Autophosphorylation-independent and -dependent Functions of Focal Adhesion Kinase during Development

Abstract

Focal adhesion kinase (FAK) regulates numerous cellular functions and is critical for processes ranging from embryo development to cancer progression. Although autophosphorylation on Tyr-397 appears required for FAK functions in vitro, its role in vivo has not been established. We addressed this question using a mutant mouse (fakΔ) deleted of exon 15, which encodes Tyr-397. The resulting mutant protein FAKΔ is an active kinase expressed at normal levels. Our results demonstrate that the requirement for FAK autophosphorylation varies during development. FAK^Δ/Δ embryos developed normally up to embryonic day (E) 12.5, contrasting with the lethality at E8.5 of FAK-null embryos. Thus, autophosphorylation on Tyr-397 is not required for FAK to achieve its functions until late mid-gestation. However, FAK^Δ/Δ embryos displayed hemorrhages, edema, delayed artery formation, vascular remodeling defects, multiple organ abnormalities, and overall developmental retardation at E13.5–14.5, and died thereafter demonstrating that FAK autophosphorylation is also necessary for normal development. Fibroblasts derived from mutant embryos had a normal stellate morphology and expression of focal adhesion proteins, Src family members, p53, and Pyk2. In contrast, in FAK^Δ/Δ fibroblasts and endothelial cells, spreading and lamellipodia formation were altered with an increased size and number of focal adhesions, enriched in FAKΔ. FAK mutation also decreased fibroblast proliferation. These results show that the physiological functions of FAK in vivo are achieved through both autophosphorylation-independent and autophosphorylation-dependent mechanisms.

Introduction

Focal adhesion kinase (FAK)² is a nonreceptor tyrosine kinase critical for processes ranging from embryo development (1) to cancer invasiveness and metastasis (2). FAK activation following integrin engagement or stimulation of a variety of transmembrane receptors triggers its phosphorylation on tyrosine and the formation of multimolecular signaling complexes (3). FAK is enriched in focal adhesions, controlling their turnover and consequently adhesion-related processes such as spreading, migration, survival, and proliferation (1).

The important physiological role of FAK is demonstrated by the lethality of its null mutation at embryonic day (E) 8.5 (4, 5). Further studies using conditional deletion showed that FAK regulates the development of the nervous system (6–9), morphogenesis of the vascular network (5, 10, 11), and cardiac development (12–15). These reports clearly established that FAK is necessary for essential processes in vivo.

In vitro studies have shown that, following its recruitment to focal adhesions, FAK autophosphorylation on Tyr-397 creates a high affinity binding site for multiple signaling proteins, including the Src family kinases (SFKs) (3). Following their binding to phospho-Tyr-397 and activation, SFKs phosphorylate other FAK residues inducing its complete activation, its interaction with other signaling proteins, and the stimulation of downstream signaling cascades (16). The FAK·SFK complexes also regulate cytoskeleton rearrangement and downstream signaling pathways by phosphorylating partner proteins such as p130Cas and paxillin (17, 18). Thus, FAK autophosphorylation on Tyr-397 appears to be critical for both FAK activation and scaffolding function in vitro, triggering the assembly of multimolecular complexes responsible for its cellular effects. Interestingly, recent results showed that pharmacological inhibition of FAK activity and autophosphorylation did not block tumor cell proliferation and apoptosis in vitro (19), suggesting that FAK may also have autophosphorylation-independent functions (20). Therefore it is particularly important to determine the role of Tyr-397 in FAK functions in vivo.

Here, we addressed this question using a mutant mouse deleted of FAK exon 15, which encodes 19 amino acids, including Tyr-397 (21). Our results show that the requirement for FAK autophosphorylation varies during development and demonstrate that the physiological functions of FAK in vivo are achieved through both autophosphorylation-independent and autophosphorylation-dependent mechanisms.

EXPERIMENTAL PROCEDURES

Generation of FAK^Δ/Δ Mice

FAK^Δ/Δ mice were generated (see also supplemental Fig. 1) in the course of experiments aimed at testing the role of FAK alternatively spliced exons (22). A DNA fragment containing exons 13–18 from the mouse Ptk2 gene was isolated from an SV129 genomic library (RPCI21MPAC, clone identification RPCIP711H19216Q2; RZPD, Berlin, Germany) and subcloned to construct the targeting vector (supplemental Fig. 1A). Embryonic stem cells isolated from male SV129 mice were electroporated with the targeting vector. Transfected cells were selected with hygromycin 24 h later and tested by PCR (data not shown) and Southern blotting (supplemental Fig. 1, A and B). For this study, we selected a clone that exhibited the homologous recombination event and an additional spontaneous deletion of a 307-bp genomic fragment, including exon 15 (nucleotides 34393031 to 34393338 of FAK gene sequence NT_039621.7). This mutant allele of FAK gene was named FAK^Δ. C57BL/6 blastocysts were injected with FAK^Δ/+ embryonic stem cells and transferred to pseudo-pregnant mice. Male chimeric mice carrying the FAK^Δ allele were mated to C57BL/6 female mice, and offspring were genotyped based on their coat color.

Animal Care

FAK^+/Δ mice were produced and maintained in the Fer-à-Moulin Institute animal house in stable conditions of temperature (22 °C) and humidity (60%) with a constant cycle of 12 h light and 12 h dark and free access to food and water. All experiments were in accordance with the guidelines of the French Agriculture and Forestry Ministry for handling animals (decree 87849, license A 75-05-22).

Genotyping by PCR

Mice and embryos were genotyped by PCR analysis (details in supplemental Fig. 4) using DNA extracted from tail biopsies.

Antibodies

Sources were as follows: pY-397-FAK, pY-418-Src, and pY421-cortactin from Biosource, Inc.; pY-118-paxillin, pY-410-p130Cas, pY-576/577-FAK, and pY-925-FAK from Cell Signaling Inc.; FAK (clone 4.47) and cortactin from Upstate Biotechnology, Inc.; Fyn (FYN3), Src (SRC2), FAK (C-20), and glyceraldehyde-3-phosphate dehydrogenase from Santa Cruz Biotechnology; rat PECAM-1 monoclonal CD31 (clone MEC13.3), CD102, p130Cas, and paxillin from BD Biosciences; biotinylated TuJI from R & D Systems; smooth muscle α-actin-fluorescein isothiocyanate (mouse clone 1A4) and vinculin from Sigma. Pyk2 rabbit antibody was produced in-house. Rhodamine-conjugated phalloidin and all secondary antibodies (AlexaFluor) were from Molecular Probes, Invitrogen, and Jackson ImmunoResearch.

Whole-mount Staining

Embryos were fixed in 4% paraformaldehyde in phosphate-buffered saline. After dehydration in a series of methanol, they were treated with 1% H₂O₂, rehydrated from methanol to phosphate-buffered saline, and blocked in 4% (w/v) bovine serum albumin, 0.1% (v/v) Triton X-100. They were then incubated with anti-PECAM-1 (1:500) in 4% bovine serum albumin, 0.1% Triton X-100 in phosphate-buffered saline at 4 °C overnight followed by peroxidase-conjugated secondary antibody. The reaction was developed in 0.03% 3,3′-diaminobenzidine with H₂O₂. For the whole-mount skin immunostaining, skins were removed after embryo fixation, permeabilized for 2 h (0.2% Triton X-100), and blocked (catalog number 11096176001, Roche Applied Science) for 2 h prior to incubation with antibodies against PECAM-1 and biotinylated TuJI (1:200 dilution) in 0.1% Triton + 1% bovine serum albumin + 1% goat serum overnight at 4 °C. After washing for 4 h, skins were incubated with secondary antibody (AlexaFluor 555 goat anti-rat) and monoclonal anti-smooth muscle α-actin-fluorescein isothiocyanate in blocking solution for 2 h. Acquisitions were performed with a confocal microscope (Leica, SP5, ×10 magnification).

Tissue Extraction and Preparation for Immunoblotting

Tissues and organs from embryos or adult mice were frozen immediately after dissection and sonicated later in 100 °C 1% (w/v) SDS containing NaVO₄ (1 mm), placed in a 100 °C heating block for 3 min, and processed as described under “Immunoblotting.”

Cell Culture

COS7 cells were grown and transfected with Lipofectamine 2000 (Invitrogen) as described previously (23). For the primary fibroblasts, four FAK^+/+ and three FAK^Δ/Δ independent mouse embryonic fibroblast (MEF) populations were prepared from seven littermates of E12.5 embryos of two different litters obtained from heterozygous crossings and cultured as described (24). All experiments were performed between passage 1 and 3. The seven MEF populations were tested separately in each experiment. Primary endothelial cells (EC) were isolated from FAK^+/+ and FAK^Δ/Δ E12.5 and E13.5 embryos using magnetic beads (Dynabead M-450; Dynal Corp.) and rat anti-mouse PECAM-1 (CD31) (5). For the fibronectin replating experiments, MEF or EC were prepared as described (25) before being plated on tissue culture dishes or glass coverslips precoated with polylysine (50 μg/ml) or fibronectin (10 μg/ml) and processed for immunoblotting or immunofluorescence at the indicated times.

Immunoblotting

After the appropriate treatment, cells were rinsed, frozen on dry ice, lysed with boiling SDS (1%) containing NaVO₄ (1 mm), sonicated, and boiled for 3 min. Protein concentration was determined with the BCA assay (Pierce), and 50–100 μg of protein were analyzed by SDS-PAGE. Quantification of immunoblots was performed with Odyssey Li-Cor. Data were normalized to the mean value of untreated controls in the same autoradiograms.

Immunofluorescence

Cells were fixed for 15 min in 4% (w/v) paraformaldehyde and permeabilized with 0.05% Triton X-100 for 5 min. After blocking and incubation with primary antibody overnight, cells were incubated with AlexaFluor 488- or Cy3-coupled secondary antibodies or rhodamine-conjugated phalloidin (1:400) for 1 h and mounted in Vectashield with 4′,6-diamidino-2-phenylindole. Acquisition of the images was performed on a Leica DM600B equipped with a numerical camera CCD Micromax (Princeton Instrument) at ×20 (N.A. 0.7, PL APO) or ×40 (N.A. 1.25, oil, PL APO) using Metamorph software. In experiments with EC, only cells positive for CD31 or CD102 were selected for analysis of immunofluorescence with antibodies recognizing other proteins.

Proliferation Studies

MEFs were plated separately in triplicate at 7500 cells/well in 24-well plates. The growing medium was changed every other day, and cells were fixed every day with cold methanol and kept at −20 °C until the end of the experiment. At that time, cells were stained with crystal violet staining and solubilized (24), and cell number was estimated with a spectrophotometer at 590 nm. Data points were the average of triplicates.

RESULTS AND DISCUSSION

Generation of fakΔ Mice Expressing an Active but Autophosphorylation-deficient Mutant Form of FAK

Mutant mice (fakΔ) were generated from embryonic stem cells in which FAK exon 15 (21) was deleted by homologous recombination (supplemental Fig. 1, A and B). Analyses of tissues from heterozygous FAK^+/Δ mice by RT-PCR (data not shown) and Western blotting (supplemental Fig. 1C) confirmed that the deletion of exon 15 did not alter FAK mRNA open reading frame, resulting in the expression of the expected full-length mutant protein FAKΔ deleted of the 19 residues (including Tyr-397) coded by exon 15 and located between the N-terminal four-point-one, ezrin, radixin, moesin (FERM), and kinase domains (21) (Fig. 1A). FAKΔ was expressed at the same levels as normal FAK encoded by the wild type allele (supplemental Fig. 1C). As expected, FAKΔ was not detected by Tyr(P)-397-specific antibodies in transfection experiments. Using a transphosphorylation assay in COS7 cells (23), we found that the basal kinase activity of FAKΔ was moderately increased (supplemental Fig. 2). Thus fakΔ mice constitute an interesting model to study the role of FAK autophosphorylation in vivo.

FIGURE 1.

Late defects and lethality in FAK^Δ/Δ embryos. A, schematic FAK and FAKΔ structure showing the N-terminal FERM domain, the kinase domain, and the C-terminal focal adhesion targeting (FAT) domain, as well as proline-rich sequences (PR1–3). FAKΔ lacks 19 amino acids in the FERM-kinase linker, including Tyr-397. B, comparison of the in utero survival curves of FAK-null and FAK^Δ/Δ embryos (red inverted triangles). The data concerning FAK-null embryos are from Ref. 26 (FAK^−/−, green squares) and from Ref. 5 (CFKO, magenta triangles). C, appearance of FAK^+/+ and FAK^Δ/Δ embryos at E13.5 and E15.5 (arrows, hemorrhages; arrowheads, edemas). Gross morphology defects of E14.5 FAK^+/+ and FAK^Δ/Δ lungs, heart, and liver (scale bars, 1 mm). E14.5 (upper panel) and E15.5 (lower panel) anterior (left) and posterior (right) footplates (scale bars, 1 mm). D, FAK^Δ/Δ embryos show delayed artery formation and vascular remodeling defects. Whole-mount triple immunofluorescence confocal microscopy using antibodies to smooth muscle actin (SMA, green), PECAM1 (red), and neurons (TUJ1) (blue) on skin from E14.5 embryos. Confocal microscopy images were acquired using ×10 magnification.

FAK^Δ/Δ Mice Display Multiple Abnormalities after Mid-gestation and Lethality between E14.5 and E16.5

The intercrossing of FAK^+/Δ mice produced only 0.25% FAK^Δ/Δ pups (Table 1) compared with the expected Mendelian ratio of 25%. This result suggested that the dizygous expression of FAKΔ was lethal for the vast majority of FAK^Δ/Δ embryos. To define the stage at which the FAK^Δ/Δ embryos died and to compare their phenotype to FAK^+/+ embryos, dated embryos were collected during development and genotyped by PCR. At E11.5 and E12.5, living FAK^Δ/Δ embryos were found at Mendelian ratios (Table 1 and Fig. 1B). They were indistinguishable from their control littermates in morphology and development and showed normal vascularization by whole-mount anti-PECAM1 immunostaining (supplemental Fig. 3A and data not shown). At E13.5, FAK^Δ/Δ embryos were smaller than FAK^+/+ embryos and showed abnormal superficial vasculature, hemorrhagic liver, multiple body hemorrhages, and edemas (Fig. 1C), indicating that FAK autophosphorylation was required for normal development after E12.5. A decrease in the ratios of living FAK^Δ/Δ embryos compared with wild type was observed between E13.5 and E15.5, and very few survived past E16.5 (Table 1 and Fig. 1B). FAK^Δ/Δ embryos also showed alteration and retardation in heart, lung, liver, and limbs development at E14.5 (Fig. 1C), and yolk sac vascularization defects were observed at E15.5 (supplemental Fig. 3B). Because vascular defects were observed in FAK^Δ/Δ embryos from E13.5, we further investigated vascular development by performing skin whole-mount immunofluorescence from E14.5 embryos. Smooth muscle actin staining was visible on arteries formed in the skin of wild type embryos, whereas FAK^Δ/Δ arteries only showed weak smooth muscle actin staining in the proximal part. In addition, there was a defect in the remodeling of the vascular plexus in FAK^Δ/Δ embryos (Fig. 1D). These results demonstrate that the homozygous FAK mutation deleting the autophosphorylation site resulted in vascular defects and developmental retardation apparent at E13.5 and embryonic lethality between E14.5 and E16.5. Although the stage of embryonic lethality caused by the early conditional deletion of FAK in endothelial cells remains controversial (5, 10), one study reported lethality between E13.5 and E14.5 with a phenotype very similar to what we observed in FAK^Δ/Δ embryos (5). Thus, it is likely that vascular abnormalities were an important factor in embryonic lethality, although we cannot exclude that failure of other organs also played a role.

TABLE 1.

Genotype of embryos obtained from crosses between FAK^+/Δ mice

The total number of embryos (alive or dead) of each genotype obtained from crosses between (FAK^+/Δ) mice is indicated at various embryonic ages and birth. The number of dead embryos is indicated in parentheses.

	Embryonic day
	11.5	12.5	13.5	14.5	15.5	16.5	18.5	Born
+/+	20 (1)	11 (1)	22	30	18 (1)	16	10	213
+/Δ	46 (2)	22	78	84 (5)	51 (1)	39 (1)	17	416
Δ/Δ	26 (7)	10	33 (11)	42 (17)	11 (5)	7 (4)	1 (1)	3
No. of litters	12	5	18	21	11	6	5	94

The mid-gestation lethality of FAK^Δ/Δ embryos contrasted with the early developmental lethality at E8.5 reported in FAK-null embryos (4, 26) even in the same genetic background as fakΔ mice (5). Comparison of the survival of FAK^Δ/Δ and FAK^−/− embryos during development confirmed that the absence of FAK led to an earlier developmental lethality than the absence of FAK autophosphorylation (Fig. 1B). Thus, FAK requirement for normal development is independent of its autophosphorylation until E13.5. Moreover, we observed rare (0.25%) verified FAK^Δ/Δ mice, which survived to adulthood and were fertile (Table 1 and supplemental Fig. 4), whereas no FAK-null adult mice have been reported (4, 5, 26). This observation demonstrates that the absence of FAK autophosphorylation can be compensated in very rare cases in vivo.

FAK^Δ/Δ Mutation Has Minor Effect on the Levels of Associated Proteins

Next we monitored the expression of FAKΔ and FAK-associated proteins in mutant embryos. FAKΔ protein was expressed at the same level as FAK in FAK^+/Δ embryos, or in FAK^Δ/Δ embryos compared with FAK^+/+ embryos at every developmental stage tested, but as expected, only FAK was phosphorylated on Tyr-397 (Fig. 2A and data not shown). Because the cytoskeleton rearrangements and downstream signaling pathways regulated by FAK are mediated by its interaction with SFKs and focal adhesions proteins such as paxillin and p130Cas (17, 18), we also monitored their expression. We found in E14.5 mutant embryos a moderate increase in the expression of paxillin and p130Cas, as well as cortactin (Fig. 2A), which facilitates cortical actin polymerization (27). In contrast, vinculin, another focal adhesion protein anchoring F-actin to the membrane, and Src and Fyn were expressed at the same levels in FAK^+/+ and FAK^Δ/Δ embryos (Fig. 2A). The FAK-related tyrosine kinase Pyk2 was barely detectable in both FAK^+/+ and FAK^Δ/Δ embryos, and no increase was observed in mutant mice at any developmental stage tested (data not shown). Interestingly, a very strong increased expression of Pyk2 and paxillin has been reported in some FAK-null tissues or cell types (11, 20, 25, 28, 29), and Pyk2 overexpression can functionally compensate FAK absence (11, 25, 28, 29) and/or have unrelated deleterious consequences (28). Based on these results, it has been proposed that FAK maintains a low level of Pyk2 expression in normal conditions in some tissues or cell types (11, 28). The lack of Pyk2 up-regulation in FAK^Δ/Δ embryos suggests that FAK autophosphorylation is not required to maintain a low level of Pyk2 expression during development in vivo. Our results also support the hypothesis that Pyk2 overexpression might contribute to the severity of FAK-null mutation (20, 28).

FIGURE 2.

Expression of FAK-associated proteins in tissue from FAK^Δ/Δ embryos and mutant MEFs. A, immunoblotting of phospho-Tyr-397 (pY397), FAK, vinculin (Vinc), p130Cas (Cas), paxillin (Pax), cortactin (Cort), Src, and Fyn in liver lysates from FAK^+/+, FAK^+/Δ, and FAK^Δ/Δ E14.5 littermate embryos. B, immunoblotting of the same proteins as in A, Pyk2, and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) in MEFs derived from E12.5 FAK^+/+ and FAK^Δ/Δ littermate embryos. C, immunoblotting for p53 in MEFs as in B. Lysates from COS7 cells expressing a fusion luciferase-p53 protein were used as a control for p53 immunoreactivity.

To characterize the cellular consequences of the mutation on the regulation of associated proteins, we studied primary MEFs from E12.5 FAK^+/+ and FAK^Δ/Δ embryos. Western blotting experiments showed that FAKΔ was expressed at the same levels in FAK^Δ/Δ MEFs as FAK in FAK^+/+ MEFs (Fig. 2B). Only a moderate increase (<2-fold) in the level of Pyk2 was observed in these cells compared with FAK^+/+ MEFs (Fig. 2B). This result contrasted with the strong overexpression of Pyk2 that we observed in FAK^−/− fibroblasts (∼10-fold, data not shown) and that had also been reported previously (25, 28, 29). The levels of vinculin, p130Cas, paxillin, cortactin, Fyn, and Src were not changed by the fak^Δ/Δ mutation in MEFs (Fig. 2B). Finally, similarly to FAK^+/+ MEFs, no p53 was detected in FAK^Δ/Δ MEFs (Fig. 2C) in contrast to its strong expression in primary FAK^−/− fibroblasts (20). Thus, FAK^Δ/Δ MEFs are an interesting model to study the effects of the FAK mutant lacking Tyr-397, in a context where its functionally related proteins and partners are expressed at levels close to normal.

Dramatic Alteration of the Phosphorylation Response to Cell Adhesion in FAK^Δ/Δ MEFs

We tested the effects of integrin engagement on the phosphorylation of FAK tyrosine residues in FAK^Δ/Δ MEFs. As expected, plating wild type MEFs on fibronectin (FN) induced the phosphorylation of wild type FAK on Tyr-397, Tyr-576/577, which allows its full enzymatic activation, and on Tyr-925, which regulates the recruitment of partners and downstream signaling pathways, as well as FAK localization to focal adhesions (Fig. 3, A and B) (16, 30). In contrast, the FN-induced phosphorylation of these residues was lost in FAK^Δ/Δ MEFs (Fig. 3, A and B), although SFKs were similarly activated in wild type and mutant MEFs (data not shown). These results show that autophosphorylation on Tyr-397 is critical for SFK-dependent tyrosine phosphorylation of FAK and confirm that the moderate increase of FAKΔ basal kinase activity (supplemental Fig. 2) has no effect on FAK activation and Src-mediated signaling. We also examined p130Cas and paxillin, whose phosphorylation by the FAK·SFK complex is important for cytoskeleton rearrangement and activation of downstream signaling pathways (17). The phosphorylation of both proteins was stimulated by FN in FAK^+/+ but not in FAK^Δ/Δ MEFs (Fig. 3, C and D) despite a similar interaction of endogenous FAK and FAKΔ with both p130Cas and paxillin in these cells or after overexpression in COS7 (data not shown). In contrast, the phosphorylation of cortactin, which is known to be catalyzed by SFKs independently of FAK (27), was unaffected in FAK^Δ/Δ cells (Fig. 3E). Previous studies proposed that the phosphorylation of p130Cas and paxillin by Src may be achieved through both FAK-dependent and -independent mechanisms (29, 31). However, the absence of any significant phosphorylation of p130Cas and paxillin in FAK^Δ/Δ cells plated on fibronectin suggests that FAK autophosphorylation is required for targeting SFKs to p130Cas and paxillin in response to integrin engagement.

FIGURE 3.

Alteration of the phosphorylation response to integrin engagement in FAK^Δ/Δ MEFs. MEFs were plated on polylysine (Pl) or fibronectin (Fn) for 30 min before lysis and immunoblottings for phospho-Tyr-397, -576, -577, -925, and total FAK (A) and quantified (B). Immunoblotting of tyrosine-phosphorylated (p, upper panels) and total levels (middle panels) of p130Cas (Cas, C), paxillin (Pax, D), and cortactin (Cort, E) in MEFs plated as in A. These experiments have been repeated with independent FAK^+/+ or FAK^Δ/Δ (n = 3 for each genotype) MEFs populations established from different littermate embryos. Quantified data (means ± S.E. n = 3) were analyzed using two-way analysis of variance. Bonferroni post-test: *, p < 0.05; **, p < 0.01, ***, p < 0.001.

Spreading and Morphology Are Altered in FAK^Δ/Δ Fibroblasts and Endothelial Cells

Integrin engagement mediates the adhesion of many cell types and induces organized actin polymerization resulting in the extension of the membrane (lamellipodia) at the periphery of spreading cells. Because FAK plays a critical role in adhesion and spreading (4, 10, 28), we investigated the consequences of exon 15 deletion on these processes. After 30 min on polylysine or FN, FAK^Δ/Δ cells showed delayed spreading compared with FAK^+/+ MEFs and lacked continuous smooth and round peripheral membrane extensions, characteristic of lamellipodia (Fig. 4A). Instead, FAK^Δ/Δ cells exhibited multiple thinner membrane protrusions containing dense actin fibers (Fig. 4A). These narrow actin-rich membrane extensions of FAK^Δ/Δ MEFs remained prominent after 1 h on FN and were identified as abnormal lamellipodia because cortactin, which normally localizes at the edges of protruding lamellipodia, was localized at their tips (Fig. 4B). A similar phenotype was reported in several cell types lacking FAK expression (10, 32, 33). Lamellipodia formation and cell spreading in response to integrin engagement have been reported to require the release by FAK of the Arp2/3 complex bound to its FERM domain, triggered by autophosphorylation of Tyr-397 (32). Thus, the defects observed in FAK^Δ/Δ cells may be due to the disruption of this mechanism. After 2 h on FN, immunostaining of both cell types for vinculin, a protein localized at focal adhesions, showed an increase in the number and size of focal adhesions in FAK^Δ/Δ MEFs compared with wild type cells (Fig. 4C). Both FAK^+/+ and FAK^Δ/Δ MEFs acquired an apparent typical fibroblastic stellate morphology with radial actin stress fibers after 20 h on polylysine or FN (Fig. 4D). However, FAK^Δ/Δ MEFs lacked the large lamellipodia observed in wild type. Furthermore, they displayed a persistent increase in the number and size of focal adhesions localized at the tips of stress fibers, characterized by a stronger enrichment of FAK immunoreactivity than in FAK^+/+ MEFs (Fig. 4D). Because FAK^Δ/Δ embryos showed important vascularization defects, we also monitored the properties of primary FAK^Δ/Δ EC in response to integrin engagement. After 2 h on FN, abnormal lamellipodia and cell spreading defects were observed in mutant EC that showed enlarged focal adhesions enriched in FAKΔ (Fig. 4, E and F). Interestingly, a similar phenotype was reported in primary FAK^−/− EC (10), which also showed no increased expression of Pyk2. The enrichment of FAKΔ compared with FAK at focal adhesions was also observed following its expression by transfection in FAK^−/− fibroblasts (data not shown) similarly to the mutant FAKY397F overexpressed in FAK^−/− cells (29). Because the phosphorylation of Tyr-925 is important for the release of FAK from focal adhesions and their turnover (30), the deficit in FAK Tyr-925 phosphorylation that we observed in FAK^Δ/Δ cells (Fig. 3, A and B) may contribute to the increased number and size of focal adhesions and their enrichment in FAKΔ protein (Fig. 4, A–F). The stellate morphology of FAK^Δ/Δ MEFs contrasts with the ultraspread morphology of FAK^−/− fibroblasts stably expressing FAKY397F in the same experimental conditions (29). The high expression of Pyk2 and FAKY397F in these cells may be responsible for their different phenotypes compared with FAK^Δ/Δ MEFs that contain close to normal levels of Pyk2 and FAKΔ. Our results also demonstrate that FAK^Δ/Δ cells have a delayed spreading characterized by disorganized lamellipodia, which never reach the same extent as in FAK^+/+ MEFs, and an increased number and size of focal adhesions, enriched in FAKΔ. These defects, also observed in the mutant endothelial cells, may underlie the vascular abnormalities observed in FAK^Δ/Δ embryos.

FIGURE 4.

Delayed spreading and abnormal lamellipodia in FAK^Δ/Δ fibroblasts and endothelial cells. FAK^+/+ and FAK^Δ/Δ MEFs or EC were plated on polylysine (Pl) or fibronectin (Fn) for 0.5–20 h as indicated. Cells were stained for actin with rhodamine-conjugated phalloidin (A–D and F) and/or antibodies recognizing FAK (A, D, and E), cortactin (B and F), and vinculin (C and F) and revealed with fluorescein isothiocyanate-conjugated secondary antibody. EC cells (E and F) were identified by CD31 immunostaining. Lamellipodia were well developed in wild type (arrows) but not mutant cells (A, B and D–F), which displayed narrow cortactin-positive protrusions (B and F). Focal adhesions were more numerous and larger in FAK^Δ/Δ than wild type cells (C–F). Scale bars, 20 μm.

Proliferation of FAK^Δ/Δ Fibroblasts Is Decreased

FAK is implicated in the control of cell growth (28, 34), although its importance depends on cell types (19, 20, 35). We compared the proliferation of FAK^+/+ and FAK^Δ/Δ MEFs in standard culture conditions (10% serum). No differences in proliferation between wild type and mutant MEFs were observed during the first 3 days. Thereafter, FAK^Δ/Δ MEFs proliferated more slowly than FAK^+/+, and they reached a lower saturation density (Fig. 5A). These results demonstrate that FAK autophosphorylation on Tyr-397 is required for normal proliferation of MEFs (Fig. 5A). However, growth of FAK^Δ/Δ MEFs was mostly restored at a higher serum concentration (Fig. 5B), suggesting that the proliferative defect due to the lack of FAK autophosphorylation could be compensated by increased proliferative signals. FAK^−/− embryonic mesodermal cells in vivo or FAK^−/− fibroblasts in vitro also exhibited a proliferation defect attributed to an up-regulation of p53 (20, 36). This up-regulation may account for the use of a p53^−/− background to establish FAK^−/− fibroblasts (4). Indeed, the FAK FERM domain binds to and causes the degradation of p53, independently of its autophosphorylation and kinase activity (20, 37). However, we showed that p53 was not increased in FAK^Δ/Δ MEFs (Fig. 2C). Thus, the lack of FAK autophosphorylation is likely to inhibit cell proliferation by an alternative pathway.

FIGURE 5.

Proliferation is slowed in FAK^Δ/Δ MEFs. A, independent MEFs populations (n = 3–4 for each genotype) established from littermate embryos were plated separately in triplicate (7500 cells/well) and grown in 10% serum for the indicated number of days. Cell number was evaluated by crystal violet staining, and the day after plating was considered as D = 0. B, FAK^+/+ and FAK^Δ/Δ MEFs populations were grown in the indicated serum concentrations and counted by crystal violet staining at D = 0 or after 7 days in culture (D = 7). Data (means ± S.E. n = 3–4) were analyzed using two-way analysis of variance. Bonferroni post-test: **, p < 0.01; ***, p < 0.001.

The contrast between the phenotype of FAK^−/− and FAK^Δ/Δ embryos provides strong evidence for autophosphorylation-independent functions of FAK in vivo. Such functions have been suggested in cultured cells, in addition to the aforementioned regulation of p53 expression. For example, pharmacological inhibition of FAK did not block proliferation of tumor cells and did not block their apoptosis in vitro (19). Durotaxis, the ability of cells, cultured on a substrate of graded stiffness, to move from softer to stiffer regions, was abolished in FAK^−/− cells and rescued by FAKY397F, which did not rescue the migration speed (38). Furthermore, abnormal axonal branching in hippocampal FAK^−/− neurons was partially rescued by the expression of FAKY397F (7). Although very few bona fide substrates of FAK have been characterized in intact cells (1), the respective role of FAK catalytic activity and scaffolding properties will have to be determined in these autophosphorylation-independent functions. Autophosphorylation-independent function of FAK is also supported by the observation that although Tyr-397 is highly conserved in most metazoans, it is not found in Caenorhabditis elegans (21).

In conclusion, the study of FAK^Δ/Δ mice provides important information about the role of FAK autophosphorylation on Tyr-397 in vivo. FAK autophosphorylation is not required for development before E13.5 but becomes necessary past that stage. The autophosphorylation of FAK on Tyr-397 is not necessary for general cell morphology in standard culture conditions or to repress p53 and Pyk2 expression. In contrast, it plays an important role in focal adhesion turnover, lamellipodia formation, cell spreading, and proliferation. Thus, our results demonstrate that the physiological functions of FAK in vivo are achieved through both autophosphorylation-independent and autophosphorylation-dependent mechanisms and that the requirements for these mechanisms vary during development. They underline that identification of the mechanisms by which FAK regulates different cellular functions will be important to improve the design of appropriate therapeutic tools.