Main

Transmembrane integrins bind to extracellular matrix proteins and generate important signals that regulate cell proliferation and migration events stimulated by receptors for soluble growth factors. Integrin and growth-factor signalling pathways can interact through several mechanisms, from membrane-proximal clustering of the two receptor types1,2 to the activation of common downstream signalling pathways3,4,5. Although there is a wealth of knowledge regarding the signalling pathways activated by both integrin and growth-factor receptors, little is known about how these signals are integrated by cells and whether there are common receptor-proximal control points that synchronize the execution of biological functions such as cell motility.

FAK is a non-receptor protein-tyrosine kinase (PTK) that indirectly localizes to sites of integrin-receptor clustering through C-terminal-domain-mediated interations6 with integrin-associated proteins such as paxillin7,8 and talin9. FAK becomes phosphorylated at seven to eight different tyrosine residues in vivo after engagement of integrin with matrix proteins10. Several studies have shown that FAK functions as part of a cytoskeleton-associated network of signalling proteins, also including the Src-family PTKs, p130Cas, Shc and Grb2, which act in combination to transduce integrin-generated signals to the ERK/JNK mitogen-activated protein (MAP) kinase cascades (reviewed in ref. 11).

Evidence indicates that tyrosine phosphorylation of FAK may be important for cell migration, as expression of the protein-tyrosine phosphatase PTEN leads to dephosphorylation of FAK and inhibition of cell motility12,13. Experiments using FAK-null cells14,15,16, FAK overexpression17 and dominant-negative FAK constructs18 have established that FAK is essential for integrin-stimulated cell migration. Null mutations of the murine fibronectin19 or fak20 genes result in a similar lethal phenotype where embryos fail to develop past day 8.5 of embryonic development because of defective gastrulation events. Cultured FAK-deficient (FAK–/–) cells exhibit a rounded morphology, enhanced formation of focal contacts with matrix proteins, and migration defects21. Expression of the FAK-related PTK Pyk2 is elevated in FAK–/– fibroblasts, and the combined phosphorylation activities of Pyk2 and Src-family PTKs facilitate integrin-stimulated ERK2 activation in the absence of FAK14. Interestingly, Pyk2 does not strongly localize to focal contact sites and its overexpression does not rescue the cell-migration defects caused by the FAK mutation14. When stably re-expressed in FAK–/– cells, epitope-tagged FAK localizes to focal-contact sites, promotes morphology changes and reverses the integrin-stimulated migration defects15.

Here we show that FAK acts as a receptor-proximal bridging protein that links growth-factor-receptor and integrin signalling pathways. By directly comparing FAK+/+, FAK–/– and FAK-reconstituted fibroblasts, we find that FAK associates with activated growth-factor receptors through its N-terminal domain and that FAK has an important function in promoting PDGF- and EGF-stimulated cell migration.

Results

FAK function is required for PDGF-stimulated cell migration.

To test the importance of FAK in promoting growth-factor-stimulated cell migration, we carried out chemotaxis assays in a modified Boyden chamber with FAK+/+ and FAK–/– fibroblasts, and with DA2 clonal fibroblasts, which re-express FAK15 (Fig. 1). At low concentrations of PDGF (2.5–10 ng ml–1), FAK+/+ and DA2 cells, but not FAK–/– cells, readily migrated to PDGF-BB in a dose-dependent manner (Fig. 1a). The lack of migration to PDGF-BB observed in FAK–/– cells was not due to the refractory responsiveness of the PDGFR, as normal levels of PDGF-stimulated tyrosine phosphorylation (10 ng ml–1, 5 min) were detected in lysates from FAK–/– cells (Fig. 1a).

Figure 1: FAK forms a complex with the activated PDGFR and is required for PDGF-stimulated cell motility.
figure 1

a,In a chemotaxis assay using a modified Boyden chamber, FAK+/+ fibroblasts and DA2 cells migrate towards PDGF-BB whereas FAK–/– cells do not. Values are means s.d. of at least three separate experiments. Starved FAK–/– fibroblasts respond to PDGF-BB stimulation with enhanced tyrosine phosphorylation, as detected by blotting of cell lysates against phosphorylated tyrosine (P–Y). Numbers in the right panel represent Mr in thousands. b, After stimulation with PDGF-BB, a phosphotyrosine-containing protein of Mr ~190K co-immunoprecipitates with antibodies to FAK in FAK+/+ cells and antibodies to HA-tagged FAK (HA–FAK) in DA2 cells, but not with antibodies to Pyk2 in FAK–/– cells. HA-tagged FAK associates with PDGFR immunoprecipitates (IPs) in DA2 fibroblasts after PDGF-BB stimulation. c, Transient co-transfection of PDGFR-null 293T cells shows the specific association of HA–FAK and PDGFR immunoprecipitates after PDGF-BB stimulation.

Full size image

Previous studies have shown that phosphorylation of FAK tyrosine residues increases upon cell stimulation involving growth factors, chemokines or G proteins (reviewed in ref. 11). Surprisingly, stimulation with PDGF caused only small increases in total phosphotyrosine levels in FAK and Pyk2. However, PDGF treatment induced the association of a phosphotyrosine-containing protein of relative molecular mass 190,000 ( Mr ~190K) with FAK in FAK+/+ and DA2 cells, but not with Pyk2 in FAK–/– cells (Fig. 1b). Reciprocal co-immunoprecipitation experiments using antibodies to the PDGFR and lysates from starved or PDGF-stimulated DA2 cells showed that HA-tagged FAK was tyrosine-phosphorylated and associated with PDGFR immunoprecipitates only after DA2 cells were stimulated with PDGF (Fig. 1b).

FAK forms a complex with activated PDGFR.

FAK has been shown to co-immunoprecipitate with an unknown phosphotyrosine-containing protein of Mr ~200K after cell stimulation with PDGF22. To confirm the ability of FAK and the PDGFR to associate, we transiently transfected human 293T cells, which do not express PDGFR, with expression vectors for the PDGFR β-subunit or HA-tagged FAK (Fig. 1c).

Expression of the PDGFR in 293T cells led to growth-factor-independent phosphorylation of the PDGFR (Fig. 1c, lanes 1, 5). In the absence of HA–FAK expression, antibodies to the HA tag did not co-immunoprecipitate PDGFR after PDGF-BB stimulation (Fig. 1c, lanes 1, 2). In cells not transfected with PDGFR (Fig. 1c, lanes 3, 4), HA-tag antibodies did not co-immunoprecipitate with either a phosphotyrosine-containing protein of Mr ~200K (data not shown) or a PDGFR-immunoreactive band. However, upon co-expression of PDGFR and HA–FAK, antibodies to the HA tag weakly pulled down the PDGFR in the absence of growth factor, whereas strong PDGFR association with HA–FAK was detected after PDGF-BB stimulation of the 293T cells (Fig. 1c, lanes 5, 6). These results indicate that PDGF-BB stimulation may promote the recruitment of FAK into an active PDGFR signalling complex.

FAK Y397 phosphorylation is required for both integrin- and PDGF-stimulated cell motility.

FAK mutants lacking the autophosphorylation/SH2-binding site (FAK(F397)), kinase activity (FAK(R454)), or the primary p130Cas SH3-binding site (FAK(A712/713)) are ineffective in promoting fibronectin-stimulated cell migration when transiently expressed in FAK–/– cells15. These data indicate that several protein–protein interactions involving FAK may be necessary to promote integrin-initiated cell motility (haptotaxis). To determine whether similar interactions with FAK are required to support PDGF-stimulated cell motility (chemotaxis), we transiently transfected FAK–/– cells with various FAK constructs and tested them for migration in response to PDGF-BB (Fig. 2a). Expression of HA-tagged FAK, FAK(R454), FAK(A712/713) or a Grb2-binding-site mutant (FAK(F925)) promoted a roughly eightfold increase in PDGF-stimulated cell migration, whereas equivalent expression of FAK(F397) did not effectively restore PDGF-stimulated cell motility (Fig. 2a). All of the FAK mutants except FAK(F397) were tyrosine-phosphorylated in PDGF-stimulated FAK–/– cells (data not shown).

Figure 2: The FAK SH2-binding site and Src-family protein-tyrosine-kinase activity are necessary for PDGF-stimulated cell migration.
figure 2

a, The FAK Y397 site is necessary to enhance PDGF-stimulated migration of FAK–/– cells, whereas kinase-inactive FAK(R454) or FAK mutants that do not bind p130Cas (FAK(A712/713)) or Grb2 (FAK(F925)) function to promote PDGF-stimulated FAK–/– motility. Values are means s.d. of at least three separate experiments. b, Effect of PDGF stimulation on the kinase activity of FAK, as measured by in vitro poly-Glu–Tyr phosphorylation and observed using an in vitro autophosphorylation (32P IVK) assay. The observed difference may reflect the recruitment of active Src-family PTKs to a FAK signalling complex. IPs, immunoprecipitates. Numbers in the central panel represent Mr in thousands. c, Inhibition of Src-family PTK activity by overexpression of p50Csk but not kinase-inactive p50Csk(K222M) prevents PDGF-stimulated motility of DA2 cells. Values are means±s.d. of two separate experiments.

Full size image

PDGF stimulates the association of Src-family PTKs with FAK.

The one common functional site in FAK needed to promote both fibronectin- and PDGF-BB-stimulated cell migration is the autophosphorylation/SH2-binding site, Y397. As the kinase activity of FAK is not required for PDGF-stimulated cell migration, it is possible that other PTKs may transphosphorylate FAK at the Y397 site after PDGF stimulation of cells. To evaluate the potential contribution of other FAK-associated kinases, we analysed HA–FAK immunoprecipitates from starved or PDGF-BB-stimulated DA2 cells for in vitro kinase activity using poly-Glu–Tyr as a substrate (Fig. 2b). In PDGF-BB-stimulated (10 ng ml–1, 5 min) DA2 cells, poly-Glu–Tyr transphosphorylation was increased sevenfold relative to serum-starved cells (Fig. 2b). Surprisingly, in parallel autophosphorylation assays, this increase in the PDGF-stimulated kinase activity of FAK was not reflected by strong PDGFR or FAK autophosphorylation activity in the HA–FAK immunoprecipitates, but was correlated with the recruitment of active Src-family PTKs to the activated FAK complex (Fig. 2b). A similar FAK–Src-PTK complex occurred after the addition of low concentrations of PDGF (1–10 ng ml–1) to Swiss 3T3 cells23.

SH2-mediated binding of Src-family proteins to the FAK Y397 site is competed by several other SH2-containing signalling proteins, such as Shc24, the p85 subunit of phosphatidylinositol-3-OH kinase25, SHP-2 (ref. 26), phospholipase C-γ1 (ref. 27) and Grb7 (ref. 28). To determine the importance of Src-family PTK activity in PDGF-stimulated migration of DA2 cells, we specifically inhibited Src kinase activity by transiently transfecting DA2 cells with an expression vector for C-terminal Src kinase (p50Csk). Overexpression of p50Csk markedly inhibited PDGF-stimulated (5 ng ml–1) motility, whereas overexpression of the kinase-inactive mutant p50Csk(K222M) did not (Fig. 2c). These results indicate that active binding of Src-family PTKs to the FAK Y397 site may be the first of several important signalling events necessary to promote PDGF-stimulated cell migration.

FAK function is required for EGF-stimulated cell migration.

To test whether the role of FAK in promoting cell migration is growth-factor-specific, we carried out chemotaxis assays of FAK+/+, FAK–/– and DA2 cells in a modified Boyden chamber using low concentrations of EGF (Fig. 3a). Whereas FAK–/– and DA2 cells expressed equivalent levels of EGFR (see Supplementary Information ), FAK–/– cells were not capable of initiating a migratory response, even though enhanced tyrosine phosphorylation of EGFR and its downstream targets, such as Shc, readily occurred in EGF-stimulated (10 ng ml–1, 5 min) FAK–/– cells (Fig. 3a). Similar to the stimulated formation of a complex containing activated PDGFR and FAK, the association of endogenous FAK with activated EGFR was detected by co-immunoprecipitation with anti-FAK antibody only after cells were stimulated with EGF (Fig. 3b). In a further analogy with the PDGF results, co-immunoprecipitation anlayses showed no detectable association of Pyk2 with an activated EGFR complex after EGF stimulation of FAK–/– cells (data not shown).

Figure 3: FAK forms a complex with the activated EGFR and is required for EGF-stimulated cell motility.
figure 3

a, In a chemotaxis assay using a modified Boyden chamber, FAK+/+ fibroblasts and DA2 cells migrate towards EGF, whereas FAK–/– cells do not migrate. Values are menas s.d. of at least three separate experiments. Starved FAK–/– fibroblasts respond to EGF stimulation with enhanced tyrosine phosphorylation, as detected by anti-phosphotyrosine (P–Y) blotting of cell lysates. Numbers in the right panel represent Mr in thousands. b, Association of EGFR and endogenous FAK upon stimulation with EGF of FAK+/+ cells, as detected by co-immunoprecipitation (IP) analyses with antibodies to FAK. c, Transfection of FAK constructs into 293T cells shows that the kinase activity of FAK, tyrosine phosphorylation, and the integrity of the FAK Y397 site are not required to mediate EGFR association in EGF-stimulated 293T cells. No EGFR association with the FAK C-terminal domain (FRNK) was detected. Ig, immunoglobulin.

Full size image

FAK forms a complex with activated EGFR.

To investigate further the association of FAK and EGFR, we transiently transfected 293T cells with various FAK mutant constructs and evaluated their association with endogenous EGFR using co-immunoprecipitation analyses after EGF stimulation (10 ng ml–1, 5 min; Fig. 3c). All the FAK mutants used, except the isolated FAK C-terminal domain (FRNK), formed complexes with activated EGFR (Fig. 3c). The fact that FAK(F397) readily formed a complex with activated EGFR, and yet showed only minimal levels of tyrosine phosphorylation, indicates that SH2-mediated binding of an adaptor protein to the FAK Y397 site may not be crucial to the FAK–EGFR interaction. By analogy with the results obtained from PDGF-stimulated FAK–/– cells (Fig. 2a), transient expression of FAK(F397) did not significantly enhance EGF-stimulated FAK–/– cell motility (data not shown). It has also been shown that inducible expression of FAK(F397) in FAK–/– cells inhibits serum-stimulated migration16. Together, these results support the hypothesis that FAK(F397) fails to promote both integrin-15 and growth-factor-stimulated cell motility because it fails to form a productive signalling complex with downstream targets such as the Src-family PTKs.

The FAK N-terminal domain is required for EGF-stimulated motility.

As the isolated FAK C-terminal domain did not detectably associate with activated EGFR in 293T cells (Fig. 3c), we investigated the possibility that the FAK–EGFR association is mediated by the FAK N-terminal domain. Previous studies have shown that truncations of the FAK N-terminal domain lead to a greatly raised level of FAK tyrosine phosphorylation10. We therefore introduced a kinase-inactive mutation (R454) into a truncated version of FAK to form FAK(Δ1–100, R454). We expressed wild-type FAK, FAK(R454) and FAK(Δ1–100, R454) in 293T cells and examined their ability to co-immunoprecipitate with activated EGFR after EGF stimulation (10 ng ml–1, 5 min; Fig. 4a). Whereas FAK(R454) was strongly associated with EGFR and also weakly tyrosine phosphorylated, FAK(Δ1–100, R454), although tyrosine phosphorylated, did not detectably associate with activated EGFR (Fig. 4a). Wild-type FAK and FAK(R454) were phosphorylated at the Y397 site, as detected by antibodies specific to phosphorylated Y397, whereas FAK(Δ1–100, R454) was only minimally phosphorylated at this site (Fig. 4a). These results indicate that the integrity of the FAK N-terminal domain may be important for mediating interactions with activated EGFR.

Figure 4: The FAK N-terminal domain is required for association with EGFR and EGF-stimulated cell motility.
figure 4

a, Wild-type FAK and kinase-inactive FAK(R454), but not FAK(Δ1–100, R454), associate with activated EGFR in 293T cells. FAK(R454) is tyrosine phosphorylated at the Y397 site after EGF stimulation, whereas FAK(Δ1–100, R454) is only minimally tyrosine phosphorylated at this site. IPs, immunoprecipitates. Number between the two panels represents Mr in thousands. b, Upper panel, blotting for HA-tagged truncated versions of FAK. Lower panel, wild-type FAK and FAK(R454) promote EGF-stimulated migration of FAK–/– cells equally, whereas FAK(Δ1–100, R454), FAK(Δ1–200, R454) and FAK(Δ1–300, R454) do not function to promote EGF-stimulated migration. Values are means s.d. of at least three separate experiments. c, FAK is phosphorylated at several sites and exhibits a high level of in vitro kinase activity when transiently expressed in FAK–/– cells. Wild-type FAK, but not FAK(Δ1–100, R454), associates with activated EGFR in FAK–/– cells. Phosphorylation of FAK(Δ1–100, R454) at the Y407 and Y950 sites is enhanced by EGF stimulation.

Full size image

To test the functional correlation between the FAK–EGFR association and EGF-stimulated cell migration, we transiently expressed mutant or truncated FAK constructs in FAK–/– cells and assessed their ability to promote migration after EGF stimulation (5 ng ml–1; Fig. 4b). Expression of either wild-type FAK or FAK(R454) promoted migration of FAK–/– cells equally well, whereas FAK(Δ1–100, R454), FAK(Δ1–200, R454) and FAK(Δ1–300, R454) did not significantly promote cell migration above control levels (Fig. 4b). Transfection of kinase-active FAK(Δ1–100) also failed to promote EGF-stimulated migration of FAK–/– cells (data not shown), but its transient expression significantly reduced FAK–/– cell adhesion, which hampered conclusions from motility assays. Nevertheless, the results obtained from FAK-truncation mutants show that formation of the FAK–EGFR complex is correlated to the promotion of EGF-stimulated cell motility. By analogy with PDGF-stimulated motility (Fig. 2a), the kinase activity of FAK was not required for EGF-stimulated cell migration. This can be explained mechanistically by the fact that FAK may be a substrate for other PTKs upon EGF stimulation of cells. This idea is supported by our finding that untruncated FAK(R454) associated with activated EGFR and was phosphorylated at the Y397/SH2-binding site when expressed in 293T cells (Fig. 4a).

To investigate further the transphosphorylation of FAK after EGF stimulation and to determine whether FAK(Δ1–100, R454) is phosphorylated at the Y397 site when transiently expressed in FAK–/– cells, we sequentially analysed immunoprecipitates of HA-tagged FAK and FAK(Δ1–100, R454), using phosphospecific antibodies that recognize different FAK phosphorylation sites (Fig. 4c). Under both starved and EGF-stimulated conditions, HA–FAK was highly tyrosine phosphorylated at the Y397, Y407, Y576, Y577 and Y861 sites, and weakly phosphorylated at the Y925 and Y950 sites. HA–FAK transiently expressed in FAK–/– cells associated with EGFR after EGF stimulation, yet it exhibited a high level of in vitro kinase activity under both starved and EGF-stimulated conditions (Fig. 4c). Association of FAK(Δ1–100, R454) with EGFR was not detected in FAK–/– cells, although the total level of tyrosine phosphorylation of FAK(Δ1–100, R454) markedly increased after EGF stimulation (Fig. 4c). Using the phosphospecific FAK antibodies, we detected FAK(Δ1–100, R454) phosphorylation at the Y407 and Y950 sites, but no significant phosphorylation at the Y397 site or in the kinase domain at Y576 and Y577 (Fig. 4c). The mechanism of tyrosine phopshorylation of FAK(Δ1–100, R454) is unknown. However, the fact that FAK(Δ1–100, R454) does not detectably associate with activated EGFR and is not phosphorylated at the Y397 site supports our previous conclusions that these two events are important for FAK function in promoting EGF-stimulated motility.

The isolated FAK N-terminal domain can associate with activated EGFR.

To determine whether the FAK–EGFR interaction can occur in the absence of the focal-adhesion-targeting signals located in the FAK C-terminal domain, we expressed FAK residues 1–402 (FAK(1–402)) as a Myc-epitope-tagged protein in 293T cells. Immunoprecipitation and blotting against the Myc tag detected a protein of Mr ~60K in cells transfected with this construct (Fig. 5a, lane 2). Co-immunoprecipitation analyses showed that Myc-tagged FAK(1–402) and HA-tagged FAK associated with activated EGFR (Fig. 5a, lanes 2, 4), whereas HA-tagged FRNK did not show a strong or reproducible association (Fig. 5a, lane 3 and Fig. 3c). Analysis of association of β1 integrin with the FAK constructs used showed that it was present in both the FRNK and the FAK immunoprecipitates but not in those of Myc-tagged FAK(1–402) (Fig. 5a, lanes 2–4). This result indicates that the association of the FAK N-terminal domain with EGFR may not involve linkages through β1 integrins; in this respect our results are not consistent with a previous report of in vitro binding of β1 integrin peptides to the FAK N-terminal domain29.

Figure 5: Exogenously expressed FAK N-terminal domain can form a complex with activated EGFR.
figure 5

a, 293T cells were transfected with the indicated constructs and stimulated with EGF. Sequential blotting of Myc-tag, HA-tag, or EGFR immunoprecipitates (IPs) shows that the Myc-tagged FAK N-terminal domain (Myc–FAK(1–402); Mr ~60K) associates with EGFR. Wild-type FAK and the FAK C-terminal domain (FRNK), but not FAK(1–402), form a complex with β1 integrins. FAK(1–402) is transphosphorylated at the Y397 site in EGF-stimulated cells. Numbers represent Mr in thousands. b, FAK(1–402) promotes EGF-stimulated migration of FAK–/– cells half as well as untruncated FAK. Values are means s.d. of two separate experiments. c, Expression of GST–FAK(1–402), but not GST–Pyk2(1–407), leads to the formation of a complex with activated EGFR in vivo upon EGF stimulation of 293T cells. Direct binding of GST–Grb2 to EGFR is much stronger than GST–FAK(1–402) association. Both GST–FAK(1–402) and GST–Pyk2(1–407) are tyrosine phosphorylated. GST–FAK(1–402) phosphorylation at the Y397 site is enhanced after EGF stimulation. Staining of the membrane with Coomassie blue shows the amounts of GST constructs associated with glutathione beads.

Full size image

To confirm the association of the FAK N-terminal domain with EGFR, we carried out reciprocal co-immunoprecipitation experiments using anti-EGFR antibodies, and blotted the immunoprecipitates for associated Myc-tagged proteins (Fig. 5a, lane 5). Myc-tagged FAK(1–402) was phosphorylated at the Y397/SH2-binding site after EGF stimulation of 293T cells, which is indicative of a transphosphorylation event (Fig. 5a, lanes 2, 5). A low level of β1-integrin-subunit signal was also detected in the EGFR immunoprecipitate (Fig. 5a, lane 5), which is consistent with previous reports of EGFR and integrin co-clustering2. FAK(1–402) expressed in FAK–/– cells promoted cell migration in response to EGF at half the level stimulated by wild-type FAK (Fig. 5b). These results support the idea that although FAK(1–402) forms a complex with activated EGFR and is phosphorylated at the Y397 site, full FAK function in promoting EGF-stimulated cell motility may also require targeting of the FAK C-terminal domain to sites of integrin-receptor clustering.

EGFR associates with the N-terminal domain of FAK but not of Pyk2.

Recent sequence comparisons have revealed that the N-terminal domains of both FAK and Pyk2 contain a divergent domain with homology to band 4.1 that has been implicated in mediating interactions with transmembrane receptors3031. To evaluate the specificity of the association of the FAK N-terminal domain with EGFR, and to show that this association occurs in an antibody-independent fashion, we amplified sequences encoding FAK(1–402), FAK(1–402, Δ1–100), the Pyk2 N-terminal domain (Pyk2(1–407)) and the Grb2 SH2/SH3 adaptor protein and cloned them into the pEBG mammalian glutathione-S-transferase (GST)-fusion expression vector. GST–FAK(1–402), GST–Pyk2(1–407) and GST–Grb2 were successfully expressed and bound to glutathione–agarose beads (Fig. 5c), but GST–FAK(1–402, Δ1–100) was proteolytically degraded and not stably expressed (data not shown). Both the GST–FAK(1–402) and the GST–Pyk2(1–407) fusion proteins were tyrosine phosphorylated in starved 293T cells (Fig. 5c, lanes 1, 2), and GST–Pyk2(1–407) and GST–Grb2 bound to an unknown phosphotyrosine-containing protein of Mr ~190K (Fig. 5c, lanes 2, 3).

As expected, EGF stimulation of 293T cells greatly enhanced the association of various phosphotyrosine-containing proteins such as Shc and EGFR with GST–Grb2 (Fig. 5c, lane 7). Significantly, EGF stimulation also promoted a clear, although weaker, association of a phosphotyrosine-containing protein of Mr ~200K with GST–FAK(1–402); this protein was positively identified as EGFR by immunoblotting (Fig. 5c, lane 5). EGF stimulation also increased the tyrosine phosphorylation of GST–FAK(1–402) at the Y397 site, but did not promote the association of GST–Pyk2(1–407) or GST alone with EGFR (Fig. 5c, lanes 6, 8). Farwestern protein–protein analyses showed that 32P-labelled GST–Grb2 bound directly to immunoprecipitated and tyrosine-phosphorylated EGFR, whereas 32P-labelled GST–FAK(1–402) did not (data not shown). Treatment of cells with cytochalasin D (1 µM, 20 min) before EGF stimulation disrupted the ability of FAK to associate with the activated EGFR complex (data not shown). These data support the idea that this FAK–EGFR association may reflect, similarly to FAK–integrin associations, an indirect interaction.

Expression of the FAK C-terminal domain inhibits EGF-stimulated cell motility.

Interactions, mediated by the C-terminal domain, of FAK with proteins such as paxillin7,8 and talin9 function to promote the targeting to, and the indirect association of, FAK with integrins at focal-contact sites. As expression and tyrosine phosphorylation of FAK(1–402) did not promote EGF-stimulated migration of FAK–/– cells to the same extent as with wild-type FAK (Fig. 5b), full activity may require C-terminal-domain-mediated targeting of FAK to focal contact sites. As C-terminal truncations of FAK were subject to proteolytic degradation when expressed in FAK–/– cells (data not shown), we exogenously expressed the FAK C-terminal domain (FRNK) as a competitive inhibitor of FAK localization to sites of integrin-receptor clustering32.

Although FRNK did not strongly associate with the activated EGFR complex when expressed in 293T cells (Fig. 3d), co-expression of FRNK with FAK in FAK–/– cells disrupted the association of FAK with the activated EGFR complex (Fig. 6a) and suppressed EGF-stimulated FAK phosphorylation at the Y397 site (Fig. 6b). Significantly, FRNK expression potently inhibited EGF-stimulated migration of DA2 cells (Fig. 6c), whereas equivalent expression of a FRNK point mutant (FRNK(S1034)) that does not strongly localize to focal-contact sites15 did not inhibit DA2 motility (Fig. 6c) or disrupt the stimulated formation of a FAK–EGFR complex (data not shown). These results support the conclusion that full FAK function in promoting EGF-stimulated motility involves a combination of distinct interactions mediated by both the N and the C termini (Fig. 7). In addition, the results obtained from FRNK overexpression indicate that a link between the FAK N terminus and EGFR may be facilitated and/or stabilized by interactions between the FAK C terminus and integrins.

Figure 6: Expression of the FAK C-terminal domain inhibits the FAK–EGFR association and EGF-stimulated cell motility.
figure 6

a, Co-expression of the FAK C-terminal domain (FRNK) with untruncated FAK in EGF-stimulated FAK–/– cells inhibits FAK association with activated EGFR complexes. IPs, immunoprecipitates. Numbers represent Mr in thousands. b, Co-expression of FRNK with untruncated FAK in EGF-stimulated FAK–/– cells suppresses FAK phosphorylation at the Y397 site. c, Expression of FRNK, but not the point mutant FRNK(S1034) that does not localize to focal contact sites, inhibits EGF-stimulated DA2 cell motility, as measured by a modified Boyden-chamber assay. Values are means±s.d. of at least three separate experiments.

Full size image
Figure 7: Model of FAK function.
figure 7

Indirect localization of FAK to sites of integrin and growth-factor-receptor clustering places FAK in a position to act as a receptor-proximal regulatory protein. Phosphorylation of FAK at the Y397 site and the integrity of the actin cytoskeleton are both required for FAK function in promoting both growth-factor (GF)- and integrin-mediated cell motility.

Full size image

Discussion

It is well documented that FAK has important functions in integrin-initiated signalling events. Changes in the tyrosine-phosphorylation status of FAK also occur after stimulation with soluble growth factor in several different cell types. We have shown here that FAK is important in linking PDGFR and EGFR activation to the cellular machinery that promotes directed cell migration. Our results support a new function for the unique FAK N-terminal domain which, upon exogenous expression, associated with an activated EGFR complex. The binding of the FAK N-terminal domain to EGFR was not as strong as direct Grb2 SH2-domain-mediated binding to the phosphorylated EGFR. In addition, the association of FAK with activated complexes of both EGFR and PDGFR could be disrupted by pretreatment of cells with cytochalasin D, which prevents actin polymerization but does not block tyrosine phosphorylation of PDGFR or EGFR. As we could not detect a direct association between the FAK N-terminal domain and EGFR in farwestern analyses, this interaction, like the co-localization of FAK with integrin receptors, may be mediated by one or more intermediary bridging proteins.

The overall significance of our biochemical findings is highlighted by the fact that FAK is recruited to sites of growth-factor-receptor and integrin clustering in intact cells (see Supplementary Information ) and that FAK expression is required for both PDGF- and EGF-stimulated cell motility. This places FAK in a critical position as a receptor-proximal component of both integrin- and growth-factor-receptor PTK signalling pathways (Fig. 7). We found that deletions of the FAK N-terminal domain disrupted the association of FAK with an activated EGFR complex. FAK constructs lacking an intact N-terminal domain did not promote EGF-stimulated motility when analysed in gain-of-function chemotaxis assays using FAK–/– cells. However, tyrosine phosphorylation of FAK and exogenous expression of its N-terminal domain only partially functioned to promote EGF-stimulated FAK–/– cell motility. These results indicate that localization of FAK to sites of integrin-receptor clustering is also important for cell motility. Indeed, overexpression of the FAK C-terminal domain acted to disrupt the association of FAK with activated EGFR complexes, promote FAK dephosphorylation at the Y397 site and inhibit EGF-stimulated cell motility. Thus, FAK connections to both growth-factor receptors and integrins are required to coordinate cell migration.

The association of FAK with an activated CCR5 receptor complex has recently been demonstrated using co-immunoprecipitation analyses33. In addition, FAK has been shown to associate with the inactive EphA2 PTK34. Although these reports did not elucidate the molecular mechanisms mediating these FAK–receptor interactions, we speculate that the involvement of FAK with other transmembrane-receptor signalling pathways may be reflected by the fact that tyrosine phosphorylation of FAK can be modulated by several different extracellular stimuli (reviewed in ref. 11). Specificity between FAK- and Pyk2-mediated signalling events may be determined by differences in the interactions undertaken by their respective N-terminal domains.

With respect to the mechanism of FAK function in cell motility, we found, surprisingly, that the kinase activity of FAK was not required to promote either PDGF- or EGF-stimulated cell migration in FAK–/– cells, even though the same cells require FAK’s kinase activity to promote fibronectin-stimulated cell motility15. Direct binding of p130Cas to FAK was also not required for PDGF-stimulated cell migration, although previous analyses have shown that FAK–p130Cas interactions are important for fibronectin- and integrin-stimulated haptotaxis15,17. These results indicate that the molecular mechanisms of FAK-mediated motility differ depending upon whether the primary cell stimulus is initiated by the activation of integrin or growth-factor receptors. Mechanistically, the requirement for FAK’s kinase activity in promoting integrin-stimulated motility can be explained by the fact that integrins lack intrinsic catalytic activity. For growth-factor-stimulated motility, FAK phosphorylation may be initiated by receptor-PTK activation. We found that untruncated kinase-inactive FAK and the FAK N-terminal domain were transphosphorylated at the Y397 site after EGF stimulation, an indication that other PTKs participate in FAK-mediated signalling events.

A common and important mechanistic connection needed to promote both integrin- and growth-factor-stimulated cell migration was the FAK Y397 site. Phosphorylation at this site creates an SH2-binding site for a number of different signalling and adaptor proteins. Our results show that PDGF stimulation promotes the recruitment of active Src-family PTKs to FAK, and that inhibition of Src-family PTK activity by overexpression of p50Csk prevents PDGF- and FAK-mediated migration. As FAK–/– cells exhibit an overabundance of focal contacts and decreased cell motility, our results also support a working model for FAK function in promoting focal-contact turnover. The catalytic activity of Src-family PTKs has been shown to promote the turnover of focal-contact structures during cell motility35, and Src–FAK interactions are important for EGF-stimulated cell migration36. Therefore, FAK-mediated localization of active Src-family PTKs to sites of growth-factor- and integrin-receptor clustering may initiate focal-contact remodelling and migratory signals.

In summary, our results show that FAK can function as an important receptor-proximal component to promote cell migration. As FAK is overexpressed in a variety of invasive human tumours37,38 and the FAK gene is amplified in a variety of human cancer cells39, strategies to reduce levels of FAK protein or phosphorylation at the Y397/SH2-binding site may provide a way to interfere with enhanced growth-factor- and integrin-stimulated cell motility in both vascular and neoplastic diseases.

Methods

Cell culture.

FAK+/+, FAK–/– and DA2 fibroblasts were cultured, serum starved and transfected as described15. FAK+/+, FAK–/– and DA2 cells have the same p53–/– genetic background. Human 293T cells were cultured and transfected as described40.

DNA constructs.

HA-tagged constructs for wild-type FAK, FAK(F397), FAK(R454), FAK(A712/713), FAK(F925), FRNK and FRNK(S1034) were used as described15. Expression vectors for p50Csk and p50Csk(K222M) were used as described14. FAK(Δ1–100) (ref. 40) was cloned into FAK(R454) as a KpnI/ClaI fragment. Myc–FAK(1–402) was created by ligating a BamHI/ClaI wild-type FAK fragment into the BglII/XbaI sites of pCS–Myc14 using the primers 5′-CGATGGTGGTTGAT and 5′-CTAGATCAACCACCAT as linkers. BamHI/ClaI-ended fragments containing the nucleotide sequence encoding FAK(Δ1–200) and FAK(Δ1–300) were created using the polymerase chain reaction (PCR) and cloned into the same sites in pcDNA encoding FAK(R454). PCR was used to create a BamHI/XhoI-ended fragment containing the nucleotide sequence encoding FAK(1–402), FAK(1–402, Δ1–100), Pyk 2(1–407) and human Grb2. The 3′-flanking PCR primer contained the nucleotide sequence encoding a protein kinase A consensus phosphorylation site (amino-acid sequence RRASVG) upstream of the XhoI site in all constructs. Constructs were cloned into the BamHI/XhoI sites of pGEX-4T and subcloned as BamHI/NotI fragments into pEBG mammalian GST-fusion expression vector. All DNA constructs created by PCR were verified by sequencing.

Chemotaxis.

For modified Boyden-chamber (Millipore, Bedford, MA) motility assays, both sides of membranes were coated with rat-tail collagen (5 µg ml–1 in PBS; Boehringer) for 2 h at 37 °C and washed with PBS; chambers were placed in 24-well dishes containing migration media (0.4 ml DMEM containing 0.5% BSA) with or without human recombinant PDGF-BB or EGF (Calbiochem) at the indicated concentrations. For transient-transfection cell-migration assays, 2.5 µg pcDNA3–lacZ were combined with 2.5 µg FAK expression plasmids to identify transfected cells. Serum-starved cells (1 × 105 cells in 0.3 ml migration media) were added to the upper compartment and, after 3 h at 37 °C, cells on the membrane upper surface were removed with a cotton-tip applicator; migratory cells on the lower membrane surface were fixed by treatment with methanol and acid and then either stained (0.1% crystal violet, 0.1 M borate pH 9.0 and 2% ethanol) or analysed for transfected β-galactosidase activity using X-gal as a substrate. Cell-migration values were determined either by dye elution and absorbence measurements at 600 nm or by counting β-gal-stained cells (measured as cells per field using a ×40 objective). Means were taken from three individual chambers. Background levels of cell migration (less than 5% of total) in the absence of chemotaxis stimuli (0.5% BSA only) were subtracted from all points.

Immunoprecipitation.

Cells on 10-cm plates were lysed with 0.5 ml modified RIPA lysis buffer15; lysates were immediately diluted with 0.5 ml HNTG buffer (50 mM HEPES pH 7.4, 150 mM NaCl, 0.1% Triton X-100 and 10% glycerol), incubated with agarose beads and cleared by centrifugation. Immunoprecipitates with the indicated antibodies were carried out for 4 h at 4 °C and collected on agarose beads with protein A (Repligen, Cambridge, MA) or protein G-plus (Calbiochem); the precipitated protein complexes were washed at 4 °C in Triton-only lysis buffer (modified RIPA without sodium deoxycholate and SDS) and then in HNTG buffer before direct analysis by SDS–PAGE. For GST-fusion proteins, lysates of 293T cells were precleared and incubated with 20 µl glutathione–agarose-bead slurry (Sigma) for 2 h at 4 °C. Beads were pelleted by centrifugation and washed three times in 1% Triton buffer before direct analysis by SDS–PAGE. Associated GST-fusion proteins were observed by staining with Coomassie blue and immobilization on polyvinylidene fluoride membranes (PVDF; Millipore).

In vitro kinase (IVK) assay.

IVK assays of FAK immunoprecipitates in the presence or absence of poly-Glu–Tyr (4:1; Sigma) were carried out as described14.

Immunoblotting.

Proteins analysed by SDS–PAGE were transferred to PVDF membranes, blocked in 2% BSA and Tris-buffered saline with 0.05% Tween-20 for 2 h at room temperature and then incubated with either 1 µg ml–1 monoclonal antibody or a 1:1,000 dilution of polyclonal antibodies for 2 h at room temperature. Bound primary antibody was observed by enhanced chemiluminescence detection and subsequent reprobing of membranes was carried out as described15. Anti-phosphotyrosine and anti-PDGFR antibodies were from Upstate Biotechnology (Lake Placid, NY); anti-Myc- and anti-HA-tag antibodies were from Covance Research (Berkeley, CA); anti-EGFR and anti-Src antibodies were from Santa Cruz. Negatively pre-adsorbed and affinity-purified polyclonal antibodies to peptides corresponding to the sequences surrounding FAK tyrosine-phosphorylation sites Y397, Y407, Y576, Y577, Y861, Y925 and Y950 were from BioSource International (Camarillo, California, USA). Affinity-purified polyclonal antibodies to FAK, Pyk2 and Csk were used as described14.