Targeted disruption of the focal adhesion kinase (FAK) gene in mice is lethal at embryonic day 8.5 (E8.5). Vascular defects in FAK-/- mice result from the inability of FAK-deficient endothelial cells to organize themselves into vascular network. We found that, although fibronectin (FN) levels were similar, its organization was less fibrillar in both FAK-/- endothelial cells and mesoderm of E8.5 FAK-/- embryos, as well as in mouse embryonic fibroblasts isolated from mutant embryos. FAK catalytic activity, proline-rich domains, and location in focal contacts were all required for proper allocation and patterning of FN matrix. Cells lacking FAK in focal adhesions fail to translocate supramolecular complexes of integrin-bound FN and focal adhesion proteins along actin filaments to form mature fibrillar adhesions. Taken together, our data suggest that proper FN allocation and organization are dependent on FAK-mediated remodeling of focal adhesions.
The glycoprotein fibronectin (FN) is a major component of the extracellular matrix (ECM) and is essential for normal development. Directional migration of mesodermal cells during early development requires positional information provided by deposits of FN matrix (Boucaut et al., 1990; Winklbauer and Nagel, 1991; Yost, 1992). Furthermore, disruption of the Fn gene in mice results in embryonic lethality. Fn-/- embryos die at E8.5-9.0 with extensive defects in the mesoderm, neural tube and cardiovascular network (George et al., 1993). Very similar defects have been described in mouse embryos deficient for the nonreceptor protein tyrosine kinase FAK (focal adhesion kinase) (Furuta et al., 1995; Ilić et al., 1995).
Focal adhesions are sites where cells interact with ECM through integrin receptors, and FAK is both a structural and an enzymatic component of focal adhesions, sites where cells interact with ECM through integrin receptors. Integrin clustering upon binding to ECM components, such as FN, results in enhanced tyrosine phosphorylation of cellular proteins. Integrins do not possess intrinsic enzymatic activity; they transduce signals following ECM binding by triggering the formation of multimolecular complexes that contain nonreceptor protein tyrosine kinases as well as cytoskeletal components. Once activated, FAK, in turn, can associate with other kinases (e.g. PI 3-kinase, Src family kinases) and adaptor proteins (e.g. Grb2, p130Cas [Cas]), enabling integrins to couple to multiple intracellular signaling pathways (Schlaepfer et al., 1999; Schaller, 2001).
Integrin binding to ECM is not only required for transduction of signals from matrix to cells but also initiates responses within the cell that make it possible for the cell to organize a fibrillar FN matrix (Wu et al., 1995). Inside-out signaling through the integrin cytoplasmic domain incites a conformational change in the integrin ligand-binding domain that switches integrin affinity from a low to a high state (Takagi et al., 2001). Such integrin activation is necessary for effective binding of FN but is still not sufficient for FN matrix assembly. Subsequent events, in which the cytoskeleton plays a crucial role, lead to exposure of an active assembly site in the FN molecule (Wu et al., 1995; Zhong et al., 1998; Sechler et al., 2001). In particular, the organization and insertion of oriented actin filaments (stress fibers) into focal adhesion sites, where integrins bind to ECM, allow cells to exert tension on the surrounding ECM. This tension is necessary to expose a cryptic site(s) in FN that influences the positioning, rate and orientation of FN matrix assembly (Zhong et al., 1998; Schoenwaelder and Burridge, 1999; Sechler et al., 2001). Newly assembled FN fibrils coalign with bundles of actin filaments at focal adhesions at the cell-substrate interface and at other types of ECM adhesion sites (e.g., fibrillar adhesions) that form on the dorsal surfaces of cells or between cells (Hynes and Destree, 1978; Ohashi et al., 1999; Katz et al., 2000; Pankov et al., 2000; Cukierman et al., 2001).
Given the locations of FN and FAK on opposite sides of the cell membrane at focal adhesion and ECM adhesion sites, it is perhaps not surprising that the phenotypes following in vivo gene targeting in mice, are so similar. This similarity strongly supports a tight functional linkage of these two molecules. The question of how signals from FN are transduced in the absence of FAK has been the focus of several studies (Sieg et al., 1998; Sieg et al., 1999). This is the first report to approach a FN-FAK signaling axis from the other perspective: using FAK-null cells to determine how the absence of FAK affects the FN matrix.
Materials and Methods
Heterozygous lines of FAK and Fn mice were housed and bred in an environmentally controlled room at UCSF according to institutional guidelines. Genotypes were determined by PCR analysis (Furuta et al., 1995; George et al., 1993). Embryos were isolated at E8.5. Placental cone and yolk sac were used for genotyping, and the embryo proper was fixed and embedded or snap frozen and kept at -80°C for subsequent analyses.
Cells and cell culture
Isolation and culture of p53-immortalized and polyoma middle T-immortalized FAK+/+ and FAK-/- endothelial cells were described earlier (Ilić et al., 1998; Ilić et al., 2003). FAK-/- and FAK+/+ cells, derived from E8.5 embryos that were also deficient in p53 function, were described initially by Ilić et al. (Ilić et al., 1995). DA2 cells are FAK-/- cells in which full-length FAK has been expressed (Sieg et al., 1999). CA3, CB4, SE6 and SX4 (Klingbeil et al., 2001) are FAK-/- cells in which chimeras composed of the C-terminal domain from FAK and the N-terminal and kinase domains from the closely related PYK2 have been expressed. SE6 and SX4 cells also have a mutation in the paxillin-binding domain that is required for localization of the chimeric proteins in focal contacts. FAK-deficient and control endothelial cells were also described previously (Ilić et al., 1998; Illić et al., 2003). All fibroblastic cell lines were cultured in complete medium: DME high glucose containing 10% FCS, nonessential amino acids, sodium pyruvate, antibiotics and 0.1 mM β-mercaptoethanol.
Antibodies and chemicals
Anti-human FN, anti-tensin, anti-β-catenin and anti-Skp1 monoclonal antibody (mAb) were purchased from BD Transduction Lab; anti-β1 integrin rat mAb 9EG7 from BD Pharmingen; anti-paxillin ZO35 mAb from Zymed; anti-human FN polyclonal antibody (pAb) and anti-actin AC-40 mAb from Sigma-Aldrich; anti-mouse FN pAb from Biogenesis; anti-FAK rabbit C-20 pAb from Santa Cruz Biotech, and phospho-specific anti-pY397 and anti-pY861 FAK from BioSource International. Goat anti-rabbit IgG-conjugated with 15 nm gold particles was from British BioCell International. All other secondary Abs and ChromPure donkey whole IgG were from Jackson ImmunoResearch Lab. Human plasma FN was from Roche, and Hoechst 33342 from Molecular Probes.
Embryos were fixed in 3.8% paraformaldehyde (PFA)/PBS overnight, washed in PBS, and infiltrated with 5-15% sucrose followed by optimal cutting temperature (OCT) compound (Miles Scientific), and frozen in liquid nitrogen. Five μm thick sections were prepared using a cryostat (Slee International), incubated with whole donkey IgG and anti-FN pAb (Sigma). After washing, sections were treated for 30 minutes with rhodamine-conjugated secondary Ab and Hoechst 33342. For electron microscopy embryos were prepared as described previously (Johkura et al., 2001).
Live cells cultured as a monolayer, or wounded monolayer for visualizing a trail of FN left on the substratum after migration, were washed in cold PBS and incubated with a mixture of 5 μg/ml anti-β1 integrin 9EG7 rat mAb, 1:200 diluted rabbit anti-FN pAb, and 280 μg/ml ChromPure donkey whole IgG for 1 hour on ice. Cells were washed with cold PBS and fixed in 3.8% PFA/PBS. The cells were incubated for 45 minutes with 5 μg/ml each of FITC-conjugated donkey anti-rat IgG, rhodamine X-conjugated anti-rabbit IgG, and 10 μg/ml Hoechst 33342. Cells were washed with PBS, mounted in Vectashield (Vector), and analyzed using a Zeiss Axiophot epifluorescence microscope equipped with appropriate filters. Images were photographed with a Spot CCD camera (Diagnostic Instruments, Inc.) and processed using Adobe Photoshop 7.0 software.
Co-staining of FN/β-catenin in permeabilized cells started with fixation in 3.8% PFA/PBS and permeabilization with cold acetone for 10 minutes before appropriate primary Abs were added. For immunolocalization of tensin, cells were fixed and permeabilized at the same time with 0.2% Triton X-100/3.8% PFA for 2 minutes, and then fixed for a further 18 minutes with 3.8% PFA only.
Antibody-chase experiments were performed as described by Pankov et al. (Pankov et al., 2000). Briefly, live cells were incubated in warm medium containing 5 μg/ml anti-β1 integrin 9EG7 rat mAb (BD Pharmingen) and 0.2 mg/ml ChromPure donkey whole IgG for 45 minutes. Cells were then washed three times and either fixed in 3.8% PFA/PBS or further cultured in medium without antibodies for 2 hours and then fixed. Samples were incubated for 45 minutes with rabbit anti-FN pAb (Sigma) at a dilution of 1:250 and 5 μg/ml FITC-conjugated donkey anti-rat IgG. After washing, cells were permeabilized for 10 minutes with cold acetone, blocked with avidin/biotin blocking kit (Vector), and incubated with 5 μg/ml mouse anti-paxillin ZO35 and 0.2 mg/ml ChromPure donkey whole IgG antibodies for 1 hour. Cells were then washed and incubated with 5 μg/ml each of rhodamin-X-conjugated donkey anti-rabbit IgG and biotin-conjugated donkey anti-mouse IgG for 45 minutes. After washing, cells were finally incubated with avidin/AMCA conjugate (Vector) for 15 minutes, washed again and mounted in Vectashield (Vector).
Actin stress fibers were stained by incubating fixed (3.8% PFA, 20 minutes) and permeabilized (cold acetone, 10 minutes) cells with phalloidin/rhodamine (Molecular Probes) for 30 minutes.
Northern and western blotting
Northern and western blotting were performed essentially as described previously (Kovačič et al., 2001).
One hour after plating on FN-coated tissue culture dishes in medium containing serum, cells were washed twice with methionine-, cysteine-, and L-glutamine-free MEM (Sigma-Aldrich). L-[35S]methionine and L-[35S]cysteine labeling mixture (ARC) was added at 1.0 mCi/ml in growth medium with methionine and cysteine at 10% of the normal level, and cells were exposed to the labeling medium for 15 minutes. Cells were then washed with normal medium and either lysed immediately or cultured for a further 2 or 15 hours. At the 2- and 15-hour time points the medium was collected and 0.1% SDS and protease inhibitors were added (0.5 M PMSF, 1 μg/ml leupeptin, 2 μg/ml aprotinin). Cells were lysed for 30 minutes in buffer L (150 mM NaCl, 50 mM Tris, pH 7.4, 1 mM EDTA, 1 mM EGTA, 10% glycerol, 0.5 M PMSF, 1 μg/ml leupeptin, 2 μg/ml aprotinin) containing 1% Triton X-100 and 1% deoxycholate (DOC). Lysates were centrifuged for 10 minutes at 15,000 g at 4°C. Supernatants were transferred into new tubes, whereas insoluble pellets were boiled for 10 minutes in buffer containing 1% SDS. After boiling, lysates were diluted 10-fold with lysis buffer without SDS. FN was immunoprecipitated overnight with 1 μg/ml anti-FN mAb. Precipitates were collected with sheep anti-mouse-conjugated Dynabeads (Dynal Biotech Inc.) and analyzed by SDS-PAGE using 6% polyacrylamide gels. Gels were dried and exposed to BioMax MS Imaging Film (Eastman Kodak). Bands were quantified by densitometry.
Production of adenoviruses used in this study was described elsewhere (Streblow et al., 1999; Hsia et al., 2003). FAK+/+ and FAK-/- cells were suspended in complete medium and infected at a MOI of 10 or 30 plaque-forming units/cell. After 30 minutes cells were plated on FN-coated dishes (10 μg/ml) and cultured in complete medium. All analyses were performed 48 hours post-infection.
In vitro actin polymerization assay
Cells were trypsinized, washed and resuspended in a buffer containing 10 mM Hepes pH 7.6, 100 mM KCl, 1 mM MnCl2, 0.1 mM EDTA, 1 mM DTT, 0.5 M PMSF, 1 μg/ml leupeptin and 2 μg/ml aprotinin. Cells were then broken using a probe sonicator and lysate was centrifuged at 3000 g for 30 minutes. Supernatant was removed from nuclear pellet, centrifuged at 40,000 g for 1 hour at 4°C, and concentrated to 10 mg/ml using Amicon-10 ultrafiltration cells (Milipore). High-speed extract was supplemented with an energy-regenerating mix (1 mM ATP, 1.25 mM MgCl2, 7.5 mM creatine phosphate) and stored at -80°C. Actin polymerization was assayed as described previously (Ma et al., 1998). Briefly, cell lysates were mixed with 1 μM pyrene-labeled rabbit skeletal muscle actin (Cytoskeleton). Fluorescence was monitored in 80 μl aliquots using a spectrofluorometer.
Endothelial cells and mouse embryos lacking FAK have altered FN organization
At the time of death at E8.5 FAK-/- mouse embryos have severely impaired mesodermal development, including defective vasculogenesis. Gene expression and in vitro differentiation studies revealed that the vascular defects result from the inability of FAK-/- endothelial cells to organize vascular networks, rather than from defects in tissue-specific differentiation (Ilić et al., 2003). Because of the importance of FN in directing migration during development, we compared FN matrix organization on FAK+/+ and FAK-/- endothelial cells by immunostaining. Compared with FAK+/+ controls, FN in permeabilized FAK-/- endothelial cells had a granular appearance and displayed an abundant, poorly organized array of short, thin fibrils. This pattern was evident regardless of endothelial cell origin (embryos or embryoid bodies) or immortalization method (p53-deficiency or polyoma middle T-mediated transformation). These findings suggested a link between altered FN organization and an inability of FAK-/- endothelial cells to support normal blood vessel morphogenesis (Fig. 1A).
To verify that the altered FN matrix organization found in FAK-/- endothelial cells was not a phenomenon observed only in cultured cells, we first evaluated the presence and organization of the FN matrix in vivo, by staining for FN on sections of E8.5 wild-type and FAK-null littermate embryos. At the light microscope level, strong staining for FN was detected in the mesoderm (Fig. 1B, arrows), and at the ectodermal-mesodermal interface (Fig. 1 B, arrowheads) in both mutant and wild-type embryos. The wild-type mesoderm contained extensive matrix-filled interstitial spaces with intense fibrillar staining for FN. In contrast, FN matrix in the mesoderm of FAK-/- mutants was comparatively sparse and had a more patchy, granular appearance (Fig. 1B).
The altered appearance of FN matrix was confirmed by immuno-electron microscopy on sections of normal and mutant E8.5 embryos (Fig. 1C). Amorphous patches, positive for FN, were frequently observed in the mesenchymal matrix or on cell surfaces in sections of FAK-/- mutant embryos. In contrast, FN was detected in a fibrillar pattern along cellular processes in sections of wild-type cells. The specificity of staining for FN was confirmed by the absence of any signal on sections of Fn-/- embryos.
To assess whether these differences were due to changes in FN expression levels, we performed northern and western blot analyses on E8.5 littermate embryos from matings of FAK heterozygous mice. The specificities of the FAK and FN bands in the western blots were verified by their absence in FAK-null and FN-null littermates, respectively. A member of the F-box family of proteins S phase kinase-associated protein-1 (Skp1) was used as a loading control. We found that FAK wild-type (+/+), FAK heterozygous (+/-) and FAK null (-/-) embryos had very similar levels of FN mRNA and protein (Fig. 1D,E). Taken together, these data suggested that the loss of FAK primarily influences matrix organization or deposition rather than Fn gene function. To analyze further this possibility we switched to using cells in culture obtained from FAK-/- p53-/- and control FAK+/+ p53-/- E8.5 embryos [referred to subsequently as FAK-/- and FAK+/+ cells, respectively (Ilić et al., 1995)]. These cells are of mesenchymal origin, since they express vimentin and lack E-cadherin. The p53-null mutation was introduced to enable propagation of primary FAK-/- cells (Tsukada et al., 1993: Ilić et al., 1995; Ilić et al., 1998).
FN matrix organization and allocation to the apical surface of cells, but not synthesis, are impaired in FAK-/- cells
We examined FN matrix organization in several ways using FAK+/+ cells, FAK-/- cells and FAK-/- cells transfected with, and re-expressing, full-length FAK [referred to as DA2 (Sieg et al., 1999) cells]. Immunostaining of FN and β1 integrin on live cells in culture showed extensive fibrillar staining for FN on the dorsal (top) surfaces of both FAK+/+ and DA2 cells. Dorsal surfaces of FAK-/- cells had greatly reduced staining for both FN and β1 integrin. Interestingly, strong staining for β1 integrin was detected in the high density of punctate focal adhesions on the ventral cell surfaces of FAK-/- cells in contact with the substratum (Fig. 2A). FN staining under cells was barely detectable in these non-permeabilized cultures. Following permeabilization, however, there was abundant FN immunostaining at the ventral surface of FAK-/- cells (Fig. 2B, upper panel). The FN pattern appeared to be punctate or to follow cell-cell borders as shown with β-catenin co-staining (Fig. 2B, lower panel). To compare the pattern of matrix deposition by FAK+/+ and FAK-/- cells in another way, we analyzed the FN trails left behind migrating cells (Fig. 2C). Although the number and density of FN fibrils appeared to be quite similar, the average length of trails left behind by FAK-/- cells was significantly shorter. The majority of FN trails (∼60-70%) left by FAK+/+ and DA2 cells was between 5 and 10 μm long, whereas the majority of FN trails (∼60%) left behind by FAK-/- cells was less than 5 μm long.
The studies described above all suggest that the defects in FN matrix observed in FAK-/- embryos and cells were probably due to changes in how FN was organized and allocated, and not to changes in synthesis or assembly of FN. We verified this first by analyzing the response of FAK+/+ and FAK-/- cells to cell density, which normally stimulates FN gene expression (Perkinson et al., 1996). We plated FAK-/- and control FAK+/+ cells at five different cell densities after transfection with a chloramphenicol transferase reporter vector that contained the rat FN promoter region between positions -1908 and +136 (Shino et al., 1997). Changes in the FN promoter activity and FN mRNA levels assessed by northern blotting confirmed that both FAK+/+ and FAK-/- cells responded to increases in cell density (data not shown). We also carried out a pulse-chase analysis of FN synthesis, secretion and assembly into a DOC-insoluble matrix. DA2 and FAK-/- cells showed similar levels of FN synthesis (not shown). FAK-/- cultures secreted a somewhat greater percentage of what they synthesized than DA2 cells, whereas the percentage incorporated into a DOC-insoluble matrix was slightly less in FAK-/- than DA2 cells (Fig. 3). Together, these data indicated that the loss of FAK affects organization and allocation of FN to the apical surface of the cells in culture, rather than its synthesis or ability to be incorporated into an insoluble matrix.
Using Abs against epitopes specific for the three alternatively spliced exons (EIIIB, EIIIA and V in rodents; EDB, EDA, and CS-1 in humans), we detected all isoforms in fixed, permeabilized FAK+/+, FAK-/- and DA2 cells (data not shown). The presence of all three isoforms suggested that no significant splicing deficiency could be linked to abnormal appearance of FN matrix in FAK-/- cells.
Next, we sought to determine which features of the FAK molecule are important for normal FN matrix organization and allocation.
Both catalytic activity and intact proline-rich (PR) domains are required for FAK-mediated FN matrix organization and allocation
To gain more insight into the role of FAK in FN matrix organization and allocation, we conducted loss-of-function experiments, in which we transduced FAK+/+ cells with adenoviruses carrying mutant or truncated forms of FAK, and gain-of-function experiments, in which we transduced FAK-/- cells with these same constructs to see which features of FAK are essential to restore normal matrix patterning. In both cases we assessed FN matrix organization by staining live cells cultured for 48 hours in complete medium after exposure to adenoviral vectors. In loss-of-function experiments with FAK+/+ cells, we found that exogenous expression of FAK with mutations in the autophosphorylation site (Y397F FAK), ATP-binding site (kinase-dead [KD] FAK), or PR domains (P→A FAK, in which alanines were substituted for prolines 712, 713, 872, 873, 876 and 877) interfered with normal FN matrix organization and allocation on the cell surface. Similarly, exogenous expression of FRNK in focal contacts interfered with FN matrix patterning, whereas FRNK with a mutation in the paxillin-binding domain (L1034S FRNK), which could not go to focal contacts, did not affect the organization of FN matrix (Fig. 4A). Interestingly, P→A FAK was highly phosphorylated at both sites examined (Y397 and Y861) suggesting that even though this molecule was catalytically active, it still required interaction with an SH3-containing protein, possibly Cas, to support normal FN matrix organization and allocation. We did not detect phosphorylation of Y861 in either FRNK or L1034S FRNK. Gain-of-function experiments supported these results. Only transduction of wild-type FAK, but not Y397F, KD FAK, P→A FAK, or FRNK, into FAK-/- cells increased FN matrix allocation on the apical surface of cells (Fig. 4B).
Data from both the gain-of-function and loss-of-function experiments indicate the importance of FAK enzymatic activity and its structural roles in the focal adhesion complex. Next we addressed the importance of the location of FAK within cells.
FAK-family tyrosine kinase activity in focal contacts rescues normal FN matrix organization and allocation
One of the most potent inhibitors of almost all FAK functions is the exogenously expressed focal adhesion targeting (FAT) domain. When expressed as a fragment, the FAT domain alone can displace full-length FAK from focal contacts and act as a dominant negative (Gilmore and Rohmer, 1996; Ilić et al., 1998; Xu et al., 1998; Almeida et al., 2000). To test the importance of the presence of FAK at sites of focal adhesions for proper FN matrix organization, we used a system of PYK2/FAK chimeras that can or cannot go to focal contacts (Klingbeil et al., 2001). PYK2 is a FAK-related tyrosine kinase with ∼40% homology to FAK in its N-terminal FERM domain and ∼60% homology in the central kinase domain. PYK2 is overexpressed in FAK-null cells but cannot substitute for FAK in functions related to migration and haptotaxis and does not go to focal contacts in the FAK-null cells (Sieg et al., 1998). Previous studies (Klingbeil et al., 2001) showed that substitution of the PYK2 C-terminal region with the FAT-containing region of FAK enabled this chimeric molecule to go to focal contacts and to rescue haptotactic defects in FAK-null cells. When the L1034 residue in the FAT domain, which is required for its association with focal contacts (Tachibana et al., 1995), was mutated, the PYK2/FAK chimera was no longer able to rescue migratory defects of FAK-null cells. A similar set of PYK2/FAK or PYK2/L1034S FAK-mutated chimeric molecules, expressed in FAK-/- cells, was used to determine whether directing PYK2 to focal contacts via the FAT domain of FAK could rescue defects in FN allocation and patterning observed in FAK-null cells (Fig. 5A). We examined FN matrix organization and allocation, and distribution of β1 integrin in all four lines by immunostaining live cells. CA3 and CB4 cells, which had localized PYK2/FAK chimeras in focal contacts and were able to migrate normally (Klingbeil et al., 2001), also had normal-appearing fibrillar FN matrix and integrin β1 distribution on both the dorsal surfaces of cells and at the cell-substratum interface. In contrast, SE6 and SX4 cells, which could not localize their chimeric PYK2/L1034S FAK molecules to focal contacts and were unable to respond to a haptotactic signal, displayed the abnormal FN matrix allocation and β1 integrin distribution observed in mock-transfected FAK-/- cells (Fig. 5B).
Plating of FAK-/- cells on organized ECM laid down by FAK+/+ or DA2 cells did neither improve their motility or have an effect on the organization of the actin cytoskeleton and number of focal adhesions. And vice versa, FAK+/+ and DA2 cells migrated at similar rate on all FN-coated surfaces: complex ECM laid down by FAK-/-, FAK+/+, or DA2 cells (data not shown). These results suggested that although organized differently, the ECM per se is not the underlying cause of impaired migration observed in cells that lack FAK.
FAK localization in focal adhesions is required for generation of fibrillar adhesions
Yamada and colleagues (Pankov et al., 2000) have shown that linear translocation of integrin FN receptors from initial focal adhesions, along actin fibers, is required for the generation of fibrillar adhesions, conformational changes in the integrin-bound FN, and subsequent FN matrix fibrillar organization. Since fibrillar adhesions are at least in part defined by the presence of tensin, we looked at their distribution by tensin immunostaining. As shown in Fig. 6A, cells that did not have either FAK or PYK2/FAK chimeras in focal contacts had poor or no tensin-positive fibrillar adhesions present.
To confirm this observation, we used an Ab-chase technique to assess β1 integrin translocation and formation of fibrillar adhesions in FAK+/+, FAK-/- and DA2 cells (Fig. 6B). Because paxillin is present in the initial focal adhesions but does not translocate with β1 integrin along the actin cytoskeleton and is not detected in fibrillar adhesions (Cukierman et al., 2001), it was used as a marker of focal adhesions. We exposed live cells to the 9EG7 rat anti-active β1 integrin mAb for 45 minutes and fixed them either immediately (t=0) or 2 hours later (t=2). The cells were then incubated with anti-FN and FITC-conjugated anti-rat Abs while still unpermeabilized, to stain cell surface FN, while avoiding detection of intracellular FN and internalized 9EG7 Ab. After permeabilization, cells were incubated with anti-paxillin and other subsequent Abs and complexes (see Materials and Methods). FN- and β1 integrin-positive, paxillin-negative fibrillar adhesions were detected on FAK+/+ but not FAK-/- cells. Fibrillar adhesions were detected with this technique also in CA3 and CB4 cells that had PYK2/FAK chimeras in focal adhesions, but not in SE6 or SX4 in which PYK2/FAK chimeras were not localized in focal adhesions (data not shown).
FAK-dependent actin stress fiber organization
Lack of fibrillar adhesions suggests that the primary defect in FAK-/- cells might be related not only to defective focal adhesion remodeling but also to an altered cytoskeletal function.
In vitro actin polymerization assay using lysates of FAK+/+, FAK-/- and DA2 cells demonstrated that actin polymerization de novo is not affected by lack of FAK (Fig. 7A). However, staining of actin stress fibers revealed that cells lacking either FAK or FAK/PYK2 chimeras in focal adhesions were unable to organize stress fibers into typical parallel bundles (Fig. 7B).
Altogether, these results demonstrate the importance of having a FAK family FERM domain and tyrosine kinase activity localized in focal adhesion complexes, not only for normal cell migration but also for proper FN matrix allocation, focal adhesion dynamics and associated cytoskeletal rearrangements.
A number of studies have addressed FN-mediated FAK activation and its role in transducing signals from ECM. Interruption of these signals caused by either deletion of FAK or displacement of FAK from focal adhesions affects cytoskeletal organization and cell migration, as well as cell survival and cell cycle progression (Schlaepfer et al., 1999; Schaller, 2001). Cytoskeletal reorganization in response to initial cell-ECM interactions is required not only for cell motility but also for proper physical organization of the ECM. The nature of the signaling that promotes matrix organization in response to adhesion and spreading is still poorly understood. Here, we examined how the absence of FAK affects FN expression and organization. The following findings are particularly striking. (1) FN matrix is less fibrillar and more punctate in FAK-deficient mouse embryos, and in FAK-/- mouse embryonic fibroblasts and endothelial cells. (2) The FN matrix is also allocated differently in FAK-/- cells, being quite sparse at the dorsal cell surface and very abundant and punctate at the substratum interface, compared with FAK+/+ and DA2 cells. (3) Defective FN matrix is due to impaired FN fibril allocation and organization (patterning) rather than to changes in Fn gene regulation. (4) FAK catalytic activity, PR regions and location in focal contacts are all required for FAK to support proper FN matrix patterning. (5) Absence of FAK from focal adhesions is associated not only with altered FN organization but also with failure in organization of fibrillar adhesions and distribution of actin stress fibers in a parallel fashion.
Taken together with data from Pankov et al. (Pankov et al., 2000) showing that an intact cytoskeleton and fibrillar adhesions are required to expose sites on FN necessary for its organization, our results suggest that the underlying cause of defective FN matrix in FAK-/- cells is the absence of appropriate post-adhesive cytoskeletal reorganization and associated adhesion site remodeling required to support normal FN matrix allocation and patterning.
Requirement of tyrosine kinase activity in focal adhesion sites for FN matrix organization
We have shown the importance of FAK autophosphorylation of Y397 and catalytic activity for FN matrix organization. Autophosphorylation of Y397 is an initial step in FAK activation and serves to recruit Src-family kinases, which then phosphorylate other FAK tyrosine residues, increasing FAK activity. Phosphorylated Y397 can also bind PI 3-kinase and PLCγ (Schaller, 2001), suggesting that FAK might be involved in lipid signaling and focal contact localization of second messengers, thereby influencing actin polymerization (Ma et al., 1998; Chen et al., 2000) and cytoskeletal tension on ECM and FN matrix assembly. We did not find a major difference in localization of PtdIns(4,5)P2 at the periphery of FAK+/+ and FAK-/- cells. Furthermore, intracellular delivery of fluorescent PtdIns(4,5)P2 using polybasic shuttle carriers did not significantly improve cytoskeletal organization in FAK-/- cells (data not shown). These data, and findings from a previous report linking FAK with Cdc42 in a PtdIns(4,5)P2 synthesis-independent manner (Linseman et al., 2000), indicate that the role of FAK in actin cytoskeleton dynamics and ECM assembly may not come through these signaling intermediates.
In support of data presented here, Wierzbicka-Patynowski and Schwarzbauer (Wierzbicka-Patynowski and Schwarzbauer, 2002) reported that blocking PI 3-kinase and Src activity promoted the loss of FAK phosphorylation and inhibited FN matrix assembly. However, they did not report a deleterious effect on actin stress fibers, suggesting that tyrosine kinase activity at focal adhesion sites may not be critical for actin cytoskeleton organization per se. However, tyrosine kinase activity could be required for the proper organization of the multimolecular complexes through which actin stress fibers exert tension, via Rho (Zhong et al., 1998), on integrin receptors to promote FN organization and allocation.
PR regions of FAK link FAK with actin cytoskeleton
Our data clearly point to an essential role for FAK PR domains in FN matrix organization. Even though FAK was localized in focal contacts and its catalytic activity was very high, FN matrix organization and allocation were abnormal in cells expressing mutated PR regions of FAK. Other studies point to a particularly important role for PR region 1 (PR1) in FAK-mediated regulation of assembly of signaling complexes and the organization of the cytoskeleton. Expression in primary newborn rat cardiac myocytes of green fluorescent protein-tagged FAK with a mutation in PR1 alone (FAK residues 638-841) triggered disbanding of sarcomeres and disorganization of actomyosin myofibrils (Kovačič et al., 2001). Polte and Hanks (Polte and Hanks, 1995) showed that the SH3 domain of the adaptor protein Cas binds preferentially and with high affinity to PR1 of FAK. Cas may interact with actin filaments indirectly in two different ways, through tensin and zyxin, linking FAK with the actin cytoskeleton. Tensin, a large focal adhesion protein with barbed-end F-actin capping activity, consists of a protein tyrosine phosphatase-like domain (Haynie and Ponting, 1996), an SH2 domain and actin homology regions that may bind directly to actin and Cas (Lo et al., 1994). Interestingly, independent expression of actin homology region 2 of tensin was able to block FN fibrillogenesis in cultured primary human foreskin fibroblasts (Pankov et al., 2000). Zyxin is a PR LIM domain focal adhesion protein that binds the SH2 domain of Cas (Yi et al., 2002). Zyxin also interacts directly with α-actinin and members of the Ena/VASP family of proteins that bind actin filaments (Drees et al., 2000). Zyxin gene silencing by small interfering RNAs caused the loss of stress fibers (Harborth et al., 2001), which would likely cause subsequent deformities in FN matrix organization. Lack of FAK might interfere with either of these two Cas-mediated links that keep actin stress fibers in sites of focal contacts. Further investigation in this direction is clearly warranted.
Lack of fibrillar adhesions and disorganized actin cytoskeleton in FAK-/- cells may explain impaired FN matrix allocation and organization
Work from the laboratories of Geiger and Yamada (Katz et al., 2000; Pankov et al., 2000; Volberg et al., 2001; Cukierman et al., 2001) advocate two distinct types of cell-ECM adhesions in cells cultured on two-dimensional substrata: classical focal adhesions, which are rich in tensin, paxillin, vinculin and tyrosine-phosphorylated proteins, and fibrillar adhesions, which are rich in tensin but low in the other components. Pankov et al. (Pankov et al., 2000) proposed that classic focal adhesions formed upon initial adhesion are transformed into fibrillar adhesions through reorganization of protein complexes in sites of focal adhesions. During cell spreading, ligand-occupied α5β1 integrin moves along actin stress fibers and, together with tensin, forms elongated fibrillar adhesions that colocalize with organized FN fibrils (Katz et al., 2000; Pankov et al., 2000). FAK-/- cells have numerous classic focal contacts rich in tensin, paxillin, vinculin and tyrosine-phosphorylated proteins. However, our data show that FAK-/- cells do not generate fibrillar adhesions. Petroll et al. (Petroll et al., 2003) have shown with time-lapse experiments that new adhesions are continuously forming at the leading edge of cells. At the same time existing adhesions are moving backward in a retrograde fashion generating fibrillar adhesions. Cytochalasin D, a pharmacological agent that has a strong effect on adhesions and the cytoskeleton also interfered with fibrillar adhesion formation and matrix organization. Similarly, we were able to obtain an FN matrix pattern resembling that of FAK-/- cells by treating either FAK+/+ or DA2 cells with the pharmacological Rho-kinase inhibitor Y-27632, which led to decreased numbers of focal adhesions and disarray of actin stress fibers (data not shown).
Taking all data together, defects in actin stress fiber organization and transition from focal adhesions to fibrillar adhesions could explain abnormal FN matrix allocation and organization in FAK-/- cells.
We thank Dr R. Hynes (MIT) for Fn+/- mice and Dr D. Barber (UCSF) for critical reading of the manuscript. This work was supported by Hellman Family Award and NIH grants CA87652 (D.I.), CA75240 and CA87038 (D.D.S.), and P60DE13958 (C.H.D.).
- Accepted August 27, 2003.
- © The Company of Biologists Limited 2004