 domain is EDDpYVEM and that the preference for Met at the +3 position is not as strong as that when a peptide library varying only residues COOH-terminal to the phosphotyrosine is used (23). An examination of FAK sequences revealed that sequences flanking Tyr-397 (TDDpYAEI) conformed well with the EDDpYVEM motif for binding the p85 NH₂-terminal SH2 domain. Two aspartic acid and one glutamic acid residue were identical to those in the motif at the 2, 1, and +2 positions, respectively. Furthermore the general pattern of pY-hydrophobic-hydrophilic-hydrophobic residues for binding p85 SH2 domains (23) was also maintained in Tyr³⁹⁷-Ala-Glu-Ile of FAK.

Tyr-397 in FAK has previously been identified as the binding site for Src through its SH2 domain (11, 36). Furthermore, significantly more FAK:Src association has been detected in v-Src transformed NIH 3T3 cells than that in normal NIH 3T3 cells (14). However, we did not detect any increases in FAK:PI 3-kinase association in v-Src-transformed NIH 3T3 cells compared with normal NIH 3T3 cells (data not shown), which suggested that PI 3-kinase binding to FAK was likely to be independent of the Src:FAK interaction. In addition, the identification of Tyr-397 for both PI 3-kinase and Src binding suggested the interesting possibility that FAK might exist in different cellular pools complexed with either Src or PI 3-kinase. In v-Src-transformed CEF cells, at least 80% of FAK is complexed with v-Src (13). Experiments are in progress to determine the fraction of FAK associated with c-Src or PI 3-kinase in NIH 3T3 cells under various conditions.

Recent studies have suggested that FAK is involved in an integrin-triggered signaling pathway leading to cell migration. Ilic et al. (5) have shown that embryonic cells from FAK-deficient mice exhibited a decreased migration on fibronectin, which was suggested to be responsible for a defect in mesodermal migration resulting in an embryonic lethal phenotype of the FAK-deficient mice. We have found that overexpression of FAK in Chinese hamster ovary cells caused a significant increase in cell migration on fibronectin and that autophosphorylation of FAK at Tyr-397 and its subsequent association with Src and Fyn were correlated with this increased migration (52). Identification of Tyr-397 as the binding site for p85 here raised the possibility that FAK:PI 3-kinase as well as FAK:Src bindings may be important for downstream signaling events leading to cell migration.

Association of FAK with p85 may lead to activation of PI 3-kinase. We found that incubation of a phosphopeptide containing Tyr-397 and its flanking residues with PI 3-kinase increased its activity by 2-3-fold in vitro (Fig. 5). This was comparable to the level of activation of PI 3-kinase by phosphopeptides derived from receptor tyrosine kinases which are believed to activate PI 3-kinase in vivo (34, 35). These results and our previous observation that FAK could phosphorylate p85 in vitro (15) suggested that PI 3-kinase may be an important downstream effector of FAK in integrin signaling leading to cell migration. In preliminary experiments, we have found that wortmannin, a specific inhibitor of PI 3-kinase, could reduce cell migration on fibronectin of Chinese hamster ovary cells overexpressing FAK (data not shown). Identification of the PI 3-kinase binding site on FAK should allow us to dissect the relative contributions of PI 3-kinase and other FAK-binding proteins to downstream signaling events leading to cell migration.