Paxillin: adapting to change.

 
  I. Introduction 
 II. Paxillin Superfamily 
     A. Paxillin 
     B. Hic-5 
     C. Leupaxin 
 III. Paxillin Structure 
      A. Paxillin LD motifs 
      B. Paxillin LIM domains 
      C. Other paxillin binding partners 
 IV. Paxillin Phosphorylation 
     A. Tyrosine phosphorylation 
     B. Serine/threonine phosphorylation 
     C. Dephosphorylation 
  V. Paxillin Function 
     A. Paxillin phosphorylation and migration 
     B. p21-GTPase regulation and migration 
     C. Integrin-actin linkages and muscle contraction 
     D. Gene expression 
 VI. Development and Disease 
     A. Paxillin family member expression 
     B. Potential roles in disease 
VII. Future Directions 

Molecular scaffold or adaptor proteins facilitate precise spatiotemporal regulation and integration of multiple signaling pathways to effect the optimal cellular response to changes in the immediate environment. Paxillin is a multidomain adaptor that recruits both structural and signaling molecules to focal adhesions, sites of integrin engagement with the extracellular matrix, where it performs a critical role in transducing adhesion and growth factor signals to elicit changes in cell migration and gene expression.

I. INTRODUCTION

It is well established that an organism's normal development and maintenance, as well as its capacity to recover from injury and infection, is dictated to a large degree by the ability of individual cells to sense and respond appropriately to changes in their immediate external environment. Thus a complex, interwoven array of intracellular signaling pathways is modulated by cell adhesion to the extracellular matrix, combined with engagement of soluble growth factors and cytokines with their cognizant receptors to control cell proliferation and survival as well as changes in cell shape and motility. Functional defects or imbalance in these pathways can result in developmental abnormalities, tissue degeneration, hypertrophy, cell transformation, and metastasis.

Cell adhesion signaling from the extracellular matrix is initiated primarily via engagement of members of the integrin family of transmembrane receptors (103), with their appropriate ligands: fibronectin, laminin, vitronectin, etc. Contributions from other matrix receptor types such as the syndecan family of proteoglycans provide an additional level of control and complexity (306). In contrast to many growth factor receptors such as receptor tyrosine kinases, integrins have no inherent enzymatic activity; instead, intracellular signaling is initiated via clustering of the receptors in the plane of the plasma membrane and the concurrent recruitment and activation of intracellular signaling molecules via association with the cytoplasmic tails of the integrin [alpha]- and [beta]-subunits (70). Structural proteins are similarly recruited to provide a physical link to the actin cytoskeleton (153). These macromolecular adhesion complexes are commonly referred to as focal complexes or focal adhesions (226). The fidelity of integrin and growth factor receptor signaling relies on several common features. In particular, multidomain adaptor or scaffold proteins are utilized to recruit the appropriate complement of signaling intermediates and effector proteins to discrete subcellular compartments. Not only does this promote efficient activation of a given pathway but also adaptor proteins provide an ideal platform for the controlled integration of multiple pathways, thereby coordinating such diverse cellular responses as changes in gene expression and reorganization of the cytoskeleton.

First described over a decade ago, paxillin is one of the prototypical adaptor proteins involved in integrin signaling. In the intervening years, two new paxillin family members have emerged, and numerous new paxillin binding proteins have been identified. This review presents our current understanding of the role for the paxillin family in the integration of integrin and growth factor signaling (Fig. 1).

[FIGURE 1 OMITTED]

II. PAXILLIN SUPERFAMILY

A. Paxillin

Paxillin was initially characterized as a 68-kDa focal adhesion protein exhibiting a significant increase in tyrosine phosphorylation upon v-src expression (72). Shortly thereafter, it was purified to homogeneity from chicken gizzard smooth muscle tissue, an abundant source of cytoskeletal proteins, and identified as a novel binding partner for the focal adhesion and actin binding protein vinculin (279). It was named paxillin, derived from the Latin paxillus, a stake or peg, consistent with its proposed function in linking actin filaments to integrin-rich cell adhesion sites.

Paxillin was cloned by [lambda]gt11 expression screening in 1994 from an avian cDNA library (281) and shown to encode for a 559-amino acid protein. This protein is organized into a series of protein binding modules (Fig. 2). The amino terminus contains five LD motifs and several SH2-binding domains while the carboxy terminus consists of four LIM domains (48, 124, 274). These domains will be discussed in further detail below. Paxillin encoding cDNAs have subsequently been isolated from human (221), mouse (81, 177), frog (193), zebrafish (43), fly (304, 311), slime mold (AAM09351) (73), and yeast (163). Interestingly in nematode, a paxillin that contains both the amino terminus and carboxy terminus has yet to be identified, although a protein with the four LIM domains well conserved (including exon splice boundaries) exists (NM065879 Caenorhabditis elegans, BQ611130 Caenorhabditis briggsae). Further work on nematode genomic content and organization will reveal whether the full-length coding sequence remains to be identified or if the unique physiology of worms mandates the expression of a paxillin lacking the amino-terminal protein interaction domains. Regardless, the evolutionary conservation of paxillin attests to the critical importance of this molecule.

[FIGURE 2 OMITTED]

In higher eukaryotes, three paxillin alternative splice isoforms have been identified. Paxillin [alpha] is the principal, ubiquitously expressed isoform, whereas the [beta]- and [gamma]-isoforms exhibit restricted expression (176). The [beta]- and [gamma]-isoforms contain a 34- and 48-amino acid insertion, respectively, between amino acids 277/278 (Fig. 2). Although initial reports suggested the lack of a murine [gamma]-isoform (177), all three isoforms appear to also be expressed in mouse (308). It is not yet known whether other splice isoforms exist in lower organisms, although in zebrafish two proteins that differ by 5 kDa are detectable by Western immunoblotting (43). A fourth paxillin isoform denoted paxillin [delta] is the product of alternative translation initiation beginning at amino acid 132 (293; unpublished observations). Importantly, the consensus Kozak for this downstream start methionine is conserved across species. In total therefore, there is the potential in higher eukaryotes for the expression of six paxillin isoforms by a combination of alternative splicing and translation initiation.

Interestingly, a novel splice variant called PDLP is expressed in Drosophila that encodes a carboxy-terminal portion of paxillin including LIM domains 1 and 2 as well as the first zinc finger of LIM3 and a novel second zinc finger to generate a new LIM3 (304, 311). A Lepidopteran PDLP ortholog, death-associated LIM-only protein (DALP), originally proposed to be a Hic-5 ortholog, has been described (100).

Insofar as genomic organization is concerned, in humans the paxillin gene is comprised of 11 exons located on chromosome 12, mapped to 12q24.2 (221). In rat, paxillin is located on chromosome 12q16, whereas mouse paxillin is positioned on chromosome 5. Zebrafish paxillin is reported at linkage group 5, in Drosophila paxillin is on the long arm of chromosome 2 at 37D and chromosome III in C. elegans.

B. Hic-5

The paxillin paralog hydrogen peroxide inducible clone-5 (Hic-5) was first identified in an analysis of mouse osteoblast transforming growth factor (TGF)-[beta]- and hydrogen peroxide-inducible cDNAs and encodes a 444-amino acid protein (247) (Fig. 2). Human Hic-5 was later cloned in an androgen receptor two-hybrid screen and named ARA55 for androgen receptor coactivator 55-kDa protein (63). A Hic-5 splice isoform that contains an additional 17 amino acids on the amino terminus (containing an LD motif) was originally identified in a screen for marine paxillin (266) and later in a differential display screen of senescent versus nonsenescent human keratinocytes (328). Based on homology with paxillin, we suggest that the "full-length" Hic-5 splice isoform containing the amino-terminal LD1 "extension" be denoted Hic-5 [alpha] and the smaller form Hic-5 [beta]. As yet, there is no evidence that Hic-5 orthologs exist in lower eukaryotes, although unpublished observations suggest Hic-5 is expressed in zebrafish (43). Interestingly, Hic-5 shares the same 11-exon genomic organization as paxillin, consistent with an evolutionary duplication event, and is located on chromosome 16p11 in humans (328) and chromosome 7 in mouse (266).

C. Leupaxin

Leupaxin is a 386-amino acid 45-kDa family member that is predominantly expressed in leukocytes, as is reflected in the naming convention (150). Leupaxin is located on chromosome 11cen-11q12.3 and is encoded by 9 exons. It lacks its paralogs exons 4 and 5 and consequently has a significantly different structural composition and therefore likely unique regulation and function (Fig. 2). No evidence exists to support leupaxin expression in lower eukaryotes.

III. PAXILLIN STRUCTURE

The multidomain structure of paxillin and the lack of identifiable enzymatic motifs first suggested it was an adaptor protein (281). Within the amino terminus are five leucine-rich regions, termed paxillin LD motifs, that function in protein recognition (23). The carboxy terminus is comprised of four lin-11, isl-1, mec-3 (LIM) domains that also mediate protein-protein interactions (48). Dispersed throughout the molecule are many serine/threonine and tyrosine phosphorylation sites that will be discussed in more detail below. In addition, several potential proline-rich SH3-binding motifs are present within the paxillin amino terminus, consistent with a described Src SH3-paxillin association (302). The paxillin amino terminus is particularly proline and glycine rich, which in addition to the existence of a multitude of phosphoisoforms, is reflected by its aberrant electrophoretic mobility of 68-75 kDa versus a calculated molecular weight of 62 kDa.

A. Paxillin LD Motifs

Thus far the most extensively characterized domains within paxillin are the LD motifs. These protein recognition domains were first identified during a biochemical microdissection of paxillin initiated to identify the binding sites for the proteins vinculin and FAK (24, 281). Truncation and deletion analyses delineated two binding sites for FAK and one for vinculin. Visual sequence gazing revealed the binding sites to share a leucine-rich motif that was found to be repeated four times within the amino terminus of paxillin. Based on this, a "consensus" LDX-LLXXL paxillin LD-motif was proposed as an evolutionarily conserved peptide docking site for FAK and vinculin (23, 24, 274). Tong et al. (269) later proposed a fifth "degenerate" LD (LD3) that was originally disregarded due to lack of the conserved "LD" start but has since been incorporated into the nomenclature (Fig. 2).

The paxillin LD1 motif is conserved across species and paralogs and is encoded by exons 1 and 2. Notably, paxillin [delta] and Hic-5 [beta] lack the LD1 motif. The paxillin LD2 motif is encoded by exon 4 and is conserved across species and in Hic-5 but is absent in leupaxin, which has evolved a second unique LD motif that is encoded by exon 2. Paxillin LD3 motif is present within the highly divergent exon 5 and is conserved only among orthologs. The paxillin LD4 motif is encoded by exon 6 and is conserved across species and paralogs. Interestingly, the coding sequence for this LD motif is the most highly evolutionarily conserved DNA sequence of the paxillin family outside of the LIM domains. The paxillin LD5 motif is encoded by exon 7 and is conserved across species and paralogs.

The individual LD motifs provide specific protein interaction interfaces (23, 24) (Fig. 3). Within paxillin, LD1 mediates interactions with actopaxin (183), the integrin-linked kinase (ILK) (184), vinculin (278), and the papillomavirus protein E6 (268, 288). LD2 binds to vinculin and FAK/PYK2 (24, 278). LD4 binds to actopaxin (183), FAK/PYK2 (24, 265, 278), the Arf-GAPs p95PKL/GIT2/ GIT1 (278), and perhaps PAK3 (90), clathrin (278), and PABP1 (307). Thus far no binding partners have been identified for LD5 or the degenerate LD3 motif. Importantly, the capacity of Hic-5 and leupaxin to interact with these LD-binding proteins has been confirmed (64, 79, 150, 186, 266, 278).

[FIGURE 3 OMITTED]

Upon their identification, the leucine-rich paxillin LD motifs were modeled and suggested to fold as amphipathic [alpha]-helices with the leucines providing a hydrophobic interface with its binding partners (23, 228, 274), a prediction subsequently confirmed experimentally (7, 91, 152). The individual paxillin (and Hic-5) LD motifs are flanked by proline- and glycine-rich segments (average composition of 25%), which may also contribute to the global folding, presentation, regulation, and function of the individual paxillin protein binding domains. This is reflected in the [beta]- and [gamma]-paxillin isoforms exhibiting reduced affinity for the LD4 binding proteins FAK and GIT (175, 176). Also interspersed between the LD motifs are potential SH3-binding domains and numerous phosphorylation sites, including SH2 binding domains, that may also regulate the activities ,and context of the LD binding partners, thereby imparting an additional means of temporal-spatial regulation of the adaptor functions of paxillin. Hic-5 and leupaxin exhibit significant differences from paxillin in the length and composition of the LD "spacing" regions and consequently likely exhibit unique regulatory and functional capacities, indeed, Hic-5 binds to PYK2 more efficiently than to FAK (174, 194), and Hic-5 binds to GIT1 more efficiently than does paxillin to GIT1 (188).

Before our identification of LD motifs as discrete domains mediating binding to vinculin and FAK, the binding sites for paxillin on the vinculin tail (305) and the FAK focal adhesion targeting (FAT) domain (95) were identified through truncation and deletion mutagenesis. Subsequently, a FAK point mutagenesis study further localized the site of paxillin binding and allowed for a comparison to the known site of paxillin binding to vinculin (305). This led to the description of paxillin-binding subdomains (PBS) as an evolutionarily conserved paxillin binding motif (257). Thus a novel protein interaction pair, LD-PBS, was discovered. These data allowed for the creation of an algorithm to search and identify the PBS sequences within the paxillin LD binding partners PKL, ILK, and actopaxin and other potential candidates (23, 183, 184, 274, 278). The general utility of LD-PBS associations has emerged with the discovery of an interaction between gelsolin and PYK2 (294) as well as between the papillomavirus E6 protein and the ubiquitin ligase E6-AP (14, 47), the RapGAP E6TP1(67), and the [Ca.sup.2+]-binding protein ERC-55/ E6BP (38), in addition to paxillin (211).

The crystal structures of the vinculin tail and the FAK FAT domain, winch contain the PBS, have been solved (7, 10, 91, 114). FAK FAT solution structures have also been reported (66, 152). The vinculin tail and FAK FAT domain share a parallel up-down-up-down four-helix bundle. This general structural organization is shared by [alpha]-catenin, apolipoprotein E, and the p130Cas family of proteins (7, 10, 91, 114), although only vinculin and FAK bind paxillin. Interestingly, structural predictions suggest that the PBS-containing PKL carboxy terminus also folds in a similar manner (7). In contrast, this structural organization is unlikely in the case of the ILK and actopaxin PBS domains (7, 183). Such differences are likely to dictate the optimal binding parameters for each protein and contribute to selection of the "appropriate" paxillin binding partner(s) in response to a particular extracellular cue.

B. Paxillin LIM Domains

LIM domains are double-zinc finger motifs first identified in the lin-11, isl-1, and mec-3 homeodomain proteins. The four LIM domains of paxillin are present in tandem on the carboxy terminus, thus designating paxillin a group 3 LIM family protein (48). The structures of several LIM domains have been solved and reveal that LIM domains are arranged such that each individual zinc finger is comprised of two antiparallel [beta]-sheets that are separated by a tight turn. In addition, the two zinc fingers pack together due to hydrophobic interactions, and each LIM domain ends with a short [alpha]-helix (48, 203, 289). Interestingly, although it is now generally accepted that the primary function for LIM domains is in mediating protein-protein interactions (235), the CRIP and CRP structures are nearly identical to the DNA-binding domains of GATA-1 and the steroid hormone receptor family (203).

The capacity of LIM domains to mediate discrete subcellular localization to the actin cytoskeleton and focal adhesions was first demonstrated for muscle LIM protein (MLP) (6) and paxillin (24) and has subsequently been found to be a common link between many group 3 LIM domain proteins including the zyxin (190), PINCH (144), and FHL families of proteins (40). Although it is not believed that tandem LIM domains physically interact, there is strong evidence for cooperatively between adjacent LIM domains (6, 24, 190).

Functionally, LIM3 (with LIM2 cooperating) mediates the localization of paxillin (24) and Hic-5 (64) to focal adhesions with serine/threonine phosphorylation of LIM2/3 regulating paxillin localization to focal adhesions and consequently cell adhesion to fibronectin (25) (Figs. 3 and 4). These sites of phosphorylation are largely conserved in paxillin orthologs. The localization of leupaxin has not been extensively studied, although it colocalizes with cortical actin in lymphoid cells attached to intracellular adhesion molecule (ICAM)-I (150), podosomes in osteoclasts (79), and focal adhesions in fibroblastic cells (D. E. Staunton, personal communication). The insect paxillin splice isoform (PDLP/DALP) that is comprised solely of LIM domains localizes to myotendonous junctions and along actin stress fibers (311), whereas the LIM domains of the yeast ortholog Pxl mediate targeting to the bud neck (163), thus offering evidence for the ancestral conservation of LIM function in subcellular compartmentalization. The identity of the paxillin focal adhesion targeting molecule has proven elusive, as it has for these other LIM domain containing proteins. It will be interesting to determine whether any relationship exists between the protein(s) that target each of the diverse LIM family proteins to their respective cytoskeletal compartments.

[FIGURE 4 OMITTED]

In addition to the identification of the LIM domains as mediating paxillin subcellular targeting, tubulin was identified as a LIM2 and LIM3 binding partner (94) and PTPPEST as a binding partner for tandem LIM3 and LIM4 of paxillin, LIM4 in Hic-5 (41,186), and leupaxin (79) (Fig. 3). The ability of Hic-5 and leupaxin to bind tubulin is unknown. Paxillin LIM2 also associates with mid precipitates a kinase that can phosphorylate T396/401 (human), and LIM3 binds to a detergent-insoluble kinase that can phosphorylate LIM3 on S455/479 (25). The identity of these kinases as well as the capacity of Hic-5 and leupaxin LIM domains to bind to and be phosphorylated by serine/ threonine kinases remains to be determined. In addition to these cytoplasmic LIM associations, Hic-5 LIM domains have been shown to be capable of binding to DNA (187), and the capacity of paxillin and Hic-5 to interact with the nuclear matrix (315) perhaps through LIM3 (121) has been described. Furthermore, both paxillin and Hic-5 can interact with the androgen and glucocorticoid receptors through LIM4 (208). The significance of these interactions and the function of paxillin and Hic-5 in gene expression is discussed below.

C. Other Paxillin Binding Partners

In addition to the LD and LIM binding partners described above, paxillin binding to many other molecules has been reported, although in many cases the sites of interaction are not clearly defined (Table 1). A direct association of paxillin with [beta]1-integrin (231, 259), [alpha]4- and [alpha]9-subunit cytoplasmic tails has been reported (153, 155, 156, 322). The [alpha]4- and [alpha]9-integrin binding site has been localized to a region …