Paxillin comes of age

The interaction of a cell with its microenvironment provides important chemical and spatial cues that contribute to the regulation of processes such as embryonic development, wound healing, immune surveillance and tissue homeostasis through the modulation of a diverse array of cellular functions, such as migration, differentiation and proliferation. The principal cell-surface proteins that are responsible for regulating the binding of a cell to components of the external environment are the integrins. Integrins are transmembrane proteins that comprise an α and a β subunit, which together produce 24 distinct heterodimers (Hynes, 2002) that serve as bridges between the extracellular matrix (ECM) and the intracellular signaling machinery and actin cytoskeleton. Upon their interaction with the ECM, integrins cluster and recruit a wide variety of intracellular proteins (Fig. 1). These macromolecular foci were originally observed in cultured fibroblasts as electron-dense regions of the plasma membrane (Abercrombie et al., 1971) that were termed attachment plaques (Heggeness et al., 1978; Hynes and Destree, 1978) and have subsequently been defined as focal complexes, focal adhesions and fibrillar adhesions (Geiger et al., 2001), depending on their size, cellular localization or dependency on different members of the Rho family of GTPases (Chrzanowska-Wodnicka and Burridge, 1992; Nobes and Hall, 1995; Small et al., 1999; Zamir et al., 2000).

Integrin cytoplasmic domains have no intrinsic enzymatic activity and so must recruit a variety of proteins to adhesion plaques or contacts to enable these foci to serve as conduits for the transmission of force that is necessary for cell migration and for bidirectional signaling between the cell interior and its local microenvironment. Originally, the list of focal-adhesion-localized proteins was limited to structural proteins, such as talin and vinculin, that were believed to mediate the anchorage of the integrin cytoplasmic domain to the actin cytoskeleton (Burridge et al., 1988). Now, approximately 37 years after the visualization of the first cell-adhesion contact, the apparent molecular complexity of focal adhesions continues to increase. Focal adhesions are now reported to comprise upwards of 125 individual proteins, many of which exhibit multiple protein-protein interactions (Turner, 2000; Zaidel-Bar et al., 2007a). Thus, understanding the interactions that govern focal-adhesion function, the regulation of these multiple interactions and their role in coordinating bidirectional signaling remains a considerable challenge.

In 1990, paxillin joined the integrins (Hynes, 1992; Zaidel-Bar et al., 2007a), talin (Burridge and Connell, 1983) and vinculin (Geiger et al., 1980) as one of the earliest known members of the `focal adhesion proteome' (Turner et al., 1990). Having first been identified as a 68 kDa protein that exhibited increased tyrosine phosphorylation following the transformation of chick embryonic fibroblasts by the Src-expressing Rous sarcoma virus (Glenney and Zokas, 1989), paxillin was subsequently purified from smooth muscle tissue and characterized as a direct binding partner for the vinculin tail domain (Turner et al., 1990). In keeping with the prevailing dogma of the time, in which focal adhesions were believed to function solely as passive, structural links between the ECM and the actin cytoskeleton, the name `paxillin' was coined: it derives from the Latin term `paxillus' (a peg or stake) and suggests a function that is somewhat analogous to a tent peg in tethering actin stress-fiber cables to the adhesion site (Brown and Turner, 2004; Turner et al., 1990). The discovery that paxillin, along with the recently described focal adhesion kinase (FAK) (Parsons, 2003; Schaller et al., 1992), was also tyrosine phosphorylated in non-transformed cells upon their adhesion to the ECM (Burridge et al., 1992) and in the developing embryo (Turner, 1991; Turner et al., 1993) provided the first indications that focal-adhesion proteins were actively signaling to the cell interior. The early embryonic lethality of mice deficient in paxillin, FAK and fibronectin further reinforced the essential role that focal adhesion proteins play in vivo (Hagel et al., 2002; Ilic et al., 1995). The subsequent analysis of fibroblasts derived from paxillin-deficient embryos has indicated that defects in cell migration might be the primary cause of the severe developmental phenotype.

In this Commentary, we first detail the basic molecular architecture of paxillin and explain its classification as a molecular scaffold. In addition, we discuss how paxillin, through tightly regulated interactions with multiple structural and signaling molecules, serves as a nexus for the control of the Rho family of GTPases in their capacity as essential regulators of the actin cytoarchitecture and adhesion dynamics. The importance of emerging microscopy technologies and model systems for elucidating paxillin function and directing future avenues of investigation will also be discussed.

The molecular cloning of paxillin and subsequent peptide-sequence analysis revealed that it comprises numerous discrete structural domains (Brown et al., 1996; Turner and Miller, 1994) that have since been identified as protein-binding modules (Brown and Turner, 2004; Turner, 2000). In turn, this has led to the current classification of paxillin as a molecular adaptor or scaffold protein, the primary function of which is to serve as a nexus to coordinate, integrate and facilitate efficient cell signaling, through direct and indirect interactions with multiple signaling and structural proteins that constitute the paxillin `interactome'.

The C-terminal half of paxillin contains four LIM (Lin11, Isl-1, Mec-3) domains, which are double-zinc-finger motifs that mediate protein-protein interactions (Perez-Alvarado et al., 1994; Schmeichel and Beckerle, 1994) and are found in all eukaryotes but are absent from prokaryotes (Kadrmas and Beckerle, 2004). The LIM2 and LIM3 domains of paxillin are essential for targeting the protein to focal adhesions (Brown et al., 1996). It has been established that phosphorylation of these domains contributes to the regulation of focal-adhesion targeting of paxillin (Brown et al., 1998b), but the identity of the docking protein for paxillin has so far remained elusive, despite the fact that the localization of paxillin at focal adhesions is an absolute requirement for most paxillin-mediated processes. The LIM domains of paxillin also serve as binding sites for several structural and regulatory proteins, including tubulin and the tyrosine phosphatase PTP-PEST (also known as protein tyrosine phosphatase non-receptor type 12, PTPN12) (Cote et al., 1999; Herreros et al., 2000), and these interactions have important roles in controlling focal-adhesion dynamics (Efimov et al., 2008; Webb et al., 2004).

The N-terminus of paxillin controls most of its signaling activity. It contains five leucine- and aspartate-rich LD motifs (LD1-LD5), which are protein-binding modules that have the consensus sequence LDxLLxxL (Tumbarello et al., 2002). LD motifs were originally identified when the binding sites on paxillin for vinculin and FAK were mapped (Brown et al., 1996; Turner and Miller, 1994); they have since been identified in several proteins of diverse function (Brown et al., 1998a). Molecular modeling predicted that the short LD peptides would probably form amphipathic α-helices in which the leucine side chains were arranged on a single, hydrophobic face of the helix (Brown et al., 1998a; Sattler et al., 2000; Tumbarello et al., 2002). This secondary structure has since been confirmed by structural nuclear magnetic resonance (NMR) studies (Hoellerer et al., 2003). Interestingly, despite their small size and high level of sequence conservation, the individual paxillin LD motifs are able to mediate multiple protein interactions that are both overlapping and specific (Brown and Turner, 2004; Turner, 2000; Turner et al., 1999). The paxillin N-terminus also contains a proline-rich region, which was originally identified as a docking site for the SH3 domain of the non-receptor tyrosine kinase Src (Weng et al., 1993). The proline-rich region has subsequently been shown to form a polyproline-II helix upon interaction with the vinculin-binding protein ponsin, and this interaction might be important for the formation of costameres (sites of actin-membrane interactions in cardiac muscle) (Gehmlich et al., 2007; Zhang et al., 2006).

Multiple tyrosine, serine and threonine phosphorylation sites exist throughout the paxillin molecule (Brown and Turner, 2004; Turner and Miller, 1994; Webb et al., 2005); these sites are targeted by a diverse array of kinases that are activated in response to various adhesion stimuli and by growth factors. These include, but are not limited to, p21-activated kinase (PAK) (Nayal et al., 2006), FAK-Src (Thomas et al., 1999), receptor for activated C kinase 1 (RACK1; also known as GNB2L1) (Doan and Huttenlocher, 2007), Jun N-terminal kinase (JNK) (Huang et al., 2004b; Huang et al., 2003), extracellular-signal-regulated kinase (ERK; also known as MAPK1) (Ishibe et al., 2003; Ku and Meier, 2000), p38 mitogen-activated protein kinase (p38 MAPK) (Huang et al., 2004a; Huang et al., 2004b), cyclin-dependent kinase 5 (CDK5) (Miyamoto et al., 2007) and Abl (Lewis and Schwartz, 1998). The phosphorylation of paxillin contributes to the complexity of its interactome by regulating the interactions of various proteins with its protein-binding modules or, as in the case of tyrosine phosphorylation, by providing additional docking sites for other structural and signaling components (Brown and Turner, 2004).

The ability of cells to coordinate the complicated and tightly regulated process of cell migration is fundamental for wound healing, the immune response, angiogenesis and embryogenesis (Li et al., 2005). The spatiotemporal regulation of the Rho family of small GTPases, which includes the proteins Cdc42, Rac1 and RhoA, and in turn the localized activation of a wide variety of effector proteins by these GTPases, is essential for controlling the dynamics of the actin cytoskeleton and of actin-associated adhesions during polarized cell migration (Raftopoulou and Hall, 2004; Ridley, 2001a; Ridley, 2001b). Specifically, Cdc42 is required for cell polarization and the formation of filopodia (narrow finger-like actin-rich projections at the front of the migrating cell), whereas Rac1 promotes the extension of broad membrane sheets or lamellipodia (also at the leading edge of the cell). RhoA stimulates actin-stress-fiber formation, which is necessary for cell contractility and translocation. Cdc42 and Rac1 also stimulate the assembly of nascent adhesion complexes, whereas RhoA promotes both the maturation of adhesions at the leading edge and their disassembly at the rear of the cell (Ridley, 2001b; Raftopoulou and Hall, 2004). In the 18 years since paxillin was first identified as a component of focal adhesions, it has emerged as a key coordinator of the Rho GTPase family and of their signaling processes in the context of cell spreading and migration (Brown and Turner, 2004; Price et al., 1998).

Members of the Rho family of GTPases function as molecular switches by cycling between an inactive (GDP-bound) and an active (GTP-bound) state. Cycling is controlled by a large group of guanine-nucleotide-exchange factors (GEFs), which catalyze the exchange of GDP for GTP, and by GTPase-activating proteins (GAPs), which promote the hydrolysis of GTP to GDP (Hoffman and Cerione, 2002). Paxillin contributes to the regulation of the Rho family of GTPases, and therefore to the coordination of their downstream signaling to the cytoskeleton, by indirectly recruiting various GEFs, GAPs and effector proteins to cell-ECM contact sites. For example, upon the ligation of an integrin to either a fibronectin or collagen substrate, paxillin becomes tyrosine phosphorylated, primarily on tyrosine residues 31 and 118 (Y31 and Y118, respectively) (Bellis et al., 1997; Burridge et al., 1992; Petit et al., 2000), in a FAK- and Src-dependent manner (Schaller and Parsons, 1995). The CrkII-DOCK180-ELMO complex, which regulates Rac1 signaling (Grimsley et al., 2004), can interact with the phosphorylated Y31 and Y118 residues of paxillin via the SH2 domain of CrkII (Birge et al., 1993; Petit et al., 2000; Valles et al., 2004) and, by means of the unconventional Rac1- and Cdc42-GEF activity of DOCK180 (Brugnera et al., 2002), activates Rac1 to promote cell migration. Interestingly, the phosphorylation of Y31 and Y118 has also been shown to regulate RhoA activity. Phosphorylated paxillin binds directly to p120RasGAP (RASA1), which diminishes the interaction of p120RasGAP with p190RhoGAP (ARHGAP5) at the plasma membrane; this culminates in the suppression of localized RhoA activity by p190RhoGAP (Tsubouchi et al., 2002). Therefore, paxillin that is phosphorylated at Y31 and Y118 can indirectly activate Rac1 and inhibit RhoA, and both of these activities are necessary for efficient leading-edge protrusion during cell migration. Because phosphorylated Y31 and Y118 cannot bind simultaneously to both CrkII and p120RasGAP, it will be important to determine what additional signals, perhaps in the form of posttranslational modifications or recruitment of additional binding partners, are necessary to control the specificity of these interactions with paxillin. In this regard, it is interesting to note that tyrosine phosphorylation of Y31 and Y118 indirectly enhances the binding of FAK to the adjacent LD motifs of paxillin and has been implicated in focal-adhesion maturation (Zaidel-Bar et al., 2007b).

The LD4 motif of paxillin is a particularly important region of the protein for the regulation of Rho GTPase signaling and focal-adhesion turnover. The motif recruits a molecular complex that comprises the proteins G-protein-coupled receptor kinase interacting protein [GIT1 or GIT2; the latter is also known as paxillin kinase linker (PKL)], PAK-interacting exchange factor (PIX), p21-activated serine/threonine kinase (PAK) and NCK to adhesion sites at the leading edge of migrating cells (Turner et al., 1999; West et al., 2001; Zhao et al., 2000). GIT2 and the closely related GIT1 bind directly to the paxillin LD4 motif (Turner et al., 1999; West et al., 2001; Zhao et al., 2000) and are members of the ArfGAP family (negative regulators of Arfs, another family of small GTPases that control membrane trafficking and can in some cases also indirectly stimulate Rac1 activity and modulate actin dynamics) (Premont et al., 2000; Zhao et al., 2000). α- and β-PIX (ARHGEF6 and ARHGEF7, respectively) (Manser et al., 1998) display Rac1- and Cdc42-GEF activity, and in turn bind to the Cdc42-Rac1 effector PAK (Bokoch, 2003; Brown et al., 2002; Feng et al., 2002; Premont et al., 2004; Turner et al., 1999; West et al., 2001; Zhao et al., 2000). Importantly, fibroblasts that express a paxillin mutant that lacks the LD4 motif, and therefore cannot recruit the GIT1/2-PIX-PAK-NCK complex to focal adhesions, exhibits abnormal membrane-protrusion dynamics that are caused by sustained global Rac1 activity (West et al., 2001) (Fig. 2). These cells are also defective in polarized cell migration (Brown and Turner, 2004; West et al., 2001) and focal-adhesion turnover (Webb et al., 2004).

The recruitment of the GIT-PIX-PAK-NCK complex to paxillin and the activity of the complex are tightly regulated (Fig. 3), which is consistent with the crucial role of paxillin in coordinating Rac1 signaling. The adhesion-dependent activation of Cdc42 and Rac1 – which is mediated, in part, by the GEF activity of PIX – combined with FAK-Src-dependent tyrosine phosphorylation of GIT2 promotes a functional unmasking of its paxillin-binding site and thus the stable recruitment of the entire GIT-PIX-PAK-NCK complex to nascent focal complexes at the leading edge of the cell (Brown et al., 2005; Brown et al., 2002; Loo et al., 2004; Turner et al., 1999). The kinase activity of PAK is also stimulated upon its binding to GTP-bound Rac1 or Cdc42 (Bokoch, 2003) and, once it is targeted to the cell periphery as a member of the GIT2 complex, PAK contributes to focal-adhesion and actin-cytoskeleton remodeling in the developing lamellipodium by phosphorylating LIM kinase and myosin-light-chain kinase (Bokoch, 2003). Recently, the kinase activity of PAK has also been shown to enhance the interaction between the related GIT1-PIX-PAK complex and paxillin; PAK directly phosphorylates serine 273 (S273) within the paxillin LD4 motif, which further stimulates local Rac1 signaling (Nayal et al., 2006). Conversely, other studies suggest that the interaction of the GIT1/2-PIX-PAK complex with paxillin might also serve as a termination signal for Rac1 activity, by means of a mechanism that requires the ArfGAP activity of the GIT proteins (Nishiya et al., 2005; West et al., 2001). It is important to note that, although they are closely related, the GIT1 and GIT2 proteins are not functionally redundant (Brown et al., 2005; Frank et al., 2006), and this distinction provides further potential for differential signaling via the binding of the two GIT proteins to paxillin. For instance, it remains to be determined whether phosphorylation of S273 influences the binding of GIT2, as well as that of GIT1.

Interestingly, the phosphorylation of the paxillin LD4 motif at S273 not only promotes GIT1 binding, but also reduces the affinity of FAK for the motif (Nayal et al., 2006). This provides a mechanism by which paxillin might promote GIT1-mediated signaling while simultaneously attenuating the activity and/or localization of FAK, which has a key role in regulating RhoA activity (Hanks et al., 2003). Structural NMR studies of the LD4 motif provide a potential mechanism for this regulation. In solution, and when phosphorylated at S273, the LD4 peptide exists primarily as a random coil, whereas the non-phosphorylated peptide folds into an ordered α-helix upon its interaction with the C-terminal focal-adhesion-targeting (FAT) domain of FAK (Bertolucci et al., 2005; Bertolucci et al., 2008; Hildebrand et al., 1995; Hoellerer et al., 2003). However, unlike GIT proteins, FAK interacts with the LD2 motif of paxillin in addition to the LD4 motif (Brown et al., 1996; Turner et al., 1999). The FAT domain of FAK is composed of a four-helical bundle, which is stabilized by hydrophobic interactions (Arold et al., 2002; Hayashi et al., 2002). The LD2 and LD4 motifs of paxillin bind preferentially to opposite faces of the FAT helical bundle and interact with hydrophobic patches at the interface of helices 1 and 4, and helices 2 and 3, respectively (Bertolucci et al., 2005), which suggests that paxillin and FAK interact with a stoichiometry of 1:1. Recently, it has been shown that FAK requires only one functional LD motif (LD2 or LD4) for its appropriate localization to focal adhesions but must bind to both for maximal activation and downstream signaling (Scheswohl et al., 2008). Therefore, it seems likely that paxillin can fine-tune the balance between the local activity of FAK and Rac1 activity through the phosphorylation of its LD4 motif, which regulates its interaction with the GIT1-PIX-PAK-NCK complex (and possibly the GIT2 complex); at the same time, paxillin can maintain its interaction with sub-optimally active FAK in adhesion contacts.

Not unexpectedly, the action of phosphatases is also essential for adhesion turnover and cell migration (Angers-Loustau et al., 1999; Garton and Tonks, 1999; Stoker, 2005; Webb et al., 2004). A further link between paxillin and Rho GTPase regulation involves the tyrosine phosphatase PTP-PEST, which is recruited to focal adhesions by binding to the LIM3 and LIM4 domains of paxillin (Cote et al., 1999). PTP-PEST regulates cell spreading, cell migration and protrusion by decreasing Rac1 activity (Sastry et al., 2002). It does this by interacting directly with paxillin, thereby inhibiting signaling cascades that are mediated by the LD4 motif and/or phosphorylation of Y31 and Y118 (Jamieson et al., 2005). Although the mechanism of action of PTP-PEST in this context is incompletely understood, GIT2 has been shown to be a substrate for PTP-PEST (Jamieson et al., 2005); through the dephosphorylation of GIT2, PTP-PEST might act to destabilize the GIT2-PIX-PAK complex and its interaction with paxillin to suppress localized Rac1 activation. PTP-PEST can also dephosphorylate and thereby inactivate both Vav2 (a Rac1 GEF) and p190RhoGAP (a RhoA GAP), and so might modulate the balance between Rac1 activity (which drives cell protrusion) and RhoA activity (which stabilizes adhesions at the leading edge and promotes detachment of the cell rear) (Sastry et al., 2006). Interestingly, preliminary evidence suggests that paxillin also binds indirectly to Vav2 (Jennifer S. Jamieson and C.E.T., unpublished observations). In addition to PTP-PEST, other phosphatases, including PP2A (Ito et al., 2000; Jackson and Young, 2003; Young et al., 2003) and SHP-2 (also known as PTPN11) (Manes et al., 1999), have been reported to bind to and dephosphorylate paxillin, and these undoubtedly have important roles in controlling paxillin interactions, for instance during cell metastasis (Young et al., 2003).

The paxillin LD1 motif regulates cell-adhesion signaling by interacting directly with the integrin-linked kinase (ILK) (Nikolopoulos and Turner, 2001) and with actopaxin (Nikolopoulos and Turner, 2000) (a member of the parvin family of focal-adhesion proteins) (Gimona et al., 2002; Olski et al., 2001) to help recruit both proteins to focal adhesions (Nikolopoulos and Turner, 2002). When complexed with the adapter protein PINCH, ILK and actopaxin play a necessary role in the stabilization of interactions between the plasma membrane and the actin cytoskeleton in motile and non-motile cells (Legate, 2006; Wu, 2004; Zervas et al., 2001), as well as in the regulation of cell survival, proliferation and tissue morphogenesis (Legate, 2006). Both actopaxin and ILK are thought to regulate cell migration by modulating Rho GTPase signaling (Clarke et al., 2004; Filipenko et al., 2005; Khyrul et al., 2004). Parvin family members bind both to PIX (Mishima et al., 2004) and CdGAP (LaLonde et al., 2006); this binding either activates (PIX) or inhibits (CdGAP) Cdc42 and Rac1 activity. Perturbation of these interactions affects the subcellular localization and activity of both PIX and CdGAP, which results in defects in cell spreading and migration (Filipenko et al., 2005; LaLonde et al., 2006; Mishima et al., 2004).

The role of paxillin in the regulation of Rho-family GTPase signaling and cell migration is evolutionarily well conserved. In the yeast Saccharomyces cerevisiae, a paxillin-like protein, Pxl1p, has been described that appears to regulate cell polarity and bud formation by coordinating the activity of RhoA and Cdc42 (Gao et al., 2004; Mackin et al., 2004). Interestingly, the Schizosaccharomyces pombe homolog of Pxl1p has also been shown to modulate RhoA activity during cytokinesis (Ge and Balasubramanian, 2008; Pinar et al., 2008). The cellular slime mold Dictyostelium discoideum expresses two isoforms of paxillin, PaxA and PaxB, of which PaxB exhibits closer homology to mammalian paxillin (Bukharova et al., 2005). The deletion of PaxB reveals that this protein has a key role in regulating cell adhesion and migration during the multicellular phases of D. discoideum development (Bukharova et al., 2005). Paxillin also plays an important role in cell migration and Rho GTPase signaling during development in Drosophila melanogaster (Chen et al., 2005; Llense and Martin-Blanco, 2008) and Xenopus laevis (Iioka et al., 2007). It is also expressed throughout zebrafish (Danio rerio) development (Crawford et al., 2003), during which it exhibits robust changes in tyrosine phosphorylation (Lemeer et al., 2007) that are similar to those observed during development in higher eukaryotes (Turner, 1991). The precise role of paxillin in Rho GTPase signaling in these model organisms deserves further study.

Adhesion contacts are dynamic structures (Ballestrem et al., 2001; Laukaitis et al., 2001; Webb et al., 2002; Wiseman et al., 2004) that must assemble, disassemble or mature at the extending leading edge and disassemble at the retracting cell rear for efficient cell migration to occur (Webb et al., 2002). Paxillin is one of the earliest proteins to be detected in nascent adhesions at the leading edge of the cell, where it is rapidly organized (Digman et al., 2007), which suggests that paxillin plays an important role in promoting the assembly of adhesions and in defining their molecular composition. Integrin clustering cannot be visualized in these paxillin-rich nascent adhesions by using standard approaches (Laukaitis et al., 2001), but the use of image correlation microscopy (ICM) and fluorescence recovery after photobleaching (FRAP) has revealed that integrins do indeed cluster at these sites and are less mobile there than in the surrounding area; this is consistent with integrin engagement with the ECM and/or the cytoskeleton (Wiseman et al., 2004). Ratio image analysis of newly formed adhesions during fibroblast spreading has revealed that paxillin is particularly enriched at the periphery of these small dot-like adhesions (Zimerman et al., 2004). By contrast, the adhesion core is enriched in actin and has high levels of tyrosine-phosphorylated protein throughout, which suggests the existence of nascent microarchitecture (Zimerman et al., 2004).

Upon the application of tensile force and the subsequent activation of RhoA, some focal complexes can mature into focal adhesions, thereby providing a physical link to the contractile actomyosin machinery that is required for cell translocation (Balaban et al., 2001; Rottner et al., 1999; Zaidel-Bar et al., 2003). The powerful combination of total internal reflection fluorescence microscopy (TIRFM) and fluorescent speckle microscopy (FSM) indicates that, despite these robust links to the cytoskeleton, individual focal-adhesion components exhibit variable dynamic retrograde flow that appears to relate to whether they interact directly with actin. The retrograde flow of paxillin has a low correlation with actin retrograde flow, whereas the flow of the actin-binding proteins vinculin and α-actinin is more tightly coupled to the motion of actin (Hu et al., 2007). This suggests that, despite biochemical evidence for a direct interaction between paxillin and vinculin (Turner et al., 1990), the dynamic behavior of the two proteins in focal adhesions can be uncoupled. FRAP analysis confirms this observation, because the dynamics of paxillin in focal adhesions appear to be independent of its association with vinculin (Humphries et al., 2007). One plausible explanation for this apparent contradiction is that the interaction of paxillin with vinculin structurally stabilizes nascent adhesions by providing an early link to the cortical actin network at the leading edge of the cell, whereas, in mature focal adhesions, their association is primarily transient and so has little or no influence on the dynamic behavior of the proteins; this does not, however, preclude a signaling role for such short-lived encounters.

In addition to its role in the assembly of focal adhesions, paxillin also regulates cell migration by contributing to efficient focal-adhesion disassembly, as indicated by the stabilization of adhesions in cells that are devoid of paxillin (Webb et al., 2004). Paxillin-mediated disassembly might be accomplished in several ways. For example, paxillin that is phosphorylated at Y31 and Y118 localizes to dynamic adhesions (Ballestrem et al., 2006; Laukaitis et al., 2001), and the mutation of Y31 and Y118 to non-phosphorylatable amino acids impairs the disassembly of adhesions at the leading edge of migrating cells (Webb et al., 2004). Expressing a paxillin-LD4-motif deletion mutant has a similar effect (Webb et al., 2004). The precise mechanism by which these regions of paxillin influence adhesion disassembly is currently unclear, but potentially involves the interactions of paxillin with ERK (Ishibe et al., 2003) and FAK-Src to regulate myosin-light-chain-kinase-dependent contractility and perhaps calpain protease activity (Carragher et al., 2003; Webb et al., 2004; Zaidel-Bar et al., 2007b). Indeed, the proteolysis of paxillin by the calcium-dependent cysteine protease calpain 2 (Yamaguchi et al., 1994) has been reported to induce the disassembly of adhesion contacts in smooth muscle cells (Carragher et al., 1999) and to stimulate protrusive activity in fibroblasts (Franco et al., 2004). Conversely, the serine phosphorylation of paxillin at residues 188 and 190 (Bellis et al., 1997) has been shown to stimulate cell migration, and presumably adhesion turnover, by preventing the polyubiquitylation of paxillin and its subsequent degradation (Abou Zeid et al., 2006). Furthermore, in X. laevis, the regulation of paxillin stability by polyubiquitylation is essential for mesodermal cell migration during the process of convergent extension (Iioka et al., 2007). Finally, it has also been suggested that the targeting of microtubules to adhesions triggers paxillin-mediated disassembly or `catastrophes' of these macromolecular foci, either directly – through the binding of tubulin to the paxillin LIM2 and LIM3 domains (Brown and Turner, 2002; Herreros et al., 2000) – or through an as-yet-unidentified paxillin-associated factor (Efimov et al., 2008).

Although it is clear that paxillin can employ several mechanisms to regulate adhesion contacts and concomitant cell migration, it has yet to be ascertained whether each of the paxillin-mediated migration-control mechanisms are employed at distinct times and subcellular locations in a single migrating cell or whether they are cell-type or environmental-context specific. Further complexity has been suggested by the use of emerging, highly sensitive, nano-scale microscopy techniques, which are providing a more detailed dissection of focal-adhesion dynamics and submolecular organization. Evidence is growing that adhesions are not homogenous clusters of proteins but instead contain sub-domains that are enriched in specific focal-adhesion components or display `hot spots' of enzymatic activity. For example, in αvβ3-integrin-mediated adhesions, local variations in the degree of colocalization between the integrins, paxillin and vinculin have been observed (Zimerman et al., 2004). FAK is another protein that is known to interact directly with paxillin and to exhibit heterogeneity in distribution within adhesions. The visualization of FAK activity using a FRET-based biosensor also reveals `hot spots' of FAK activation, which are independent of the fluorescence intensity of FAK or the focal-adhesion size and might therefore indicate potential sub-domains of kinase activity (Cai et al., 2008). Vinculin and paxillin, when conjugated to the spectrally distinct photoactivatable fluorescent proteins Dronpa (Ando et al., 2004) and EosFP (Wiedenmann et al., 2004), can be localized to spatially distinct aggregates within adhesion contacts using high-resolution two-color photoactivated localization microscopy (PALM) (Shroff et al., 2007). It is also of note that the adhesion disassembly that is induced by microtubule targeting, as visualized by time-lapse microscopy of GFP-paxillin expressed in fibroblasts, does not occur homogenously within adhesion contacts, but rather in a punctate manner (Ezratty et al., 2005). The dynamic motion of adhesion proteins, as analyzed by fluorescence speckle microscopy, also shows heterogeneity within individual focal adhesions (Hu et al., 2007). There is therefore the potential for a high level of complexity in the spatiotemporal regulation of paxillin and its interactions within a single adhesion contact.

The evolving picture of focal adhesions, from mere homogenous structural plaques to complex, discretely heterogeneous, signaling foci, emphasizes the need for continued study to enable greater understanding of the dynamics of paxillin and the architecture of adhesion contacts. This will provide invaluable information on how adhesion contacts are regulated during the cell-migration-dependent processes of wound healing, the immune response and embryogenesis, as well as pathological processes such as invasion, metastasis and angiogenesis.

Regulated programmed cell death, or apoptosis, and cell proliferation are essential for normal embryonic development. Apoptosis also provides protection against cancer, which can be induced by genetic instability (Hanahan and Weinberg, 2000). Members of the Bcl-2 family of proteins have been shown to provide anti- and pro-apoptotic signals (Youle and Strasser, 2008) and are important in embryogenesis, the immune response and tissue morphogenesis. The interaction of Bcl-2 with the LD4 motif of paxillin promotes cell survival and is required for nephrogenesis during kidney development, although the mechanism of action is as yet unclear (Sheibani et al., 2008; Sorenson, 2004). The structure of Bcl-2 has been solved and, interestingly, reveals an amphipathic helical-bundle-like topology (Petros et al., 2001) that is closely analogous to the characterized LD-motif-binding proteins vinculin (Bakolitsa et al., 1999; Izard et al., 2004), GIT1 (Schmalzigaug et al., 2007) and FAK (Hayashi et al., 2002; Liu et al., 2002).

The interaction of vinculin with the LD1 and LD2 motifs of paxillin (Turner et al., 1999) has also been shown to regulate cell survival. It has been proposed that the tail domain of vinculin competes with FAK for paxillin binding and thereby promotes ERK signaling through FAK or paxillin to prevent apoptosis (Subauste et al., 2004). Paxillin has also been linked to apoptosis through its identification as a substrate for caspase-3. The cleavage of paxillin at six different aspartic acid residues by caspase-3 (Chay et al., 2002) was proposed to inhibit integrin-mediated cell-survival signals (Frisch and Francis, 1994) and promote apoptosis or anoikis caused by the detachment of cells from the ECM. In cardiomyocytes, apoptosis that is induced by the overexpression of the FAK homolog PYK2 can be inhibited by the co-expression of paxillin, which further implies that pro-survival signals are mediated by paxillin (Melendez et al., 2004). Thus, the involvement of paxillin in regulating cell survival and apoptosis suggests that it plays an essential role during normal developmental processes and could also provide a novel cellular target for cancer therapeutics.

In this brief review of paxillin's formative years, from its humble origins as a structural `peg', we have focused primarily on highlighting advances in our understanding of its role in coordinating the regulation of the motile machinery of the cell, because it is through this function that paxillin exerts the most significant impact on development and on tissue morphogenesis (Brown and Turner, 2004; Hagel et al., 2002; Ishibe et al., 2003), as well as on pathologies such as tumor-cell invasion (Azuma et al., 2005), fibrosis and atherosclerosis (Mehta and Griendling, 2007). However, the role of paxillin in regulating gene expression (Brown and Turner, 2004) through its interactions with ERK (Ishibe et al., 2003), poly-A-binding protein (Woods et al., 2002), Abl (Lewis et al., 1996) and steroid receptors, as well as through its own ability to undergo nucleocytoplasmic shuttling (Brown and Turner, 2004), should not be overlooked.

As paxillin comes of age, it is clear that it is important as a nexus for the coordination of numerous signaling pathways through its function as a molecular scaffold. However, there are several instances in which a number of proteins must compete for a single binding domain or motif on the paxillin molecule (Turner, 2000). Thus, an important focus of future studies must be the identification of the mechanism(s) by which the cell orchestrates these interactions in a spatiotemporal and context-specific manner. To that end, invaluable information will be obtained from the complete structural determination of paxillin in complex with its interacting partners: this remains a significant technical challenge owing to the inherent flexibility of the paxillin N-terminus. In combination with biochemical and microscopy-based direct-binding analyses, this will enable the development of bioinformatics-based modeling of paxillin and its interactome. The recent identification of adhesion sub-domains underscores the complexity of the interactions of paxillin with other proteins, and high-resolution microscopy-based technologies will be required to fully understand the spatiotemporal modulation of protein-protein interactions by paxillin, and therefore its regulation of downstream signaling pathways.

The increasing use of more in-vivo-relevant, 3D ECM model systems has identified aspects of cell morphology, focal-adhesion organization and signaling processes that are distinct from those observed using 2D substrates (Fig. 4) (Berrier and Yamada, 2007; Cukierman et al., 2001; Damianova et al., 2007; Pankov et al., 2005; Wozniak et al., 2003). It will be important to employ these 3D systems in conjunction with whole-animal models to ascertain the precise role of paxillin and its interactome in cell migration and survival in vivo.

The authors wish to apologize to colleagues whose work could not be cited because of space limitations. Work in the authors' laboratory was supported by grants GM47607 and HL070244 from the NIH.