Focal adhesion kinase: Structure and signalling

Cell adhesion and motility play a central role in a diverse array of cellular events, including cellular differentiation, develop- ment and cancer (Albelda and Buck, 1990; Hynes, 1992). An experimental entry into the study of the molecular events trig- gering cell adhesion comes from the analysis of cell attach- ment and spreading, a process that is driven by the formation of molecular structures called focal adhesions. Focal adhesions (also referred to as focal contacts) are points of close apposition between the cell membrane and the extracel- lular matrix (ECM), which is comprised of proteins such as collagen, fibronectin or vitronectin (Burridge et al., 1988; Luna and Hitt, 1992). The structural organization of focal adhesions is complex. Integrins, heterodimeric transmem- brane receptors comprised of 㬡 and β subunits (Albelda and Buck, 1990; Hynes, 1992) bridge the cell membrane, the extracellular ligand-binding domains engaging the ECM on the outside of the cell and the short cytoplasmic tails inter- acting with the cytoplasmic cytoskeleton. Thus, integrins physically link the ECM to the cytoplasmic actin cytoskeletal network and may function to transmit signals from the extra- cellular matrix to the cytoplasm (Turner and Burridge, 1991; Schwartz, 1992). The actual linkage between integrin cyto- plasmic tails and actin bundles or stress fibers appears to be mediated by an intricate structure comprised of focal adhesion-associated proteins. Considerable evidence suggests that at least two of these focal adhesion-associated proteins, talin and α-actinin, interact directly with the cytoplasmic domain of β integrin subunits (Tapley et al., 1989a; Otey et al., 1990). Both talin and β-actinin have also been shown to bind to the actin-binding protein vinculin, supporting the idea that protein-protein interactions are responsible in large part for the ordered structure of the focal adhesion (Burridge et al., 1988).

Several lines of evidence point to the importance of tyrosine phosphorylation in the formation and organization of focal adhesions. In cells transformed by the tyrosine kinase oncogene pp60^src, two focal adhesion-associated proteins, tensin and paxillin, are highly phosphorylated on tyrosine (Turner et al., 1990; Davis et al., 1991). In addition, in Src- transformed cells, other focal adhesion proteins, talin, vinculin and 0i integrin subunits have been reported to be tyrosine phosphorylated, albeit at low stoichiometry (Sefton and Hunter, 1981; DeClue and Martin, 1987; Tapley et al., 1989b). Thus the dramatic alterations in cytoskeletal structure induced by Src transformation may be due in part to the tyrosine phos- phorylation of focal adhesion-associated proteins. In normal cells, immunofluorescence analysis with antibodies to phos- photyrosine reveals prominent staining of focal adhesions, indicating the presence of significant levels of tyrosine phos- phorylated proteins (Maher et al., 1985; Burridge et al., 1988). The attachment and spreading of rodent fibroblasts in culture leads to the increased tyrosine phosphorylation of both paxillin and tensin (Burridge et al., 1992; Bockholt and Burridge, 1993), while treatment of cells with inhibitors of protein tyrosine kinases blocks spreading of fibroblasts in culture (Burridge et al., 1992).

Studies from our own laboratory have led to the identifica- tion of a major pp60^src substrate, of M_r 125,000 (pp125), which localizes to focal adhesions of normal adherent chicken embryo cells (Schaller et al., 1992). The isolation and charac- terization of cDNA clones encoding ppi25 revealed that ppi25 was a novel protein tyrosine kinase, which we designated focal adhesion kinase, or pp125^FAK. Clues to the function of pp125^FAK come from numerous studies showing that the tyrosine phosphorylation of pp125^FAK is increased as a conse- quence of either the engagement of integrins with the extra- cellular matrix, for example the attachment and spreading of embryo fibroblasts onto a fibronectin matrix (Guan et al., 1991 ; Burridge et al., 1992; Schaller et al., 1993) or the cross-linking of surface integrins with integrin-specific antibodies (Kornberg et al., 1991, 1992). In addition, activation of fibrinogen- dependent platelet aggregation also induces tyrosine phospho-rylation of pp125^FAK in vivo and an increase in pp125^FAK tyrosine kinase activity in vitro (Lipfert et al., 1992). Thus, the increased tyrosine phosphorylation of pp125^FAK appears to be closely coupled with binding and activation of cell surface integrin receptors. In this brief review, we consider recent experimental data indicating that pp125^FAK plays a role in reg- ulating cellular events leading to the assembly of focal adhesions. In addition, we speculate on the possible role of pp 1 25^FAK in cellular signalling via pathways that modulate or control cellular gene expression.

Songyang et al., 1993). SH3 domains also appear to mediate protein-protein interactions, directing binding to proteins with proline-rich peptide sequence motifs (Ren et al., 1993). The lack of SH2 and SH3 domains in pp125^FAK suggests that FAK may play a role in cell signalling distinct from previously char- acterized non-receptor protein tyrosine kinases. In addition, as we will discuss below, it is likely that the non-catalytic domains of pp125^FAK participate in directing the protein- protein interactions that regulate and control pp125^FAK function.

To date, pp125^FAK homologues have been identified in mouse, human and Xenopus (Hanks et al., 1992; Andre and Becker- Andre, 1993; Whitney et al., 1993; M. Hens andD. DeSimone, personal communication). The structure of the pp!25^FAK in each of these species is highly conserved and is distinct from all other known protein tyrosine kinases. The catalytic domain exhibits most of the structural hallmarks of a typical tyrosine kinase, however, in the case of pp125^FAK the catalytic domain is flanked by two non-catalytic domains that exhibit little sequence similarity to other proteins (or gene products) present in the existing data bases (Fig. 1). FAK is expressed in most cell lines and tissues examined to date (Hanks et al., 1992; Andre and Becker-Andre, 1993; Turner et al., 1993). In some cells the carboxyl-terminal domain of pp125^FAK is expressed autonomously as a 41,000 M_r protein (called FRNK - FAK- related non-kinase; Schaller et al., 1993). In avian cells and tissues, FRNK is encoded by an alternatively processed 2.4 kb mRNA (Schaller et al., 1993). A similar sized mRNA has been detected in human tissues, but it remains to be determined if this mRNA encodes p41^FRNK. A notable feature of pp125^FAK structure is the absence of SH2 and SH3 domains, domains present in the Src family of kinases and other cytoplasmic protein tyrosine kinases, as well as many protein components of receptor-directed signalling pathways (reviewed by Pawson and G₁sh, 1992). In most SH2-containing proteins, the SH2 domains appear to direct protein-protein interactions, promoting stable interactions with unique phosphotyrosine- containing peptide sequence motifs (Pawson and G₁sh, 1992;

Little information is available as to how focal adhesion-asso- ciated proteins are directed to the existing or newly formed focal adhesions. The first clues as to how pp125^FAK is targeted to focal adhesions came from the analysis of a series of deletion mutations within the amino- and carboxyl- terminal non-catalytic domains (Hildebrand et al., 1993). Deletion of sequences between residues 853 and 1012 greatly diminished the translocation of retrovirally expressed FAK protein to the focal adhesions of chicken embryo cells grown in culture. In contrast, deletion of sequences within the amino-terminal non-catalytic domain or small deletions within a region of the C-terminal domain proximal to the kinase domain had no effect on the efficient localization of pp!25^FAK to focal adhesions. These data indicate that residues 853 to 1012 comprise a targeting sequence (termed the ‘focal adhesion targeting’ or ‘FAT’ sequence) necessary for the efficient localization of pp125^FAK to focal adhesions (Fig. 2). Further evidence for the importance of the FAT sequence comes from studies analyzing hybrid proteins comprised of unmyristylated, cytosolic pp60^src fused to a polypeptide containing residues 853 to 1052 of pp125^FAK. Immunofluorescence staining of chicken embryo cells infected with a retrovirus encoding the Src-FAT fusion protein showed efficient localization of Src-FAT protein to focal contacts, providing additional evidence that FAT sequences direct the translocation of pp125^FAK to focal adhesions.

Recent evidence indicates that within the cell there is a direct interaction between ppi25^FAK and the focal adhesion-associ- ated protein paxillin. Immunoprecipitation of pp!25^FAK from extracts of cells expressing wild-type pp125^FAK demonstrates the efficient co- immunoprecipitation of ppi 25^FAK and paxillin (M. Schaller, J. Hildebrand and J. T. Parsons, unpublished observations). The stable association of these two proteins was not observed when cells expressing FAK mutants lacking the FAT sequence or mutants lacking the C-terminal 11 residues of pp125^FAK were subjected to a similar analysis. Parallel in vitro experiments using glutathione S-transferase fused to FAK peptides containing sequences present in residues 687 to 1052 confirmed that paxillin could efficiently bind to sequences present in the carboxyl-terminal non-catalytic domain of pp 1 25^FAK (J. Hildebrand, M. Schaller and J. T. Parsons, unpublished observations). Furthermore, the binding to paxillin appears to be direct, since isotopically labelled GST- FAK bound to paxillin immobilized on a filter matrix (a ‘south- western’ blot). A careful analysis of a series of GST fusion proteins containing deletions of residues within the carboxyl- terminal domains shows that paxillin binding was functionally distinct from sequences necessary for focal adhesion targeting, although sequences required for paxillin binding appear to overlap, in part, the sequences required for focal adhesion targeting (Fig. 2). These results provide evidence for a role for the carboxyl-terminal non-catalytic domain of pp125^FAK in both the localization of pp 125^FAK to focal adhesions as well as directing the binding of pp125^FAK to a potential cellular substrate.

The cell adhesion-dependent activation of pp125^FAK tyrosine phosphorylation suggests that integrins may directly regulate, in some fashion, the activation of pp125^FAK kinase activity. Previous experiments by Otey et al. (1990) showed that the interactions of focal adhesion proteins and integrin cytoplas- mic domains can be analyzed in vitro. The focal adhesion-asso- ciated protein, α-actinin, binds in vitro to synthetic peptides mimicking the 47 amino acid cytoplasmic domain of the β ₁ and β ₃ integrins. A similar experimental approach reveals that pp125^FAK also binds efficiently to peptides mimicking the complete cytoplasmic domain of Pi and p.t integrins (M. Schaller, C. Otey and J. T. Parsons, unpublished observations). Further, analysis of pp125^FAK binding to a set of four over- lapping peptides comprising the total cytoplasmic domain sequence of Pi shows that pp125^FAK interacts preferentially with a peptide sequence representative of the first 13 residues adjacent to the transmembrane domain of β₁ (Fig. 3). Binding of pp125^FAK to peptide-containing beads can be blocked by preincubation with excess soluble peptide and pp125^FAK does not bind to beads containing a ‘scrambled’ short peptide. To determine where in pp125^FAK the integrin peptide-binding sequences resides, individual domains of pp125^FAK were expressed in Escherichia coli and used in the in vitro binding assay. Significant binding activity was observed with peptides derived from the amino-terminal non-catalytic domain, whereas no binding activity was observed with peptides derived from the carboxyl-terminal region of pp125^FAK. These data argue convincingly that pp125^FAK is capable of directly binding to integrin cytoplasmic domain sequences in vitro. Interestingly a comparison of the sequences of individual β cytoplasmic domains shows a a high degree of sequence con- servation within sequences corresponding to the β₁ pp125^FAK- binding regions (Fig. 3). Whether such sequences direct the binding of pp125^FAK and integrins in vivo, whether pp125^FAK interacts with different β integrins via a conserved sequence motif, and how such interactions regulate pp 125^FAK activity are issues under current investigation.

In Src-transformed cells the tyrosine phosphorylation of pp I 25^FAK is increased several fold, an observation that lead to its original identification as a Src-substrate (Kanner et al., 1990). In these cells the majority (>80%) of pp 125^FAK is stably associated with pp60^src (Cobb et al., 1994). Genetic experiments using retroviruses expressing mutants of Src, as well as in vitro analysis of pp60^src-pp125^FAK complex formation, clearly indicate that the assembly of stable FAK-Src complexes requires both the SH2 domain of pp60^src and the autophos- phorylation site of pp125^FAK. Peptide mapping experiments, coupled with site-directed mutagenesis of potential phospho- rylation sites have identified the major site of pp125^FAK autophosphorylation as Tyr³⁹⁷ (Fig. 2) (Schaller et al., 1994). Mutation of Tyr³⁹⁷ to Phe efficiently blocks pp60^src-pp125^FAK interactions in vivo and in vitro. Several features of the Tyr³⁹⁷ autophosphorylation site are of interest. The position of Tyr³⁹⁷ within pp125^FAK distinguishes it from other receptor and non- receptor tyrosine kinases. In most instances tyrosine kinase autophosphorylation occurs at a highly conserved tyrosine within the catalytic domain (equivalent to Tyr⁵⁷⁶ in pp125^FAK; Tyr⁴¹⁶ in pp60^src), within a kinase insert domain, which is a nonconserved insert found within the catalytic domains of some receptor protein tyrosine kinases (but not in pp 125^FAK) or distal to the catalytic domain at sites near the C terminus (a region found in many growth factor protein tyrosine kinases). Tyr³⁹⁷ resides immediately amino-terminal to the catalytic domain, in relative proximity to the ATP-binding site. In addition Tyr³⁹⁷ is embedded in the sequence DDYAEI, a sequence very similar to the consensus of a high-affinity Src SH2-binding peptide, YEEI (Songyang et al., 1993). These observations pose the possibility that in normal cells, integrin engagement may trigger autophosphorylation of pp125^FAK, which may, in turn, direct the translocation and concomitant activation of Src or other Src-like tyrosine kinases. Experi- mental support for such a model comes from the identification of pp 125^FAK-p59^f>‘ⁿ complexes in extracts of normal adherent cultures of chicken embryo cells (Cobb et al., 1994).

One important function of the integrins is to translate extra- cellular cues into cytoplasmic signals, a function that is pre-sumably important for the biological activities of integrins. On the basis of the data summarized above we are led to speculate that integrin engagement with the extracellular matrix may result in either the direct clustering of pp125^FAK, allosteric changes in pp125^FAK or the stimulation of a regulator protein(s) that triggers pp125^FAK activation. A direct conse- quence of such an activation step is the autophosphorylation of pp125^FAK and generation of a high affinity binding site for Src and Src-family kinases. In normal cells the enzymatic activity of pp60^src and pp59^f>‘ⁿ is repressed through the action of a negative regulatory phosphorylation site at the C terminus of these kinases (Fig. 4). Phosphorylation of a highly conserved tyrosine within this region by a regulatory protein tyrosine kinase (Csk) is critical for down-regulation of catalytic activity (reviewed by Cooper and Howell, 1993). Current models for Src regulation suggest that the tyrosine phosphorylated C- terminal sequence binds in an intramolecular interaction to its own SH2 domain (Cantley et al., 1991; Cooper and Howell, 1993). The amino acid sequence flanking this C-terminal tyrosine does not resemble the consensus high affinity binding site and while a tyrosine phosphorylated C-terminal peptide can bind to the SH2 domain of pp60^src, it does so poorly (Songyang et al., 1993). In vitro, pp60^src can be enzymatically activated by incubation with a synthetic phosphopeptide con- taining the consensus, high affinity, Src SH2-binding site, pre- sumably by binding more efficiently to the SH2 domain than the regulatory C-terminal peptide (Liu et al., 1993). It is intriguing to speculate that autophosphorylation of Tyr³⁹⁷ of pp125^FAK may create a high affinity binding site for pp60^src and pp59^f>‘ⁿ and that these kinases may bind to pp125^FAK resulting in the displacement of their C-termini from their SH2 domains. Thus, binding to pp125^FAK may be a mechanism by which pp60^src and pp59^fyn are enzymatically activated in addition to a mechanism for the recruitment of these kinases to a highly localized site within the cell.

What might be the consequences of the activation of pp125^FAK or the translocation-dependent activation of Src or Fyn? The adhesion-dependent increase in tyrosine phosphory- lation of paxillin and tensin suggests that either or both of these focal adhesion proteins may be direct substrates for pp125^FAK or the pp125^FAK -Src/Fyn complex. The association of pp125^FAK and paxillin is interesting in this context and is con- sistent with the idea that pp125^FAK may play a direct role in bringing paxillin into the tyrosine kinase complex. It is inter- esting to speculate that activation of both pp125^FAK and Src/Fyn may be necessary for catalyzing the formation of focal adhesion assembly and for initiating signals that may direct the activation of other cellular signalling pathways. For example it is well established that cell adhesion and spreading can trigger the expression of cellular genes (Damsky and Werb, 1992). The association of pp125^FAK with either Src or Fyn may be sufficient to activate cellular signalling pathways that in turn lead to the activation of cellular genes. What these pathways are and how they function to regulate adhesion-dependent phenomena remain to be elucidated.

The studies from the author’s laboratory were supported by NIH- NCI grants, P01 CA 40042, R37 CA 29243.